ie

Lp
Aa

L
aenstelereatigeaenl:

p
eerie
RPE parr
c

rs

J
PIRIPPR

Ket
Tee ru

WCB PCs
aren AEG

erect
ras eeeVM Brera

eo
PUPCRTC NG PRUEHER pierr

Mseerast
igestee
are

tes VONb
eee
EEE

PMP
Lear atare okey
jority
POT

Pe
vowel
Cera
NPL Pier Seay

SC
yy) Tet Sete
CNP
Srre

rar
ibaa ee arare

aaatghe
Exphene! ere
AL
te

rere y
or.

Ce eva ries
S ede
tet oa
etre
7 hieig

sdyhid
Let
Aa
seratoatyne
Shirk
hide
pada

Wave
ne ty ee
Bee

9
ms

ereae
se
eT? Ni
e4ay
Mt
2s
é bye CU
Ate
ge
ee
er
ny
era)
er
eater

:
ite
data
tatadtdy

Bins
areas
awe
Say
err
PY
AT
rt

eete
eet ears

seat
x

ceeh enecee WEDD

Le SRO

enn
Carp

eer
ney!

Ean

apm}
LPO

Lae
4

ear b
LenS Z Weer
aria
rer ee

Lb
Tae
eee
esr

arctor
ye:
aren a a) ROT

at
ereray

er

Ee
Paper
ARM (ots

one htehye hurSash

wy rT

tae
vets
o

ara
ne oePetree
4. ra Harr

a
ny NY
See
RENeae
OL a
SPT
NCS
ORaE
HSt
trae

Pe Vr
een
nay:he
by
hi)
eee eae va

<
er a

Talla
varee St
}
at eerene
eee

upseaee
elt orirrererg!
soe

q
Wee
oty

aE
ae

ane
iene
Ary

vt aS Pee
ae

Sgn

Pane
vePes)
Sts

a
LOOP
:
ota
BE
ete
Aan
eriere
oa

a
ant
are

an
PNT Re. eyar

“ tiverineks
ireoa

ST
ae

Se
i
aY

aarti: Me
Seay
Cort

WOOP Pate

sn Sr
DN ty

or
aso
Dee

Meee Tee
eee Re cs
potaphy
eT. Nt

va
SEL
SY ee
ete
ana

es SeeMeL
ye
rertans

NIN
a) eryba

ia

andaudt
aeae
yet

poner

cee
or
at Nirdien

eee
cay
ROR
Les
ey
Cay
ye:

+,
pei
ahi
eve

ao
veer

‘ wore
ae tae
AeSCr
SOY
Whe.
¥
ESECE

wae
armen
ee
eae
ay

ran
SoraST

xu
ois PyPens

Senn
v8 tote SSP Darseahen

Perey
Teer
[CVPR
Lyte

seh Yet \
a as
SAYER

58

This baets
Dae OR

con
eS
De
peere

piesa
ior

Ovetee
URveRr
rete
ee

:
ear eerie shee

NT: ari ana

ony TY
Siunies
SEDiy
ererh Nae VaThe er

eran
\eeeate
ON

Navn
=
oe
PSAy

a
AVA
rs
oe

Toneyaa

«Nutri
yeh
my ee
Pela
ii Fe

Neath
Ay

ats
aN
We a Gn Fa

prey d
Serr
ay

ties
Peasy

Petre
ae
wr

Grhe
mre
ey

pal

Ne
eerie
Keener
Pree
FETE
RIGS
wre
yy
Steve
Nara

\
el)

‘ ny 4
i

Peay ie
OTL
Sree re ear eta §
ewes ear uta te

ea

dats
Nt TLL)

Ruts
ese

\
yey
POST
GS a SCTE PURE
rei

Penman’
ESO
Neer

ues

aren
See
Seah ae
wits Ser

‘
nana
TECINES
WHY er
ATESsy Sa

ESTE
aA

TneSe
SY mune
ST

Rea)
Sete

yy
Coe
Tens

erence

N
Adis

Dede

r§
Ae wes

te.
aa
eras

reery

ESET)
ee
Pete
ea
rerery
oi si ysvans

5
45 $s
ATSIC Y

on
NEON

5 eds
Eaede
ve Seen Syswy Sasa
Sewn
ESSE

Needs

ete
TAM
wens

SE
Ve

ETE
aye
ae.

Si
Lya yee
ae
eneue
k
aeRO

tase
ot) SEE

ENTE
eres eae plea
ereteiesSiete
eae
eexefeaSie
Sete

Sad,
;

a8
ASE Sis
Nek
ua
Ee

Ae
bt
>

Ney
Sra
Ses PVE
aeSyah SS iHey
NOSES ET ETS Tat rt

FLTi StyPya
PEE eee See Pa

2
SSS

eee
yFS 2 SIOYe ee YRS,

ENS OE Ya ae
essay NS

NESS

¥ a TLieee seeder’
te Vesey

Sey
an aS yatta)

xyes
Siriesr eesesnies Petree
Serene
epee

eats
Jeoth
We

TTL
ES Ee

Weeks
nS
eres ey

:
a
Sonn VESee
be

aaron oak
eShata
acta
eae)
See iNe64

eee
per

iei zeuseas ee
e ses

toe
RN
ets
+
3
Ceramic Science for Materials Technologists

WITHDRAWA-UNL
To my wife, Dorothy
 Ceramic Science
for Materials Technologists

1. J. McColm, B.Sc., D.Phil.

Senior Lecturer in Industrial Technology
University of Bradford

Leonard Hill
Distributed in the USA by
Chapman and Hall
New York
 Published by Leonard Hill
a member of the Blackie Group
Bishopbriggs
Glasgow G64 2NZ

Furnival House
14-18 High Holborn
London WC1V 6BX

Distributed in the USA by

Chapman and Hall, 733 Third Avenue,
New York, N.Y. 10017,
in association with Methuen, Inc.

© 1983 Blackie & Son Ltd.

First published 1983

All rights reserved

No part of this publication may be reproduced,
stored in a retrieval system, or transmitted,
in any form or by any means,
electronic, mechanical, recording or otherwise,
without prior permission of the Publishers.

British Library Cataloguing in Publication Data

McColm, I. J.
Ceramic science for materials technologists.
1. Ceramics
|. Title
666 TP807
ISBN 0-249-44163-2

Library of Congress Cataloging in Publication Data

McColm, |. J.
Ceramic science for materials technologists.

Includes index.
1. Ceramics. |. Title.
TP807.M38 1983 666 82-22205
ISBN 0-412-00351-1

Filmset by Advanced Filmsetters (Glasgow) Ltd.

Printed by Thomson Litho Ltd., East Kilbride, Scotland
Preface

This book is designed to cover those topics within ceramics that have in my
experience provided the necessary knowledge, excitement and breadth for
students following materials science and materials technology courses. It
should also provide a stimulus to anyone entering the industry. It is one
man’s view of the coverage of ceramics that can be achieved in a general
materials course, other parts of which consider detailed metallurgy, polymers,
crystallography and techniques of physical examination.
Knowledge of other topics common to materials courses is assumed in this
text—for example, dislocation theory, phase diagrams and sintering, all of
which are best covered in a metallurgy course. The whole area of crystallo-
graphy, its conventions and notations, which are a vital part of materials
science, should be dealt with separately and fully in a course on physical
techniques.
As ceramics in its widest sense is the one area of materials science where a
student encounters all the types of chemical bond, it should be the area where
these are emphasized. Thus the approach used here is to develop these more
fully than in most materials texts and to use them to investigate many of the
structures and properties encountered.
The following pages present a large amount of factual information,
providing a broad basis on which the reader may build. Chapter 1 develops
traditional silicate and aluminosilicate ceramics which have inherent restric-
tions resulting from the way silicate structures are formed, on the way they
can transform, and most importantly on the fact that they are glass-matrix
based mixtures. The materials can be improved by changing the crystal phase
and this forms the substance of chapter 2 where special lightweight ceramic
elements, compounds and composites are discussed. The covalent bond is the
dominant bonding concept in these early chapters and it is dealt with in
chapter 2.
Chapter 3 is devoted to glass science and examines how the materials of the
first chapter can be improved. Crystallization of the glass is a possible route
and so chapter 4 deals with this topic in the context of glass-ceramic science.
Vv
Vi PREFACE

In chapter 5 an alternative answer to the question of improvement posed by

chapter 1 is the development of binary ceramic oxides that are not silicates,
in a glass matrix. At this point the materials are best approached using the
concepts of the ionic bond model, which is dealt with in the necessary detail.
The final chapter demonstrates that those ceramics termed “special” must be
seen from a mixed bonding model in order to classify the properties
encountered in such materials.
There are diversions and applications that are analysed in terms of the
main themes of this text; where they interlink, the reader is referred to other
sections that develop the ideas. Attempts are made to analyse the old arts for
their science, and to provide examples of new technological developments
stemming from scientific research.
The following are gratefully acknowledged for permission to reproduce
previously published material: The Royal Society of Chemistry (Fig. 2.1),
Chapman and Hall Ltd (Fig. 3.15) and Taylor and Francis Ltd (Fig. 5.7).
Finally I must acknowledge that although “I have written what I have
written”, it is not without great assistance. First, I wish to thank my many
classes of students for their enthusiasm and benevolence. I am grateful to Dr
R. Steadman whose vision made my interest in the subject possible. Help in
preparation of figures was cheerfully given by Mr D. Mistry, Mr V. Kotroczo
and Mr H. Clay, whilst Miss M. Cobb and Mrs M. Walker were foremost
amongst those on whose typing skills I relied.
I.J.M.
Contents

1 Traditional ceramics
1.1 Introduction
1.2 Study area defined
1.3. Role of heat—heat work
1.4 Translucency
1.5 Silicates
1.5.1 Silicates and aluminosilicates containing isolated anion groups
1.5.2 Chain silicates
1.5.3 Sheet silicates
1.5.4 Framework silicates
1.5.5 Sodium silicates
1.5.6 Cements
1.5.7 Thermal transformations of ceramic raw materials—topotaxy
1.5.8 Phosphate-bonded refractories
References and further reading

2 Special light ceramics for modern applications

2.1 Analysis of a strength equation
2.1.1 Surface energy
2.1.2 Origin of Young’s modulus
2.2 The covalent bond model
2.2.1 Nature of the problem
2.2.2 Matter waves
2.2.3 The wave nature of the atom
2.2.4 Atomic orbitals and their specification
2.2.5 Atom groundstates, aufbau principle and the periodic table
2.2.6 Valence bond description of covalent bonding
2.2.7 Molecular orbital theory of covalent bonding
2.3 Identification of potential high-strength materials
2.4 Boron
2.4.1 Structure and bonding
2.5 Carbon and graphite
2.5.1 Structure and bonding
2.5.2 Carbon fibres for engineering use
2.5.3. Graphite reactivity
2.6 Silicon
2.6.1 Preparation and structure
2.6.2 Hole conductivity
2.6.3. p—njunctions

Vil
Vill CONTENTS

20 Boron carbide 100

2.7.1 ‘Structure 100
2.8 Boron nitride 101
2.8.1 Preparation and properties of pyrolytic boron nitride 101
aS Boric oxide 103
2.10 Silicon carbide 104
2.10.1 Structure 105
2.10.2 Oxidation 106
all Silicon nitride 107
2.11.1 Structure 107
2.11.2 Properties and fabrication 108
2.11.3 a-Silicon nitride and silicon oxynitride 110
2.11.4 Fabrication with silicon nitride 111
2.11.5 Densification aids in Si,N, sintering 112
Dale Sialons 114
2.12.1 Representation of the Si-Al-O—-N system
2.12.2 Si-Al-O-N phase relationships 117
2.12.3 Sialon preparation 120
2.12.4 Metal-sialons
2.12.5 Nitrogen glasses 122
Dall Ww Realization of high strength and the development of toughness through fibre-
composites 123
2.13.1 Composite elastic properties
2.13.2 Composite strength
2.13.3 Specific strength and specific stiffness
2.13.4 Composite toughness
References and further reading

3 Glass
oul Introduction
3.1.1 Definition
3.1.2 Viscosity
Br Structural models for glass
3.2.1 Zachariasen random network
3.2.2 Porai-Koshits crystallite model
3.2.3 Discrete-ion concept
3.2.4 Five-fold symmetry model
3.2.5 Goodman’s strained mixed cluster model
3.3 Predictions of glass formation
3.3.1 Bond strength and type
3.3.2 Coordination environment
3.3.3. Strained mixed cluster model
3.3.4 Kinetics and concept of free volume
3.3.5 Cooling rate
3.4 Mechanical properties
3.4.1 Group 1: flawless glass
3.4.2 Group 2: flaws exist—how do they propagate?
3.4.3. Characterization of Griffith flaws
3.4.4 Fracture mechanics
3.4.5 Plastic flow in glass
3.4.6 Group 3: flaw generation hypotheses
3.4.7 Crack propagation and static fatigue
3.4.8 Methods of glass strengthening
3.4.9 Load-duration analysis for glass design
3h Glass/metal and glass/ceramic bonding
3.5.1 Technical procedures
3.5.2 Theories of adherence 187
3.5.3 The glass/metal interface 192
 CONTENTS

3.6 Durability 194

3.6.1 General observations
3.6.2 Mechanisms and rate laws
3.6.3 Methods of corrosion protection
References and further reading

Glass-ceramics
4.1 Introduction
4.2 Technical processes
4.3 Nucleation
4.3.1 Homogeneous nucleation
4.3.2 Heterogeneous nucleation
4.3.3. Nucleant characteristics
4.4 _ Liquid phase separation in glass-forming systems
4.4.1 Occurrence of the spinodal region
4.5 Crystal growth
4.5.1 Screw dislocation mechanism for crystal growth
4.5.2 Spherulitic and dendritic crystal growth
4.6 Glass ceramics arising from control of process variables
4.6.1 Lithia-silica (Li,0-SiO,) system
4.6.2 Li,O-Al,03-SiO, system
4.6.3 MgO-Al,O,-SiO, system
46.4 Na,O-Al,0;-SiO, system
4.6.5 Tetrasilicic-mica machinable ceramics
4.7 Role ofa residual glass phase
4.7.1 Strength
4.7.2 Control of fracture toughness and thermal shock resistance
4.7.3 Creep resistance
4.7.4 Optical transparency
4.8 Summary
References and further reading 233

Refractory oxide ceramics 230

5.1 Introduction 235
5.2 Prediction of environmental stability 238
5.3. The ionic bond 239
5.3.1 Production of isolated cations 239
5.3.2 Production of isolated anions 244
5.3.3. Coulombic attraction 245
5.3.4 Lattice energies 247
5.3.5 Kapustinskii equation 249
5.3.6 Polarization 253
5.3.7 Geometrical aspects of ionic bonds 254
5.4 Alumina, Al,O,
5.4.1 Sources
5.4.2 Manufacture
5.4.3 Structure
5.4.4 a-Alumina
5.4.5 The ruby laser—a new use for an old material
5.4.6 B-Alumina
5.4.7 y-Alumina
5.5 Zirconia, ZrO,
5.5.1 Sources and manufacture
5.5.2 Bonding and structure
5.5.3. Mechanical properties oF,
5.5.4 Problem of the monoclinic/tetragonal phase transition
5.5.5 Stabilized zirconia
5.5.6 Ceramic steel: a use for the zirconia phase change
 CONTENTS

5.6 Magnesia, MgO 284

5.6.1 Sources and manufacture 284
5.6.2 Structure and strength 286
5.6.3 Role of impurities in magnesia 288
Ded Urania, UO, 252,
5.7.1 Source and preparation 293
5.7.2 Structure 293
5.7.3 Strength 293
5.8 Titania, TiO, 295
5.8.1 Manufacture 295
5.8.2 Structure 296
5.8.3 Bonding 298
5.8.4 Mechanical properties 299
5 Spinels 299
5.9.1 Structure 300
5.9.2 Factors determining degree of inversion 302
5.9.3 Preparation and reactivity 305
5.10 ABO,-type oxides 306
5.10.1 Structures 306
5.10.2 Properties 307
sitll Ferrites 308
References and further reading 309

Special ceramics—refractory hard metals 311

6.1 Interstitial carbides and nitrides of classes 1 and 2 312
6.1.1 Preparation 312
6.1.2 Compositions and non-stoichiometry 313
6.1.3. Properties and uses 317
6.1.4 Uses of cemented hard-metals 321
6.1.5 Bonding 323
6.2 Borides 330
6.2.1 Structures 332
6.2.2 Non-stoichiometry 336
6.2.3 Bonding and valency 337
6.2.4 Preparation 340
6.2.5 Properties 341
6.2.6 Uses 343
6.3 Silicides 344
References and further reading 345 |

Index 347
1 Traditional ceramics

1.1 Introduction
Traditional ceramic techniques consist in the main of working and heating
china clay, ball clay, calcined flint, and a feldspathic mineral known as
Cornish stone. The starting mix commonly includes 50-60% of clay, and
when fired it becomes microporous, consisting of crystalline phases bonded
by a complex glassy phase. The porosity is controlled to a degree by adding
up to 50% of animal bone, which produces an essentially fully vitrified
non-porous china characteristic of English ware. A similar vitrified ware is
produced abroad for domestic items, and in Britain for electrical insulators,
by adding feldspar as the flux in place of bone, to produce porcelain goods.
The production of artefacts from this restricted range of minerals constitutes
a large branch, in financial terms, of the ceramics industry.
The second main branch of the traditional ceramics industry uses clay
minerals without additions, to produce building materials such as bricks,
tiles, pipes and floor quarries to the extent of over 10° items every hour of the
year when in full production. The traditional ceramics industry is therefore of
considerable importance to the economy.
FLINT SiO,
Porosity increasing

FELDSPAR

Ease of
forming
Vitrification
increasing

FELDSPAR CLAY LEUCITE Maar:

(a) :
K,0.AIO, 4S 4Si0, (b) 3AI,0, 2Si
Si 0

Figure 1.1 Composition of main products arising from the use of standard raw materials:
(a) schematic diagram; (b) same information showing compositions of some materials. A,wall
tile; B, vitreous white ware; C, floor tiles; D, electrical porcelain; E, hard porcelain; F, semi-
vitreous porcelain.
2 CERAMIC SCIENCE

Comparing energy costs per tonne of product shows that it takes approxi-
mately four times as much energy to produce pottery-ware compared to
glassware (cf. 3.1), and clearly this will become increasingly important with
present energy costs.
Figure 1.1 indicates where common ceramic products lie with respect to
composition. This scheme emphasizes the mixed nature of the products and
indicates the changes effected by control of composition.

1.2 Study area defined

Four main topics emerge as deserving most attention. These are (a) clays;
(b) transformations in clays; (c) the role of glass bonding phases; and (d) the
structure and properties of silicates and aluminosilicates derivable from the
common raw materials.
Major sections of this chapter will be devoted to (a), (b) and (d), whilst (c)
is dealt with in sections 3.1, 3.4, 3.5, and 4.7.
The following flow-chart summarizes the main theme to be developed in
this chapter and gives some indication of the reasons leading to the
restrictions encountered in the properties of ceramic materials.

Baked Clays = |(TraditionalCeramics)

Ceramics) = Heat on earthy raw materials
2 ee
cca
oe

Layer structures One structural family = silicates or aluminosilicates

with chance fit of
of Mg(OH), and 3
Al(OH); give glass
limited range of ——-» _ Limited Structures (arises
Composition and Transformations | from restricted linkages due
tOstrone Sig eee Sin|
| repulsions)

Changes governed
by existing crystal Narrow band of
structures (topotaxy) properties

Traditional ceramic phases can be considered using the features inherent

in two principal areas of bonding, and this approach will be adopted
throughout. The two types of bond encountered, ionic and covalent, should
be appreciated and understood (see 2.2 and 5.3). Knowledge of other areas
essential to a well-balanced materials course will be necessary, principally
crystallography and phase diagrams.
 TRADITIONAL CERAMICS 3
1.3 Role of heat—heat work!!
Towards the end of the 18th century, Josiah Wedgwood began to investigate
materials used in pottery manufacture, and so became one of the first to
attempt to obtain a basic understanding of the ancient craft of processing and
firing clays. Wedgwood developed forms of dilatometry, thermogravimetry,
and evolved gas analysis, which were applied to studying the changes
occurring during the heating process necessary to cohere a pre-formed shape.
Ceramic ware is heat-treated to the point of customer acceptability, and
therefore compositions and microstructure reflect arrested reactions. The
technological control considered above all others is the heating cycle.
Although it is so important, the heating cycle is still the least understood step
in ceramic manufacture, and is often surrounded by myth and regarded as art
rather than science. This conception largely persists, despite work, over forty
years old, by Meyer who produced vitrification charts relating maturing of
products to length and temperature of kilning. Such a presentation allows
deduction of a variety of possible treatments to obtain a given degree of
vitrification.
In practice, most attention is given to the time-temperature profile during
ceramic firing, which gives rise to the loose concept of “heat work” (from an
integration of the area under a time-temperature curve). It is felt that this
“amount of energy” put into the system represents the work required to
achieve densification. This is far from the truth, as a simple consideration
below will show, and the total energy supplied probably bears no relation-
ship to the extent of reaction or maturation of the artefacts. Instead, it relates
to the thermal mass of the kiln and the rate of energy input. This can be seen
from the following order-of-magnitude analysis.
First, the energy needed by the various processes occurring during kilning
is assessed.

(i) Sintering or maturing: this process does not require energy, but evolves it, since the surface
energy is decreasing (see 2.1). The effect is small, approximately 4kJ kg ', for clay-sized powders.
(ii) Raising the mixture to temperature: since one is dealing with a restricted range of raw
materials of similar composition and structure, the heat capacities of most of the resulting
ceramics are similar and require about 80kJkg~' per 100°C, or about 1000kJ kg~! to reach
1200°C.
(iii) Chemical decomposition: parent raw materials are transformed to silicates or mixed oxides
as shown in 1.5.7. These reactions are usually exothermic, and can contribute energy up to
1400kJkg~+, but since such reactions are rarely allowed to progress to completion a more
realistic contribution is about 200kJ kg7?.
(iv) Production of a liquid phase: several transformations produce a liquid phase which eventually
becomes aglass, and as a result a latent heat of fusion must be supplied. At the most, 20 % of the
body will be liquefied and so the energy need will be about 60kJ kg~}.
(v) Structural phase changes: solid state transformations typically involve energies in the region
of 40kJ kg~?.

The summation of (i)—(v) suggests that although considerable energy is

required to raise the ware to temperature, the overall balance shows only a
small overall energy requirement to effect transformation. Thus the idea that
4 CERAMIC SCIENCE

energy is being supplied during firing to drive the product to its final state is
not easily defended.
Information from kilning operations shows that for sanitaryware typically
8000 kJ kg! is used, for chinaware 4000 kJ kg~', and about 6000 kJ kg? for
basic bricks. This suggests that most energy is expended in heating kilns.
It seems then that attention should be focused elsewhere, probably on the
concept of activation energy. High temperatures are required to overcome
the energy barrier to processes involving the transfer of material. Activated
processes frequently conform to the Arrhenius equation, (3.1, 3.6), thus since
the controlling feature has an exponential form, time spent at temperatures
below the critical value to overcome the activation energy is to a large degree
wasted. For example, since typical activation energies in traditional ceramic
systems are about 300kJ mol~', a process that requires 1 hour at 1500°K
needs nearly three years at 1000°K. Thus time at peak temperature is the vital
parameter in ceramic firing and ceramic ware should be heated as fast as
possible to achieve peak temperatures.

1.4 Translucency
One of the properties that firing schedules seek to control is the degree of
translucency in fine white china,'* but another feature in firing should also be
considered here, namely the furnace atmosphere. Translucency results from
two processes—light scattering and absorption—and two parameters, S, the
scattering coefficient, and K, the absorption coefficient, must be extracted
from the measured reflectance and transmittance of fired samples.
Kubelka and Munk have derived equations relating the coefficients and
measured properties.

e1Sa
Rites
© a+b Coth bsd oe
1
Ea Se Se !
aSinh bsd + b Cosh bsd 2)
where Rg is the corrected reflectance of the sample over a black background,
T is the corrected transmittance, s is the scattering coefficient, d is the sample
thickness, and a and b are constants given by equations (1.3) and (1.4):

a= Ae +R 1.3

i od|
b=~|——--—R :

R,, 1s the reflectance of a very thick sample and has the form
S
Ro eee
e (1.5)
SH 42/K(Kos)
 TRADITIONAL CERAMICS 5

Because ceramics have internal as well as external surfaces to.cause reflection,

corrections have to be applied to the measured reflectance and transmittance

Rol —r,)(1—r,) (1.6)

1—r,Ro

t= (l= tyRe) =r
(1.7)
where r and t are measured reflectance and transmittance respectively, Ry and
T are the corrected values for use in (1.1) to (1.5), and r,, r, are the external
and internal reflection coefficients. The choice of values for r,; and r, is not
critical, and general values obtained for opal glasses of r, = 0.06 and
ry = 0.50 can be adopted.
Using equations (1.1) to (1.7) allows values of S and K to be obtained for
translucent ceramics. Figure 1.2 shows some results for two types of hard
porcelain. Hard porcelain is fired under reducing conditions, and as a result
iron impurities are reduced to Fe**. Since Fe?* in silicates has a strong
absorption near to 1200 nm, the tail of this is responsible for the rise of K in
the red region. A difference in titanium content is responsible for the
increased absorption in the blue end occurring in the Czech sample; charge
transfer Fe**—O-Ti** — Fe? *—O-Ti?* being responsible for strong ultra-
violet absorption at 350nm. No difference can be detected in the scattering
coefficient, indicating that the reducing environment does not affect the
microstructure.
Soft porcelain differs from the hard variety in that it is heated to a lower
peak temperature in a kiln which, because of the firing method chosen, does
not subject it to reducing conditions. The result is that Fe** remains in the
structure, bestowing a strong absorption in the ultraviolet, and producing a
tail into the visible blue giving a pinkish hue. Scattering coefficient values are

0:22

iczechoslovakian

@english clay

4
400 500 600 700 800 900
Wavelength nm

Figure 1.2 Scattering and absorption coefficients as a function of wavelength for hard porcelain
made from Czechoslovakian or English clay.
6 CERAMIC SCIENCE

higher for soft porcelain as a result of the lower peak temperature achieved,
which leaves a more porous microstructure.
Bone china when devoid of manganese is very white, with very low K
values at 400—S00 nm. This is because the phosphate group produces a charge
transfer peak, with the Fe** shifted to shorter wavelengths to give little
tail-over. From this one can conclude that bone china is fired in non-reducing
atmospheres.
Separation of the two coefficients in this way is useful because K changes
are indicative of sample chemistry and crystal spectra, while S changes can be
used to probe microstructural changes during firing.

1.5 Silicates
When the compositions of naturally-occurring or synthetic silicates are
considered, they appear to be a complex and almost infinite series of
materials which, despite their large numbers, show arelatively narrow range
of properties. Early attempts at classification based on chemical analysis were
of almost no value for the following reasons:
(a) Frequently the homogeneity of a sample is in doubt.
(b) Because of the mixed bonding present the ionic components are capable of isomorphous
replacement; Mg?*, Ca?*, Fe?*, etc. may replace each other; Al?*, Fe?+ and Na‘, Kt,
OH ,F , etc. may all replace each other in endless variety.
(c) Cation radii (5.3) for Si** and Al?* are close (at 0.5nm and 0.55nm respectively) and
therefore Al?* can replace Si** as long as a balancing cation is inserted or M"* is iso-
morphously replaced by M“*1)*,
(d) When Al** replaces Si**, Al?* can become octahedral with 6 oxygen near neighbours, even
though Si** is constrained to a 4-coordination with oxygen (table 5.6, 5.3.7).

Rationalization of the silicate and aluminosilicate structures was achieved

with a classification based on structural units. This approach has given us the
well-known categories of discrete group, chain, sheet and framework silicates
as described below. Sometimes the linked polyhedra become so deformed
that this otherwise powerful descriptive method becomes restrictive, and .
structural descriptions based on layer packing sequences have been developed.
Both approaches are used in this text; (1.5.4), (2.10) contain examples.
The SiO, tetrahedron has been shown by X-rays to exist in silicates, and
can be theoretically justified from electronic structure arguments (2.2). It is
often drawn schematically (Fig. 1.3) and can be considered in an ionic way as
Si** tetrahedrally coordinated by O?~, while bearing in mind that covalent
bonding between Si-O must be important and can have dominant structural
effects (2.2). In almost all silicates, the dimensions of this unit are little
changed, with Si-O = 0.16nm and O-O = 0.27nm.
Linking of tetrahedra occurs to generate new structures by a process of
condensation, leading to a sharing of oxygen at the tetrahedral apices.
Hence not only does the new structural unit have a larger size, but the
Si: O ratio decreases. From the cation repulsion argument in 5.3.7, we arrive
at an immediate restriction in that any two tetrahedra can only share at one
 TRADITIONAL CERAMICS We

Si 3a |

Ac
[s',] -
[st0, one
ial

oe ey
0-1 3 5
nm
[S06]

Figure 1.3 Size and structure of isolated Si-O anions.

apex, never at two or three, that is, they can corner-share but not edge- or
face-share. Application of this restriction gives rise to the limited classes as
follows.
(a) Discrete anionic silicates
Single SiO{~ ions can occur in structures, two tetrahedral anions can
condense to form Si,O$~ ions, three will condense to form (Si,;0,)°~, four
give (Si,O,,)®~ and six condense to form (Si,O,,)'?~. These have been
identified in structures that will be considered more fully below. The principle
of condensation to form larger anions is not followed indefinitely, because the
apex sharing quickly leads to very large sizes and, consequent upon equation
(5.13), to a decreasing lattice energy to stabilize the structure. There is also a
restriction on packing because the Si-O-Si bonds have a preferred angle of
180° (2.2), and this leads either to too much strain in the anion if they are to
assume a compact shape, or to a very unusual shape to pack into a crystal.
(b) Extended anions
(i) Single chains. Condensation of n(SiO,)*~ units at two apices gives rise to
a chain sequence (SiO,)?~ found in pyroxenes (1.5.2). The chains shown in
Fig. 1.7 can be packed into a crystal with M”* ions forming the ionic bonds
holding the crystal together. Structural variation frequently arises when the
different coordination requirements of the M”* ions (5.3.7) cause variation in
the chain packing.
(ii) Double chains. Condensation can be achieved at an average of 27 oxygens
per tetrahedron to produce the infinite double chains shown in Fig. 1.7 and
encountered in amphiboles and other asbestos minerals. The double chains
have the composition [(Si,O,,)°~],,.-
(iii) Sheet silicates. Condensation of n(SiO,)*~ units at all three oxygens in
the plane of the equilateral triangle, that is any one face of the tetrahedron,
8 CERAMIC SCIENCE

produces a 2-D sheet with overall composition [(Si,O;)?~ ],- These sheets
form the basis of many of the flaky minerals discussed below as clays (1.5.3).
(c) Three-dimensional covalent networks
When all four oxygens in tetrahedra are involved in the condensation, then
the large family of feldspars, granites, man-made zeolites (1.5.4) and glasses
(chapter 3) is formed.
This classification is summarized in Table 1.1. As the materials change
from possessing strong non-directional ionic bonding to three-dimensional
“siant-molecule” covalent bonding, there is a fall and subsequent rise in
hardness. Clearly if the Si:O ratio in a material is known, then Table 1.1
allows classification to be made and, as shown shortly, a good idea of the
crystal structure will be gained. Expression of the composition of a ceramic
in the old form as the ratio of oxides, for example 2MgO -SiO, for forsterite,
is inhibiting, suggesting some uncertainty about the material; expressed as
Mg,SiO,, however, we can see the Si:O ratio of 4 indicates an orthosilicate
which must contain discrete SiO{~ anions and have an isotropic 3-D
ionic bonding mechanism producing a hard, refractory material. Similarly,
diopside, CaMgSi,O,, having a Si:O ratio of 1:3, will contain linear single
chains, and any amount of isomorphous replacement of Ca?* or Mg?* will
not confuse this conclusion.
The frequent appearance of aluminium in the composition can lead to
alternative suggestions. For example, there is little difficulty with albite
(NaAlSi,0,) because the Si:O ratio is not encountered in the classification,
but the (Al+ Si): O ratio of 4:8 suggests a 3-D net framework, indicating that
Al?* is a network cation and Na* is a charge balancing cation, making this
a predominantly covalent solid. A similar argument could be applied to
anorthite (CaAl,Si,O,), suggesting that it is in the same class and iso-
morphous with albite; but alternatively the Si:O ratio alone at 1:4 predicts
an orthosilicate. The complete range of solid solutions between albite and
anorthite strongly suggests that anorthite is isostructural with albite and°

Table 1.1 Silicate families classified according to SiO, tetrahedra condensation products

SiO, Condensation Charge per Hardness

Unit structure per Si Si Example (mhos)

ESO = monomer 0 4 orthosilicates 5-8

St Ones dimer 4 3 pyrosilicates 5
[SiZOslce trimeric ring 1 2 benitoite 6
[Si,0,,]8~ tetrameric ring 1 2 axinite 5
[SizOy nace hexamer ring 1 2 beryl, emerald 4-7
[(SiO3),17"— linear chain 1 2 pyroxenes 4-7
K(SL,O >, e2s — “chain ofrine it 14 amphiboles 5
: oe 7 talc, mica,
[Si,05] sheet iS 1 ashestosclays 1-3
[Si,0,] net framework 2 0 (granites), zeolites 3-5
SiO, net framework 2 0 quartz 8
 TRADITIONAL CERAMICS 9
therefore is not an orthosilicate. Thus, however the formula ‘is quoted, either
the old way of oxide ratios or the crystallographic way intended to show the
silicate or aluminosilicate grouping, the important step is to identify the
possible Si:O or (Si+Al):O ratio. When the family has been identified from
the appropriate ratio, then the broad features of the crystal structure can be
inferred. Within any one of the limited number of silicate-aluminosilicate
families the following variables are responsible for final minor structural
differences:
(a) Coordination requirement of charge balancing cations (5.3.7).
(b) Variations from the ideal tetrahedral angles as predicted in (2.2) in order to accommodate
the needs in (a).
(c) Some variation in the Si-O-Si bridging angle, which from (2.2) would ideally be 180°;
usually values from 120—160° are encountered.
(d) Rotation of tetrahedra about common shared oxygen atoms, from eclipsed to varying
degrees of staggered, as shown in Fig. 1.4.
(e) Tilting of the tetrahedra about the triad axis.
(f) When SiO, tetrahedra are partly substituted by Al, alternatives can arise, depending on
whether the replacement is random or ordered throughout the structure.
Examples of (a)—(f) can be found in the following paragraphs where some
of the more common and important silicates and aluminosilicates are
discussed. It should be noted, however, that from this analysis wide variations
would not be expected in the types of crystalline phase encountered in
traditional ceramic products, all the more so because the layer silicates, which
are traditionally used because of their moulding and forming properties,
when subjected to heat treatments transform to the extreme family members,
either ionic or covalent, as listed in Table 1.1.

1.5.1 Silicates and aluminosilicates containing isolated anion groups

The forms and relative sizes of the isolated silicon-oxygen groups found in
ceramic systems are shown in Fig. 1.3. Direct evidence for these groupings
comes mainly from detailed X-ray analysis, but chemical methods have
proved useful, especially where amorphous silicates such as glasses (chapter
3) and metallurgical slags are encountered. The method was introduced by
Lenz in 1964, and is known as trimethylsilylation. Silicates that can be made
soluble in acid are leached in the presence of chlorotrimethylsilane to yield
stable trimethylsilyl derivatives, such as SiO,[Si(CH3)3],, that are identified
by chromatography. Good supporting evidence for the X-ray work has been
proved in the cases of hemimorphite, Zn,(OH),Si,0,:H,O, aluminium
ordering in melilites, and the phases present in Portland cement pastes, (1.5.6),
and the method has given independent evidence for the non-network
structure of some glasses (3.2).
(i) Phases containing isolated SiO, tetrahedra. These are known as ortho-
silicates, and amongst the best known are the olivine group. The group name
is applied to members of the forsterite-fayalite series (Mg,Fe),SiO,. Structural
features can be envisaged in a number of ways: first, close packing of SiOZ~
such that the tetrahedra point alternatively in opposite ways along the a and
10 CERAMIC SCIENCE

b crystal directions, and occupation of interstices so produced by the Mg?’*

and/or Fe?*. Arising from this close packing it is possible to discern that the
oxygen atoms are approximately in two types of sheet, alternating ABAB, to
produce an arrangement close to that of hexagonal close packing. A second
possible description is occupation of tetrahedral sites in the close-packed
oxygen array by silicon and occupation of the octahedral sites by Mg?* or
Fe2+. Two kinds of octahedra are occupied by the Mg’*, with slightly
different dimensions. Given the general formula A,SiO, the similarity of this
description to the spinel structure (5.9) becomes obvious.
Since there are two kinds of cation-occupied octahedra in the structure, a
third description is sometimes used: Mg? *-filled octahedra of one kind share
edges with one another to form ribbons parallel to [001], and the ribbons are
linked together by Mg-filled octahedra of a second kind by sharing edges.
Within this 3-D structure, the Si atoms occupy interstices to form tetrahedra
that also link the MgO, ribbons in the (010) plane. On this model the sharing
of edges by octahedra causes a shortening of these edges (5.3.7), with the
result that the oxygen atom layers are drawn from perfect hexagonal packing.
In the olivine series, when Fe** is replaced by Ca?* the mineral monti-
cellite, CaMgSiO,, results. In natural samples the Ca : Mg ratio is close to 1:1
in all cases, mainly because this silicate prefers to be ordered with Mg?*
occupying only the same type of alternative octahedra so that they are at the
centres of symmetry. In oxides, Ca** has a radius ratio (5.3.7) such that it is
predisposed to a coordination number of 8, and this seems to introduce a
strain in the bonding, making this olivine softer than others in the group.
Continuing with this argument, it is not surprising to find that complete
replacement of the octahedral cations by Ca** leads to even greater
instability in y-Ca,SiO,, where the presence of the large Ca?*, the tendency
to 8-coordination, and the resultant “stretched” SiO{~ tetrahedra, lead to
easy transformations and the existence of several polymorphs, some of which
have cementing properties (1.5.6).
Two ions, Zr** and Th**, that must be packed in such a way that they
have eight oxygen neighbours, constitute a special tetragonal sub-group of
orthosilicates consisting of zircon, ZrSiO,, and thorite, ThSiO,. Zircon,
as well as having gem qualities, is used in refractories and zirconia
production (5.5).
The third group within this classification is the garnet group consist-
ing of two solid solution series: the pyralspite series named after its
members pyrope, Mg3Al,Si,0,,, almandite, Fe,Al,Si,0,,, and spessarite,
Mn,Al1,Si,O,,, and the ugrandite series containing uvarovite, Ca,Cr,Si,0, 5,
grossularite, Ca,Al,Si,0,, and andradite, Ca,Fe,Si,0,,. Two series do
not form solid solutions with each other. Structural variations occur because
the SiO{~ tetrahedra are arranged to provide 8-fold coordination for Ca**,
Fe?*, Mg?* and Mn?* and octahedral coordination for Al?*, Fe** and
Cr**. As well as these natural minerals, many man-made garnets are
 TRADITIONAL CERAMICS 11

known (5.9). The garnets are cubic and all are hard and brittle with no
distinct cleavage planes.
The structure appears to be complex because of the large size of the cubic
unit cell (1.15—1.20nm) and the various kinds of cation coordination to be
encompassed. It can be regarded as composed of corner-linked SiO{7
tetrahedra and Al octahedra, with Ca”* in interstices having 8 oxygens at the
corners of a cube. In grossularite the unit cell contains eight Ca,Al,Si,0,,
formula units, within which each oxygen atom takes part in the coordination
requirements of one Al?*, one Si** and two Ca?*.
An important group of minerals and raw materials containing the isolated
SiO, group, but also other oxygen ions, is known as the subsaturate group
and includes the important minerals kyanite, sillimanite, and andalusite, all
with the same composition Al,OSiO,. All three sillimanite minerals are
refractory themselves, occur in useful deposits in India, Africa and the U.S.A.,
and when heated to fusion yield mullite. The porcelain made from them has
a high melting point, low thermal expansion and low electrical conductivity,
making it ideal for spark plugs, refractories and cement-kiln linings. Their
structures are related, containing principally AlO, octahedra linked by the
isolated SiO, tetrahedra that bring them into our particular classification at
this point. The simplest structure is found in sillimanite, which has chains of
AlO, octahedra sharing edges with AlO, octahedra above and below to form
chains parallel to the c axis at the unit cell corners and centre. Alternate SiO,
and AlO, tetrahedra link the chains sideways. Andalusite has the same
octahedral chains but the sideways links are formed by the SiO, tetrahedra
and an unusual polyhedron in which Al has 5 oxygen neighbours.
Ions other than oxygen can be present in members of the subsaturate
group—a common example is topaz, Al,(SiO,)F,. Here the SiO, groups
provide four oxygens around each Al in the structure while two fluorines
complete the six near neighbours for each Al.
From the examples given of isolated SiO, silicates, it is possible to see the
major factor in determining structural variation is (a) in the list above,
namely the changing coordination requirements of the cations. This same
factor plays a decisive part in the solidification and solid solution behaviour

a
—_—

9 Loeb ie a

iomratntre : ia
Oo 0 4
Li
(a) (b) (c)

Figure 1.4 Two symmetries of [Si,O,]°~: (a) staggered, (b) eclipsed, (c) packing of [SiOrice
anions in thortveitite.
2 CERAMIC SCIENCE

of these silicates. Hence willemite, Zn,SiO,, and forsterite, Mg,SiO,, in

which Zn2* has tetrahedral requirements and Mg’* octahedral, form limited
solid solutions in a peritectic system. The actual morphology of such a system
as willemite-forsterite is then dependent on the rate of cooling giving rise to
non-equilibrium phase assemblages, since the rate of reaction to form the
solid solution is slow.*°
(ii) Double tetrahedral silicates. The (Si,O,)°~ pyrosilicate unit (Fig. 1.4) is
made up of two tetrahedra sharing one oxygen; the remaining six oxygens in
the unit carry one negative charge each which must be balanced by the
positively charged ions that hold the crystal together. Because the (Si505)°"
unit has an unusual shape for packing into crystals the structures are
complex.
Some of the possible variations that might occur in the pyrosilicate anion
itself may now be considered. Since the tetrahedra are joined at an apex,
rotation about the common oxygen is possible between the two extremes of
staggered and eclipsed (Fig. 1.4a).
Variation can also occur with respect to the Si-O—Si angle at the bridging
oxygen; angles in the range 100-180° are found. These variations fall
naturally into two groups.
(i) In the thortveitite group the conformation is staggered, and the Si-O-Si
angle is in excess of 140°, reaching 180° in thortveitite itself. As a result, the
bridging oxygen is shielded and does not take part in satisfying the coordina-
tion needs of charge balancing cations in the crystal. Oxygen p-orbitals
available for z bonding influence the bond angle as shown in (2.2).
(ii) In the dichromate type, the pyrosilicate group is eclipsed, the angle is
< 140°, the bridging oxygen is normally within the first coordination sphere
of at least one cation, and one cation bonds across both halves of the
pyroanion.
Balancing cations in the crystal and their coordination numbers determine
to some extent which type of pyrosilicate arrangement occurs. Generalizing,
it seems that cations with radii greater than 0.10nm cause adoption of '
dichromate characteristics and smaller cations produce the thortveitite
arrangement.
The classic member is thortveitite itself, Sc,Si,O,, a rare mineral in which
each scandium atom is surrounded by a somewhat distorted octahedron of
oxygens. Except for the oxygen at the centre of each (Si,O,)°~ group, each
oxygen is coordinated to one Si and two Sc ions. In the structure the
(Si,O,)°~ groups form dovetailed columns along the a axis of the cell, as
shown in Fig. 1.4.
Another example in this group which is of some importance as a zinc ore
is hemimorphite, Zn4Si,0,(OH),:H,O; the Si-O-Si angle in this pyrosilicate
is 151° and these somewhat bent units are packed at the corners and centre
of the orthorhombic unit cell. Each Zn** has tetrahedral coordination
consisting of three oxygens and one OH.
 TRADITIONAL CERAMICS 13

The largest group in this class of silicates is the melilites, which include
a complete solid solution series between gehlenite, Ca,Al(AlSiO,), and
akermanite, Ca,Mg(Si,O,). Many synthetic melilites are known, for example
akermanite is usually found in slags, and in nitrogen-containing ceramics
(2.12), and as a result the more general family composition is given as
(Ca,Na),(Mg,Al,Si),O-. All the melilites are in the dichromate class, and the
structure can be viewed as twisted layers of (Si,O,)°~ parallel to (001) held
together by Ca** or Na*. Bonding between the sheets appears to be weaker
than the bonding in the sheets, resulting in moderate cleavage parallel to
(001). When the structure contains Mg?* it is in tetrahedral coordination,
which introduces a potential weakness, since Mg”* prefers 6 oxygen neigh-
bours and as a result only moderate temperatures are required for trans-
formations, whence Ca,Mg(Si,O,) transforms to the orthosilicate garnet,
monticellite (1.5.1), and chain silicate wollastonite (1.5.2) on moderate
heating.
(iti) Epidote silicates. This is a complex and interesting group because it
contains both (SiO,)*~ and (Si,O,)°~ discrete anions, packed together and
ionically bonded with Al°*, Fe?*, Mn** and Ca?*. Epidote itself has the
composition Ca,Fe(Al,O)(OH)(Si,O,)(SiO,) and so from this formula must
contain features of all the groups so far discussed.
(iv) Silicates with Si,;0§~ ring groups. A silicate ring is formed when each
SiO, tetrahedron shares two oxygens with adjacent tetrahedra (Fig. 1.3). The
rings are held together in a crystal by cations coordinated with the oxygen
ions of the ring, but because of the relatively large size of the silicate anions
these ionic bonds are rather weak, (5.3.4), as well as being uniformly
distributed, with the result that these silicates do not exhibit well-developed
cleavage. Benitoite, BaTiSi,0,, and wadeite, K,ZrSi,;O., are “idealized”
examples of this class of silicate, idealized because silicon can be partially
replaced by Al, zirconium by Ti, Mg and Na, and the potassium by Ca, Ba
and Na. In these minerals the rings are arranged in layers with two layers in
the unit cell, and differences arise in the relationship of the rings in the two
layers; in wadeite one ring is related to another by a c/2 translation and a
180° rotation. Large cations like K* lie above and below the centres of the
(Si,;0,)°~ rings. Zirconium ions have a regular octahedral grouping of
oxygens, each being supplied by a different Si,0§~ ring, so that they bond
the structure vertically as well as horizontally.
(v) Discrete (Si,O,,)°®~ rings. This is not acommon group of tetrahedra in sili-
cates with isolated anions, although the four-tetrahedron ring is morecommon
in framework structure silicates. Axinite, Ca, (Fe? *,Mn)Al,(OH)BO,Si,0, »,
is one reported member. The anions which determine the structure can
be seen as two pyrosilicate units sharing an edge to produce a square ring
anion in Fig. 1.5a. Packing of the (Si,O,,)®~ anions with their general
plane parallel to (010) and all parallel to each other is the main structural
feature.
14 CERAMIC SCIENCE

O Be

@aA

A Si OQ,

A Tetrahedra

{a} (b]

Figure 1.5 (a) tetrahedral model of ring silicate, [Si,O,,]®~. (b) Beryl structure.

(vi) Sig0, ring silicates. This class contains not only silicates with elegant
structures, but also the common industrial source of beryllium and BeO as
well as many types of gemstones arising from the main mineral member,
beryl, Be;Al,Si,O,,. It also contains a most important ceramic composition
and structure in cordierite, Al,(Mg”*,Fe** ),(SisA1)O,3.
Beryl is found in granites and may form crystals weighing up to 200 tons.
Clear coloured varieties are emerald, aquamarine and heliodor. It has a
hexagonal structure with a = c = 0.919 nm, which when projected on to the
basal plane reveals the channels running through the structure formed from
positioning (SigO,,)'?~ anions directly above each other and then rotating
them by 30° as shown in Fig. 1.5b. Hexagonal rings formed from six SiO,
tetrahedra form sheets at 0, c/2, c with the Si and shared oxygens lying in the
planes at these heights in the unit cell. Aluminium ions and Be** lie on sheets
halfway between the (Si;O,,)'*~ sheets in 6-fold and 4-fold coordination
respectively. These ions link oxygens from sheets above and below, thus
producing vertical and horizontal bonding. The large open channels in the
structure can be used as sites for atoms and molecules in forms of beryl with
complex compositions.
A closer examination of the structure dispels the view of great regularity
gained from the projection shown in Fig. 1.5b. For example, the SiO,
tetrahedra have Si-O lengths 0.159 and 0.162 nm, O—Si—O angles of 108.24°,
108.42°, 110.40° and 110.88° and the Si-O-Si angle about the shared oxygens
is 168.2°; the BeO, tetrahedra are very distorted from the ideal, containing
angles of 90.90°, 131.40° and 108.85°; the AlO, octahedra have similar large
variations from ideal angles. These are the usual conditions for a material to
have an anisotropic thermal expansion coefficient?° and it is not surprising
to find that beryl has this property. When beryl is heated, the a axis expands
linearly in the range 0—800°C, but the c-axis contracts in the range 0—300°C
before expanding thereafter; average expansivities are 2.6 x 10° °C“! for the
aaxis and —2.9x 10~° C~* for the c direction. Average thermal expansion is
low and so thermal shock resistance is high. These desirable properties are
 TRADITIONAL CERAMICS 15
bestowed on some traditional ceramics by developing a structurally iso-
morphous phase, cordierite, from the standard raw materials of the industry.
Cordierite and mullite are two representatives of the isolated group silicates
which have an important role in determining the properties and micro-
structures of many ceramic products.
(vii) Cordierite and mullite ceramics. A formulae comparison between cordierite
(Mg,Al,Si;O,,) and beryl shows how closely related their structures must
be:
beryl Be, Al, Sig Oe
cordierite Al, (Mg, Fe?*), Si5Alm Onn
The inclusion of one Al per anion ring introduces a change in the tetrahedral
distortion, as too does the Al** in the tetrahedral sites bonding the anion
sheets vertically and laterally. These changes produce only small effects on
the thermal expansion behaviour outlined for beryl;'> the a axis expansion
for beryl and cordierite is nearly identical, but the c-axis contraction is
approximately twice as large and does not change sign until 400°C. Because
of variations in AlO, :SiO, ratios and resultant variations in the polyhedral
building units, three structural modifications are known:
a-cordierite, which is the natural and high temperature form.
B-cordierite, a low temperature form made from « or p below 830°.
u-cordierite, a metastable form found in the recrystallization products of
some glasses heated to 800—900°C (see 4.6).
Heating the y-form in the range 925-1150° produces the a-form with a
volume increase that improves the toughness (2.13) of certain glass ceramics
(4.6). As a result of the thermal expansion effect, ceramics that are either pure
cordierite, or contain high percentages of the phase, have wide applications
as refractories with high resistance to thermal shock and as high frequency
electrical insulators.
Pure cordierite is formed in a solid state reaction at 1000—1400°C between
MgO, Al,O, and SiO,. Much work has gone into using the layer silicate raw
materials because of their moulding characteristics, but this has all the
difficulties to be described in 1.5.7. Reaction follows the routes below which
give rise to cordierite in a mixture of decomposition products.

Talc +Al,0,1°SMgSi0,+SiO,2%>Mg,
Al,Si,O,, + MgSiO;
Kaolinite + Al,O;+“>spinel +SiO,+mullite—+Mg,Al,Si;O,, + spinel + mullite
Benitoite +Al,0,4+spinel +SiO, + mullite >Mg,Al,Si;O,,
+spinel + mullite.

Overfiring in excess of 1200° increases the mullite, forsterite and glassy

phases, thus increasing the thermal expansion coefficient of the product,
whilst underfiring, below 1170°C, produces insufficient cordierite. Addition
of PbSiO, or BaCO, fluxing agents extends this narrow working range but
produces more glass, which adversely affects the electrical resistivity.
Cordierite ceramics cannot be used above 1460°C because of the in-
congruent decomposition to mullite.
16 CERAMIC SCIENCE
Mullite, (3A1,0;-2Si0,—2A1,0;-SiO,) is the only aluminosilicate stable
at very high temperatures, and is therefore common in porcelain and
refractories and has since 1960 assumed a major role as a high-temperature
material. In fact, even though special ceramics (see chapters 2 and 6) and
binary oxides (chapter 5) have been rapidly developed, on a tonnage basis
alone the Al,O,—-SiO, system is the most important binary system in
high-temperature technology.
Commercially it is prepared by fusing alumina (5.4) with silica-gel (1.5.5)
to form a product with either a granular or an acicular habit. Inevitably when
formed in the presence of a liquid phase it adopts the acicular habit, the
needle-like crystals common to potteryware.
As long ago as 1924, Bowen showed that mullite was a phase in the
Al,O,-SiO, system, and ever since it has been the centre of study and
controversy which has focused on the following questions:
(i) Does it melt congruently?
(ii) What is its composition?
(iii) What are the details of its structure? (Since its composition has been in doubt, this has
taken the longest time to resolve.)

One factor contributing to these problems is the marked volatility of SiO, at

high temperatures in reducing conditions, due to the formation of SiO. How-
ever, the main difficulty seems to lie in the range of non-stoichiometry the
phase can exhibit and in the 2/1, (2Al,0,:SiO,), composition which,
although metastable, is commonly formed by a topotactic process, (1.5.7),
below 1500°C.
It seems that a range of mullites exist between the limits 3A1,03/2Si0O, to
2A1,0,/SiO,, with the equilibrium composition lying in the 60-63 mole %
Al,O, range. A range of lattice parameters have been reported for the
orthorhombic unit cell, centring on a=0.750nm, b = 0.768nm and c =
0.2885 nm, with variations depending on composition and heat treatment of
the sample. The heat treatment change is thought to occur through Al-Si

Mullite+Liquid
2000
Bees
Al,0.
Mullite (ss) 1850 4 +Li 8aay
Las2ar
sie oh ° ne
o 1840
it
oS AS ~o /
1700 miLOs
oi
Eke
Isolated Jae Ae
ZY,
Mullite (ss) SiQ, & ty eae
AIO oo oso
ietranoara
1400. alternating 1

Al,O, Mole % AlOg Octahedra

(4) (b)

Figure 1.6 (a) The Al,O,—SiO, binary system. (b) A simplified view of the mullite structure.
 TRADITIONAL CERAMICS 17
disordering in tetrahedral sites at high temperatures. A simplified view of the
structure showing the SiO, group is given in Fig. 1.6b. A commonly accepted
binary diagram is shown as Fig. 1.6a, in which the steep, then flattened,
liquidus line shows considerable non-ideality near to the equilibrium mullite
composition, probably reflecting the presence of large SiO ,-rich polyanions
in a less viscous Al,O,-rich solvent.
By analogy with sillimanite, (see (i) above), the composition of the mullite
solid solution phase is written as

AMLSI2 ALO Gi) <5]

where x varies from 1.25 to 1.40 for mullites in the 3/2 to 2/1 range.
For example, 3Al,0,:2SiO, has x =1.25 and the formula becomes
Al(Sip 75Al, 2504.75). The analogy with sillimanite follows from early X-ray
work, but it was not until 1962, when a synthetic crystal of 1/2 mullite was
analysed, that the way it could be derived from the sillimanite structure was
seen. The unit cell of the crystal examined had a formula unit of Al, gSi, >Oo¢
which compares to a corresponding volume of sillimanite at Al,Si,O,. Thus,
while both structures have the same number of cations per similar volume,
the mullite is short of oxygen. The mullite structure is derived from the
sillimanite structure in the following way:
(a) 0.8 Si per sillimanite unit cell is replaced by 0.8 Al at random.
(b) 0.4 tetrahedra linking oxygen per unit cell are removed which “destroys”
20% of the tetrahedral cation positions and therefore Al?* that were
in these sites must move to positions close by to regain an approximately
tetrahedral coordination.
(c) 0.8 Si** is replaced by Al?* in order to compensate for the charge lost
when the oxygen vacancies are created.
When sillimanite is heated it transforms to a 3/2 mullite and silica, and since
this is true also for many metastable low Al,O, mullites, refractories made to
contain the phase aim at 3/2 compositions. A good mullite refractory should
contain an average of 85% mullite, 3-6% unreacted Al,O, and a glass
content no greater than 5%. Furthermore the glass should have a high SiO,
content in order to have high viscosity (3.1). The ASTM classification of a
mullite refractory specifies Al,O, contents in the 56—79 % range, maximum
impurities of 5% and a load deformation at 1595°C no greater than 5%.
Refractories conforming to these specifications are widely used in glass tank
linings, electric furnace roofs, hot metal mixers, and pottery kiln load-bearing
slabs. Since mullite is so important and ubiquitous in traditional ceramic
ware, some of its properties are collected in Table 1.2.
From Table 1.2 it should be noted that mullite is not stable in vacuum and
must be used considerably below its melting point. Decomposition in
atmospheres with low oxygen potentials produces Al,O, particles in the
18 CERAMIC SCIENCE

Table 1.2 Mullite properties

Property Value

Theoretical density (kg:m7

°) 3.16-3.22 x 103
Hardness (mhos) TES)
Crystal habit acicular
Chemical nature slightly acid
Working limit °C
in air 1800
in vacuum 1500-1700
M. pt. °C 1850
Stability
in reducing atmosphere only fair
acid slags good
acids insoluble
Young’s modulus GN m~?
12 % porosity 160
0.3 % porosity 220
0.3 % porosity at 1200°C 48
Flexural strength MN m~?
5.0% porosity ey
2.0 % porosity 132
hot pressed zero porosity 240
Creep at 8.27 MNm ~? 1100°C
m hr? 9.8x10~?
Deformation under load (%)
0.172 MNm ~” at 1850°C Z
0.345 MNm ~” at 1450°C 0
Expansion (x 107° K~*)
25-1500°C 5.6
Thermal conductivity Wm~!K_~? at 1200°C 0.543

sample. Direct evaporation of SiO, at high temperatures is ruled out because

the material is stable to much higher temperatures in the air. Decomposition
is believed to follow the sequence
3A1,03 ‘2810 9s) = 3A1,036) SF 2Si0 @) a Or) (1.8),

2810g)+ Org) = 2810 x). (1.9)

For (1.8) and (1.9) we can write, following (5.2)
AG° = —RT2.303 log [(Psio)” x (Po,)] (1.10)
and
AG? = RT 303 log PSO.) (1.11)
(Dsio) (Po,)
and values calculated from available data give at 1650°C for (1.8)
AG® = 642.9kJ mol”* when log [(psio)” * Po,] = — 17.50
and for (1.9)
AG° = —123.1kJmol~* and _ log(psio,)?/(Psio)? (Po,) = 3.36.
 TRADITIONAL CERAMICS 19
These can be used to demonstrate the role played by ambient oxygen
pressures on the decomposition of mullite. For example in air at 1650°C,
Po, = 9.2 atm. and from (1.10) we can write
— 17.50 = log pigio)0.2
which gives psig = 1.25 x 107? atm. Using this value in (1.11) gives

386 DSi
1og=
Psio Po,
which is equivalent to
3.36 = log pio, + 17.50
from which psi9, = 1.18 x 1078 atm.
Now in a vacuum, the oxygen partial pressure will be controlled by
refractory metals used as heaters or heat shields, for example at 1650°C in the
presence of a tantalum heater the oxygen pressure will be 3.89 x 107 '> atm,
and a repeat calculation gives psi9 as 9.02 x 10° atm. The SiO, pressure
will of course remain 1.18 x 10°8atm. From this calculation it appears that
in air the psig near the sample must reach 1.25 x 10~? atm very quickly to
prevent further decomposition. In a “vacuum” however, the equilibrium
pressure of SiO, that is 9.02 x 107? atm, is quite high and is not reached,
hence the mullite decomposes. This is responsible for the limitations on the
use of this refractory in vacuum or reducing conditions. Simple calculations
of this kind are instructive in considering the uses of ceramics, and several
examples are given in the text, e.g. (2.8), (6.3) and (5.2) contains equations
plus a table of data for such uses.
Finally, as the uses for mullite expand, some attention is being given to
producing pure stoichiometric 3/2 mullite.2* A method using «-Al,O; (5.4)
and amorphous SiO, (1.5.5) in a gel mixing technique followed by firing at
1420°C has been described. A co-precipitation method based on organo-
silicon hydrolysis is also being considered:
AICI, +3NH,OH = Al(OH), +3NH,Cl
Si(OC,H;),+4H,0 —Si(OH), +4C,H,OH.
After filtration the mixed precipitates are heated to 1400°C. This type of gel
preparation method is being more widely applied in ceramic and glass
preparation procedures: (3.1), (5.5).

1.5.2 Chain silicates

In the classification of silicates summarized in Table 1.3, two groups of chains
arising from infinite condensation of the SiO, units at 2 or 25 oxygen were
identified. These are considered as two groups: the pyroxene single chain
silicates and the amphibole, double chain silicates, the dimensions and
structure of which are shown in Fig. 1.7.
20 CERAMIC SCIENCE

Figure 1.7 (a) Idealized single chain [(SiO;)* ~],,. (b) End-on view of single chains. (c) Idealized
double chain [(Si,O,,)°~],. (2) End-on view of the double chain.

Though the chains are usually pictured in pyroxenes as straight, in real

crystals they are commonly twisted to accommodate the coordination
requirements of the cations binding them into a structure. The most common
binding ions are Mg”*, Fe** and Ca?*, which have to be located close to the
two non-shared oxygens per tetrahedron at which are located the two
negative charges per SiO%~ unit. The single chains, as well as being twisted,
can be packed together in different ways to accommodate cation size and
charge requirements, and as a result the pyroxenes are more abundant than
ring silicates (1.5.1), and have wide variations in composition. However,
despite the compositional variations they have only a limited range of
properties. Pyroxenes of many compositions form in industrial processes,
when layered silicates are heated (1.5.7), and in glass ceramics (4.6). Some of
the more common and important ones are discussed below and their solid
solution relationships are shown in Fig. 1.8. Spodumene, LiA1Si,O,, which is
of considerable importance in traditional and the newer glass-ceramics has
a non-chain cation composition quite different from the more common
pyroxenes shown in Fig. 1.8.

CaSiO, Wollastonite

No
solid

(cm)
solution

‘Augite solid solutio: (care Si03

Diopside Hedenbergite
No solid solution

MgSi0, Hypersthene solid solution FeSiO,

Enstatite Ferrosilite
(MaFe)(sio,)
2

Figure 1.8 Common pyroxene composition relationships.

 TRADITIONAL CERAMICS ol
Because the covalent bonds within the chains are stronger than the ionic
forces provided by the charge balancing cations, some anisotropic behaviour
is expected and is observed in the characteristic cleavage of pyroxenes.
Cleavage occurs parallel to the chain lengths and produces a characteristic
angle of 93°. This can be visualized by examining the structure in projection
on the (001) plane, when it has the appearance of the close packing of the
end-on units of the chain shown in Fig. 1.7. Such a packing is shown in Fig.
1.9, where it can be seen that cleavage occurs diagonally to avoid cutting
through the chains, with the result that cleavage planes are parallel to (110);
the 93° angle is then expected.
In the amphiboles the infinite chains shown in Fig. 1.7c contain hexagonal
rings, and it is found that these must be occupied by OH~, F~, or O2~ to
stabilize the unit. Thus amphiboles most commonly have the composition
[(Si,0,,)OH]’~ as the repeating chain links. The seven negative charges
must be neutralized locally. Because many different ions may be present and
many isomorphous replacements are possible, the compositional variations
in the amphiboles are remarkable. Once again, however, this variation is not
reflected in the physical properties.
A general formula for the amphiboles is A,_3B;(Si,Al)g0,(OH,F,O),,
where A and Bare cations that locally neutralize the chain-unit charges at
sites X,, X3, X, and X, in the diagram of the basal section of amphibole
shown in Fig. 1.9. The seven cation sites have sizes decreasing from the two
X, sites to the smallest (X}) site at the centre of the chain units. The A cations
in the general formula occupy the X, sites and they can be any mono- or
divalent cation with radius from Mg?* (0.078 nm) to K* (0.133 nm).
The B cations occupy the 2X,, 2X, and X; sites and can be any di- or
tri-valent cations of radius 0.057 nm (AI? *) to 0.091 nm (Mn?* ). Immediately
a bonding difference arises, because all the A cations on X, sites will be
8-coordinated, while the B cations have 6-coordination and therefore the A
cation to chain bonds are on average weaker. Orthorhombic amphiboles
have all A and B sites occupied by Mg?*, but up to three Fe** can go into
X, and X, sites without disturbing the packing. Transformation to the

x4
X2
X4
Taya Fe
0000000

0000000 X3
ay hy er A x2
0000000 0000000 X4
UEEENGNGS SoleS

Figure 1.9 Plan of basal section of the amphibole structure; X,, Xj, X3, X4 are cation sites with
size decreasing from X, to X3.
22 CERAMIC SCIENCE

clinoforms of these minerals, that is the monoclinic modifications, occurs

with the presence of four Fe?* which, being larger than Mg’*, occupy the
2X,+2X, sites with Mg** in the inner sites. When a large ion occupies X4,
then a small cation like Fe?*+ must occupy X,, followed by (say) Fe** on X,
and Mg?+ on X3, because of the size order. It is believed that this
juxtaposition of Fe** ---Fe?* in crocidolite, for example, is responsible for
the distinctive blue colour of this asbestos mineral, as an Fe**—Fe** charge
transfer occurs on sites X, and X3.
The projection of the chain ends on to (001) demonstrates, as for the
pyroxenes, an easy cleavage in the planes parallel to (110). However, due to
the elongated cross-sectional aspect of these chains, the angle between
cleavage cracks is 56° and the crack proceeds between X, sites and the chains
where the weakest bonding occurs.
(i) Examples. The simplest pyroxene with respect to composition is enstatite
(MgSiO;) which has three polymorphs, protoenstatite, enstatite and clino-
enstatite, of which protoenstatite is stable only at high temperatures and is
the only form containing fully extended chains. Enstatite and clinoenstatite
have chains slightly different in shape because they are not fully extended.
Diopside, CaMg(SiO;),, is a common silicate formed at high temperatures
from dolomite (5.6). The SiO, chains running parallel to the c-axis are held
by Ca?* and Mg?*, each of which have six non-shared oxygens from the
chain as their coordination sphere, but the Ca?* is also close to two oxygens
that link tetrahedra, giving its preferred environment of eight. Each chain
oxygen not involved in linking tetrahedra and on which the charge resides is
linked to one Ca?* and one Mg’*. The requirement of eight oxygens for
Ca?* causes the chains to be crumpled and puts this pyroxene into the
monoclinic group.
Spodumene, (LiAl)(SiO3)>, is the chief source of lithium from particularly
rich deposits in South Dakota containing crystals over 40 feet in length. It is
an important phase in several glass-ceramics (4.6) where its low thermal
expansion coefficient makes it a suitable crystal in products requiring good
thermal shock resistance. Although it contains no Ca?*, it is still a monoclinic
pyroxene because the small size of Li* requires considerable chain distortion
to provide a satisfactory coordination environment for that ion.
Tremolite, Ca,Mg5(Si,O,,),(OH,F),, crystallizes in the monoclinic system
as fibrous masses, like many amphiboles. It is known as Italian asbestos, and
is frequently employed in wall insulation, lagging, and in acid filtration
systems. In accord with the above rules relating to the position of the cations,
the large Ca”* ions occupy the X, sites where the 8-fold coordination
environment satisfies them and the Mg?* ions occupy the X,, X, and X,
sites.

(ii) Fibrous silicates

Many of the amphiboles have a fibrous habit that has led over many years to
 TRADITIONAL CERAMICS 23

ASBESTOS

CHRYSOTILE AMPHIBOLES
(white . asbestos)
Mg3[Si,0;|OH),

ACTINOLITE CROCIDOLITE
CaFeg(Si,0,] 9OHF), (blue asbestos)
Riebeckite
Na,Fe,|Si,0,,| (OH
TREMOLITE AMOSITE 7 s[SuPn]JM),
CagMag/Si40,1| (OHF), Grunerite
Fes.<Mg,. 5[Siqr1] 3(OH)

ANTHOPHYLLITE
Mg7[Siq0,,| HF),

Figure 1.10 Asbestiform minerals.

their widespread use in industry. As Fig. 1.10 shows, they form a major class
of the asbestos minerals. The texture of asbestos fibres can be quite variable,
but is generally related to the composition—those containing high percent-
ages of Mg, such as chrysotile and tremolite, form white silky fibres with
small diameters down to 0.01 um. At the other extreme is amosite with harsh
spiky brittle fibres, grey in colour and thicker than 0.1 pm, features related to
a high iron content. Crocidolite has a texture between these extremes and has
an intermediate iron content. Fibrous habit arises in the amphiboles from the
lath-like units in the unit cell comprising the double chains separated by the
bands of cations. Chrysotile, the most abundant asbestos, has a sheet silicate
structure (1.5.3) but the two-dimensional layer is rolled about an axis to form
a narrow tubular fibre. The production of chrysotile in excess of 4 million
tons from massive deposits in Canada and Russia far outweighs the pro-
duction of all other asbestos minerals.”°
Applications of fibrous silicates. The largest use of the fibres is in asbestos-
cement products such as flat sheets, tiles, pipes, and gutterings which usually
contain 10-16 % of asbestos. Of that percentage most is chrysotile, but up to
50% may be amphibole fibres when enhanced reinforcement is sought. These
natural fibres are mainly being used as fibre reinforcement (2.13). Amphibole
fibres are specifically used in reinforcing plastic materials for structural
applications because they have superior strength and chemical resistance
compared to chrysotile and glass fibres; many battery casings have amphibole
fibres as reinforcement. Amosite fibres up to 35% in an autoclaved calcium
silicate matrix provide good thermal insulating tiles. Other thermal insulators
contain amosite in magnesia (5.6), and in sodium silicate (1.5.5). The longer
fibres found in chrysotile are used in the manufacture of asbestos textiles for
B
24 CERAMIC SCIENCE

Table 1.3 Some properties of asbestiform minerals

Property Chrysotile Amosite Tremolite Crocidolite

Tensile Sekee iN) 3100 (RT) 2500 (RT) 5000 (RT) 3500 (RT)
1860(550°C) 325 (550°C) _ 700 (550°C)
Young’s Modulus (GN m ~”) 165 165 = 190
Flexibility Good Fair Brittle Good
Hardness (mhos) 3-4 5-6 aD 4
Density (kg m7 °) 2550 3430 3100 3370
Specific heat (kJ kg~*K~') 1060 780 1080 830
Acid resistance (4.0 M HCl)—
% weight loss, 2 hrs refluxing ey - : i
Colour Colourless Pale fawn Colourless Blue

safety clothes, belting and chemical plant filters; blue asbestos was also used
in these applications before the health hazards were known. The third major
use involves the favourable frictional properties, when embedded in phenolic
resins, to make brake and clutch-linings. Some of the properties that lead to
these uses are collected in Table 1.3.
Chemical and thermal reactivity. All forms of asbestos decompose when heated
in the range 600—1000°C to give products such as pyroxenes and silica, and
therefore reactions between asbestos as such and other materials do not
exist.17 For example, between 600° and 900°C amosite becomes a mixture of
the pyroxene hypersthene, magnesia, Fe,O,, and silica, and it is these
products that react with other phases. This is a demonstration of the principle
to be enlarged upon later (1.5.7) that thermal transformations of silicates in
the traditional ceramics area produce only a limited type of product and thus
only a restricted range of properties.
All asbestos types resist attack by refluxing alkalis but show only limited
and variable resistance to attack by acids. Acid resistance sharply demarcates
the chrysotile and amphibole asbestos groups, for even hot acetic acid will
destroy chrysotile fibres. Examination of the chrysotile structure (1.5.3)
places this weakness on the easy access the acid has to the (MgOH) layer,
which is then easily removed leaving a silica skeleton.
Variations in the acid resistance of the amphibole asbestos fibres can be
ascribed to the way in which the cations occupy the seven sites X ,—X, (see (i)
above); decomposition is a process of selective leaching of the cations:in these
sites, a process which becomes progressively harder from site X, inwards.
Tensile strength. Failure begins at surface flaws (3.4) and proceeds by the
rupture of weak interfibre bonds to expose new surface for fracture propaga-
tion. Quoted strength values are therefore sensitive to fibre length and the
values in Table 1.3 are averages for standard fibres of 4mm length and
10—20 um diameter. In order to realize the high strengths of these fibres,
approximately twice that of glass fibre (3.1), (2.1), good bonding to matrix
materials must be present. In this respect amphibole asbestos is superior to
 TRADITIONAL CERAMICS 25

the chrysotile fibre when embedded in polyester resins, because amphibole

surfaces have low polarities, like the polymer, whilst chrysotile has a polar
surface. Unexpectedly, from this simple model, the amphiboles are still
superior to chrysotile in cement matrices. A reason for this is believed to be
the provision of nucleation sites on the silica amphibole surface for the
process of cement hydration (1.5.6). Several micrographs exist which show
tobermorite, Ca;(Si;0,OH),-4H,O, growing at 90° from fibres of amosite,
an observation that supports the nucleation idea and suggests that the strong
matrix-fibre bonds are partly chemical and partly mechanical.

1.5.3 Sheet silicates

Here each SiO, tetrahedron shares three oxygens with adjacent tetrahedra to
produce a sheet extending infinitely in a plane. Each oxygen that is not shared
is at an apex forming a layer above the plane of the sheet, and each has a
negative charge. Clearly many silicates can be envisaged by stacking these
sheets together in various ways, and holding them together by inserting layers
of cations between them to neutralize the charge on the apical oxygen layer.
However, despite the wide formula variations, only a limited number of
arrangements do occur and these can be grouped into a few classes. It is
instructive to consider how this arises.
First, what variations in the sheet geometry occur? In Fig. 1.11 the two
alternative arrangements that are encountered are shown; of these only the
hexagonal arrangement is common in layer silicates, and that where the
tetrahedra form 4- and 8-membered rings is only encountered in a distorted
form in a few minerals such as apophyllite, KCa,(Si,0;),(F,OH):8H,O. In
the rare structures such as apophyllite, a more compact arrangement is
adopted by rotation of the rings formed from four shared SiO, tetrahedra.
The formation of 4-membered rings places two pyrosilicate units (1.5.1) ina
position where they must share a common edge, which introduces some

0-52 nm
——————

Figure 1.11 (a) The hexagonal type silicon-oxygen sheet. (b) The idealized tetragonal type
silicon-oxygen sheet.
26 CERAMIC SCIENCE

Figure 1.12 (a) Section of a sheet of SiO, tetrahedra linked to give (Si,O,)?~ in which the
vertices of each tetrahedron point upwards (heavy circles). (b) Close-packing of OH” ions in a
layer of Al(OH), or Mg(OH),. Distance between O™ ions in (a) is approximately equal to
separation of OH ions in (b).

strain in the structure (5.3.7); it is probably this effect that accounts for the
rarity of the tetragonal layer silicates. This then reduces the number of
possibilities for layer silicates and means that we need only consider the
hexagonal type of silicon-oxygen sheet.
Secondly, the simple sheets, those that contain only a silicon and oxygen,
are never just held together by sheets of cations. They do in fact only occur
with Al or Mg in the structure, but these are involved in a more intimate
bonding arrangement than just as alternating ionic sheets. The occurrence of '
such specific bonding occurs as a geological and geometrical coincidence, and
is outlined here with the aid of the sketch in Fig. 1.12. Above the hexagonal
sheet of linked SiO, tetrahedra, the “free” oxygens carrying the negative
charges form a layer of hexagonal rings as shown in Fig. 1.12a. The
hydroxides, gibbsite and brucite, Al(OH), and Mg(OH),, have a close-
packed structure in which Al** or Mg’** occupy octahedral interstices in the
close-packed OH™ layers (Fig. 1.12b). In the close-packed OH layers
six-membered rings can be distinguished with one OH group at the centre of
each ring. By coincidence, the O—O separations in the layers in (a) and (b) are
almost identical, and if (b) is superimposed on (a) then one OH™ will sit in
the centre of the hexagonal ring of O~ in the (Si,O2~ ), sheet and every O7
will have an OH in its coordination sphere. After this, condensation over
geological time, in which 150, molecules and 3H,O molecules per silicate
 TRADITIONAL CERAMICS D7]
ring are removed, produces a fused layer consisting of a tetrahedral layer with
OH, and an octahedral layer of Al?* or Mg?* and OH~. The positive
charges on the Al and Mg ions neutralize the sheet charges and the OH~ ions
that now form a layer below them to complete their octahedral coordination.
The now electrically neutral double layers are held together only by weak van
der Waal’s forces, making the materials very anisotropic with easy inter-layer
slippage, a greasy feel and excellent moulding characteristics in the presence
of water.
Thus there are two basic formulae for this fused double layer. The first class
emerges as the two-layer silicates, serpentine Mg,Si,0;(OH), and kaolinite
Al,Si;0;(OH),. The layers in the serpentine and kaolinite group are shown
in elevation in Fig. 1.13 in order to emphasize how the next group emerges.
In Fig. 1.13 the bottom layer of OH™ ions are still close-packed as in Fig.
1.12b and therefore can act as the original brucite or gibbsite with another
(Si,0;)*~ sheet if such a sheet is brought to the OH™ layer so that the
tetrahedral vertices are pointing upward. Then the same condensation
process will produce a triple layer: a tetrahedral layer with OH™, an
octahedral layer of Al?* or Mg?* and a repeat tetrahedral layer with OH.
These three-layer silicates comprise talc, Mg3(Si,0,),(OH),, and pyrophyllite,
Al, (Si,0;).(OH),. Again, since the three fused layers are electrically neutral
they are stacked together by van der Waal’s forces, and the resultant
properties are very similar to the kaolinite-serpentine group.
Variations in structure between members of the two groups can readily
occur by changing the stacking of the weakly bonded layers, thus causing
changes in symmetry with little variation in properties. This is how the group
membership is built up.
However, another way exists to extend the group of layer silicates. We
recall the common occurrence whereby Al** replaces a Si** (1.5), so

oO
Si
0,0H 8 8
}r
Mg orAl @ @ oy Corie

oH © @ 6 6 6 8

Figure 1.13 (a) The double fused layers in serpentine and kaolinite; (b) the three-fused-layer
structure of talc and pyrophyllite.
28 CERAMIC SCIENCE

(S1:05)" |,
Si,05)
2—

HEXAGONAL

Mg(OH),

SERPENTINE GROUP KAOLINITE GROUP

Mg3Si,0;(OH), Al,Si,0; (OH),

TALC GROUP PYROPHYLLITE GROUP

Ma,(Siz05), (OH), Al,
(8,05 ),(OH),

BRITTLE MICA GROUP MICA GROUP

CaAl,(siatos)(oH), KAL(5, ALO, 0

MONTMORILLONITES (Ca,Na)(Mg,Al,Fe), (Si,AIO.), (OH),

CHLORITES (Mg,Al,Fe)g(Si,AI0; ),(OH),
VERMICULITES (Mg,Ca )(Mg,Fe,Al\s(Al,SiO,)(OH\,

Figure 1.14 Group relationships in the hexagonal layer silicates.

requiring the inclusion of charge-balancing cations. Thus, the tetrahedral

layers no longer have the composition (Si,O,;)?~, but have a composition
close to (Si, sAly ;0;)?°~ or (SiAIO,)°~, and occasionally [(Si,Al),05]*~
when more general replacement occurs. This type of compositional variation
appears to be limited to the three-layer group with charge-balancing cations
of large size located between the layers, providing an ionic bond linking the’
layers. Although the interlayer bonding is now stronger, it is weak compared
to intralayer bonds, and these materials show obvious cleavage characteristics.
When K* is an interlayer cation, members are classed as micas, e.g.
muscovite, KAI,(Si, ;Alp ;0;),OH;. If Ca?* and Na* are balancing cations,
the materials are brittle micas, e.g. margarite, CaAl,(SiAlO;),(OH),. Fig. 1.14
shows the main groups of layer silicates stemming from the primary (Si,O;)*~
layer.
The serpentine and kaolinite group. The serpentine group is so named because
minerals with this structure have a mottled green pattern like that of a snake.
Much of the material we know as asbestos (1.5.2) is a serpentine called
chrysotile with the almost ideal composition Mg,Si,0;,(OH),. The fibrous
nature of chrysotile is an interesting feature of the structure, believed to result
from dimensional differences between brucite layers and Si—O sheets.
 TRADITIONAL CERAMICS 29

The basic structure, as explained, is a composite layer consisting of an

Si-O sheet on one side and an Mg—OH layer resembling brucite on the other.
These two layers have the following dimensions: Si-O (side) a = 0.516 nm,
b = 0.89nm; the Mg—OH sheet in brucite has a = 0.539nm, b = 0.93nm.
There is therefore a strong expansive force on the OH™ side tending to curl
each sheet into a cylinder with the Si—O sheet on its inside. Only weak forces
bind the sheets, insufficient to prevent the curling force from turning the
structure into a stack of curved sheets, or in the extreme a series of hollow
tubes. The inner and outer diameters of such tubes are typically 10.0nm and
25.0 nm. The a axis of the unit cell lies on the curved sheet, while the b axis is
perpendicular to the curve, with the result that structures of successive sheets
are in alignment in the a direction but this cannot be true along the b
direction. Thus the crystal can grow easily only along the a direction, either
because the layers are out of step, or because a full circle eventually results;
this accounts for the fibrous nature of chrysotile.
Variaticns on this curling effect are responsible for a number of forms of
serpentine. For example, if periodically the layers are arranged so that the
tetrahedra in the (Si,O;)?~ layers point either up or down, then a corrugated
wavelike microstructure will result, with a platey habit. This is found in
antigorites.
Despite the range of layer positions giving rise to the many family
members, the unit cell dimensions of serpentines are very similar; Table 1.4
compares them to kaolinites and micas. The a-axis and b-axis spacings are
those of the Si-O sheet common to all these structures.
At high temperatures serpentines lose water and recrystallize as a mixture
of the discrete silicate, olivine, (1.5.1), and enstatite (1.5.2). The nature of such
transitions is discussed more fully in (1.5.7). This mixture of products is
ubiquitous and fired serpentines make good consistent firebricks.
The kaolinite group includes kaolinite and three varieties with the same
composition but different variations of the same basic structural two-layer
unit. Kaolinite is commonly called china-clay, and all deposits in the world
are the product of the decomposition of feldspars in granite (1.5.4), a
decomposition effected by CO, and superheated water ascending from the
lower strata leaching out the potash in the feldspar. In Britain there are
massive, high quality deposits in Devon and Cornwall, which form an
important export. Kaolinite is an essential constituent in the bodies and

Table 1.4 X-ray lattice parameters of sheet

silicates

a(nm) b(nm) c (nm)

Serpentine 0.53 0.92 1.46

Kaolinite 0.516 0.890 1.451
Mica 0.518 0.902 2.009
30 CERAMIC SCIENCE
glazes of cups, plates, insulators, and whiteware. It is used widely in the paper
industry, as a filler in rubber products, paints and textiles, and more recently
as a source of pure alumina (5.4).
Kaolinite is the archetypal double fused layer material, and like the
serpentines the strongly polar nature of the layers influences the ability to
stack them together easily, due to the same bending tendency as described for
serpentine. The result is that the crystals of kaolinite are very small and
imperfect, and since they are plastic when wet, this material fulfils the criteria
for classification as a clay mineral (see below).
Other forms of kaolinite are dickite, nacrite and halloysite. Of these, nacrite
and dickite are most like kaolinite, except that the layer stacking is sufficiently
different to provide monoclinic symmetry with two layers in dickite in the c
repeat and six layers in the unit cell of nacrite. Halloysite is a more weathered
form, some examples of which have the tubular structure of serpentine, and
it is the one member of this group that does not become plastic when wet; in
general, samples of this mineral contain water molecules between the layers
and usually a more disordered stacking of the crystal planes. The different c
axes are: kaolinite 0.737 nm, dickite 1.442 nm, nacrite 4.3nm and halloysite
1.0 nm.
The action of heat on this group of minerals forms the basis of much of the
ceramics industry and is dealt with separately in (1.5.7).
Three-layer groups—talcs and pyrophyllites. The three-layer sheets, Fig. 1.13,
are electrically neutral in both series, and because the OH™ layer is
sandwiched between two (Si,O,;)?~ layers the bending tendency is removed.
Thus the stacking of the neutral triple layer sheets is easier and larger crystals
result. The van der Waal’s bonds are weaker between the sheets, so talc,
Mg;3(Si,0;),(OH),, and steatite or soapstone, Al,(Si,O;),(OH),, are better
lubricants than kaolinite. They are the softest of minerals, can be scratched by
a fingernail, with perfect cleavage in one direction and irregular fracture. Heat
effects transformation into hard, refractory mixtures following the dictates of
the crystal structure as discussed in (1.5.7). The structure bestows low’
conductivity of heat and electricity, resistance to acids, and (when powdered)
a good white colour. These properties account for their wide use in industries
other than the traditional ceramics industry. Many of these requirements are
met by supplies from the U.S.A. and, because the purest deposits are found in
association with dolomite (5.6), France and Italy are also important sources.
Mica and brittle mica group. Micas are amongst the most abundant materials
in the earth’s crust and are derived from the ideal structure of talc or
pyrophyllite by isomorphous replacement of Si** by Al?* in the tetrahedral
positions in the sheets. The (Si,O;)*~ layers on either side of the Mg(OH),
or Al(OH); layers are bonded in the way already described for this class of
silicates, but they are staggered relative to each other as shown in Fig. 1.15.
Most crystallize in the monoclinic system as tubular crystals dispersed as
“books” in rocks, or as clay particles in sedimentary rocks. There is perfect
 TRADITIONAL CERAMICS SM

i dodo obo eed

i brucite
H BP Soar S, |layer

=—A/A

0,0H 2 ° ° ° talc
9 ° 8 ° layer

(a) (b)

Figure 1.15 (a) The offset layers of the mica structure on the (010) plane. (b) The chlorite
structure showing the intermediate brucite layer. A and B show where water layers are found in
vermiculites and montmorillonites.

cleavage between the layers of double sheets by breaking the weak bonds
between K* and Na* and O~. Cleavage is not so easy when Ca** is present,
which accounts for the differentiation into the brittle micas. Mica is valued
industrially on account of its perfect cleavage, which enables it to be
separated easily into thin sheets with excellent electrical insulating properties.
India is the main supplier of sheet mica, but particulates come from many
sources.
Muscovite, KAI,(Si, ;Alp ;05).(OH),, is the commonest mica. Its structure
is derived from pyrophyllite by replacement of 25 % of the tetrahedral silicons
by Al?*. It cannot be used where the temperature exceeds 500°C. Biotite,
K(Fe,Mg),(Si, sAlp ;0;),(OH)>, is derived from the tale structure in
the same way that muscovite is derived from pyrophyllite. Phlogopite,
KMg,(Al, ;Si,;.;O5;)2(OH),, contains some Fe** as a replacement for Mg?*
in the octahedral sites, but only a modest amount. It is therefore essentially a
biotite and can be thought of as an end member of that series. Phlogopite can
withstand temperatures of up to 1000°C before transformation is a serious
problem, and therefore has several uses as a thermal window material.
Lepidolite, K(Li,Al)5_¢(Sig_7Al,-2)O2.(OH,F)4, is an interesting mica
having the muscovite structure but in which some of the octahedral Al** are
replaced by Li*. It is a common source of lithium, and has value as a raw
material which is added to glasses and enamels to lower their thermal
expansion coefficient. It is commonly found in association with spodumene
(15.2);
Mixed class derived from the talc, pyrophyllite and mica groups. This class
includes (a) chlorites, and (b) montmorillonites. Chlorites arise as weathering
products of pyroxenes, amphiboles and biotites (1.5.3). The basic structure is
the three-layer talc unit, but instead of K* binding these units, there is an
32 CERAMIC SCIENCE
intermediate brucite layer. Compositional fluctuations are large, as Al?* or
Fe** replace some Mg?* in the extra brucite layer (Fig. 1.15b), which is
associated with excess water molecules that ensure that the chlorites are
invariably plastic.
Vermiculite, (Mg,Fe,Al);(SiAlO;),:4H,O, is a typical and useful chlorite.
The water molecules form intermediate layers between the extra brucite layer
and the (Si,O,)?~ layers, their presence causes a rapid volume expansion and
exfoliation when heated as the steam is driven off. This makes it useful as a
packing agent.
Members of the montmorillonite group inevitably (on account of the basic
structure) possess very small and imperfect crystals packed in massive strata.
They are soft and plastic when wet but brittle when dry. It is easiest to view
these clays as having hybrid structures and compositions between talc and
pyrophyllite. Thus the octahedral sites in talc or pyrophyllite are occupied by
a mixture of Mg”* and Al?* ions, a fact that has three consequences. First,
the stacking of the three-layer sheets becomes quite disordered, leading to
poor crystallization and clay-like behaviour. Second, cations are required to
correct charge imbalances in the (Si,O,)*~ sheets; this gives the structure
mica-like features. Finally, the irregular packing makes the inter-layer
spacing the home of varying numbers of water molecules. Thus, in humid
environments they expand in the c direction as water is taken in, and on
warming, or under reduced humidity, they contract, making them problem
materials as far as civil engineering is concerned. A typical montmorillonite
has the composition Al; 33Mgp ¢7(Si,0;),(OH),0.67Na.
Clay minerals. The name “clay” is generally applied to any rock or soil having
a significant percentage of very small particles and possessing the ability to
be moulded when wet.* Particle sizes are less than 2.01m diameter, which
make them difficult to study, and as a result the structure and chemistry of
clays were unknown until the middle of this century. X-ray and electron
microscopic studies have since led to the following tentative definition:
Crystalline clay minerals are hydrated silicates with layer or chain lattices consisting of sheets :
of silica tetrahedra arranged in hexagonal form condensed with octahedral layers; they are
usually of small particle size.

This definition is a mineralogical one and omits what ceramists would say
was the most important property, namely plasticity. The definition includes
some layer lattice minerals described above that do not have this property,
such as talc, pyrophyllite and mica. It is not easy to resolve this point since
clays are inevitably mixtures and may contain plastic components. The
chemical composition of a clay depends both on the mineral from which it
was formed and the chemical environment of weathering. Thus in basic soils
in warm climates, weathering removes alkalis and silica and leaves Al and Fe;
the result is kaolinite formation. In acid soils of temperate climates all but
silica is removed and the soil has a low clay mineral content. Incomplete
weathering in arid climates causes montmorillonites and chlorites to be
 TRADITIONAL CERAMICS 33

Table 1.5 Clay mineral classification and use

Structure type Mineral Typical formula — Uses

2-sheet Kaolinite Al,Si,0;(OH), Principal component

Nacrite Al,Si,0;(OH), fire brick, china,
Dickite Al,Si,0;(OH), whiteware, tiles,
Halloysite Al,Si,0,(OH), porcelain. Paper
filler.
3-sheet Illite KAI, (Si;A1)O,.(OH), Moulding sands;
non-expandable (Mica) house brick, sewer
pipes.
Chlorite (MgFeAl),(Si,Al),0;.(OH), Not so abundant as
: to be useful.
3-sheet Montmorillonite (Ca,Na)o 35-9.7 Fuller’s earth,
expandable (Al,Mg,Fe),(Si,Al),0,.9(O0H),:nH,O — suspending agent,
filters, filler.
Vermiculite (Mg,Fe,Al)3(SiAIO;),-4H,O Thermal insulation,
packing material.
Fibrous Sepiolite Mg,(Si,0;),(OH), - 10H,O Carved into
ornaments.
Attapulgite Mg,;(Si,0;),(OH),-°8H,O Suspending agent for
slips.

present as clay minerals. The classification given in Table 1.5 is a compromise

with respect to mouldability, but does demonstrate the necessary limited
range of structural features that give rise to useful clay minerals.
An ice-like layer of water molecules around the small hexagonal platey
crystals of clays is postulated as the source of clay plasticity. The water is
believed to be structured in this way because the large O” and OH™ ions
forming the apices of the tetrahedral sheet and the base of the octahedral
sheet respectively give the particles a negative aspect to surrounding water,
since they effectively bury the small Sit*, Mg** or Al** in the tetrahedral
and octahedral positions within the two-sheet structure. Thus the ice-like
water layer is formed when H®+ dipoles on the H,O are attracted to the
particle surface. This general idea does not, however, suggest why the
non-plastic layer minerals like talc, pyrophyllite and mica do not strongly
bind water to themselves. Plastic clay as normally used contains between 15
and 30% of water by weight. Deformation in almost any mode produces
some orientation of the crystallites so that the c-axis of the platey crystals is
normal to the lines of the stress field. Clays are generally regarded as
Bingham-like materials.

1.5.4 Framework silicates

In framework silicates, the oxygen ions at all four corners of the SiOZ7
tetrahedra are shared with adjacent tetrahedra, so satisfying all oxygen
valencies and leaving a unit with no residual charge. The prototypes of these
silicates are the various forms of silica minerals having the formula SiO).
It is important to note that a framework structure has two characteristics:
34 CERAMIC SCIENCE
first, the structure is necessarily “open” with large interstitial spaces, and the
stabilities of the polymorphs are not determined by the packing of their
atoms. Secondly, the number of possible arrangements is large as the
variables, already mentioned, of tetrahedral angle, Si-O-Si angle, rotation of
tetrahedra and tilting of tetrahedra can be induced by the requirements of a
particular structure. Each structure on the whole is a compromise between
the conflicting requirements of tetrahedral shape, ideal tetrahedral linking
angle of 180°, and the symmetry demands of the structure.
It seems in the main that those linkage patterns are favoured that have a
small repeat unit and high symmetry, at least in their idealized forms.
Deviations from the idealized forms leads to families of structures based on
the idealized SiO, structures. A form of classification into groups is given
below that allows once again for the extension of the silicate families to
aluminosilicates by isomorphous replacement of Si** by Al?* in the tetra-
hedra with necessary balancing cations occupying the open network cavities.
Silica group
SiO,
(several forms)

25-50 %Al3* Na*\ (50%+) AB*

Nat Ke
Ca? +

Feldspar group Feldspathoid group

S
NS
K* Nain@ace Ss
\ \H20
Potassium feldspars Plagioclase feldspars N
ei eet oP
‘a
Zeolite group
Mixed feldspars

Because of the ease with which the framework can be modified to fit ions
of varying sizes without changing the main tetrahedral linkage scheme or the
nature of the chemical bonding, there is extensive solid solution amongst
these groups of structural families. Feldspars are the most abundant mineral
group on earth, and from them the layer and chain silicates have emerged by
weathering, and apart from being the precursors of clays they are much used
as fluxing agents in the traditional ceramics industry. Some features of all the
above groups, particularly the family precursors, the silica minerals, now
follow.
(i) The silica group. Three main polymorphic forms have been used and
studied for many years; in order of increasing stability at higher temperatures
we can list them as
 TRADITIONAL CERAMICS 35

(i) quartz up to 870°C

(ii) tridymite 870-1470°C
(ui) cristobalite 1470°—melting point 1730°C.
However, since there are fundamental differences in the tetrahedral linkage
patterns which imply differences in bonding, then changes will require sub-
stantial bonding disruption; the transformations from one form to another
are therefore very sluggish. Hence all three forms are regularly encountered
in minerals or quenched during ceramic transformations, or in glass-ceramic
systems (4.6). Each of the three forms has a high and low temperature
modification in which the structure is only slightly altered by a modest
change in the orientation of the SiO, tetrahedra relative to each other. The
high-temperature modifications have a higher symmetry or smaller unit cell
than the low temperature forms:
a-quartz (trigonal) a B-quartz (hexagonal)
«-tridymite (large hexagonal) Zs B-tridymite (small hexagonal cell)
a-cristobalite (tetragonal) oS B-cristobalite (cubic).

These transformations are rapid and reversible in contrast to the inter-type

changes.
There are other rarer forms of SiO, found at high pressures: coesite,
keatite, and stishovite. Stishovite is notable in that it contains Si in 6-
coordination in a rutile structure (5.8).
Quartz is used extensively in the ceramic and metallurgical industries.
Large single crystals, found in hydrothermal veins in Brazil, are suitable for
utilizing the piezoelectric effect found in this mineral. As a refractory (5.1) it
is used in the manufacture of firebrick, and as a network former (3.2) it is used
on a massive scale in the glass industry. When pure, it is a colourless hard
material given to conchoidal fracture. When impure it has many colours that
make it an attractive gemstone: agate, onyx, amethyst, topaz, and opal, a gel
form containing appreciable water, are examples.
The structure is based on a hexagonal lattice with a=0.49nm, c=
0.539nm, and with three SiO, units in the unit cell. A projection of the
structure of both low, (a-) and high, B-quartz on to (0001) is shown in Fig.
1.16 which shows the spiralling formation of the linked tetrahedra. The
spiralling is up axes about (4, 4) and (4, 4), in a left-handed fashion. For
example, starting from A in Fig. 1.16, the rise is from Si at O, to oxygen at
1 to Si at 2, to oxygen at 2, to Si at %, to oxygen at ~ above the break,
making a left-hand spiral of three SiO, units. In Fig. 1.16a the sequence is in
zioths (0, 11, 33, 45, 67, 78, 100), which gives some indication of the
increased symmetry of the higher-temperature polymorph.
It is notable that the linkage pattern remains unchanged in the two forms
which differ by just a small amount of spiral crumpling throughout. The
tetrahedra in both forms remain nearly regular, the change occurring by
rotation of the tetrahedra about their diad axis in (0001). The rotation and
36 CERAMIC SCIENCE

Figure 1.16 (a) «-quartz projected on (0001). Heights are in units of c/100. Breaks in Si-O
lines indicate that Si is joined to an O that is a translational repeat. (b) B-quartz projected on
(0001) with heights in units of c/6.

crumpling can be achieved in this open type of structure without atomic

hindrance. In both forms the Si-O-—Si bond angle linking the tetrahedra is
about 145°; this is considerably different to 180° and is characteristic of
semi-polar bonding. In some way this is part justification for using the
convention Sit* O77 to explain some patterns of behaviour in chapter 3.
Differences between the « and £ structures are reflected in the thermal
expansion behaviour of the two forms.’ As Fig. 1.17 shows, the « form
expands steadily at first and then rapidly at the transition temperature to be
followed by a zero then negative volume expansion for f-quartz. In the
a-form, as the temperature increases the Si-O-—Si bond angle increases from
about 143° to 150°, which improves the z-bonding overlap between Si—O,
(Fig. 2.3), and so decreases the Si-O bond lengths. Thus if the Si-O bonds
are decreasing in length then the observed volume expansion must be entirely
due to the Si-O-Si angle change. In the f-form, the zero expansion is
attributed to highly preferred atomic vibrations towards the open spaces in

240
Molar
Volume
m3 mole! B
x 10

23:5

400 500 600 1400

Temperature C

Figure 1.17 Volume expansion behaviour of «- and $-quartz.

 TRADITIONAL CERAMICS 37

Figure 1.18 Structure of 6-tridymite drawn to emphasize the pseudo or puckered layer structure
formed by the corner-sharing tetrahedra.

the structure such that an increase in Si-O distance is compensated for by a

decrease in Si-O-Si angle. Measurements of Young’s modulus in the a and c
directions and of the microhardness indicate a marked strengthening of
quartz after the transformation.
a-tridymite has an hexagonal cell six times larger than the B-form, with
a = 3.008 nm and c = 4.908 nm, and the 864 formula units in it have not been
precisely placed. The hexagonal sub-cell (a = 0.501 nm, c = 0.818 nm) is close
to the true unit cell of B-tridymite (a = 0.503 nm, c = 0.822 nm) for which the
structure shown in Fig. 1.18 has been derived. From the way this structure is
projected, it is possible to distinguish layers of puckered hexagonal rings
formed from the SiO, tetrahedra. In the a, b direction the puckered hexagons
have the “chair” configuration, whilst in the c direction they have the “boat”
configuration, This type of pseudo-layer structural description is analogous
to that used for SiC (2.10) and emphasizes the relationship to the wurtzite
structure.7®
Cristobalite is the form of silica that is stable at high temperatures. It
crystallizes in the cubic system with an open structure quite different to
quartz.‘© When it appears metastably at low temperatures, the structure is
slightly less open and becomes tetragonal. The existence of cristobalite down
to room temperature is dependent on environment, and if other materials
(notably Na,O and Al,O,) are present it may not be formed, for solid
solutions and impurities can cause quartz or tridymite to exist to higher
temperatures.
The idealized B-cristobalite structure is cubic, with a = 0.705 nm and eight
formula units in the cell. Each Si tetrahedron shares each of its corners with
another tetrahedron, and the centres of its four neighbouring tetrahedra are
arranged tetrahedrally around it. The silicon atoms alone are arranged
exactly like the Si atoms in silicon or the carbon atoms in diamond (2.5.1),
(2.6.1). In the idealized structure, the O atoms are midway between two Si in
this diamond-type structure, with a 180° Si-O-Si angle as shown in Fig.
1.19a. In the actual structure of high cristobalite, the Si-Si distances are not
all exactly equal and the oxygen atoms are displaced from the Si-Si axis
reducing the Si-O-Si angle to 140°; the structure however remains cubic.
38 CERAMIC SCIENCE

Figure 1.19 (a) Idealised B-cristobalite projected on (001). Heights are marked in units of c/8.
(b) a-cristobalite projected on (001). Heights represent c/100 for oxygen.

Below 200°C, the structure transforms to a tetragonal modification whilst

retaining the same linkage pattern.
When described in a manner emphasizing the pseudo-layer nature consist-
ing of puckered hexagons, the cristobalite forms have a zinc blende arrange-
ment of boat and chair rings (2.5.1). Furthermore, in the layer description
tridymite and crystobalite are related in the same way as are the two classic
types, zinc blende and wurtzite; both these forms are more open than
quartz, which is reflected by the densities: quartz 2.655, tridymite 2.37 and
cristobalite 2.33 x 10? kgm >.
f-cristobalite has an extremely low value for the linear coefficient of
thermal expansion at 2.08 x 10° '* m/mK~!; «-cristobalite has much larger
coefficients: 9.27 x 107! m/m K~! for the a-axis and 3.51 x 10712m/mK~}
for the c-axis. In the « form the Si and O are not fixed by symmetry, and
change steadily towards the positions they would occupy in $ with increasing
temperature; the geometry of the tetrahedra is constant, but their relative
orientation changes and this is the prime cause for the modest thermal
expansion coefficient of a. In f there is negligible thermal expansion because

B - cristobalite

6 —quartz

1:0

Expansion

—cristobalite silica glass

200
Temperature °C 1008

Figure 1.20 Thermal expansion of SiO, polymorphs.

 TRADITIONAL CERAMICS 39
as the Si-O bonds decrease in length the Si-O-Si angle increases with
increasing temperature. Each of these movements has an opposite effect on
the Si-Si separation, producing the null effect. Large effects occur at the
phase transition, as Fig. 1.20 shows.
(ii) The feldspar group. Feldspars are stuffed framework structures. When
large ions like K* or Ba** occupy the cavities, they keep the framework
extended; smaller ions like Na* and Ca?* allow the framework to collapse
somewhat, but for even smaller ions no stable configuration seems to exist.
All are low symmetry structures and because of this it is not easy to define the
variation in rotation of tetrahedra which occurs between one type and
composition and another. The problem is even further complicated by the
occurrence of order and disorder between the tetrahedral Si+*+ and Al3* ions.
Microcline, K(Alo 5Sio.75O)4, is the most common potassium feldspar,
and its formation at high temperatures with slow cooling produces an
ordering of the Al** and Si** in the tetrahedral sites. When heated above
900°C, the Si and Al become disordered, and this form is called sanidine.
Under special conditions a partial ordering can occur to yield orthoclase.
Anorthoclase, (K,Na)(Alp.5Sig.75O2)4 is a solid solution of potassium and
sodium feldspar formed by rapid cooling. When cooled slowly, perthite
results: this is an intergrowth of microcline and albite due to exsolution
occurring as Na* aggregate in bands oriented in selected crystallographic
directions.
(iii) Plagioclase feldspars form a series from albite to anorthite which
is difficult to distinguish from the potassium feldspars. Albite,
Na(Alo 25Sio.75O2)4, with smaller Na* ions in the cavity, has a crumpled
structure and a decrease to triclinic symmetry. This is the precursor of
kaolinite clay when weathering in basic conditions occurs. Albite fuses to a
colourless glass and finds use in the glass manufacturing industry. Anorthite,
Ca(Alp 5Sip.sO)4, is the end member of the series formed by isomorphous
replacement of Na* by Ca** and the simultaneous replacement of a Si** by
Al?* in albite. It weathers rapidly to bauxite (5.4.1) and clay (1.5.3).
(iv) The feldspathoid group are formed in an environment of low silica
content, and as a result the framework contains less silicon and more
aluminium than the feldspars. They form a family with the network structure
based on f-tridymite (see (i) above) with a variety of small distortions
produced by tetrahedra tilting and crumpling in order to reduce the volume
of the cavities occupied by large cations.
Nepheline, KNa3(Al.5Sig 502)s, gives its name as the alternative one for
this whole group. It occurs naturally in good deposits and is an important
rock mineral for use in the manufacture of ceramic whiteware bodies and
glazes. It crystallizes in the hexagonal system witha = 1.001 nm, c = 0.840 nm,
which shows its close relationship to tridymite. Framework tetrahedra have
nearly the same tilt about the triad axes as in tridymite, but they are also
tilted about a horizontal axis so that they no longer point exactly up or down.
40 CERAMIC SCIENCE

a b c

Figure 1.21 (a) Tetrahedral framework in f-tridymite. Each @ represents the centre of an SiO,
tetrahedron. (b) Tetrahedral framework in nepheline; dotted line shows the base of the unit cell.
(c) Idealized structure of nepheline projected on (0001), alternate tetrahedra point up and down.

Furthermore, the tetrahedral ions Sit* and Al?* have moved off the triad
axes. Fig. 1.21 shows the arrangement of the tetrahedra centres as a result of
this; when compared to the idealized £-tridymite structure it reveals that
there are three narrowed cavities for every one that is little affected. The K*
ions are located in the wide channels and the Na” in the elongated channels.
Kalsilite, K(Alo.;Sip.;O)>, is a nepheline where the absence of Na* ions is
accompanied by Al** and Si** ordering in alternate tetrahedra. Every
oxygen has one silicon and one aluminium nearest neighbour, but the Al and
Si remain on triad axes. The hexagonal unit cell has a=0.515nm, c=
0.867 nm, again showing its derivation from f-tridymite. Figure 1.22 shows
the slight rotation of the tetrahedra relative to each other.
Eucryptite, Li(Alg 5Sig.;O,)2, although an uncommon mineral, is frequently
encountered in lithium-containing ceramic products with low thermal expan-
sion coefficients (4.6.2). This desirable feature arises because, like nepheline,
it has a tetrahedral network in which half the cations are Al?* and a
balancing single charge cation in the cavities, but because Li* is much smaller
than Na* there is a crumpling to accommodate its coordination requirements,
and the basic linkage scheme becomes that of B-quartz (see (1) above).
(v) Zeolites: the ceramic catalysts. Most members of this group are synthetic,
having been made to modify and exploit the three-dimensional network of
channels or linked cavities found in the framework silicate structures. There

On---gge---=-9
K K

/ . :
/ ’
7 ’
/ : ’
ae coreeh aa
{riCPN @ eelf ’

fu shg
US ’
‘ Ve 4

K K

Figure 1.22 Kalsilite projection on (0001), triangles show the bases of the alternately Al- and
Si-filled tetrahedra at ------ =0,---- =4, = 1.0.
 TRADITIONAL CERAMICS 41

Figure 1.23 (a) The framework of 24 linked Si or Al tetrahedra in the sodalite structure; at the
intersection of the lines is an oxygen ion, thus apices a, b, c, d, e, f, define a hexagonal face
containing a channel; x denotes a 4-membered oxygen ring. (b) Linking of sodalite units across
the 4-membered ring to form zeolite-A. The channel at A is formed by an 8-membered oxygen
ring. (c) Linking across the hexagonal faces to form faujasite produces linked cavities with
entrances formed by 12 oxygen rings.

are about 120 synthetic (compared to approximately 30 natural) representa-

tives of the hydrated aluminosilicates known as zeolites.* Since interest is
focused on the cavities in these materials, a useful approach to their structures
is by means of the basic structure of a feldspathoid known as sodalite,
Nag(Alo.5Sip,502);2Cl,. This can be considered as a cubo-octahedral build-
ing block consisting of twenty-four SiO, or AlO, tetrahedra, which can be
linked across the four- and 6-membered rings present in such a framework.
The sodalite structure is outlined in Fig. 1.23, where some linked units
generating zeolites are also shown. When the sodalite units are connected by
bridging oxygens between the tetrahedra forming the four-membered rings,
the structure generated is zeolite A. When the sodalite units are linked so that
the apical oxygens at the tetrahedra forming the hexagonal rings are the
oxygen bridges, a structure known as faujasite results.
Each zeolite (Fig. 1.23) has a 3-D structure of linked cavities able to
absorb molecules of the correct dimensions. For example, in zeolite A the
cavity entrance measures 0.42nm and the internal cavity 1.14nm; faujasite
has a wider channel, 0.74 nm leading to a 1.2 nm cavity. The channel openings
and the internal cavities can be varied by changing the Al:Si ratio. Some
zeolites have been synthesized with channels in one direction only, or with
interconnected nets of different-sized cavities and channels. Because they are
aluminosilicates the channels and cavities cannot be completely empty, since
they must contain the cations needed to offset the charge loss when Al?*
replaces SI** in the sodalite units. This led to their first large-scale use, as ion
exchangers. Recently the ability of zeolites to exchange ions has been used to
moderate and control channel and cavity size.
The open porous structure leads to large internal surface areas and high
sorption capacities for molecules small enough to pass through the entrance
ports. Thus members of this class of aluminosilicate are extremely selective
42 CERAMIC SCIENCE

adsorbents, able to “sieve” many systems, with various industrial uses

ranging from the production of pure oxygen and nitrogen to the isolation of
p-xylene from a mixed xylene product.** The nature of the catalytic activity
of zeolites is intimately connected with the large internal surface area, part of
which has aluminium ions in the near surface region. Since this is Al*?*, the
tetrahedron has a negative charge which is neutralized by an ion like Na*
close by on the cavity surface; hydrogen ion exchange can remove the Na*
and leaves an acid H* surface that has catalytic activity. Raising the
temperature above 500°C removes the H* layer as water, and to do this an
Al-O-Si bridging oxygen must be removed, exposing an Al** ion at the
surface which has the property of accepting electron pairs, i.e. making Lewis
acid sites at the surface available for catalytic use. Such catalysts have several
well-defined advantages.
(i) High density of catalytic sites.
(ii) Well-defined active sites.
(iii) Molecular sieving properties which can be effective in at least three ways. First, only re-
actants below a certain critical size have access to the catalytic sites. Second, product-
selective catalysis occurs, since only products up to a limiting size can escape from the
sites. Thirdly, there is transition state selectivity, for if the transition state of a certain
reaction pathway is too large it cannot be accommodated in the available space. For
example, in a zeolite with pipe-like tunnels, CH,0OH can be dehydrated in such a way that
CH, units are constrained to link together in long chains to form aliphatic oils.
(iv) Controllable electrostatic fields. If M"* cations are exchanged for the Na“, strong electro-
static fields can be generated near to the active sites, and these fields can help to overcome
process activation energies.
(v ~ Selective sites provided for other catalytic species. For example, if Ru** ions are exchanged
on to the Na* ‘sites and the zeolite is then reduced with hydrogen, ruthenium metal is
produced at those sites. The metal now acts catalytically as it normally does, but being in a
constricted tunnel, selectivity occurs, and (for example) CO+H, produces only C,—C,,
hydrocarbons, thus providing good yields of diesel fuel.

Over 300000 tonnes of zeolite catalysts are now in use, and this quantity
can only increase as new zeolites with novel structures and new catalytic
properties are synthesized to meet the demand of the chemical industry for
new routes not requiring oil-based feedstocks. However, the route to new ‘
zeolites lies along old roads, for they are manufactured via the sodium
silicates, which we must now consider.

1.5.5 Sodium silicates

In Britain, 190000 tonnes per annum of these silicates are manufactured to
be used in a wide variety of industries. For example, detergent manufacture
takes 24%, paint manufacture 24 %, zeolites (1.5.4) and other catalysts 11 °%,
foundry binders 11° and cements 7 %.
Most sodium silicates are produced by Gossage’s method from silica sand
(1.5.4) by reaction with sodium carbonate to yield a glass:
Na,CO, +xSiO, <> Na,O *xSiO, Sta CO,,

Here x specifies the mole ratio, but the similar formulae weights of Na,O and
 TRADITIONAL CERAMICS 43
H20

kanemite
NaOH soluble powders

kenyaite @
mags!O4 Nag SiOg

Na20 SiO9
NagSi07 Na Si20, ‘NaySi307

manufactured range

Figure 1.24 Sodium silicate products.

SiO, make this almost identical to weight percentages. The reaction requires
temperatures above 1500°C, and so alternatives have been sought to reduce
energy costs which at present make the product sell at around £180 per ton.
The SiO, can be dissolved in NaOH solution when a weight ratio below 2.5
is acceptable; products of weight ratio around 3.3 are called neutral glass, and
around 2.0, alkaline glass. Direct reaction of salt, silica and water has so far
proved uneconomical but is feasible.
Crystalline products and the compositions of technical products are shown
in Fig. 1.24. Highly siliceous phases are slow to attain equilibrium structures
and compositions, but a few of the naturally occurring members (kenyaite,
maketite, and kanernite) have attained equilibrium over geological time. Of
these, maketite has been investigated crystallographically and shown to be
based on the single silicate sheet structure (Si,O;)?~ (1.5.3). Thus they
behave like clay minerals (1.5.3) and variable basal plane spacing occurs
according to the degree of hydration.
Relationships nearer to the soda side of Fig. 1.2.4 have been established
with more certainty. The metasilicate hydrates Na,SiO;:xH,O have all been
shown to contain the ion (H,SiO,)?~ and should therefore be expressed as
NaH,Si0O,:(x—1)H,O. Thus, despite their empirical formulae, they are not
chain silicates in contrast to what was said at the beginning of 1.5.
The area on the diagram denoting sols and gels is a scientifically poorly-
understood region, particularly with respect to the structure and nature of the
precipitated phases. Nevertheless, products arising from addition of salts to
compositions in the unstable gel area find wide use.

(a) Precipitates. These are made from concentrated, high-ratio silicate solutions by adding
soluble salts such as CaCl, or NaCl. They are precipitated from alkaline solutions, are non-
gelatinous and non-compacted, and their morphologies can be controlled by the salt added
44 CERAMIC SCIENCE

and the washing cycle used. They are mixtures of silica and silicates. When Al,(SO4); is
used as the precipitant it requires no additional mineral acid to give materials that are widely
used in paints, as plastic fillers, heat insulators and additives to aid powder flow.
(b) Gels. Attempts to precipitate in more acid conditions leads to a morphologically different
and extremely useful material broadly termed a gel. The first stage in the process gives rise
to a sol, as polysilicic acids with molecular weights around 2000 separate from solution.
The high concentration of salt causes a compression and spheroidization of the acid
particles. As the spheroidization occurs, the small particles of about 1.5 nm attract a sheath
of water molecules and pack with a 6-coordination sphere (5.3.2). All the spaces in the
close-packed arrangement so formed become filled with water, turning the sol into a gel with
evolution of heat. At this stage it is known as a hydrogel and has the appearance of a
glassy-solid. Different washing and neutralizing treatments bring about removal of inter-
particle water, salt, and particle sheath water which in turn affect the mean particle sizes to
give particles up to 8.0nm in diameter. It is this control that gives the dried materials
their varied uses as drying agents, abrasives, catalyst carriers, and food clarifiers. Because
their porosity can be closely controlled during the drying operation their density can
be adjusted to make them amongst the lightest of minerals: a gel with pore volume
0.0012 m?kg~' has a particle density of 600kgm_ ° and pores of mean diameter 12.0nm.

1.5.6 Cements
Since the main cement in use today consists mainly of calcium silicate phases,
some space is now devoted to considering the chemistry, mineralogy,
microstructure and mechanical properties of cements. All cements in common
use are termed “hydraulic”, because they set and harden due to chemical
reactions that occur on mixing with water.
(i) Portland cement. The name originates from the early description of the
product as a material resembling Portland stone. It is by far the most
important cement in terms of the quantity used in manufacturing concrete
products (3.5 x 10? tonnes of concrete per year). It is manufactured in large
rotary kilns lined with magnesia refractory (5.6). The kiln rotates slowly
about its own axis, which is a few degrees from the horizontal, and is fed a
mixture of powdered limestone and clay (1.5.3). As the mixture slowly
progresses down the kiln, it meets a flame from the opposite end which
gradually raises its temperature; by 1000°C, free and combined water and
CO, are removed. Near the end the temperature is 1300—1500°C, when
approximately 30% of the mixture is molten and reacts to produce the
cement mixture in the form of a lumpy product known as cement clinker.
The clinker is cooled and then mixed and ground with 2-5% gypsum,
CaSO,:2H,O. The gypsum controls the setting rate. Oxidizing conditions
ensure that any iron present remains as Fe(III). It is believed that conditions
in such kilns allow the clinker to approach phase equilibrium and so Fig.
1.25 allows prediction of the main constituents of the mixture.
The compositions in Fig. 1.25 are given in cement-chemical notation,
where C= CaO, S=SiO,, A= AI,0,, F =Fe,0, and H = H,O; tri-
calcium silicate is designated C,S, dicalcium silicate C,S, tricalcium aluminate
C,A and a ferrite phase C,AF (this last material reflects the impure nature of
the raw materials). Relative amounts in Portland cement are C3S 50-70%,
C,S 20-30%, C3A 5-12% and C,AF 5-12%. Most microscopic examina-
 TRADITIONAL CERAMICS 45
SiO,
Cos = Ca,SiQ, 7
Portland cement compositions

c,s = Ca3Si0,

C2 ta0 Alp0, =A
C3ZA = Cag,

Figure 1.25 Part of the CaO—Al,O;-SiO, system.

tions confirm this mixture and show that the C,S and C,S crystals are
relatively large (10-50 pm) having been formed at the clinkering temperature,
and are embedded in a matrix consisting mainly of C,A formed from the
liquid phase; the ferrite solid solutions occur as part of the matrix.
All formulae quoted so far are approximations because, as expected from
the discussion above, isomorphous replacement by impurity cations occurs.
This is reflected in both phase stabilities*’ and the product names: C,S is
often called alite, while C,S and C,A are belite and aluminate respectively.
The main constituents of Portland cement have been studied separately.
Alite has the ionic constitution Ca,(SiO,)O, and the structural principles of
(1.5.1) apply. Regions can be distinguished in the structure where Ca?* and
O?~ pack together as in CaO. Several slightly different structural modifica-
tions exist, depending on the degree of isomorphous replacement of Ca?* by
Mg?*. When made into a paste, C,S develops high compressive strength
relatively rapidly, ie. 70% of the final value in 28 days. Belite is one of four
known polymorphs of the orthosilicate Ca,SiO, and is the only one which
reacts with water to form a paste that develops compressive strength. It is
unstable at room temperature relative to «-Ca,SiO, but may be stabilized
when some impurity ions replace Ca”*. On forming the clinker, « or a’ arises
and then changes to 8 on cooling. Occasionally the « or y forms persist in the

Table 1.6 Some Portland cements of different types

Cement Phase composition %

Type Description C3S C,S C3;A C,AF MgO CaO CaSO,

I General 45 27 11 8 29 a OS ame sol

II Moderate heat ofhardening 44 31 5 13 Za 0.4 2.8
Ill ’ High duty strength 53 19 11 9 ID Of AKO
Vv Sulphate-resisting 38 43 4 9 Oa Os cum 201,
46 CERAMIC SCIENCE

Table 1.7 Relative phase development in Portland cement as a function of atmosphere

Atmosphere C3A B-C,S C,AF CaO

Oxidizing Low Medium Medium Medium

Neutral Medium Medium Medium Medium
Humid Low Medium High Low
Reducing Very high Medium Low High

clinker, and since they have no cementing properties they are to be avoided.
B-C,S develops high compressive strength when formed into a paste, but
much more slowly than alite, i.e. 75 °% of final value in 180 days. Neither C,A
nor C,AF develop much strength when ina paste.
To a first approximation, a Portland cement behaves as the sum of the four
main constituents, and, since the rate of attaining high strength is determined
by the C3S content, special clinker compositions have been developed for
particular applications: some examples are given in Table 1.6. In order to
produce the different types of cement, control of the kiln atmosphere is
important; this is shown in Table 1.7. Belite (ic. B-C,S) content appears to
be little affected, but C;A development can be controlled via the reaction

Ca,Al,Fe,0,9 23 Ca,Al,O, + CaO + 2Fe+3H,0.

Under humid conditions the overall reactivity is increased.
(ii) Solidification of cement. The following factors control the ability of a
substance to act as a hydraulic cement when it is finely powdered and mixed
with water :??
(i) reactivity to water
(ii) solubility of the water reaction product
(iii) the microstructure of the product after the water reaction.
For example, y-C,S is not a cement because it does not react with water, but
B-C,S does, and CA, although reacting readily with water does not yield a’
product with the necessary microstructure of interlocking fibres like that
depicted in Fig. 1.26.
Several types of particle have been identified in the microstructure of
Portland cement pastes:'°
(a) Unreacted cement grains diminishing in size with time but attaining roughly equal
dimensions around 20 um.

mixing 6 Hours 7 Days

Figure 1.26 Development of structure in cement pastes.

 TRADITIONAL CERAMICS 47
(b) Platey crystals—composition near CH, (ice. Ca(OH),). Size around 10 ym.
(c) Smaller platey crystals similar to (b), but composition typified by (Ca,Al(OH),),SO,:6H,O,
and known as AFm phases (here the initials stand for Al-F e-mono).
(d) Unbranched hexagonal prisms or needles 2-3 pm long, known as AFt phases and typified by
ettringite, Ca;(Al(OH),),:24H,O(SO,)3:2H,O.
(e) Fibrous particles, 0.5—2.0 um long, sometimes described as tubular or rolled sheets. These
have an indeterminate composition, denoted as C-S-H to indicate some form of calcium
silicate hydrate.
(f) Reticular network on a 1.0—2.0 um scale, also a form of C-S—H.
(g) Small platey grains > 0.3 um: another form of C-S-H.
(h) Compact, very small grain size material formed in the space originally occupied by cement
grains. Again this is a form of C-S—H.

The relative proportions of (a)-(h) depend upon the age of the cement paste
and the water :cement ratio.
(i) Up to 5 days; fibres or needles of (b) and (e) radiate from the cement
grains. ;
(ii) After 5 days; (d) becomes apparent.
(iii) 2-3 weeks; the (d) phase disappears as (g) replaces it and fills in
remaining spaces.
(iv) 4 weeks onwards; (g) forms an almost featureless mass engulfing the
particles of all other phases.
(v) At all stages; if the H,O/cement ration is low, ie. around 0.3, much
unreacted cement remains, but if it is above 0.7, little original cement can
be detected.
All the above changes in microstructure reflect the reactions occurring
between cement and water and are accompanied by changes in temperature
arising from the exothermic reactions.*! Under normal conditions a paste
may experience a 10°C temperature rise—most of the heat is evolved in the
first three days and then tails off exponentially. Typically 50° of the
400-600J g~' is evolved in three days, and of this amount 80% is released in
the first day. Figure 1.27 shows the plot of heat liberation against time to be

=
lo
TeoD
'> 6-0
S
e
2
Pe
2
°
>
®
2:0
eS
©
©
he

200 600 1000

Time mins.

Figure 1.27 Evolution of heat from a cement-water paste.

48 CERAMIC SCIENCE

a curve with two maxima. The sharpest maximum, occurring after 10 minutes
of mixing, is due mainly to the hydration of free lime and is followed by the
dormant period believed to be caused by a coating of gelatinous product on
the grains. Eventually the gelatinous layer loses its protective efficiency and
the temperature rises to another maximum after about 10 hours. The shape
of the curve in Fig. 1.27 is reminiscent of a nucleation and growth process
(4.3).
It should be noted that a very different heat evolution curve is obtained in
the absence of gypsum, for in gypsum-free Portland cement very rapid
heating and setting occurs as C3A dissolves quickly, causing calcium
aluminate hydrate precipitation. The retarding action of gypsum is thought
to proceed by formation of ettringite or monosulphate, a phase formed from
gypsum and CH. The setting action and strength development depend on
C-S-H fibre entanglement, and such fibres occur in “flash-set” as well as
normal setting. The difference in the final cement strength therefore lies in
where they are produced; in flash conditions they precipitate from solution
and mat whilst still leaving the C,A surfaces exposed, but in normal setting
with gypsum they are formed on C,A surfaces and thus afford it some pro-
tection and bind all components of the microstructure together.
A model for the hydration process has been proposed that provides good
correlation with the microstructural and other evidence. The essentials of the
model are as follows.
Stage 1 (up to 20 minutes): rapid leaching of Ca** into the water from lime,
impurity phases, and to a limited extent, the main cement phases. To a much
smaller extent, Al>* is leached from cement grains, and this has an inhibiting
effect on the dissolution of silicate anions from the particles. Thus loss of
Ca’**, Al?* and only a little “SiO” from the particles produces a thin surface
layer around them that is continuous in the sense that it links all particles
together. When this happens the first stage is complete and the cement is set.
Stage 2 (20 minutes—3 hours). This is the induction period coinciding with the
minimum in heat evolution in Fig. 1.27. Ca?* continues to dissolve from the’
cement grains, but at a much lower rate as the ions traverse the protective

1000+ C
Compressive
strength MPa.
100 A
Log scale.

10
D
Oo 10 20 30 40 50

Porosity %

Figure 1.28 Log strength v. porosity of Portland cements.

 TRADITIONAL CERAMICS 49
layer through which silicates cannot move. At this point each cement particle
is surrounded byasilicon-rich material which is itself separated from the
solution by a membrane whose surface is relatively rich in Ca?*, and the
concentration of Ca?* in the aqueous phase is steadily rising. Eventually the
solubility product of Ca(OH), is exceeded and Stage 2 ends.
Stage 3 (3 hours—1 day). Precipitation of Ca(OH), occurs near the surface of
the membrane layer, causing localized heating and stress at the membrane,
and these combine to initiate local fibrous growth at points on its surface.
Eventually the tips of the fibres are broken, and the inside membrane
contents flow into contact with the outer solution. What are now hollow
tubes grow by precipitation from this process. Frequently the crystals
growing at this stage are ettringite. This process continues with steady heat
evolution until all the membrane has been destroyed.
Stage 4. Destruction of the membrane exposes the silica-rich interior to the
Ca**-rich exterior and C-S-H forms rapidly as thin sheets all entwining
from particle to particle and giving strength to the cement. Many of these
sheet-silicate hydrates curl up to form tubes in the manner of some clays
(1.5.3). As the C-S—H particles intertwine, (e), (g) and (h) structures in the
classification above become apparent and the unreacted core continues to
react at an ever-decreasing rate.
(i1) Microstructure and strength. This discussion on the hydration of cement
clearly indicates a complex microstructure that must be taken into account
when analysing the strength of the material. One approach is to consider
cement as a composite of three components: (i) coarse, dense and crystalline
particles called H component; (ii) low density, poorly crystalline fibrous
particles called L component; and (iii) pores which are controlled in the first
instance by the H,O:cement ratio. When the porosity is high, the intrinsic
strength of the particles is not as important as the strength of the attractive
forces between the particles. This last variable is clearly dependent on the
surface energy, (2.1.1), and will be much greater for the L particles. An
optimum blend that has the highest strength will exist on this model, and the
results in Fig. 1.28 show this is probably the case because curve CD, which is
for H-type material, drops below curve AB for L-material at high porosities.
(iv) Slag cements. If molten slag from a blast furnace is cooled slowly,
crystalline aluminosilicates result, but if it is cooled rapidly a glassy form
results. The crystalline material has no value as a cement, but the glassy form
has cementing properties when finely ground and mixed with a suitable
activator. Suitable activators are strongly alkaline materials such as CaO,
NaOH, Na,CO, or Portland cement. It seems that the activators do not
change the composition a great deal, but act by disrupting a passivating
surface layer.
The main hydration product seems to be C-S—H, but probably with a
lower Ca/Si ratio; ettringite is frequently produced together with a variety of
platey hexagonal phases such as C,ASHg. When CaSO, :2H,0O is added, the
50 CERAMIC SCIENCE

products have good resistance to mildly acid solutions, sea water, and
sulphate solutions, probably as a result of the absence of Ca(OH), and the
fact that much of the alumina is present as ettringite.
(v) Expansive cements. These are formulated to help overcome the problems
caused by structural shrinkage on hydration. A mixture of CaCOs,
CaSO, :2H,O, and bauxite when heated gives rise to a sulphoaluminate with
a framework structure closely resembling sodalite (1.5.4). The sulpho-
aluminate clinker has the formula Ca,Al,O,,:SO,4, which can be compared
to sodalite, Na,Al,Si,0,,-°Cl, and is rapidly attacked by water. A mixture of
the sulphoaluminate clinker and Portland clinker has controllable expansive
properties attributed to the presence of ettringite. However, the ettringite in
this case must have a particular texture, since it does not always produce
expansion when present in other formulations such as Portland cement.
(vi) Pozzolanic cements. According to Lea “pozzolanas are siliceous materials
which, though not cementitious themselves, contain constituents which at
ordinary temperatures will combine with lime in the presence of water to
form compounds which have a low solubility and possess cementing proper-
ties.” The Romans used CaO—pozzalana mortars with the lasting effects seen
to this day; they include glassy material of volcanic origin, fired clays, and
fuel ash. SiO, (40—90 %) and Al,O, (5—35 %) are the main components.
In practice these cements hydrate only very slowly, even in warm climates,
to C—S-H and hydrogarnets such as Ca,Al,(SiO,);(OH),. More useful
cements are mixtures of pozzolana-lime and Portland cement. Hydration of
this mixture removes all Ca(OH), through reaction with pozzolana, and so
produces a more chemically resistant, stronger cement.
(vii) Autoclaved calcium silicate cements. Lime (4-12%), quartz sand and
sufficient water for moulding are mixed. After moulding at 500kgcm ~~”, the
mixture is autoclaved in high pressure steam at 200°C for 15 hours. All
Ca(OH), is removed and an amorphous C—S—H bonding phase is produced
from the silica sand. Over-autoclaving tends to crystallize the bonding phase
to tobermorite, C;S,Hs. ;
A useful cement can also be obtained by replacing the lime with Portland
cement in the autoclaving process. Absence of Ca(OH), in the final cement
or concrete artefact leads to greater chemical stability and greater crystallinity
in the hydrate phase, which in turn leads to increased strength.
An important development of these autoclaved products with respect to
strength and potential use in non-compressive situations is the preparation of
a suitable glass fibre to enable lightweight glass-reinforced composites to be
made.*°
(viii) High alumina cement. The main constituent of this product has the
composition CA. It is easily hydrated under normal conditions to CAH,. It
is a product that was specifically designed to resist corrosive attack by
sulphates so damaging to Portland cements. A lesser degree of hydration
results in C, AH, anda gel. The relative amounts of the two products depend
 TRADITIONAL CERAMICS 51
on the temperature and the cement stoichiometry. Setting and hardening are
much faster than Portland cement. However, (1.5.6), depending on tempera-
ture and water content, conversion occurs leading to dramatic decreases in
strength which caused severe problems in the building industry in the
mid-1970s. The conversion becomes rapid at high temperatures in moist
surroundings, proceeding as it does through a solution phase:
3CAH,,) > C,;AH, +2AH, + 18H
3CAH, > 2C,AH,+AH,+9H .
The conversion product CAH, is cubic, and arising from the less dense,
phases CAH,, and C,AHg, produces many small voids in the concrete
structure and serious structural failure.

1.5.7 Thermal transformations of ceramic raw materials—topotaxy

The concept of heat work in ceramic manufacture has been discussed (1.3),
but the role of heat energy in changing the plastic formable structures of
ceramic clays into hard, strong ceramics has been left until this section. This
was a deliberate decision taken to emphasize the fact that a limited range of
silicate structures with considerable similarities between them (1.5) can only
give rise to a selected range of fired ceramic compositions and micro-
structures. It has become increasingly clear that solid state transformations
occurring in layered silicates are dominated by their particular parent
structure. When a layer-silicate is heated it does not undergo general
decomposition, from which new silicate phases can emerge as dictated. by
thermodynamics and the composition variations brought about by adding
other compounds, but rather the starting material is converted into a product
that has certain definite three-dimensional orientation relationships between
itself and the starting material. The term topotaxy has been used to describe
such reactions.
Because of the structural control of the decomposition of many silicates,
investigations aimed at producing new ceramic formulations by making
additions at the dehydroxylation stage are unrewarding.*° In many instances,
it was thought that kaolinite might become more reactive at its dehydroxyla-
tion temperature and as a result form new compounds, but in general no
significant reactions have been found, for reasons which become clearer
below.
The study of topotactic changes involves a detailed examination of the
thermal decomposition of individual single crystals. To study a typical
reaction, a single crystal that has an easily recognizable shape is mounted in
the camera, its position recorded, and it is examined by X-rays before and
after the reaction. Comparison of the spot positions from reactant and
product gives the angular relationships between the crystallographic axes of
the two lattices. Considerable experimental skill is required to reset the
crystal in the X-ray camera. If the reaction can be stopped at an intermediate
ay CERAMIC SCIENCE

stage of transformation the film will contain diffraction spots from reactant
and products, and this helps considerably by eliminating errors involved in
crystal re-setting.
(i) Kaolinite to mullite transformation sequence. This most important trans-
formation has been studied intensively by utilizing a great range of tech-
niques,*” but the most illuminating study has involved the topotactic
method.°
Overall, the change can be written as
6Al,Si,0;(OH), > 2(3A1,03-2SiO,) + 8SiO, + 12H,0.
Some general observations are:

(i) Up to 1000°C, surface water continues to desorb, which makes thermogravimetric interpre-
tation more complicated than usual.
(ii) Above 450°C, combined water begins to desorb, that is, dehydroxylation begins.
(iii) At 650°C, as far as X-ray powder methods are concerned, an amorphous product
(metakaolinite) results.
(iv) Above 980°C, an ionic spinel phase appears (5.9.1). The spinel is quite defective, because
the composition (around AISi, ,,O,) is far removed from AB,O,. An amorphous form of
silica (1.5.4) is also detected.
(v — At 1100°C, the spinel is transformed into needle-like crystals of mullite (1.5.1), which have a
composition nearer to the 1:2 end of the range and are therefore to some extent meta-
stable. These mullite crystals are embedded in a matrix of silica which can be amorphous or
have the cristobalite structure (1.5.4). Preferred orientation of the mullite crystals occurs in
three directions from each clay particle, and if the clay particles have preferred orientation
due to moulding (1.5.3) then this is reflected in the microstructure.
(vi) Variations in the properties of pottery are mainly due to the degree of crystallinity and
porosity of the silica matrix phase.

A topotactic investigation by Brindley as long ago as 1959 has given the

following more detailed information on the transformation. Stage 1 occurs
when kaolinite is totally dehydroxylated at 500°C. At this point the original
(Si,O,)2"" sheets are maintained, Al? * coordination changes from octahedral
to tetrahedral, and water molecules are formed by a homogeneous mech-
anism. The following axial relationships hold: aq, = dq) and bg = bg) in
size and direction, but c;,,, does not equal cq) because of structural collapse in
this direction. Furthermore, since the structural collapse is at different stages,
the material is quasi-crystalline. It is this disorder that must be responsible
for the amorphous appearance of metakaolinite when investigated by X-ray
powder methods.
Stage 2 begins when metakaolinite layers perpendicular to the c-axis
condense out SiO, to form a defect spinel in which the [111] direction is
perpendicular to the (001) plane in the original kaolinite and <110)
directions are parallel to the b-axis of kaolinite.
Stage 3 is when mullite is formed with its c-axis parallel to <110) directions
in the spinel, and since there are three of these, there are three alignments of
mullite crystals. The mullite structure has close similarities to the sillimanite
structure (1.5.1) in which it is interesting to note that there are chains of AlO,
 TRADITIONAL CERAMICS a3

octahedra and (Si,Al)O, tetrahedra parallel to the c-axis whilst similar chains
run in the ¢110) directions in spinel.
From these, it is instructive to consider the transformation in terms of the
layer sequence along the related directions, i.e. perpendicular to the (001)
plane in kaolinite and metakaolinite and parallel to [111] in spinel. A
schematic representation of a structural block containing 48 oxygen atoms for
these structures in the related directions is shown in Fig. 1.29. The figure
shows that to go from the layer sequence in kaolinite to the spinel sequence, a
100% increase in Al** is required along with a decrease of 23 Si** and
213 H* for a volume of 48 oxygen atoms; this is extremely severe, and need
not occur since the correct sequence can be obtained via metakaolinite. First
a homogeneous nucleation of H,O in the OH layers leads to the metakaolinite
c-axis sequence shown in Fig. 1.29, and then an expulsion of the condensation
product from 5} O and 23 Si per 48 oxygens i.e. 23 SiO,, produces the
correct sequence for the spinel layers along [111]. All that is required finally
in a small lateral migration of the Al?* and remaining Si** to produce the
spinel phase. By this topotactic mechanism the transformation proceeds to
the spinel and mullite without total decomposition. This mechanism also
helps to overcome the inherent problem posed by the thermodynamics,
namely that the reactions of dehydroxylation to mullite and to sillimanite
have nearly equal free energy changes, and yet sillimanite is never encountered.

LLL,\ OM
[i
023190 O00. O

© e 0102/3S; 8Sit51/20) 6 eeu 6

O70 O O 1023 Al 1023A1 QO O O O

\
«’LEZIZ 16(OH)
Kaolinite .
\
(001)
160)
Spinel [111]

Condense
H20
Homogeneously
ee e @ 102/3Si .

0O O O 102/3Al wa
2 4

530) 7
Metakaolinite _b [001]

Figure 1.29 Schematic representation of the kaolinite, metakaolinite and spinel structures
showing how the transformation from clay to fired ceramic is achieved. Rectangular blocks
represent oxygen layers.
54 CERAMIC SCIENCE

Table 1.8 Effect of firing atmosphere on mullite

formation from kaolinite at 1100°C for 4 hours

Pressure
Atmosphere (kN m7?) % Mullite

Oxygen 101.325 12.0

Co, 101.325 10.0
Air 101.325 17.0
N, 101.325 28.0
H, 101.325 33.0
H,0 e 60.0
Vacuum 1.013 x 107° 52.0

The topotactic mechanism can form the basis for attempting to understand
the mixed results that have been obtained by numerous workers who have
attempted to improve the yield of mullite at lower temperatures by the use of
mineralizers. It appears that ionic fluoride mineralizers enhance the mullitiza-
tion process, and we might expect this at the metakaolinite condensation
stage when F~ will react with SiO,, so removing it from the system and in a
sense catalysing the transition. Other, non-fluoride, mineralizers probably
only change the size and specific surface of the mullite crystals, thereby
making them more apparent, by a fluxing mechanism.
Mineralizers, particularly oxide phases, may also act in local control of
oxygen potentials. This is important since mullite is inherently oxygen
deficient (1.5.1.), and low oxygen potentials are conducive to its formation
and so help to control the spinel to mullite step in the process. There is no
doubt that firing atmospheres markedly affect the formation of mullite, as
Table 1.8 shows. This control through the local partial pressure of oxygen can
be compared to the later example (5.6.3) of how pitch can protect magnesia
refractories.
Since mullite improves almost all the ceramic properties of clay-based
ceramics, its accelerated production is desirable within the limited time given
to firing campaigns, and Table 1.8 emphasizes the important role kiln
atmosphere can play in ceramic production, as the earlier bone china
example also demonstrates (1.4).
The topotactic process has since been questioned and refined at two points.
First, several experimental techniques have been used to investigate the
nature of the defect spinel phase: infra-red spectroscopy, X-ray spectroscopy,
and radial electron density distribution as used in glass structure studies
(3.2.2). None of the evidence is absolutely convincing, but it does suggest that
the spinel phase is silicon-free and is in fact y-Al,O3 (5.4.8).
The second point of refinement is the nature of the transformation to
mullite, which must involve a further condensation of silica from the defect
spinel as the mechanism whereby the mullite is formed with an oxygen
 TRADITIONAL CERAMICS 55
deficiency or, of course, if the spinel phase is mainly y-Al,O3, then a reaction
to produce secondary mullite from the SiO, and y-Al,O3.
An interesting exercise is to see how a theoretical calculation of disorder
entropy for intermediate phases can throw some light on the possible stages
of the overall transformation.”* The following steps are involved.
(i) Assume the phases have close-packed oxide structures that are fixed but allow the cationic
distribution to be varied.
(ii) Select a reproducible unit of close-packed oxygens, say a block of 6x6, and note that
because of corner sharing, etc., this is reduced from 36 to 25 equivalent oxygens.
(iit) The number of octahedral and tetrahedral sites have to be calculated for the chosen block,
making the same allowances for their number as in (ii).
(iv) The number of distributions of the cations over the available sites is used to determine the
disorder entropy S, by using (1.12):
S,=RinW (1.12)
where R is the gas constant and W is the probability of a configuration which is itself
proportional to the total number of atomic arrangements. Thus one value for W arises
from the number of ways of distributing 12 cations over 56 tetrahedral and 28 octahedral
sites provided by 35 equivalent oxygens in close packing as:
[(56 + 28)!/(56
+28 —12)!]2!4,
The factor 2** occurs because there is a restriction placed on the remaining eleven cations
as soon as the first cation has been placed.
(v) A number of models for each phase in the transformation has to be devised. For example,
metakaolinite has 30 oxygens in rows of 6 with 54 Si and 54+ Al atoms distributed
over the available tetrahedral and octahedral sites in various ways, e.g. Si and Al in
octahedral and tetrahedral sites randomly, or Si and Al only in tetrahedral sites randomly,
or Alin octahedral and tetrahedral randomly but Si in ordered tetrahedral sites, etc. At this
stage, in order to facilitate the calculation the number of cations is raised to a whole
number and a correction applied to the available sites. For the Al—Si spinel phase a model
of 24 oxygens in rows of 6 providing sites for 8 Al?* and 6 Si** ions will suffice.
(vi) An important assumption is now made, that the vibrational entropy contribution is similar
in the reactants and products and so entropy changes reflect almost totally the changes
in the disorder between various phases.
(vil) Experimental values of entropy changes for the various stages of the transformation as
determined from heat capacity data are collected. The following data are available for this
sequence:
2 metakaolinite > spinel + cristobalite S° = —27.3 J mol™!
3 metakaolinite > mullite + 4 cristobalite S° = +11.8Jmol~*
3 spinel — 2 mullite + 5 cristobaliteS° = + 105.4J mol ?.
(viii) Disorder entropy changes for reactions between the various theoretical models are calcu-
lated as the algebraic sums of the calculated S, values after these have been corrected to
the number of oxygens in the molar formula from the number in the arbitrarily chosen
unit. This of course is a simple multiplication by the fractional ratio.
(ix) Transformations between all the possible models that give the closest agreement to the
experimental values given in (vii) are deemed to represent the possible transformation
routes.

Application of this process leads to the scheme set out in Fig. 1.30, and the
interpretation that metakaolinite is converted to a spinel structure; both
contain some tetrahedral Al°*. Then, as the spinel structure becomes
increasingly mullite-like, the lattice becomes progressively random and
separation of a silica phase causes depletion of oxygen in the mullite
framework.
Cc
56 CERAMIC SCIENCE

4 Al(oct+tet) RANDOM 4 Al(oct+tet) RANDOM

4 Si (tet) RANDOM 3Si (tet) ORDERING

4 Al(oct+tet) RANDOM 4Al (oct+tet) RANDOM

4 Si(tet) RANDOM 3Si (tet) ORDERING

METAKAOLINITE SPINEL
980°C

74Al(oct+tet) RANDOM 4 Al(oct or tet) RANDOM

¥3si(tet) ORDERED 3 Si (tet) ORDERED

ys4Al(oct+ tet) RANDOM 4 Al (oct or tet) RANDOM

ee ee eee ee
eee
Geren

83si(tet) ORDERED 3Si (tet) ORDERED

SPINEL

2Al(oct) RANDOM
CRISTOBALITE
2Al+2Si(tet) DISORDERING

2Al(oct) RANDOM

2Al+2Si(tet)DISORDERING

MULLITE ALL CATIONS DISORDERED

Figure 1.30 Representation of kaolinite-mullite reaction sequence. Lines represent rows of

close-packed oxygens.

As the temperature is raised the SiO, phases form a glassy matrix which
ultimately becomes dominant in controlling the properties.
(ii) Serpentine to forsterite transformation. This is another example of a layer
silicate transforming topotactically whilst losing water on firing. The sheet
structure gives rise to an orthosilicate with isolated tetrahedra (1.5.1). Overall
the reaction can be written:
2Mg,(Si,0;)(OH), > 3Mg,SiO, +4H,0 +SiO,.
The SiO, is either amorphous or finely divided, for it cannot be identified on
X-ray photographs. Brindley and Zussman have studied this transformation
by the topotactic method and have produced the following facts and the
suggested reaction sequence:
(a) There are three sets of orthorhombic forsterite crystallization products, preferentially
oriented with the <001> and <013) directions parallel to the a axis of serpentine.
(b) The unit cell parameters have close similarities: 2d) ~ by, 2by) ~ 3c, and twice the
separation of (001) planes in serpentine ~ 3a,. These relationships result in 8 cell volumes of
serpentine being equal to 9 cell volumes of forsterite.
 TRADITIONAL CERAMICS Sy)

(c) The orientation relationships between forsterite and serpentine are a,//b p and b,//c,.
(d) A simple rotation of all the linked tetrahedra in the serpentine sheets, followed by small
lateral migrations of some Si**, brings the oxygen atoms to those positions found for
them in forsterite when the SiO{~ tetrahedra are close packed. This is indicated in Fig.
1.31a.

These findings are characteristic of topotaxy, and by considering the two

structures in related orientations the near-equality of the layers becomes
evident. The orientations chosen are 2d(100), and 3a,, as shown in Fig. 1.31b.
For the transformation, 64H* must diffuse out of the volume indicated in
Fig. 1.31b and be replaced by 24 Mg?* and 4 Si** diffusing in. In outer
zones the H* react with OH™ to nucleate water at dislocations and voids;
thus the process is heterogeneous as compared to kaolinite, where H,O is
produced in the layers that are transforming.** Heterogeneous nucleation of
the water destroys the crystallinity in the outer regions, leaving an amorphous
silica layer around each growing forsterite crystal. Clearly each forsterite
crystal must be surrounded by SiO, as Mg** ions have been drawn into the
growing crystal from the surrounding sefpentine matrix.
Ceramics made from serpentine can also contain enstatite (1.5.2) at higher
temperatures of decomposition and again at even higher temperatures
forsterite will be reformed by the incongruent melting of MgSiO3. In all cases
the product crystals are in a matrix of glassy siliceous material which again
dominates the properties of the fired product.
(iii) Dehydroxylation of talc. A third example involves another layer silicate,
talc (1.5.3) which is an important raw material in the fabrication of ceramics,

TENEMENT 24(0) 24(0) XXXAYYANAY

e oO ;

e A
YAY TY 24(OH) O e
d(100)} O O O 24(Mg) GEN
SLXTUMIIMY 16(0),8(0H) 6
ee 16(Si) =
WLLL 24(0) eee af
@

2A ) AYALA
d(100)| XYWYXYNTty24(0H) ih) Oars
OOO 24(0) NYNXXINAY |ay
@
®@

TIS XYUITNNNT
SERPENTINE _ FORSTERITE
2d(100) = 3a,

a b

Figure 1.31 (a) Tetrahedral diagram of serpentine and forsterite; solid lines are linked serpentine
tetrahedra, dotted triangles are SiO{~ in forsterite. <~rotations; ---> —, Sis migrations.
(b) Layers in serpentine and forsterite in a spacing of 2d(100), and 3a,. The near equality of these
layers in position and oxygen content is evident.
58 CERAMIC SCIENCE

in particular steatite insulators. This has recently been shown to be topo-

tactically controlled.'?
In all the above transitions, the product crystals are limited in type, have
preferred orientations, and, most important, must occur with the production
of a glassy phase. Thus properties such as strength are determined predomi-
nantly by thermal matching of the crystalline and glassy phases and by the
distribution of flaws in both phases. As a result, the physical properties of
traditional ceramics are texture-sensitive, and the actual crystal structures of
the non-glassy components have little influence. We conclude this chapter
with attempts to improve the glassy bonding phase, and in later chapters
show how this inherent problem has led to the development of glass ceramics
(4) and new and exciting materials (2).
1.5.8 Phosphate-bonded refractories
The problem associated with traditional ceramics, namely that such materials
are mixtures of crystals bound together by glassy intergranular phases as we
have seen in this chapter, cannot be completely overcome by utilizing the
pure materials described in (5). This is because refractories such as bricks
have to be “cemented” into structures and held together in difficult or severe
conditions. Industrial users also have the problem of how to make small
repairs to refractory structures during the lifetime of a given unit. These
considerations have motivated considerable progress into the study and
use of phosphate-bonded refractories. Phosphate-bonded refractory bricks,
mortars, repairing-pastes, and cold-setting castables utilizing most of the
known refractories are now widely used. Figure 1.32 gives some indication of
how successful the technique can be when the essential requirement of a
correct choice of phosphate bonding agent and aggregate is met.
There are now many phosphate-bonding agents suggested for use in
MOR.
MNm2

20-68
27:58

13-8 c
1:38
6:89
A A
Cub B
400 800 1200 1600 200 600 1000 1400
Temperature °C

Figure 1.32 Hot strengths of some phosphate bonded refractories.

Top: A = sodium polyphosphate bonded MgO, B = periclase brick, C = direct bonded MgO-
chrome spinel.
Bottom: A = Al,O3, B = Al,O; bonded with H,PO,, C = Al,O, bonded with SiO,.
 TRADITIONAL CERAMICS 59

ALUMINIUM ACID PHOSPHATE /,0 BINDER

AIPO,
(Ber)+(cRis)

Al,0, + Py Os

Figure 1.33 Development of phosphate binder phases as a function of temperature. Ber =

berlinite, Trid = tridymite, Cris = cristobalite, all of which are isostructural with high-
temperature polymorphs of SiO.

different conditions and as a result there are several mechanisms to explain

the processes involved. Originally, when nearly all phosphate bonding was
achieved with phosphoric acid, H,;PO,, three methods were postulated:
(i) reaction between siliceous materials and H;PO,
(ii) reactions between oxides and H,PO,
(iii) direct formulation of acid phosphates.
Since the advent of systems based on, for example, ammonium phosphate,
alkali metaphosphates or aluminium chlorophosphate, several of which are
solid, not liquid systems, wider explanations for effective phosphate bond
formation have had to be sought. It should be noted that all systems need to
be considered individually since: (a) real refractory systems contain an excess
of such reactive phase as Al,O3, ZrO, etc.; (b) P,O;:M,O, ratios can be
changed by using different members of the available bonding systems; (c)
different phase assemblages will result in different temperature regions. While
60 CERAMIC SCIENCE

acknowledging these restrictions, it is useful to have a scheme which describes

the effect of heat treatment on the most commonly encountered binder
compositions based on the acid phosphate Al(H,PO,.),; this is shown in Fig.
1.33.
(i) Phosphoric acid. Phosphoric acid, H,PO,, reacts with metal oxides and
hydroxides in the temperature range 20—200°C to form hard solid shapes, for
example tabular Al,O, (5.4.4), begins to react at 127°C. It is generally agreed
that the process involves, at the initial stage, the formation of Al(H,PO,);3.
Inevitably on heating in the presence of excess Al,O3, changes occur, and the
final bonding phase is AIPO,.
Mixtures of dried alumina, NH,F, and phosphoric acid will form castable
refractories, i.e. compositions that set without firing and are very useful for
making repairs to furnace linings and ladles, etc.
(ii) Aluminium acid phosphates. A range of liquid binding agents can be
prepared under this label but the most important parameter in deciding their
use is the acidity, defined as x, the molar ratio P,O;:Al,0;. Common
commercial products have x ~ 3; smaller ratios than this result in products
with poor storage life as they harden on keeping. With x ~ 3 the aqueous
solutions are approximately 50% W/W Al(H,PO,),. Typical reaction paths
leading to binding are shown in Fig. 1.33. Further bonding of a chemical type
takes place when the acid phosphate reacts with basic or amphoteric oxides
(5.1) with formation of crystalline orthophosphates that become the bonding
phases:

2MgHPO, -3H,O +4AlPO, -3H,0.

(iti) Alkali polymetaphosphates. Sodium hexametaphosphate Na,P,O,, has
been used as a bond for refractory mortars since 1947; it is one of a series of
polymers:

te eed
| |
Na—O— p—
=O ne O—Pp—
—ONa

O 6 O
| | |
Na Na/n Na
where n varies from 6—21.
Stability on keeping increases with the value of n as the hydrolysis
resistance increases.
These binding polymers are particularly effective when used with the
magnesite and chrome-magnesite refractories (5.6.2) where chemical binding
seems to be important with the presence of CaNaPO,, sodium rhenanite, in
the intergranular material.
 TRADITIONAL CERAMICS 61

References and further reading

i Akermann, R. J. and C. A. Sorrell. Thermal expansion and the high-low transformation in
quartz. J. Appl. Cryst. 7, 461 (1974).
. Aramaki, S. and R. Roy. Revised phase diagram for the system Al,O,—SiO,. J. Am. Ceram.
Soc., 45, 229 (1962).
WwW . Breck, D. W. Zeolite Molecular Sieves. Wiley, New York (1974).
. Brindley, G. W. X-Ray Identification and Crystal Structures of Clay Minerals. Min. Soc.
London (1951).
. Brindley, G. W. and M. Nakahira. New concepts of the transformation sequence of kaolinite
to mullite. Nature, 181, 1333 (1958).
. Brindley, G. W. and H. A. McKinstry. Kaolinite-mullite reaction series IV. J. Am. Ceram.
Soc., 44, 506 (1961).
. Brindley, G. W. Progress in Ceramic Science, Vol. 3, Pergamon Press, Oxford (1963).
. Calhoun, H. P. and C. R. Masson. Trimethylsilyl derivatives for the study of silicate
own

structures. J. Chem. Soc., Dalton, 1282 (1980).

. Chaudhuri, S. P. Induced mullitization of kaolinite—a review. Trans. J. Brit. Ceram. Soc., 76,
113 (1977).
10. Clark, G. M. and R. Morley. Inorganic pyro-compounds. Chem. Soc. Rev., 5, 269 (1976).
14: Cooper, C. F. Heat work. Trans. J. Brit. Ceram. Soc., 80, 60 (1981).
12: Daw, J. D., P. S. Nicholson and J. D. Embury. Inhomogeneous dehydroxylation of talc. J.
Am. Ceram. Soc., 55, 149 (1972).
133 Dinsdale, A. Translucency of tableware bodies. Bull. Am. Ceram. Soc., 55, 993 (1976).
14. Eitel, W. Silicate Science. Academic Press, New York (1965).
15: Evans, D. L., G. R. Fischer, J. E. Geiger and F. W. Martin. Thermal expansion and
chemical modifications of cordierite. J. Am. Ceram. Soc., 63, 629 (1980).
16. Hansen, S., L. Falth and S. Andersson. Structural relationships in tetrahedral framework:
reflections on cristobalite. J. Solid State Chem., 39, 137 (1981).
17. Hodgson, A. A., A. G. Freeman and H. F. W. Taylor. The thermal decomposition of
amosite. Mineralog. Mag., 35, 445 (1965).
18. Jenkins, G. F. Industrial Minerals and Rocks. Am. Inst. Min. Metall. Pet. Eng. (1960).
19. Jennings, H. M., B. J. Dalgleish and P. L. Pratt. Morphological development of hydrating
tricalcium silicate as examined by electron microscopy techniques. J. Am. Ceram. Soc.,
64, 567 (1981).
20. Kingery, W. D. Introduction to Ceramics. J. Wiley, New York (1960).
Dae Lea, F. M. and C. H. Desch. The Chemistry of Cement and Concrete. London: Edward
Arnold (1956).
De Leonard, A. Structural analysis of the transition phases in the kaolinite—mullite thermal
sequence. J. Am. Ceram. Soc., 60, 37 (1977).
23: Mackenzie, K. J. D. and J. F. Duncan. Calculation of the theoretical disorder entropies of
the intermediate phases formed during thermal transformations of aluminosilicate
minerals. Trans. J. Brit. Ceram. Soc., 70, 181 (1971).
24. Metcalfe, B. L. and J. H. Sant. The synthesis, microstructure and physical properties of high
purity mullite. Trans. J. Brit. Ceram. Soc., 74, 193 (1975).
Zs Morosin, B. Structure and thermal expansion of beryl. Acta Cryst. B28, 1899 (1972).
26. Ney, P. Asbeste. Naturwissenschaften, 68, 597 (1981).
Da Odler, I. and S. Abdul-Maula. Structure and properties of Portland cement clinker doped
with CaF,. J. Am. Ceram. Soc., 63, 654 (1980).
28. O’Keefe, M. and and B. G. Hyde. Plane nets in crystal chemistry. Proc. Roy. Soc., 295,
553 (1980).
29. Rado, P. Slip casting of non-clay minerals. Trans. J. Brit. Ceram. Soc., 72, 291 (1973).
30. Segnit, E. R. and T. Gelb. Kaolinite-CaCO3, kaolinite-MgO and kaolinite-ZnO reactions
below 1000°C. J. Austral. Ceram. Soc.,7, 1 (1971).
Sie Skalny, J., I. Jawed and H. F. W. Taylor. Studies on hydration of cement—recent develop-
ments. World Cement Technology, 183, (1978).
32: Taylor, H. F. W. Homogeneous and inhomogeneous mechanism in the dehydroxylation of
minerals. Clay Miner. Bull., 5, 45 (1962).
33: Taylor, H. F. W. The Chemistry of Cement. Academic Press. London and New York (1964).
34. Vaughan, D. E. W. Properties and Applications of Zeolites. London, R.S.C. (1980).
62 CERAMIC SCIENCE

35. Weber, L. and E. Grauer-Carstensen. The solidification behaviour of melts in the

ZnSiO,—Mg,SiO, system. J. Mater. Sci., 12, 1988 (1977).
36. West, J. M., A. J. Majumdar and R. C. De Vekey. Lightweight glass-reinforced cement.
Composites, 19, (1980).
2 Special light ceramics for
modern applications

Chapter 1 demonstrated that traditional ceramics have restricted properties

attractive only to specific users. Apart from intrinsic restrictions within the
crystalline silicates, the overriding problem is their existence as mixtures and
therefore the dominating influence microstructure has on their properties,
and this makes them an unattractive source of engineering materials, new
high-performance materials, or stable super-refractories (5.1). Here the aim is
to first analyse the material parameters required for intrinsic high-strength
and high-performance materials, then to use them to identify potential
materials for development. Some of these ceramics will be discussed in terms
of their production and properties. The chapter ends by considering how
most of the attractive properties can best be realized within the framework of
fibre composites.
Frequent cross-references will be given to important aspects of strength
developed in chapters 3 and 5; to comparable silicates in chapter 1 and to
glass science in chapter 3. The section on covalent bonding will become
essential for use in other chapters, mainly 1, 3 and 6. Information from other
parts of a general materials course will be necessary to benefit fully from this
chapter, for example, the phase rule, phase diagrams, introductory X-ray
crystallography and some basic dislocation theory.

2.1 Analysis of a strength equation

2.1.1 Surface energy

In 3.4.1, equation (3.28) is developed as an expression of the theoretical
strength of a material. The expression (equation 2.1) is quite general since no
specific bonding or structure was assumed for its development; it is therefore
valuable in identifying the criteria to be met in a ceramic of exceptional
strength.

a Ey
0
(2.1)
63
64 CERAMIC SCIENCE

Equation 2.1 contains the term y, the excess surface free energy, which has to
be large to give high intrinsic strength. However, it is not a frequently-
encountered property outside mechanics courses, and it is not immediately
obvious which ceramics will have large values for y. Therefore it is instructive
to derive an expression to replace y in (2.1).
Consider a crystal containing N atoms of which N, are at the surface,
leaving N—N, as internal atoms. We now make the grossly simplifying
assumption that only near-neighbour atoms bond to each other, and the total
binding energy (TBE) of the crystal can then be found by summing each
bond. Each inside atom has twelve near neighbours, six in its own layer,
three above and three below. Each surface atom has fewer near neighbours,
eight if the unit cell is fec. Hence
TBE = $AE[12(N—N,)+8N,] = 4AE[12N —4N,] (2.2)
In (2.2) AE is the binding energy between any two near-neighbour atoms,
which explains the factor 3—without it, (2.2) would be counting each
interaction twice. Now a thought experiment can be done: what would the
TBE be if the solid had no surface? The answer is expressed in (2.3):
TBE = $AE12N. (2.3)
Clearly the difference between (2.2) and (2.3) is the effect a surface has on the
solid. Thus the solid has a surface energy of —2AEN,, and since the bond
energy AE is negative, this is a positive term, meaning that the presence of a
surface raises the energy of the system, making it less stable. Thus production
of new surface absorbs energy, a fact used in fracture mechanics (3.4.1). The
units are kgs~* for surface tension and specific free energy, “specific”
referring to unit area. For liquids the two quantities are numerically equal,
but not for solids; atoms in a solid surface are not usually in equilibrium
positions after cleavage because of their low mobility.
This is partial progress, for we can now write:
yA = 2AEN, (2.4) '
where A is the surface area.
For 1 cm? containing z, the number of atoms per square centimetre, we can
write (2.4) as:
y = 2AEz (2.5)
Two steps are now necessary, first to replace z, and then AE, by using a model
crystal. To replace z we must consider the volume of an fcc solid: the side of
the unit cell 2a produces a cell volume 8a°, and since the cell contains four
atoms, the volume per atom is 2a. If the material has a molecular weight M
then one atom or molecule has mass M/No, where N, is Avogadro’s number.
If the density of the material is p then
M
= act
PNo
 SPECIAL LIGHT CERAMICS 65

from which

M
Gi (2.6)
2Nop
The face of the unit cell, which now represents a surface, has an area 4a? and
contains two atoms. Therefore each atom occupies an area 2a’.

Zzex
= number percm* os= Fa?

which from (2.6) gives

1 (2Nop\?!?
2=5( | (2.7)
In (2.7) we have expressed z in terms of the readily comprehended and
assessed quantities of mass and density.
In order to express AE in more meaningful terms we return to the model.
What must be supplied if one internal atom is removed to the atmosphere?
The answer is clearly six units of AE, since it has twelve nearest neighbours.
When one mole of a solid is sublimed, the energy needed is AH,,,, and so
one atom would require AH,,,/No. Therefore
AH
—_sub — 6AE
0

and
AR = AH sub
Nn.
Ng (2.8)

Putting (2.8) and (2.7) into (2.5) gives

027A aepe
a
eescey Go
Substitution of (2.9) and (2.6) into (2.1) gives the required equation (2.10)
0.33AH yppE
O4= fate (2.10)

Equation (2.10) is interesting because it leads to the criteria that a material

must fulfil in order to have high intrinsic strength. Perhaps the most
important and unexpected prediction is that suitable materials must have a
low molecular weight. This means that we can restrict further consideration
to the first two rows of the Periodic Table of the elements. (2.10) also tells us
that a suitable material must have (i) high values for the enthalpy of
sublimation; (ii) high density, and (iii) a large modulus of elasticity E. It is
fairly obvious that (i) will be found in materials possessing high melting and
boiling points. It is not so obvious which class of materials will have the
66 CERAMIC SCIENCE

highest values of Young’s modulus without first considering the nature of this
property.

2.1.2 Origin of Young’s modulus

When a load is applied to a solid then, in accord with Newton’s third law, if
its position looks unchanged the solid must be exerting an equal and opposite
force. The question of how the solid generates this force interested early
scientists, although they could not answer it. Hooke’s law of 1679, stating that
extension is proportional to force, was the first step towards an understanding
of the problem. Hooke realized that a spring was a special case of the
behaviour of all solids, but it is not obvious how clearly he distinguished
between the extension due to the geometrical shape and size of the object
under stress and the extension arising from the intrinsic properties of the
material, that is the elastic strain. Nevertheless, his work associated the
resisting force within a solid object and deflections within that material, and
the idea that a solid deflects just far enough to generate forces to oppose
externally applied loads was born.
It took about a century for this idea to be formulated more precisely.
Around 1800, Young showed that stresses and strains in a material should be
considered, rather than gross deflections, to probe the intrinsic resisting
force of a solid material. Stress and strain were formulated to make scientific
comparison possible. Tensile stress (c) is the load per unit area:

o=7. (2.11)

Nominal tensile strain (¢) is the extension under load per unit length:
_ AL
6 ==. (2.12)
Both stress and strain are independent of the size of the specimen, and strain
is even independent of the units of measurement, since it is a ratio of two
lengths. Young re-wrote Hooke’s law as

ao constant. (2.13)
&

He seems to have realised that the constant was an intrinsic property of a

material related to its ability to deflect and so generate resistive forces. The
constant became known as Young’s modulus or E, and (since it equals stress
divided by a dimensionless number) must itself be a stress. In theory it is that
stress which would double the length of a specimen, and since most solids
fracture at elastic strains around 1% it must be large: in the region of 100
times measured breaking stresses. It can be thought of as the constant that
generates resistance to applied stress, and as determining the stiffness of a
 SPECIAL LIGHT CERAMICS 67

material, since, from (2.13). Ee equals stress, and if E is large ¢ will be very
small, which is the same as saying any deflection will be very small.
Although early scientists were not aware of the molecular dimension, it is
now obvious that the forces generated during a deflection must arise from the
chemical bonds between atoms. That is, the elastic modulus E must be a
direct reflection of the interatomic bonding forces in a solid.
It is possible to express the bonding most simply in the case of ions in ionic
solids (5.3) from which equation (5.8) shows that force is inversely pro-
portional to the square of the interatomic separation. Thus any deflection
involving these bonds would not be expected to produce a linear stress v.
distance plot. This is an apparent contradiction to the Hooke’s law observa-
tions of a linear o v. ¢. Other types of bonding (covalent, metallic, and van
der Waal’s) all produce more complex relationships between force and
distance. This seems to imply that Hooke’s law is an approximation, and a
plot of o v. ¢ should not be linear. This prediction has been justified by studies
on fibres and whiskers. The stress-strain relationship for silicon whiskers
shows that elastic strains of over 3 % can be achieved, and that at these large
strains the plot is clearly curved, as expected if the modulus is an expression
of interatomic force relationships; the non-curvature for most materials is
just a reflection of the fact that they either break or flow and stop being elastic
before 1.0% strain. However, it is important to emphasize this last point since
observations of non-linearity for stress-strain relationships do confirm the
direct relationship between E and the nature of the dominant bonding forces
in a solid. E is found to vary by a factor of 2 x 10° over the range of solids

Table 2.1 Some values of strength, Young’s modulus and temperature of use

Density Tensile strength E Temperature

Material (Mgm_~*) (GPa) (GPa) (ee)

Aluminium 2.70 0.06 71 600

“Steel” 7.87 2.0 aa, 1200
Magnesium 1.74 0.04 42 200
Organic crystals _ — it 50
Al,O, (bulk) 4.0 0.28 400 1800
Al,O, (fibre) 4.0 2.0 470 800
Al,O, (whisker) 4.0 15 2250 1200
SiC (fibre) 3.4 2.3 480 900
SiC (whisker) 3.2 oN 840 1600
Diamond _ _ 1000 800
Carbon (fibre) a 23 377 2500
Boron (fibre) 2.6 35 1170 700
Glass (bulk) DS 0.070 70 700
E-glass (fibre) 2.54 3.4 72 550
S-glass (fibre) 2.48 4.8 85 650
DD, 5.8 72 750
SiO, (fibre)
3.3 4.1 190 300
Amphibole asbestos
29 aD 160 500
Chrysotile asbestos
68 CERAMIC SCIENCE

tested, and (broadly speaking) it is often a measure of the type of chemical

bonds in the solid. The order covalent > ionic ~ metallic > van der Waal’s
has been clearly established for E values. Table 2.1 helps to confirm this
classification. These observations can be applied to equation (2.10), from
which it can be seen that intrinsically strong materials must contain covalent
bonds, or, if mixed bonding occurs, then a substantial contribution should be
covalent.
In summary, then, ultra-strong and stiff materials should consist of the low
atomic weight elements that have a disposition to covalent bond formation.
Furthermore, these materials should have high melting and boiling points.
Before these criteria are applied to analyse potential engineering materials,
some consideration should be given to theories concerning the nature of the
covalent bond (2.2).

2.2 The covalent bond model>

2.2.1 Nature of the problem

We may say that two or more atoms form bonds because there isa total
lowering of the energy of the system when they remain in close proximity. In
practice, as shown in Table 2.2, this means a lowering of the electronic energy
of the system, since rotational and vibrational energies are small relative to
the energy of electrons. Hence species like H, are commonly encountered,
and not H;, because the total electronic energy of H; is greater than (H, +H)
even though it is lower than (H+H-+H). Similarly, the H-O-H angle in a
water molecule is 104.5° and both O-H distances are equal at 0.097nm
because the electron energy is lowest at this configuration. Thus any theory
of covalent bonding must be able to show how electron energies depend upon
angles and distances between nuclei. When stated like this the problem at first|
seems simple because it should be possible to use Newtonian mechanics and
the Coulombic energy of the system, as can be done for macroscopic bodies,

Table 2.2 Energies in the system H+H > H,

Energy contribution Energy (kJ mol~ *)

Total electronic for H, — 3095.3

Total electronic for 2H — 2640.1
Binding — 455.2
Zero point for H, 25.9
Rotational for H, 2D
Translational for H, 2.8
pv correction for H, 2.1
Bond energy — 432.2
 SPECIAL LIGHT CERAMICS 69

to obtain the electron energies. However, this contains two presuppositions

that closer inspection shows to be invalid: firstly, that one is able to define the
position and velocity of an electron in an atom accurately, and secondly that
one can follow the motion of each individual electron simultaneously. In any
attempt to measure both the position and momentum of an electron a
measuring device must be used, and such a device will involve electromagnetic
radiation, photons of which must collide with, and be reflected from the
electron, to enable it to be “seen”. Photons and electrons have nearly the
same energy, and so the “seeing” process results in changes in the electron’s
momentum that are not contained in the returning photon. Thus it is not
possible to locate or to measure the momentum of an electron exactly, which
obviously precludes the use of Newtonian mechanics. Heisenberg expressed
this problem in 1927 in the principle that bears his name: “Related pairs of
properties cannot be determined with as high a precision as we would wish.”
Equations (2.14) and (2.15) are expressions of the Heisenberg Uncertainty
Principle.

(Ap,)(Ax) 2 = (2.14)

h
(AE)(At) 2 an (2.15)

The second says that if a species exists for a very short time, that is, its
position in time is known very precisely, then its energy must be imprecisely
known since AE will be large. This leads to broadening of spectral lines.
Conversely, if one chooses to be imprecise about time, that is a relatively long
time interval elapses when the electron is in ground or excited states, then
energies can be determined with high precision, which is a useful point since
binding energies are the difference between two large energy terms as shown
in Table 2.2.
Although it is true that all systems must conform to the Uncertainty
Principle, it is not an important consideration for large bodies because the
measurements are not precise enough to make it obvious. For example,
consider a stone weighing 0.2 kg falling with a velocity of 30ms~ 1 The stone
has a momentum of 6.0kgms~‘ which we can assume it is possible to
measure with a precision of | part an 10°. The uncertainty in the momentum
would be no greater than 6.0 x 10°’ gems °. 1 Equation (2.14) tells us that the
position can be known with a precision of around 107 24m, which is much
greater than a measurement could possibly achieve. Now an electron, mass
10~ 3° kg, which has the same velocity as the stone, would have a momentum
of 3.0x10~2°kgms~‘. Allowing the same precision on the momentum
measurement would make Ap equal to 3.0x 10~3°gcms~*, and now the
uncertainty in position would be approximately 10m, which effectively
means that in atomic terms we have no idea where the electron is.
70 CERAMIC SCIENCE

2.2.2 Matter waves

The application of the Uncertainty Principle, as outlined above, makes it
clear that Newtonian concepts cannot be used to describe electrons in atoms.
An alternative description, in keeping with what can be found experimentally,
is required. Great theoretical and conceptual advances in this direction were
made in the 1920’s, but rather than describing this work we shall consider a
simple system that points to a suitable description. When light passes from a
source through two suitably-sized slits, a pattern is developed on a screen
behind them which is explained by considering the superposition of two
waves, one originating from slit 1, and the other from slit 2. At the point on
the screen corresponding to the mid-point of the two slits, the wave
amplitude is doubled, and the measured intensity is four times larger,
producing a bright line. The bright line is separated by dark regions from a
series of bright lines of decreasing intensity. Light waves from the slits
conform to an equation of the form

Y = Bsin2n>—vt (2.16)

where ‘Y = wave amplitude, commonly called the wave function. B is a

constant, x is the distance in the direction of the wave motion, / is the
wavelength, v the frequency, and t the time.
The intensity is proportional to ¥*, which is the number of photons per
second for large values of the source intensity. Consider what happens when
the source intensity is decreased: it reaches a state where W? appears to be
non-integral, say 3.62. However, fundamental particles such as photons
cannot exist in less than integral numbers. This paradox arises in the
following way: in one second perhaps 1 or 4 or 5 photons arrive at a
particular area on the screen, but every 100 seconds 362 arrive, giving rise to
the 3.62s~+1.Hence ‘¥? can only represent particle probability, and a wave |
description automatically contains the inability to predict when a particle
will arrive, that is an inability to predict its position exactly. This type of
language or model might be suitable for electrons in atoms. Pressing the
model further by decreasing the intensity of the source so that it emits one
photon per second, we find that the interference pattern is still produced over
a period of time. One might now ask which slit any one photon passed
through, and how did it help to produce a pattern which is a consequence of
using both slits? Here we see that as the system becomes more specific (i.e.
one photon at a time) we are more uncertain about how the system works. In
the wave model, there appears to be a greater uncertainty for small systems
arising naturally, that is, a built-in difference between macroscopic and
microscopic systems. Wave mechanics has the right kind of language (un-
certainties and probabilities) but we now need to know whether electrons can
be ascribed a wave nature.
 SPECIAL LIGHT CERAMICS al
2.2.3 The wave nature of the atom
De Broglie in 1924 postulated that electrons possessed wave properties with
a wavelength given by the equation

ae (2.17)
where p is the momentum of the electron. Three years later the wave nature
of the electron was demonstrated in a series of diffraction experiments. This
theoretical and practical work provided the impetus for an entirely new
theory concerning electrons in atoms and the nature of atomic interactions,
which firmly established the quantum properties of energy, (the values of
energy possessed by a system are not continuous, but discrete). The new
theory was logically called quantum mechanics, and modern concepts of the
atom and of chemical bonding arise as applications of quantum mechanics to
the description of electronic structures of atoms and molecules.
Waves can be described by equations, indeed we have already done this
in (2.16). How are electron waves described mathematically? Several descrip-
tions could be given, but the formalism given by Erwin Schrédinger is
the simplest and most convenient for solving molecular problems. The
Schrodinger equation, given in shorthand form as (2.18), relates the energy of
a system to its wave properties:

Hw = Ey. (2.18)
In (2.18), w is the function that describes the waveform of the electron, E is
the energy of the system, and ¥% is a mathematical instruction or operator.
The incentive for solving this equation is great, since it provides the answer
to our original question about how one determines the energy of an electron.
The wave functions that are solutions to the wave equations are called
orbitals, and so orbitals are mathematical descriptions of electron waves in
atoms. In all wave equations, the square of the wave function is the property
which has physical significance, and in the case of electronic wave functions
this represents the probability of finding the electron in a volume of space,
dx dy dz. In other words, the greater the amplitude or value of y in a given
region, the greater the probability that the electron is in that region.
Precise solutions of the Schrédinger equation, and hence exact wave
functions, have only been obtained for atoms containing one electron, that is
hydrogen, or ions containing only one electron. The process of solution
involves transforming from Cartesian coordinates to spherical polar co-
ordinates, separation of variables, solution of the three smaller equations, and
re-combination to find y. Each separated equation provides discrete solutions
that are physically meaningful, that is, the solutions have quantized values,
and the quantum numbers for the system arise naturally from the mathematics
and are not imposed arbitrarily as earlier theories found it necessary so to do.
7. CERAMIC SCIENCE

The orbitals can be written as

W(r,0,0) = Ry (t)O1,m(O)On(P): (2.20)

From (2.20) it is possible to state that orbitals, or electron wave equations,
can be represented as the product of three functions, two of which depend on
the direction only and are determined by the quantum numbers / and m. The
third depends only on the distance r from the nucleus and is specified by the
quantum numbers n and I. This fact that the wave function for a hydrogen-
like atom may be separated into radial and angular components is often used
to simplify the use of a wave function to describe an electron. The nature of
the wave equation and its solution impose restrictions on the values of the
four quantum numbers that arise. n is the principal quantum number and can
have values = 1,2,3...00;/is the orbital-shape quantum number and equals
0, 1,2,...(n—1) for any n value; mis the orbital-orientation quantum number
and equals —/, —1+1,...0,...J—1, | for any value of /; and s is the spin
quantum number = +3.

2.2.4 Atomic orbitals and their specification®

The wave function, w, mathematically describes the amplitude of vibration of
a standing wave within the volume of an atom that we take to represent an
electron. As already stated, it is more useful to consider 7 since this may be
regarded as proportional to the electron density, or as a representation of a
charge cloud about the nucleus. The orbital having the largest values for W
closest to the nucleus and hence having the lowest energy value, has n = 1.
From the above rules relating to quantum numbers, / and m must be zero;
hence there is only one orbital with n = 1 and this is designated as 1s in the
shorthand notation used to describe orbitals.
For 1s, y has the wave function given in (2.21)

Was = Ne" (2.21)

N is a constant known as the normalization constant, chosen so that the
probability of the electron existing in all space becomes one, and has the
value x '/?; r is the distance from the nucleus expressed in units of ay where
Ay = h?/(4n?m,e”), in which h is Planck’s constant, m, the mass of the
electron, and e is the charge associated with an electron.
Equation (2.21) gives the electron density, y*, associated with this orbital
at points in space as e 7”"/z, hence for the 1s orbital, and all other s orbitals,
there is a finite probability of finding an electron at the nucleus. It is
furthermore apparent from equation (2.21) that y can have values for all
values of r, but the region of space referred to when discussing orbitals is
usually restricted to a cross-section contour of constant probability density
such that the electron is within the volume about 90 % of the time. In the case
of a 1s orbital, the contour is a circle which represents a cross-section of the
spherical 1s orbital equation (2.21).
 SPECIAL LIGHT CERAMICS 13
When the principal quantum number is 2, | can have values 0 or 1. The
orbital with / = 0 is the 2s orbital and has the mathematical form given in
(222)
Woes =4(2n)
12 O—re-™. (2.22)
This shows that the 2s orbital is larger than 1s, and when r = 2 it has a zero
value. When a wave function becomes zero this is known as a node, thus the
2s orbital has a nodal sphere with a radius = 2 x 0.0529 nm; the electron
probability density has the appearance of an onion. This type of structure is
emphasized at n = 3,! = 0,n = 4,1 = 0 etc. which have two and three nodes
respectively. All s orbitals are spherically symmetrical.
There is a second possible set of orbitals when n = 2, for | can equal 1 and
then the quantum number m can have values —1, 0 and 1. This set of
solutions is not spherically symmetrical, and unlike the s orbitals, the three
orbitals have directional properties. They are designated 2p,, 2p, and 2p,.
When m = 0,
io
W(2p,) = a cos Ore-"?, O73)

The 2p, orbital has regions of greatest electron probability concentration

along the z-axis. By substituting 6 = 90° we see that w = 0, and hence there
is a nodal plane, the xy plane, where there is no probability of finding an
electron described by this function.
Further inspection of the p, function indicates that mathematical signs
should be ascribed to it above and below the nodal plane, since cos @ changes
sign in the range 0—180°. It is a general property of all p orbitals that they
change sign when inverted at the atomic nucleus (they are said to be
anti-symmetric). The p orbital is correctly drawn as two equal circles
touching at one point with a positive sign in one lobe and a negative sign in
the other lobe, whilst the probability function y* is dumb-bell shaped with
no sign differences in the two lobes. (It should be noted that the two concepts
are mixed in many diagrams by showing the probability surface, wy”, but
implying the wave function by showing different signs in the two lobes of the
p orbital.) The three equivalent p orbitals differ only in their spatial
orientations, not their energies; they are illustrated as the angular part only
of the wave-functions in Fig. 2.1 along with other orbitals.
To recapitulate: we have one solution to the wave equation (2.18) when
n = 1 and one waveform or state of lowest energy designated 1s. When n = 2,
there are four solutions from wave functions leading to four equal energy
states 2s, 2p,, 2p, and 2p,;this level is said to be “4-fold degenerate”.
This process can be continued to produce a nine-fold degeneracy, that is
nine orbitals representing the same electronic energy value, when n = 3. These
will be 3s when n = 3,1 = 0 which has the spherical form of 1s but contains two
nodes, and 3p when n = 3, /! = 1, 0 or —1. The 2p and 3p orbitals with the
74 CERAMIC SCIENCE

(g) (h)

Figure 2.1 Angular wave functions (a) = s, (b) = py, (c) = pz, (d) = dy, (e) = dz, (f) = dz_y,
(g) = fra, (A) = fie2—x2)y» (i) = feyes From A. H. Norbury, Education in Chemistry, 5, 28 (1968).

same m value have the same angular dependence, even though the boundary
contours of the 3p orbitals are more complicated than those for 2p orbitals
because they contain an additional node. Nevertheless the outer part of a 3p
 SPECIAL LIGHT CERAMICS 75
orbital looks like a 2p orbital and this is true for 4p etc; hence the generalized
angular wave functions shown in Fig. 2.1 can be used without specifying
principal quantum number when looking at bonding possibilities on an
atom.
When n = 3, from the rules above regarding quantum numbers, it is
possible for m to have five values 2, 1, 0, —1 and —2. This gives rise to a
further five solutions to the wave equation and increases the degeneracy of
this level to the nine already stated. The five orbitals associated with n = 3,
!= 2 and m+2+1, 0 have mathematical representations more complicated
than s or p orbitals, and are called d orbitals (Fig. 2.1). Although the lobes of
d orbitals carry a sign, the functions are symmetrical because inversion at the
origin does not result in a change of sign.
Models of the angular part of electron wave-functions as shown in Fig. 2.1
have several valuable uses. First, they emphasize the directional properties of
electron waves and may thus lead to useful discussions of molecular geometry.
Secondly, they are useful in simply demonstrating degeneracy breakdown in
the presence of fields arising from ions in crystals; an example of this is in the
inversion characteristics of spinels (5.9.2). A third direct use is in the guidance
they provide in the construction of correct combinations of orbitals, that is
hybridization (2.2.6), necessary for overlap and covalent bond formation, for
instance between non-metals and metals.
So far it is possible to solve the wave equation for hydrogen and
hydrogen-like one-electron systems exactly, to obtain the discrete energy
levels available for electron occupation and to picture the distribution of
electron density in such atoms through the angular parts of the wave
functions. The following energy order arises:
ls < 2s = 2p, = 2p, = 2p, < 3s = 3p, = 3p, = 3p, = 3d, = 3d,, = 3dzy
= 3452.2 = aay, < 4s = 4p, etc.
How are solutions obtained for more complex atoms? The Schrodinger
equation can be set up for such atoms, but it cannot be solved exactly,
because second and subsequent electrons introduce electron—electron repul-
sion terms. As a result, approximate analytical functions have been described
that allow solution and permit the determination of electron energies. The
commonest functions are Slater orbitals’? constructed according to rules
involving the atomic number of the element, the effective nuclear charge, and
the screening constant. Solutions using these hydrogen-like orbitals show
that the degeneracy of the energy levels is lifted, producing the following
sequence of electron wave energies:
1s < 2s < 2p, = 2p, = 2p, < 3s < 3p, = 3p, = 3p, < 4s < 3d < 4pete.
This sequence emphasizes that for all atoms, other than hydrogen, the
energy levels characteristic of each orbital depend not only on n but also on
1. It is possible to visualize these variations occurring because of interpenetra-
tion of the various orbitals.
76 CERAMIC SCIENCE

Returning to the original problem regarding determination of the energies

of atoms, we can see that we have arrived at a solution, since by assigning
each electron to an orbital we assign its energy, and the total atom energy is
then the sum of all the orbital energies.

2.2.5 Atom groundstates, aufbau principle and the periodic table

In 2.2.4, the allowed orbitals and their relative energies were established. The
question now is how they are filled. The most stable atom would have all the
electrons in 1s orbitals in order to minimize its energy. However, this is not
possible because two principles must be observed. The first is the Pauli
Exclusion Principle, which states that a maximum of two electrons can
occupy an orbital and only then if the spins are opposed, or “paired”. The
second principle is Hund’s rule, which states that one electron is allocated to
each orbital of equal energy before two electrons are placed in any one of a
degenerate set, and the electrons in the singly occupied orbitals have parallel
spins.
Since we have seen that intrinsically strong materials must have low atomic
or molecular weights (2.1.1) it is worthwhile now to consider the aufbau or
building-up process for the first few elements in the periodic classification.
(a) Hydrogen. This element contains only one electron, which must therefore
be in 1s, and so the electron configuration is given either as 1s or J, in which
+ stands for s = +4, and conversely | stands for s= —%, or *S4j.. The
latter is the spectroscopic notation and is derived by, determining the
resultant angular momentum quantum number L = Li” and noting that
L = 0,1,2,3,4,... are the states S, P, D, F,... respectively. Each electronic
state of the atom can be coded as: @S+)7, where S is the resultant spin
quantum number = Ys, J = L +S . The number of different J values for an
atom is called the multiplicity. Spectroscopic notation is useful in the way
excited states of atoms can be specified for subsequent discussion.
(b) Helium. In accord with the building-up principle, first one electron is
placed in the 1s orbital leading to He*. Now we must ask what the difference .
in repulsion energy arising from two electrons in 1s will be compared to the
second electron obtaining the energy necessary for it to become the 2s
orbital? The repulsion energy in the small volume represented by 1s is
2880kJ mol~!, whilst the gap 1s > 2s is 3940kJ mol”'. This determines the
groundstate electronic configuration for He as 1s”, or T|, or *So.
The first possible excited state for helium is 1s2s=°S, which is
1912kJ mol~! above 'S,, and so to achieve this level two covalent bonds
each stronger than 960 kJ mol + would be required. The most stable covalent
bonds encountered have an energy around 656kJ mol~' and therefore He
cannot interact covalently. The very high ionization energy precludes He”. It
is these facts that result in the inertness of helium, not any magic number of
electrons. Continuing this process the electronic structures of the early
members of the Periodic Table can be deduced (Table 2.3).
 SPECIAL LIGHT CERAMICS Wa

Table 2.3 Electronic configuration of the light elements

Spectroscopic
Element Ground state State Notes

Lithium 1s?2s 2
Si Even though half-filled level, covalent
bonds are rare; ionization energy
only 500kJ mol™?.
Beryllium 1s?2s? So
1s?2s2p state only 263 kJ mol~! above
'S so two covalent bonds common.
Boron 1s?2s?2p 2
Py). Excited state 1s?2s2p,p, only
345kJ mol~* above *P,,, .”. three
covalent bonds common.
Carbon 1s?2s?2p,2p, WP
Mutually perpendicular p orbitals give
no shielding .. no cations expected.
Anions possible when p, is used.
Excited state 1s?2s2p,p,p, is
404 kJ mol~1 above 3P, .”. four
covalent bonds formed as long as
each > 101kJmol7?.
4
Nitrogen 1s?2s?2p,D,D, S3)2 Anion formation difficult (5.3). Three
covalent bonds from the groundstate.
Oxygen 1s?2s?2p2p,p, SIP
(5.3) discusses anion formation. Only
two covalent bonds because excited
state 1s?2s?2p,p,p,3s is 883 kJ mol”!
above ?P,.
Fluorine 1s?2s?2p2p?p, 2
P3/2 Promotion to 1s?2s?2p*3s requires
1226kJ mol~? .*. no trivalence only
monovalence.
Neon 1s°2s?2p2D5Dz 1s?2s?2p°3s is 1602 kJ mol~ + above 'S,
.. no covalent bond formation.
Sodium 1's*2572p°3s 2
Sij2
Magnesium 15725727 °3s2 1
0 First excited state is only 262kJ mol7!
above 'S, but covalent bonds not
common because large lattice energy
favours ionic solids (5.6).
Aluminium 1s?2s?2p°3s?3p 2
Pi) Excited state *P,,. needs 347kJ mol” *
.. three covalent bonds are possible
(5.4.5). Mostly polarized electro-
valency (5.3).
Silicon 1s?2s?2p°3s3p,p, 2P6
Four covalent bonds via the °S, excited
state is easier than for carbon.
Phosphorus 1s?2s?2p°3s?3p,p,D, 4
S3/2 Three-valent but the 3d level makes
five-covalency possible.
Sulphur 1s?2s*2p°3s?3p2p,p, Ps

Chlorine 1s?2s?2p°3s?3p2p2p, 3
P32 Promotion to *P;,. needs 862 kJ mol~ *
.. three covalent bonds with fluorine
are possible as well as monovalency.
Argon 1s?2s?2p°3s73p° 256

2.2.6 Valence bond (localized molecular orbital) description of covalent bonding

This approach to the rationalization and interpretation of molecule and
compound formation begins from the ground state and excited state electronic
configurations of atoms as described in 2.2.5. One considers the atomic
orbitals and allows combinations of those that only contain one electron to
form a localized molecular orbital. Atomic orbitals containing just one
78 CERAMIC SCIENCE

electron are termed valence orbitals and the process of combining singly
occupied orbitals is referred to as pairing.
Pairing cannot leave the original waveforms unaffected, and therefore
produces a joint wave function between the atoms, given by the simple linear
combination w = Wa(1)Wp(2)+Wpg(1)Wa(2), where wa(l1) is the wave
function for electron 1 defined by the nucleus of atom A, etc. When this
combined wave function is substituted into equation (2.18) the solution
shows that there is an overall lowering of the electronic energy in the system
for favourable combinations. Few systems are solved in this way because of
the mathematical complexity, and frequently decisions as to which orbitals
are paired invoke the concept of maximum overlap. For simple atomic wave
functions, the maximum overlap of orbitals is achieved in the following order
S—S>Dg—Po > Pr—Pn; this order is changed when hybrid orbitals are
considered below. The symbolism p, and p, requires a little explanation. If
we are here considering the overlap, or common occupancy of space, of
atomic orbitals then it is clear from Fig. 2.1 that overlap can be brought
about end-on (e.g. two p, orbitals approaching along the x-axis), or sideways
on (when two p, orbitals overlap as the atoms approach along the x-axis).
The end-on overlap has the symbol o and produces a o-bond. Side-on has the
symbol z and produces a z bond which has wave amplitude maxima above
and below the molecular axis and is therefore more diffuse. This concept leads
quite easily to directed valencies and a few examples will show how to apply
this.
(a) Hydrogen. The electronic structure is simple, having the required half-
filled orbital, and so hydrogen atoms will be more stable when combined into
pairs with a strong bond arising from s—s interaction. This will be a o-bond.
(b) Nitrogen. Nitrogen will exist as a very stable diatomic molecule with three
electron pair bonds formed from the 2p half-filled orbitals, whilst leaving
alone the internally paired 1s? and 2s” orbitals. The presence of three bonds
in the N, molecule accounts for the apparent inertness of nitrogen, but we are
led to expect bonds of different strength from the maximum overlap principle.
This arises because the p orbitals are mutually perpendicular. A sigma bond
formed from say p,—p, overlap along the x axis brings the p, and p, orbitals
on respective atoms into parallel alignment and side-on wave amplification
above and below as well as on each side of the N—N o bond. Thus the
molecule has two x bonds and one sigma bond.
(c) Oxygen. From the atom ground state we would expect oxygen to exist as
a double-bonded molecule with one o bond from say p,—p, overlap and a x
bond from the resultant side-on overlap of p,—p,, leaving internally paired 2s”
and 2p? orbitals. In the O, molecule, in this description, all the electrons are
paired which should give rise to a diamagnetic state. In practice it is found
that oxygen is paramagnetic, indicating the presence of unpaired electrons;
the molecular orbital method, described below, is more successful in inter-
preting this fact.
 SPECIAL LIGHT CERAMICS 719
Oxygen, in ceramics, is frequently a bridge-forming element because of its
two unpaired p orbitals. However, it is not common to find the two o-bonds
formed from the p-orbitals at 90° to each other as might be expected. Usually,
the o-bonds forming the bridges are oriented closer to tetrahedral angles in
order to minimize repulsions from the internal electron pairs on the oxygen
atom. Oxygen differs from the rest of the group six elements in the un-
availability of d orbitals which makes the role of the internal electron pairs
different from that in sulphur and selenium. The repulsive force between the
electron pairs that cannot “de-localize” through d-orbitals precludes the
polymerization of O, molecules and the existence of a solid form at normal
temperatures.
When involved with atoms containing completely empty orbitals of only
moderately high energy, the oxygen atom can display donor covalent character
from its 2s* or 2p? pairs. This type of behaviour can influence the character
of resultant bonds and goes some way to interpreting the preferred molecular
geometries found in oxygen-containing ceramics. The concept of donor
covalent bonds is a modification to this approach; both electrons for a pair
bond come from one atom as long as the orbital pair overlaps significantly
with the space volume defined by an unused orbital on the second atom. A
more radical concept is introduced into this idea of covalency when the
electronic structures of polyatomic molecules are considered, and that is the
concept of hybridization.
(d) Hybridization. Hybridization is the mixing of orbitals on the same atom
to produce the appropriate number of valence orbitals on that atom. The
mixing of orbitals on the same atom is a favourable process because, as
emphasized by Linus Pauling, the hybrid orbitals which result from the
process have much better bonding properties, that is allow a larger volume of
overlap, than the pure s, p, d etc. orbitals. Hybrid orbitals are atomic orbitals
insofar as they are restricted to a single atom. Consider for example the
hybridization of 2s and 2p valence orbitals on, say, beryllium. These can be
mixed as follows:

Wny-1) = N(W25 + AW
ap,)
Why-2) = N(Av25 — Wp,) (2.24)
where Nis the normalizing constant and / is the weighting factor (0 < A < 1).
The physical picture of hybridization is represented in Fig. 2.2, which shows
that the process produces two sp hybrid orbitals on the Be atom, oriented at
180° from each other.
When J is 1 the orbitals are described as equivalent hybrid orbitals. These
two hybrids can now be half-filled, and are able to participate in a pair bond
by overlapping with (say) one hydrogen 1s orbital each, to form BeH,.
Because of the shape of the sp orbital, the overlap and interaction between it
and a spherical 1s orbital on the hydrogen is very large, and therefore the
80 CERAMIC SCIENCE

2s 4d

Ss

S u
Peete sai
Why (2s =2p)

Figure 2.2 Formation of two sp hybrid orbitals. Only when containing equal parts of s and p
(i.e. A = 1) do they have this shape.

bond formed is a strong one, able to more than compensate for the energy
involved in attaining the excited state implied by the hybridization and
unpairing scheme. Thus the previous order given for orbital overlap preference
is now radically changed.
The concept of hybridization is a pragmatic one which depends on the
balance of two factors if it is to be invoked. These are (i) the relative energies
of the orbitals to be mixed (they must not be too widely separated), and (ii)
the overlap qualities of the resulting hybrids.
Another example which emphasizes the pragmatic approach is H,O.
Water has an H-O-H bond angle of 104.5° which, if the bonds were formed
between pure p orbitals on the oxygen and the hydrogen 1s, we might expect
to be 90°. Polarization (5.3.6) of the O-H bonds and repulsion of the partly
charged H atoms would be expected to increase the 90° angle, but not beyond
97°. The total increase is then attributable to stronger bonding which will
arise from some mixing of the 2s orbital in the bonding, half-filled orbitals on
the oxygen atom. An equal admixture to produce equivalent sp hybrids would
lead to the expectation of 180° bond angles. A solution is forthcoming by ~
allowing intermediate degrees of hybridization, providing that not all the
resulting orbitals are to be equivalent. They must however be orthogonal, so
that they can be paired to form real pair bonds. According to quantum
mechanics, in order to be proper molecular wave-functions of the same
system the product of the two wave functions integrated over all space must
equal zero; this is the condition for orthogonality.
For equivalent hybrid orbitals, which will of course produce equivalent
bonds from the central atom, a number of well-defined mixtures have been
formulated and these are given in Table 2.4. The usefulness of these
hybridization schemes is in the molecular or solid geometries they predict, a
satisfying aspect of this theoretical approach. Two examples pertinent to
ceramics are given here, but others will be found in relevant sections of the
text.
 SPECIAL LIGHT CERAMICS 81

Table 2.4 Orbital geometries for hybridization producing equivalent valence

orbitals

No. of bonds Representation Shape Example

2 sp linear BeH,, HgCl,

3 sp? trigonal B,O;, BN, graphite
4 sp? tetrahedral SiO,, diamond, CH,
dsp* square planar AuBr,
5 dsp>,d°sp _ trigonal bipyramid PCI,
d’sp?,d*s squarepyramid _ IF,
6 d*sp? octahedral MoO,
d*sp trigonal prism Mo§, in MoS,
8 d*sp? dodecahedral —
d°p? square antiprism —

We first consider the electron configuration of the Si-O—Si oxygen bridge

encountered in so many ceramics. If the oxygen is considered to manifest sp*
hybridization and so too the silicon atoms, then a bridge with an angle of
109° is predicted and no oxygen orbitals or lone pairs are directed towards
available d orbitals on the silicon atoms. If sp” hybridization of the oxygen
occurs, then the bond angle would be 120° from overlap of sp” with silicon
sp orbitals, but this time the 2s orbital on oxygen can achieve some donor
covalency with a suitably oriented d orbital on each Si and so increase the
bond order and strength, thus shortening the Si-O distance. If the oxygen
achieves sp hybridization then the sp—sp*® o bonds will be linear but this time
2s and 2p orbitals “remaining” from the mixing process can overlap with two
of the d orbitals on each Si atom and so increasing the 7 component of the
bonding. Hence silicates and glasses can be envisaged as having Si-O-Si
angles varying between 109-180°, with bond distances varying as the x

(b)

hybridization.
Figure 2.3 Configuration of Si-O-Si oxygen bridge. (a) sp* hybridization. (b) sp
wave
The large dotted lobes are the relative positions of silicon d orbitals. Shaded areas indicate
amplification and bonding.
82 CERAMIC SCIENCE

component changes asa function of hybridization. Another way of looking at

that is to say that the oxygen hybridization can alter to accommodate the
requirements of other groups bonded to the silicon atoms (Fig. 2.3).
A second example concerns the probable structure of B,O, glass (2.9).
The electronic structure of boron indicates that sp” hybridization can occur
to produce three planar half-filled sp* hybrid valency orbitals whilst leaving
one empty p orbital perpendicular to the plane of the sp’ orbitals. An
equivalent hybridization of the oxygen orbitals would allow superposition of
the three boron sp” hybrids with oxygen sp” hybrids to produce a planar net
of BO, groups with each oxygen acting as a 120° bridge. The strong
interaction of o sp*—sp? bonds would then bring into overlap ina side-on Le.
m manner, the vacant boron p orbital with the parallel p orbital on oxygen
containing an electron pair to give a second component to the bonding. Boric
oxide glass is considered to be a disordering of this basic structure (3.2.3).

2.2.7 Molecular orbital theory of covalent bonding

This is a powerful method of analysis that begins by assuming that the main
ideas and methods developed to determine the electronic structures of atoms
and their energies (2.2.5) can be applied to groups of nuclei when they remain
in close proximity. The principles can be set out as follows.
(a) Each electron in a molecule is described by a wave function now called a
molecular orbital, w,,,, which is defined by quantum numbers governing
its energy and probability boundaries, i.e. its shape.
(b) Each y,,, has a definite energy value which represents the energy required
to remove an electron from the molecule. The total energy of the
molecule is close to the sum of the energies represented by occupied
molecular orbitals.
(c) Since each electron can have a spin of only +3, each y,,, can only
define two electrons.
(d) When building a molecule we must determine the allowed orbitals then,
adopt the aufbau or building-up process already described in 2.2.5, using
the same Pauli and Hund principles.
All this appears identical to the isolated atom case but differences do arise
when attempts are made to formulate the molecular orbitals (M.O.’s).
The method known as the linear combination of atomic orbitals, LCAO, first
suggested by R. S. Mulliken to describe molecular orbitals, is most commonly
used, perhaps because it is easily visualized. For example in diatomic
molecules we can see that (to a good approximation) the M.O. has the
character of an atomic orbital when close to one nucleus; similarly, when
close to the second nucleus the M.O. has the character of an atomic orbital
and therefore it seems natural to describe the M.O. as a combination of
atomic orbitals:
Wino = NW, +A)
 O- CD
SPECIAL LIGHT CERAMICS 83

1
1
'
t
'

Figure 2.4 The addition and subtraction of two 1s orbitals.

and since for most cases we can assume normalized wave function this can be
simplified to

Wino = Wa t Avg. (2.25)

A measures the polarity of the orbital and has a value chosen to minimize the
function

JWino A WinoOt

JVino
which is a re-write of equation (2.18) and is therefore the energy of the M.O.
Solution of the quadratic gives two roots or two energies, E,,, = E, +B,
where £ is the resonance integral [4% Wpgdt. Thus a second molecular
orbital must be generated by the addition process. In general, the LCAO of n
atomic orbitals must generate n molecular orbitals, which seems reasonable
if we recall that each A.O. can accommodate a maximum of 2 electrons; if n
atomic orbitals combine, the combination must result in sufficient molecular
orbitals to accommodate a maximum of 2n electrons.
The second linear combination can be obtained by subtracting one A.O.
from the other: y,,. = Wa —AWg. Figure 2.4 shows the boundary surfaces
generated by this procedure for 1s orbitals. In the M.O. obtained by
subtraction in Fig. 2.4, the probability of finding the electrons at exactly half
the distance between the nuclei is zero, as indicated by the nodal plane. On
average electrons in this M.O. are further from either of the nuclei than they
would be in isolated A.O.’s and so atoms would be energetically more
favourable if separated. Because electrons in such an orbital would act to

*
o
coo

is a
of ‘
Wp —~ eo ad)
\ /
‘ if
s /Y,
o

Figure 2.5 The molecular orbital energy diagram for H.

84 CERAMIC SCIENCE

separate the atoms it is called an antibonding orbital and given the subscript
a or superscript *. Conversely, electrons in the orbital (or wave function)
obtained by the addition have lower energy than atomic functions and so
act to bind the molecule. These are given the subscript b. The process is
represented by an energy diagram, Fig. 2.5.
The following rules have to be used to decide which atomic orbitals can be
added together to give an effective combination giving rise to a bonding and
antibonding M.O.:
(i) The energies of w, and Wg in their respective atoms must be com-
parable in magnitude.
(ii) Electron probability densities should overlap as much as possible.
(iii) Wa, Wg must have the same symmetry relative to the molecular axis
(sign of wave amplitude to be considered).
When porbitals are used in the linear combination process, again both
addition and subtraction of the atomic orbitals has to be made to generate
the correct number of molecular orbitals. This is shown schematically in Fig.
2.6, in which we see that there are two possibilities leading to o and x
molecular orbitals. The three atomic 2p levels in an isolated atom are
degenerate (of equal energy), but when the first atom is brought into the field
of another atom, the p, orbitals pointing towards the other atom begin to
interact to form a o bond between the two atoms. This of course generates
the corresponding o* orbital. The orbitals o(2p,) and o*(2p,) are still o
M.O.’s, even though generated by p orbitals, because they are symmetric with
respect to rotation around the internuclear axis, like o(s) orbitals. The
bonding interaction is large and the splitting between bonding and anti-

ee ee

Figure 2.6 The addition and subtraction of 2p atomic orbitals.

 SPECIAL LIGHT CERAMICS 85

: a"(2p)
+ TT p>.
4 nip)

20 HI ia
= ae 2p
aa

Figure 2.7 The molecular orbital energy diagram for O,. Unhybridized form.

bonding orbitals is large. As the process develops, the p, and p, orbitals on

each atom combine in the manner shown in Fig. 2.6b to form a z set; 2,,(2p,),
13(2px); T,(2py), 77 (2p,). These two pairs of bonding and antibonding
molecular orbitals must be perpendicular to each other. The side-on approach
does not produce as effective an electron wave interaction as the o overlap,
so that the x—n* splitting is not so large; the resultant molecular orbitals have
nodes where there is no electron density giving rise to their designation z, the
Greek equivalent of p.
In this manner a complete molecular orbital energy ladder can be built up
for a molecule and the total number of electrons then allocated in the manner
used to build up atomic structure. The example of oxygen (Fig. 2.7) is
interesting, because it shows that application of Hund’s rule leads to a
prediction of unpaired electrons and hence of paramagnetic behaviour, and
not the diamagnetism suggested by the Valence Bond Approach (2.2.6).
However, it should be stressed that the two bond models are not in
competition, but are complementary in the sense that the Valence Bond
Approach is useful for gaining insights into molecular and solid shapes, while
the molecular orbital approach provides a more detailed model of metallic—
non-metallic, magnetic, and spectral behaviour of materials.
The following examples using the molecular orbital method are chosen to
illustrate this situation, and further examples are included in the special
ceramics chapter 6.
(a) Oxygen. For this molecule, the molecular orbital energy diagram is as
shown in Fig. 2.7 for simple unhybridized atomic orbitals. The 16 electrons
from the two oxygen atoms must now be accommodated: 2 each in o(1s),
a*(1s), o(2s), ¢*(2s) and o(2p) accounts for 10 of the electrons and shows that
no binding contributions are made by the 1s and 2s molecular orbitals. The
next lowest energy level is the degenerate pair 2(2p) which will account for
another four electrons, leaving the final two. The assignment of these two
electrons is interesting, since the next lowest level is a twofold degenerate
86 CERAMIC SCIENCE

antibonding level, and so in accord with Hund’s rule each orbital is assigned
one electron with parallel spin. Thus the paramagnetic property of O, is
realized. The occupation of antibonding orbitals by two electrons cancels
some of the bonding between the atoms brought about by the total
occupation of three bonding orbitals o(2p) and 2(2p,), m(2p,). Thus the net
bonding in oxygen can be considered to result from a double bond as two
n-bonding contributions are cancelled. In the molecule the bond distance is
0.1207 nm, but when ionized to O} by loss of an electron from the highest
level, ic. the antibonding level, the bond distance decreases to 0.1122 nm
indicating a stronger bond in this molecule-ion as anticipated in this theory.
(b) Anionic carbon species. The molecular orbital energy diagram for oxygen
in Fig. 2.7 can be used to discuss the catenation possibilities for carbon, since
the same set of molecular orbitals will be generated, although with relatively
different energy values. In the carbon case, only 12 electrons must be
allocated to the molecular orbitals: a pair each in 04,), O45), F(25)» Fas) and
6(2p) uses 10 electrons, leaving two electrons, and once again two equal energy
orbitals which this time are m2) and 72,) each of which will accept one
electron with parallel spin. This arrangement leaves two half-filled bonding
molecular orbitals, occupation of which will strengthen the bonding in the C,
species. Thus the acceptance of two electrons to give C3~ is an energetically
favourable process and accounts for the existence of many ionic dicarbides
(6.1).
(c) Carbon monoxide. In this molecule the atoms are different, and we are able
to see the effects on the molecular orbital energy diagram brought about by
the increased nuclear charge on the oxygen atom relative to the carbon atom.
In constructing the M.O. diagram, the 2s and 2p orbitals of the one with the
higher nuclear charge are lower than the corresponding orbitals of its

lame
/ \
ise
fo NN
2p =e —— =a \

/
WUny ne H Y----G———— 2p
447 W \—_y
ae hy, \ Tr Ze Why
2

Figure 2.8 Molecular orbital energy diagram for CO.

 SPECIAL LIGHT CERAMICS 87

partner. This is shown in Fig. 2.8, which also contains the further sophistica-
tion of first producing sp hybrids on both the carbon and oxygen atoms.
Figure 2.2 shows that W,, for sp combinations produce large lobes at 180°,
and therefore if one such on oxygen is combined with one on carbon to form
the o bond then one on each atom must point away from the molecule and
have no chance of overlap. Such orbitals are called non-bonding and are
shown in Fig. 2.8 as being energetically unchanged in the molecular orbital
energy diagram. It can be seen from Fig. 2.8 that the two hybrid orbitals on
oxygen do not match the energies of those on carbon; indeed the lower
energy sp hybrid on carbon approximately matches the energy of the higher
energy hybrid on oxygen with the result that the two sp non-bonding orbitals
in the molecule contain lone pairs of electrons which play an important role
in the donor-acceptor properties of carbon monoxide. One lone pair is on
oxygen in a hybrid orbital, mostly s in character, and is thus more stable (less
reactive) than the lone pair on the carbon atom which is in a hybrid with a
large amount of p character. The bonding between carbon and oxygen arises
from the o bond and the z,,7, bond. Antibonding x* orbitals are shown in
the diagram lying close to the carbon p orbitals and therefore concentrated
on carbon. It is these empty antibonding orbitals that are important when
considering the acceptor properties of CO in a variety of compounds. This is
mentioned to emphasize the fact that antibonding orbitals are important and
can, despite their name, take part in chemical bonding in favourable
situations; some carbides are an example of this (6.1).
(d) Lithium. In this model an entire block of crystal is considered to bea giant
molecule. All the atomic orbitals manifest their allowed interactions to form
a set of delocalized orbitals extending throughout the solid. Nearest neigh-
bours produce most orbital overlap, and therefore the widest separation
between the bonding and anti-bonding molecular orbitals. All other bonding
and antibonding levels formed from next-nearest interactions, etc., must fall
within this envelope. If we assume that the number of valence orbitals is of
the order of 107° in a crystal, then the molecular orbital energy levels within
the envelope must have infinitesimally small separations, and because of this

oa a unfilled 2p

A —— a; Bioneers
2p ~ Sa

1023 vim 4023 half-filled 2s delocalised

——- molecular orbitals
ett in band

pore) mY, 1022 filled 1s delocalised

pres orbitals
1s
band

Figure 2.9 Delocalized molecular orbital bands in lithium.

D
88 CERAMIC SCIENCE

we use the term “energy band”. Figure 2.9 shows the three bands formed from
1s, 2s and 2p orbitals in lithium. The 1s molecular orbitals are completely
filled because the 1s atomic orbitals in isolated lithium atoms are full. Thus
the 1s electrons make no contribution to bonding. The lithium 2s orbitals
contain only 1 electron and therefore the 107° levels in the 2s band are only
half-filled, or rather, the lower half of the 2s band is filled. It is this
partially-filled band that accounts for the bonding in the metal. For a metal
to conduct an electric current, electrons must be excited to unfilled delocalized
orbitals which we see in this case are readily available.

2.3 Identification of potential high strength materials

After our study of covalent bond models, we can now focus the ideas so far
developed on ceramics that will become increasingly important in modern
technology. To recapitulate, we are searching for low atomic- or molecular-
weight materials, bound by covalent bonds. Section 2.2.5 suggests the
following potential elements:
(i) beryllium
(ii) boron
(iii) carbon
(iv) silicon

Combinations of these are also suitable:

(v) boron carbide
(vi) silicon carbide

as are combinations of these elements with non-metals:

(vii) boron oxide
(viii) boron nitride
(ix) silicon nitride
(x) silicon dioxide
(xi) aluminium oxide
(xii) some metallic carbides and nitrides.

Alumina, Al,O,, is dealt with in detail in 5.4; the metallic carbides and
nitrides form important parts of 6.1; silica is considered more fully in 1.5.4;
beryllium compounds, because of their poisonous nature, have a limited
appeal and will be ignored here. All the other members of the list will be
discussed in more or less detail in the remainder of this chapter.

2.4 Boron
Boron minerals are many, but only a few are of commercial interest. Their
origin is closely connected with present or past volcanic activity, and
arid climates are required for the formation of all commercial deposits of
 SPECIAL LIGHT CERAMICS 89

borates. The borax industry has often been controlled by the USA, mainly
because the mineral kernite (Na,B,O,:4H,O) was found in huge deposits in
Kern County, California. It dissolves easily in water, from which borax
(Na,B,0,-10H,O) crystallizes out on concentrating to make it an easily
extracted mineral. Other less useful sources are colemanite (Ca,B;O,,:5H,O)
and boracite (Mg,Cl,B,,03) from the Stassfurt deposits. The chief use of
borax is in the manufacture of porcelain enamels (3.5) and glass compositions
(3.1). So far little boron has been used in its elemental form, even though it is
amongst the hardest and lightest of materials, probably because it is one of
the most difficult of all elements to obtain ina state of purity. This arises from
the very vigorous conditions that must be employed to liberate it, the fact
that so many elements form borides (6.2), and even though the crystalline
element is very inert, in the amorphous state it is very reactive and frequently
combines with the reducing agent employed, with the crucible materials,
and with nitrogen. Reduction of B,O, or borax by magnesium, calcium,
aluminium or iron is used to obtain a crude product suitable for further
purification.'”
In engineering, boron is not usually used in the massive form but rather as
reinforcing and stiffening fibres. These are made by deposition on to a very
fine tungsten wire, in the region of 10-12 um diameter, utilizing the reaction
2BCl, + 3H, — 2B+6HCI. The tungsten wire is resistively heated to 1100°C,
which is the optimum temperature for producing fibres with very small
crystals in the 2—3nm region. It is difficult to maintain close control of the
process as the coating thickens and normal production gives a product about
0.1mm thick. Table 2.1 shows that fibres obtained in this way are extremely
strong and stiff, but they cannot be used above 700°C, far below the normal
melting point of the element at 2300°C. They are usually available as tapes or
spun fabrics for inclusion in composites, but the method of production makes
them expensive—around £250 per kg. High-technology applications where
weight saving is important, such as the boron/aluminium mid-fuselage truss
members used for their high strength and stiffness to weight ratio in the
American space shuttle, are therefore the only type of application in which
they are encountered to date.

2.4.1 Structure and bonding

Very many allotropic modifications of boron have been claimed, but this
probably reflects the extreme reactivity of hot boron and the ease with which
it forms binary borides. It has been suggested that the initial deposit of boron
reacts to form a boride with its own predetermined structure which persists
during the rapid deposition stage. If the boron layer grows much more
rapidly than the substrate metal, atoms can diffuse outward, and an outside
piece would analyse as “pure” boron whilst still having the template structure
of the initial boride. Working from this hypothesis to eliminate several forms,
it seems that there are at the most three structural modifications of the
90 CERAMIC SCIENCE

Figure 2.10 An icosahedron, as found in boron.

element. When prepared in the temperature range 800—1100°C the «-rhombo-

hedral form occurs. When prepared above 1250°C, or when the «-form is
heated above this temperature, the B-rhombohedral form is obtained. Finally
when prepared by H, reduction at 1300°C, a tetragonal form exists. Only the
structures of the «-rhombohedral and tetragonal forms have been established
and it is in these that we must perceive the reasons for the high strength,
stiffness and melting point.
Both structures are unique amongst the elements in that they consist of
icosahedral units formed from 12 boron atoms. An icosahedron can be
described as two pentagonal pyramids superimposed at the base, but with
one basal plane rotated through 36° with respect to the other so that the
corners do not lie above each other (Fig. 2.10).
A molecular orbital treatment (2.2.7), of this unit provides some insight
into the stability of the material. Each boron atom has one 2s and three 2p
atomic orbitals; the 2s and the 2p, are hybridized, still providing 4 orbitals
on each boron, but one of the sp hybrids is directed away from the
icosahedron. One of the sp hybrids and the p,., p, orbitals on each boron atom
point in some degree to the inside of the icosahedron and as a result of their
close proximity must mix when symmetry allows. The result of this is that
there are thirty-six molecular orbitals formed within the structure, only
thirteen of which are bonding, the remainder being non-bonding or anti-
bonding. Now considering the electron structure of boron (2.2.5), we can see
that the twelve atoms have thirty-six electrons to be allocated to the available
M.O.’s. If one is assigned to the outward-pointing sp hybrid ready to combine
with other such hybrids from nearby icosahedra in the form of localized
valence bonds linking the structure together (2.2.6), then there are twenty-
four for the inner icosahedral orbitals. Clearly all these can be accommodated
in the bonding levels, still leaving two unoccupied, and thus all twenty-four
electrons contribute to the high stability of the B,, unit.
The structure of the tetragonal form then follows simply from this theory.
Close-packing of the icosahedra in the cubic manner means that each unit
has six icosahedron near neighbours in one plane, three above and three
below. In the plane with the six nearest icosahedra the structure can be
 SPECIAL LIGHT CERAMICS 91
viewed as planes of regular hexagons, atoms 1, 2, 3, 4, 5 and 6 in Fig. 2.10;
and above and below these planes are the equilateral triangles formed by
atoms 7, 8, 9. The localized orbital of each of the hexagonal ring atoms
overlaps with two similar orbitals from separate adjacent icosahedra which
have appropriate symmetry to form a tricentric molecular orbital. Since the
tricentric M.O. contains only two electrons, it is not as strongly binding as
the intra-icosahedral bonds, and B-B distances are about 15% longer
between icosahedra than the B-B distance within them.
The structure of the a-rhombohedral form is more complex because the
unit cell contains two isolated B atoms together with four of the B,, groups.
Each isolated B atom has four icosahedra near neighbours distributed
tetrahedrally to which it is bonded by the sp localized hybrids. In order to
achieve this there must be sp? hybridization involved on the isolated atoms.

2.5 Carbon and graphite

Carbon can be obtained from natural deposits in three different crystalline
modifications: «- and B-graphite, and diamond.
Graphite is found mainly in metamorphic rocks such as gneisses, schists
and metamorphosed limestones which have been subjected to heat and
pressure. Some naturally-occurring graphites contain as much as 30% of the
rhombohedral £-form, but synthetic graphites contain only the «-form. The
a-form can be converted to the B by mechanical working of the material, but
when heated above 1000°C the f is converted to the a-form.
The formation of natural diamonds in kimberlite pipes is a matter of some
speculation. Diamond (Fig. 2.11), the hardest known material, is transparent
to translucent, and may contain inclusions of graphite or other minerals
leading to colours ranging from black to colourless with examples of yellow,
red, brown, green and blue. Graphite inclusions arise because it is the
thermodynamically more stable form of carbon; transformation is extremely
slow because the structural change is profound even though, as Fig. 2.12
shows, both forms contain networks of six-membered carbon rings. The
reverse transition from graphite to diamond has been much sought after since

Figure 2.11 Structure of diamond. Six-membered rings in a folded chair configuration can be
discerned in horizontal and vertical planes.
92 CERAMIC SCIENCE

Table 2.5 Examples of technical carbons and graphites

Material Organic precursor Preparation

Metallurgical coke coal carbonization in retorts

: mixing
Electrode gtaphites petroleum coke filler with extrusion
Nuclear graphites coal tar pitch binder baking
Refractory carbons ) graphitization
Carbon blacks hydrocarbon gases incomplete combustion
Pyrolytic carbon thermal decomposition
Pyrolytic graphite hydrocarbon gases 1250-2750°C
Carbon fibres polyacrylonitrile fibres —
Glassy carbon phenolic resin ~

the energy difference between the two forms is only of the order of 2100J/g
atom. Experiments have shown the equilibrium curve to be defined by the
equation P = 7100+27T, where P is in atmospheres and T is in °K; thus
pressures above 50000 atmospheres, 5.07 x 10°kNm ”, are required above
1300°C to effect the change.
The terms “carbon” and “graphite” are also generic terms for a wide range
of carbonaceous materials.*° Table 2.5 gives examples of some technically
important carbons and graphites. The products in Table 2.5 have several
structures ranging from the near perfect crystalline structure of pyrolytic
graphite (see below), to the essentially amorphous structure of glassy carbon
and carbon blacks. All manufactured graphites consist of microcrystalline
graphitic regions in an amorphous matrix. In the graphitic areas there is no
ordered stacking sequence of graphitic planes within the crystallites and such
structures are called “turbostratic”.
The route to graphite is via carbonization, which is a two-stage process:
primary carbonization takes place up to 550°C involving the breakdown of
the organic precursor and expulsion of volatiles. Secondary carbonization
then occurs in the range 550—1150°C when polynuclear aromatic compounds |
condense to form a carbon residue by elimination of hydrogen and side chain
aliphatics as methane.
During the primary phase it is often found that anisotropic droplets
separate from the fluid mass. Such droplets are called the mesophase and
contain polynuclear aromatic molecules oriented to lie parallel to the equator
of the droplet. Extensive graphitization of carbon occurs only if the mass
forms a coalesced mesophase. Hence mesophased carbons are called “soft”,
or graphitizing, while non-mesoforming masses produce carbons termed
“hard”. Heavily crosslinked precursors usually produce hard carbons.
The transition from carbonization to graphitization is a continuous one
which varies widely with the starting material and is accompanied by a
decrease in interlayer spacing of the product from 0.344 nm towards 0.335 nm.
Self-diffusion of carbon atoms by a vacancy mechanism parallel to basal
 SPECIAL LIGHT CERAMICS 93
planes is the dominant process giving rise to graphitization. The classical
view is that the degree of graphitization is determined by temperature rather
than by time at a given temperature, which implies that the process has a
spectrum of activation energies which become larger as the degree of
graphitization increases.
A major field of interest in carbons and graphites is the development of
porosity, since many of the uses of these materials depend on this feature of
their structure. Three types of pore are denoted: (a) macropores, those visible
in the optical microscope, (b) micropores, less than 3.0nm in size, and (c)
mesopores which lie in the range 3.0—200 nm. Of these, micropores make the
greatest contribution to the total surface areas, values of which lie in the
range 2 x 10°-5 x 10° m*kg~* for “hard” carbons to 250-500m*kg~! for
“soft” carbons. The low surface areas of soft-graphitizing carbons arise from
low porosity which is usually of the macropore type. The difference is
associated with the presence of liquid mesophase bubbles in the soft carbons.
For the microporous, hard carbons, there is no evidence for loss of porosity
on heat treatment; it seems that open micropores simply close, with little
change in total pore volume. Graphitizing carbons, however, do densify on
heating as pores between parallel crystallites are eliminated.

2.5.1 Structure and bonding

(a) Diamond. When diamond is considered in terms of the molecular orbital
theory, the observed tetrahedral distribution of four carbon neighbours
around each carbon in the structure is most readily interpreted in terms of
localized sp? hybrid orbital overlap from each carbon atom. Complete filling
of the bonding molecular orbital band by the carbon 2s and 2p electrons
leads to the expectation of very hard, colourless crystals.
Diamond has a cubic structure arising from occupation of the two
interpenetrating lattices of the zinc blende structure by one type of atom
(Fig. 2.11).
The layer sequence of carbon atoms is AA’ BB’ CC’ A etc. where A, B and
C denote different possible orientations of layers with respect to each other.
This is equivalent to A B C, indicating the cubic nature of the carbon layer
stacking. Fig. 2.11 shows the consistent tetrahedral co-ordination of carbon
atoms bonded with sp? hybrid orbitals giving rise to pure o bonds and
accounting for the 0.1555 nm bond length throughout the puckered hexagonal
network.
The cubic structure of diamond has a high symmetry, and so any section
through it crosses large numbers of bonds, each difficult to break; no easy
cleavage therefore exists.
(b) Graphite. The hexagonal form, that is the «-form, is the most commonly
encountered and has the structure illustrated in Fig. 2.12, where planar
hexagonal rings are clearly seen to be stacked in the AB AB... sequence
which gives rise to the hexagonal structure. Not unexpectedly, the f-
94 CERAMIC SCIENCE

Figure 2.12 (a) The structure of «-graphite. (b) The relation between cubic diamond and
rhombohedral f-graphite.

rhombohedral form has the same planar hexagonal nets, but stacked in the
sequence ABC ABC etc. Within each hexagonal layer the carbon atoms are
separated from each near neighbour by 0.1415 nm, implying a stronger bond
than in diamond, but the distance between layers at 0.3354 nm is too far to
permit serious bonding to occur. This loose layer packing accounts for the
low density of 2.22 x 10° kgm 3, compared to diamond (3.51 x 10? kgm °).
Such a highly anisotropic structure for graphite gives rise to the black
colour of the polycrystalline material or metallic appearance of the single
crystals, and to the softness and easy cleavage. It is worth examining it-in.
more detail in terms of the molecular orbital and valence bonding theories to
seek some rationalization. The dominant feature of the structure is the strong
hexagonal planar network. From Table 2.4, we can see that such a shape can
arise from localized overlap of half-filled sp? hybrid orbitals. The electronic
structure of carbon (2.2.5) allows sp? hybridization, since the energy gap
2s > 2p is only modest and allows unpairing of the 2s orbital to achieve the
excited state. Hybridization (2.2.6) from this excited state produces three’
half-filled sp? hybrids directed at 120° from each other, and leaves on each
carbon atom a half-filled 2p, orbital. Overlap of the sp? hybrids to produce
localized bonds between carbon atoms, directed at 120° from each other, will
give rise to the hexagonal network. Such strong bonds however will bring
each carbon close to its inter-layer neighbour and must then bring each 2p,
orbital, which is aligned at 90° to the plane of the sp” hybrids, into z-type
overlap. The close proximity of each carbon will mean that the resultant
M.O. energy level diagram is not discrete but broadens into a a band. 2N
carbon atoms will produce 2N levels, but since each 2p, orbital was only
half-filled, the bonding band will only be half full. This type of molecular
orbital formation is thus visualized as producing a completely delocalized
system which permits electron interaction throughout an entire hexagonal
layer. As a result, each carbon atom in the layers is bonded by 22 electrons,
 SPECIAL LIGHT CERAMICS 95

as opposed to two electrons per carbon in diamond. Hence the shorter C—C
distance in the layer is to be expected. Furthermore, all electrons are involved
in carbon-carbon o or x bonding in the layers, leaving none for inter-layer
bonding, hence the easy cleavage as the layers are bonded by weak van der
Waal’s forces.
Above the filled part of the z-band there are available many closely spaced
energy levels, which allow most radiation or imparted energies to be
absorbed in orbital promotion. This leads to the black colour and electrical
conductivity. The conductivity of graphite decreases with increasing tempera-
ture 1.e., it is metallic as predicted by this model—at room temperature there
are 107 +-10~° electrons per atom in the conduction band.
In single crystals the anisotropic behaviour arising from this structure is
clearly manifest. Along the layer stacking direction, that is the c axis
direction, graphite can behave as an insulator, while in the a, b directions it
has the metallic properties outlined. In the bulk form graphite has modified
anisotropic properties (Table 2.6) dependent on the average grain orientation.
Figure 2.12b shows how the operation of collapsing the puckered hexagonal
rings discernible in the cubic diamond structures converts it to rhombohedral
B-graphite. It follows from our discussion so far that such a transformation
corresponds to transforming the bonding orbitals from sp* to sp* + p, which
accounts for the slow rate of transformation from metastable diamond to
graphite.

2.5.2 Carbon fibres for engineering use

Carbon fibres are often called graphite fibres, and they can have very high
strengths and moduli (Table 2.1). They clearly, when comparing the properties
of bulk graphite, represent highly oriented forms of the element. Such fibres
are usually made from polymer fibre by careful heat treatment. For example,
polyacrylonitrile (PAN) fibres are heated at relatively low temperatures,

Table 2.6 Some properties of bulk graphite

Temperature With-the-grain Across-grains

(°C) average average

Young’s modulus R.T. 20.62 5.10

(GNm~?) 1400 23:92 5.79
2800 25.92 Tel
Tensile strength RoI; 30.37 10.55
(MN m7?) 1500 43.23 13.27
2500 69.46 16.86
Compressive strength Reis 48.85 86.87
(MNm ~~”)
Thermal conductivity R.T. 216.3 87.02
(Wm7!K~?)
Specific resistance eas 6.97 21.87
(ohm)
96 CERAMIC SCIENCE

about 230°C, in air under tension to oxidize and stabilize the fibres. They are
then heated to 2500°C under tension so that graphite planes are produced
oriented along the fibre axis.

230 °C
or

N N N N N
CN CN CN CN CN CN
1400-2400 °C
-Ny -HCN

Graphite layer

Depending on the degree of orientation of the starting polymer, variations

in the modulus and strength of the fibres can be achieved from a highly
oriented high modulus fibre (E about 400GNm~”) with lower strength,
about 2.0GNm~?, to a high strength (3.5GNm~*) but low modulus
(200GNm~*) from a less oriented structure. These quoted strengths of
commercially available fibres are far below the theoretical strengths derived
in 2.1 as a result of stress-raising surface flaws (3.4.2), a fact attested by the
factor of 10 drop in measured tensile strength of carbon fibres when the fibre
length is increased from 0.5cm to 5cm. Fibres obtained by this route are
supplied as clusters of fibres, called tows. These cost approximately £50 per
kg, and an attempt to avoid this high cost is the manufacture of graphitic
fibre from mesophase pitch.
Any carbon fibres as yet available, usually from 8—11 zm diameter, are far
from 100% crystalline; they must always be regarded as composite non-
homogeneous systems consisting of graphite crystals in amorphous carbon,
so that the final properties will depend on the degree of preferred orientation, '
the surface flaws, and the texture. Residual amorphous areas are responsible
for chemical degradations of strength as outlined below.

2.5.3. Graphite reactivity

Because diamond has an excess energy compared to graphite, heats of
reaction between the diamond form of carbon and other elements are greater.
However, because of the relatively large surface area of graphite it always
reacts faster than diamond. There are also z electrons in the easily accessible
region between the layers, which makes graphite much more reactive. Many
materials can be intercalated between the layers to modify the structure and
properties of graphite. For example, when graphite is oxidized in an aqueous
system, a so-called graphitic oxide is produced. In this material, the colour
and conductivity is lost as oxygen between the layers induces sp? hybridization
 SPECIAL LIGHT CERAMICS 97

with resultant buckling of the planar rings. Fluorine gas intercalates in this
way to produce the insulating, colourless material carbon monofluoride, CF.
Other materials such as potassium, Cl,, MoO,, and FeS, occupy between-
layer positions in an ordered way, giving rise to a C,X series of materials in
which the conductivity and colour of the graphite is modified but not
destroyed and the hexagonal layers are maintained. Intercalated materials
that add electrons to the graphite x band can enhance the electrical
conductivity.
Gas reactions of graphites are of some importance in several technologies,
such as metallurgy and nuclear power. The reactivity of a given carbon is
structure-dependent, because attack by an oxidizing gas has to occur at
atoms situated at the edge of graphite planes. Attack at atoms within the
basal planes is very slow, but residual amorphous areas are easily attacked.
All graphites show a useful and at first sight unexpected feature: rates of
oxidation fall sharply with increase in temperature. This is important to
nuclear power engineers in the oxidation of carbon in CO, atmospheres.

2.6 Silicon
Although exceedingly abundant throughout the world when in combination
with oxygen, silicon does not occur as the free element. It has been identified
(2.3) as a potentially useful high-strength engineering material but has
not been exploited, probably because no readily available route to fibre
production exists. Whisker crystals have been produced, but with ultimate
tensile strengths no greater than 3.799GNm* which compares poorly with
the carbon whisker values around 20.68GNm~?. Furthermore, although
they can be obtained in the form of whisker crystals, and ceramic whiskers
are potentially the strongest of all reinforcing materials, because of their
reactivity with metal matrices, their manufacturing difficulties and their high
cost, they excite little interest as components in modern composites. However,
the lack of engineering applications for silicon pales into insignificance when
its development as a semiconductor fuelling the revolution in electronics is
considered.

2.6.1 Preparation and structure

The element is obtained from the reduction of silica by carbon. It is relatively
easily melted, (m.pt. 1410°C), and from the liquid large crystals can be pulled
by the Czochralski method.
Although much single crystal material is used in device manufacture,
polycrystalline silicon (polysilicon) is extensively used too, particularly in the
form of thin films formed by pyrolytic deposition on to silica substrates.
Deposition is from silane, SiH,, in argon at 650—800°C by a three-step
process involving adsorption and decomposition of SiH, on, and H, desorp-
tion off, the silica surface.*®
98 CERAMIC SCIENCE

Silicon crystallizes with the same structure as diamond (2.5). It has

covalent bonds formed from sp? hybridization of the silicon ground state
(Table 2.3). The molecular orbital description developed for diamond (2.2)
applies directly to silicon, but because in Si the third quantum ievel is
involved, the gap between the filled valence band and the upper empty
conduction band is reduced to 1.09eV. A simple calculation will show that
this gap can be crossed by appreciable numbers of electrons when the silicon
is heated.
Assume a Boltzmann distribution of the electrons in the upper levels of the
valence band. Then at 1000°K, the ratio of the number of electrons thermally
promoted to those remaining will be
exp (— 1.09 x 1.602 x 107 !7/1.3805 x 107 *® x 2 x 1000) = 1.78 x 1K debs
Thus about one in 700 electrons will acquire enough thermal energy to cross
the band gap. In the conduction band these electrons will have many energy
levels available, making it possible for them to contribute to conductivity.
They will only contribute to the conductivity, and not be the sole reason for
it, since the levels they leave in the valence band will be available for
lower-lying electrons in that band and so permit excitation of electrons into
these “holes”.
Silicon is therefore in the M.O. model an intrinsic semiconductor with an
electronic conductivity increasing exponentially with increasing temperature.
The room temperature conductivity is about 10~* ohm~ + cm“ + which can be
compared to the values for metals around 10° ohm~‘cm~? and insulators
like NaCl at 107!? ohm~!cm~?.

2.6.2 Hole conductivity

Instead of regarding the conductivity of silicon as the sum of electron
movements in the conduction band and in the valence band, it is common to
regard the filling of holes in the valence band by lower energy electrons as the
movement downward of the “hole”. A hole can then be regarded as a point’
positive charge, and the holes can be thought of as occupying a nearly empty
“hole band”, thus becoming the mirror image of the nearly empty conduction
band with its few electrons. Therefore hole charge carriers have positive
charge to distinguish them from electron charge carriers; they are also
distinguished by their different mobilities; holes, because of their occupancy
of a lower energy band, appear to be more mobile than electrons.
The ratio of electrons to holes in a silicon sample can be greatly altered if
impurity atoms are put into the silicon structure. For example, if a phosphorus
atom replaces a silicon atom it has five valence electrons (2.2), four of which
will be accommodated in the valence band, but the fifth must become a higher
energy orbital. In practice this new energy state is just below the conduction
band, only about 0.012 eV below, and so it becomes very easy to excite this
electron up to the conduction band. In this way a semiconductor with an
 SPECIAL LIGHT CERAMICS 99

enhanced number of conduction electrons is produced. This is an n-type

semiconductor because it has an excess of negative electrons compared to
positive holes. Atoms like P, Sb, As are known as donors and the new levels
they produce in the molecular orbital band structure are donor levels.
If a dopant such as boron is used, the electronic structure, (Table 2.3),
indicates that it cannot completely partake in sp* band formation, with the
result that there are new levels introduced into the band gap just above the
top of the valence band. Electrons from the valence band can easily be excited
into these new levels, thereby generating more holes in the valence band to
be current carriers. This will be a p-type semiconductor.

2.6.3. p—n junctions

When an n-type silicon contacts a p-type silicon, excess electrons in the
conduction band of the n-type spill over and positive holes on the p side move
to the valence band of the n-type, making the n-side positive and the p-side
negative. This double layer of charge builds up to prevent further transfer and
establish equilibrium. At equilibrium the flow of electrons from n to p equals
the flow from p to n. If a positive terminal from a battery is connected to the
p side, the effect is to lower the double layer barrier so that electrons can
move again from n to p and holes fromp to n. As the voltage is increased, the
current across the junction increases. Should the polarity be reversed, then
the double layer barrier will be heightened and electron-hole transfer is
drastically reduced. In this way the p—n junction acts as a rectifier. This is an
important use for this ceramic.
A more important use is as an amplifier or transistor, when p—n—p or
n—p—n junctions are made. Usually a very thin piece of n-type silicon, called
the base, separates two p-type pieces called the emitter and collector
respectively. A positive bias at the emitter side injects positive holes from the
p-type into the n-type base, and since the base is thin, these pass through
aided by a negative bias at the collector. Adjustment of resistances on either
side makes this device a voltage amplifier (Fig. 2.13). This powerful device can
be made very small by selective doping of regions on the same silicon chip,
and is the basis of a large and important industry.

emitter base collector

Figure 2.13 A schematic p—n—p junction transistor.

100 CERAMIC SCIENCE

2.7 Boron carbide

Boron carbide, B,C, is used for its mechanical properties of strength and
extreme hardness in the manufacture of armour-plated body protection. It is
also used in rod form as a neutron absorber in nuclear power applications
where, although it bears little structural loading, its ability to withstand high
thermal stress fields is important. Its most promising strength and modulus
properties have not been realized as yet in fibre form because of the failure to
develop a fibre-forming process. It does, however, sometimes appear as
vapour-deposited coatings on boron fibres, affording them greater protection
from matrix interactions at elevated temperatures.
B,C is manufactured by reduction of B,O3 by carbon in an electric
furnace. Dense artefacts are made by hot pressing to porosities in the range
0-8°%. Hot pressed samples appear to retain their room temperature
strengths of 275 to 415 MNm ~ ? up to 1400°C.? They have values for Young’s
modulus around 310 GNm ~”, but this depends on porosity according to the
relationship

E = 460[(1—p)/(1+3p)]GNm
~? (2.26)
where pis the porosity.

2.7.1 Structure
Not unexpectedly, the rigid three-dimensional structure of B,C resembles
boron itself, in which regular icosahedra are present. The structural units can
be viewed as linear chains of three carbon atoms along with the B,,
icosahedra distributed as if in a NaCl structure; at each Na®* site there is a
B,, icosahedron centred on the site, and the central carbon atom of each
carbon chain occupies the Cl~ site. As described for boron, (2.4), the
icosahedra are connected by strong sp hybrid orbital overlap to other
icosahedra, but there is in this case also some B-—C covalent bonding of the
sp-sp hybrid type. Bond distances are B—B = 0.174 to 0.180nm, B—-C =
0.164 nm and C—C = 0.139 nm.
The molecular orbital description of the intra-icosahedral orbitals (2.5)
accounts for the electrical conductivity and overall strong bonding, leading
to hardness and chemical inertness.
The composition B,C is an idealized one, for it is easy to see that boron
can replace the central carbon of the chain of three in order to produce more
effective sp—sp bonding with the neighbouring icosahedra, or carbon can be
missing. When this occurs B,3C, becomes the formula and it is possible to
get a range from B,,C; to B,C, by this mechanism.
Since carbon clearly shows sp hybridization, some excess carbon might be
expected to be accommodated in the B,, icosahedra to produce a carbon rich
B,C, ,, variant. However, there is some doubt if such a phase exists, and such
compositions are not just graphite + B,C mixtures.
 SPECIAL LIGHT CERAMICS 101

2.8 Boron nitride

This material (often known as white graphite) is obtainable in two crystalline
forms: cubic diamond and, most commonly, a hexagonal layer structure.
Reasons for this are easily seen in the bonding treatment given below. It is
obtainable in hot-pressed, pyrolytic vapour-deposited, and fibre forms. Very
few other materials have comparable hot strength to the dense form.
BN is isoelectronic with carbon, and so, not surprisingly, the structures of
the two forms are analogous to graphite and diamond. The diamond form is
made only at very high pressures, and as a result is not often encountered.
The hexagonal layer form consists of alternate boron and nitrogen atoms
bound by three bonds in the plane of the layers. In their ground states, neither
element has three half-filled orbitals necessary for coplanar bonding, and
therefore consideration must be given to excited hybridized states for each.
For this it is possible to envisage the scheme:
B = 1s?2s?2p > B* = 1s?2s2p,2p,2p, > 1s°2p,2(sp?)'
(sp) (sp)?
which has the appropriate three valence orbitals, and
N = 1s?2s?2p3 + N* = 1s?2s2p,p,p2 > 1s?2p?2(sp”)* (sp?)*(sp?)’.
Covalent pair bonds between the half-filled sp* valence orbitals will lead to a
hexagonal BN planar network of the type:
i |
a NG
| | |
7 B_ Ni = N~ =

The p, orbitals on B and Nare, like graphite, perpendicular to the plane of

the network, but whilst the boron orbital is empty, the nitrogen orbital is
filled. Because the energies of the p orbitals on boron and nitrogen are quite
different, they do not conform to the orbital interaction rules (2.2) and so
there is little interaction, with no delocalization as was the case in graphite.
Thus, although it has the same structure, BN is an insulator, not a conductor,
and is white or colourless.
The presence of the lone electron pair on each nitrogen atom does lead to
a slight difference in structure between BN and graphite. It appears to order
the layer sequence so that they are directly superimposed, and below each
nitrogen is a boron atom; presumably in this configuration a little weak
donor-covalence from N-—B is possible. B—N distance in the layers is
0.145 nm and B-N distance between layers is 0.333 nm. Such weak interlayer
bonds account for the soft feel of the material. The hexagonal unit cell has
a = 0.2504 nm, c = 0.661 nm with a density of 2.27 x 10° kgm~°.

2.8.1 Preparation and properties of pyrolytic boron nitride’

Preparation of this material is often by vapour deposition from one of the
102 CERAMIC SCIENCE

reactions given below. When produced this way, it contains no binder

material, and has an increased degree of crystal orientation.

(a) B,H,+2NH, > 2BN + 6H,

(b) BCL,+NH, —°-BN+3HCI
(c) B,N3,H,Cl, 2" 3BN+3HCl
(B-trichloroborazole)

At temperatures above those indicated, the only product of these reactions is

BN. Below 1000°C the product is amorphous, between 1000—1200°C it is
polycrystalline but has little orientation and above 1800°C the product is
highly oriented and as a result has obvious anisotropic properties. Above
2100°C, the dissociation pressure of N, becomes appreciable, and to sublime
the material one atmosphere of N, is needed at 2330°C. The microstructure
shows cone-like features growing at 90° to the substrate surface; high growth
rates lead to more cones and internal stresses are generated between them.
The combination of stress and the planar structure can lead to failure by
delamination.
(a) Mechanical strength. Unlike many ceramics, thin layers and narrow strips
of BN are quite flexible. When Young’s modulus is measured in compression
it shows considerable anisotropy, as expected from the structure, with values
of 7 x 10° Nm “?in the c-direction and 2 x 10? Nm~? in the a-direction. In the
a-direction the UTS increases from 4.2MNm~? at room temperature to
10.6 MNm ? at 2200°C. This is a feature unlike almost all ceramics.
(b) Thermal properties. It has extremely good thermal shock characteristics,
again arising from the anisotropic nature. The c-axis expansion is linear but
the a-axis contracts from 0—800°C prior to expansion in this direction.
In the a-direction the thermal conductivity is as good as metallic iron at
approximately 2.5Wcm~'K~!, but only one-hundredth that value in the
c-direction. Only hot pressed BeO exceeds the value for the a-direction and
combines such thermal conductivity with electrical insulating behaviour. As
a result of this feature, oriented BN is replacing toxic BeO where thermally
conducting insulators are required.
Pyrolytic BN crucibles are a natural development from this combination
of properties, for if the a-direction is parallel to the crucible walls, then heat
flows round them easily and so minimizes the effects of furnace hot spots.
(c) Chemical properties. Like the majority of refractories, BN is not thermo-
dynamically stable in service environments, and kinetic barriers have to be
erected against deterioration (5.2). In the case of BN, that barrier is a glassy
film of B,O3, and subsequent decomposition rates depend on oxygen
diffusion through the layer. A simple calculation using the data in Table 5.2
for standard free energies for the reactions

BN, Btn (2.27)

B,03q) > 2B) +30 2¢) (2.28)
 SPECIAL LIGHT CERAMICS 103

0-166 Pyrolytic

0-0166

Oxidation rate
Kg ms! Pyrolytic BN
0-00166

0:000166
1400 - 2000
Temperature C

Figure 2.14 Oxidation characteristics of graphite and boron nitride.

confirms the potential instability of the system and the vital role played by
the kinetic barrier. The overall reaction

2BN) +30 2/9) > B2O3q) + No@)

is obtained from the above decomposition reactions by multiplying (2.27) by
two and subtracting (2.28). Using equation (5.5) and the average values of
AH, and AS, for 1000°K to obtain AG° for the overall reaction at 1000°K we
get

2[250382 — 1000(87.53)] — [1227666 — 1000(209.84)] = —692.1 KJ

as the thermodynamic driving force for the oxidation process. Hence the low
rate of oxidation up to 1300°C is a strong indication of the protective nature
of the oxide film and the strong kinetic barrier it raises (Fig. 2.14).
As well as having oxidation resistance, BN is chemically very inert—most
molten metals do not wet or attack it. It is inert to most acids or acidic fused
salts, but over long periods of exposure to molten alkalis, borates, or aqueous
alkaline solutions it is gradually attacked. Moisture in the atmosphere above
100°C causes slow hydrolysis.
Strong sp?—sp? covalent bonds and the B,O, outer layer are responsible
for these favourable properties.

2.9 Boric oxide

This is obtained by heating boric acid, H,BO3;, to red heat. Although B,O,
is plentiful, its only large scale use is in the glass industry, because of two
obvious problems. First, it has a low melting point at 450°C, unexpected in
view of the suggested sp?—sp* bonding (2.2) and the structure which suggests
B-O bonds of high strength—see (a), p. 104. Secondly, it takes up water very
rapidly to become the layer compound B(OH);.
104 CERAMIC SCIENCE

Structural complexities are probably encountered in the system, because

only a small extra energy is needed to change boron hybridization from sp”
to sp? with a resultant change in boron coordination from three to four. This
change is effected by pressure or the presence of polarizable oxygen atoms
introduced by addition of oxide. Boric oxide glasses contain large percentages
of tetrahedral boron as well as triangular planar boron atoms. This picture
leads to structural models containing boroxole ring groups linked by planar
triangles (b). Further structural complications arise from the ability of the
BO, triangular units to share common edges and common faces as the
temperature rises (c).

In view of the stability of BN and its relative structural simplicity

compared to B,O,, it is perhaps not surprising to find that B,O, glasses’
absorb large amounts of nitrogen as B—N—B groups are formed. This

B
behaviour leads to eventual devitrification. Hence, although this is an
important ceramic/glass material, it only finds use (in the engineering sense
of this chapter) when forming part of the composition of strong glass fibres
such as E glass (3.1).

2.10 Silicon carbide

This ceramic is available as reaction-sintered or hot-pressed material, or in
the form of fibres made by chemical vapour deposition.'* It has a greenish
appearance when impure, but is black when pure.
SiC fibres are mostly made by a process like that used for the manufacture
 SPECIAL LIGHT CERAMICS 105
of boron fibres and chemical vapour deposition is achieved on a thin, hot
tungsten filament. Chlorinated silicones are used with a hydrogen carrier gas.

CH,SiHCl, aa SiC + 2HC+H,

methyl dichlorosilane

Fibres with very acceptable mechanical properties (Table 2.1) are produced
by this method, but they cannot be used above 900°C because of their
chemical reaction with the tungsten core. This is a drawback, because SiC has
widespread applications as a refractory material in oxidizing environments.
Fibres from this route are usually quite thick, around 0.1 mm diameter, and
cost more than boron or carbon fibre—around £550kg~‘. Attempts to
produce thinner fibres without W inserts resulted in the polycarbosilanes,
which can be melt spun into continuous fibre before vacuum pyrolysis at
1300°C to produce the B-SiC form of fibre. Made this way the fibres are
about 15um in diameter and have E values of 245GNm~? and tensile
strengths around 3.0GNm ~? as compared to the values in Table 2.1. These
fibres can be used at higher temperatures, and even though they are
well-wetted by many molten metals appear to be very inert chemically.

2.10.1 Structure
The electronic structures of the two elements indicate that the valency bond
approach (2.2) might be most rewarding. Both elements have relatively easy
access to hybridized states and sp? hybrids, each half-filled, which allows the
formation of three-dimensional covalent bonds imparting giant molecule
characteristics. In this arrangement each atom will be surrounded by four
tetrahedrally disposed neighbours. The crystalline forms so far identified
reveal that the Si and C atoms alternate throughout the structures. In
principle the crystalline forms that exist are related to the cubic diamond
structure and the hexagonal wurtzite structure. Figure 2.15 shows that these
two structures are themselves closely related. The a-forms of SiC have
structures related to both zinc blende and wurtzite, the simplest being
wurtzite-like alone. The f-form of SiC has the cubic zinc blende structure.
The atomic positions in zinc blende are the same as in diamond, but alternate
atoms are Zn and S.

A rom “Lp v Cc
if bi a
1
B eS B
iT

Figure 2.15 Pseudo-layer representation of (a) wurtzite form of SiC in which horizontal 6-
membered Si-C rings have the chair configuration and vertical puckered hexagonal rings have a
boat configuration; (b) zinc blende structure of SiC in which both horizontal and vertical rings
have the chair form.
106 CERAMIC SCIENCE

The zinc blende and wurtzite structures are three-dimensional, not layer
structures in the sense of BN and graphite, but we can discern within them
for descriptive purposes layers comprising puckered 6-membered rings, as
Fig. 2.15 shows. When viewed in this pseudo-layer manner it is possible to see
howa variety of SiC structures might exist as periodic interweaving of faults
in the packing sequence which occur during crystal growth. For example, the
AB sequence giving rise to the wurtzite form of «-SiC, having the Ramsdell
symbol 2H (5.10), may be transformed to 6H or ABCBAC layer alternation
if such layer displacement faults occur preferentially on every third layer:
2H = ABABABAB...
faulted 2H = ABABCBAB...
and with preferential 3rd layer faulting:
2H = ABABABABAB...
6H = ABCBACABCEB...
Thus there are several forms of «-SiC reported which have varying c-axis
parameters, depending upon the number of layers in the X-ray repeat units,
e.g. 2 for 2H, 6 for 6H as shown. Some polymorphs have c-axes as large as
8.29 nm.
The diamond structure and the silicon structures have been described in
molecular orbital terms (2.5, 2.6), from which we see that the band gap in
diamond is large, whilst that in silicon is small enough to produce intrinsic
semiconduction. It is easy from this to anticipate that f-SiC at least will have
intermediate behaviour which can be manipulated by suitable doping to
provide an electrically conducting refractory. Herein lies one of the uses of the
material as an electrical furnace heating element.
The strong covalent bonding throughout all forms of SiC make it easy to
see why it is so refractory and also is hard enough to be used extensively as
an abrasive.

2.10.2 Oxidation
Like all non-oxide ceramics, the apparent lack of oxidation and chemical
inertness is based on a kinetic barrier. In this case it is a coating of silica that
all samples used in the air soon acquire.
Oxidation of SiC in the air takes place according to the simple equation
SiC) = 20) =». S10 9(s)ar CO).

The silica is formed as a coherent layer on the surface and slows down further
oxidation. Because the protection depends on silica, SiC refractories must
only be used when water vapour is severely restricted. The solubility of H,O
in SiO, is 1000 times greater than oxygen because of its facility to break
oxygen bridges? (3.2).
 SPECIAL LIGHT CERAMICS 107

ve NK
—Si—O—Si— + H,O = 2—Si—OH.
we aX Yo
The enhanced transfer of H,O to the underlying SiC facilitates the oxidation
reaction:

2.11 Silicon nitride

For some time, silicon nitride has been at the forefront of developments in
high-strength, high-temperature materials, because it has been the centre of
efforts to develop a more efficient, ceramic-based gas turbine able to run at
temperatures beyond the capability of metal alloys.'°

2.11.1 Structure
Electron energy levels in Si suggest relatively easy unpairing and promotion
to an excited state sp* configuration. Such a configuration produces the usual
tetrahedral arrangement of valence orbitals for covalent bond formation with
four nitrogen atoms, resulting in an SiN, tetrahedron that becomes the
building unit of the Si,N, structure. These tetrahedra form a 3-D network by
sharing corners so that each N atom is common to three tetrahedra. It is
worth contrasting the silicates, where SiO, tetrahedra are connected by
corners, but each oxygen is common to only two tetrahedra (1.5). A weakness
in silicate ceramics is that of the bridging link between tetrahedra which gives
rise to variable Si-O-Si bond angles and tilted and twisted tetrahedra,
leading to a variety of silicate polymorphic structures. This weakness is
removed in silicon nitride, where the greater rigidity of the tetrahedral
bridging units gives rise to the outstanding properties of the materials and
the occurrence of only one crystal structure, namely hexagonal f-Si,N,,
a = 0.7600 nm, c = 0.2908 nm. Although there is only one polymorph it has
the designation f, because it was thought, probably incorrectly, that another
form, a-Si,N,, existed. Now the favoured opinion is that the «-form is in fact
an oxynitride; this has led to the rapid development of a completely new
range of ceramics, a development outlined below. The structure of B-Si,N, is
shown in Fig. 2.16 which indicates that the SiN, tetrahedra are linked with
one edge parallel to, and one edge perpendicular to, the (001) plane, with very
little distortion of the tetrahedra or Si,N groups. A slight distortion occurs
because the Si-Si and N—N distances are forced to be equal in the c-direction
by small deviations in the N—Si—N and Si-N-Si angles, so that Si,;N groups
perpendicular to the (001) plane deviate slightly from a strict planar arrange-
ment. One view of the f-structure is that of superimposed layers of linked,
puckered rings formed by joining six almost unstrained SiN, tetrahedra. This
108 CERAMIC SCIENCE

GB = silicon

@ = nitrogen

Figure 2.16 The crystal structure of B—Si,;N, showing two formula units per unit cell.

description has parallels with those of BN, graphite, and SiC already given in
2.5, 2.8 and 2.10.

2.11.2 Properties and fabrication

The intrinsic properties of B-Si,N, arising from the covalent bonding and
crystal structure are excellent, but quoted properties are variable, reflecting
the fabrication problems.
(a) High strength. Values in excess of 1000 MN m “? for flexural strength are
obtained, and can be maintained to temperatures about 1000°C. Above
1100°C where ceramics must operate to surpass metallic materials in strength
properties, considerable degradation of strength occurs. This is reminiscent
of Al,O, (5.4, Fig. 5.6), where viscous flow of the grain boundary phase
nucleates microflaws for subsequent propagation.
The high-temperature strength degradation is due to the onset of slow
crack growth. Slow crack growth processes are characterized by the crack
velocity and stress intensity factor K; as described in (3.4) and (5.4). In a :
typical crack growth rate curve (Fig. 2.17) the K;c values for Si,N, rise from
5MNm_?? at 200°C to 11.5 MNm_7 ? at 1350°C.
(b) High wear resistance.
(c) High decomposition temperature. At 1900°C appreciable vapour pressures
of Si and N, above the solid are detected. Data in Table 5.2 show that the
reaction Si;3Naj) = 3Siq) + 2Nog) in the range 1410-1700°C is endothermic
to the extent of 873.6 — 1973 x 0.405 = 75.3kJ mol™?.
(d) Oxidation and corrosion resistance. The notable corrosion resistance and
stability at high temperatures in oxygen can only come from kinetic sources.
This can be shown using two of the equations given in Table 5.2:
SiO 4/5) = Si) + Og) (2.29)
 SPECIAL LIGHT CERAMICS 109

°
1200 C °
1250 © 4350 °C

crack
velocity

3 5 Tie 3%
K. I MNm2

Figure 2.17 Crack growth characteristics in hot pressed Si,N,.

Reversing (2.29) and multiplying by three gives: Si,N,+30, = 3SiO,+2N,

which denotes instability when the expression AG = AH —TAS is used to
calculate the mean energy changes. The mean values for the temperature
ranges given in Table 5.2 lead to — 1750kJ for the reaction or — 291.6kJ/gram
atom of oxygen, as such reactions are often expressed. Clearly then such a
large negative free energy demands a strong kinetic barrier to bestow
resistance.
Oxidation resistance is paralleled by the stability of the nitride in acid
media and the excellent compatibility with molten metals; there is no sign of
attack after 30 minutes in iron at 1600°C.
(e) Thermal shock resistance. The excellent thermal shock resistance of Si,N,
is obviously related to the strength of the Si-N bonds and the low thermal
expansion coefficient.
This arises mainly because the nitrogen atoms are in three-fold coordination
with silicon, and the tetrahedra are rigidly linked. There is no tilting or
twisting of tetrahedra like that found in many silicates (1.5) which can give
rise to relatively large thermal expansion effects as the tetrahedra untwist,
untilt or even tilt further. The a and c axis expansion of f-Si,N, is given by
the expressions:
a = 0.75999(1 + 7.95 x 10° 7T 42.43 x 10° T’)
and
c = 0.29059(1 + 1.955 x 107°
T+ 1.762 x 10°? T”)
where T is the temperature in the range 0—1000°C. Even though the structure
is fully extended and rigid, the thermal expansion is seen from the above
expressions not to be isotropic. The slight degree of anisotropy is thought to
be due to the small difference in tetrahedral distortion in the a and c
directions. A mean value for the thermal expansion coefficient arising from
Si—N bond extension is 3.7 x 10°° K~?.
110 CERAMIC SCIENCE

These properties should make f-Si;N, an ideal material, and indeed it has
been called “the ceramic of the decade” for over 20 years. Why then has it
failed to meet its expectations? The answer lies in a particular fabrication
problem that will be dealt with in 2.11.4.

2.11.3 «-Silicon nitride and silicon oxynitride

Opinions differ at the moment concerning the composition and structure of
a-SizN,. There is the well-supported view of Jack et al. that «-Si,N, is a
nitride of high oxygen potential, ic. a ternary phase that could be called an
oxynitride, in which oxygen replaces some nitrogen and some silicon sites are
vacant to maintain electrical neutrality. About one in 30 nitrogens are
replaced by oxygen and about one silicon site per 120 is vacant. Because of
the non-stoichiometry, the material is quoted as having a range of com-
positions from Si, , 4N,5Oo.3 to Si,;.5N 500.5. It has a hexagonal unit cell,
like the B-phase, but with a = 0.7749 nm and c = 0.5619 nm.
Other workers believe «-Si,;N, to be a true binary phase with a different
stacking arrangement of the types of layer found in f-Si,N,. Resolution of
this disagreement seems to lie in a simple oxygen analysis, but this is
complicated by the SiO, layer that must be present on all samples of Si;N,
as the protective barrier against total oxidation (2.11.2).
In the oxynitride model, it is suggested that the average tetrahedral
building unit has the composition SiN; .,O,.,; that « is always the precursor
to B-Si,;N, because of the SiO, layer inevitably present on Si used in the
preparation, and that the «— 8 change which occurs around 1300°C is
irreversible because SiO is evaporated from the system.
Whether the structural units are SIN; 9O,.;, or SiN,, in «-Si,N, they must
clearly be joined together in a way different to that in fB-Si,N,. In order to
form the a-structure; both models propose that the tetrahedra must be
distorted and tilted to the extent that none of the Si,N groups remain planar
but all are pyramidal. The «-structure can be described as two distorted
B-type layers rotated with respect to each other and joined in the c-axis .
direction. This arrangement produces a greater internal strain that always
makes the £-form thermodynamically more stable, but a higher entropy in
the «-form produces a kinetic factor which favours it in any silicon nitriding
reaction, and it remains in this form until a high enough temperature is
reached to enable the gross reconstructive change necessary to form the
f-structure. Thermal expansion of the «- and B-forms is very similar, despite
the distortion and strain.
That it is possible to have mixed non-metal tetrahedra in the Si-N system
is shown by the well-documented existence of the potentially useful ceramic
Si,N,O. In this orthorhombic material with a = 0.8873 nm, b = 0.5495 nm,
c = 0.4849 nm, the building unit is the SiN,O tetrahedron. These units are
linked to form a planar structure consisting of parallel sheets of Si and N
atoms in the (011) plane linked by Si-O-Si bridges along the a-direction.
 SPECIAL LIGHT CERAMICS 111
As expected from this structure, the thermal expansion along the b and c
directions is very close to « and $-Si,N, values, but there is a marked
anisotropy in the a-direction. In fact, because the Si-O-Si bridge is relatively
weak it is more easily deformed by increasing the oxygen displacement about
the Si-O-Si bond (that is decreasing the mean bond angle) rather than
stretching it. Such thermal anharmonicity means that the thermal expansion
in the a-direction is only about 4 that in the b and c directions.
This structure manifests several other obvious anisotropic properties.

2.11.4 Fabrication with silicon nitride

The nitride can be made comparatively easily by a number of routes.?’!?
381) 4:2N5 are Si, Ni,
3SiO, +6C + 2N, ——> Si,N,+6CO
SiCl, +6NH, —?°>Si(NH),+4NH,Cl
nSi(NH), 22 —- [Si(NH,)], 222 o-Si,N,,.
However, because it is a covalent solid it has only a small concentration of
vacancies and cannot be sintered to high densities just by the application of
heat. Several routes have been tried to obtain useful high-density material.
(a) Hot pressing. First the elements are allowed to react at 1400°C to form
a-Si,N, powder. The powdered Si,N, is mixed with densification aids such
as MgO, MnO, or Y,O; (the role of these aids is discussed separately below).
Then the mixture is heated to 1700-1800°C at 20 MNm ? pressure in a
graphite die. These conditions after 30 minutes produce f-Si;N, of high
density. The drawback is the high cost and the fact that only a few simple
shapes can be produced.
(b) Reaction bonding. This method has the advantage that complex shapes
can be made from the silicon powder compacted to high density. The shaped
article is then heated in N, to 1400°C, which turns the Si to a mixture of «-
and fB-Si,;N, but develops about 25% porosity. In the process the rate of
a-phase formation is independent of the rate of B-production, appearing to
grow linearly up to 20% of the product. After an initially slow growth rate,
the B-phase appearance accelerates after two hours to reach a maximum of
80% after 20 hours.
Addition of about 2% of metal fluorides to the silicon enhances both the
growth rate and the «/B ratio by a mechanism that involves devitrification
(3.1) of the silica coating on the silicon.
(c) Vapour deposition. This process uses volatile SiCl, and NH, gas and reacts
them over a heated substrate. High densities can be achieved, but only thin
layers of an amorphous product result.
(d) Warm moulding. This has been developed to produce a wide variety of
shapes using compression moulding machinery from the plastics industry.
Silicon powder is mixed with 15-25 % of a thermoplastic such as polystyrene,
1D CERAMIC SCIENCE

1000
Hot pressed
Modulus
of
rupture
GNm2

Reaction bonded

1500 2500 3500

Density kgm

Figure 2.18 Ranges of observed strengths for Si,N, as a function of density.

granulated, and the granules subsequently compression-moulded to shape.

The shaped article is air-heated to remove the plastic before sintering at
1400°C in nitrogen. This is a good production route, but unfortunately again
gives a high porosity product.
(e) Foaming. Here a deliberate decision is made to encourage porosity in
order to manufacture a material suitable for insulation, disregarding the
mechanical properties. Silicon powder is used as one component of a rigid
polyurethane system. When the second component is added, a gas is evolved
which expands the foam and sets it. The mass is debonded in air and heated
in nitrogen to 1350°C. Porosities up to 90% are achieved this way.
With the exception of hot pressed articles, few artefacts exhibit the real
mechanical advantages of Si,N,, and there is a wide range of properties (Fig.
2.18).

2.11.5 Densification aids in Si,N, sintering

Below 1700°C, unless additives are used, densification of silicon nitride is .
negligible. Above 1700°C the outer surface silica might be considered as an
additive as it becomes vitreous. However, it reacts with Si,N, to form an
oxynitride before its viscosity can become low enough to assist in the
sintering process. Herein lies the main function of any additive, namely to
assist the silica in providing a liquid phase of sufficiently low viscosity to
further the process of liquid-phase sintering. Densification occurs by three
stages:
(i) particle re-arrangement;
(ii) solution precipitation (this involves solution of a-Si;N, and precipitation of the B-form);
and
(iii) coalescence.

To enable a sample to reach complete densification by these stages, three

requirements must be met: (a) an appreciable amount of liquid; (b) good
 SPECIAL LIGHT CERAMICS 113

solubility of the solid and (c) complete wetting (3.5) of the «-Si,;N,. Additives
can be assessed in terms of these requirements.
(a) Magnesia. MgO was the first widely-used additive. Reaction with an outer
SiO, layer of the Si,N, produces a silicate liquid near to the MgSiO,-SiO,
eutectic composition. Further reaction ensues between the eutectic liquid and
Si;N, to produce an Mg-Si-O-N glass which wets Si,;N, and dissolves
a-Si,N,. Thus all three conditions are met. Precipitation of B-Si,N, produces
densification. Other impurities, in particular alkali and alkaline earth oxides,
lower the viscosity still further, so aiding the process.
(b) Yttria. Reaction between SiO,, Y,O,; and Si,N, produces a liquid phase
able to bring about densification with « > £ transformation. This additive
improves strength because the sintering liquid reacts further with Si,N, to
produce a refractory grain boundary phase,”’ instead of glass as in the MgO
case, which means that viscous shear in the grain boundaries is a much less
serious problem.
The advantage over using MgO is shown in Fig. 2.19; it is gained at the
expense of another desirable property, oxidation resistance. When some
compositions of high-strength, creep-resistant Si,N, hot-pressed with Y,O,
are heated to 1000—1200°C in oxidizing environments, they fail catastrophi-
cally. Extensive cracking is caused by easy oxidation of the grain boundary
phase to give cristobalite (1.5.4) and Y,Si,O, (1.5.1) with such a volume
change that no protective layers can remain intact. The drastic oxidation
behaviour does not occur in all samples; only those with overall composition
outside the compatibility triangle Si, N,—Si, N,O—Y,Si,O, are susceptible.
(c) Zirconia. Zirconia (5.5) forms a liquid phase, capable of dissolving
a-Si,N,, at the silica-rich grain boundaries of Siz;N, when 5wt % ZrO, is
used as a densification aid. Final products with a composition within the
compatibility triangle, Si, N,-Si, N,O-—ZrO,, contain little residual glass and
therefore have a good oxidation resistance enabling them to maintain high
strength at temperature.

1400

1000 A
Flexural
strength

MN m2

200

200 1000 1400

3 :
Temperature C

Figure 2.19 Strength v. temperature for (a) SisN, hot pressed with Y,O3; (b) SizN, hot pressed
with MgO.
114 CERAMIC SCIENCE

From the cases so far considered it is possible to generalize and say that all
additives produce additional crystalline or vitreous phases in the product.
Inevitably these phases impair the strength, creep resistance, or oxidation
resistance of that product. Uncertainties, such as the amount of silica
impurity in the starting Si,N, powder, make it difficult to achieve consistent
properties, since this variable changes the composition of the impurity phase
in the final product. Thus we have returned to the ultimate problem of
chapter 1: we have materials that are mixtures in which texture and
microstructure dominate the intrinsic properties of the major phase.
An attempt to overcome this problem by Jack and his colleagues stemmed
from their belief that «-Si,N, was a non-stoichiometric compound, and led
in 1972 to a major discovery of a new class of materials with considerable and
fascinating potential. The basic reasoning was that if «-Si,N, contained
intrinsic defects which are necessary to aid pressureless sintering, then Si,N,
could perhaps be induced to contain more vacancies by replacement of N by
O, with simultaneous charge compensation with (say) Al?*, in the same way
as in the aluminosilicates (1.5.1). Regardless of whether «-Si,N, is an
oxynitride or a true polymorph of Si,N,, the concept of a non-stoichiometric
covalent phase in the Si,N, system has led to a useful development, the
sialons.

2.12 Sialons
By analogy with the silicates (1.5.1), it was thought possible to replace N?~
in Si,N, by O77 if at the same time Si** was replaced by Al>*, giving rise to
the (Si, Al)(O, N), tetrahedron as a structural unit.1! This, it was felt, might
lead to new materials built up in the same way as mineral silicates are built
up (1.5), but this time from (Si,Al)(ON), tetrahedra. Such predictions did not
include the surprising finding that up to 70 wt % Al,O, could be dissolved in
Si;N, by hot pressing at 1700—2000°C, without significant alteration to the
X-ray diffraction pattern of B-Si,N,.
Because the f-structure was little changed, only scheme (A) below was
considered in the early work to account for the dissolution of the alumina.

(A) B (=Sig** Ng*~)

| upto 70% Al,O,

B (=Sig 0735 Alove7x02"" Nt

| theoretically for 100 % replacement

Al5%305°—
Al5°308° is equivalent to a very defective alumina (5.4) with the
B-Si;N, hexagonal structure. With such defective f’-phases, it was felt, it
should be possible to sinter to high density without pressure. Since only a
single X-ray phase was identified after additions of up to 70 % of Al,O,, it was
concluded that this scheme applied.
 SPECIAL LIGHT CERAMICS iS

However, if volatilization occurs, or if a glassy phase not detectable by

X-rays is present, then an alternative scheme (B) exists:

(By OB = Sig’ Nat”

{
Br = Si2454)*
Ay OP-NZ4=39-
| by loss of SiO,N,,CO or glassy SiO,
Ali8+Q12-N&~- = 2(Al,O,AIN).
For example, for y = 4, the reaction between Si,N, and Al,O, at 1750°C
could be represented by the following:
4Si,N,+6Al,0, = 3Si1,Al,0,N,+ 6SiO + 2N,
Si,N,+2Al,0;+C = Si,Al,0,N,+Si0+CO
(C derived from hot pressing crucible)
SizN,+2Al1,0; = $1,A1,0,N, +Si0, (glass)
Doubts as to whether f’ is a widely non-stoichiometric phase were aroused
when the unexpectedly bad high-temperature creep properties of the “Sialons”
were observed, that is except when the (Al+Si)/(N +O) ratio was exactly 2.
Further examination showed scheme (B) to be correct, with little range of
existence away from the Si,N,—A1,03N tie line for these new materials.

2.12.1 Representation of the Si-Al-O—N system

There are four components in the Si-AlI-O-—N system and therefore a
tetrahedron is needed to represent it (Fig. 2.20). Within the tetrahedron, as
long as all the elements have their normal valencies, Si(IV), Al(III), O(II) and
N(IID), all the solid compounds lie on the irregular quadrilateral plane shown
in Fig. 2.20. These elements, as seen in the discussions on bonding and
valence in (2.1 and 5.3), would not be expected to show anything but such
fixed valencies, and as a result the irregularity of the plane can be removed by
expressing concentrations in equivalents, that is by making the corners of the

Figure 2.20 Representation of the Si-Al-O-N system.

116 CERAMIC SCIENCE

EQ.%0

Si,N,0 wer

$).
0 Equivalent %AI 100

Figure 2.21 The equivalent representation of the Si-Al-O-N system.

tetrahedron 3Si!7*, 4Al!?~, 60!2~ and 4N??~, whereby all the solid com-
pounds lie on a square plane (Fig. 2.21).
By convention, Si,N, is placed at the bottom left corner, and any point
within the square represents a combination of 12 positive and 12 negative
valencies. Going from left to right represents the replacement of silicon by its
equivalent of aluminium, and moving from bottom to top represents re-
placement of nitrogen by its equivalent of oxygen. The simple representation
in Fig. 2.21 is obtained by expressing the concentrations in equivalents, so
that the concentration of any mixture can be expressed by the two quantities
cee [o]2
Al and, “7.0'=
~ [Si]4+[Al]3 [N]3+[0]2’
The following two examples show how these are used.
(i) What composition does the centre of the square represent?

eaee
Ape [Si] | l,
a ae4[Si] == 3[Al] and so Si/A? l ==3¢== Si,A
a=
/o 5"
2 TsA J4+[ANS
4aT Al]3

1 2[O
L ] ..2[0] = 3[N] and O/N =3 = O,N,.
OY ===
2 3IN\ 210)
So far then the answer is Si,Al,(O3N,),., and to obtain x the charges must be
made to balance. There are 24 positive charges, and so x must 2 and the
composition is Si,Al,O,N,.
(11) Where does Si,N,O appear on the diagram?
V7Ale,
 SPECIAL LIGHT CERAMICS 117

Hence the oxynitride appears one-quarter the way up the side joining Si,N,
to SiO,. Figure 2.22 is a diagram obtained by Jack and co-workers in this
equivalent form, representing the behaviour of the Si-Al-O-N system at
1750°C.

2.12.2 Si-Al—O-N phase relationships

In this system the principal phases are shown in Fig. 2.22. It is believed
that the bond length similarities and differences Si-N (0.175nm) = Al-O
(0.175nm), and AI-N (0.187nm) # Si-O (0.162nm), partly control the homo-
geneity ranges. Thus, extension along the direction of constant metal :non-
metal. atom ratio occurs when an Si-—N is replaced by an Al-—O, but
perpendicular to this, when Al-N would be replaced by Si—O, there is little
evidence for homogeneity.
(a) The f’ phase. This extends essentially along the 3M :4X line with a range
Si,_,Al,O,N,_, where z reaches ~4 at 1700°C. It is the only sialon
examined in detail, but because it usually contains vitreous or crystalline
phases, for example X, or 15R in Fig. 2.22, the intrinsic properties have not
been definitely determined. Other measurements have been on compositions
with z = 3, ie. Si,Al,0,N;. Because of its structure, its physical and
mechanical properties are similar to those of B-Si,N,, but chemically it is
closer to Al,O, in behaviour. One advantage of sialon over Si,N, is that
conventional ceramic techniques such as extrusion, pressing, and slip casting
can be used to produce shapes of the mixed components, and these can be
sintered to near-theoretical density in most atmospheres at 1600°C. Oxidation
resistance is better than Si,N,, probably because a coherent surface layer of
mullite (1.5.1) is formed.

Y(3A1,0,A1N)

Hy 7A(AIN.
Al,03)

Figure 2.22 The sialon system at 1750°C.

118 CERAMIC SCIENCE .

(b) The O’-phase. This has the structure of silicon oxynitride, Si, NO (2.11.3),
and extends to a limited extent along the 2M :3X line.
(c) The X-phase. The designation of X arose because it was some time before
its X-ray diffraction pattern was indexed. A suggested composition is
Si,Al,0,N. From the most recent X-ray work, which identifies the unit cell
as monoclinic, there is a suggestion that X may be a nitrogen-mullite.
(d) H and Rsialons. These occur near to the AIN corner of the Si-Al-O-N
diagram. Each phase extends along a direction of constant M:X ratio in
which Si+N is gradually replaced by Al+O. Rationalization of their
structures and compositions is achieved by considering them to be based on
the AIN structure. AIN has, like «-SiC (2.10), a wurtzite-type hexagonal
structure with a = 0.3114nm, c = 0.4986 nm. In 2.10 it is demonstrated how
a number of SiC polymorphs can exist by periodic interweaving of faults in
the layer stacking sequence; the same reasoning applies here, except that the
layer sequence is determined by the composition, and not by chance, that is
by the (Al+ Si)/(N + O) ratio.
The polytypes reported to date have the compositions (Al + Si),,(N + O)m +15
where m is an integer in the series 4—9. Lattice parameters and layer spacings
are given in Table 2.7. Most of the increase in c-repeat distance is due to AlO,
octahedra inserted half-way between each M(O+N) layer. The octahedral
layer introduces an ABC .. . layer sequence into the otherwise ABAB...
metal stacking of a wurtzite structure. An even number of metal layers
between each ABC ... unit gives a rhombohedral unit cell whereas an
odd number results in an hexagonal structure, hence the R and H symbols in
the Ramsdell notation (5.10). Cubic close packing sequences beyond one unit
of ABC are not observed.
Here again it should be emphasized that although we have mentioned
layers, the sialons are not layer-structure compounds like BN or graphite. We
are using the terminology loosely as in the «-SiC polytype case (2.10.1), in
that puckered layers of (Al,Si)(O,N), tetrahedra can be visualized. The
structure and bonding throughout is three-dimensional, making these sialon
ceramics potentially very useful. We are essentially saying that there are

Table 2.7 AIN-type sialons

Atom ratio Cell dimensions Layer spacing

(Al+Si)/ Ramsdell a c c/n
(N+0O) symbol (nm) (nm) (nm)

$ 8H 0.2988 2.302 0.288

2 15R 0.3010 4.181 0.279
$ 12H 0.3029 3.291 0.274
4 21R 0.3048 5.719 0.272
a 27R 0.3059 ~—s_7.198 0.267
oH? 0.3079 0.53 0.265
AIN 2H 0.3114 0.4986 0.249
 SPECIAL LIGHT CERAMICS 119

>
@o0

2H 3R
Figure 2.23 Puckered layer sequence and Ramsdell symbols for (a) AIN and (b) AIN-type sialon.

discernible MX layers stacked in an ABAB .. . sequence as in wurtzite, but

each “layer” is covalently bonded to those above and below through Si or Al
atoms with sp* hybridized orbitals. The number of puckered layers, or
double-layers as they are denoted in some descriptions, along the c-direction
gives the Ramsdell numeral 2, 3, etc. A 3R arrangement is shown in Fig. 2.23
which can be compared to the “layer” arrangement in isostructural p-SiC
(Fig. 2.15).
Each block of m layers, where m is determined by the composition,
contains an extra non-metal atom, so converting one MX layer to an MX,
layer. This forces a change in stacking sequence because a continuation of
ABAB. ... would cause two tetrahedra to share a common base and we have
seen already (1.5) how SiO, and AIO, tetrahedra cannot do this readily
because of M”* repulsion terms (5.3.2). Thus the ABC . . . sequence arises.
When the composition of these sialons gets close to AIN, the concentration
of M(O+N), layers becomes so small that an ordered arrangement does not
occur but only an expanded, disordered 2H? structure. The orderliness of the

Figure 2.24 (001) lattice fringes of 15 R Si-Al-O-N. By courtesy of Dr A. Hendry.

E
120 CERAMIC SCIENCE -

MX, stacking fault sequence is demonstrated by the pattern of (001) lattice

fringes obtained for a 15R sialon examined in the electron microscope (Fig.
2.24).
None of this group of sialons have been examined as thoroughly as
B’-sialons, but they do offer potentially more suitable materials for high-
strength, high-temperature applications. More interestingly, the region f’ +
15R in Fig. 2.22 may offer good prospects, because many of the AIN-polytype
sialons such as 15R have a needle-like habit, and when embedded in a f’-
matrix, free from glassy phases from the SiO,-rich regions, could provide a
microstructure capable of realizing the full potential of these materials.

2.12.3 Sialon preparation

Because of the potential large-scale technological use of these materials,
several routes to their preparation have been described.
(i) Simple heating or hot pressing of mixtures of SizN,, Al,O3, AIN, SiO,
and Si,N,O in N, have been most frequently used. In this case, in contrast
to the Si,N, case, high densities have been achieved without high pressure.
(ii) Kaolinite with ammonia—this represents attempts to make the products
commercially more attractive. First the kaolinite (1.5.3) is heated to 500°C to
produce metakaolin (1.5.7.). The temperature is then raised to 1400°C in the
presence of NH;. The overall reaction is summarized by the equation
4(14—3x) NH, > 8(3—x)AIN
2A1,0, °4Si0, +
6-x
+o
ag r
Ste-nAlO.Nis-2) +
(14—3x)
6(14-—3
20.
6—x
If the pure material is required, the AIN is removed by solution in warm
dilute NaOH. This may be the route to f’ + 15R products.
(iii) Volcanic ash heated with aluminium metal and N, has been shown
to produce sialons.
(iv) White sialons are made when co-precipitated SiO, + Al,O; gels are
heated to 1700°C in NH3.

2.12.4 Metal-sialons
Consideration of the work to date on sialons leads to two simple principles
that can be applied in order to engineer new materials in this field of ceramics.
1 First, it is clear that a family of sialons exist in which the fundamental
building unit is the AIN, tetrahedron; these are the AIN-type polytypes
15R, etc. Replacement of the AIN, tetrahedra is possible by SiN, and
therefore binary phases known to contain MN, tetrahedra may also be
used. The compounds MgSiN,, Mg3N,, LiSi,N3, for example, are
known to have wurtzite structures, to contain MN, tetrahedra, and
therefore to be capable of becoming network formers that could perhaps
form sialons.
 SPECIAL LIGHT CERAMICS 1A

34Mg,Si0,

Si3Nq4 4,A1303N

(b)

Figure 2.25 (a) A section of the Mg—Si-O-N diagram. (b) The 3M:4(O0+N) plane of
Mg-Si-Al-O-N system.

2 A second principle is to extend the charge compensation (needed when

nitrogen is replaced by oxygen in B-Si,N,) to ions other than Al°*.
Application of both 1 and 2 has rapidly led to many new and potentially
useful ceramics.
(a) Magnesia-sialons. Si;N, reacts with spinel (5.9) to give f’-sialons and R
and H polytypes containing magnesium. Mixtures including a magnesium-
containing phase with the «-Si,N, structure have been produced. This latter
phase has a range of stoichiometry and is labelled «’. A nitrogen-containing
spinel of wide homogeneity range is also reported.
In the Mg—Si—O-N region of the system there occurs an extensive liquid
region above 1500°C which is responsible for the success of MgO as a
densification aid in hot-pressing Si,N,. This liquid cools to form a glass.
f’-sialons containing magnesium have a_ general composition
Mo. 5nSi¢ —x+n/2OxNg -x- Two representative sections of the Mg-Si-Al-O-N
system are shown in Fig. 2.25.
(b) Yttrium sialons. In the case of Y,03, only the second of the above
principles is applied. This is because Y** does not have tetrahedral co-
ordination and cannot become a network cation (3.3) but can only act as a
modifier cation (3.2). However, as a modifier and charge balance cation it
gives rise to several new types of oxynitride. Four quaternary phases have
been identified.
(i) A nitrogen-containing melilite, Y,03-Si;N,, which in structural terms is Y,Si(Si,03N,)
whereby we can see the comparison with akermanite, Ca,Mg(Si,O,), a discrete ionic
silicate (1.5.1). N-melilite contains sheets of linked Si(ON), tetrahedra stacked on top of
each other and held together by Y°* cations. Being isostructural with akermanite and
gehlenite (1.5.1), it forms a complete series of high-melting solid solutions with them. In this
manner, impurities containing Ca, Mg and Al that might be present in Si,N, and could
form low melting glasses are “gettered”, to preserve a desirable high temperature strength.
(ii) A nitrogen apatite Y;(SiO,),N with a structure similar to the naturally-occurring apatite
Ca;(PO,)3;0OH. i
(iii) A phase Y,Si,0,N, which is isostructural with yttrium aluminate Y,Al,Ob, with which it
forms a complete range of solid solutions.
(iv) A nitrogen wollastonite, YSiO,N (1.5.2).
122 CERAMIC SCIENCE .

This example serves to show the great potential available in this new area for
producing nitrogen analogues of the known silicate phases. Y,O, is such a
useful densifying agent in hot pressing because of the availability of these
quaternary phases to scavenge impurities in Si,N,, together with the existence
of liquid phases formed in the SiO ,-rich regions present at the grain surfaces
of Si;Nyq.
(c) Beryllium-sialons. Beryllium has tetrahedral coordination possibilities,
and like magnesia, would be expected to act according to principles 1 and 2
above. There is another factor in this case because Be,SiO,, known as
phenacite, has the same structure as B-Si,;N, and as a result is much more
soluble in B-Si,N, or f’-sialons than forsterite, Mg,SiO,. The consequence of
this appears to be that the homogeneity range of the f’-phase is greatly
extended, bringing with it possible technological advantages, even though
beryllium is a hazardous element.
(d) Lithium-sialons. The ternary nitride LiSi,H contains nitrogen tetrahedra
and therefore principle 1 might be applied. Furthermore, LiSi,N; is iso-
structural with Si,N,O and forms a range of solid solutions with it. Reaction
of the spinel LiAl,O, (5.9) with B-Si,N, gives a f’-lithium sialon. Reaction
of LiAlO, with B-Si,N, interestingly leads to a sialon with the a-Si,N,
structure known as «’. The polytype 15R sialon occurs in the system, along
with spinel and cristobalite-based lithium sialons. This is another example of
the rich diversity of new materials that can be produced.
A third principle emerges at this stage.
3 Spinels will react with Si,N, to give f’-phases, since one component of
the spinel exists in tetrahedral co-ordination (5.9). Thus zinc, copper,
iron II, etc., will produce sialons when introduced via the appropriate
spinel.

2.12.5 Nitrogen glasses

These are nitrogen-containing glasses formed when Si,N, is hot pressed with
magnesia as a densifying aid. The glass so formed devitrifies on heat
treatment to produce enstatite, MgSiO, (1.5.2), and silicon oxynitride, Si, N,O.
It appears that this glass can only be maintained in the glassy form (3.1)
above 1500°C. At 1700°C, 20wt % MgO+Si0O, > MgSiO;, but when
10 wt % Si,N, is added to the mix the result is a completely vitreous product
with a probable composition Mg,,Si,,0;;N,, which devitrifies (3.1) at
1500°C to give a mixture of cristobalite (1.5.4), enstatite and Si,N,O.
Additions of other oxides extend the glassy region. Al,O; gives a glass
containing 10at.%N and Y,O, extends this to 12at. %N. To date these
glasses have not been stabilized and characterized.
It is known that small additions of nitrogen to oxide glasses increase T,
(3.1.1) and so glasses with 10 at. % N when “stabilized” might have unusual
properties or they might become the base for new highly refractory glass
ceramics (chapter 4).
 SPECIAL LIGHT CERAMICS 123

2.13 Realization of high strength and the development of toughness through

fibre-composites!*
Early sections of this chapter were concerned with establishing criteria for
high strength in ceramics and in identifying high-strength ceramics. To date
the overriding objective has been to make stronger and stronger ceramics in
the sense that higher fracture-initiation stresses are required. However, once
the fracture is initiated, there is still minimal resistance to crack propagation
in these super-ceramics and spectacular catastrophic failure occurs. Some
thought must now be given to developing tough ceramics from these strong
materials.
Here we are considering toughness as a material property which gives a
measure of resistance to crack propagation when that material is subjected to
mechanical or thermal stress.’ One approach to developing this property is
to consider the energy demand curve for crack propagation in conjunction
with the energy release curve; this statement suggests strongly a Griffith
approach (3.4.2).
Griffith’s energy balance implies that the energy demand curve will be a
straight line:
U, = 4cy G31)
where c = crack length, y is the surface energy per unit area, and U, is the
energy required to increase the crack length.
Equation 2.30 shows that the energy release curve will be a parabola
U, = (nc?o2)/E, where U, is the energy release, o, is the applied stress, and
E is Young’s modulus.
Catastrophic crack propagation occurs when the slopes of the two curves
are equal, in this case when

er =enn 232)
Oc 0c

We can differentiate 2.30 to obtain

210 ,C
E <= 4y,y (2.33)
DBS

Since in a pure homogeneous material, on this model, the energy demand

is constant but the slope of the parabola increases, then once the crack is
initiated, equation (2.32) must be fulfilled and instantaneous failure results.
To develop a tough ceramic where stable crack propagation occurs, it follows
that the energy demand curve must be made non-linear and become concave-
upwards so that the gradient of the parabolic curve does not “catch-up” with
the demand curve, or at least higher applied stress levels will be needed to
make it do so. This situation is impossible for single-phase homogeneous
brittle materials however strong they may be, but can be achieved with
multiphase brittle ceramics in the following way.
124 CERAMIC SCIENCE

In a multiphase ceramic, a crack in the weaker component will propagate,

but it may encounter the stronger component and then either cut across this
phase or detour around it. Either course means that the energy demand will
increase as a step function, and crack propagation will stop, unless the
applied stress is appropriately increased. A whole series of such encounters
produces aseries of steps in energy demand and the linear demand curve is
changed, becoming in the limit curved. If the second phase is stronger than
the matrix phase, then the energy demand curve shape is the desired
concave-up, but the reverse is true for a weak second phase in a strong
matrix.
Thus in a multiphase ceramic, an initial crack of length cy will begin to
grow under astress Go, but it will be stopped at length c, when the slope of
the energy demand curve at c, exceeds the slope of the energy release curve.
For the crack to grow further, the stress has to be raised to o,, when the crack
will extend from c, to c, at which point the demand slope curve again exceeds
the slope of the release curve. Repeated application of this step mechanism,
by which the crack proceeds from instability (moving), to stability (static),
eventually produces a situation where the energy release curve cuts the
energy demand curve as shown in Fig. 2.26. At the point of intersection, the
slopes are equal and the crack will grow spontaneously and catastrophically.
It is the value of the rate of energy absorption at the point of crack instability,
i.e. at the point of intersection of the two curves, that is important when
considering the toughness of a multiphase ceramic. This value is designated
as y,, the critical strain energy release rate, and has the value of the maximum
slope attained by the energy demand curve.
y, can be regarded as a material constant which depends on an interplay of
factors:
(i) the type of second phase (tougher or less tough than the matrix);
(ii) the size of second phase particles (too small particles become “invisible” to the crack, too
large ones may depend too much on the interfacial bonds); and
(iii) volume (increase here results in an increase in y, until it becomes large enough to produce
microscopic defects).

Energy demand curve

when a_ stronger second
phase is in the matrix

Energy U

Gy Cc, Cy crack length C

Figure 2.26 Energy demand and energy release curves in a multiphase ceramic.
 SPECIAL LIGHT CERAMICS 125

It is essential in this discussion to realize how important the interphase

bonding is: it must be very strong in order to gain the toughness without
losing the high strength. Weak interfacial bonding will increase the energy
demand, i.e. toughness by the definition we are using, as the crack detours,
but the crack is easier to initiate and ultimate strength is therefore reduced.
The inescapable conclusion of this chapter is that the super-strength elements
and compounds so far discussed will usually provide attractive engineering
materials if they are dispersed as a second phase in a matrix material, as long
as the inter-phase bonding has the same order of strength as that present in
the strong phase.
Table 2.1 suggests that the best form of the strong materials will be either
thin fibres or short whisker crystals. It is obvious to see that such a form will
restrict the effect of the strength-reducing flaws. Fine fibres have a small
surface, and if this is protected during manufacture then the probability of
existence of a strength-damaging flaw becomes very small—this argument is
developed in 3.4. Whisker crystals have no dislocations and so have high
strength, and furthermore, this type of strength degrader is absent if the
strong material has no crystal structure, i.e. it is a glass.
Thus the development of stiff, strong and tough engineering ceramics
probably lies in developing fibre composites containing the materials de-
scribed in this chapter. Some considerations applicable to such a development
now follow.

2.13.1 Composite elastic properties

Composite geometry determines the form of the relationship.'’ Three extreme
cases can be considered:
(a) long thin fibres aligned along the direction of applied tensile stress;
(b) fibres aligned at 90° to the applied stress;
(c) short pieces of fibre, or small whisker crystals, in random orientation to the direction of
applied tensile stress.

For (a), considering the sample to have unit cross-sectional area and
assuming that the matrix-fibre bonds are the strongest present, then from the
rule of mixtures we can write:

6, =for+(1—f)o, (2.34)

where f is the volume fraction of the fibre, and o,, o, and o,, are the tensile
stresses in the composite, fibre and matrix respectively.
Because fibre matrix bonds are the strongest, and assuming no cracking,
then the strain in the fibres will equal that in the matrix. For the region where
Hooke’s Law applies (2.1), equation (2.34) becomes

ES =fE-é¢+(1—f)E nem :
or E, =fBy +O =f Ep (2.35)
126 CERAMIC SCIENCE |
Equation (2.35) describes a straight line, and calculation of composite
stiffness is simple.
For model (b), the stress in both components will be equal but the strains
may be different. The rule of mixtures tells how the total strain is shared.
Applying Hooke’s Law again,

Ce 6,
ey
E, ale De
but in this mode o, = o, = o,, and
aed E Em
(2.36)
een
This is a parabolic equation, from which it can be noted that in this
orientation more stiff fibre is required to obtain an equal degree of stiffness to
mode (a).
Model (c), random short fibres, does not easily yield an expression from
first principles,* but since the composite modulus is likely to be somewhere
between the extremes of cases (a) and (b) it is easily quantified by taking the
simplest expression, (2.35), and introducing a factor to give practical results:
E, =fE;n+(1—f)Em (2.37)
n is the efficiency of reinforcement, which is found by experiment to have the
form cos*@ where ¢ is the mean angle made by the short fibres to the
direction of applied stress.

2.13.2 Composite strength

Considering first model (a), we shall now describe some general cases.
(1) Brittle fibres in a brittle matrix with the strain at failure of the fibre greater
than the failure strain in the matrix. Glass fibre in a resin matrix is an example.
Here a simple law of mixtures applies, and
C= fo.+(l—7)o,- (2.38)

fibre

stress composite
Ss secondar
modulus .
--- matrix
primary
modulus

strain “f strain
a b

Figure 2.27 Stress-strain curves for (a) brittle fibres in brittle matrix; (b) brittle fibres in ductile
matrix.
 SPECIAL LIGHT CERAMICS a

Figure 2.28 Composite stress for a brittle ceramic fibre in a ductile matrix as a function of fibre
volume.

But since the fibres still remain intact when the matrix fractures, it is
possible to distinguish two elastic moduli, as shown in Fig. 2.27a.
(ii) Brittle fibres in a ductile matrix showing some strain hardening. Boron in
aluminium is an example’®—Fig. 2.27b shows all the relevant information.
Consider two situations inherent in Fig. 2.27b. First, when the composite is
strained to ¢, and the fibres remain intact, the stress will be given by the rule
of mixtures, equation (2.38). Second, at the same strain when all the fibres
have failed and the load is taken by the matrix, but still using the total
composite cross-section, the stress will be given by:

On = (1 =) vag (2.39)

Both (2.38) and (2.39) represent straight lines and can be plotted as Fig. 2.28.
In this figure BA represents (2.38) and CE (2.39).
Some interesting features emerge from this example: if f</f, (at E) the
composite strength is determined by the matrix and the fibres, however
strong, contribute nothing. If f> f, (at E) then the strength is controlled by
the ceramic fibres. It is noteworthy that only when f > f, does the composite
strength exceed the strength obtainable in the matrix alone. Thus there is a
critical volume fraction of fibre, in contrast to the stiffness case, where as soon
as a fibre is added to the matrix, E increases. A very little fibre can be danger-
ous with respect to strength. It is therefore important to be able to define the
critical volume fraction, for reasons of safety and cost, since fibres are
expensive, and it would be desirable to have f, as small as possible. This can
be achieved by considering point F on Fig. 2.28. At F, 0, = fio,+ (1 —f.)Om
which is Ong = Ne Ion
and so
Ce 6:
if ae
ae IO 2.40
(2.40)

Equation (2.43) shows what must be done to minimize the critical volume
fraction: first ,,,—@, must be as small as possible and so the matrix must
128 CERAMIC SCIENCE

be chosen to have poor strain-hardening characteristics which might be

achieved with suitable alloying additions; then o,,—o,, must be as large as
possible, which implies using the strongest and probably the most expensive,
thinnest fibres available.
Returning to the original general model (b), there are no expressions used
to calculate strength in this case; in this model, delamination occurs in
practice since it is not possible to ensure 100% efficient fibre-to-matrix
bonding despite our working assumption.
General case (c) is acommon one when chopped strand fibre is used because
of its lower cost. In order to derive an expression for composite strength
in this case, further simplifications have to be made by assuming all the
fibres are parallel to the applied stress and are cylindrical, of diameter d and
length I.
In this mode the first question to be considered is how the fibre receives a
share of the applied stress. The answer follows from our assumption of very
strong matrix—fibre bonds which do not therefore break but become sheared
as the matrix is strained by the applied stress. Thus the surface of each fibre
is subjected to a shear force.
Total shear force transmitted to the fibre is the surface area times the
maximum interfacial shear stress, zdxt,,, where t,, is the maximum interfacial
shear stress, and x is any distance along the fibre measured from the fibre end.
Therefore, the shear force applied to the fibre is
dz
Op area = o,nr? = o,T ae

Hence
@
RANT = Oye
and
4XT
of = 7 (2.41)

Equation (2.41) contains two interesting features. First, the thinner the fibres
the more of the applied stress they bear, secondly, the stress varies along the
fibre length increasing from zero to a maximum at //2. The implication is that
if the fibre is going to reach its maximum stress loading, o,,,, then there will
be a minimum length that is acceptable and it may not be worthwhile buying
the most expensive fibre if it is very short.
The critical length can be found by putting stress in (2.41) equal to o,,:
do py,
l= x= ote
(2.42)
Note that the critical length depends on the matrix which the fibres are
strengthening, and is therefore not an intrinsic property of the material. It
 SPECIAL LIGHT CERAMICS 129

also depends on the fibre thickness and strength, but since these are not
independent, thicker fibres having in general a lower o,,, they can only be
varied to obtain an optimum value. .
It is usual to rearrange (2.42) to obtain the fibre aspect ratio.
l : Cry
ae aspect ratio = 2. (2.43)

The aspect ratio of the fibre must be exceeded if the fibres are to exhibit their
ultimate strength. If 1/d < o,,/2t,, then the matrix will break without breaking
the fibres, and the desirable property of ultra-high strength for which these
special materials have been developed will not be realized.
Now, returning to the task of deriving a composite strength equation, since
each fibre has a stress dependent on its length we must adopt an expression
for the average stress in a fibre by first assuming / >/,. From this last
assumption it follows that at I/2, 0, = o,,, and we can write
s: I,
Of == on(1-2).

Substitution of this last expression into a rule of mixtures equation gives

l
=f O4(1-5)+ (1
Pm (2.44)

This predicts that discontinuous fibres give less strengthening than long
continuous fibres.

2.13.3 Specific strength and specific stiffness

The expressions for composite modulus and composite strength derived in

10
@ carbon whisker
Specific
strength Al,O, whisker» @ SiC whisker

m? 5? Aln03// @ Si,N, whisker

~—. SS e@ Carbon fibre
Glass plastic/pjasti Ss eboron fibre
=

\§ QS oles
Ry
: N N fibre/piasti
RN “Boron epoxy
JS \Boronp,
AIO, in Ni
23 \ Boronjy

Wood
0-1
10 100 1000

Specific stiffness hae

ms

Figure 2.29 Specific strengths and stiffnesses of some fibre-reinforced composites.

130 CERAMIC SCIENCE .

the previous sections are commonly presented in a form which facilitates

comparisons: specific strength and specific stiffness are quoted. The specific
strength is o,/p and specific stiffness is E,/p where p is the composite density.
This is a useful form since it emphasizes the strength-to-weight ratio. Some
typical values for these properties are shown in Fig. 2.29.

2.13.4 Composite toughness

At the beginning of section 2.13, the requirements for improving ceramic
toughness were discussed, and the conclusion was reached that formation of
a multiphase material would improve toughness as long as ideal interphase
bonding was achieved. This must apply to model (a), for a crack generated in
the matrix must either detour round the strong fibres (this is called pull-out)
or it must break them. Either case, or a combination of both, must increase
the energy demand for crack propagation and produce the “concave-up”
curve that is needed to improve toughness.
If there are extremely strong bonds between the fibre and the matrix, it is
probable that the major contribution to toughness will be the pull-out work.
Following the reasoning given in 2.13.2 for the strength contribution arising
from short fibres in the matrix, it can be seen that pull-out work is guaranteed
in a chopped-strand composite by ensuring that the fibre length does not
exceed |.. If |. is not exceeded then the fibre UTS will not be reached at the
applied stress and any crack will have to detour and cause pull-out.
It is possible to express this pull-out contribution to toughness is the
following manner.
Work to pull out a length of fibre / =
l
|Rdtx dk =n dt,, (2) (2.45)
0

where the symbols have the meaning already defined in 2.13.1.

Number of fibres per unit area = N = a

Now assume that pull-out lengths are distributed uniformly betwéen 0 and
!,/2 then the total pull-out work is:
1/2
4f/nd one 2 dl
| tdt,,(17/2) figl?
1/2 i
| “ik 6d
0

and using equation (2.42) to replace l.:

d-o%,
Pull-out work =
24t,,
(2.46)
 SPECIAL LIGHT CERAMICS 131

This can be a large contribution to composite toughness, since o,, has

large values for the materials discussed in this chapter. It is worth noting that
thick fibres, where d is large, contribute more to toughness by this mechanism,
but (2.41) shows that thick fibres will have a lower strength. It follows from
this that a combination of long, very thin fibres and short, thick fibres with
| < 1,, made from the materials described in this chapter and embedded in a
matrix that produces extremely good matrix-fibre bonds, will produce
acceptable properties of high strength with good toughness in a light weight
structure. Perhaps this is the way that these modern ceramics will eventually
be developed.®

References and further reading

1. Archer, N. J. The preparation and properties of pyrolytic BN, in High Temperature
Chemistry of Inorganic and Ceramic Materials, eds. F. P. Glasser and P. E. Potter,
Chemical Society (1977), p. 167.
2. Campos-Loriz, D., S. P. Howlett, F. L. Riley and F. Yusaf. Fluoride accelerated nitridation
of silicon. J. Mater. Sci., 14, 2325 (1979).
3. Cappelen, H., K. H. Johansen and K. Motzfeldt. Oxidation of SiC in oxygen and water
vapour at 1500°C. Acta Chem. Scand., A35, 247 (1981).
4. Chang, D.C. and G. J. Weng. Elastic moduli of randomly oriented chopped-fibre composites
with filled resin. J. Mater. Sci., 14, 2183 (1979).
5. Gray, H. B. Chemical Bonds. W. A. Benjamin Inc. (1973).
6. Greenwood, N.N. Principles of Atomic Orbitals. Chemical Society, London (1973).
7. Gupta, T. K. A qualitative model for the development of tough ceramics. J. Mater. Sci., 9,
1585 (1974).
8. Harris, B. The promise and realities of fibre composites. Met. and Mater. Technol., 77,
(1981).
9. Hollenberg, W. and G. Walther. The elastic modulus and fracture of boron carbide. J. Am.
Ceram. Soc., 63, 610 (1980).
10. Jack, K. H. Nitrogen Ceramics. (17th Mellor memorial lecture). Institute of Ceramics (1973).
11. Jack, K. H. Sialons and related nitrogen ceramics. J. Mater. Sci., 11, 1135 (1976).
12. Mazdiyasni, K. S. and C. M. Cooke. Synthesis, characterisation and consolidation of Si;N,
obtained from ammonolysis of SiCl,. J. Am. Ceram. Soc., 56, 628 (1973).
13. Mocellin, A. Ceramics for high temperature applications. Rev. int. Htes. Temp. et. Refract.,
14, 46 (1977).
14. Piggott, M. R. Load Bearing Fibre Composites. Pergamon Press, Oxford (1980).
15. Ramsteiner, F. Elastic behaviour of unidirectional short fibre-reinforced thermoplastics.
Composites, 65, (1981).
16. Rich, G. and A. R. Bunsell. The behaviour of boron-aluminium composites during and
after impact. J. Mater. Sci., 14, 2373 (1979).
17. Ross, V. F. and J. O. Edwards. The Chemistry of Boron and its Compounds, ed. E. L.
Muetterties, Wiley, New York (1967).
18. Seto, J. Y. W. Deposition of polycrystalline silicon by pyrolysis of silane in argon. J.
Electrochem. Soc., 122, 701 (1975).
19. Slater, J.C. Quantum Theory of Matter (2nd edn.) McGraw-Hill, New York (1968).
20. Sykov, K. I. and E. I. Jelikhovskaya. On the formation and composition of pyrolytic
carbon. Carbon, 5, 201 (1967).
21. Tsuge, A., K. Nishida and M. Komatsu. The effect of crystallising the grainboundary glass
phase on the high temperature strength of hot-pressed Si,N, containing Y,O,. J. Am.
Ceram. Soc., 58, 323 (1975).
3 Glass

3.1 Introduction
There are two predominant types of chemical bonding in glass science,
covalent and ionic, and both play an equally important role in interpreting
behaviour. The main features of covalent bonding are discussed in 2.2, and
the ionic bond model is the subject of 5.3. Both these sections should be
referred to in order to gain more insight into this chapter.
Man-made glasses appeared around 4000 BC in Egypt and Mesopotamia
as decorative glazes, and glass working was known by around 1500 BC as an
art and a technology. Glass science was not further developed until the work
of Faraday, and later Zeiss, Abbé and Schott, who in 1881 began to develop
new optical glasses. By 1900 these workers had used some thirty-four
elements in experimental glass formulations. Although seventy elements have
now been tried, only three major commercial glass systems constitute 99 % by
weight of all glass production. These systems are:
(i) Soda (Na,O)}lime (CaO)-silica (SiO,)
(ii) Lead crystal glass, PhO-SiO,
(iii) Low expansion borosilicate glass B,0,;—-SiO,—-Na,O-CaO.

It is significant that only the borosilicate system was developed this century;
illustrating the development potential in the soda—lime-silica system, and the
inertia to major change present in such a massive industry.
There is also the possibility that restrictive definitions and dominant views
on glass structure have further delayed modern exploitation of new areas in
what is now scientifically a large and growing field. The changes in the
soda—lime-silica system with technological development illustrate the con-
tinuing competitiveness of the three main systems with alternative materials.
Initially, it was essential to have high soda contents in the melt to produce
satisfactorily low melting temperatures. The resulting glass had poor cor-
rosion resistance, that is poor durability’® (3.6) and relatively low strength.
The advent of improved furnace refractories containing magnesia and
zirconia (5.5, 5.6) and increased furnace efficiency, have enabled reductions in
soda content producing increased durability, better strength, and more
 GLASS 133

widespread applications. Further improvements in the durability and strength

were achieved with the addition of 1-2 % of Al,O,; structural models must
be able to interpret the dramatic effect such small amounts of additive can
have on glass properties. Another major development was the discovery that
replacement of part of the lime by magnesia produced a lower liquidus range
and better sheet glass drawing ability, with less risk of leaving the glassy state
through recrystallization (devitrification).
The development of high strength fibre composites (2.13) has required the
basic systems to provide high modulus glasses that can be fiberized. Fibre can
only be commercially drawn in a certain viscosity range, and several general
principles concerning composition and elastic moduli of the base glasses have
emerged from efforts in this area.
(i) Oxides of periodic group I elements decrease the modulus; the reduction is greatest for the
heavier elements.
(ii) Oxides of group II elements, particularly BeO, increase the modulus.
(iii) In group III, B,O, lowers the modulus but Al,O; raises it.
(iv) The silica content should be as low as possible.

These guidelines have led to the development of M-glass from the soda—
lime-silica system as a source of high modulus fibres: in weight percent
consisting of SiO, (53.7), CaO (12.9), MgO (9.0), BeO (8.0), TiO, (7.9),
CeO, (3.0), Li,O (3.0), ZrO, (2.0) and Fe,O; (0.50). Clearly this is a far cry
from the basic system. Another example is the most important glass so far
developed for use in fibre manufacture, namely E-glass. This is a development
of the borosilicate base founded on a eutectic in the CaO-Al,O,;-SiO,
system known to occur at SiO, (62.0), Al,O3 (14.7) and CaO (23.3). Total
alkali oxide content is kept below 2% to achieve good corrosion resistance
as well as high modulus. Up to 13 % BO, is used to replace SiO, in order to
reach a lower liquidus temperature for glass working, which isa feature lost
on reducing the Na,O content so severely.
These developments have paralleled major advances in surface treatments
to improve strength and durability. Whether the material will continue to be
constantly competitive is open to question, in the face of demands from the
electrical industry for better metal sealing glasses and for glass able to
withstand attack at high temperature by metal vapours; the electronics
industry for electron-conducting glasses; the optical industry for high re-
fractive index glasses for fibre optics; and the ceramics industry in general for
compositions suitable for the manufacture of glass ceramics (4.1).
Continued competitiveness requires serious reappraisal of glass structure
and glass science, together with a willingness to explore new raw materials
such as La,O3, V,O, and non-oxide materials.
Glass is worthy of more than the single chapter we are confined to here.
Our study will be structured as shown in Fig. 3.1. This scheme highlights the
way important physical and chemical properties, and the prediction of glass
formation, relate to ideas of structure; how the ideas of structure are related
134 CERAMIC SCIENCE

Glass (definition, role of viscosity)

Structure models

Prediction of glass formation <> Crystallization

Physical properties (mechanical property improvements)
1
Chemical properties Glass ceramics
Durability
Glass to metal and ceramic bonding

Figure 3.1 Study paths.

to the important concept of viscosity, and how the material is defined. All the
study paths focus on glass ceramics, which are considered separately in
chapter 4.

3.1.1 Definition
A common, and yet too restrictive, definition of glass is “an inorganic product
of fusion which has cooled to a rigid condition without crystallizing.” This
definition has several drawbacks. First, it suggests that glasses have to be
inorganic, and so excludes the many organic glasses. It fails to point the way
to useful common areas between polymers and glasses. Most of all, it focuses
on one method of preparation, that used in the glass industry, and this has
probably deflected effort from seeking alternatives. For example, in 1978 it
was shown that good silica glasses can be made from gels without melting.°*°
This method depends on the hydrolysis of silicon alkoxide solution:
Si(OCH;), dissolved in methanol is reacted with water containing around
0.008 % NH, with vigorous stirring to ensure reaction throughout. Initial
hydrolysis followed by polymerization as the solvent evaporates produces a
gel according to the equations
Si(OCH,),+xH,O0 — Si(OCH;),_,(OH), +xROH.
2Si(OCH3),4 -,(OH), > (OH),—:(OR),4_,-Si-O-Si(OR), _,(OH),_ ; + H,0.
Further dehydration-polymerization occurs as the solvent evaporates, until a
thick mouldable gel is formed that only needs heating to 550°C to produce
silica artefacts.
A more acceptable definition of a glass is “a non-crystalline elastic solid, ice.
2.0nm maximum order, with a viscosity > 10'*:> poise (10!7-5>Nsm_~7).
This definition has several implications worth considering. First, why is
viscosity important? This arises because there is a time dependence super-
imposed on the temperature dependence of glass properties. It is the viscosity
at which the material obeys Hooke’s Law, and after the application of a stress
it will return to the original state through bond stretching and not molecular
flow (we shall soon return to this more quantitatively). In the meantime, to
 GLASS 135

liquid state

PROPERTY

TEMPERATURE

Figure 3.2 Schematic plot used to determine glass formation.

further amplify the definition, one can consider what happens during glass
formation by cooling a liquid. The usual approach is to consider how a given
property changes with temperature, a schematic plot of which is given in Fig.
3.2; this shows a transition range when a supercooled liquid, that is one in
thermodynamic equilibrium, exhibits a transition commonly known as the
glass transition. On further plotting of a thermodynamic function such as C,
against T (Fig. 3.3) the second order nature of the change is emphasized by
the break in thermodynamic properties associated with glass formation.
Depending upon the rate of cooling, the glass transition temperature, also
known as the fictive temperature, can have different values within the trans-
formation range because it is determined by relaxation processes in the
cooling liquid.** Within the glass transition range, the Maxwell relaxation
time is increasing as the viscosity rises in accord with equation (3.1).

se (3.1)
Here t is the relaxation time, 7 is the coefficient of viscosity, and G is the
instantaneous shear modulus in the absence of viscous flow. We can see from

normal mpt. Ty;

Cp
J°K' mole’

Figure 3.3 Plot to determine the glass transition temperature.

136 CERAMIC SCIENCE

Figure 3.4 Graphical method to determine the absolute glass transition temperature.

equation (3.1) that at the viscosity referred to in the above definition,

t ~ 300s for materials with typical G values around 30GNm~?, and
molecular adjustments in order to assume the “correct” i.e. thermodynamic,
structure take minutes as opposed to small fractions of a second at viscosities
in the region of 107Nsm”. It is easy to visualize a given cooling rate
becoming too high, and the system falling out of equilibrium with the
resultant curvature of Fig. 3.3.
From this picture we might assume that it is possible to cool a liquid so
slowly that it can adjust with no T, being observed, that is the transition
range extends from T,, down to as low a temperature as one cares to go.
However, if the plot is remade as C, v. In T as in Fig. 3.4, it is possible to see
that this cannot occur and there is a lower limit to the transition range
defined by a temperature T,,. Because Fig. 3.4 has a logarithmic temperature
scale, and since
a C
s= | — dT,
ras
the entropy of a given phase at any temperature is given by the area under
the curve up to that temperature plus any entropy increments due to
first-order phase changes. Since the supercooled liquid in Fig. 3.4 has a higher
heat capacity than the crystalline phase, it follows that on cooling from the
melting point the rate of loss of entropy in the liquid exceeds that from the
crystal. Hence a temperature would be reached on cooling when the total
entropy in the supercooled liquid would be less than that in the material
following the thermodynamically favoured route, i.e. freezing with a large
entropy loss followed by vibrational entropy loss in the crystalline phase.
This result is barely conceivable, since entropy is a measure of probability
associated with molecular disorder, and a liquid less disordered than the
crystal would be hard to visualize. Hence the lowest temperature to which the
liquid can be cooled before transition is that which makes the area between
the liquid and crystal curves match the area representing the entropy of
fusion when drawn on Fig. 3.4. It is bounded by 7;, the extended super-
 GLASS ue
cooling line, the crystal line, and the T g, line which then defines the absolute
transition temperature. The absolute transition temperature, T,,, is that at
which the heat capacity of the supercooled liquid must change drastically
because the entropy of the supercooled liquid equals that of the crystalline
solid.” It must be emphasized that the heat contents would not be equal at
this temperature, because
T,

AH = | C, aT
Ty

and a linear temperature scale would be required. Thus the free energy of the
glass at T,, would still exceed that of the crystal and the supercooled liquid
could not pass continuously into the crystalline state.
Even though it is not possible to observe T,, in practice, it is always
possible to find it graphically in the above manner. What use is it when
found? There are some applications. First, the frozen-in entropy of a glass
can be calculated from the relationship Sy = AC, InT,/T,,, where AC, is
the change in heat capacity at T,, T, is the observed transformation
temperature at the rate of cooling used in the process, and S, is the disorder
entropy. When wishing to compare the properties of different glasses, it is
then possible to compare only glasses with the same degree of disorder. A
second empirical application is that useful technological glass formers have
T,/T,, > 1.3, which can be used as a test to grade systems with respect to
potential exploitation. Some values of absolute and measured transformation
temperatures are given in Table 3.1.
It should now be clear that it is incorrect to call a glass a supercooled
liquid, since a supercooled liquid is in thermodynamic equilibrium and has
the “correct” structure for any given temperature, whereas a glass occurs after
a second order transition and has a structure expected for a higher tem-
perature at all temperatures below T,. It is not sufficient to state the
composition of a glass, since its experimental transition temperature and
absolute transition temperature are needed to define it.

Table 3.1 Measured and calculated glass _ transition

temperature

Substance Deis) Typical T, (°K)

ethyl alcohol 63 90
H-SO);. 3H5O 135 158
Ca(NO;), °4H,O 202 220
B,O, 335 523
SiO, +Na,O0 753
(75% :25%) Sue
SiO, +Na,0+CaO 2
(70 :20: 10) oe oe
pyrex 625 823
PbO-SiO, 426 713
138 CERAMIC SCIENCE

3.1.2 Viscosity
It is important to consider viscosity, since its value is specifically mentioned
in the definition of glass and since it is vital in determining the usefulness of
glass as a material.
In the definition, the value 10!7°° Nsm? for the coefficient of viscosity
was chosen as the line demarcating behaviour like a Newtonian liquid or a
Hookean solid. How realistic is the division? A glass might be assumed
always to flow viscously, since there is no sharp change in the coefficient of
viscosity, as there is when aliquid crystallizes. Here the time aspect must be
re-emphasized, since the time for permanent structural adjustment towards
equilibrium is 300 seconds at the defined viscosity. This means that if stress is
applied quickly and resultant strain is measured, the material will appear to
behave like an elastic solid, with stress proportional to strain. Long applica-
tion of stress would, however, produce permanent deformation as the
structure flowed viscously. As the temperature is raised and viscosity falls, it
soon becomes obvious that the time span when the material would appear
Hookean becomes too small to be useful; for example, at about 500°C when
n = 107Nsm~?, molecular adjustment would be about 3 x 10~*s. The value
10/25 Nsm~?, giving minutes for structural adjustment, is taken as con-
venient for denoting the change from liquid-like to solid-like behaviour. At
room temperature, coefficients of viscosity as high as 10’°Nsm~? are
encountered for glasses, which implies that loads near to the fracture stress
would have to be applied for aeons of time to deform the glass. Thus, one is
justified in viewing glass as a Hookean elastic solid at ambient and low
temperatures. This is very important to scientists, because if glass does not
behave viscously at temperatures not much below T,, and since it does not
have a recognizable ordered structure (3.2) through which dislocations can
induce plastic deformation, then we have the perfect model material for
brittle mechanical behaviour. Glass has been used for the development of
fracture mechanics models (3.4.).
Above T,, the material becomes a liquid exhibiting Newtonian viscosity, '
the consequence of which, as shown below, is the easy drawing of fibre
and sheet glass. By Newtonian we mean that shear viscosity 4 =
(stress)/(strain rate). Under the influence of stress a Newtonian fibre extends
whilst the volume remains constant.
For a-cylindrical fibre, ¢, = kné where o, is the tensile stress, é is the strain
rate, and k is a proportionality constant.
Hence
F
a kné. (3.2)
At constant volume the increase in length must equal the decrease in cross-
sectional area, dé = —dA, but since strain is dimensionless, this is written
. —dA
dé = ——.
A
 GLASS 139
In order to get strain rate, divide through by dt.

dé_dA/A_dA 1
a dt. ae A

ae=— an d e = — :
Aé=-—-A

From equation (3.2), since F = knéA

then F = —knA
or A= —F /kn. (3.3)

Thus the rate at which the fibre gets thinner is directly proportional to the
applied force, not the applied stress, and therefore thinner regions decrease in
cross-section at the same rate as the thicker regions. Equation (3.3) indicates
that as the coefficient of viscosity rises the axial strain rate decreases. Also,
since thinner regions of the fibre will cool more rapidly, their viscosity will
rise, producing a decrease in the rate at which the thin region gets thinner,
hence allowing the thicker region to catch up. In this way the fibre keeps
extending without serious necking-down.
Commercial glass fibre manufacture exploits this by extruding molten glass
through platinum orifices, about 0.79 mm in diameter, heated by the passage
of a direct current. As the glass emerges it is rapidly pulled to draw it to a fibre
3.6-19.1 x 10° mm in diameter. Each continuous filament made this way is
protected, by coating it with a size of starch in oil, before combining it into a
strand with other fibres on a winding machine.
The float glass process exploits the fact that glass poured on to molten tin
spreads out to a perfectly flat parallel sheet as a result of the viscous
properties.*° Glass sheet formed this way has an equilibrium thickness of
approximately 6.8mm determined by the surface tensions of glass, tin, and
the glass-tin interface, and by the densities of glass and tin according to the
expression
[as fo sae tsa) ;
g° Pall ae Pal Psn)

where °yg1, gy, and "ys, are the surface tensions of the liquid glass,
glass-tin interface, and liquid tin respectively and p,, ps, are the glass and tin
densities. Thinner glass sheet of uniform thickness is obtained by mechanically
stretching the initial ribbon when its viscosity reaches 10? Nsm“?.
The analysis of such favourable stretching behaviour depends on the fact
that the viscosity falls with rise in temperature. What form has the temperature
dependence? It is possible to obtain it from the simple model shown in Fig.
3.5 in the following manner. For molecule 1 to move to position 2 it must
breakaway from its surrounding molecules, i.e. there is an energy barrier ¢
that must be overcome. This can be achieved by utilizing thermal fluctuations.
140 CERAMIC SCIENCE

® ——> Layer 2 moving —_—_

©) © Layer 1 stationary
©

Figure 3.5 Model used to derive viscosity v. temperature relationship for general liquid
structural units.

No is the frequency of molecular jumps from 1 to 2 under thermal activation,

which equals
kT ene é
heer):
There will be no net transfer in practice, since activated molecules can move
equally either way from the top of the energy barrier. In layer 2, the applied
shear stress t produces a force tx on molecule 1 with cross-sectional area a,
which on taking the molecule to the top of the energy barrier involves energy
taA/2 in the form of work done, where J is the distance from molecule 1 to 2.
Hence the thermal energy required to surmount the barrier is reduced from ¢
to e—(taA/2). From the top of the barrier to return to position 1 requires
extra energy, ¢+(ya/2), and so there is now a net rate of movement to the
right given by (3.4):
Ne ieE (aie 7 sie .[e a

athe me THA (—yad) 3.4

hk LET Ore el Sey
The expansion of the square bracket can be neglected beyond the first term:

ae kT 8 —e)\| 2tad
Oe NRT
THA —€é
No S =—h exp(—
=)}. (3.5)

In Fig. 3.5 the velocity gradient across the two layers as a result of the net rate
of jumping is
(distance per jump in layer 2) x (number of jumps per second) AN
Ay Ay
Since viscosity
____ Shear stress
| velocity gradient
 GLASS 141
then
z Ay
=—=T
Nae
Substituting for No,

e hh, (é)
= ie ae kT

Given the isotropic nature of liquids and the nearly close-packed structure,
A, =~Aand «A will be the effective volume of the molecule ~ «*. Thus

ae
eer
Expanding this to molar proportions:

zy ex (E) 3.6
m

where N is Avogadro’s number, V,, is the gram molecular volume of the

liquid, and E is the molar activation energy needed to surmount the potential
barrier between equilibrium positions; the activation energy for viscous flow
can be used for this term.
Equation (3.6) is the Eyring Equation, and shows that the viscosity always
increases as the temperature decreases. It is found that this expression
describes quite well the behaviour of organic liquids with only van der Waal’s
bonds. Highly polar molecules, or those with directed covalent bonds, do not
conform well to equation (3.6). For example, a plot of logy against 1/T for
25% soda-silica glass does not produce a straight line with slope E/R as
predicted, but a curve with slopes varying from 80-600kJ mol~*. The only
known exceptions for oxide liquids to the observation of curved logy v. 1/T
plots are the tetrahedral network liquids SiO,, GeO, and albite (1.5.4). All
these display Arrhenian behaviour over the full range of measured viscosities.
In order to make the plot linear, log is plotted against 1/(T— T,) when it
is found that Ty has a value close to T,,. This indicates that the determination
of viscosity can be used to find the absolute transformation temperature. The
same experiment can be used to obtain values for E,, the activation energy
for viscous flow, which, from the model we have used, ought to be comparable
to the latent heat of vaporization of the liquid, i.e. the energy required to form
a hole. Since a liquid has some free volume, the actual value is around 40% of
the enthalpy of vaporization.
The Eyring Theory has difficulty in predicting the non-Arrhenian behaviour
of the viscosity of liquids mainly because it concentrates on one requirement,
namely that a molecule must have sufficient energy to break away, whilst
assuming that an empty site of constant volume large enough for it to move
into is available.
142 CERAMIC SCIENCE

Turnbull?! has highlighted this deficiency by showing that the sizes of the
holes vary statistically and that the free volume in the liquid only occasionally
opens up voids large enough to allow diffusive displacements. Combination
of the two ideas, Eyring’s plus Turnbull’s, leads to equation (3.7) which
provides a better fit to experimental data for materials ranging in complexity
from molten soda glass to liquid argon.
n = Ay exp[E/RT +)V,/V]. (37)
Here y = aconstant in the range 0.5—1.0, V) = van der Waal’s volume of the
molecule, and V = average volume per molecule in the liquid.
Determination of glass viscosity. It is quite common to use the traction
technique to determine the dynamic viscosity 7 using a small glass fibre in a
vertical tube furnace. Fibres, typically 1 mm in diameter, are butt-welded to
silica extension rods. Elongation rates are measured, using a cathetometer, by
measuring the time taken for a mark on the silica rod to move given
distances.
In the limit of small shear stresses, y = ns applies, where s is the shear rate.
For viscous deformation of a filament under uniaxial tensile stress,
o = 3né (3.8)
where éis the axial strain rate.
Integration of equation (3.8), assuming 7 and V are constant, gives:
mgl,lot
ee ee
To obtain 7 in poise, the mass of the applied load m is in grams; g is in cm
s~?;1, and |,, the filament length at the start and finish of the experiment, are
in centimetres; t, the duration of the test, is in seconds, and the fibre volume
is in cm*. Appropriate changes will give y in Nsm_~?.
Determination of 7 in the above manner at a range of temperatures will
lead to a value of E,. An estimate of E, can be achieved from the empirical |
relationship between melting point and activation energy for viscous
flow (3.10):
E, = 3.7RTyy. (3.10)

3.2 Structural models for glass

Several descriptions of the structure of glass have been developed** since
Goldschmidt’s (1926), but one appears to be dominant, namely the model of
Zachariasen and Warren. Goldschmidt noted that in simple oxides glass-
forming ability and the radius ratio (5.3.7) were correlated: glass-forming
ability was associated with a ratio of 0.2—0.4. This, in ionic solids, is
associated with tetrahedral distribution of four anions around each cation,
 GLASS 143
the implication of which is that glasses have structures like those oxides
containing tetrahedral polyhedra (5.3.7).

3.2.1 Zachariasen random network model

This was developed following the Goldschmidt observation, which
Zachariasen found to be unsatisfactory even on an empirical basis, since (for
example) BeO with a tetrahedral coordination of Be would not form a glass.
The following observations were put forward in support of this model.
(i) In materials that can exist as glasses and crystals, the mechanical
properties, the surface energies, the hardness, and the fracture stress deter-
mined under special conditions, are so similar that the interatomic forces
must be very similar.
(ii) The interatomic forces in glass-forming crystals lead to extended 3-D
structures, and therefore from (i) the atoms or ions in glasses must form
extended 3-D networks.
(ii) X-ray diffraction experiments with glasses show that the network is not
periodic and symmetrical.
(iv) The randomness of the network results in widespread values for M—O
distances and angles, so that on heating an increasing proportion of the
bonds are broken, giving a gradual decrease in viscosity and no sharp melting
point.
(v) The randomness of the oxide network in glass results in the glass having
a higher internal energy than the crystal. Zachariasen argued that the
difference could not exceed 10-14%, or devitrification would always occur
on cooling.
(vi) To ensure that no X-ray pattern would be achieved under normal
experimental conditions, an arbitrary value of +17%% was selected for the
randomness in bond angles.
Using the values placed on internal energy differences and bond random-
ness, a method can be developed to test the ability of crystal structures to
support a random network structure. Crystalline oxides may be conveniently
considered in terms of the nature of the metal-oxygen polyhedra they contain.
A sizeable unit of these linked polyhedra, say 100, can be considered in an
ionic sense in order to calculate the electrostatic potential at a point in the
centre of the unit as in 5.3.4. The bond angles in the chosen unit can then be
randomized within the range +17 % and the electrostatic potentials for the
crystal and random arrangement recalculated. If the difference is less than
14%, then such a random structure can exist, i.e. a glass with that structure
can be formed. Figure 3.6 demonstrates how this method can be used to
distinguish glass structures. Such calculations lead to Zachariasen’s rules,
which can be used to determine glass formers and define their structures from
a knowledge of oxide structures (5.3.7).
(1) No oxygen atom may be linked to more than two metal atoms.
(2) The metal must have a low coordination number.
144 CERAMIC SCIENCE

we
RANDOM *17%

@—C distances

angles

Figure 3.6 Glass-forming possibilities from planar triangular structural units.

(3) The M-—O,. polyhedra share corners with each other, never edges
or faces.
(4) For glasses with 3-D structures, at least three corners of a polyhedron
must be shared.
When these rules are applied to the formula types of known oxides, it is found
that:
(a) M,O, MO cannot satisfy the rules and therefore cannot be formed into
glasses—for example, MgO has a structure containing MO, polyhedra
sharing faces, so breaking rule (3), and PbO has a structure in which chains
of PbO, share O—O, edges also contravening rule (3);
(b) M,O, oxides can satisfy the rules if the oxygens form planar triangles
around M—clearly sp” hybridization of boron (2.2) will produce this, and
boron compounds like B,O, will form glasses;
(c) MO,, M,O, will satisfy the rules only when the oxygens form
tetrahedra around the metal atom;
 GLASS 145

(d) MO3, M,O,, MO, will satisfy all the rules only when the oxygens form
octahedra round the metal atom and the octahedra are linked at the corners.
(At the time Zachariasen was not aware of oxides that formed glasses from
group (d), and so modified rule (2) to restrict the number of oxygen atoms
to three or four around a metal atom.)
The success of these rules and the general acceptance of the random
network structure of glass by glass technologists, it might be argued, have
been responsible for some of the development restrictions that are now
evident when considering modern work. Since Zachariasen, diffraction and
spectroscopic data have indicated unambiguously that the basic structure
unit in glassy SiO, is the regular tetrahedron consisting of four oxygens
about a Si atom such that the O-Si—O angle is the tetrahedral angle
(cos~} 1/3).3° Each Si has four oxygens at a fixed distance of 0.162 nm. Each
oxygen is 2-fold coordinated by Si atoms and acts as a bridging atom between
two SiO, tetrahedra. However, the Si-O-Si angle joining tetrahedra exhibits
a range of angles (120°-180°) with a maximum around 144°. It is this
variation in angle that gives rise to the topological disorder in silica.
In order to accommodate the vast range of more complex glasses,
Zachariasen’s rules had to be modified as follows to describe systems that
formed glasses when a glass-forming oxide was mixed with a non-glass-
forming oxide.
(a) An oxide glass is formed if the sample contains a high percentage of
cations surrounded by oxygen tetrahedra or triangles.
(b) The tetrahedra or triangles only share corners.
(c) Some oxygens are linked to only two network cations, i.e. there are
bridging oxygens and non-bridging oxygens (Fig. 3.7).
In this way, the terms “network-forming oxides” and “network-modifying
oxides” arose: oxides such as Na,O break network bonds and “open up” the
structure. Zachariasen suggested that modifying oxides would contain large
cations of low charge so that they do not exert excessive polarization forces

Vers
a one oxygen

_—_e * eee @
ROGz GN ° mg
BO
on iD pane
: Oo
ela ree
éo-

Figure 3.7 Glass structure formed from network oxide plus modifier oxide.
146 CERAMIC SCIENCE

(5.3.6) and completely disrupt the random network. Obviously, when the
concentration of the modifier exceeds 50% a fully-connected 3-D network
cannot be maintained, and some other structure must be envisaged for glasses
with such compositions.
The validity of this structural model for pure binary glasses such as SiO,
(of a 3-D random network modified by addition of non-glass forming oxides)
can be tested by investigating the properties it implies. A linear decrease in
viscosity as the structure is “opened up”, then the increasing ability to
conduct electricity by cation movement, and the increase in thermal expan-
sion coefficient as the modifier content increases, are all expected. It is only
when some of these properties are viewed more carefully that some anomalies
arise.
(i) Viscosity. Equation (3.6), relating viscosity changes to temperature, leads
to a derivation of the activation energy for viscous flow. When the activation
energy for viscous flow is determined for silica glasses containing a variety of
modifying oxides, a drop of 100 % is noted when 0-15 mole % is added, but in
the range 15-60% the activation energy changes by only 50%. Thus two
approximately linear regions are observed, with an abrupt change occurring
in the 10-15 % region which is not predicted by a simple “opening up” of the
network by the modifier.
(ii) Heat radiation. Transmission of heat through glass at high temperatures
has considerable technical importance, and as a result an expression of
radiative conductivity has been defined:
8n?20T?
Kyad =
a

where nis the refractive index, o is the Stephan-Boltzmann constant, « is the

absorption coefficient of the glass, and T is the temperature. From this it can
be seen that K,,q is strongly dependent on the temperature, on wavelength
since it determines «, and on structure since that determines n. A plot of Kraq
against frequency shows that the absorption maximum shifts to lower:
frequencies and the peak becomes broader as the temperature increases.
Addition of modifier to SiO, has the same effect as a temperature increase,
with a gradual broadening and peak maximum shift until 13°% Na,O is
reached, when there is an abrupt increase in the peak width and absorption
index shift. Here again the change is not uniform.
(ii) Partial molar volume. When KO or Na,O are added to SiO,, the partial
molar volume of oxide decreases sharply to a minimum at 13 mole % K,O,
then rises again with increasing concentration of K,O. Accurate studies of
Li,O-SiO, and K,O-SiO, melts at 1400°C show that the thermal expansion
coefficients have a marked inflection at 12 mole % which is indicative of a
fundamental change in the structure of the liquid at that composition. Such
a change might be expected to continue into the glassy state.
(iv) Diffusion. Even in glasses formed from a network oxide only, data for the
 GLASS 147
diffusion of radioactive ?*Na into the solid suggest the process is more
complex than might be expected from movement of Na* through a random
network. An Arrhenius plot of the diffusion coefficient shows discontinuities
at 573°C and 250°C, with changes in both absolute values and in activation
energies at both temperatures. It is common to evaluate such a situation by
assuming two different diffusion sites or different phases in the material. Such
an interpretation is not obviously in accord with a statistical random
network structure. Despite general acceptance of the Zachariasen model, the
structure of glass is perhaps not so clear-cut as it suggests. Some other models
are now considered.

3.2.2 Porai-Koshits crystallite model

The Russian school of glass scientists has derived an alternative to the
random network model based on several experimental observations.”°
(i) Glasses appear to devitrify gradually, beginning at a number of centres
in the sample. This was felt to be in conflict with the predictions of the
3-D network theory, which suggested that the continuous network must be
broken with all pieces forming crystallization centres, and therefore de-
vitrification should not proceed gradually, but all at once.
(ii) Prolonged heat treatment of Na,O-—SiO, glasses followed by X-ray
investigation showed a gradual sharpening of the “peaks”, indicating a
gradual transition from glass to crystalline state.
(iii) The X-ray diagrams of soda-silica glasses could be constructed by
adding together the glassy diagrams of SiO, and Na,SiO3, each being given
a weighting factor proportional to the composition and structure. One set of
experimental results is shown in Fig. 3.8.

np a °o

nN ° °o

=~a °o

100

a o

[in
units
|Intensity
electron
absolute
7 [Na2S'0,] + [2] f

SiO, we
0-1
Sin

Figure 3.8 X-ray intensity pattern of vitreous Na,O-2SiO, ; Na,SiO, —-—-; SiO, -——
and the summation of SiO, + Na,SiO3"°:-°.
148 CERAMIC SCIENCE

(iv) Critical examination of SiO, and K,O-SiO, glasses, using Fourier

methods to get radial distribution curves from X-ray data, showed that as the
concentration of K,O increased the “peaks” move to smaller r values and
there is some increased ordering apparent from the resolution of “peaks” at
higher r values. It was reasoned that if the K* enters the SiO, network in a
random manner, the resultant expansion should shift the peaks to higher
values of r. Hence the results are more compatible with an unchanged SiO,
structure plus a new structure in which K * ions are located.
(v) Recent low-angle X-ray scattering between 0°—2° indicates the possible
existence of inhomogeneities in complex glasses. Vycor (4% Na,O, B,Os,
96% SiO,) has heterogeneities around 15.0nm in diameter. Fluoride glasses
based on MF, have heterogeneities around 40.0 nm in diameter.
Based on these observations, the crystallite theory was proposed, which
postulates glass as zoned into large areas of disorder separating regions of
relatively higher order, regarded as small distorted crystallites. The size of
the crystallites will increase with decreasing temperature on cooling from the
melt until the increasing viscosity limits diffusion.
This concept of structure is quite different to the random network, but still
implies a uniform change in properties with increasing modifier content, and
not the behaviour noted above. The crystallite theory proposes a non-
homogeneous structure for glass on a molecular scale that is permanent and
determines the fundamental behaviour of the material. Unfortunately it is
difficult to prove, since a number of possible inhomogeneities may cause
confusion when seeking experimental confirmation. Some examples of in-
homogeneities other than the suggested structural one are:
(a) Technological inhomogeneity. This encompasses crystalline inclusions, bubbles and “cords”,
and was known to ancient glass makers, who adjusted composition accordingly, e.g. by adding
arsenic for bubble removal. It is commonly believed that the absence of visible cords, crystals or
bubbles means that this inhomogeneity is absent, but recent electron microscope investigations
have shown that “micro-incomplete melting” can occur, resulting in its retention. This type of
inhomogeneity, whilst inherent in the production method, is not a necessary characteristic of the
vitreous state, although special preparation conditions (temperature and time) must be employed’
to ensure its removal.
(b) Thermal density fluctuations. These cause scattering of X-rays in a manner weakly dependent
on the scattering angle. A sign that these are influencing any experimental observations on glassy
systems is the linear dependence of their intensity on temperature and coefficient of com-
pressibility £.
(c) Liquid phase separation. This can occur in any liquid or solid solution when such a separation
leads to a decrease in the free energy in the system. It is extremely important in glass-ceramic
formation and is discussed in 4.4.

3.2.3. Discrete-ion concept

In an attempt to explain the variation of physical properties of liquid silicates
over the 10-15 mole % modifier oxide range, Bockris postulated the presence
of the discrete ions Si,O$; and SigO8 5, which, if retained in glass, would
cause non-random distributions of modifier cation, since charge has to be
neutralized locally.
 GLASS 149

At the metasilicate composition Na,SiO, in the melt, two types of ring

anions, Si,0§~ and Si,0%;, are envisaged, formed by the polymerization
of SiO{” tetrahedral anions as detailed in 1.5. The Si,O$~ and Si,O8;
units are preferred because they have the angles around 140° needed to satisfy
partial molar volume data. The ideas developed for the estimation of
inter-ion attraction (5.3) are applied, and condensation polymerization is
postulated as occurring between the units SiO{~’ Si,O§~, Si,O8%; and
resultant polymers, in such a way that two conditions hold: first, the products
must have Si-O-Si bond angles close to 140°, and secondly, the Si-O bond
lengths must only vary over experimentally observed values. Structural
evolution of the glass-forming liquid, and any resultant glass can then be
envisaged as follows. As more SiO, is added to the Na,SiO, liquid, the
oxygens at A and Bin Fig. 1.3 can polymerize with SiO{~ ions to form cages
where the Si-O-Si angles are still about 140°, whilst producing a complex
ion whose linear size is little increased but which now has a composition of
either Si,O$5 or SigO8>. Further condensation polymerization of these
new units at oxygens C and Din Fig. 1.3 leads to Si,,O$; and Si,,083,
with linear sizes 2.3-2.8nm at about 11 mole % Na,O. The increased linear
size can now be thought to bring about instability as the cation—anion
distance in the ion—ion interaction equation (5.13) causes a sharp drop in the
Coulomb stabilization factor for the system, and as more SiO, is added no
further polymerization occurs, but rather a radical change in structure. This
sequence of events may in fact predict a change from discrete ion crystallite
structure to 3-D network at this composition, and certainly it would help to
explain the reported changes in physical properties of glass and glass-forming
systems around 10-15 % modifying oxide content.
The relative amounts of the limited number of polymers given in Table 3.2
are composition-dependent, and suggest a glass structure in which there is
not one type of crystallite, but rather where the crystallites vary with
temperature.
An alternative to the Zachariasen random network model of corner-linked
planar BO, groups for the structure of B,O, glass, (2.9), is also based on
“molecular cages”. The cages are taken to be B,O, groups which “cluster”

Table 3.2 Condensation polymer anions present in glass-forming

liquids

%M,0 Polymer Anion Linear size (nm)

50 3 x SiOZ~ SiOge 0.6

4x SiOZ7 SyOrn 0.7
33 (3x Si0$-)+Si;,0§- — SigO95 0.6
(4 x SiOZ~)+Si,Of> Si,0$5 0.7
11 4x Si,O85 $i,,02; 23
4x SigO85 Si;,085 2.8
150 CERAMIC SCIENCE

together as a result of lone pair interactions from oxygen to boron. The

strength of these inter-cage bonds is about 50kJ mol~*. For geometrical
reasons, all four boron atoms in a cage cannot simultaneously form this type
of covalent bond, which results in non-crystalline structures. Changes in the
total number of such bonds with temperature lead to a variation in the
activation energy of viscous flow.
Addition of alkali oxide causes the formation of linked pairs of B,O, cages
through a bridging oxygen, and formation of quadrivalent borons which
must carry a negative charge and so fix a pair of cations near each cage-pair.
A result of this orientation effect is a collapse of the inter-cage covalent
bonding, with a subsequent fall in viscosity. As more cage-pairs are formed
on subsequent alkali addition, the viscosity rises, as it is inherently more
difficult to cause the strongly-linked cages to flow. In this way an anomalous
viscosity effect which has been observed for B,O3° can be explained.
A viscosity decrease when LiF is added to B,O;, is also anticipated, as
(B,O.F) Li* groups are formed, destroying covalent bonding so that no
oxygen-bridged cage-pairs are formed. Isolated B,O, cages are unstable
without the inter-cage covalent bonds, which may account for the ease with
which H,O causes devitrification of boric oxide glass (2.9): this would occur
by decreasing the number of such inter-cage bonds with preferential co-
ordinate covalent bond formation from the oxygen atom of water.

3.2.4 Five-fold symmetry model

In this model, silica glass is considered to have an arrangement of connected
SiO, tetrahedra in which the SiO, tetrahedra have minimum density. This
they can only obtain if the Si-O-Si bonds form angles of 180°, and therefore
when tetrahedra link to form rings, then the rings must be planar. Five
tetrahedra can form planar pentagonal rings, because the interior angles of a
tetrahedron and a regular pentagon are little different (109° 28’ v. 108°).
Growth of this structure proceeds in six directions from the sides of an initial
tetrahedron to form a network containing dodecahedral tunnels. The five-
fold symmetry of the model precludes crystal formation.*°

a b

Figure 3.9 (a) Dodecahedral element formed from SiO, tetrahedra sharing corners to form
planar pentagonal rings. (b) Dodecahedral element viewed to emphasize the six axes of five-fold
symmetry possessed by such a unit.
 GLASS yl
The dodecahedral arrangement of pentagonal rings is shown in Fig. 3.9. If
on each dodecahedral face a like unit is attached, then none in this second
shell can touch any of its five neighbours whilst maintaining planar arrange-
ments; however, only a modest radial tension could bring about contact.
When the process is repeated to further increase the cluster size by addition
of thirty-two cages in the next shell, it would require considerable tension in
the Si-O bonds to allow cages to touch. Eventually the Si-O bond tension
would be sufficient to prevent further cage addition at outer pentagonal faces,
at least not without some bond breakage. Thus the clusters would not grow
indefinitely. On the larger clusters, broken Si-O bonds would be available for
inter-cluster linking.
In some ways this model provides a connection between the random
network and the crystallite models. Some support is claimed for the model
from the predictions it makes with respect to density, selective gas diffusion
through the pentagonal tunnels, curvature of lattice planes and tensile
strength behaviour.

3.2.5 Goodman’s strained mixed cluster model

In 1975, while considering how the models already proposed for glass
structure might be unified, Goodman’* noted the fact (not hitherto
emphasized) that most glass-forming systems have as a major, or only,
constituent a material that exists in two or more polymorphic forms differing
only slightly in free energy. For example, SiO, (1.5.4), PbO, TiO, (5.8), S, Se
and CdP, are polymorphic and form glasses.
Goodman used this fact to develop a plausible model. He suggests that
above T,, transient cluster of all possible polymorphs coexist at sizes below
that critical to form aself-sustaining nucleus (4.3). Because of the mutual
impediment between clusters of different types of polymorph, no clusters
could achieve a mutually epitaxial orientation. As the melt cools, the clusters
become bonded across parts of their surface, leaving more liquid-like material
to occupy the interstices in such a structure. This stage corresponds to T,,
and one can sense that stress is caused by the different thermal contraction
characteristics of the polymorph clusters across the bonded areas. It is
suggested that the role of the more liquid-like inter-cluster material is to ease
the stresses, and might explain why apparently small concentrations of
modifier can sometimes have important and large effects on glass formation.
However, if more than the optimum amount of stress-relieving impurity were
present, the glass might become less stable, because there would be a
tendency for a new phase to form in the “extra-cluster” material.
The aggregated clusters build up a three-dimensional network throughout
the liquid. A two-dimensional representation is shown as Fig. 3.10. Better
packing will be achieved for lower rates of cooling, and so a transformation
range results. Around each polymorph in Fig. 3.10 a halo of solidified liquid
is shown, and the cross-hatched region is the final cooled liquid. This region
F
S52 CERAMIC SCIENCE .

RN
oe $
ac

Figure 3.10 Strained mixed cluster model for a two-polymorph glass. A = polymorph A,
sub-critical nucleus. B = polymorph B, sub-critical nucleus. C = A plating out. D = residual
liquid, last to solidify. -B—B-—B-— = cluster.

must contain some voids, since most liquids contract on cooling. The
liquid-like region is interconnected in three dimensions, which provides a
path for ionic conductivity to occur in complex glasses.
It should be emphasized that since the associations of linked clusters all
consist of sub-critical nuclei, they cannot be sintered to become “grains”,
since an increase in size would lead to an increase in free energy (4.3). The
presence of the chain-like, sub-critical clusters should give rise to low angle
optical scattering, and this has indeed been recently reported for a number of
optical glasses.

This section has tried to show that ideas on the structure of glass are still
evolving, but that glass structure is not as simple as the crystalline state where
models can be checked against positive experimental findings. The truth is
probably that there are a number of glass structures in the same way that
there are many crystalline structures; hence it is important to accept that all
the models described above represent reality to some degree and not to
concentrate on one to the exclusion of others. From the study scheme shown
in Fig. 3.1, it is now possible from this common base of structure to continue
by several routes into glass science.

3.3 Predictions of glass formation

One of the most puzzling questions about glasses is why some systems form
glasses on slow cooling while others do not, and many solutions have
appeared, based on viscosity-temperature relationships, diffusion rates,
thermodynamic barriers, or structural requisites. Generally, hypotheses pro-
liferate without increasing the understanding—some of these have already
been developed in 3.2.
 GLASS 153

3.3.1 Bond strength and type

These are non-quantitative approaches based on the non-directional nature
of the metallic bond. In sucha situation, there is no geometrical reference to
allow for a description of local disorder. Thus the requirement for glass
formation is predicted as the presence of strongly directional chemical bonds
in the material, which implies covalency (2.2). Clearly, this is too broad since
most covalent materials do not form glasses, and several ionic and mixed
bonding solids can be vitrified.

3.3.2 Coordination environment

Here is enshrined the work of Zachariasen, culminating in the Zachariasen
rules (3.2.1), which should now be reconsidered. It is true that too slavish an
adherence to these rules has limited the the development of new glass-forming
systems; while a good guide, they do not predict many of the new glass
formers in for example the chalcogenide series, and in the metallic alloys.

3.3.3. Strained mixed cluster model'*

Predictions here are based on a knowledge of phase diagrams, for a major
constituent of a glass must show two or more polymorphic modifications.
However, the polymorphs must not be too closely related in structure; for
example, the wurtzite and zinc blende structures not only have the same
coordination but differ only in layer stacking, which may not be sufficient to
suppress nucleation.
An interesting possibility arising from this is that liquid carbon may be
induced to give a glass, since at one atmosphere, graphite and diamond
phases differ only by about 2kJ mol™* (2.5).

3.3.4 Kinetics and the concept offree volume?!

The object of this method is to offer an explanation based on kinetic
considerations for the failure of some liquids to crystallize. It is a nucleation
and crystal-growth theory; for more detailed quantitative considerations of
nucleation, see (4.3). A brief summary will suffice here.
For a pure supercooled liquid system, the frequency of homogeneous
nucleation of crystals, I, is derived as equation (4.8) and rearranged

I = nvexp eon (5:11)

RT
where J = number of nuclei formed in 1cm* of liquid in one second,
n= number of atoms cm~°, v = vibrational frequency of atoms at the
nucleus—liquid interface, No is Avogadro’s number, and AG, = free energy of
activation per mole for an atom to jump across the liquid nucleus interface.
This is a kinetic barrier, and W* is the thermodynamic barrier to nucleation:
K(*y,)° VS (3.12)
ENE GKG
* oes
154 CERAMIC SCIENCE

In (3.12), K is the factor to accommodate different-shaped nuclei, *y, is the

tension of the nucleus—liquid interface, AG, is the molar free energy of
crystallization, and V is the molar volume of the crystal.
The linear velocity of growth U ofa finite crystal is derived in 4.5 as:
G 7

U= ivexp|RT i exp (AG,/RT)] (3.13)

where J is the interatomic spacing and AG” is the kinetic barrier faced by an
atom moving from liquid to crystal.
The linear velocity of growth just after nucleation will have a form similar
to (3.13), with AG” changed to AG,. Here it should be noted that the free
energies of activation for growth ofa finite crystal and a nucleus respectively
are not likely to be equal, and frequently are not of same order of magnitude.
Using these expressions for nucleation rate, nucleus growth, and crystal
growth it is possible to arrive at criteria for glass formation. In order to do
this one must envisage a temperature T’ and AT’ the degree of undercooling,
at which the homogeneous nucleation frequency is 1cm~*s~*. A liquid will
fail to crystallize if no T’ exists. If it does exist, a glass will still result if the
rate of nucleus or crystal growth <10° * atom spacings per second at T’.
The general relationship AG, = AH,;—AS,T can be re-written, using
AS, = AH,/T,, as:
AH SOP
m

by neglecting the variation of AH, and AS, with temperature. This can be
substituted into (3.12):

W* = K(y)°V?
A 2)
(a1, aL )
When I = 1, equation (3.11), taking logs, becomes
—(N,.W*+AG,)
log1g =1 ogm+| RT ae

or
—N,.W* —AG
O=] 2 |ee
a RT RT 2 |
RT' logy = NoW*+AG,
and
Sy \3 2 Sa \3 72
RT’logv—AG, = NoK(y) Ms = Nok! :
(a1,- wee) AH}(1—T'/Tmm)
T.m
 GLASS 5

when

AH;(RT' logv—AG,) = a y)V? _ NoK(y)°V?

~TT a. A leet
Tm
and since T,,—T’ = AT’
AT’ » Nek Vv a2

(3.14)
T,, |AH?(RT'lognv—AG,)
m

where AH, is the heat of crystallization and T,, is the melting point. In
equation (3.14), the tension in the liquid-nucleus interface is a troublesome
unknown. It is found, however, that ‘y, is related to measurable variables by
Go15):
aAH ,
Ss
i= Nie ye (3.15)

where « is a constant ranging between 4—

Continuing to approximate, it is own that AH, can be set equal to PRT,,
where the constant fis close to unity for simple liquids.
Substituting all these into (3.14)

| Kee BRT, i.
Tm RT (log (nv)
—AG,) 218)

Now consider the requirements for the existence of T’ when AG, = 0. Assume
the nucleus is spherical when K = 16z/3, n approximately 107? atoms cm~*,
and y= 10!2s~‘. Then by substituting into (3.16), T’ will exist for all
aB*/3 < 0.9. For liquids containing simple molecules, B*/* is in the range
1.0-1.5, hence homogeneous nucleation should occur if « < 0.6, so one can
take « at the upper end of its experimental range, i.e.3. In practice, for pure
substances that do crystallize, it is found that «f1/> rarely exceeds+ and
therefore one can set «B1/*= 3 in (3.16):
AT’ 2h. Ri Us
Te. |3
m
el
In order to satisfy this relationship, a value can be found for T’ only if
AG, <40RT,,. Thus for I < 1cm~*s~' at any temperature below Tit is
necessary that AG, > ~40RT,,.
Some values?! of the free energy of activation for viscous flow are:
SiO, = 30RT,,, GeO, = 25RT,, B,O3; = 25RT,,, H,O=10RT,, and
glycerol = 15RT,,, (but this increases rapidly with decreasing temperature to
25RT,,). The prediction follows that, of the above, only water will not form a
glass, which is in accord with experimental findings to date.
156 CERAMIC SCIENCE

3.3.5 Cooling rate*®

A semi-empirical approach, based on several simplifying assumptions, is used
here, beginning by considering the equation for lowering the temperature of
a body in contact with a heat sink:
Tee, (tK,,) (3.17)
Tok, ee)
where T, = temperature at time t, T, = substrate temperature, K,, = thermal
conductivity at the melting point, C, = specific heat at constant pressure,
p = density, and d = thickness or radius.
This equation holds for ideal cooling where surface effects are small. When
surface effects are large, K,, = heat transfer coefficient, and (3.17) has the
Newtonian form where K,, is replaced by hd and h is the heat transfer
coefficient.
If hd/K,, < 0.015 Newtonian cooling occurs, but if hd/K,, > 30 the cooling
is ideal.
Where attempts are made to quench ceramics rapidly, the methods
employed may involve thin films moving over a cold substrate, and some
situation between the two extremes above applies. Thus the first approxima-
tion to be made is that Newtonian cooling always applies.
Now if hd is substituted for K,, in equation (3.17), logs are taken and it is
differentiated with respect to t, the instantaneous cooling rate Q is obtained:
dT h
Onn
t
a (3.18)
Substitution of the appropriate values into (3.18) give an order of magnitude
for the quench rate, for example SiO, has T,, = 2010K°, T = 300K°,
h=0.167Jcm~*s~1K~*, p at T,, = 2.1. g/em* and C, = 0.84J mol" *K—},
which predicts for a 201m thick sheet a cooling rate of 8 x 10*Ks~*. This
can be compared toa value of 8 x 107K s~* for H,O.
The question now is whether one can estimate the critical cooling rate below
which a glass will not form. If so, then an assessment of glass forming ability’
can be made by comparing the critical values to the cooling rates achieved on
a cold substrate i.e. those predicted by (3.18). The following approach can be
applied. To a first approximation
*—2Tf (3.19)
where z is a constant, T,,, is the melting point of the crystalline solid, and f is
the jump rate of the structural units in the melt and is related to the
vibrational frequency.
A standard material like NaCl can be used to determine z because T,,, and
the Restrahlen frequency (f) are well known for rock salt. An estimate of
Q* for NaCl must be made from the experimental observation that the
crystallization front advances into molten NaCl at 10-15cms~* when it is
cooled at 10*-10° Ks" '. If the lower limit of detectable crystalline ordering
 GLASS 17)

occurs with crystallites of 0.1 nm, then Q* is that rate which will now allow
the growth of a 0.1 nm nucleus within a 10°K interval of undercooling:
12Scms: *<10°K
PAD G10?
K sah,
10°7cm
Substitution of this value in (3.19) with T,, and f for NaCl gives z as
2 XLOe>.
While a relaxation time for NaCl is easily obtained from infrared absorption
measurements, in general for other materials this is not the case, and so an
approximating equation is useful. From a variety of sources, e.g. Eyring
(3.1.6), f = KT/A%y and the Debye equation for relaxation of a polar molecule
in a liquid in an alternating electric field which gives f = KT/4nya° it can be
assumed that relaxation time relates the same parameters in like dependency:
i NoKT,, RTS
Vn Vn
where R is the gas constant, T,, the melting point, V is the molar volume, and
n is the viscosity.
Substitution of the above into (3.19) gives:
RT. RE
Q* =2x10 °— -T,, =2x10°°—. (3.20)
Vn Vn
Considering the assumptions made, it is useful at this stage to check (3.20) by
substituting values for silica.
TE, =2010K,, V=28.5x 1ie?m??
4 =3x10°Nsm*, R=8.14JK~' mol“,
which gives a value of Q* of the right order of magnitude, around 1°K s~ A
Therefore it is worth pursuing this line of prediction. Using data in Table 3.3,
the critical cooling rates have been obtained for a range of materials. Further
inspection of Table 3.3 shows that these are in good agreement with
experimental observations for cooling rates of 10°K s~*.

Table 3.3 Data for use in critical cooling rate equation

SiO, MgSiO, Mg,Si0, BaTiO; MgAlo, Pb

Tonks) 2010 1816 2163 1910 2378 600

n(Nsm~?) 3 x 10° 20 5.0 1.0 0.0023 0.003
M (kg) 60.1 100.4 140.4 233.3 142.3 207.2
p(k gm?) 2100 3100 3400 4900 3300 11200
V (m? x 107°) 28.5 32:3 13.8 47.5 6.2 18.9
calculated Q* (Ks *) 4 8.5x10* 56x105 1.3x10° 6.6.x 10° 8 x 10
Experimental phases
at cooling rate of glass glass crystals crystals crystals crystals
10° Kesa2 + glass + glass
158 CERAMIC SCIENCE

This exercise shows that it appears to be unlikely that simple purely ionic
or metallic materials can be cooled rapidly enough to form true glasses, but
on quenching a high disorder may arise from small crystallite size.
Equation (3.20) can be rearranged to develop the concept of glass
number.”® Taking logs of (3.20) gives
ball
log er? —log2 x 10~° = log1

log Q*Vn
RT2 +5 pf ea): (3.21)
Bt

Now replace Q* in (3.21) by R, the cooling rate available in any experiment,

substitute all relevant data, and if the equation yields a positive number it
shows that the critical cooling rate for the melt would be exceeded and a glass
would result. For example, SiO, at R= 10Ks~', 7 =10°Nsm 7, V=
3x10°-°m?, T,, = 2010K produces a glass number around +1 which
accords with glass formation at this rate of cooling.
Spinel (5.9) with R at 10°Ks-1,”7=2x10°-7Nsm 7,V =6x 10° °m “?,
T,, = 2378 K has a glass number of — 3.6 which predicts only crystals being
formed even at this fast rate of cooling.

3.4 Mechanical properties

Glass has been considered to be the archetypal brittle solid, and as a result it
has been a rich area for experimental and theoretical studies of brittle fracture
behaviour. Being free from the restrictions of an ordered crystalline state, it
has become the model material for advancing the study of brittle behaviour,
with the result that there is a diverse literature that is beginning to show that
perhaps glass is not as “ideal” a brittle model as first thought and may in fact
have some ductility. This section considers how developments in theories of
glass strength have led to greater insight into the material, and attempts to,
link these with the models in 3.2.
There are several theories of glass strength, but not one is wholly
satisfactory, especially when compared to the dominance of dislocation
theory of plastic deformation in crystalline structures. There are three basic
kinds of strength theory, which differ in the way they approach the question
of flaws in the structure.’ The oldest of the theories postulates no flaws at all
and is concerned with determining the absolute strength of perfect glass. Such
approaches have been numerous,* beginning in the 1920’s; and are associated
with workers such as Polanyi and Orowan. It is proposed to begin the
consideration of glass strength by outlining the Orowan approach to the
problem, which is typical of many.
A second group of theories postulates that flaws do exist in the material,
defines them, and considers how they propagate to cause fracture.’® It is
 GLASS 159

proposed to consider the highlights in some of these approaches, beginning

with the early classic Griffith analysis, and finally noting some of the very
recent lattice trapping ideas.
The third group of theories adopts the interesting approach that applied
stress generates flaws which then propagate.? Theories in this area are
usually very closely linked to the current views on the structure of glass (3.2).

3.4.1 Group 1: flawless glass

The method outlined below is basically that developed by Orowan; no
realistic model of glass structure is involved. When atensile stress is applied
to a glass, the force being applied to produce the strain is opposed by bonding
forces between sheets of atoms perpendicular to the applied stress. In the
region of small extensions, that is the elastic region, the applied force and the
restoring forces in the chemical bonds are equal and opposite.
If the original inter-sheet spacing is dy) and extension in the direction of
applied force is x, then from (2.13)
Ex
Capp = as (3:22)
A
Here E is Young’s modulus, and a problem now arises that (3.22) only applies
to very small extensions. If the model is to extend to total inter-sheet
separation then we cannot expect E to remain constant (2.1). For example,
5.3 shows ionic bond forces to be proportional to 1/r? and 2.2 shows that
covalent bond forces are exponentially related to separation.
It is still worth pursuing Orowan’s logic, however, bearing these points in
mind. If the applied stress is now increased, the restoring stress due to the
bond forces also increases and this situation continues until the bond forces
reach their maximum value o,, (molecular strength), and once Oapp > Om,
breakage occurs.
A form for the restoring stress has to be postulated, and from the many
possibilities a sine function was chosen:

Sapp = Tm sin (3:23)

where J is a distance less than an interplanar separation. Thus when the

planes are extended by x = 4/4, the restoring force will be at its greatest, i.e.
Om. When x is extended to 1/2, equation (3.23) predicts zero attractive force
and planar separation (that is, fracture) occurs.
For small values of x, sin x ~ x and (3.23) differentiates to
do, Be 200m 3.24)
dx A
Equation (3.22) differentiates to
cee! (3.25)
160 CERAMIC SCIENCE

and from (3.24) and (3.25), 276,,d) = EA, and

EA
theoretical strength o,, = (3.26)
21a,
This is almost useful, except that it contains the unknown J, and so the next
stage is to eliminate it by using the fact that on fracture two new surfaces are
formed, each with surface energy per unit area, y, that must be supplied (2.1).
Hence the work done per unit area by the applied stress must equal 2y,
assuming that all other energy losses e.g. sound, heat, are negligible. From
(3.23), fracture occurs at x = 4/2.
A/2 4/2
| odx = | on sin dx am (3.27)
0 0 Tl

Hence Ao,, = 2ny. (3.26) and (3.27) can be combined to eliminate 4 thus
arriving at an expression for the theoretical strength:

Cn = a (3.28)
Ao
This expression is quite general, in that no structure model or bonding
arrangement was assumed, but just a mode of sudden catastrophic failure
across a plane. It is useful for setting the scene, so to speak, and forms the
basis for material comparisons.
Glass is useful as a model material because values for y can be obtained
from experiments with molten drops near T, that allow the surface tension to
be measured and approximated to the surface energy of solid glass. Young’s
modulus can easily be measured and the interatomic separation taken as half
the Si-O distance in an SiO, tetrahedron. In this way a theoretical strength
of around 2 x 10* MN m “? is reached, which is far greater than the practical
strength. Other methods lead to similar answers, and so reasons for the
weakness of glass must be sought. This is of considerable importance, since
we cannot look for a microstructural or textural cause (e.g. grain boundaries),
and there can be no dislocation shear mechanism in view of the proposed
structures (3.2). Why then is glass not the strongest material we have at our
disposal? The classic Griffith analysis, born of this question, points to the
dominant feature of glass science: the existence of microcracks (3.4.2).

3.4.2 Group 2: flaws exist—how do they propagate?

Griffith noted what Inglis had previously shown, that a small crack acted as a
stress concentrator, and for a surface crack inaflat plate

a= (1 ao ae (3.29)
where ga, is the tensile stress at the crack tip, o, is the applied stress, c is the
crack length and ris the crack tip radius.
 GLASS 161

A series of experiments on freshly-prepared glass fibres of decreasing

diameters showed dramatic strength increases below a critical diameter,
which led Griffith to believe that decreasing flaw numbers might be
responsible. Many experiments of this kind assisted his acceptance of the idea
that flaws present in glass were responsible for its weakness. An elliptical
crack within the area of a glass plate of length 2c ~ 10°°m would lead to
stress enhancement around a hundred times. Therefore tiny cracks in the
glass could overcome the theoretical strength by intensifying the applied
stress on the chemical bonds in the tip of the crack.
The next main step was to accept the existence of such cracks and postulate
that they remained immobile until catastrophic propagation. Propagation
could not proceed until the stored elastic strain energy released on fracture
equalled the excess surface energy gained by the production of the two new
surfaces.
In the region of the crack shown in Fig. 3.11, the total energy of the system
is U where
U = elastic strain energy released — surface energy of the crack
The stored strain energy in the crack volume that was released is

(strain x stress) = E Xo,

. Og

x (volume of ellipse, long axis = 1, short axis = 2c.)

Surface energy of the crack = 2 x 2cy = 4cy. Hence

poe NerGe
—Acy. (3.30)

The Griffith condition for crack propagation is 0u/dc = 0, i.e. no net energy
change for an increment of crack length. Thus
21co;2) eae

aC x
a

0, ee 1.0

oo

g to
Figure 3.11 Model of elliptical surface crack showing strained bonds gradually diminishin
unstrained state ahead of the crack.
162 CERAMIC SCIENCE _
and, for fracture,

= oe (3.31)
7C
Hence a crack of about 2x 10~°m, so small it could not be seen, would
reduce the theoretical strength to the values commonly measured for glass.
This approach deals with crack length, while other work is more concerned
with the crack tip radius; it limits itself to a catastrophic situation by not
considering slow crack growth at sub-critical stress values. Nevertheless, it is
a tribute to its viability that Griffith’s crack hypothesis for glasses has
survived and grown in acceptance and importance with application to other
brittle materials (5.4.4), in spite of any real experimental evidence for the
existence of such cracks. Originally, Griffith believed that the cracks existed
throughout the volume of the glass, but it soon became customary to restrict
their existence to the surface, where fracture in glass invariably originates.*°

3.4.3. Characterization of Griffith flaws’? and surface features of brittle fracture

This can be achieved in a number of ways, one of which is easily done in the
laboratory: a glass slide is immersed in a molten eutectic mixture of LINO;
and KNO,. The Li* ions in the melt then exchange for the Na* in the glass
surface under the driving force of the Coulomb potential (5.3). Exchange is
limited to a thin zone, approximately 1 ym deep, where shrinkage occurs as
the smaller Li* replaces the larger Na*. The shrinkage results in a tensile
stress and propagation of the existing surface cracks; each crack so produced
contains the pre-existing Griffith flaw. A glass slide previously deep-etched
with HF to remove the surface will produce considerably fewer cracks, and if
the etching is deep enough, the treatment will fail to produce a cracked slide.
An elegant method has been developed whereby supersonically fast water
drops are made to hit a glass sheet. This results in controlled compression
pulses followed by relaxation tension pulses traversing the glass. Examination,
of glass plates so treated shows a slowly developing pattern of surface cracks
as Griffith flaws are caused to propagate locally. There is no evidence from
such experiments for any cracks starting from within the glass.
So far we have a model in which a brittle material like glass is stable in the
sense that cracks have zero velocity until fracture is initiated, and then failure
is instantaneous. A more realistic view of the process is that atom separation
across the fracture plane does not occur simultaneously but by propagation
of the crack, and therefore there must be some limiting speed bounding the
process. Several methods show that the maximum speed of a crack in a brittle
solid is given by (3.32), where a, is the yield stress, p the density and TFV the
terminal fracture velocity.
TEV = as, [%. (3.32)
p
 GLASS 163

For glass (3.32) predicts a consistently higher value (3 x 10*ms

*) than is
observed. Crack velocity is determined from the fracture surface morphology;
several features consistently appear on the surface of a glass fracture
(Fig. 3.13).
(a) Rib marks. Smooth rippled surfaces are a feature of glass chipped from a
larger piece. The ripples are always curved, with the concave side nearer to the
fracture initiation, or, conversely, they have a convex side which faces the
direction the crack front took; thus they indicate the origin of the fatal
Griffith flaw. Rib marks are produced by transitory changes in the direction
of maximum tension in the glass at the crack tip, caused by elastic stress
pulses. An elastic stress pulse is produced whenever an externally applied
force is changed suddenly, or by the sudden change brought about in local
stress when a Griffith flaw starts to propagate. Stress waves have a velocity
expressed as (elastic constant/p)'/* which therefore must exceed the terminal
velocity of the crack tip as expressed by (3.32). Thus the stress wave will reach
the surface and be reflected back before rupture is complete, and must interact
with the crack tip. Very obvious rib marks probably indicate where the crack
propagation was halted. Less obvious ones demonstrate where the interacting
stress fields produced a change in direction of maximum tension. Since there
are no regular planes of atoms in glass (3.2), there are no preferred fracture
directions within the material and the crack will propagate along a surface at
90° to the local maximum stress at its tip, hence the change in direction and
rib-formation.
(b) Striations. These occur parallel to the local direction of the crack tip and
so appear to radiate from the curvature of the rib marks. They are steps in
fracture surface caused when adjacent sections of a crack front follow different
levels within the body. If the axis of maximum tensile stress becomes rotated
in a plane perpendicular to the direction of crack front propagation because
of the complex tensile stress field that is generated in fracturing glass, a
continuous twisting of that front would be required. To accommodate this,
the original crack front breaks up into separate small cracks that can
individually adjust to the new orientation. These break-aways produce the
separations showing as striations.
(c) Wallner lines. These are another, less obvious, rib-like feature formed from
the interaction of the crack front and stress waves. They provide a method for
determining the TFV of cracks in brittle solids, and are best seen in a low-
power microscope, slightly out of focus, in reflected light mode. Surface
features cause a stress pulse to be generated by the crack front itself as it
passes by, and since this pulse will travel faster than the crack front, it causes
interference with the direction of maximum tensile stress and so deflects the
original crack tip. Figure 3.12 illustrates the process schematically. The crack
front is moving left to right; AA’, BB’ etc. represent positions of the crack
front. At A, a surface flaw exists that causes a stress pulse to be generated from
the changing local stress as the crack front AA’ passes by. Since the stress
164 CERAMIC SCIENCE

ABE C 2D
Vex"
Crack front LF,

A B Cc D

(a (b

Figure 3.12 Wallner line formation as-described in the text.

pulse from A will travel faster than the crack front it “catches up” with the
front at b, c etc. as it travels the longer distances Ab etc. in the same time as the
crack front goes a distance AB. Interaction of the stress fields at b, c, ete.
causes the crack tip to be deflected, and a rib-line Abc is produced, known as
a Wallner line after the author of the explanation just given.
The speed of a crack as it grows can be found if the speed of stress waves in
the fracture surface is known. Figure 3.12b represents the interaction with the
stress pulse generated at A with two positions of the crack front, at BB’ and
CC’, from which it is possible to derive the expression given in (3.33).
sin 0
crack speed = C, (3.33)
sin %
where C, is the speed of the stress pulse, @ is the angle of intersection between
the crack front at the stress pulse, and « is the angle between the direction of
propagation of the stress pulse from A and the Wallner line on the fracture
surface.
(d) Mirror zone. Rib marks and Wallner lines are mainly visible in a smooth
area of fracture emanating from the fracture origin, known as the mirror
region. The mirror region is perpendicular to the applied tensile stress and
extends on an arc from the origin of fracture to a rougher zone called the
“mist”. An empirical relationship exists between the area of the mirror surface
and the fracture stress:
O71, = constant. (3.34)

In (3.34), r,, is the radius of the mirror area, Fig. 3.13, and the constant
varies from one material to another; for soda—lime-silica glass it is
4x 10? GN? m~?, whilst for alumina (5.4) it is 1.5 x 10° GN? m73.
Although the origin of the relationship in (3.34) is obscure, it is used to
solve a number of problems. In particular, since the relationship has been
shown in studies with alumina to involve the local stress at the fracture origin,
it can be used to determine the compressive stresses present in quenched
surfaces and the stress profile across thermally tempered ceramics (3.4.8).
Studies of the mirror zone provide proof that in glass the Griffith flaws
 165

Fracture origin

Rib- marks

Figure 3.13 Schematic representation of surface features on tensile fracture surface of glass and
other brittle solids.

occur in the surface only, whilst in crystalline ceramics they can occur
throughout the volume.
(e) Mist and hackle. The mist zone is an arc of rougher fracture surface
separating the mirror from the region of heavy striations in the hackle zone,
as in Fig. 3.13. It seems most likely that mist and hackle zones denote the
onset of crack branching as the strain energy release rate (3.4.4), which
increases with crack length, becomes a limiting feature. Congleton and
Petch’s equation for crack length C, at crack branching has a form
remarkably similar to (3.34), if E and y are assumed constant over the
temperature range usually encountered in these investigations. The crack
branching equation is given as

a ,Chl? = 2(Ey)""" (3.35)

7

3.4.4 Fracture mechanics

Griffith’s energy approach to estimating strength degradation in glass was the
beginning of what is now alarge subject, fracture mechanics. This is the most
successful approach to explaining why materials fail at stresses smaller than
more
the theoretical, and deals with flaws in materials. As materials become
complex (in a microstructural sense) than glass, then a variety of flaws must
be considered; this aspect is dealt with in 5.4.5, on polycrystalline alumina as
a typical material, and in 2.13.4, on toughness of brittle materials.
In the form of equation (3.31), the relation holds for purely brittle solids,
is
with glass as an archetype, where the energy needed to create new surfaces
is
solely that needed to rupture bonds; hence the surface energy term in (3.31)
the thermodynamic surface energy.
166 CERAMIC SCIENCE

A natural advance has been made by questioning the nature of the surface
energy term when a crack propagates through a more tough material.1> The
energy absorbing processes are now amalgamated into a single term called
the fracture energy. Because of the extra dissipation terms, we can only say
that the energy used in crack propagation is proportional to, not equal to, the
new crack area, and

6) =G,5A (3.36)
where dy is the fracture energy and contains all the terms necessary to create
the fracture surface, 6A is the new crack area and G, is the specific fracture
energy, which is often called the strain energy release rate. G, may be viewed as
a composite of several effects, (G.)g +(G.)ke+(G.)p+(Ge)vc, where B is
brittle behaviour giving rise to thermodynamic surface energy, KE is kinetic
energy, P is plastic energy, and VC is viscoelastic energy, and as a result the
fracture energy is a composite term: yg+yxe+ypty)vc. If the composite
expression for surface energy is used in (3.31), then the equation applies to any
homogeneous material tested in tension in the presence ofa flaw.
Further development of this relationship is based on G. R. Irwin’s
observation that the stress field in the vicinity of a crack tip can be adequately
defined, for purposes of considering crack extension, by a single intensity
factor K which is a function of applied load and pre-existing crack size. K
increases with load and when K(the stress at the crack tip) attains the critical
value K,, the crack will propagate abruptly.
K, is a material constant, since crack propagation always occurs at a given
value of local stress at the crack tip, and its value is defined as the fracture
toughness of the material. Fracture toughness is a measure of the resistance of
a material to failure by crack propagation. In plane strain conditions it is
denoted by K7¢ to indicate the crack opening mode of loading. Kc is related
to temperature, strain rate and mode of failure i.e. tensile or shear.
The strain energy release rate, G, is the way to relate Kyc to the Griffith .
brittle-fracture strength through Young’s modulus E, and Poisson’s ratio v:

_ K?(1-v?)
G for plane strain
ve E
and

K2
G= =zfor plane stress (3.37)

G is a measure of the strain energy release rate at the crack tip and is
sometimes called the crack extension force. G,, the value attained when
K = Kjc, is defined as the energy required to extend a crack of unit width by
one unit of length. G has been shown to be equal to zo*c/E for a large plate
under remotely applied stress, where 2c is the crack length, from which it can
 GLASS 167
C=B
W4=2B P

H=1.2B
W=1.25B W.

2c

7,

Figure 3.14 (a) Compact tension specimen with eccentrically loaded crack for determination of
Kc. (b) 3-point bend specimen for determination of K;c.

be seen that at constant stress, the strain energy release rate increases with
increasing crack length, with the result that the system is unstable.
Plane-strain fracture toughness K;c is a parameter of considerable
engineering significance and is measured by determining the deformation
(compliance) of a specimen at the instant a crack of known size in it starts to
propagate. An alternative is to measure G by determining the work needed to
extend a crack slowly through a known distance in the material. It is essential
that the crack be extremely sharp, otherwise G and Ky,c become
overestimates.
Figure 3.14a shows how a compact tension specimen is used and 3.14b
shows how a bend specimen is used. In both modes, the load v. displacement
curve is recorded. It must be nearly linear and the crack front must be
straight. Under these conditions the fracture toughness is calculated from
equations (3.38) and (3.39).

P
Ka Bye ) for Fig. 3.14a (3.38)

= waz 9(cl™) for Fig. 3.14b (3.39)

where Pis the load at fracture, W the specimen width, c the crack length, B the
specimen thickness, s the span length and f(c/w), g(c/w) are calibration
functions dependent on the given geometry. (ASTM standard test method
E399 gives precise definitions of the functions).
Kjc is used to elucidate microstructural effects on ceramic strength (5.4.5),
and to probe the effects of environmental parameters on glass strength (3.4.5).

3.4.5 Plastic flow in glass

Griffith’s relationship (3.31), containing the thermodynamic surface energy,
has been considered sufficient for understanding and calculating glass
168 CERAMIC SCIENCE

strength; the fracture toughness approach, introducing energy absorbing

modes other than new surface production, has been developed for other
brittle systems. During the 1960’s some of the anomalous physical behaviour
of glass was raised as evidence for the existence of plastic flow in the vicinity of
cracks in glass which might make a fracture toughness approach more
relevant. This group 2 type of theory, in which flaws are again taken to exist, 1s
mainly associated with Marsh.?! Evidence for plastic flow is in the main
visual.
(a) Two intersecting furrows formed by drawing a lightly loaded point across a glass plate result in
the first furrow being filled in.
(b) Curled shavings can be found.
(c) Photoelastic observations show that large residual stresses exist at the root of a crack in glass.
(d) Some experimental data does show that the energy absorbed in glass fracture is higher than the
surface energy required for the area of surface produced.
(e) Most crucial, it is possible to obtain Vickers pyramid diamond indentations in glass, and these
have sharp edges and show no cracks.

All of these points taken together present a prima-facie case for plastic
behaviour, but each can be, and has been, contested.
Working from the values for Vickers hardness in glass and using Tabor’s
relationship (3.40), a value for the plastic flow stress can be obtained:
VPH
— |3. (3.40)
y

Unfortunately values of around 37.7M Nm? are obtained, well below the
observed values for glass strength (Table 2.1), and therefore it would be
expected to show itself in any fracture surface. Since it does not, there is an
impasse which is attacked by examining the Tabor relationship, (3.40). Tabor
assumed that indentation is bya flat rigid die penetrating an elastic-plastic
material so that material is pushed towards the surface. If, however, the
material under the indenter is compressed radially and not pushed to the
surface, then the relationship in equation (3.41) applies:
VPN
=C+KBInZ (3.41)
Oy

where B and Z are appropriate functions of J and py:

(1—2v)o, fa ap lle yVPN
A= Sen raG ilarene ayes,

From a series of hardness tests on plastics, metals and alloys, Marsh

determined C to be 0.28 and K to be 0.60. Since typical hardness values
for soda and pyrex glass are 543 and 595kgmm~* respectively, then
o, = 67.4 and 79.7M Nm~* for these glasses, instead of 37.7M Nm? and
40.6 M Nm? for the rigid die calculation. These new values are close to the
measured breaking stresses and suggest why flow preceding fracture is not
obvious; however the values are much smaller, about 0.05E, than the
 GLASS 169

Figure 3.15 Subsurface damage in glass produced by a microhardness indentation. From J. T.

Hagan, J. Mater. Sci., 15, 1417 (1980).

theoretical breaking stresses around 0.1E (about 8G Nm ? from Table 2.1)

and should play an important role in fracture modelling and theoretical
strength calculations.
The terminal crack propagation velocity (equation 3.32) indicates that
when o, first reaches o,,,, flow will just disappear, and the crack velocity will
just reach the theoretical brittle fracture velocity. Then an extrapolated flow
stress at brittle fracture velocity is a measure of cohesive strength; this comes
to 0.34E.
Arguments about this theory centre on whether or not microindentations
in glass are caused by plastic flow or densification of the open-network
structure, or by viscous flow at localized temperature increases. Despite the
realities of these arguments it seems that some additional energy absorption
mechanism, other than new surface formation, can usefully be considered
when modelling the glass fracture process. Figure 3.15, taken from work by
Hagan,'° demonstrates that flow lines do occur beneath an indenter. These
lines are shear faults of negligible thickness, produced by genuine shear
displacements made apparent at the intersection points of some flow lines. It
does seem that voids or cracks can be formed at the intersection points of, or
along such flow lines, which can then become sources for brittle failure.

3.4.6 Group 3: flaw generation hypotheses

The evolution of this class of glass strength theory is due to workers such as
Cox? who felt the need to give a more conceptual meaning to the hypothetical
Griffiths flaws, and to explain the time-dependence features of glass strength.
More about this last feature will be found in 3.4.7. It is a satisfying approach
for materials scientists to adopt, since it begins where one feels most “at
170 CERAMIC SCIENCE

home”, with an accepted model for the structure of the material. In this case
the model chosen is that of 3.2.1. It is recognized that when dealing with glass
fracture, one is concerned with a specific and local event which develops as a
chain reaction into a catastrophic failure; thus the theory proposes three
factors working together: (1) an external tensile stress, (2) local strain round
migrating Na‘, and (3) chance local high concentrations of Na*. The result
is, at present, the most comprehensive quantitative theory of glass strength
yet proposed. In outline, the arguments leading to the development of
formulae are as follows.
(a) The Gaussian distribution of temperature, allied to the variable Si-O
bond lengths in the Zachariasen model, cause local bond rupture to occur.
Such an event is termed an abortive bond rupture, because the Si and O
atoms do not move away and therefore reform, on a time scale of about
10-5
(b) An applied stress gives a direction, for reference and abortive bonds
parallel to the stress have a greater probability of remaining parted.
(c) In the glass each Na™ has a “natural” site formed at T, (3.1.1); however,
there will always be some Na* in “interstitial” sites relative to these, as the
configurational entropy of the solid is increased, thereby lowering the free
energy in accord with equation (5.5). The interstitial Na* ions migrate,
occasionally finding a “normal” site, which they occupy, releasing their excess
energy, termed strain energy, into surrounding Si-O bonds. The number of
bonds affected is seen to be localized by considering the velocity of
propagation of elastic waves in glass; this is about 5 x 10?>ms_', so a wave
travels about 5 x 10~° m in the lifetime of an abortive event, 10° 1” s; this is
about 25ion diameters, and it is not a serious error to confine attention to a
static distribution of the stress around migrating Na* ions.
(d) From the static situation the problem can be considered in terms of a
square grid with each side representing a Si-O bond. On removing n bonds,
the stress distribution will be the same as it would for an elastic body
containing an elliptical void of area formed by circumscribing the missing:
bonds. Thus local stress on Si-O bonds at the periphery of the ellipse where n
bonds fail is (1+ 2n)o, for a thin sheet, or

“[1+2(n/n)!"
Jo,
for a bulk specimen.
(e) The probability of a bond being in the abortive state is given by a
Boltzmann function, and the probability of n bonds being in this state
together is the product of such functions (3.42):
N' (a D—aS)
q = =M exp— kT (3.42)

where N’=concentration of dissociated Na*t, M =concentration of

 GLASS 171
“interstitial” sites, which is approximately equal to the number of Si-O
bonds, « = bond energy of an Si-O bond, D = energy required to dissociate a
Na’* from its site, a = work done per unit applied stress when a bond length is
doubled, S = local stress, k = Boltzmann constant, T = temperature, and
q = probability of a bond breaking abortively.
(f) The expression for q gives small values and the product of such
expressions will be smaller. However, the probability of n abortive events
occurring together does not decrease indefinitely as the peripheral bonds are
weakened; rather, after reaching a minimum, it increases indefinitely. This
results from chance non-uniform distributions of Na* giving up binding
energy simultaneously, plus the apparent static situation. Therefore there is a
value P = 1, and breakage will occur as the general relaxation in the area of
the simultaneous events causes the abortive Si-O breakages to become
permanent. In this way a flaw is generated in the glass.
(g) Generation of flaws will be more likely in the surface, as some bonds
there could well already be broken as a result of macroscopic damage which
might account for the usual observation of failure from the surface of glass.
(h) Thermal energy is redeployed in every vibration, which implies that
about 10’? searches per second are made for these chance agglomerations of
Na*. It is worth noting here that other models of glass structure (e.g. 3.2.3)
already postulate such agglomerations. Mobile Na* ions move every
5x 10~°s and so there are t/(5 x 10°) sodium ion arrangements in a test of
t seconds. Hence the probability of breaking the glass equals the probability
of a flaw containing n broken bonds at one site times the number of Na*
rearrangements. Defining the tensile stress as that at which P =4 where
n

P=[|4@
ue hen Bis cd (3.43)
oF Ila =
Here n = number of broken bonds in a developing flaw, no = number of
broken bonds in a pre-existing flaw, t = duration of the test, and t = dwell
time for a mobile ion. The result of applying this logic to the Zachariasen
model is expression (3.44) for the theoretical strength of pristine glass, and
(3.45) for the practical strength of bulk glass:

one Saal) (3.44)

Pape OC TN
: a Ye 3Pt

Gane
pract
G—log
x
QN'/M)
Apot
(3.45)
2H ny log Pr
where the symbols not as yet defined are:
G = («—D)/KT, L = log (4nn’)'/?N'/M, H = «/KT, V = volume of the test
 2: CERAMIC SCIENCE

sample, t = dwell time of Na* which is 1/fexp(J/KT), P = probability of

sample breaking, no = mean severity of pre-existing flaws, A = area under
test, p) = original surface density of pre-existing flaws, J = activation energy
for a mobile ion to jump, and f = vibrational frequency.
Substitution of known data into equation (3.44) gives a value for the
theoretical strength ,, of the same order of magnitude as all other approaches
used in group 1, which is some justification for the method. More convincing
evidence for the Cox model comes from considering the effects of the various
parameters in the model.
Time. The test duration time is in the denominator of (3.44) and (3.45), hence
glass should be stronger in short-term tests. However, the variation will show
in orders of magnitude since t occurs in a log term.
From many sources’! an empirical formula relating glass strength to
time is
a, = B’/(A' +logt)
which equals (3.47) if

B’ = G—log(2N’/M)/2Hn, and A’ = log Po.

Thus empirical results of many years’ standing can be given theoretical status.
Size. In equation (3.45) sample area is again in the denominator, and so
strength should rise as the sample size decreases, as observed when fibre
strengths are compared to bulk strengths.
Environment. Moisture, for example, can be seen to have two effects in the
above expressions for strength. First, (« —D) is reduced by about 50% as the
surface is covered in water and the ionic bond strength is reduced. Secondly,
J is lowered and so more sodium ion arrangements per unit of time will
occur, giving an increased probability of finding a self-generating flaw. This
latter effect implies a possible delayed fracture in wet specimens, i.e. a fatigue
effect, and it does this without presupposing any preferred chemical attack at:
flaws, but rather a general decrease in the ionic bond energy (5.3).
Temperature. This effect can be investigated by plotting calculated strength
against log(Ap,t/Pt) keeping everything constant except for temperature
and making ny = 0, 1, 10 and 100, as in Fig. 3.16. A noticeable feature is the
sigmoid shape of the curve for real glasses containing flaws; such a sigmoid
shape has been reported in more recent strength-temperature studies.

3.4.7 Crack propagation and static fatigue’

Static fatigue in glass is well known:® strength deteriorates with time under a
sustained load. Although we have to some extent accounted for this in 3.4.6
when discussing the flaw genesis theory of glass strength, a complete
quantitative explanation which takes into account all the observed phen-
omena is not available.
 GLASS 173

Tensile
Strength 8
GNm2

200 400 600 800

°
Temperature “K

Figure 3.16 Strength—temperature data for two glasses.

(a) Experimental observations. Early workers demonstrated the sensitivity of

fracture stress to load duration, particularly in humid conditions: conversely,
early work also showed the absence of such an effect for baked-out specimens
tested in a vacuum. The effect was also absent for specimens tested at the
temperature of liquid nitrogen. Time to failure was shown to be sensitive to
temperature within the range —190—520°C, and for soda-lime glasses,
delayed failure was most apparent in saturated water vapour at 150°C.
Because no static fatigue effects are observed at the temperature of liquid
nitrogen, it is possible to express the experimental results in terms of the
reduced parameters, o/oy and log (t;/to.5), where oy is the strength at liquid
nitrogen temperatures and ft). is the time when the condition o,/ay =4
obtains. When this is done, a universal fatigue curve results (Fig. 3.17).
From the studies needed to establish universal fatigue curves, the major
factors affecting fatigue sensitivity were found to be:
(i) humidity
(ii) composition of the atmosphere at the surface 17,23

(iii) the extent of ageing prior to loading

(iv) temperature
(v) the degree of surface abrasion.

0-8

Bees
on

0:2

-4 0 6
log tA
0-5

Figure 3.17 Universal fatigue curve for abraded soda-lime silica glass.
174 CERAMIC SCIENCE

Perhaps unexpectedly, pH is missing from this list, but it has been found that
strength is invariant to pH in the range |to 13.
(b) Interpretation of static fatigue. At the moment none of the proposed
theories gives a complete description of all the observed static fatigue
phenomena, and so the outlines of several are given here.
Perhaps the earliest of the semi-empirical approaches is the expression

logt, = at+bloga, (3.46)

where t, is the time to fracture and @, is the critical macroscopic breaking

stress. This expression was introduced to give an empirical expression which
encompassed most of the measured data. Several others have been proposed;
for example
b
logt, = Stearee Ge (3.47)

However, none give a model to work from, neither do they successfully

explain all the data.
The names best associated with the essentially chemical stress corrosion
theories are Charles and Hillig, who independently proposed that corrosion
mechanisms occurring at the tips of micro- or macro-cracks were responsible
for the phenomena. Their names are jointly associated with perhaps the most
widely known theory for static fatigue.
The model assumes that hydrogen ions from absorbed water effect an ion
exchange with glass modifier cations; the driving force for this is discussed in
5.3. Two effects are immediately apparent from equation (3.48):

SiONa* +H* > SiOH* +Na’*. (3.48)

First, stresses will be generated within the glass surface by the volume
change which will act with the applied stress on surface cracks. Second, the
hydrolysis produces an alkaline solution which will attack and weaken.
bridging Si-O bonds. The basic assumption of the Charles—Hillig model for
slow crack growth and static fatigue is that the activation energy for the
corrosion of the surface by the alkaline solution is a function of tensile stress
acting on the surface:
OE
Ea) = Eo ae Sapp (=) (3.49)

where E, is the activation energy in the absence of a net stress and o is the
effective stress at the point in question. At a surface flaw, one more term is
added to take into account the fact that surface energy of a solid is a function
of its radius of curvature. Finally in (3.49) the applied stress, o,,,, must be
modified by introduction of the Inglis expression, equation (3.29), to obtain
the flaw tip stress.
 GLASS Wes
The velocity associated with such a corrosion process may then be written
in the form of (3.50):

y= Kexp—|Eotyvqinte,(1+ [2)(2) [xr (3.50)

Po) NO Galina
where V is the corrosion rate normal to the surface, V,, = molar volume,
y = surface free energy, r = flaw tip radius, c = flaw length and o, = applied
stress.
For areas of surface with very small r, the second term in the exponential
of (3.50) may control the rate; if no applied stress is present this will be true,
with the result that existing cracks will blunt. This is an interesting prediction
since it gives some idea as to why pre-ageing can favourably affect static
fatigue. In the presence of stress, there will be a critical stress at which the
third term becomes more important than the second, and preferential
corrosion at the flaw tip will then cause it to grow with resultant strength
degradation. Because the basis for this theory is chemical, the process will be
an activated one and hence variations in fatigue behaviour must be expected
at different temperatures in accord with observations.
To make time v. strength predictions, equation (3.50) can be integrated for
a particular stress. This gives reasonable predictions, together with an
activation volume similar to that associated with alkali ion diffusion in
soda-lime glass.
Crack velocity v. stress intensity factor curves have the form shown in Fig.
3.18, where region I is considered to be limited by the rate of reaction of H,O
at the crack tip; II results from control passing to transport of H,O to the
crack tip; while III where very rapid crack growth occurs is difficult to
explain.
Thus the prevalent view arising from the success in interpretation of
experimental data by corrosion models is that static fatigue may be identified
as an activated process!* dependent primarily on atmospheric water content.
Unfortunately, this does not explain slow crack growth which occurs in
a vacuum.°**
10

Crack
velocity
ms!

4-0 6-0 8-0

“3 5
Ky Nm2 x10

Figure 3.18 Curve of crack velocity as a function of stress intensity.

176 CERAMIC SCIENCE

Finally, we consider lattice trapping theories. The failure of stress-corrosion

mechanisms to explain recent experimental observations of slow crack
growth in vacuum implies that another process must be acting besides the
stress-corrosion mechanism; such a process is probably intrinsic to glass.
Progress in elucidating the intrinsic process is made by an atomistic
approach to the problem?* combined with fracture mechanics (3.4.4). This
approach begins with the energy balance concept of Griffith (3.4.2), but
instead of the surface energy term being directly proportional to crack area
alone, the expression is as shown in (3.51).

where W,, is any external work done, Uz, the stored elastic energy and Us
the surface energy term.
Us is not linearly proportional to crack area but contains a periodically
varying part introduced to account for the stepwise breaking of chemical
bonds at the crack tip:

SAS Pa alec271A (3:52)

Tl a

Equation (3.52) is one such attempt to express this state; a = area of bonds
broken, ) = thermodynamic surface energy, y, is a trapping term, andAis the
area of fresh fracture surface arising from crack growth. From (3.52) the
fracture surface energy that must be supplied to produce crack extension is
dUs cos (277A)
da = 27 = 2| Ey—y,
——= anos |. (3.53)
:

Returning to (3.51), we can re-write

U = —(stored energy) + (fracture surface energy)
and
dU
ae (energy release rate) + 2I
a
that is
dU
— = —G42I.
da
which from equation (3.53) is:
dU 2nA
—
Be = —G+2]y-
st > vAcos(——]
. )|. (3.54)

Following all the earlier ideas, for mechanical equilibrium of the crack,
dU/da = 0. Hence it will become “chemical bond trapped” within the range

y=)
= G.S2(-+y,)
 GLASS a,
and if the strain energy release rate is less than (y +y,) or greater than () — Veds
then the crack will not propagate. Section (3.4.3) contains the relationship
between strain energy release rate, G, and the stress intensity K;.
In the “lattice” or “chemical bond trapped” range the activation energies
for forward or backward motion of the crack-tip as bonds are broken or
healed will vary, but the activation energy for forward motion, further bond
breaking, will be at a minimum at the upper end of the trapping range and
forward motion will occur. We can write for the velocity of crack motion:

(3:55)
Assuming that E, = E)(1—G/G,) describes the variation of activation energy
in the trapping region, then
V, = Voexp(—E,(1—G/G,)/RT ). (3.56)
Here Vy, Ey and G, are constants. Using the expression G = K?/E from 3.4,
equation (3.56) can be re-expressed as:

In(V/Vo) = —[E,(1—K*/K?)]/RT (3.57)

where K, is the stress intensity at which E, = 0. Equation (3.57) gives the
condition for spontaneous crack propagation in a vacuum, and leads to an
expected crack growth such that the logarithm of the forward velocity is
directly proportional to the square of the stress intensity for any constant
temperature. This has been confirmed experimentally.

Table 3.4 Summary of glass strength theories

Theory Concept leading to failure Suggested control

(Group 2) Griffith Microcracks in the surface. (a) Compress cracks

Instantaneous break: flaw (b) Limit crack length
length important (c) Remove surface
(Group 2) Inglis, Griffith Microcracks in the surface: (a) Compress cracks
Jer important (b) Blunt cracks
(c) Remove surface
(Group 2) Charles Microcracks in the surface: (a) Compress cracks
stress corrosion at crack tip. (b) Corrode in absence of stress
to blunt cracks
(c) Remove surface
(d) Keep surface dry and
chemically clean
(Group 2) Marsh Microcracks exist but plastic (a) Compress cracks
flow occurs at crack tip. (b) Increase density in regions
be
Crack velocity K,/o,/p. of crackurn
ish Renie

(Group 3) Cox Na? diffusion leads to flaws; (a) Restrict diffusion process
this effect enhanced by surface (b) Control size
flaws. Size important, time and (c) Remove surface
temperature dependent. (d) Compress to limit
micro-flaw effect
178 CERAMIC SCIENCE

3.4.8 Methods of glass strengthening

Before outlining the processes used it is a worthwhile exercise to summarize
the features that each of the glass strength theories so far outlined emphasizes
as controlling the strength. This is done in Table 3.4.
In Table 3.4, the most commonly suggested control feature is removal of
the surface containing the micro-cracks. This is easily achieved by etching in
HF, but cannot be regarded as a practical way of glass strengthening because
the strength improvement does not last very long.
The second line of attack suggested by all theories is to compress the
surface, and this has been and will continue to be the practical route to glass
strengthening. There are two classes of surface-compression strengthening:
thermal and chemical.
Thermal strengthening —tempering. Glass strengthened by thermal quenching
is common. It is produced by rapid cooling from above T, by directing cold
air jets on to the surface. From 3.1.1 it will be seen that this results in
T, (surface) > T, (inside), and so volume of outside > volume inside. Hence
the outer layers are in compression while the centre is in tension. This process
is (unfortunately, when compared to metallurgical use) known as tempering.
It is so well understood that computer modelling will give good answers for
almost any question regarding flat plates. Practical limits of the compressions
attainable by this method are set by the steepness of the temperature gradient
that can be obtained, which is in turn determined by the magnitude of the
heat transfer coefficient that can be achieved. Typical values of surface
compression are around 140MNm ’, leading to mean strengths of
210M Nm ”, and ratios of surface compression to centre tension around 2.2.
Figure 3.19 shows how the thermal stress profile is developed during a
quench. For thermally toughened glasses, the stress distribution throughout
the thickness is parabolic and the presence of the internal tensile stress field
can cause problems.” One particular problem is associated with the presence
of inclusions in the glass from the manufacture. Nickel sulphide, NiS,
can occur in soda-lime glass, generally as a spherical-shaped inclusion,
and usually in the high-temperature hexagonal «-form. The stable low-

Quench Time t[sec]

20 16 109876543210

Sheet
0-61 cm

; Stress ,
Tension »—+—— Compression
400 0 400

Figure 3.19 Stress distribution through the thickness of a glass plate during quenching.'*
 GLASS 179
temperature phase is B-NiS, which has a cell volume of 2.805 nm? compared
to 2.699nm°* for the «-form. The resultant 3.9° volume increase, as the
inclusion transforms with time, can produce microcracks around the peri-
phery of the NiS that propagate under the influence of the centre tension and
cause spontaneous failure of the toughened glass.?° This should be compared
to zirconia toughening (5.5.6).
Chemical strengthening. There are two kinds of chemical strengthening: one
based on increasing the specific volume of the surface and in so doing
producing a surface compression; the other relying on producing a decrease
in the thermal expansion coefficient of the surface which, if the process is done
at elevated temperature, generates a surface compression on cooling. The
second kind has been practised longest, can be done in a wide variety of ways,
but is often of limited value as the compressed layer produced is very thin
indeed. Representative schemes for this route to strengthening are:

(a) Flame polishing of glass when the hot surface evaporates Na,O, leaving a surface glass
less rich in alkali and so having a lower expansion coefficient.
(b) Treatment of glass articles in wet SO, to produce soluble salts that are leached to cause
depletion of alkali.
Na,0+SO,+H,0+0,2"S Na,SO, solution.
(c) Treatment at elevated temperatures with stannous chloride, producing a soluble sodium
stannate that can be leached.
(d) Selective recrystallization of the surface to produce crystals of low expansion coefficient
with a refractive index close to that of the parent glass. Lithium aluminosilicate glasses
are particularly useful since they crystallize to a low expansion f-eucryptite phase (4.6).

Chemical strengthening by producing an increase in specific volume at

temperatures below T, can be very effective. It is commonly achieved by
exchanging small ions for large ions, for example, by immersing soda-lime
glass in molten KNO, at about 400°C. Clearly, from 5.3, Coulombic forces
oppose this exchange, but configurational entropy (4.4) produces a small
driving force. Only a 20m layer of compression is produced after several
hours.
Two approaches to speed up the exchange process have been developed.
First, control of base composition: it has been found that addition of Al,O,
and Li,O to the soda-lime system produces a glass that can be strengthened
rapidly in a NaNO, bath by Li* exchange. These special glasses are used
commercially in supersonic aircraft and in toughened spectacle lenses. The
second line of development utilizes electric fields to promote diffusion into the
glass surface to aid the ion-stuffing process. The main weakness here is that
some area of the object must remain outside the salt melt to maintain
insulation between anode and cathode.
It must be borne in mind that surface compression achieved by ion
exchange may be much reduced if the glass is later used at high temperatures
and the foreign ions, at high concentration in the surface, diffuse down the
concentration gradient. This effect can be limited by selecting a system in
180 CERAMIC SCIENCE

Figure 3.20 Vickers indentation fracture pattern in tempered glass surface.

which the activation energy for cation diffusion is high, either by using large
cations such as K*, or by adjusting the base-glass composition.
When established, the surface compression stress must often be evaluated.
Traditionally, this is done by optical birefringence techniques, but these have
limitations, arising mainly from the manufacturing inhomogeneities in glass
plate which confine measurements to a small portion near to the centre plane
of the plate. In the method only the centre tension is being measured, and the
outer compression must be inferred from the parabolic profile (Fig. 3.19); it
is extrapolated as twice the centre tension.
An alternative way to find the outer compression is to combine micro-
hardness indentation?” with a fracture-mechanics analysis (3.4.4). When the
Vickers diamond in Fig. 3.20 causes the surface to fracture, cracks are
initiated from the central deformed zone, developing as half-penny shapes
along indentation diagonals. The indentation field drives the cracks and the
surface compression field opposes them, hence the stress intensity factor for
the cracks can be written as

k= a 2Baglc/n)! (3.58)

A is acontact constant, P is the indenter load, c is crack length; in the second

term which represents the compression resistance, B is another experimental
constant and cp is the residual compressive stress in the surface layer.
During growth, the crack remains in stable equilibrium if K = K, =
constant. Hence
K
(P/c?!?) = —*(1+2Boge"?/n"!?K,) (3.59)
Since K,/A is a constant, equation (3.59) shows that og can be found from
the slope of a plot of Pc” */? against c'/? when preliminary experiments are
performed on untempered glass-sheet to determine A, K, and B.
Other methods of strengthening. Table 3.4 suggests that strengthening could be
achieved by increasing the glass density; so far this does not appear to be a
practical route.
Cox’s theory suggests that limitations on diffusion processes may be a
possibility, but again no practical applications have been described. However,
 GLASS 181

1: Impact striking section Wi

2: Transition section
3: Impact absorbing section

Figure 3.21 Section view of bullet-resistant glass.

arising from the Cox equation (3.45), size is an important feature in achieving
higher strengths. This has been realized by using glass in the form of fibres in
a composite (2.13), but it is interesting to note here that the strongest glass,
bullet-proof glass, is a composite on the grand scale (Fig. 3.21).

3.4.9 Load-duration analysis for glass design

The analysis of glass strength so far should have convinced us that we cannot
tabulate the stress at which a glass fails for design purposes because:
(a) stress-raising flaws are dominant
(b) strength changes with time of loading
(c) strength is temperature-dependent
(d) relative humidity affects strength.

Cox (3.4) has shown that an equation can be derived which takes into
consideration all the above effects, e.g. (3.45), but it is not a practical equation
in the sense that glass-sheet design parameters can be calculated from it. The
question is whether the four observations (a) to (d) above can be pragmatic-
ally built into a practical formula for glass design. It is worthwhile
considering one approach which demonstrates how scientific development
can be interfaced with engineering need.
Step 1: acceptance of a flawed surface containing a wide range of flaw sizes,
one of which can be critical. Then the stress concentration by the critical flaw
will be given by an Inglis-type expression:
aN te
O, = 26, (<) (3.60)

Step 2: acceptance of the fact that flaw geometry changes with time due to
chemical corrosion, and that the corrosion rate also depends on the applied
stress:

A(cir)
dt _ p om.
= B°RHA T exo (— A/RT)
exp(— (3.61)
182 CERAMIC SCIENCE

Table 3.5 Measured and predicted failure rates for soda-silica glass plates

Loading rate Measured failure stress Predicted failure stress

(B)M Nm? (MNm”) (B17) M Nm?

6.895 1.66 1.68

68.948 1.86 1.92
5515.81 2.43 2.49

where B is a constant, RH is the relative humidity, n, A are unknown

constants and T is the absolute temperature.
Step 3: substitute in (3.61) for o,, from (3.60), and integrate from zero time up
to time to failure, T,.

i RH pr (Sah dt stant
, ERT
2 -ex —— = con nN

which becomes
RH -exp(—A/RT) exp. = constant. (3.62)
(n+1)T" B
Here ox, is the experimental measured breaking stress, and f is the rate of
stress increase. Equation (3.62) makes the statement that the cumulative
effect of time-dependent stress o,, applied to a specimen until it fails at any
time T,, is constant. It also indicates that the failure stress decreases with
lower rates of stress loading B. For Na,O-SiO, glass plates at constant rate
of stress application, tests by Charles gave n = 16, and so experimental
breaking stresses should be approaching f*/'’. This prediction can be tested
at this stage to see if the method is worth pursuing further. Table 3.5 shows
that data from the literature correlates well with the f1/'’ prediction.
Therefore Step 4 can be taken.

100

P; =cumulative failure
P probability
fo
% which fail ! (a)
| o, o,
|
|

|
Oo *

/ %%

Figure 3.22 Cumulative failure probability curve for plate glass. (a) Probability of withstanding
o, = (1—P,). (b) Probability of withstanding o, = (1—P,,)’.
 GLASS 183

Step 4: This is to recognize that size must be important since c and r will be
random, causing specimens to fail at different applied stress loadings. The
failures probability has the form shown in Fig. 3.22.
Figure 3.22a,b show that since the withstanding probability for Ay =
(1—Py,), and each half of 2A, has this probability, we can generalize to state
the withstanding probability of any area A as
Peel Py, (3.63)
From (3.63), we note that for constant failure probability P, the failure stress
decreases with increasing area; for example, for a failure probability of 40 %,
0.4 = (1—P,) and so P; must be 0.6, which is equivalent to o* in Fig. 3.22
that must not be exceeded to maintain this probability. Now, doubling the
plate area, 0.4 = (1 —P;,)’, from which P, must be 0.37, which has a lower
o, that must not be exceeded to obtain the same breaking probability.
Labouring the last point was necessary to bring home the truth that glass
areas of practical size (such as windows) are ~ 10° larger than laboratory
specimens, and so the mean breaking stresses will only be 1/4—1/5 of those
obtained from the samples used for laboratory tests. This can be expressed in
the general statement that tests on a limited number of small samples give no
useful information for large areas.
Step 5 is to put the size effect into the load duration equation (3.62), after first
introducing applied pressure and sheet dimensions. To do this, edge loading
is assumed for a plate thickness h and width a under uniform pressure p, in
which case, for a limited range, equation (3.64) applies:
G, A q s/n

Eta/hy lao aoe

where Y and S are constants.
Adopting a Weibull distribution to describe the form of the measured
failure stress,
Py = deo K(A/Ap)oexp]. (3.65)
The result of substitution of (3.64) into (3.62) is now substituted into (3.65):

C1A/Ao): \e"ayhy
Pp, =1—exp |

feRH/T" exp (A/RTasdt)}B/n+ | (3.66)

t

Here C is a combined constant and all other terms are defined. This last
equation, (3.66), is the desired working equation if the several constants can
be determined. Experiments on large glass plates from 3-30 m? have been
carried out to determine the constants, and s = 12.3, B = 7.3.
are
Equation (3.66) can be used, then, on design problems; some examples
184 CERAMIC SCIENCE ~-

Table 3.6 Plate thickness calculated from equation (3.66)

to withstand a uniform load of 1.4MNm ? for a failure
probability of 1 in 1000.

Size(m) 0.3x0.3 B35

60s 3 years 60s 3 years
at Zire. at 2ir€ atoleCmuatlce’
TADS ico a 28.2 44.2
(mm)

given in Table 3.6 for a failure probability of 1 in 1000. The load of

1.4MNm ?” for 60s is equivalent to a wind load of 100 mph and applied for
3 years it approximates to a cumulative snow load in 30 years. The size and
time effects are clearly demonstrated.

3.5 Glass/metal and glass/ceramic bonding

This subject is of great importance, because structural and electronic devices
frequently require the combined use of ceramics and metals, and in many of
these applications vacuum-tight seals are required. There are also the large
industries of lamp manufacture, porcelain enamelling and ceramic-ware
glazing, all dependent on the proper production of adherent interfaces
between metal and glass or oxide ceramic. Indeed, the technological need for
such combinations has been so great that development has occurred steadily
despite incomplete understanding of the structural, chemical or mechanical
nature of glass-to-metal interfaces. The empirical developments in these crafts
have been successful, but have left many complicated, and occasionally
contradictory, observations for scientists to understand and reconcile. Several
of the more consistent observations have provided starting points for
theoretical developments. This section first discusses some of the procedures,
both old and new, that are employed, and then explores how adherence
theories are developed from them.

3.5.1 Technical procedures

(a) Glass-metal seals. The metals commonly used in making such seals are Pt,
W, Mo, Cu, Fe, Ni, and Fe alloys. Specific sealing glasses are needed for each
metal and the pair has to be selected on the basis of specific properties and
requirements.
The most important physical requirement is a near match of thermal
expansion coefficients, that is within 10%, with the metal having the larger
coefficient. This matching assumes a greater importance as the bulk of glass
increases, but does not appear to be critical in the formation of thin glass
coatings such as enamels.
Initially, in the formation of the seal, the glass is molten on or around the
 GLASS 185

ae)
Elongation on
any a normal Ty

"002 delayed Ty produces

larger de
-001

200 400 $00

Temperature cc

Figure 3.23 Relative thermal expansion coefficients for tungsten and glass.

metal and therefore does not support any stress. As the glass cools, strain
occurs, but the expansion coefficients chosen to suit the requirements above
cause compression of the glass parallel and tangential to the interface, which
3.4.8 shows is a satisfactory condition to strengthen the glass. During the
development of the compression in the glass, the critical factor is the amount
of strain developed on cooling through the fictive temperature T, (3.1.1).
Because T, varies according to the heat treatment of the glass, Fig. 3.23 shows
that the amount of strain developed in a seal depends upon the skill of the
glass worker.
The critical point after the seal is made is cooling below T,, because the
composite must be cooled slowly, particularly when large thicknesses of glass
are involved.
Following the choice of metal and glass pair the metal must be thoroughly
cleaned to remove all materials from the surface that may not be soluble in
glass at the sealing temperature. These would prevent “wetting”, 1.e. intimate
contact of the glassy-liquid and the metal. This cleaning is both mechanical
and chemical and may achieve some surface roughening, believed by some to
be vital for good bonding.
The third stage seems paradoxical following the cleaning process, for it
involves “dirtying” the surface by preoxidation if the amount of oxidation
that will occur during the sealing process is felt to be insufficient. In some
respects this aspect is subjective and judged by appearance, but can be

‘03
Weight gain

kg m2
-02

V.good adherence
-01 vacuum tight seal

formation.
Figure 3.24 Degree of pre-oxidation of kovar and subsequent seal
186 CERAMIC SCIENCE

supported by experiment as demonstrated in Fig. 3.24. This shows that the

amount of oxide, irrespective of the time and temperature of its formation, is
the main factor in developing good adherence.
The final stage is when either bulk glass, powdered frit, or a thin glass coat
are heated and applied to the metal at the sealing temperature until most, if
not all, of the oxide on the surface appears to dissolve in the glass.
(b) Porcelain enamels. The basic technology here differs interestingly from
glass metal sealing because adherence is not dependent upon a proper
pre-oxidation of the metal surface, but on adherence-promoting oxides added
to the glass composition. Titania, TiO, (5.8), or cobalt oxide, CoO, are
commonly used, with the bonus that TiO, also acts as an opacifier. This is
largely because steel is the metal commonly enamelled, and heating above
910°C causes the ay transition with subsequent warping and therefore is to
be avoided. Furthermore, the steel is easily oxidized and so short firing
periods are necessary. Hence special low melting glasses must be used, to
which oxides must be added to obtain the correct conditions at the steel-glass
interface.
First the metal surface is electropolished to remove the surface layer. An
enamel frit is then prepared whose composition depends on the end use of the
finished product. Typical compositions include members of the borosilicate
system (3.1), with suitable high concentrations of soda to produce a low-
melting glass, and titania or other suitable colouring agent. After fusion and
mixing, the bulk glass is pulverized to a fine powder from which any solid
iron is removed magnetically. The frit is then spread on to the prepared iron
or steel and heated. Before the frit melts, it allows the passage of air which
oxidizes the underlying metal. At this stage, the weight increase due to
oxidation is maintained in the range 0.005kgm~? to 0.020kgm ?. The frit is
now melted, and because of the oxide now present, the glass wets the metal;
oxidation stops, and the newly-formed oxide layer is dissolved. During the
dissolution process strong glass-to-metal bonds are formed.
(c) Ceramic-to-metal bonds. A major problem in joining metals to ceramics is:
finding bonding materials that will wet ceramic surfaces well, and bonding is
usually achieved by applying a thin intermediate coating that will adhere to
the ceramic.’ If this transition layer is not readily wet by the normal bonding
metals listed above, additional coatings are applied until a suitable outer
layer is achieved.
Transition layers can be formed by:
(i) applying oxide, metal or metal-oxide mixtures by vacuum evaporation, painting, dipping,
spraying, flame spraying or electrodeposition, followed by firing in a suitable atmosphere;
(ii) partially or completely oxidizing a metal layer previously placed on the ceramic;
(iii) partially or completely reducing a ceramic oxide to the metallic state on a ceramic surface.

A common technique used to produce seals suitable for severe service is the
moly-manganese technique developed in Germany in World War II. The
initial step is the mixing of Mn and Mo powders of 5 um particle size and high
 GLASS 187

Table 3.7 Components of metal-ceramic systems

Ordinary use Severe use

Metals Cu, Ni No, Ni, W,

Stainless steel,
Inconel
Ceramics and Boron Al,O;, mullite,
glasses glasses steatite, forsterite
(1.5.1)
Transition silver-glass Mo-Mn plated
layer frit Cu or Cu/Zr

purity in a methyl methacrylate binder to hold the powder on the ceramic

surface as it is painted on. The painted ceramic is fired in a hydrogen
atmosphere at 1400°C for 5 to 20 minutes and furnace cooled to room
temperature. The component is then electroplated with copper or nickel so
that molten brazing metal in any subsequent operation will not dissolve away
too much of the transition layer which typically consists of Mo diffused in a
spinel, MnAl,O, (5.9) when alumina is used.
Tabie 3.7 lists some common materials used for metal-ceramic structures.
(d) Ceramic-metal reaction welding. Recently a series of patents has been filed
describing simpler ways of directly joining a wide range of ceramics and
glasses to many of the transition metals by a simple heating process.® In
contrast to brazing or welding, the metal must under no circumstances be
heated above its melting point. In fact, the system must be kept below the
melting point of the lowest-melting component. The ceramic is placed in
contact with a thin metal foil and heated in an induction or resistance-heated
furnace to below the melting point of the metal or glass, while applying a
small pressure of about 100kN m7”. This technique produces vacuum-
tight, extremely strong joins with no deformation, between, for example,
MgO/Pt/MgO, BeO/Pt/graphite, SiO,/Au/glass, Al,O,/Ti/Al,O; and
Cu/BeO/Cu. The bonding processes in this method are wholly confined to
solid state mechanisms and in many cases still need to be explored.
(e) Field-assisted bonding. This is another more recent development* that
allows glass and metal to be bonded at temperatures well below the softening
point of the glass by applying a dc voltage in the range 20-500v across the
substrate-glass couple so that the substrate metal is the anode and Pt in
contact with the glass is the cathode. The whole assembly is heated to 500°C
in vacuum, argon, air or nitrogen. It is an interesting feature of this method
that both the substrate and glass surfaces to be joined have to polished clean
and optically flat in order to achieve a good strong metal-glass seal.

3.5.2 Theories of adherence

Based on analysis of the technical steps outlined in 3.5.1, two general
on
categories of theory have been proposed —first, mechanical theories based
188 CERAMIC SCIENCE

glass
metal

Figure 3.25 Generalized glass—metal dovetail.

the development of a mechanical bond caused by the irregularities at the

interface, and secondly, chemical theories based on the development of a
chemical bond from glass to metal, or more commonly through an inter-
mediate layer.
(a) Mechanical theories. Figure 3.25 shows a generalized glass-metal dovetail.
The metal is viewed as keying the glass to its surface in a sort of dovetail
arrangement. The narrowest cross-section of the keys will be the controlling
area. The strength of the metal is reasonably easy to predict, but the strength
of the glass is unpredictable (3.4). However, if the glass at the interface
behaves like annealed glass in the bulk, then the maximum strength of a
mechanical bond would be about 34.5M Nm ? depending on whether the
conditions in the dovetails increase or decrease local flaws (this figure can
only be a rough guide).
Intuitively, adherence should correlate with surface roughness if the
mechanical theory is correct, and would be determined by the anchor point
density. Microscopic examination is however inconclusive, and leads to the
view that surface roughness does not account completely for adherence.
The preparation of the metals as described in 3.5.1 is often aimed at
producing smooth surfaces, and so some attention has been given to
describing how suitable roughness may arise during the joining process. Two
approaches have been proposed.
The dendrite theory is based on observations in some experiments of
dendrites, irregular metallic growths of iron and cobalt, emanating from the
glass towards the substrate metal. The dendrites were believed to originate
from reduction of the appropriate oxide in the glass composition either,
present originally, or dissolved by the glass from the substrate during the
process; certainly most of the methods practised involve production of
substrate oxide.

Co* + Fe® — Co? +Fe?t (3.67)

(glass) (surface) (dendrite)

Removal of surface iron in accord with (3.67) causes surface roughening for
mechanical bonding. Systematic tests, involving a standard frit containing
various added oxides fired on to polished metals in an argon atmosphere for
15 minutes, showed that metallic particles were produced whenever an oxide
was present in the glass which could be reduced by the substrate metal, i.e.
normal thermodynamic data apply (5.2). Many experiments showed that
adherence was almost non-existent in spite of the presence of a multitude of
 GLASS 189

dendrites. Thus metallic particles precipitated in an enamel can be considered

to be indicators of reactions which have taken place, and can be predicted
with reasonable accuracy from the thermodynamic stability of oxides and
metals, but their presence is neither indicative of, nor essential to, the
evelopment of good adherence.
The electrolytic theory derives from Dietzel, who in 1945 proposed a
echanism for substrate roughening based on electrolytic corrosion. For
xample, cobalt or nickel precipitate from the glass coming into contact with,
say, an iron substrate from a short-circuited local cell in which Fe is the
anode. The current flows from iron through the melt to cobalt and back to
the iron, but the cells are not exhausted because there is an abundance of iron,
and oxygen diffusing from the atmosphere depolarizes the cathode side.
Overall, this results in iron going into solution continuously, producing holes
into which the glass anchors. The overall mechanism is:

Fe® +CoO > FeO +Co° (establishes cell)

2Co® +0, > 2Co7* +207-
Co?*++2e7 —Co®
Fe® -—2e — Fe?’* (into glass)

and this is schematically shown as Fig. 3.26.

Three major objections prevent full acceptance of mechanical theories of
adherence. First, excellent adherence can be obtained on extremely smooth
metal surfaces even when no liquid phase has been present (3.5.1). Secondly,
the presence of a roughened interface does not ensure adherence under
conditions deviating more than slightly from those found to be necessary to
promote chemical adherence. Finally, when good adherence is achieved the
measured strengths of the seals would necessitate the assignment of extremely
high values for the strength of glass or the metal forming the seal, because the
total cross-section of anchor points is always a relatively small proportion of
the total interfacial area. It seems that surface roughness makes a contribu-
tion to seal strength of up to 25%, and is therefore beneficial, but not
essential.
(b) Chemical theories. These theories stem from the requirement for oxygen

ing in the presence of CoO.

Figure 3.26 Electrolytic surface roughening of iron on enamell
190 CERAMIC SCIENCE .

in making a seal, and from the identification of an oxide at the interface of

many seals which show good adherence.
A series of experiments, in argon or vacuum atmospheres, on carefully-
cleaned iron substrates using frits containing oxides more stable than FeO,
showed that no glass-metal seals could be produced. In similar tests with a
variety of metal substrates, as long as oxides that could be reduced were
absent, the glass never formed a bond, and in fact formed beads rather than
spread or wet the metal. From such experimental evidence it can be
concluded that available oxygen is essential for seal formation.°
When oxygen was introduced into these experimental systems, the glass
contact angle decreased, and wetting and then bonding occurred; similarly, if
the experiments were conducted in vacuum but a reducible oxide was present
in the glass composition, good sealing was achieved. In the case of enamelling
the important oxide was determined to be FeO, which must be present in the
glass. Electropolished Fe plus glass saturated with FeO (i.e. 43 % FeO) when
fired under argon produces an enamel with excellent adherence, but no
interfacial roughness. When the FeO content in this system was reduced
below 40 % then almost no adherence occurred.
From this it can be concluded that (1) the glass or enamel frit must wet the
metal, (2) the glass in the seal interface must be saturated with an oxide of the
substrate metal, and (3) the oxide should be one which, when in solution in
the glass, will not be reduced by the metal.
It is probably the second of these requirements that the various sealing
techniques are controlling in some roundabout way. Oxide saturation is
required to promote wetting, and to ensure that all available sites within the
glass structure are occupied: therefore additional M"* ions are forced into
sites where they are influenced both by the oxide ions in the glass and by the
metal atoms at the substrate surface, forming an intermediate layer linking
glass to metal (Fig. 3.27). Requirement (2) ensures that good adherence is
achieved with a standardized technique:

(a) The glass is saturated with the lowest oxide of the substrate metal.
(b) It is melted on to the metal, which is either polished or left rough, but must be clean of
oxide and grease; any remaining oxide would not be dissolved and would weaken the seal,
since many oxides (particularly iron oxides) are weakly bonded to the metal.
(c) The process is carried out under argon or vacuum.

An oxidizing atmosphere is required only when the bonding glass contains

non-reducible oxides. Too much oxygen can lead to oxide beyond the
saturation requirement.
(c) Role of adherence-promoting oxides and glass structure. If the standard
conditions just defined could be used to ensure adherence, then one might ask
what is the role of the adherence-promoting oxides, in whose addition lies
much of the craft of the enameller? Three main roles (other than decorative
ones) can be discerned from the discussion above.
 GLASS 191

|intermediate layer of iron

between Fe? and Fe2*

Figure 3.27 Transition layer development in FeO saturated glass. This layer chemically bonds
the glass to the metal substrate.

(1) They provide available oxygen by reduction processes. This may be

crucial when atmospheric oxygen is becoming depleted as the molten
glass spreads across the substrate.
(2) They will lower the saturation value for the substrate metal oxide in the
glass, which means that less oxide scale needs to be dissolved from the
surface.
3 — It is quite probable that they speed up the dissolution of the oxide from
the substrate by a pre-oxidation process. For example, whilst the satura-
tion value for FeO in glass is 43 wt % and it dissolves only slowly, the
saturation value for Fe,O, is only 23 wt % and it dissolves rapidly. Hence
a scheme can be postulated, involving a promoting oxide, showing how
this may help to clean the surface more rapidly:
3FeO+ NiO — Fe,0, +WNi
(surface) (promoter) (fast dissolution)

Ni+ Fe,0,; > 2Fe,0, + NiO

(surface) (reformed for further attack)

An example serves to show that the structures peculiar to glass (3.2) are
essential in forming strong interfacial bonds. The example is fayalite, Fe,SiO,,
a crystalline silicate (1.5.1), which when melted has an FeO concentration
above the saturation value and will easily “wet” metals, but on cooling does
not produce any adherence. This is probably because all the Fe?* ions are
shielded below the fayalite surface, and are not available for intermediate
layer formation because the crystalline structure cannot adjust as the glass
structure can. Local fluctuations in glass structure at the interface allow
M-M bonding from the substrate into the glass.
(d) Bonding mechanism in the field-assisted method.* As the two optically flat
surfaces (3.5.1) are heated in air, a thin oxide layer forms on the metal
192 CERAMIC SCIENCE

substrate. When non-oxygen atmospheres are used, oxygen is provided by

decomposition of glass at the anode caused by the electric field.
The electric field increases mechanical contact, and electrochemically
etches away interfacial asperities, so causing even more intimate contact.
After substantial contact between oxidized substrate and glass has been
brought about, dissolution of the oxide layer begins at a controlled slow rate
compared to conventional sealing processes. When all the oxide has dissolved,
it is concentrated in the glass close to the interface, because of the low
temperature and resultant slow diffusion. Electrochemical oxidation of the
clean metal begins when the last oxide is dissolved, and this continues to
introduce metal ions from the substrate into the glass, and eventually
produces the correct concentrations and orientations for intermediate bond-
ing layer formation, and hence a strong seal. It can be seen that this type of
seal and the mechanism of its formation provide very good support for the
concept of sealing by way of chemical bond formation, and no support at all
for the idea of a mechanical bond.
In the other type of solid state sealing, the reaction welding method (3.5.1),
two types of mechanisms of sealing have been observed directly, using
electron microscopy. In type 1, between oxide ceramic and noble metals a
surface reaction between metal and oxide occurs at depths of no more than a
few unit cells. With such small distances involved, there appears to be a sharp
time-independent discontinuity at the interface. Type 2 reaction is a bulk
reaction between oxide ceramics and other transition metals. It is time-
dependent and macroscopic in character, leading to a diffuse interface as the
composition varies with depth. The intermediate layer, and the diffusion
mechanisms leading to its formation, may be extremely complex as our
knowledge of mixed-oxide chemistry suggests (cf. spinels, 5.9).

3.5.3. The glass-metal interface

The main conclusion from 3.5.2 is that the chemistry and structure of an
interfacial layer is of paramount importance in controlling the effectiveness of
glass- or ceramic-to-metal seals. The nature of such an interface can be
discussed and described without a detailed knowledge of its contents and
structure. Theoretically, the strength of a seal can be analysed on the basis of
the energy required to cleanly separate the glass and metal at the interface.
This was formulated as long ago as 1869 by Dupré, who introduced the
concepts of work of adhesion and work of cohesion. The work of adhesion,
Wa, 18 equivalent to the change in energy on separating the glass from the
substrate:

Waa a K(y, + fy Par *y,) (3.68)

where ‘y, is the free energy of the solid-vapour interface, 'y, the free energy of
the liquid-vapour interface, *y, is a free energy of the solid—liquid interface and
 GLASS 193
K is a constant. There is a free energy change (2.9) associated with the
formation of each interface.
After cooling a glass-metal seal, both phases are solids, and so equation
(3.68) must be modified to take account of the strain in the interface:
W.a = K(y, +'y, —%y, — interfacial strain).
Figure 3.27 shows that the interfacial strain is a function of the thermal
expansion coefficients, therefore:

Wo Kyoto), Na (Ag)A (3.69)

where A has been introduced to take into account surface roughness which
will affect the true contact area.
Clearly from equation (3.69) a strong seal will be achieved if the negative
components are minimized. Careful choice of glass composition and substrate
metal will minimize f(A), so that the interfacial energy or tension between
the glass and metal is now the most important; when this is large, the seal
becomes weak. Ideally, the interfacial energy should be zero, but the
discontinuity in structure at the glass-metal interface makes this impossible,
and a minimum discontinuity is required. As shown above, the desired
minimum can be achieved through the presence of an intermediate layer
brought about by saturation of the glass with the substrate-metal or
reducible metal oxide.
Prior to setting, the interface is formed from a liquid glass in contact with
the metal, and consideration of this fact can help to further develop equation
(3.69). At the sealing temperatures, an interface will form and overall wetting
occur if the free energy of the system becomes lowered, or if the interfacial free
energy is less than the sum of the surface free energies of the molten glass and
metal for equivalent areas. The driving force due to a decrease in the surface
energy of the solid, *y,—‘y,, shows itself as a tensile force acting on the
periphery of the drop. Restraining the spreading action of this force is the
surface tension of the liquid glass. The balance of all forces acting on the glass
drop determines its shape, which is reflected in the contact angle (Fig. 3.28).
Horizontal resolution of the forces at A in Fig. 3.28 gives

y. 7 *y = oe cos 6 (3.70)

"
Vapour

Pie eceys:
a Solid

Figure 3.28 The shape of a drop of liquid in contact with a solid when @ < 90° and the forces
existing in the three phases contact at the periphery.
194 CERAMIC SCIENCE -

which is the Young—Dupré equation. It contains the assumption that the

vertical component of 'y, does not distort the surface of the substrate at A,
which for a thick solid is a reasonable assumption.
The Young equation can be used to replace the interfacial energy term in
(3.69) to give
W. a , = KA[!y,(1+c0s 0)— F(Aa)]. (3.71)
From (3.71) we can see that for strong seals 'y, must be large, that is glasses
with large values of surface tension must be chosen. The contact angle 6 must
be close to zero, which is the complete wetting requirement. The term
“wetting” is often used loosely—for practical purposes it is usually said if
8 > 90° the liquid does not wet the solid, and only if 6 =0 is wetting
considered to occur.
Summarizing the requirements for good seals, we have:
(a) the metal must have a large surface energy;
(b) the glass must have as large a value of surface tension as possible (this is adjustable by
changing the glass composition);
(c) the contact angle should be close to zero (this is achieved through dissolution of substrate
oxides to achieve saturation);
(d) thermal expansion mismatch between the components should not exceed 10%.

The craft of the enameller and glass-worker is directed to achieving these

aims, albeit in an empirical way.

3.6 Durability
Durability is the ability of a glass to resist corrosion; most silicate glasses
have a high resistance to corrosion, and are often thought of as inert and
having high durability. However, earlier considerations of static fatigue and
crack propagation (3.4.5) should help to dispel the idea that corrosion does
not occur or that it is of little importance in silicate glasses. “Blooming”, that
is the loss of shine and clarity of glasses and bottles following frequent
washing in detergent, is an easily observable phenomenon that also attests to
the cumulative effects of glass corrosion. Since the important glass properties
of optical clarity and mechanical strength are vitally affected by corrosion, it
is necessary to consider some of the mechanisms and time laws controlling
the process. The aim of this section is first to consider attack by water, acid,
alkali, salt solutions, and gases. Then an attempt can be made to develop the
rate laws for such corroding reagents, bearing in mind that corrosion may
occur by one, or by a combination of three routes:
(i) Reaction with the corrosive agent to form a new compound surface layer.
(ii) Preferential leaching of components of the glass, leaving a leached surface layer.
(iii) By total dissolution, continuously exposing fresh glass.

Finally, some consideration is given to the possibility of protecting glass from

corrosive attack.
 GLASS 195

3.6.1 General observations

(a) Water and steam. In water at pH 7.0, pure silica is regarded as insoluble
at temperatures below 250°C, and only very slightly soluble above this
temperature. This is a crucial observation, since it suggests that water attack
on complex glasses is due only to the existence of soluble components such
as alkali oxides. Simple alkali oxide glasses, such as Na,O - 2SiO,, are of low
durability and dissolve relatively quickly. This observation supports the key
role played by alkali ions in the glass.
Addition of other modifying oxides can either enhance the corrosion rate,
e.g. Al,O3, or can increase the durability, e.g. CaO. Visual evidence has been
obtained showing that water attack produces a surface film; this would be
expected if corrosion followed route (ii) as the alkali is extracted. Electron
microscopy has revealed that the leached layer is porous and not completely
protective. The degree of protection afforded by the surface film must to some
extent depend on whether a static or a flow test system is used. Inastatic test,
preferential leaching of Na* will raise the pH of the water so that other glass
components might be attacked later. Complications arise which might lead
to conflicting expectations, for example, when B,O, is present in the glass it
will be extracted as the pH rises, and it will then act as a buffer to prevent the
pH rising further, which will result in protection of the silicon-oxygen
network. Static tests are used in trade durability gradings (Table 3.8).
(b) Alkaline solutions. These solutions supply OH™ which will degrade Si-O
polymers or networks (3.2) to produce complete structural breakdown with
no protective leached layer formation.
NaOH is more corrosive than KOH and NaCl has a synergistic effect when
added to alkali solutions. From this it might be expected that alkali attack is
total and straight line relationships between mass dissolved and time should
exist; however, this is only found to be the case when the concentration of
OH is closely controlled. As the concentration of OH™ decreases, the
relationship becomes parabolic due to the formation of silicates and hydro-
silicates (1.5.5). In practice, the rate of attack is as much determined by the
concentration of SiO, in solution as by the structure of the glass or the
solution in contact with it. This last fact has suggested that corrosion
inhibitors can be added to solutions to control the rate of attack.
The rate of attack has been shown to be temperature-dependent, that is the
process is an activated one. An activation energy for the process has been
measured for a range of glasses, including pyrex, borosilicate, and window

Table 3.8 Water durability gradings of glasses

Grading Glass

Excellent SiO,, Vycor, borosilicate, high PbO

Good soda-lime, lead-alkali, lead-borosilicate
Poor high alkali, borosilicate sealing, calcium aluminate, phosphate, borate
196 CERAMIC SCIENCE ~

glass, and has an average value around 72kJ mol” *. This experimental value
is close to that found for the activation energy for self-diffusion of Na*
soda-silica glasses, and suggests that Na* diffusion may control the corrosion
process even though there is an excess of OH” ions for general network
attack.
(c) Acid solutions. Corrosion by acid is essentially different to corrosion by
water in that any leached alkali is neutralized, so preventing general Si-O
attack. However, alkali and basic oxides are more easily leached out of the
glass to leave the Si-O layer to reduce the subsequent rate of attack. This
may not always be the case, for high modifying oxide glasses may not contain
enough SiO, to form a protective layer and they will completely disintegrate.
Optical glasses containing high concentrations of PbO are an example of
easily disintegrated glasses, while pyrex and soda—lime-silica glasses rapidly
form thin highly protective layers.
The type of acid also affects the rate of corrosion, depending on the
solubility of its salts, and therefore HNO; is more corrosive than H,SO, and
HCI (nitric acid is in fact used to produce anti-reflection films on optical
instruments by controlled corrosion). Two acids behave completely differently
to all others because of their ability to degrade the Si-O network completely;
they are orthophosphoric acid, H,PO, (1.5.8) and hydrofluoric acid.
One practical application of solubility in the milder acids is the production
of vycor, a glass of very high silica content. Vycor is made by utilizing, first,
phase separation (4.3), and then the differential solubility of the resultant two
glasses. A glass of composition 75 wt % SiO,, 5wt % Na,O, 20 wt % BOs,
when reheated in the range 525—650°C separates into two glasses, one
containing 96% SiO, and one rich in B,O, and Na,O. Nitric acid easily
dissolves the high soda glass, leaving a porous structure which, when heated
to 1000°C, densifies into vycor, a transparent, high melting point, acid
resistant glass that is often used as a cheap alternative to fused silica.
Table 3.9 gives an indication of the extent of the corrosion problem oa
glasses are kept in contact with aqueous acids.
(d) Salt solutions. In general, the presence of ions in solution produces only a
small perturbation of the behaviour observed for water attack. An important

Table 3.9 Acid resistance characteristics of some glasses

Plate kept for Powdered glass in

24 hours at 100°C M/25 H,SO, at 90°C
in5 % HCI: for 4 hours:
Glass residue (kg 1~*) residue (kg |” *)

Vycor (96% SiO.) 4x1071° ax:

Pyrex 10m SSolOee
Na,O-SiO, i 10ss 8x 1078
High PbO Disintegrates
 GLASS 197

class of salts are the detergents. Detergents contain silicates and tripoly-
phosphates, expressed as the P,O;/SiO, ratio. Their effect is two-fold: first,
as the ratio increases, the solution pH decreases, and secondly, many metal
ions that may be present in glass form complex coordination compounds
with polyphosphates. Control of the P,O,/SiO, ratio appears to be crucial,
with minimum corrosion effect occurring at the ratio 1.0, where the pH is 11.9.
Ratios producing higher pH values cause alkali attack on SiO,, and below
pH 11.9, silica deposition from the solution can occur.

3.6.2. Mechanisms and rate laws

(a) Mechanisms. There is considerable evidence to support the view that the
mechanism of glass corrosion depends essentially on ion exchange.!°:*” This
must be an important feature, since modifying cations (3.2) cannot be leached
out without balancing the electric charge so removed. As long ago as 1927,
Kraus noted the ion-exchange properties of glass wool prewashed with
distilled water, on which he showed that Ca?*, Ht, K*, Na*, NHj and
quinine could all be absorbed reversibly.
The ion exchange between molten salts and glass produces surface cracking
(3.4.3); Ag* ions are an exception to this, probably because Na* and Ag*
are close in size. This latter phenomenon is useful for silvering glass to
provide mirrors. Another manifestation of ion exchange can be found in the
staining of wet glass when iron or manganese are left in contact with it. The
mechanism for this staining is partly galvanic:
Fe+H,O — Fe’ +Fe2i,
Glass+Fe®- | —> conduction on the glass surface
Glass®- + Fe?* > Fe? stain.
Even this unsightly reaction can be put to practical use if sodium stearate is
present during the process, because the ion-exchanged chemisorbed layer will
produce Fe-OOCC,H,,,CH; with the oriented CH, group forming a pro-
tective water-repellent layer.
As a result of such observations, it is felt that the first stage of the corrosion

f
!

A ee Si
, Strained Si—O bond
3d orbital ¢ ~~
————__L» Oxygen hybridisation approaching
on silicon |
' sp2— overlapping with Si sp? hybrid

H
sie °

drogen bond
o+ nycre’

Figure 3.29 Schematic representation of the orienting effect of a hydrogen bond between H,O
and bridging oxygen bringing a water molecule lone pair into the Si-3d orbital volume.
198 CERAMIC SCIENCE

mechanism in water, acid, and probably alkali, is an ion exchange process in

which H* replaces Na*. The driving force for the exchange is the gain in
Coulomb energy (5.3.3).
Charles (3.4.7b) has developed this idea, and presents corrosion as a
function of ion exchange and ion diffusion such that degradation of the total
glass structure is dependent on the ionic terminal structures in glass.
Schematically the Charles mechanism is:

| |
(1) —Si—O-...Na* +H,O ~ —Si—O®- ...H®+ +Na* +OH™
| |
F |
(2) —Si—O—Si—+OH™
7 ——Si—O°*-
eee —
...H°?+—Si—O
| | | |
| |
(3) —Si—O- +H,O + —Si—O°- ...H®+ +OH™
|

Step (1) is essentially a typical hydrolysis reaction which must precede step
(2), which is responsible for producing a silanol end capable of further attack
by H,O. The process is autocatalytic, producing excess OH , but the rate
will not continually increase, because as the pH rises silicate ions will form
and remove OH~ from solution (1.5.5).
Whilst ion exchange must provide the dominant mechanism in glass
corrosion, the fact that slow crack growth rate in pure silica is enhanced by
the presence of water vapour means that a second corrosion mechanism must
be available. On the molecular scale the following mechanism has been
suggested.”? First, bridging Si-O—Si bonds at the crack tip are strained up to
20 % and thus have reduced electron densities in the bonding region. Perhaps
most importantly, the z overlap between the oxygen sp hybrid orbitals and
the silicon 3d orbitals as shown in Fig. 2.3b is greatly reduced, making the 3d '
orbital more accessible for donor covalent bond formation from the p,
electron pair on the oxygen of a water molecule. A correct orientation for this
process is provided when a hydrogen bond between one hydrogen from the
water molecule and the bridging oxygen is formed (Fig. 3.29). The hydrogen-
bonded, back-donated bond between H,O and the complex of the bridging
oxygen and one silicon atom is a low energy intermediate, providing a
pathway to complete bond rupture, leaving two silanol end groups in place
of the Si-O-Si bridge. In this manner the crack tip advances to the next link
in the SiO, structure. A corroding molecule must possess a proton separated
from an orbital containing a lone pair of electrons; certainly NH will cause
crack growth in SiO,, but CO (2.2.7), which contains lone pairs but no
protons, does not.
(b) Corrosion rate laws. Many of the direct observations reported above, and
 GLASS 199
the concept of ion exchange, suggest that the rate of attack in water and many
acid solutions will be controlled by diffusion of Na* through a depleted layer
to the H,O glass interface—the concept of diffusion through a hydrated layer.
If C, is the concentration of leachable species at the hydrated layer/glass
interface and C; is the concentration of the same species at the water/hydrated
layer interface, then (3.72) represents the rate of transfer of soluble component
into the H,O.
ds AD(C,—C;)
Pica ieer aa (3.72)

where s = mass of dissolved materials, A = surface area, D = diffusion co-

efficient of the soluble species through the hydrated layer, and d = layer
thickness in time t.
Stirring the solution in order to keep C; small compared to Cy reduces
(3.72) to
ds ADC,
(3.73)
de ad
The amount of soluble material lost by the glass at time t = hydrated layer
volume x mean concentration of the diffusing species. Therefore § = AdC,/2
and d = 2s/AC, which can be substituted into equation (3.73):

ds_ D(AC,)
a ape Er: (3.74)

Integration of (3.74) gives

s? = D(AC,)*t +k. (3.75)

Att= 0, s=0,2 -k.=0.and

s* = D(AC,)*t. (3.76)

Substitution ofA -d-C,/2 for s into (3.76) gives

a = DA*C?t andso d* = 4Dt. (3.77)

These are parabolic rate laws which tell us that under these conditions the
film thickness is independent of the concentration of the soluble material, say
Na,O concentration, in the glass, but the amount extracted into solution
does depend on the glass composition.
Since this process is diffusion-controlled, it is an activated process and must
be temperature-dependent:
D = Kexp(—E/RT). (3.78)
200 CERAMIC SCIENCE

Substitution of this into (3.76) or (3.77) and subsequent rearrangement

leads to
B
Dinogs == log logt—-—
Tre ;

wt
GE
2logdog = og logt— TT

which give the temperature dependence of the corrosion process as log K v. 1/T
linear. This type of temperature dependence and the parabolic law have been
found in many experimental investigations of glass-water and glass-acid
systems.
A second model, dissolution of the hydrated layer, is important in the case
of water (because the OH™ concentration can build up), for alkali solutions,
and in HF or H,PO,.
The rate of layer formation will be given by:

is V, 3.79
dtcecidens Jit Se
where V, is the rate of dissolution of the layer whose thickness is d. When
Ga. suck that K/d,= V,, then dd/dt = 0 and the film will have a constant
thickness as the corrosion proceeds. d at any time t can be found as follows:
dt d
Rearrange (3.79): ia Kava

Integrate:
g * t=el ——dd+C.
Ke iVi d 3.80
( : )

whence i== ev.ae am

v2 og (K+V,d)+M.
)+ (3.3.81 )

Att = 0,d = 0, then

v, 7Digak +C=0. (3.82)

Subtract (3.82) from (3.81)

LAK
“7
Poa kesv, 7 log (K —V,d)

AR oe la 3.83
T Vin Ravi ve oe)
A plot of (3.83) shows that the film thickness is no longer parabolic with
time, but it should be noted that since there is a hydrated layer and glass is
 GLASS 201
corroded by Na* diffusing through this layer, then the mass of glass dissolved
will still be parabolic with respect to time.
Experimental findings agree with these predictions.

3.6.3. Methods of corrosion protection

The preceding discussion concerning the mechanism of corrosion suggests
two approaches to protection that might be successful; first, to prevent the
H*—M"* exchange by producing a surface containing no M"* ions, and
secondly to make it more difficult for M"* to diffuse through the hydrated
layer.
Early attempts to control corrosion centred on changing the glass com-
position to hinder Na* diffusion. This process is not well understood, but
addition of some oxides, CaO for example, will do this; it might be that the
addition of such modifiers greatly affects the composition of the less ordered
inter-cluster material in the Goodman model (3.2.4) making it less rich in
Na” for diffusion. More complete packing in the inter-cluster material as a
result of different ion sizes could also contribute to restricted Na* ion
diffusion.
Na*-depleted surfaces for corrosion inhibition have been produced in a
variety of ways, some of which are:

(a) Flame polishing. Rapid heating of a glass surface volatilizes Na,O, and as a bonus heals, at
least temporarily, the surface microcracks (3.4.2) producing some improvement in the
mechanical properties.
(b) Gas corrosion treatment. Exposure of glass surfaces to moist SO, at elevated temperatures
removes Na* by formation of Na,SO,, which can be washed away:

2Na’* ... glass + SO, +40,+H,O > 2H* ... glass + Na,SO,.

After this ion-exchange step, the glass must be heated to T, to remove H,O
and leave the surface not only depleted in Na* but with a compressive stress
to bring about mechanical improvements (3.4.8). Other methods are used
when special glasses are involved; for example, glass lenses are sometimes
heated to 700°C with powdered zirconia (5.5) on the surface. This results in
a layer of soluble sodium zirconate being produced which can be washed
away leaving the Na* depleted surface layer.
More mechanical methods can of course be used, and common amongst
these is the vapour deposition of a very thin ceramic film. MgF,, V,O3;,
Ta,O, and TiO, have all been used in this respect. The major drawback of
this method is the existence of pinholes or scratches which eventually allow
moisture through. Another problem is the one already discussed in detail in
3.5.1(c) when trying to produce glass-ceramic seals. In an attempt to
overcome both these problems, the silicone treatment method has been
developed. Methylchlorosilane and water are applied to the glass surface, at
202 CERAMIC SCIENCE

which first the water reacts, removing Na* and leaving H*. The silanol glass
reacts with the methylchlorosilane:
—Si—O*- ... H* +CI—Si-——- > HCI + Si—O—_Si—_CH.
This reaction produces an oriented surface layer of CH; groups at the surface
to repel water.

References
1. Adams, R. and P. W. McMillan. Static fatigue in glass. J. Mater. Sci., 12, 643 (1977).
2. Angell, C. A. and K. J. Rao. Configurational excitations in condensed matter and the “bond
lattice” model for the liquid-glass transition. J. Chem. Phys., 57, 470 (1972).
3. Bobkova, N. M. The chemical bond in glass: its theoretical and real strengths. Neorg.
Materially, 11, 328 (1975).
4. Borom, M. P. Electron microprobe study of field-assisted bonding of glasses to metals.
J. Am. Ceram. Soc., 56, 254 (1973).
5. Brennan, J. J. and J. A. Pask. Effect of composition on glass-metal interface reactions and
adherence. J. Am. Ceram. Soc., 56, 58 (1973).
6. DeBruin, H. J., A. F. Moodie and C. E. Warble. Ceramic-metal reaction welding. J. Mater.
Sci., 1, 909 (1972).
7. Burgess, J. F., C. A. Neugebauer and G. Flanagan. The direct bonding of metals to ceramics
by the gas-metal eutectic method. J. Electrochem. Soc., 122, 688 (1975).
8. Cekirge, H. M., W. R. Tyson and A. S. Krauz. Static corrosion and static fatigue of glass.
J. Am. Ceram. Soc., 58, 265 (1976).
9. Cox, S. M. Glass strength and ion mobility. Phys. Chem. Glass, 10, 226 (1969).
10. Douglas, R. W. and T. M. M. El-Shamy. Reactions of glasses with aqueous solutions. J. Am.
Ceram. Soc., 50, 1 (1967).
11. Ernsberger, F. M. Strength of brittle ceramic materials. Am. Ceram. Soc. Bull., 52, 240 (1973).
12. Freeguard, G. F. and D. Marshall. Bullet-resistant glass. Composites, Jan. (1980), p. 25.
13. Gerberich, W. W. and M. Stout. Discussion of thermally activated approaches to glass
fracture. J. Am. Ceram. Soc., 58, 222 (1976).
14. Goodman, C. H. L. Strained mixed-cluster model for glass structure. Nature, 257, 370 (1975).
15. Gupta, T. K. A qualitative model for the development of tough ceramics. J. Mater. Sci., 9,
1585 (1974).
16. Hagan, J. T. Shear deformation under pyramidal indentation in soda-lime glass. J. Mater.
Sci., 15, 1417 (1980).
17. Hsiad, C. C. Spontaneous fracture of tempered glass. Fracture, 3, 985 (1977).
18. Lawn, B. R. Van der Waal’s interactions between brittle cracks and gaseous environments.
J. Mater. Sci., 12, 1950 (1977).
19. Lawn, B. R., B. J. Hockey and S. M. Wiederhorn. Atomically sharp cracks in brittle solids:
an electron microscopy study. J. Mater. Sci., 15, 1207 (1980).
20. Macedo, P. B. and A. Napolitano. Inadequacies of viscosity theories for B,O3. J. Chem.
Phys., 49, 1887 (1968).
21. Marsh, R. D. Plastic flow in glass. Proc. Roy. Soc. A., 279, 420 (1964).
22. Marshall, D. B. and B. R. Lawn. An indentation technique for measuring stresses in
tempered glass surfaces. J. Am. Ceram. Soc., 60, 86 (1977).
23. Michalske, T. and S. Freiman. A molecular interpretation of stress corrosion in silica.
Nature, 295, 511 (1982).
24. Moynihan, C. T., A. J. Easteal, M. A. DeBolt and J. Tucker. Dependence of the fictive
temperature of glass on cooling rate. J. Am. Ceram. Soc., 59, 12 (1976).
25. Narayanaswamy, O. S. A one-dimensional model of stretching float glass. J. Am. Ceram.
Soc., 60, 1 (1977).
26. Porai-Koshits, E. A., S. P. Zhdanov and N. S. Andreev. The Vitreous State. Izd-Vo Akad.
Nauk. SSR Moscow-Leningrad (1960), p. 517.
 GLASS 203
Ze Sanders, D. M. and L. L. Hench. Mechanisms of glass corrosion. J. Am. Ceram. Soc., 56,
373 (1973).
28. Sarjeant, P. T. and R. Roy. A new approach to the prediction of glass formation. Mater. Res.
Bull., 3, 265 (1968).
29. Swain, M. V. Nickel sulphide inclusions in glass: an example of microcracking induced by a
volumetric expanding phase change. J. Mater. Sci., 16, 151 (1981).
30. Tilton, L. W. Noncrystal ionic model for silica glass. J. Res. Nat. Bur. Stand., 59, 139 (1957).
Suite Turnbull, D. and M. H. Cohen. Concerning reconstructive transformation and formation of
glass. J. Chem. Phys., 29, 1049 (1958).
32: Tyson, W. R., H. M. Cekirge and A. S. Krauz. Thermally activated fracture of glass. J.
Mater. Sci., 11, 780 (1976).
33. Uhlmann, D. R. Polymer glasses and oxide glasses. J. Non-Cryst. Solids, 42, 119 (1980).
. Wiederhorn, S. M., H. Johnson, A. M. Diness and A. H. Hever. Fracture of glass in
vacuum. J. Am. Ceram. Soc., 57, 336 (1974).
35: Wiech, G., E. Zopf, H. U. Chun and R. Bruckner. X-ray spectroscopic investigation of the
structure of silica, silicates and oxides in the crystalline and vitreous state. J. Non-Cryst.
Solids, 21, 251 (1976).
36. Yamane, M., S. Aso and T. Sakaino. Preparation of a gel from metal alkoxide and its
properties as a precursor of oxide glass. J. Mater. Sci., 13, 865 (1978).
4 Glass-ceramics

The theme connecting the topics in chapter 1 is an analysis of the restrictions

encountered in traditional ceramics arising from the silicate structures, phase
transitions, and the microstructure. The glassy phases encountered in tradi-
tional ceramics dominate the physical properties of artefacts. Consequently,
it is not surprising to find that many attempts have been made to improve the
properties of the inter-crystalline glassy phases; phosphate bonding (1.5.8) is
one example. A major advance in improving high-temperature strength and
creep can be achieved by causing the glassy phase to crystallize completely,
especially if the grain size of the crystalline phases can be made very small,
since strength is inversely related to grain size (5.6.2). However, many
glass-forming compositions are resistant to devitrification (3.3), and special
efforts have to be made to overcome this problem. The outcome of research
in this field has led to a new area of ceramics—the science of glass ceramics.
This new technology has produced insights into areas such as phase separa-
tion and crystal nucleation. This chapter analyses the main steps in the process
of formation: the role of nucleation; nucleation catalysts; phase separation;
spinodal decomposition; and the controlling variables, namely glass com-
position, temperature, and nucleating agents. To assist in understanding this '
chapter, sections 3.1 to 3.3 should be consulted; some silicate structures (1.5)
are re-encountered, and a knowledge of spinodal decomposition of alloy
systems, as encountered in metallurgy, is relevant.

4.1 Introduction
A glass-ceramic is initially a glass in which, at some stage, the formation of
nuclei is enhanced either by the addition of a nucleating agent or by using
special compositions which are self-nucleating. The resulting material
contains very small crystals.*
This definition points immediately to some advantageous general
properties, and directs our attention to important areas to consider in more
detail.
204
 GLASS-CERAMICS 205
Glass manufacturing techniques have the advantage that any shape is
easily produced with close control of dimensions, and high-speed automation
can be applied. The method of production leads to’ zero porosity and an
outstanding uniformity of properties in the finished ceramic, because of the
molten state first achieved and the nature of the nucleation process. Dimen-
sional changes in manufacture are consequently small compared to any other
ceramic process.
Experience has shown that many practical advantages arise. It is in
principle possible to engineer materials to order,” with specific properties,
from any composition that can be cooled to the glassy state (3.3). Properties
which can be built into ceramics by design at the base composition stage
include thermal expansions in the range —20x 1077 to 200x1077C7},
strengths in the range 6x 10’-10° Nm’, any degree of transparency,
durabilities (3.6) from soluble to inert,’° and electrical properties from semi-
conducting to insulating. New crystalline phases unobtainable by other
routes are sometimes produced, which adds to the excitement of this field.*®
The factors that influence the final properties of a glass ceramic and
therefore those which the technologist seeks to control can be listed as
follows.
(i) Properties of crystalline phases. In this case, as compared to traditional ceramics, the intrinsic
properties of the crystals will have a major effect on the final properties of the artefact and
the role of texture will be less dominant.
(ii) Grain size.
(iii) Intergranular bonding. This property, together with grain size, dominates in the final strength
and appearance of the product.
(iv) Crystal orientation.
(v) Percentage crystallinity and distribution of any remaining glassy phase. Although glass
ceramics should be 100% crystalline, this is not always possible to attain, and indeed for
some applications it is desirable to have a residual glass phase.

Glass ceramic Thermal expansion

(100 % crystals) control over (b) Strength
(c) Opacity
(d) Electrical properties
Durability

By varying

(i) Crystalline phases

(a) Base compositon (i) Grain size
(b) Choice of nucleant (iii) Inter-grain bonds
(c) Heat treatment (iv) Crystal orientation
(v) % residual glass

Figure 4.1 Controlling variables in glass-ceramic production.

206 CERAMIC SCIENCE

The materials technologist can influence these factors only by control of

(a) base composition, (b) the choice of nucleant,! and (c) the heat treatment.
The following sections explain how this is effected, and direct attention to
requirements for (a), (b) and (c) following a closer look at the underlying
science. A series of examples illustrates how these three controls are exercised.
Figure 4.1 summarizes this philosophy.

4.2 Technical processes

The process of glass ceramic production invariably involves a two-stage
heating cycle in order to develop a suitable crystalline texture. Formation of
the crystals depends on (a) the number of nuclei formed, (b) the rate of crystal
growth, and (c) the viscosity of the glass. It is these dependencies that dictate
the heating cycle that is used, for if the temperature range of nuclei formation
and crystal growth overlap, little control is possible. This is shown in Fig.
4.2a. Between temperatures T, and T3, as soon as a nucleus forms it grows
and therefore only a few large crystals would be produced. However, if the
nucleation and growth processes can be separated, as in Fig. 4.2b, then
between T, and T, only nuclei form, and so by maintaining any temperature
between T, and T,, it is possible to control the number of nuclei. When
sufficient nuclei have been produced, the temperature is quickly altered to
one between T; and T,, which promotes rapid growth, thus explaining the
commonly-used two-stage heat treatment.
The processes of both nucleation and growth are activated, and therefore
temperature-controlled—these are considered in some detail in (4.4) and
(4.5). However, the implication here is that practical results will depend on
both the temperature and the time spent at that temperature. This is
represented in a working diagram which plots temperature against log time
(Fig. 4.3). Such a diagram is established experimentally using X-ray, micro-

Nucleation Nucleation

Growth
Rate Rate

4 — — a =<
3 ant aa
Temp. Temp.

(a) (b)

Figure 4.2 Schematic plot of nucleation and growth rates as a function of temperature.
 GLASS-CERAMICS 207

Liquidus
1200
Region ©
Temperature
°

c Region ©

1000

Region @

800 Annealing Strain

Point point
3
1-0 Log (Hrs) 10

Figure 4.3 Heat treatment diagram used to aid selection of technical process conditions.

scopy and thermal-analysis techniques. In Fig. 4.3, conditions of time and

temperature chosen to fall in region (A) allow phase separation and/or
nucleation to occur. In region (B), one major crystal phase is growing.
Conditions specified by region (C) bring about crystallization of a second
crystalline phase, either as a new nucleant or by solid-state transformation
from the first phase. It is easy to see how, once established for a particular
glass composition, Fig. 4.3 enables decisions to be made that influence the
product microstructure, properties and the economics of the process.

4.3 Nucleation
Crystallization is usually modelled as a two-step process, the first step
involving the formation of a stable nucleus, and the second, growth of a
nucleus to form a crystal. Nucleation is classified as homogeneous if it occurs
within a pure phase, and as heterogeneous if the nucleus is formed at an
interface, such as container walls, bubbles, or particles of a second phase.
Since glasses form because of the large energy barrier to homogeneous
nucleation (3.3.4) then it is likely that heterogeneous nucleation is more
important in the formation of glass-ceramics, and therefore nucleation
catalysts are added to the base glass composition. In order to consider the
addition of such agents less empirically, and to elucidate the features they
should possess, it is useful to compare the two nucleation mechanisms on a
quantitative basis.

4.3.1 Homogeneous nucleation

When aliquid is supercooled, formation of a crystal nucleus is possible. This
involves two changes: first, a liquid to solid phase transformation; second,
208 CERAMIC SCIENCE

the formation of a solid-liquid interface. If it is assumed that a spherical

nucleus is formed, the free energy change accompanying the reaction can be
written down, since the phase change evolves energy whilst the interface
formation requires energy.

AG, = 4nr?5y, +977 AG, (4.1)

where r = radius of the nucleus, *y, = solid-liquid interfacial energy (3.5.3)

and AG, = free energy change per unit volume accompanying a liquid-solid
phase change.
The two functions on the right-hand side of (4.1) are plotted as a function
of radius in Fig. 4.4.
The shape of Fig. 4.4 can be explained as follows. For very small particles,
the ratio of surface area to volume is large, and so the interfacial energy term
predominates and AG, will be positive. For large particles, the surface term is
small compared to the volume energy term and AG, is negative.
In Fig. 4.4 there is a critical radius r* associated with the maximum in AG,,
and this can be found by differentiating AG, with respect to r and setting this
equal to zero at the turning point.

d{(AG, yi. Lene

i an SET Vi deaa AG,.

Atr*:
12
0 = 8ar**y, + mrAG,.

4mr2Sy
' |

Radius r

Figure 4.4 Free energy change occurring on spherical nucleus formation.

 GLASS-CERAMICS 209

Hence

eee
oe KGa (4.2)

Clearly from Fig. 4.4 particles of radius smaller than r* are unstable because
their free energy becomes more negative, that is more stable, as their radius
decreases. Such particles are called embryos. Particles larger than r* become
more stable as their size increases; these are nuclei. Further examination of
Fig. 4.4 suggests that nuclei are obtained by adding atoms to embryos, and
since embryo formation involves a positive free energy change, then the
probability of such an occurrence should be very low indeed since few
embryos would be present.
This dilemma is overcome by considering the entropy of the solid-liquid
system, because the entropy of that system can be increased by having some
atom clusters in equilibrium with the atoms of the liquid. Assuming ideal
mixing, it becomes the entropy of mixing, which contributes to the free energy
change of the system in the manner shown in equation (4.3):

AG,= nA6,+kT|—*In(
n,+n
2 + heen
n,+n
: in : )
non
(4.3)
where n, = number of clusters of size r per unit volume, and n = number of
atoms or molecules per unit volume.
The number of clusters is expected to be small relative to n, but since AG,
will be at a minimum for some volume, this term guarantees the existence of
some embryos.
The occurrence of mixing entropy terms is seen to play an important part
in molecular properties several times in this text, e.g. (3.4.6), (3.4.8), (4.4).
The number of the embryos of a given size is:
—AG (4.4)
n, = nexp| ie ‘|
Embryos of critical size r* can add one atom from the melt to become nuclei
available for crystal formation. Therefore the rate of nuclei formation can be
written down as the product of the number of critical-sized embryos present
in a unit volume of liquid and the rate at which atoms are attached to the
embryos. Equation (4.4) gives the first term in such a product, while the
second term comes from considering a model of the process. An atom near
the embryo-liquid interface must acquire an activation energy AG, in order
to leave the liquid and join the solid surface; each second, an atom will
attempt the crossing a number of times determined by the vibrational
frequency Vo, with a probability of success given by (4.5).
P = v,exp(—AG,/kT). (4.5)
Only atoms adjacent to the interface are available to make the jump to the
210 CERAMIC SCIENCE -

embryo and if this number is n,, then the number of atoms crossing the
interface per second is
dn:
a = n.v, exp (—AG,/kT). (4.6)
Replacing vo by kT/h in the manner of reaction rate theory, and admitting
the possibility that atoms from the embryo can make the opposite jump, then
(4.6) becomes
an eikE
~ = —n,— —AG,/kT). 4.7

Now by completing the product, the rate of nucleation, I, is

kT
= nexp(—AG,« (kT) —exp (EAC TET (4.8)
Equation (4.8) gives a qualitative interpretation of the variation in nucleation
rate with temperature, as shown in Fig. 4.2, by considering the exponential
terms. AG,« is found by substituting the expression for r*, (4.2), into (4.1)
16n(*y,)°
AED BAG (4.9)

Now AG, is proportional to the undercooling AT, and therefore AG,« varies
as 1/(AT)*. Hence, at the liquidus temperature exp (— AG,«/kT) will be zero,
and increases as AT increases. The second exponential, exp (— AG,/kT), will
become dominant at large undercooling, when it becomes more difficult to
get an atom to cross the interface as AT increases. Hence the nucleation rate
will decrease with further cooling after having passed through a maximum.
We have now developed two useful expressions for the homogeneous
process, namely J and AG,«, which can be compared to expressions developed
for a heterogeneous system.

4.3.2 Heterogeneous nucleation

The model used is shown in Fig. 4.5 as a solid substrate in a supercooled
liquid on which an embryo is growing. In the model the change in free energy
associated with the formation of the two new interfaces is

AG, = "yA a mr? (°y. aa 1) (4.10)

Liquid

Substrate
g Embryo radius

Figure 4.5 Nucleus formation from a supercooled liquid onto a solid substrate.
 GLASS-CERAMICS Dit

where A is the interfacial area of the liquid-embryo interface, and 'y,, °y, and
'y, are the interfacial energies of the liquid-embryo, embryo-substrate, and
liquid substrate respectively.
Already in (4.10) it is apparent that if 'y, is larger than °y,, then AG, is less
than 'y,A, and the total surface energy required to form an embryo on a
substrate is less than that required for homogeneous nucleation.
The contact angle @ in Fig. 4.5 is related to the interfacial energies by (4.11):
1 e
cos 6 aoa (4.11)
e

Substitution of (4.11) into (4.10) leads to the expression

AG, ='y,A—nr?'y, cos 0. (4.12)
In Fig. 4.5, the embryo is assumed to be a spherical cap from a sphere of
radius R, and the volume of the cap is

y=ar(2 Seonseo’).
2—3cos0 20
(4.13)
The surface area of the embryo A is
A = 2nR?(1— cos 6@) (4.14)
and the radius of the contact area r is
r= sing. (4.15)
The free energy change occurring when the volume V of embryo is formed is
AG = VAG; (4.16)
Using the turning point argument as for the homogeneous case,
ao ie
Re= AG, (4.17)
4.17

which can be rewritten, using (4.15), as

2'y, sin 0
*— es 4.18
’ AG. ee)
Now we have the critical-sized embryo or nucleus, and the total free energy
change to form this is
AG* ='y,A—ar*?'y, cos 0+ VAG,,. (4.19)

Substitution of equations (4.11), (4.13), (4.14), (4.15) and (4.18) into (4.19) and
subsequent rearrangement leads to
16n(*y,)° [Pos 0)(1—cos =| (4.20)
xe 2s
oo 3(AG,)? 4
212 CERAMIC SCIENCE -

Immediately one sees that the first term on the right-hand side of (4.20) is
identical to that determined for the homogeneous case, (4.9). Consideration
of the second term is revealing; if 9 = 0, which represents complete wetting of
the substrate by the nucleus in the presence of the liquid phase, cos @ = 1 and
AG* then becomes zero; there is no barrier to nucleation. When @ = 90°,
cos@=0 and the energy barrier to nucleation is reduced to half the
homogeneous case. At 6 = 180° no wetting occurs, cos@ = —1 and (4.20)
reduces to the homogeneous case.

4.3.3. Nucleant characteristics.

The first and most important characteristic of an effective nucleating agent is
immediately clear: the nucleus—catalyst interfacial energy should be low
(3.5.3). Thus a similar atomic arrangement between nucleus and nucleating
agent, with strong attractive forces across the interface, is required. It follows
from this that the crystal structure and lattice parameters of the nucleant and
the crystal phase to be developed should be as close as possible.
Furthermore, it follows from the technological route to glass ceramic
manufacture that three other characteristics of a nucleant must be sought.
(i) It should be readily soluble at glass-forming temperatures, but sparingly soluble at low
temperatures.
(ii) It should have a low free energy of homogeneous nucleation.
(iii) Nucleant atoms or ions must diffuse rapidly at low temperature; that is at least compared
to the base glass constituents.

It is not surprising, in view of the decrease in free energy requirements

following the introduction of a nucleating agent, that all precursors to glass
ceramics were thought to be colloidal crystals of the nucleant phase. Such a
conclusion was reinforced by much of the early researches, which were
undertaken on base glasses containing metal nucleants such as Ag, Au, Pt
and Cu. A typical base composition in weight % is SiO, (81.5%), Li,O
(1207), K;0 3.5%), Al,O, G0Y%), CeO, (0.03%) and Ag (0.027):
compositions such as this melt to form a glass which can be irradiated with
photons of a suitable wavelength and then heat-treated in the normal
manner. During the nucleation stage, colloidal metal particles are first
nucleated, then grow to about 8.0nm and form nuclei for crystals of lithium
metasilicate to grow (1.5.2).
It was assumed for some time that catalysts such as TiO, acted in the same
way as the metal in the above example, but more recent investigations have
shown that the precursor stage involves amorphous phase separation of the
base glass into two glassy phases. In fact, much evidence is now forthcoming
that many glass ceramic systems crystallize via a glass in glass phase
separation.
The alternative routes from homogeneous glass to major phase crystalliza-
tion are shown schematically in Fig. 4.6. From this figure it is possible to
state that there is little evidence for path (1). Route (6)—(5) is well established
 GLASS-CERAMICS DS

Homogeneous glassLeo Amorphous phase separation

Crystallisation of eS Colloidal crystal of nucleant

major phase phase

Figure 4.6 Schematic routes of the process of formation of a glass-ceramic.

as described above. Route (2)—(3) is now most common, probably because

the interfacial energy between two glassy phases is less than that between
glass and crystals, which as demonstrated in equation 4.10, will reduce the
barrier to nucleation.
Perhaps as important is the possibility that the composition of one of the
new glassy phases may be closer to the primary crystalline phase than the
original glass. Recently there have been several attempts to exploit this route,
with additions to base glass compositions that will promote phase separation;
for example P,O, in the Li,O-—SiO, system which causes separation into a
Li,O-2SiO, and a 2Li,0-P,O; mixture. An added advantage of this
process is that finer-grained glass ceramics can be produced.
Route (2)-(4)-(5) is now recognized as a common one for some of
the technically important glass ceramics. For example, in the system
Li,O-Al,O,-SiO,-TiO,, phase separation of the original glass occurs on
cooling into 5.0nm regions. On reheating, crystalline Al,Ti,O, of about this
particle size is produced, which in turn acts as nuclei on further heating at a
higher temperature for crystallization of the major phase.
From this analysis of nucleation routes the importance of phase separation
is apparent.

4.4 Liquid-phase separation in glass-forming systems*°

Here we are concerned with metastable or sub-liquidus immiscibility,> not
the stable kind that can occur in some systems in the liquidus state. Schematic
binary phase diagrams are shown in Fig. 4.7 which demonstrate the two-
liquid immiscibility regions. In systems like MgO-SiO, and PbO-B,O3;,
which behave as in Fig. 4.7a, the liquidus has a horizontal portion. An
entirely sub-liquidus separation system is shown in Fig. 4.7b, where there is a
flattening of the liquidus to an S-shape, which is often an indication of the
occurrence of metastable immiscibility. BaO-SiO, and Li,O-SiO, are
examples of metastable immiscible systems.
Separation into two phases implies that the free energy of the system has
been lowered by the process. Such a lowering may arise from mixing. Normal
free energy of mixing can be expressed as
AG,, = AH,,= TAS... (4.21)
214 CERAMIC SCIENCE ~

Metastable ‘
2 liquids

Composition Composition

(a (b)

Figure 4.7 Schematic binary phase diagrams showing (a) both stable and metastable 2-liquid
immiscibility, and (b) only sub-liquidus metastable immiscibility.

Figure 4.8a shows this equation plotted for the normal case when AH,, is
negative, and Fig. 4.8b shows a situation which arises if AH,, has positive
values. In Fig. 4.85 it is notable that two minima occur in AG,,, and so
separation of any composition in the region between B-—C into two phase of
composition B and C is to be expected.
It is clear from equation (4.21) that as T increases, then the entropy term
becomes more important, making — TAS,, deeper, with the result that B and

AH

1:0

(a) (b)

Figure 4.8 Schematic plot of free energy of mixing for systems having (a) negative enthalpy of
mixing, (b) positive enthalpy of mixing.
 GLASS-CERAMICS 215

C approach each other in Fig. 4.8b. Eventually a temperature is reached when

B and C coincide and miscibility occurs. The locus of B and C with
temperature describes a dome known as the binodal.
For the simplest model, the regular solution model, which assumes that
atoms or structural units are randomly mixed on a regular lattice, the shape
of the binodal should be symmetrical about a mole fraction of 0.5. Since
glasses are far more complex than such a simple model (3.2), it is not
surprising to find that immiscibility domes in these systems are far from
symmetrical. For example, in Li,O-SiO, and Na,O-SiO, systems the
critical composition occurs at a mole fraction of alkali oxide of about 0.1, not
Saray
For separation to occur within the binodal on cooling, then local fluctua-
tions in composition must occur, and by considering the probability of such
fluctuations occurring a major feature of the dome area becomes apparent,
namely the spinodal region.

4.4.1 The occurrence of the spinodal region**

Consider a homogeneous solution of n atoms with average composition Xg.
Deviations from Xz by an amount AX, will occur as a result of local
thermal fluctuations. Such composition fluctuations will be characterized by
local free energy fluctuations. If G is the average free energy of the n atoms
and G is free energy of any one atom, then the probability that any one atom
will have a G value different from G is given by Boltzmann statistics:
P = Bexp[—(G—G)/kT] = Bexp[—AG/kT]]. (4.22)
The probability that n atoms will have the same G simultaneously will be
P,, = Bexp[—nAG/kT]. (4.23)

Suppose the composition fluctuation AX, is small, then AG will be small,

and so (4.23) can be expanded as a Taylor series about X g:
21 Wi d-G _
Ga— 6 +5(Sy)_ esa. (4.24)

The higher orders in (4.24) can be neglected if xg — xg < 1.

Substitution of (4.24) into (4.23) yields
kn Xp)’ (d?G/dx}) 5
P= B exp] (4.25)
2kT ;

From equation (4.25) it can be seen that a high probability of a composition

fluctuation is associated with small values of d?G/dxz. More interestingly, if
this second derivative term is negative then the exponent will become
positive, making the probability P, > 1, and separation will occur. The
second derivative can become negative in a phase-separated solid as shown
schematically in Fig. 4.9. In Fig. 4.9 the composition under the immiscibility
H
216 CERAMIC SCIENCE
L M

dG
dx, :

dG 6
d x2 =

Figure 4.9 Schematic free energy functions for a phase-separated solid.

dome, the binodal, now has a critical area defined by the compositions at the
inflexion points L and M, between which d?G/dx} is negative and spon-
taneous non-nucleated separation will occur. This inner dome is called the
spinodal. If the glass composition is outside the spinodal but inside the
binodal it will not separate spontaneously but will require nucleation; the
necessary nuclei arise from composition variations in the sample and must be
of critical size, but the critical size decreases as the composition gets closer to
the spinodal region.
Another way of looking at the situation is to consider Fig. 4.10, which
shows the equilibrium compositions of the separating phases to be x and y.

G mole!

A xa y B
Mole fraction B

Figure 4.10 Schematic free energy-composition diagram showing the graphical method of
determining the thermodynamic driving force for liquid-phase separation.
 GLASS-CERAMICS DA:

Consider the case where the glass composition is at a and a small region of
the equilibrium composition, y, separates out. The free energy decrease per
mole of y on separation is given by the distance DE on the figure. This
distance represents the driving force for separation, and is the height above
the free energy curve at y of the tangent drawn to the curve at composition
a. What change in free energy is expected for a small compositional change
to separate out that is not y? The driving force is again the difference between
the tangent and the free energy curve at the composition point of interest,
which in all cases up to b is clearly an increase in free energy per mole. Thus
a composition fluctuation must exceed the composition at b before the free
energy will decrease and provide a positive driving force. There is therefore a
thermodynamic barrier to overcome before phase separation will occur, and
this corresponds to the region between the binodal and the spinodal. Now if
the initial composition, a, lies just to the right of the inflexion point L on the
free energy curve, then even for an infinitesimal composition fluctuation the
free energy change is negative, and there is no thermodynamic barrier.
Between the points of inflexion, therefore, the system is unstable to infinitesi-
mal fluctuations in composition, and separation occurs with a continuous fall
in free energy controlled only by diffusion. This corresponds to the region
inside the spinodal, and the process is called spinodal decomposition.
Whichever of the mechanisms operates within the binodal, the end result
of separation will be the equilibrium phases x and y; the relative proportions
can be determined by the Lever Rule.
Isothermal holding at a temperature below the spinodal in silicate systems
produces fluctuations in the order 3.0—10.0 nm for undercoolings about 10%
of the spinodal temperature; even in viscous liquids decomposition can be
rapid under such conditions, ic. about 1 second. As the temperature is
lowered, diffusion rates will become so slow that phase separation is
inhibited. Cooling rates needed to prevent the phase separation are controlled
by diffusion coefficients; for example if D = 10°'°m?s~*, cooling at 1°s~*
will suffice, but 1000° s~! is required if Dis 107*® m*s~*.
The two types of phase separation within the binodal can be distinguished
microscopically as shown in Fig. 4.11; spinodal phase separation has no
regular form, while nucleated phase separation produces uniform, nearly
spherical droplets. More sophisticated models of the form of the composition
fluctuation have been presented, principally by Cahn.*
In order to establish the position of the spinodal within the binodal (Fig.
4.12), experimental data on free energy in glass-forming oxide systems is
needed. This is not often available, and a method of estimating it is required.
Equation (4.26) has been found useful in making the estimation.
X~—X, =(x—x,)/3 (4.26)

where x, x, and x, are the equilibrium, spinodal and critical compositions

respectively.
218 CERAMIC SCIENCE

Figure 4.11 (Top) spinodal phase separation; (bottom) nucleated phase separation.

So far the simple picture that has been developed to account for glass in
glass phase separation seems quite satisfactory, with its emphasis on a
positive enthalpy of mixing term AH,,,. This leads to the question of why only
some systems have a positive AH,, and whether they can be predicted. Several
attempts have been made to explain the extent of the immiscibility region,
and here we will examine one based on ionic bond strengths.
A term (electrostatic bonding energy) is defined from 5.3.3 as —z,z,e7/r,.
The Si-O bond is considered to be stable, as Si**, O?~ with r,, = 0.162nm,
compared to Na—O with Na*, O?~ and r,, = 0.235nm. It is predicted that
immiscibility occurs when the modifier (3.2.1) has an EBE close to the Si-O
 GLASS-CERAMICS 219

one liquid

Temperature

Mole fraction B

Figure 4.12 Two-liquid immiscibility showing the binodal and spinodal.

value, and the coordination number (5.3.2) around the modifier cation is
different to the value of 4 for silicon. Thus there is competition to determine
the structure. In Table 4.1, the EBE expression has been reduced to z/r since
O77 is constant in oxide glass-forming systems; it can be seen that z/r values
in excess of 10 predict immiscibility through the occurrence of positive AH,
values.
It is therefore a geometrical approach to the problem that can provide
quite satisfactory quantitative predictions. This is illustrated by the example
of the CaO-SiO, system, which according to Table 4.1 produces an
immiscible system. Each Ca?* attempts to complete its coordination sphere
from the non-bridging oxygens (3.2.1). There are twice as many non-bridging
oxygens as Ca?* and so it is necessary for each non-bridging oxygen to be
bonded to more than one Ca’*.
Assume that if two Ca?+ ions are bonded to the same O77
(Ca2+-O?~-Ca?*) they cannot be further apart than (2x0.106)+

Table 4.1 Miscibility in SiO, —M,O, systems

Modifier cation __r (nm) Z z/r(nm~*) — Liquidus curve shape

Cs 0.165 1 6.06 Straight

Rb 0.149 1 6.71 Straight
K 0.133 1 7.58 Straight
Na 0.099 1 10.1 S-shaped
Ei 0.078 1 12.8 Immiscibility dome
Ba 0.143 y 14.0 Immiscibility dome
Sr 0.127 Z SH) Immiscibility dome
Ca 0.106 2 18.9 Immiscibility dome
Mg 0.078 2 25.6 Immiscibility dome
220 CERAMIC SCIENCE

(2x 0.132)nm. Taking the cube of this distance gives the approximate
volume of glass that contains one Ca?* as 0.1079 nm.
Now consider 1 kg of the glass, which will have a volume

1
V =—m? I 10?’
=—x 27 (nm)’.3
p
For xCaO + (1—x) SiO, weight % glass, the number of Ca?* ions will be
__ xx Avogadro constant _ x
* O02 x 10e3.
Ca molecular weight CaO 56.1 x 107°
Hence the volume of glass per Ca?*
Ve eel eee
"EN)pO. 0ooChUex x.
This last equality must from the earlier assumption equal 0.1079 (nm)?.
Putting p = 2.6 x 10° as an average value gives x = 0.33 = 33 wt %, which is
a reasonable prediction of the experimental immiscibility limit of 28 wt %,
CaO.
The EBE approach can be used further in an empirical way, since a plot of
z/r against mole °% of the limiting composition of modifier at the immiscibility
boundary produces a series of parallel lines for the M?*, M** and M**
modifiers. These lines can be used to predict the limiting composition for
immiscibility in other systems. Some other generalizations can also be made.
If z/r < 30, some common features in the phase diagrams are present: on
separation one liquid has a composition close to SiO, ; the primary crystalline
phase field below the two-liquid region is cristobalite. If z/r > 30, the primary
crystalline phase will not be SiO, but either a silicate or the M,O,, end
member itself.

4.5 Crystal growth®:'? .

Once a stable nucleus has been formed in the glass, with or without the aid
of phase separation, crystal growth may proceed. This section deals with
some of the fundamental aspects of crystal growth models.
Growth may proceed by the addition of structural units, atoms or
molecules, to the nucleus at a rate determined by conditions within the melt
and at the interface. In order to move across the interface, 4 in Fig. 4.13, a
unit must acquire an activation energy E, to be able to break from the liquid
and diffuse across the interface to the crystal. We can then write the rate of
transfer of units from liquid to crystal as (number of atoms in the interface) x
(attempted jump frequency) x (fraction of atoms with sufficient thermal
energy), or
dn, —E,
pete: Sv, exp (7) (4.27)
 GLASS-CERAMICS 221

OO" 3c) on
LOLOL G6 0

crystal A
0° Oo
melt

Figure 4.13 Schematic arrangement at the crystal-melt interface.

For the reverse movement there is a larger activation energy to be overcome

since the crystal is a thermodynamically more stable state than the liquid.
This activation energy will be E+VAG,, where VAG, is the free energy
change associated with dissolving the volume V of the nucleus. Thus the net
rate of transfer from liquid to crystal is

dn_ [dng dm) (Be) en (Tet VAG.)

Fre opPacryad besuey hea mas kT
Le,
dn E. —VAG
dé = SV oo (iT ‘|1= exp( LT ele )|
Se (4.28 )

The linear growth of the crystal can be found by dividing the number of
atoms transferred per unit of time by the interfacial area S, and multiplying
the result by the increase in length 4
Ea —VAG,
v= 26syroen(Fe
kT *)|1-ex( LT

which is
E, —VAG
Ue= fagexp(Zz2)| 1 ex ia :)| (4.29)

When the undercooling is only small, VAG, is a small negative number and
VAG,/kT can be expanded as a series, 1+x+ (x?/2!)+ ..., in which only the
first terms need be considered. Hence:

a) VAG,
exp =1+
kT kT
and
— VAG, — VAG,
1—exp kT becomes LT
222 CERAMIC SCIENCE

and
—E VAG
U =A exp(=) OEP . (4.30)

For molar quantities if the molar free energy change, AG, for the liquid to
solid transition is used, the last term in (4.30) becomes — AG/RT. Now
AG = AH—-TAS

and at the liquidus temperature T,, AG = 0

. .AH = T,AS
AG = T,AS — TAS
or AG = ASAT,
assuming AH is constant over the temperature range of interest. Substitution
of this last equality into (4.30) yields
DEY ASNT
U=Ay, exp( eat (4.31)
RT
For small undercoolings, the exponential term changes only slightly and
therefore the crystal growth rate varies as AT. This has already been assumed
in Fig. 4.2. For large undercoolings, the value of AG, becomes large and the
1—exp(—VAG,/kT) in (4.29) approaches unity. Then the growth rate is
limited only by the rate’’ at which atoms can diffuse across 4, which may be
written as
—E
D = A*vo. exp (7). (4.32)

Using this expression in (4.29) produces a simplified growth expression.

D
Us= 7 (4.33)

This implies that the growth rate for large undercoolings would vary
exponentially as —1/T.
The growth rate equation (4.29) holds well when large undercoolings are
involved, but when crystal growth occurs from melts held just below the
liquidus, it needs to be amended by inclusion of a factor that takes into
account the fact that all sites on the crystal surface many not be available for
adding on the necessary building units. This is achieved by multiplying by an
f factor. The f factor can have various forms depending on the temperature
and the mechanism for growth. One example is discussed below.

4.5.1 Screw dislocation mechanism for crystal growth

This model has been developed by a number of workers. It considers an
 GLASS-CERAMICS 223
Archimedean spiral, the equation for which is r = a0, where @ is the angular
coordinate of the spiral, a is the separation of the spirals, and r is the radius.
The central island of the spiral is assumed to be equal to the critical radius
for a two-dimensional nucleus at a degree of undercooling AT, r* = a/2 and
the spiral equation becomes r = 26r*.
For a two-dimensional nucleus,

ees 4
SAR (4.34)

The f factor can be investigated by dividing the total number of atom sites on
the spiral step, of height one atom, by the total number of atom sites on the
area beneath the spiral:

f= wae =e. (4.35)

24%

Substituting for r* from (4.34) gives

pe ao ASAT
amyV
This factor is now used to multiply the growth equation (4.29):

U= a(AS)(AT)? 8 exp(=7") (4.36)

2n°y,VRT kT
It can be seen from this that the growth rate is proportional to (AT)? for this
mechanism, and it has been found that in many systems crystal growth rate
is proportional to the square of the undercooling.
In ceramic systems the maximum growth rate usually ocurs at approxi-
mately 100°C of undercooling, and is usually of the order of 10°-*ms7?.

4.5.2 Spherulitic and dendritic crystal growth

When the nuclei are further apart than about 2 um in the glass, the common
morphologies of developing crystals are spherulitic or dendritic. In these
modes the crystals grow in the form of fibres separated by a residual glass or
crystalline matrix. The fibres constituting the spherulites are usually very fine
and twisted, but in the dendrites the fibres are straighter.
These morphologies are most commonly encountered when the com-
position of the earliest crystals differs significantly from the overall
composition, for in these conditions at the crystal—glass interface the crystal
becomes depleted of the components required for growth and is then
dependent for continued growth on diffusion from elsewhere. Equation (4.33)
shows that the depleted region has a width of D/U. The accessible volume
from which the crystal surface can draw the necessary components for
growth is greatest at protuberances on the crystal growth front, and in
particular at the tip, which accounts for the fibrous character.
224 CERAMIC SCIENCE

If the fibrous growth is in a fast-growing direction, perturbations on its

faces will make it branch. The directions of growth of the branches are
controlled by the interference of the diffusion fields of other branches, which
results in spherulitic morphology eventually becoming radial. At lower
undercooling values, dendrites are more common than spherulites: dendrites
produce branches at angles to the original fibre which themselves then
become branched to re-align with the original fibre. This gives the appearance
of parallel rather than radiating growth. Spherulitic and dendritic growth are
seldom found side by side in the same region.
It should be noted that if the separation of the nuclei produced at the heat
treatment stage is less than D/U, the growing crystals will be in each other’s
diffusion fields, causing growth without the development of dendritic or
spherulitic forms. This is quite normal for fine-grained glass ceramics,
particularly those involving glass in glass phase separation.'®

4.6 Glass ceramics arising from control of process variables

Since glass ceramics begin life as glasses, this must somewhat restrict the
range of materials to be found in this area, since all base compositions must
be glass-forming. This requirement imposes quite strict limitations as set out
in 3.3. In general, therefore, the available glass ceramics are to be found in the
silica plus modifier oxide system, and as shown throughout chapter 3, such
glasses tend to be quite stable once formed. The problem then emerges of how
to make them crystallize. Consideration of sections 4.1—4.5 above leads to the
answer.
(1) Add a catalyst; called hereafter a nucleant.*
(2) Add a destabilizing component to the base glass composition.
(3) Control the temperature and the time at temperature.
These are the three controls the technologist has at his disposal to obtain
a product and to vary its properties. However, since the controls are limited:
to a few base compositions and nucleants, there are (at this stage in the
development of the subject) only a few crystal types to be considered. It is the
aim of this section to summarize and collect the information together in terms
of the effect that the variables (1) to (3), have on producing developments
from the lithia—silica system.

4.6.1 Lithia—silica (Li,O—SiO,) system

Early work was concentrated in this system because the glasses are less stable
than most silica systems, and nuclei form readily in the pure glasses. Use of
(3) produces limited results—below 500°C crystallization proceeds too
slowly to be useful and at temperatures above 580°C only silica-rich solid
solutions based on lithium disilicate, Li,Si,O,; (1.5.5), can be generated.
Above 800°C these solid solutions decompose into stoichiometric Li,Si,O,
 GLASS-CERAMICS 225
and silica polymorphs (1.5.4). Between 500°C and 580°C lithium metasilicate,
Li,SiO,, crystals (1.5.2) can be produced. The small temperature range for
metasilicate and the low temperature for crystallization means that glass-
ceramics in this system inevitably contain only the disilicate which grows in
the glass as duplex fibres with dendritic habit (4.5.2).
Use of (1), nucleant addition, does not produce crystals other than
disilicate but does alter the microstructure, producing a finer scale nucleation.
The common nucleant is photochemically precipitated silver.
Use of (2) can cause, or rather aid, glass in glass separation in this system
to produce a crystalline system containing again Li,Si,O, crystals perhaps
nucleated on lithium phosphate crystals. One percent of P,O; in the
Li,O-SiO, glass stimulates glass in glass phase separation and nucleation to
produce an extremely fine-grained ceramic.
Although only one type of crystal occurs in this system, Li,Si,O,, because
it is of low durability (3.6) compared to the parent glass, and because it can
be nucleated photochemically, a useful material can be developed which
can be chemically machined by preferentially dissolving away the crystals
developed after the heat treatment in areas that were exposed to the correct
incident radiation.

4.6.2 Li,O—Al,03—SiO, system

Here (2) is being applied to the basic system and Al,O, is added to the base
glass to further destabilize the system. Al** substitutes for Sit* in the
network structure and Li* is held nearby to maintain charge balance. The
presence of the Li* ions prevents the Si-O-Si and Si-O-—AlI angles from
deviating from 180°, with the result that solid solutions based on f-quartz
(1.5.4) occur as the crystalline phase. Sometimes these solid solutions are
described as f-eucryptite because the composition range includes that of
naturally occurring f-eucryptite (1.5.4).
High quartz has a zero thermal expansion coefficient, and so glass-
ceramics containing this phase have ideal thermal shock characteristics and

Table 4.2 Some compositions of low expansion glass

ceramics expressed as molar ratios relative to Al,O,

Commercial __Hercuvit
Basic system Corning 9608 Cooker top

SiO, 4 6.5 6.1

Al,O; 1 1 1
Li,O 1 0.50 0.68
MgO = 0.40 =
ZnO ea 0.07 0.13
Na,O+K,0 ea 0.05 0.04
TiO, 0.18 0.31 0.11
ZrO, a Be 0.07
226 CERAMIC SCIENCE
1300 liquidus

1200 keatite + liquid

solidus

Temperature keatite solid solution

¢ [8- spodumene|
356 B- quartz...
+keatite..

700

ee Log time (Hrs.) ae

Figure 4.14 Temperature-time diagram for the Li,O—Al,O,—SiO, system.

excellent strength at high temperatures as a result of which they find use in

cooking ware, cooker tops and telescope mirrors. Typical base compositions
leading to these products are shown in Table 4.2.
Figure 4.14 shows that use of (3) enables a second series of solid solutions
to be nucleated and grown by raising the temperature. The second crystalline
phase is often described as B-spodumene, because although based on keatite,
the composition range spans the naturally-occurring silicate B-spodumene
Li,Al,Si,O,,. In this silicate, as expected from the formula (1.5.1) there is a
3-D network of SiO, and AlO, tetrahedra sharing all corners and not
ordered. Each Li* ion incorporated into the structure to maintain the
electrical charge balance has a distorted oxygen coordination environment
consisting of two oxygens at 0.184nm, two at 0.254nm and two at 0.265 nm
at ambient temperatures. As the temperature increases, the two nearest
oxygens move away and the others move in, producing an almost constant
volume and no thermal expansion. Thus, like high quartz, these crystals have
good thermal shock characteristics but are stronger and more durable (3.6).
Glass ceramics containing keatite crystals therefore have advantages that
may compensate for the higher temperatures needed to obtain them.
Because higher temperatures and longer times are required to get the
keatites in this system some improvement is attempted via control (1),
varying the nucleating agent. TiO, and ZrO, are used and nucleate both
B-quartz’? and keatite, whereas ZnO nucleates only a very fine grained p-
spodumene’! together with a few larger spinel crystals (5.9) based on gahnite,
ZnAl,O,; these are located in keatite grain boundaries.

4.6.3 MgO-AI,0,—SiO, system

Once again control (2) is being exercised by replacing the lithia completely
 GLASS-CERAMICS Doe,

by MgO. There are three main reasons for this; first, magnesia is considerably
cheaper than Li,O; secondly, it improves considerably the glass working
properties of the melt; thirdly, it provides access to another crystal system!4
based on cordierite, Mg,Al,SisO,, (1.5.1). The penalty paid for removal of
Li,O is considerably more difficult nucleation. Nucleation is achieved by
TiO,,”° ZrO, and SnO,, whereby a mixed crystal system of B-quartz plus
cordierite results. The ratio of B-quartz to cordierite depends upon how near
the base glass composition is to 2MgO :2A1,0, :5SiO,.° The cordierite is a
good electrical insulator and has high strength, but suffers from a relatively
large expansion coefficient. Glass-ceramics in this system find use as insulators.

4.6.4 Na,O-—Al,0,—SiO, system

Here again, complete replacement of the lithia, by soda this time, gives access
to another crystal system based on the « or low quartz structure idealized in
nepheline, NaAISiO, (1.5.4). Being related to low quartz, this product has a
high expansion coefficient and is only of modest strength, with poor thermal
shock characteristics. The presence of Na* in these crystals is a good basis
for glaze application (3.5) and as a result such glass-ceramics find a use in
dinnerware.
Alongside the nepheline, other crystalline phases can be produced, depend-
ing on the use of control (1), for example a mixed nucleant of CdO+ TiO,
nucleates CdTiO,, whilst CdO alone gives rise to the spinel CdAl,O, and
TiO, alone produces a mixture of Na,TiO,; and Na,Ti,0,. Niobium
pentoxide is also a successful nucleant,'* producing NaNbO, in the structure.
Since several of the above crystals are perovskites (5.10) or spinels (5.9), the
electrical behaviour of glass-ceramics in this system can be varied by using
control (1).

4.6.5 Tetrasilicic-mica machinable ceramics.

Recently a more sophisticated use has been made of control (2), base glass
composition, by replacing Li,O with a mixture of K,O and MgF, in the base
system containing magnesia and silica.’ Typical compositions in the simple
quaternary system are in wt. % K,O/(11.0), MgF, (10.6), MgO (16.4) and
SiO, (62.0), and glasses can be produced at 1400°C that are transparent to
faintly opalescent, depending on the rate of quenching. Glass-ceramics are
formed by reheating the glasses in the range 650°-1150°C, and contain a
randomly oriented dispersion of tetrasilicic-rnica crystals.’ These crystals
have a formula KMg, ,Si,O,,)F, and a structure similar to the trisilicic mica
fluorophlogopite, KMg,AI1Si,0,,)F,. Thus the structure is analogous to the
natural mica mineral phlogopite, (1.5.3), with fluorine replacing hydroxyl and
the silicate sheets being formed entirely from SiO, tetrahedra. The sheets are
joined into double layers by octahedrally coordinated Mg’* ions, and the
double layers are themselves weakly bound by planes of K” ions. Cleavage
or rotation in the K* planes is relatively easy, and since the crystals in the
228 CERAMIC SCIENCE

glass-ceramic are in random orientations then propagating cracks in the

material (3.4.2) are continuously deflected in different directions with the
result that the propagation energy is quickly absorbed. The fracture paths
follow either the mica cleavage planes or the mica-glass interface, removing in
the process very small fragments, so that a good machined finish is easily
obtained.
The machinability of the material depends on:

(i) the ease of cleavage of the mica structure,

(ii) the aspect ratio of the crystals, and
(iii) volume fraction of the crystals.

It seems possible to control all three factors, for example, if Al°* is intro-
duced to replace some Mg?* in the octahedral layer, the number of K*
required for charge balance decreases and so weakens the double-layer
adhesion. The aspect ratio of the crystals depends on residual glass viscosity
and so can be controlled by base glass composition changes. The volume
fraction of the crystals can be maximized by using an overall composition
close to that of the desired crystals.
Crystal phase development is more complex than usual in these potentially
useful systems. It begins with glass in glass phase separation on cooling (4.4),
hence the common opalescent appearance of the glass. The separated phases
appear to be either SiO,-rich or F-rich, and the matrix MgO-rich. Several
precursor crystals to the final mica crystals have been claimed, one of which
is norbergite, MgF, -Mg,SiO,, on which the fluorophlogopite grows. The
MgO-rich matrix means that enstatite, MgSiO, (1.5.2) is acommon crystal in
this system, and bestows a degree of refractoriness to the product.
Although not occurring in this machinable system, an advantage accruing
from adding fluoride to glass compositions in those systems, producing
B-spodumene crystals, has recently been described. It is worthy of note here
since it relates to “ceramic steels” (5.5.5) and to glass and ceramic strengthen-
ing procedures (3.4.7).
A 0.4% difference in the unit cell volume of B-spodumene occurs between
the exterior surface and the interior of a fluorine-containing glass-ceramic.
The crystals inside the ceramic have a smaller volume, which places the
outside in compression within a 140 pm-thick layer depleted in fluorine.
The magnitude of the compressive stress can be calculated as follows:

Since o,=0, thenassume ¢,=é,=6.

Then ¢, = —[2ve/(1—v)]. (4.37)

The stress in the plane of the surface layer is

r==0,=0= z
5,=0=8| =| 4.38
(4.38)
 GLASS-CERAMICS 229

Now
Av
| meet bytes = ete, (4.39)

where v is the volume.

Hence
Av
— (~)[(1 —v)/2(1 —2v)] (4.40)

and substitution of this into equation (4.38) gives

ae Av E 441
Ne |)2 2y) ue)
where v is Poisson’s ratio.
This last equation can be applied to strengthening and toughening systems
based on tetragonal to monoclinic transformations in zirconia (5.4.5), at least
to a first approximation, since more detailed equations are given in (5.4.5).

4.7 Role of a residual glass phase'*

In (4.1) the definition of a glass-ceramic implied a wholly crystalline product
arising from a precursor glass. In practice, even in cases where the heat
treatment cycle is optimized, a residual glass phase will usually be present.
The role of the residual glass phase can be dominant in that the desirable
properties of chosen crystal phases may never be realized. Even in those
glass-ceramics that do not contain a glass, the presence of a glass phase
during critical stages of microstructural development may be important. For
example, the crystal growth stage normally involves a volume change, and if
the residual glass is unable to relieve resultant stress by viscous flow, (3.1.2),
then micro- or even macro-sized cracking can occur. It is this last effect that
accounts for the frequent addition of Na,O or K,O to the Li,O—Al,0,—SiO,
base by ceramic manufacturers.
A good knowledge of oxide phase diagrams is essential in order to
manipulate the composition of the residual glass phase, since during
crystallization the glass usually becomes enriched in some components of the
base glass, and the composition can be estimated from knowledge of the
stoichiometry of the crystal phases. Indeed, to form a glass-ceramic which is
fully crystalline, the base glass composition should be chosen so that the
residual composition after primary crystallization falls in a region of the
phase diagram where glasses have been shown never to occur. An example of
this control of product can be found in the Li,O0—-ZnO-SiO, system: as
expected from (4.6.1) the first crystals to precipitate are of the disilicate,
Li,Si,O,, which leaves a glass rich in Zn”* surrounding them. In order for
the primary crystals to grow, the Zn?* ions have to migrate away from the
230 CERAMIC SCIENCE

interface and so growth rates are controlled. Eventually B-Li,Si,O, crystals

are nucleated, which now deplete the glass phase of Li,O until a composition
of ZnO-SiO, is reached which will then crystallize as zinc silicate and
cristobalite. It is very hard in practice to prove that any residual glass occurs
in as well a chosen system as the above example, but it is often the case, for
manufacturing reasons, that such careful composition control cannot be
exercised and some of the effects of the residual glass on the properties of
what would then be a composite must be considered.

4.7.1 Strength
From 3.4 we must expect that the mechanical strength of a glass ceramic will
be dependent upon the severity and distribution of microcracks. Since there
are two phases, a glass and a crystalline phase, there must be some
uncertainty as to where the strength-damaging flaws reside.
One suggestion is that the critical flaws are present in the crystalline phase;
this arises from reports that, like normal polycrystalline ceramics (5.6.2 and
Fig. 5.17), strength is proportional to mean grain diameter in the manner of
(4.42).
o = Kd- "2, (4.42)
Now comparing this to the Griffith equation (3.31) would imply that crack
length c is proportional or equal to the grain diameter.
The second alternative is that the fracture-initiating flaws exist in the
residual glass phase, making it the controlling feature of the strength of glass-
ceramics. If this is the case, the mechanical strength must be related to the
volume and distribution of the glass, in particular the mean free path of the
glass 2. Equation (4.43) gives an expression for the mean free path:

A= oe (4.43)
v
where v is the volume fraction of the crystalline phase and d is the mean ,
crystal diameter. At high volume fractions and small particle size, the mean
free path becomes very small, probably smaller than the original flaw size in
the glass phase. When this occurs, the flaw size in the glass-ceramic will be
controlled by the inter-crystal spacing and strength will be proportional to
A~ 1/2, Some data for strength measurements in Li,O—SiO, systems do show
that an increase in 4 */* is accompanied by an increase in mechanical
strength.
Further support for the controlling flaws lying in the residual glass phase
may be found in the fact that glass-ceramics are known to exhibit static
fatigue and antifatigue which are prominent features in the mechanical
behaviour of glass systems (3.4.7).

4.7.2 Control offracture toughness and thermal shock resistance

If the crack, wherever nucleated, takes a tortuous path through the glass-
 GLASS-CERAMICS 231
ceramic, then more energy will be absorbed in surface production (3.4.2,
2.13)
and fracture toughness will be improved (2.13, 3.4.4). Clearly, the presenc
e of
the residual glass phase will be of major importance, because
stresses
developed at the interface between crystal and glass will to a large
extent
determine the crack path. Stresses between the two phases arise from therma
l
expansion mismatch, and consideration of this leads to expectations
for
improved toughness.
(i) crystal < Ogiass- In this case the stresses in the interface are compres
sive
in the radial directions in both phases but in the circumferential directions
they are tensile in the glass and compressive in the crystal. The compressive
stress in the interface would cause a crack approaching from the glass phase
to propagate across the interface to cause transgranular fracture. This more
direct path results in lower fracture toughness. Glass-ceramics containing low
expansion quartz crystals (4.6.2) will inevitably be in this class and will not
have the best fracture toughness characteristics.
(ii) Mcrystal > %giass- In this arrangement of thermal expansion coefficients,
the residual stress in the interface is tensile and fracture will always be
expected to be intergranular, with a longer path length and higher fracture
toughness.
From this it can be seen that suitable oxide additions can be made to the
original base glasses so that the residual glass phase will concentrate them
and achieve improved fracture toughness.
It is not only to regulate the fracture toughness by thermal expansion
coefficient control that additions are made to glass-ceramic base com-
positions, they are also made to help generate materials with a high resistance
to thermal shock. Ideally, for thermal shock resistance the material should
have zero expansion. In zero expansion materials the net expansion of the
crystalline phase is negative and therefore the expansion of any residual-glass
phase must be positive. From the above considerations this will result in the
generation of very high microstresses in the two phase interfaces, which might
result in crack generation. It is then necessary to ensure that no oxides
become concentrated in the residual glass phase that will raise its thermal
expansion coefficient to too high a value if a material with good thermal
shock characteristics is required.

4.7.3 Creep resistance

Viscous fiow is a major characteristic of glass systems (3.1.2); it is not
surprising to find therefore that any residual glass phase significantly affects
the creep resistance of a glass-ceramic.
In general, materials have curves of deformation under stress against time
that comprise three sections:
(a) Primary creep, beginning at the instant of loading and decreasing with
time.
252 CERAMIC SCIENCE

(b) Secondary creep, where a steady creep rate is observed.

(c) Tertiary creep, where the creep rate rises rapidly prior to failure.
The steady rate section of secondary creep has been analysed and shown
to be revealing with respect to the deformation mechanism.
It is usual to assume that the secondary creep rate is a separable function
of three independent variables: temperature, stress, and a composite function
known as the structure factor, included in which are the effects of porosity,
grain-size, and dislocation density. Thus
dé
E o f(T) f(a) f(s).
From this it can be seen that keeping the temperature constant for a given
specimen allows de/dt as a function of stress to be investigated.
It is found that the dependence of creep rate on stress is not linear but
proportional to o”. Experimental and theoretical studies have shown that the
value of the stress exponent, n, is characteristic of the deformation mechanism.
A plot of logde/dt against logo is a straight line with a slope equal to n.
Deformation by viscous flow is associated with values of n close to unity and
values in the range 3—4 are associated with processes such as dislocation
climb.
Further information can be gained by considering the effect of temperature
on creep at constant applied stress. The process is thermally activated and so
the f(T) form of the creep rate equation is proportional to exp(—E,/RT),
where E, is the activation energy for creep. Once again a log plot is useful, for
log de/dt against 1/T produces a straight line from which the activation
energy for the process can be determined. Low values for E,, in the range
360-530 kJ mol~!, are associated with viscous mechanisms. This is in accord
with the values derived for the variation of viscosity with temperature (3.1.2).
Much higher values, above 650kJ mol”’, are found for processes involving
dislocation movements in polycrystalline ceramics.
Armed with the above information one is able to consider creep data;
where available, from glass-ceramic systems. It seems that few glass-ceramics
have been prepared and tested which do not contain a residual glass phase,
and that glass-phase then dominates the characteristics, for most reports give
n = land E, around 410kJ mol”! for the secondary creep region.
Important improvements in high temperature creep resistance can only be
obtained if there is no residual glass phase. Some improvement can be
effected by base glass composition changes that provide residual glass of high
viscosity. Even distribution of the residual glass is preferred, since pockets of
glass might allow large void nucleation and subsequent poor mechanical
performance.

4.7.4 Optical transparency

It is sometimes a desirable property in glass ceramics that they maintain the
 GLASS-CERAMICS 233
transparency of the parent glass. In order to achieve this one of two
conditions must be satisfied:
(i)Crystalline phases of all species must be smaller than the wavelength of visible light.
(ii) The refractive index difference between crystals and residual glass must be very small.

If this second condition can be fulfilled, the size limitation on the crystals
becomes less critical and more easily met. Thus aresidual glass phase in a
glass-ceramic is usually aimed for if transparency is sought. This implies from
the above discussion that transparency is acquired at the expense of strength
and creep resistance.

4.8 Summary
At the start of this chapter it was suggested that progress towards better
ceramics could be made by removal of the glassy phase in traditional
ceramics. Glass-ceramics have been developed to this end, but (4.7) should
make it clear that it is still difficult to achieve and the residual glass-phase
often dominates. It has also been stressed in (4.6) that the requirement of a
parent glass introduces the restriction of a limited number of crystalline
phases available to the materials technologist in this area. It is a field capable
of considerable development. However, certain removal of the glassy phase
and the access to a wider range of properties can only come from the use of
purer materials.

References and further reading

1. Barry, T. I., J. M. Cox and R. Morrell. Cordierite glass-ceramics—effect of TiO, and ZrO,
content on phase sequence during heat treatment. J. Mater. Sci., 13, 594 (1978).
2. Barry, T.I., L. A. Lay and R. Morrell. Refractory glass ceramics. Proc. Brit. Ceram. Soc., 22,
27 (1973).
Ww. Cahn, J. W. Spinodal decomposition. Trans. Met. Soc. AIME, 242, 166 (1968).
4. DeVekey, R. C. and A. J. Majumdar. Metastable zoning in cordierite-based glass ceramics.
Mat. Res. Bull., 8, 1073 (1973).
5. Goldsborough, J. The spinodal—its impact on metallurgy, glass technology and polymer
science. Sci. Prog. Oxf., 60, 281 (1972).
6. Gostelow, C. R. and J. E. Restall. The production and properties of cordierite glass-ceramics.
Special Ceramics 6, Brit. Ceramic Soc., Stoke-on-Trent (1975), p. 23.
7. Grossman, D. G. Machinable glass-ceramics based on tetrasilicic mica. J. Am. Ceram. Soc.,
55, 446 (1972).
8. Harper, H. and P. W. McMillan. The formation of glass-ceramic microstructures. Phys.
Chem. Glasses, 13, 97 (1972).
9. Hilling, W. B. and D. Turnbull. Theory of crystal growth in undercooled pure liquids. J.
Chem. Phys., 24, 914 (1956).
10. James, P. F. Liquid-phase separation in glass forming systems. J. Mater. Sci., 10, 1802 (1975).
11. Kay. J. F. and R. H. Doremus. Strength and durability of a glass-ceramic containing
spodumene crystals. J. Am. Ceram. Soc., 57, 480 (1974).
12. Kirkpatrick, R. J. Crystal growth from a melt. Amer. Mineral., 60, 798 (1975).
13. Kivlighn, H. D. and M. A. Russak. Formation of nepheline glass-ceramics using Nb,O, as a
nucleation catalyst. J. Am. Ceram. Soc., 57, 382 (1974).
234 CERAMIC SCIENCE

14. McMillan, P. W. Glass-Ceramics. Academic Press, London (1964).

By McMillan, P. W. The glass phase in glass-ceramics. Glass Tech., 15, 5 (1974).
16. McMillan, P. W. The constitution, microstructure and properties of glass-ceramics. Phys.
Chem. Glasses, 11, 224 (1970).
life Meiling, G. S. and D. R. Uhlmann. Crystallisation kinetics of sodium disilicate, in Int. Conf.
on Crystal Growth, Boston (1966), ed. H. S. Peiser, Pergamon Press, N.Y. (1967).
_ Scherer, G. W. and D. R. Uhlmann. Effects of phase separation on crystallisation behaviour.
J. Non-Cryst. Sol., 21, 199 (1976).
. Stewart, D. R. TiO, and ZrO, as nucleants ina lithia aluminosilicate glass-ceramic, in Adv.
in Nucleation and Crystallisation from Glasses, Am. Ceram. Soc. Spec. Publ. 5, ed. L. L.
Hench, Cleveland (1972), p. 83.
. Zdaniewski, W. Crystallisation and structure of a MgO-Al,0,—SiO,—-TiO, glass-ceramics.
J. Mater. Sci., 8, 192 (1973).
5 Refractory oxide ceramics

5.1 Introduction
A refractory is a material that can withstand the action of corrosive solids,
liquids and gases and high temperatures. The materials available can be
broadly classified, according to their fusion range, as refractories melting in
the range 1580-1780°C, high refractories in the range 1780—2000°C, and
super refractories above 2000°C. The increasing use by industry of purer and
better refractories is apparent in Fig. 5.1, where the production of refractory
shapes in the period 1955-1973 in the U.S.A. is compared to raw steel
produced in that country. (The steel production rate is shown because it is an
index of industrial activity, and that industry consumes more than 60 % of
refractories.?) While steel production during the period covered by Fig. 5.1

180 000 <— Raw steel

=— Refractories

eapoo
Lan 3.
eae =— Fireclay shapes

«— Basic shapes

(tons)
Log eal _—High alumina
shapes

100
1955 56 57 \ 53/59
\S§ 60 61 62 63 64 65 66 67 68 69 70 71 72 73

Figure 5.1 Shaped refractories shipped and raw steel produced in the U.S.A.

Za5)
236 CERAMIC SCIENCE

increased by about 3% per year, the total manufacture of refractories

remained constant; the weight of refractory produced per ton of steel has
fallen from approximately 45.5 kg to 28.9 kg.
Output of the main product, fireclay brick, has declined over the past 25
years by 4% per year, basic brick involving magnesia has held steady, while
high alumina brick and pure Al,O, products have shown a steady rate of
increase in use. These trends demonstrate an improving quality of refractory
as more pure binaries are used.
Traditional raw materials that have been available to meet refractory
demand are:
(a) Aluminosilicates. The disadvantages of these materials are highlighted in
the discussion in chapter 1, but their main advantages of cost and availability
should not be forgotten. Andalusite, kyanite and sillimanite (1.5.1) are the
most widely used minerals. All three are converted to mixtures of mullite
(1.5.1) embedded in silica (1.5.4) in the temperature range 1250—1500°C.
Andalusite and sillimanite expand by only 4 and 6 % respectively during the
transformation, and are often used without prior calcination. Kyanite has
a 17% volume increase and so must be prefired before refractory use.
Andalusites are generally quite pure compared to most other aluminosilicate
minerals.
(b) Fire clays. These are found with coal deposits and consist of intimate
mixtures of kaolinite, silica and carbonaceous material.
(c) Chamottes are fire clay, reconstituted with reclaimed used materials
known as grog.
(d) Kaolins are formed from china clay minerals (1.5.3), and used as a source
of high alumina refractories.
(e) Dinas refractories are 92-98 % pure silica (1.5.4), much used in Europe.
(f) Alumina materials from bauxite. These form a special section and are
extensively discussed in this chapter along with magnesites.
It is apparent that more attention is being given to the development of pure
oxides for ceramic use. The question is where this development will centre,’
and what criteria can be used to direct our attention to possible refractories.
There are about twenty-two refractory oxides, which can give rise to over
9000 binaries, but since in any system increasing the number of elements
tends to lower the maximum fusion point of compounds, it is probable that
only about 1000 of these will be refractories as far as the fusion temperatures
are concerned. About 150000 ternary oxides could be formed from the
original twenty-two, but most of these have not been made, and only about
2 ¥ are likely to have refractory melting points, so our search for refractory
materials can be confined to the pure oxides and some binary systems.
Although an apparently extremely large area still remains to be investigated,
this can be reduced to realistic proportions by reconsidering the full definition
given at the beginning. Satisfactory candidates must be able to withstand the
action of corrosive solids, liquids and gases at high temperatures. We can use
 REFRACTORY OXIDES De,
this requirement to make a selection from among the possibilities in a
semi-quantitative way.

H,0+MO=>MOH= MOH ve
solid solid solution EOS

Measurement of equilibria of this type, and in particular the relative values

of K, and K,, allow a classification of oxides. When K, is large, the oxide is
classed as basic. When K, is large, the oxide is acidic, but when K, ~ K, and
neither is large, the oxide is amphoteric. The effects of polymerization and
insolubility make these statements gross over-simplifications, but they do
show that K, and K, will be of the same order if polarization of the O-H
group by M”* is neither strong nor weak.
One way to quantify this effect is to consider the r/z ratio (4.4) values of
which can be used to define the amphoteric region; i.e., when 0.03 > r/z > 0.01,
then oxides will be neither acidic nor basic but amphoteric. An amphoteric
oxide will not readily react with acidic or basic media and will therefore be a
potentially good ceramic material.
Acidic oxides which have low r/z values will be useless within the definition
of ceramics, because weak inter-molecular bond forces lead to low thermal
stability and poor chemical stability. Strongly basic oxides have r/z > 0.03,
which leads to highly ionic 3-D structures with high melting points, but they
react easily with acidic slags and even with acidic components of the air, e.g.,
CaO + CO, = CaCO. Hence these oxides cannot be considered seriously as
ceramics.
Now we can see that for ceramic use we can mainly confine our attention
to binary oxides with r/z values in the range 0.01—0.03. We can further confine
our attention to systems which are stable in air or vacuum, which effectively
removes many transition-element oxides from consideration, and reduces the
twenty-two original candidates to more manageable proportions. A further
requirement is for no extensive polymorphism in otherwise suitable oxides.
This finally reduces the field for detailed consideration to Al,O3, ZrO,, MgO,
and to a lesser extent UO, and TiO,. Because of their importance, spinels
and ferrites from amongst the ternary oxides need to be considered too.
The approach here is, first, to examine in some detail the bond model most
useful for interpreting the behaviour of these materials, namely the ionic
bond, and so 2.2 should be consulted for details concerning electronic
structure. At appropriate places, the discussions on strength in 2.1, 2.13 and
3.4 are developed more fully by concentrating on the effects of microstructure
and impurities on this property. At the end of each section, some modern
developments and uses of these materials are given. Finally, since “refractory”
238 CERAMIC SCIENCE

implies use in aggressive atmospheres at high temperatures, some useful

thermodynamic equations, simplifications, and data are given; this material
has been used and referred to already in earlier chapters.

5.2 Prediction of environmental stability

Clay-based ceramics, as the introduction states, have been commonly used at
high temperatures for many years, despite the limitations we have already
considered. It is now worth emphasizing that the majority of refractories,
whether clay-based (chapter 1), special (chapter 6) or oxide-based (this
section), are not thermodynamically stable in their service environments, and
development has focused on erecting kinetic barriers to the deterioration that
is ultimately inevitable. Thus materials scientists are often asked questions
such as: is A stable if in contact with B at x°C? How much AB would arise
from BC +AC? What temperature causes appreciable vaporization losses of
X from XY + AB in contact?
By assuming that equilibrium is approached, answers can be given by
applying the principles of equilibrium thermodynamics and using in particular
a selected few thermodynamic functions. The development of, and detailed
studies of the functions listed below, can be found in all classic thermodynamic
texts.
For our purposes the following will suffice:
ae Tr AC?
AG° = AH%o5 +|
298 298

where AG° is the standard free-energy change as a function of temperature in

terms of the standard enthalpy, entropy, and heat capacity changes.

Co =a+bT—CT~? (52h

AG° = A+BT log T+ CT. (5.3)

Here (5.2) and (5.3) are empirical relations for C and AG® as a function of T.

AG = —RTIn K,. (5.4)

Equation (5.4) is a versatile equation, much used (see for example 1.5.1,
2.11, 2.8 and 6.3) and relates the free-energy change to the reaction equilibrium
constant as defined by gaseous product partial pressure measurements.
Equation (5.1) is an accurate description of the temperature dependence of
the standard free energy over a wide range of temperatures; however, it is
possible to use a linear equation with average values of AH® and AS° and
obtain satisfactory answers, because AC, for most reactions is small.
 REFRACTORY OXIDES 239
Even when AC, is not negligible, the changes in AH° and AS° with
temperature are similar and tend to cancel each other, and so an average
equation is quite adequate. Making this assumption, and ignoring small
enthalpy changes encountered in allotropic phase transformations, allows a
general and useful table of data to be presented for the mean equation (5.5),

AG° = AH°— TAS° (5.5)

and this applies throughout the temperature range listed in Table 5.1.
The data in Table 5.1 immediately indicate the stabilities of a variety of
ceramics, and can be used, as examples show in the text, to make estimates of
the stability of chosen systems. However, it should be remembered, as much
of chapter 6 shows, the kinetic barriers that can be erected against reactivity
and the attainment of suitable microstructures can often make the predictions
appear to be incorrect in practice; this is of course an illusion.

5.3 The ionic bond

The ionic bond model is intrinsically simple in its approach: it basically
ignores the effects in atoms caused by the existence of atomic orbitals (2.2)
but stresses an electrostatic interaction of charged species. It assumes that
ions, that is charged atoms, exist as hard incompressible spheres, which is a
view in sharp contrast to the accepted one of atomic structure. Given these
entities, then application of electrostatic principles and the concept of
space-filling by close packing of hard spheres lead to useful predictions
concerning stabilities and structures. To a satisfying degree, considering the
above assumptions, the model is successful in predicting so many modes of
behaviour.
There is much pseudo-evidence given for the existence of ions in certain
solids, mostly based on features such as the limited number of structures
known to exist, the electrical conductivity, optical spectra, magnetic prop-
erties, some detailed X-ray analyses, and Mossbauer spectroscopy. In prac-
tice, one considers a variety of evidence from a number of sources before
accepting the starting hypothesis of the existence of hard spherical charged
atoms (i.e. ions) in certain solids. The most convincing evidence and justifica-
tion for the ionic bond model must ultimately lie in an analysis based on
energy considerations, where the driving force for ion formation is ascribed
to the strength of bonds in an ionic assembly. We can perhaps best see this
from a stepwise analysis.

5.3.1 Production of isolated cations

Octet theory suggests that the attainment of certain stable electron configura-
tions with magic numbers 2, 8 and 18 is the driving force for ion formation,
a view still unchallenged in many minds despite the preparation in 1962 by
240 CERAMIC SCIENCE

Table 5.1 Energy values for use in equation 5.5

Reaction AH° (J) AS° (J) Range (°C)

Al) > Ala 10 784 11.53 25-660

Aly > Alg) 304 346 109.4 660-2520
1506 305 Ds. 660-1276
AIF 36) fareAly oT 3/2F x)
1228 377 79.29 660—2000
AIF3¢) = Alq) ats 3/2F x)
219 784 7.44 660-2200
AIBi) > Alg + 12Be)
AIN« > Ala+ INo@) 326751 115.4 660-2000
Al,O; > 2Alq + 3/2026) 1 685 627 326.5 660-2050
Al,O@) > 2Alq) + $0246) 170 544 — 49.32 660-2000
3A1,03-281026) > 3A1,03¢) + 2810 x) 8 590 — 17.39 25-1750
ALSiO5,) > Al,O36) + Si024) 8 803 3.89 25-1700
Be > Bay 50 160 21.81 25-2030
By) > Be) 569 609 147.7 25-2030
By — Bg) 499 176 M72 2030-4000
B,C, > 4B) + Ce, 41 465 5.56 25-2030
BNg > Be) +3Nq) 250 382 87.53 25-2030
B,O3\) > B2O30) 24035 33°23 25-450
B,O3q) > 2B) + 3/202) 1227 666 209.8 450-2040
BaAl,O4) > BaOg) + AlnO3¢5) 124 146 — 6.69 25-1830
Ba,SiO4,) > 2BaO + Si0 26) 259 578 —5.85 25-1760
Be,C,) = 2Be«) a Cw) 93214 13.79 25-1285

Be;N>4) > 3Beg) + Now 586 872 185.2 25-1285

Be3No¢) =?. SBeq. te Noe) 615714 203.0 1285-2200

BeO,, > Beg) + 202) 607 604 97.60 25-1285

BeO« > Beq + 50g) 613038 100.8 1285-2000
BeAl,O4,) > BeOw) + Al,O3) 4180 — 2.93 25-1870
Be,SiO4) > 2BeO%) + Si0 4) 10 868 7A1 25-1560
graphite > diamond 1442 —4.47 25-900
Ca ' Ce 712774 155.3 1750-3800
CHa) > Ce) + 2H) 90957 110.6 500-2000
COg) => Cats 402) 114281 — 85.69 500-2000

CO r@) > Ce)+ Or) 394 968 —0.54 500-2000

CaC x) > Cag + 2C¢) 60 192 — 26.25 840-1484
Ca,Si«) > 2Cag + Sig 209 000 ZIAD 25-840
CaSi«) > Cag + Sig) 150 480 15.47 25-840
CaSi.) > Cag) +2Si 150 480 28.42 25-840
CaO) > Cag +402) 639 540 108.5 840-1484 |
Ca3Al,06) > 3CaOg) + Al,O3¢) 12540 — 24.66 500-1535
CaAl,O4.) > CaO + Al, 036) 17974 — 18.81 500-1605
CaAl,07,) ~ CaO) + 2A1,0;3 16720 — 25.50 $00—1750
Ca;SiOs,) > 3CaO,) + S104) 118 712 — 6.69 25-1500
Ca,SiO4,) + 2CaO@) + S102) 118712 — 11.29 25-2130
CaSiO34) > CaSiO3,) 56012 33.02 1540
CaSi034) =< CaO + S102) 92 378 Zoi 25-1540

CaZrO,,,) > CaO, + Z1O 46) 39 292 0.42 25-2000

CaMgSiO, > CaO, + MgOq) + SiO 2) 124 564 3.76 25-1200
Ca,MgSi,O7,) > 2CaO,) + MgO@) + 281026) 204 820 0. 25-1454
CaMg(SiO3)2) > CaMg(SiO3;)2q) 128 326 77.08 1392
CaMg(SiO3)2.) ~ CaO) + MgO@ + 2810445) 162 602 18.81 25-1392
Ca3A1,Si30 12%)> 3CaO,,) + Al, O03) + 3810 26) 388 322 100.3 25-1400
Ca,Al,Si07,) > 2CaO«) + Al,O3¢) + S10 246) 170 962 8.78 25-1500
CaAlSiOg.) > CaOg + Al,03) + S10 46) 105 754 14.21 25-1400
CaALSi,O; > CaO.) + Al,O4) + 2810 44) 138 776 17.14 25-1550
CeBeis) > Cea + 6Bi) 376 200 55.18 798-2190
 REFRACTORY OXIDES 241
Table 5.1—continued

Reaction AH? (J) sie AS °() Range (°C)

CeCz4) > Ceq +2Cg) 85 147 — 26.96 798-2250

CeNg > Ceg) +3No) 487 806 176.9 2000-2575
CeSi + Ceq + 2Sig 200640 16.30 798-1425
Co2B,) — 2CO@) te Bw) 129 580 10.87 25-1200
CoBg) > Cog) + Be) 96 140 7.52 25-1460
CrBy) > Cre) + Beg 75 240 6.69 25-1000
CrBy) > Cr) + 2B) 91960 8.36 25-1000
Cr3Cr 6) => 3Cr¢) ar2Cw@) 79 002 —17.64 25-1850
Cr3Si¢) =e 3Cr¢) aigSig) 106172 3.39 25-1410
Cr;Sizg) > SCI + 38i¢ MOOI 45 25-1410
CrSig) > Cr@ + Sig) 53 504 — 3.89 25-1410
CrSi, > Cre + 2Sigg 79 420 1.96 25-1410
Fe,Be) > 2Fe) +Be) 87780 18.39 25-1390
FeB,) > Feg) + Be 79 420 10.45 25-1650
FeOq > Feg +402) 255 816 53.63 1371-2000
Feo9s04) rd 0.95Fe +402) 263 465 64.29 25-1371
Fe304,) as 3Fe,) +202) 1101 137 307.1 25-1597
FeAl, Og) > Fe,Ow + (1—x)Feg +AlOe 33 106 6.10 1550-1750
Fe,SiO4q) > Fe SiOug, 91960 61.61 1220
Fe,SiO4g > 2FeO + SiOx) 36 199 20.92 25-1220
H,0q > Hyg) +2028) 247 247 55.80 25-2000
HfC,) > Hf.) + Ce) 229 900 (se? 25-2000
25-1700
HfO (45) => HfO4(gys) 10 450 ail 1700
KAIF) > 3Kq) + Al) + 3F 26) 3318920 544.2 66-660
K,Si03,) — K,Si0O3q) 50 160 40.17 976
KiS:0,9- K;0g 4510), 279642 —0.46 25-976
LaNg > Lag +4N24) 296 780 105.8 25-920
Li,SiO3,) > Li,SiO xq) 28 006 19.02 1201
Li)Si03¢) Si Li,O%) are Si0 45) 143 792 3.76 25-1201
Li,Si,05,) —? Li.Si,05q) 53 922 41.13 1034
E60: > 11,0— 4 S105, 146 300 0 25-1034
Mg3Ny@) > 3Mg@) + No) 459800 202.7 25-649
Mg,Si«) > 2Mgq) +Sie 100320 39.29 649-1090
MgOw > Mg,) #40a 600 373 107.5 25-649
MgOq) > Mgq) +402) 728 908 203.9 1090-2000
MgAlsOaq) > MgO, +ALO4,) 35 530 —2.09 25-1400
Mg2Si0 4.) — Mg,SiO4q) 71 060 32.60 1898
Mg,SiOgq > 2MgZ0w + SiO 76) 67131 431 25-1898
MgsiO3,) =F Mgsi03,) 75 240 40.55 1577
MgSiO;,) > MgO) + SiOx, 41089 6.10 25-1577
MnsSi3z¢) > MnSi3,) 172 634 109.8 1300
MnSi3/) =? S5Mnw) aT 38i¢s) 204 820 = 15.05 25-1240
MnAI;O4, > MnO, + Al,O) 48070 7.32 500-1200
MnSi03;) =. MnSi03\) 66 880 42.76 1291
MnSiO3 > MnO + SiO 28006 2.76 25-1291
Mo,Cw > 2Mo@) + Co) 45562 —4.18 25-1100
MoC,, > Mow + Co 7524 — 5.43 25-700
Mo,No) > 2Mow) +4No¢) 60610 14.63 25-700
Mo;Sig) > 3Moq) + Sig) 118712 2.51 25-500
M05Siyq) > 5Moq)+ 3Sig) 310992 4.10 25-1410
MOSix) > Mog)+ 2Si¢ 132 506 2.80 25-1410
NH3¢) => FNo@) ote 3/2H 2) 53 671 1 16.4 25-2000
Na3AlF 6) = Na; AlF 107 175 83.41 1012
 242 CERAMIC SCIENCE

Table 5.1 —continued

Reaction AH° (J) AS° (J) Range (°C)

Na, AIF 6.) > 3Naq +Aly) + 3F x6) 3 305 544 553.2 98-660
Na,AlF 4) > 3Na) + Aly + 3F 2) 3 374932 622.8 1012-2000
Na,SiOx,) + Na,SiOx,) 51748 38.00 1089
Na,SiO,,) =. Na,O,) + SiO) 237 424 8.82 25-1089
NajSi,05) radNa,Si.O5,) 35 530 30.97 874
Na,Si,05) ~ Na,O,) + 2810446) 233 244 — 3.85 25-874
NbB»s) > Nby) + 2Be) 250 800 9.20 25-1100
Nb,C,) > 2Nbe) + Cg) 193 534 11.70 25-1500
25-1500
Nb,N,) > 2Nb,) +4No) 250 800 83.18 25-2400
NbNg) > Nbg) +5No¢¢) 229 900 as 25-2050
NiBg > Nig) + Be 175 560 8.36 25-1000
992
NiSij) > Nig) + Sig) 91124 79.42 25-992
NiO.) > Nig) +402) 235 376 85.98 25-1984
NiAl,O4,) > NiOg) + Al, 036) 4180 —12.54 700-1300
Ni,SiO4,) > 2NiO,) + S104) 15 466 9.20 25-1545
Pb, SiO 4.) > Pb,SiO4,) 50996 50.20 743
Pb, SiO4) > 2PbO,q + Si02,) 33 440 — 6.69 885-1500
PbSiO 5; > PbO + SiO.) 25080 1225 885-1500
764
ScNq > Scw) +3N2) 313 500 98.23 25-1539
1412
Sig) > Sig) 395010 1S 1412-3280
a-Si3;Na) > 38i¢s) +2No) 723 140 314.8 25-1412
a-Si; Na) > 3Sig) + 2No¢) 873 620 404.6 1412-1700
25-1412
1412-2000
Si0Oq) > Si) +302) 104 082 — 82.43 25-1412
quartz > SiOrq) 7691 4.51 1423
quartz > Sig) + Org) 906 224 175.6 25-1412
cristobalite > SiO») 9572 4.81 1723
cristobalite fourSivs) +Ong) 904 970 175.4 25-1412
SnO2) > Snqy +40, 574 332 198.2 232-1632
SrAl,O4,) > SrO%) + Al,03,¢) 71060 —4.18 25-1300
Sr2Si04) =? 2S8rO4) at Si04,) 213 180 5.85 25-900
SrSiO,,) => SrO% + S104)
133 760 4.18 25-900 |
TaBy,) > Tay) + 2B) 209 000 9.61 25-1700
Ta,C) => 2Tag) a Ce 200 640 2.09 25-1700
Tato > Tag) + Ce 142120 1.21 25-1700
Ta,Ng) > 2Tag) + ZNr) 263 340 90.71 25-1700
TaNg > Ta) +3Nre@) 246 620 81.09 25-1412
TaSi¢) = 2Ta) otSig) 125 400 — 3.34 25-1412
25-1412
TaSing) > Tag + 2Si¢) 117040 2257 25-1412
ThC,) > The) +2Cg 125 400 —8.36 25-1755
ThNg) > The) +3No¢) 376 200 86.11 25-1755
ThSiyg) + The, + 2Si¢, 175 560 10.45 25-1412
ThOx) > The) + Org) 1220560 183.1 25-2000
TiBy) > Tig) + Bo) 163 020 5.85 25-1670
TiB24) > Tig) + 2Bg) 284 240 20.48 25-1670
TiCy > Ti) + Ce)
TiN) > Tig) +4N22)
184588 12.54 25-1670
335 988 93.17 25-1670
TiOg, > Tig +307) 514140 73.99 25-1670
 REFRACTORY OXIDES 243
Table 5.1—continued

Reaction AH° (J) AS° (J) Range (°C)

Rutile > Tig) + Org) 940 082 177.4 25-1670

Ug > Ug) 8527 6.06 1132
UBy) > Ug) +2Bg) 179 740 15.55 1132-2430
UC, = UgtCe 109 516 1.84 1132-2520
U2C36) > 2Ug) + 3Cw) 221 540 Dan 1132-1700
UC; 938) + Ug + 1.93C,) 90 288 9.20 1132-1700
UN > Ug) +4No@) 292 600 80.67 25-1132
USi > Ug) + Si) 87780 5.27 25-1132
UO, > Ug + Org) 1085546 172.1 1132-2000
WC) > Wa) +Ce 42 218 4.97 900-1302
W046) > We) +3/202¢) 832 656 245.2 25-1472
YN) > Y@) tN) 296 780 99.48 25-1526
Y2ZrO54) > Y.034) + ZrO) 20900 0 25-1200
ZnFe,0 4) > ZnO@ + Fe304¢) 9614 3.76 25-700
Zn SiO 44) > 2ZNO«) + SiO) 36 366 2.84 25-1512
Z1By4) > Zr@) + 2Be) 327712 23.41 25-1850
Z1Cw > Zr~)+Ce 196 460 9.20 25-1850
Z1N@ > Zt) +4N2) 363 242 91.96 25-1850
ZrSig) > Zt) + Sig) 158 840 4.18 25-1412
ZrO, > Zrq) + Or) 1090980 183.5 25-1850
ZrSiO4.) > ZrO) + SiO.) 26 752 12.54 25-1707

Bartlett of dioxygenylhexafluoroplatinate, OS PtF;, and XePtF,. The

xenon compound is a charge-transfer compound, stable at room temperature,
which takes from xenon one of the configurations to which other atoms are
said to aspire. Indeed to achieve a “stable” electron configuration, formation
of a positively charged atom must in all cases be energetically unfavourable
as an isolated step.
M > M’* +ze™. (5.6)
Equation (5.6) as written must always be positive in its energy requirement,
and AH for the process is the ionization energy (I.E.). When this energy is
expressed in electron volts (eV), it is called the ionization potential. The
ionization energy decreases with the increase in atomic number of non-
transition elements and increases rapidly with increasing values of z. Further-
more, the energy required to form cations in those ions with full d orbitals will
be greater than others due to orbital penetration effects. In this way K and
Cu form ions by the loss of a 4s electron, but the ionization energies are
respectively 418.3kJmol~' and 744.7kJ mol~! from outer electron con-
figurations of 3d°4s! and 3d'°4s'. Stable ceramic oxides containing sub-
stantial ionic bonding might therefore be expected to have large cations of
low charge. We have seen, however, at the beginning of this chapter, that
chemical instability, as predicted by the r/z ratio, would result. This anomaly
must be resolved by considering the problem more closely.
244 CERAMIC SCIENCE

5.3.2 Production of isolated anions

Here the process is seen as the gain of electrons and is expressed by equation
(S21):
X+ze” — X?-. (5.7)
The energy associated with this step is called the electron affinity (E.A.). It
might be pictured as the accommodation of a high-energy electron in an
available low-energy orbital (2.2). It follows that, since the electron must
occupy an atomic orbital, then the type of electron wave it is to become (2.2.4)
plays a vital role. For example, s electrons within the same orbital have poor
shielding for each other and therefore each is significantly attracted to the
nuclear charge, with a net stabilization of the system. However, when p
orbitals are involved there is better shielding from the nucleus by an electron
already in the same orbital, with the result that there is less net attraction for
the electron. The repulsion effect caused by the close confinement of two
electron waves requires energy to be supplied, and the system may assume an
overall higher energy state. In most cases, the system does not achieve this
higher energy state, because addition of an electron as a p function usually
increases the total orbital angular momentum L (2.2.4), with consequent
energy loss from the system to balance the electron—electron repulsion effect.
The exception is when the ion would gain the np? or half-filled configuration
which is an S state (2.2.5), has no net angular momentum, and therefore
does not provide stabilization of the electron repulsion within the orbital.
Hence formation of the N~ ion is always an energy-gaining process with no
driving force causing it to occur.
When the number of electrons to be added in equation (5.7) is greater than
one, then there is always an energy gain in the system with resultant loss of
driving force to form anions. We are concerned in this chapter mainly with
the existence of O?~ ions, and we see from Table 5.2 that their formation will
always require energy to be supplied; it is not a spontaneous process.
From the discussion so far, we would not expect O*~ ions to occur in

Table 5.2 Some electron

affinities

Ion E.A. kJ mol?

Ba — 332.3
Cis — 357.4
Br- —336:5
is — 307.6
On — 141.2
im =207-3
On +719
S25 +418
Na~ alley
 REFRACTORY OXIDES 245

solids. Furthermore, since I.E. > E.A. for all ions with a single charge, we
would not expect ions to exist at all. Indeed we would conclude that we
should encounter Na™ rather than Na* ions in compounds. The solution to
this paradox is found when we consider that if the cations and anions can be
produced, then having opposite charges they attract each other, and that
process reduces the total energy of the system.

5.3.3, Coulombic attraction

The relationship describing the force between charged particles was estab-
lished in 1784 by Coulomb and, for oppositely charged particles, given the
form:
F= —kz,z,e?
:: (5.8)
where z, and z, are the charges on the particles, e is the unit of charge, and r
is the distance apart of the centres of the spheres. When the charges are
expressed in esu, the proportionality constant k equals 1 because of the
definition of the electrostatic unit as that charge which repels an equal charge
of the same sign with a force of 1dyne when the charges are separated by
lcm. The potential energy is a quantity of greater importance than is force,
and is found by utilizing the fact that it is the work required to bring the
particles to the distance r from a separation of infinity.
O 7 25e" —2Z422e 2
Energy = | ee aS. (5.9)
r
i r
The potential energy expression is seen to be negative and thus can provide
the stabilization to overcome the positive ionization energy. At this stage, for
ion formation in the gas phase, we can summarize the overall energy equation
as

AH =1E.+E.A,—* 1%) (5.10)

The (+) or (—) before the E.A. term depends on the number of electrons
removed. In a simplistic way equation (5.10) can be used to consider the
possibility of ion pairs existing in the gaseous state above solids, by making
AH equal to zero and using the known values of I.E., E.A. and 2,, z,. If the
calculated distance of separation is reasonable then such species may exist in
the atmosphere.
Clearly we cannot stop here, because the Coulomb term is attractive, with
negative energy, and AH will become infinitely large exothermically when r
becomes very small. Therefore a fourth term must be introduced to represent
the repulsion of ions at short distances. The repulsive energy is usually
assumed to be given by an empirical expression, the most commonly
246 CERAMIC SCIENCE

encountered of which is B/r” or Be~?!". The first of these is known as the

Born Repulsion Term, and the second is a more recent term introduced to
reflect the wave model (2.2) rather than the hard sphere model of the atom.
Now we can extend the energy equation to
Za ane
AH =LE.+ EA. — ath Be~?"", (5.11)

This expression must have a turning point at which we can write 0AH/0r = 0
and use that fact to remove one of the unknown constants in the repulsive
term. The Coulomb and the repulsive term become, at the equilibrium
distance r, which is the turning point, equal to
—z,z,e? (1—p
le le

Equation (5.11) is now

2
MB LE+BA.~225 (1-2) (5.12)
e
iGe

Now (5.12) can be used in a more sensible way than equation (5.10). For
example, for a mole of NaCl we can write LE. = 495.3kJ, E.A. = —357.4kJ,
Z4,25 = 1, e = 1,602 x 10719C, N= 6 10** (needed to bring the third term
in (5.12) to mole terms), r,=0.251nm (this term can be found from
the turning point above), p = 0.035nm, and the constant 8.98 x 10° is
needed to convert from Asm~! to kJ. Substitution in (5.12) gives AH as
—337.8kJ mol”, which is a considerable lowering in energy for the system.
It is more significant when compared to the lowering of energy that would
occur from covalent bond formation arising from Na 3s interaction with Cl
3p, —83.6kJ mol”!. Furthermore, if the system is considered as Na —Cl*,
even though the electron affinity of Na is negative, the considerably increased
ionization energy term needed to produce Cl* would make AH positive.
Here we see the inherent strength of the ionic bond model in the simple
calculations that follow from the initial assumptions and allow one to explore
quickly a series of alternatives for a system.
If we were to repeat the calculation for MgO, then AH would have a value
around +368kJmol™~! because of the positive value for E.A., even though
the charge on the ions is raised to two. The vapour above MgO must
therefore not contain ion pairs, but consists of Mg +50,, which will produce
a limiting oxygen potential at a given temperature when a magnesia refractory
is used in the system. This will be true for all the oxides considered in this
chapter.
Continuing to develop this theme, we now see that the existence of some
ions can be justified in the gaseous state on energy grounds and not on the
magic electron number system. Unfortunately, for the materials of interest to
us, amphoteric oxides basic to industry, this is not the case, and we must take
 REFRACTORY OXIDES 247

the considerations further to prove what effect the structure of a solid has on
the possibility of the existence of ions.

5.3.4 Lattice energies**

It is immediately clear that the existence of a repetitive close-packed crystal
structure containing ions must have a large effect on the calculations done so
far, which were limited to pair interactions. In a solid, each ion will be
surrounded by several ions of opposite charge as near neighbours, which
means that there is in effect a magnification of the energy of attraction by the
number of nearest neighbour ions of opposite charge. For example, for ions
of the same size this could be six. To stop now, however, produces the wrong
answer, because at a not much larger distance will be a greater number of ions
of the same charge, which for a rock salt cubic structure would be 12 at a/or.
These would produce repulsion and net destabilization. Clearly we have to go
on examining the structure, finding at a somewhat greater distance again a
number of oppositely-charged ions which for the rock salt structure would be
8 at ./3r. Inevitably this process has to be continued in order to obtain the
correct multiplier for the attractive force between ions in a solid which will
reflect the part played by the structure in ion formation.
Each structure type produces its own series to be summed, and the
constant so obtained is called the Madelung constant (A) for that structure
type. The practical difficulties are that the series always decreases slowly and
presents mathematical problems that have to be overcome; some materials
contain complex non-spherical ions; and on occasion the real structures are
not known.
By and large the summation problem has been solved, and we can find
tabulated Madelung constants for most structure types (Table 5.3). Madelung
constants then show us how the existence of a structure can itself provide the
driving force for the existence of ions and, in the case of MgO cited above,
multiplication of the Coulomb term by 1.74756 brings the energy balance to
a favourable negative value. Clearly the large Madelung constant for the
AlL,O, structure plays an important role in determining the ionic nature of

Table 5.3 Some Madelung constants

Type A

NaCl 1.74756
CsCl 1.7626
ZnS 1.63806
ZnO 1.64132
CaF, 5.0387
TiO, 4.816
Al,O; 25.0312
248 CERAMIC SCIENCE

this ceramic and controlling the overall stability of the solid; it is able to offset
the large positive energy needed to produce Al** and O7~.
It has been common since the problem was first formulated by Born to
write for the lattice energy of a solid
2
+ NAe* 2125
E= + (repulsive energy)
r
where the first term represents the electrostatic attractions and repulsions of
ions, and the second arises from the repulsion of electrons at short distances,
N is Avogadro’s number and A the Madelung constant. At the equilibrium
distance r = r, when E becomes U, the lattice energy is

te
r
e re
dU ee
eee
At equilibrium
dr
— NAC? 72, NB
‘oO 2 ~ J (n+1)
Le re

B NAez,2z,
ree nr?
eee cies
Pe nr,

ee nr e

—NAe? 1
ay eee“i2a(1 2). (5.13)
ife n

Equation 5.13 is known as the simple lattice energy equation. So far all the
terms have been described, except n, which is called the Born exponent, and is
found experimentally from crystal compressibility experiments at relatively
low pressures. Although in general compressibility falls with increasing
pressure, the values of n are taken as whole numbers related to the ion
electron structure, as shown in Table 5.4. When necessary, these n values are
averaged, e.g. LiF = (5+ 7)/2 = 6.
Even though this is an expression for lattice energy arising from bold
assumptions and using only two adjustable constants, the answers it yields
are remarkably good for many materials. For example, NaCl has N =
6.02 x 107 molecules mol~', = 4.8 x 107
!°esu, r, = 2.814 x 107 !°m, n = 8,
which yields Uyjac} = —753.7kJ mol”* and can be compared to the experi-
mental value of —768.3kJ mol“.
A variety of extended equations have been proposed in an attempt to
 REFRACTORY OXIDES 249

Table 5.4 Values of the Born exponent

Ion Electron structure n

Lie He 5
BieNas Ne 7
Clank Gut Ar 9
Br, Rb*, Ag* Kr 10
Te CsunAta Xe 12

improve the agreement between calculated and experimental values. The

extended equations have terms added to the Madelung and repulsive parts,
and have the form U = —U,,+Upg—U,+Uz where U,, is the Madelung
attractive energy, Up is the repulsive energy term, Born or other, U, is the
London or dispersive term arising from what might be seen as a synchronous
vibration producing induced dipoles, and is proportional to r~°, and Uz is
the zero point energy; at 0°K it is not possible for all the vibrational energy
states to be in the lowest single vibrational mode of the crystal. The value for
this term is 2.26 NhVmax-
For ionic crystals containing transition metal cations, such as the spinels
(5.9), an important extra term has to be added, known as the crystal field
stabilization energy. It was noted in 2.2 that the d-orbital is five-fold
degenerate in the gas phase atom, but in a crystal the close proximity of the
anion X” and the directed nature of the d-orbitals lift this degeneracy,
making some of the d orbitals more stable relative to the degenerate free atom
state and making others more unstable. Now, depending upon the geo-
metrical arrangement of the anions and the number of d electrons in the
cation, the system can attain overall lower energies for some distributions.
When the cation has 0, 5 or 10 d electrons, this is not a source of stabiliza-
tion, but for other numbers of d electrons this can be between — 83.6 to
—209kJ mol~! depending on the type of anion;I~ < Cl” < F~ ~ O?~. For
4f, 5f elements the orbitals are not measurably affected, and incompletely
filled p orbitals in an ion lead to covalency (2.2). When the extended form of
the lattice energy equation is used, extremely close agreement between
calculated and experimental values can be obtained. Two examples are NaCl
and MgO.
Ue 02 U
NaCl 859.4 98.7 12.1 fel 765.7 kimole ”
MgO 4631 698.1 63 184 —3920.8kJ mol”!

These figures show the relative importance of the terms in the equation, and
also the large lattice stability of an array of Mg?* and O?~ ions in magnesia.

5.3.5 Kapustinskii equation

What if one does not know the structure and hence the Madelung constant,
or the structure has non-spherical ions??° Here again, the essentially simple
250 CERAMIC SCIENCE

nature of this bond model comes to our aid. It was noted that the difference
between the Madelung constant A for known crystals and for the rock salt
structure was compensated by differences between actual interionic distances
and those calculated from tables of six coordinate ionic radii, i.e. the
calculated value for an ion’s size if it was always surrounded by six
neighbours of opposite charge. Taking this into account led to the empirical
Kapustinskii equation:

Pieeoey oat (.- we

34 ) (5.14)
hearts LGA

Here v is the number of ions in the formula, e.g. three for CaCl, or Na,CO3,
and rc, r4 are the six coordinate radii of the ions in Angstrom units. This
expression has often proved valuable when used in conjunction with cycles of
the Born-Haber type to evaluate the possible existence and nature of new
materials. It has also given rise to a concept of thermochemical radii of
complex ions, as being those which such an ion would have in order to make
the Kapustinskii equation produce correct values of lattice energy; some of
these values are given below, but it must be emphasized that they have
significance only in lattice energy calculations. The Kapustinskii equation is
criticized for emphasizing (r4 +rc) when in large complex ions close-packing
in fact determines the interatomic distances. Nevertheless one is forced to
recognize that it works well in practice.
Calculated values from either the simple lattice energy equation, the
extended versions, or the Kapustinskii equation are inevitably compared to
experimental values obtained indirectly by a thermodynamic cycle—the
Born-Haber cycle—which involve the combination of a number of experi-
mental results.
The interaction of the Kapustinskii equation, Born-Haber cycles and the
ionic model is well demonstrated by their use in studying the technically
important system NaF-AIF;.** In this system, there has been controversy
with respect to the stability of sodium tetrafluoroaluminate as a product
arising from the dissociation of molten cryolite:
Na3AlF¢q) a= 2NaF tsNaAlF4q) :

Table 5.5 Thermochemical

radii

Ion Radius (A)

CrO2- 2.4
MoQ27 2.5
SiOZ- 2.4
OH- 1.4
CO?" 1.85
Ba(H,0)?* 1.61
 REFRACTORY OXIDES 251

This proposed dissociation is not supported from observations in the solid

state when conflicting reports describe NaAIF, or eutectics of chiolite,
Na,;Al3F,,4, and aluminium fluoride arising from the decomposition of the
tetrafluoroaluminate:

One approach to the problem is to calculate the enthalpy of formation of

NaAIF, and compare it to values for other compounds in the system. The
enthalpy of NaAIF, can be calculated, whether or not it exists at 298°K, from
the cycle

(AH3pA(AIFse
toes
ENA) y@) + L-E.(Na) +4D.E.F,)
2DeePa). Nag)
Nag) a 2F x¢) = Als) oe AIF3(g) Se Fg)
E.A.
AH? LE. | |
AIF 3g) + F¢g)
ee ee

U(NaAlIF,) FA.)
NaAIF 4) + Nag) + AIF 4@)

We need to calculate the lattice energy U by use of the Kapustinskii equation

and to calculate the fluoride affinity, F.A., of AIF3,which will require a cycle
of its own. If these two goals are achieved then, since all other values are
known, the enthalpy of formation of NaAIF, can be calculated. The first step
is to use the Kapustinskii equation and an estimated thermochemical radius
of the tetrahedral AIF; ion of 2.60A:
VZ425 0.345
Oe ~2672 724(1 )
koamol.
Gland rot,

With v = 2,z, =z) = 1,rc = 0.98A, then

U,, = 146kcal mol~' which is corrected to — 606.1 kJ mol” * at 298°K.
The second step is to estimate the fluoride affinity, which can be done through
the following cycle:

AH
NaFg + AIF 3g) +> NaAlF 4)
7 AH,
{ AH, { AH, F.A. (AIF . is

Following this cycle round, one can write

F.A.(AIF;) = AH,—AH, —AH,—AH,

where, from the literature, AH, = 635.4kJ, AH, =0, AH; = 317.7kJ, but
D2 CERAMIC SCIENCE

AH, has to be calculated from the Coulomb attraction and Born repulsive
term:

AH, = (1—p/po).
d(NaAIF,)
Using d = 0.310 nm, then AH, = —405.5kJ mol” ', which in turn gives
F.A.(AIF,) = —317.7—636.4+ 405.5 = —547.6kJ mol *.
We can now return to the first cycle:

— F.A.(arrz) — Unaalr)
AH¥. = —1212.24+108.7+ 75.24 493.2 — 334.4 — 547.6 — 606.1.

Using this calculated value, and literature values for chiolite, AIF; and NaF,
the following enthalpies of reaction are obtained:
Na3AlF¢q) = 3NaF) ao AIF 3) AH = 92 ae 4kJ

Na3AlF 6a) = 2NaF ) =F NaAlF 4q) Meow) ah 4kJ

Thus the most probable dissociation scheme for the tetrafluoroaluminate is to

chiolite and AIF,,which supports the observation of a eutectic microstructure
at 50:50 NaF: AIF3.
These calculations then support the form of the phase diagram shown in
Fig. 5.2.

Cryolite
994 1010

Liquid AIF.
a 3
695

N acAl.Aa + AIF,

mole % AIF,

Figure 5.2 NaF: AIF, phase diagram.

 REFRACTORY OXIDES D3

5.3.6 Polarization
Polarization to some extent must occur in all ionic systems, since we cannot
envisage that the close proximity of charged ions in an ionic crystal has no
effect on the negative electron waves in other ions. We see this as small
cations of high charge distorting the orbitals of the anion, with a resultant
apparent reduction in the overall positive charge on the cation and negative
charge on the anion. Small, highly-charged cations are said to have large
polarizing power. More subtle features can make some cations highly
polarizing; for example, ions with partly filled d-orbitals are more polarizing
because the poor shielding provided by the d-orbitals allows the nuclear
charge a more direct effect on the anion orbitals.

High polarisability + Large polarising power ———————» Polarisation of

the ionic bond

(1) (2) (3)

Decrease in Z+, Z— Decrease in interatomic Extreme

.. decrease in U as distance .. U leads to
% covalency rises. rises as % covalency rises. 100% covalency.

Polarization is the term used to describe the effect charge density has on
anions, and it has its greatest value for large anions with electrons in orbitals
defined by large quantum numbers. Because of trends (1) and (2) above we
can see that the model is self-compensating. To a certain extent, the model
can be improved by using one consistent set of ionic radii to calculate the
cation-anion separation and not using the measured values. There are several
sets of ionic radii because the real electron density on ions tends to infinity
(2.2), and therefore empirical methods have to be used to define the real radii.
Such radii reproduce with reasonable accuracy the observed interionic
distances found in a wide range of ionic crystals.

Figure 5.3 Determination of radius ratio.

254 CERAMIC SCIENCE _
Table 5.6 Radius ratios for polyhedra

Coordination Coordination Minimum radius

polyhedron number ratio

cubo-octahedron 12 1.00
cube 8 0.732
square antiprism 8 0.645
octahedron 6 0.414
trigonal bipyramid 5 0.155
tetrahedron 4 0.225
triangle 3 0.155

It is the calculated radii for ions that are used to carry forward the inherent
simplicity of the model into the area of geometry and structure, through the
concepts of radius ratio and polyhedral linking to form structures.

5.3.7 Geometrical aspects of ionic bonds

(i) Radius ratio. Ignoring the effects of polarization and concentrating on the
hard-sphere picture of ions having fixed radii places limits on the types of
geometry predicted for ionic structures. In Fig. 5.3 we can see that application
of simple geometry soon leads to relationship between the radius of the cation
r, and the anion r, if all the spheres are to touch the central cation.

(eGo
r= (/2-1)rg
= 0.414,
r./Tq = 0.414.
In this way coordination polyhedra are predicted and are listed in Table 5.6.
Close agreements with observed structures are obtained using this concept,
some examples of which are given in Table 5.7. Such comparisons clearly
establish the usefulness of the concept. It also alerts one to potential
polymorphism when radius ratio values are calculated to be close to the
minimum limiting values between polyhedral types.

Table 5.7 Observed coordination numbers

Cation Anion
Compound Radiusratio coordination coordination

BeO 0.22 ol 4
MgO 0.46 6 6
SiO, 0.29 4 2
GeO, 0.38 4 or 6 2 or 3
TiO, 0.49 6 3
SnO, 0.51 6 3
PbO, 0.60 6 3
CeO, 0.72 8 “A
 REFRACTORY OXIDES D5

(ii) Pauling’s rules.** Pauling realized the power of this concept when he
formulated a series of rules for determining the number of anions in a
polyhedron and the manner in which different polyhedra are arranged
relative to each other. Pauling’s rules are chiefly useful in discussing oxides
and fluorides in which the overall structure is determined by the packing of
large anions. Each positive ion in the structure is surrounded by a number of
negative ions, known as its coordination polyhedron.
Rule 1. The cation-anion distances are determined by the sum of the radii,
and the coordination number of the cation by the radius ratio. This rule leads
to the data in Tables 5.6 and 5.7 from which the coordination number 5 is
absent. There is no regular arrangement with minimum radius ratio between
4 and 6, therefore such an arrangement is rare in ionic crystals and only
occurs when the structure is primarily determined by another ion.
Rule 2. This rule implies that charges in a crystal are neutralized locally and
states that in a stable ionic structure the valency of each anion with changed
sign is exactly or nearly equal to the sum of the strengths of the electrostatic
bonds to it from adjacent cations. The strength of an electrostatic bond is
defined as the ratio of the charges on a cation to its coordination number. For
example CeO,, with Ce** and O? ", has a radius ratio of 0.72 with predicted
coordination number of 8 and a strength of electrostatic bond = § = 5.
Rule 3. The presence of shared edges and especially of shared faces in a
coordinated structure decreases its stability. This effect is large for cations
with high charge and small coordination number. It arises because of
increased cation-cation repulsions within the crystal as demonstrated in Fig.
5.4. For face sharing, A — B = 0.58r which increases the cation-cation repulsion
terms appreciably for cations with large charge.
Rule 4. In a crystal containing different cations, those with large valency
and small coordination number tend not to share polyhedron elements with
each other, since cations with large electric charges tend to be as far apart as
possible in order to reduce the repulsion effect in the Coulomb term of the
lattice energy equation. When for example the oxygen :silicon ratio equals 4
in a compound, then we would safely predict that silicon-oxygen tetrahedra
share no elements with each other (1.5).
Through the radius ratio concept, the recognition of polyhedra which

MH
O-7ir
r
Sd

yp €6
Ai iB

corner sharing A-B=r edge sharing A-B=0-71Ir

Figure 5.4 Repulsion due to polyhedra sharing corners and edges.

256 CERAMIC SCIENCE _

become the building blocks for ionic structures is a powerful aid to structural
analysis and property interpretation through the idea of structural families,
as already shown in 1.5. Structural families are essentially structures with the
same building units, octahedra, tetrahedra, cube etc., and the same linkage
pattern. Within families framework cations and cavity cations can be
identified. Framework cations form polyhedra, as defined by the radius ratio
with the anion, which then can share corners, edges or faces to produce 2- or
3-dimensional networks.
The type of sharing is indicated by the Pauling rules, particularly rule 4.
Generally the framework cations are all the same, but can be different as long
as the two cations have nearly the same size and electrostatic bond strength
(1.5.1). Clearly, when framework cations are of more than one kind, they may
occupy different framework sites in an ordered way or a random way, but the
ordered occupation is far more common.
Frameworks from octahedra sharing edges give rise to close-packed rigid
structures with small tetrahedral interstices only. Such a structure forms the
spinel family (5.9). When octahedra share both corners and edges a somewhat
more open structure is encountered, with important members such as the
rutile family (5.8) and the olivines (1.5.1). Corner sharing by polyhedra
produce more open and flexible structures with cavities of varying size and
shape each typifying a family, e.g. perovskite (5.10), tridymite and chalazite
(1.5.4).
The type of polyhedron forming the family is arrived at from the radius
ratio. Naturally we expect distortions to occur in practice, and even though
the agreement is usually very good, distortions to the polyhedra do occur for
the following reasons:
(i) Different framework cations of different size produce alternating polyhedra, each of which
has to adjust its bond length and angles to allow each cation to approach as near as
possible to its ideal values.
(ii) Large polarization can lead to substantial covalent character in the electrostatic bonds from
the cation, and hybridization of the covalent bonds (2.2) may lead to conflicting demands of
length and angle. ‘
(iii) Framework cations of high charge can become off-centred in the adjacent polyhedra in
order to minimize their repulsions for each other over the crystal as a whole. This effect can
give rise to interesting and useful electrical properties.
(iv) A crumpling of the framework can be achieved to minimize overall strains by tilting or
rotating the polyhedra when they are linked by corner sharing. This effect gives rise to a
considerable opportunity for polymorphs to exist within one compound, and perhaps the
classic case of this is the variety of forms found for SiO, (1.5.4).

Many framework families have cavities within the structure housing cations
or groups. Cavity ions must obviously be of the right size to fit the site
provided, which can bearestriction when the sites are small and the
framework electrostatic bond strengths are high. Spinels and olivines fall in
this class, while rutile has no cavity sites. Some corner-shared polyhedra
produce cavities so large that molecules can occupy them, e.g. zeolites (1.5.4).
Essentially the bonds between the framework and the cavity ions or groups
 REFRACTORY OXIDES QS

will be weak, which together with the large number of cavities that can exist
leads to very wide possibilities of disorder and composition variation. It is
here we see arising the apparently endless compositional variations in some
families of structures.
We will be using this basic view of ionic bonding and structure to consider
specific examples of basic refractory oxides.

5.4 Alumina, Al,O,

5.4.1 Sources
Although alumina represents about 25 % of the earth’s crust, it does not very
frequently occur in the free form for hydrolytic and other reasons already
alluded to. Firstly, Al?* is a small ion of high charge, close in size and
electrostatic bond strength to Si**. This makes it an ideal candidate for part
of the framework in the silicate family of compounds (the aluminosilicates).
Secondly, because of the coincidence in O—O distances in Al(OH), and
(Si,O,)?~ networks (1.5.3), it is “intergrown” with the layer silicates—
kaolinite and micas. Finally, where it has not undergone the above changes
it is subject to the hydration reactions which we have used to classify oxides.
It follows from this that the commonest sources of alumina are hydrargillite
or gibbsite (Al(OH),, frequently written incorrectly as Al,03; -3H,O) which
represents complete hydration; bauxite, Al,O(OH),, often written as
AlL,O, -2H,O; diaspor, Al,O; -H,O, or more correctly AIOOH. These forms
represent different stages of hydration and frequently occur mixed together.
Bauxite is the principle ore worked for alumina recovery. By 1977, world pro-
duction of bauxite was over 920000 tonnes with over two-thirds of this coming
from Guyana and the remainder from Surinam, America and China. Since
demand is likely to exceed supply by over 200000 tons in the coming years,
some areas of low-alumina bauxite in Brazil may be used, but a more
immediate prospect is development of the huge deposits in China. There are
anhydrous deposits in parts of Africa mixed with up to 20% of SiO, and
Fe,O3. Such material is known as emery, which is often used without pre-
treatment. All these sources are far from Europe, a fact that has resulted over
the years in the development of methods to win Al,O, from its commonest
source here, that is from kaolinite (1.5.3)

5.4.2 Alumina manufacture

Processes differ according to the type of raw material available, the relative
arts of processing chemicals, and energy sources in different countries.
Examples are:
(a) Bayer process. Most of the relatively inexpensive alumina made from
bauxite is being refined each year by this process at costs currently ranging
258 CERAMIC SCIENCE .

from 7 to 20p per kg, depending on the degree of purification and grinding
required by the purchaser. The process consists of five stages: raw material
preparation, digestion, clarification, precipitation, and calcination (5.15).
Equation (5.15) summarizes the reactions.
4 atmospheres
Al,O(OH), ++ NaOH + H,O eS sodium aluminate in solution
160—170°C
+waste (5.15)

The separated sodium aluminate liquor is concentrated to saturation, and

after seeding with a charge of fine gibbsite, a heavy precipitate of Al(OH), is
produced by hydrolysis for subsequent calcining. The conditions of seeding,
temperature, agitation and time during precipitation are controlled to obtain
gibbsite most economically and to control the particle size, texture and
purity. It is these properties that ultimately control the quality of calcined
ceramic-grade aluminas. Final calcination is achieved at temperatures around
1200°C to convert gibbsite into ceramic grades of alumina. The principal
impurities removed at the digestion stage are iron, silicon and titanium, most
of which are discarded to mud lakes.
(b) Pyrogenic process. Because of the high energy input and relatively
expensive chemicals this process has been largely replaced by the Bayer
method. Bauxite ore is processed before reaction with Na,CO, powder at
1200°C. The ore is converted to a series of sodium salts with the evolution of
CO,. Treatment with water takes the sodium aluminate and sodium silicate
into solution and produces a precipitate known as red-mud from the
hydrolysis of sodium ferrite. The solution is then treated as in the Bayer
process to obtain alumina.
Recently a process has been described which might make this method
cheaper by using the red-mud waste. An efficient, low-cost sound-absorbent
product which is pleasing in appearance, resistant to the spread of fire and
suitable for outdoor use can be made by binding calcined red mud grains with
powdered waste glass at 950°C.
(c) Peniakoff method. This gives less pure aluminas but by utilizing an
exothermic reaction reduces the energy costs. Once again, bauxite is dressed,
but then mixed with Na,SO, and coke whereby heating to 900—1000°C
produces an exothermic reaction, evolution of SO, and CO and production
of sodium salts for treatment as in the above methods. European variations
of these preparations are based on clay as the raw material.
(d) Clay method. Kaolinite is heated to 500°C to effect the transformation to
metakaolinite (1.5.7), and treated with either HCl] or H,SO, to effect solution
of the alumina. Concentration and treatment of the liquor follows the above
methods. Although two moles of silica waste are produced per mole of
alumina, the clay and the acids are relatively cheap.
Al,Si,0,(OH), > metakaolinite + H,SO, > Al,(SO,); soln.
(e) Chemico process. Purer forms of alumina can be made from clay this way.
 REFRACTORY OXIDES 259

Kaolinite is reacted with ammonium hydrogen sulphate to produce an

ammonium-alum salt. The alum is soluble but can be crystallized from
concentrated solution any number of times until the desired purity is reached.
Finally, the alum solution is treated with ammonia to produce gibbsite. In all
these processes the washed aluminium trihydroxide is calcined to produce the
alumina. The loss of combined water is very temperature-sensitive at lower
temperatures. About 2.5 moles of water are lost before a plateau dehydration
rate is reached at 350—400°C. In the plateau region Al,O, forms, without
much shrinkage, to create a porous structure with high surface area. These
powders are known as activated aluminas and have many uses in drying gases,
chromatography, and as catalyst carriers.”7”
Development of «-Al,O, starts in the range 900—950°C, but temperatures
of 1100-1150°C are needed to provide powder with sufficiently low H,O
sorption to be used as metal-grade alumina. Complete conversion to a-
Al,O3, the phase of most interest to the ceramics industry, is essentially
complete in 1 hr at 1200—1300°C, but temperatures up to 1650°C are required
to obtain powders with 2.511m mean particle size. This temperature can be
brought down by use of halide or boron mineralizers. Boron has the
advantage that it produces crystals up to 15 um as well as removing sodium
as volatile sodium metaborate. Temperatures above 1800°C are needed to

Table 5.8 Alumina grades

Mean crystal Na,0 %

size ee
(um) <0.15 0.15—0.40 > 0.40

<1.0 A-16 A-13 A-1

A-3
OKs A-14 A-12 A-2
A-15
A-17
SOD A-10 A-5

Notes:
A-1 is a low-calcined- product for addition to
glass, glazes, fibreglass and to increase Al,O,
content of fireclays.
A-2 is used to improve heavy china and electrical
porcelain.
A-3 being uniformly fine is used as a metal polish.
A-5 is also used for grinding and ceramic
compositions.
A-10 has low sodium content for use in spark plug
insulators.
A-12 is predominantly used in spark plug
manufacture.
A-13 is used for high strength catalyst carriers.
A-15 represents a compromise between low shrink-
age and maximum reactivity and as such is used
~ in general production of high-alumina ceramics.
260 CERAMIC SCIENCE

produce large plate-like crystals several hundred microns in diameter, and

such stable material is known as tabular alumina. Stability rises as the surface
area decreases (see 2.1).
Commercially-available materials are listed with respect to soda content
and crystal size for discussion of the potential uses. Table 5.8 gives the
grading and the notes below it list some typical uses.

5.4.3 Structure
Mainly because of the proliferation of preparative methods and the range of
calcining temperatures, many forms of alumina have been described. It is
clear now that these mainly refer to partially dehydrated and impure forms
that are basically ternary oxides. Thus if we restrict our thoughts to pure
alumina, there is only one crystal modification, «-Al,O3;, which has a
hexagonal structure with DS, space group and two Al,O, molecules per unit
cell. Other forms such as B-Al,O, are ternary oxides that represent the effects
impurity cations can have on the basic hexagonal structure of alumina. If we
confine our thoughts to Al,O, as an ionic solid, then from the radius ratio
0.36, we would expect six-fold coordination round the Al?* ion, but the value
is clearly near to a boundary zone (5.3.7). Furthermore, the low melting point
arising from a material with +3 and —2 charges on the ions and a Madelung
constant as large as 23.04 (which should lead to large crystal lattice energies)
should alert us to the fact that all is not so simple. Hence the structural
description, as close-packed O?~ ions in an hexagonal arrangement with
two-thirds of the interstitial octahedral sites occupied by Al** so that each
Al?* has six nearest O*~ neighbours and each O?~ has four Al** disposed
towards four of the corners of a trigonal prism, must be an oversimplification
because it suggests too symmetrical an arrangement. Al?* is small and highly
charged and must therefore provide polarizing power necessary to distort this
simple arrangement. There is sufficient polarization to view this structure as
far from ideally ionic and possessing considerable covalent character. Using
Pauling’s electronegativity values, 37 °% covalent character is predicted.
A more rewarding model of the structure is obtained by considering Al,O,
units as three oxygens forming an equilateral triangle with aluminiums above
and below the centre of the triangle. One of these groups is placed at each
corner of a cube while another is placed at the centre of the cube giving the
two formula units. Now the cube is stretched along the [111] direction, the
body diagonal, until a rhombohedral angle of 55° 17’ and a unit cell of
0.542 nm are obtained. This emphasizes the distorted nature of the structure
and gives rise to the distorted arrangement of oxygens around each aluminium
as shown in Fig. 5.5, where the closeness of three oxygens to each Al3* is
demonstrated.
The structures of the other so-called forms of alumina, y and f, will be
described separately when dealing with these materials, in order not to lose
the point that there is only one form of the pure material. There is no
 REFRACTORY OXIDES 261

05 > 1 0-50Nae \
> el ob sae Hey nm

& PO, 2, |
oe 7 1 Py %e 0137 nm
o” ee i
1 Y

Figure 5.5 Oxygen arrangement around Al?* ion in Al,O,.

polymorphism of pure alumina. It is sufficient at this stage to say that

y-Al,O3 precedes the «-form during calcination of gibbsite, while B-Al,O,
results from reacting Al,O, with Na,CO, around 1100°C. The alumina types
A-1 to A-15 described above contain a, B and y forms of alumina.

5.4.4 a-Alumina
From the description of the structure, it is not surprising to find that because
of the partial covalency arising from the polarization this is not a very
refractory material (melting point 2050°C). In addition, the vapour pressure
rises rapidly from 0.133kNm? at 2148°C to 101.325kNm ? at 2977°C to
give a boiling point of 2980°C and a vapour containing various aluminium-
oxygen species.
(a) Strength of alumina—strength and structure generally.‘®?! The structure of
a-Al,O, does have implications for the mechanical properties of the material.
The key mechanical property is strength, o, but other important properties
such as fracture toughness,’ effective surface energy, and work of fracture
are important (see 3.4), and are related directly to strength. Thus for the
whole class of oxide refractories one can make substantial progress in
rationalization by concentrating on an equation of the type of (5.16), which
is itself a modification of (3.32).

bie (=)\1/2 (5.16)

y €

where y is a geometrical constant, E is Young’s modulus, (2Ey;,)"/” is equal to

the stress intensity factory K;c (3.4.3) and y,; equals the effective surface
energy. Also related to y; is the work of fracture y, which is the energy needed
to produce unit area of fracture surface. Because ceramics are brittle and
262 CERAMIC SCIENCE

catastrophic failure is common once the fracture stress has been reached,
Vp le ;
The mechanical behaviour of ceramics is affected by two main groups of
factors.

Material parameters Environmental parameters

Composition Temperature
Crystal structure Atmosphere
Specimen shape and size Strain rate
Microstructure: pore size Static or cyclic fatigue
grain size, cracks, surface condition _— State of stress
Internal strains

An advantage of using equation (5.16) is that it allows strength to be

discussed in terms of two key variables, effective surface energy and flaw size,
and hence generates an understanding of strength in terms of microstructural
variables. By ignoring y; all-encompassing relationships between flaw size c
and grain or pore size have been sought. This is not a wise aim, because it
belittles the role of y; and presupposes that (in all but some special
circumstances) such all-encompassing relationships exist.
(b) Strength and surface energy. For a given material, y; can be measured
reliably by a number of techniques. However, because of the great variations
in purity, grain boundary types, and heat treatments, it is usually useless to
consider data for similar materials from different origins. Alumina presents
one of the best self-consistent data sets from experiments using notched
beams (Fig. 3.14a), where all the fracture paths were intergranular. In other
materials this is not always the case, for usually a material shows intergranular
failure at small grain sizes and some grain cleavage fracture at coarse grain
sizes. Al,O,; data show a fall in y; from 45Jm~? at about 1 um grain size to
below 30Jm~? with 20pm grains. In contrast, y; has been reported as
increasing in Lucalox alumina (a very high-density material), as the grain size
increases. In the first case, the fall is attributed to residual strain energy in
large grains due to anisotropic thermal contraction on cooling, which cannot
be relieved by slip, in turn producing high stresses near grain boundaries. In
the Lucalox case, the increase in ;; is associated with increasing amounts of
transgranular fracture as the grain size increases. As a result of these and
other findings it has been observed that numerous factors affect the grain size
dependence of y; and so no general conclusion can be drawn, and it seems
likely that no estimate of surface energy can be made for some time from
microstructural information alone.
One generalization can be made about y, and grain size: atomic bonding
in grain boundaries is weaker than in grains as a result of porosity being
concentrated at grain boundaries, and because in grain boundaries around
larger grains there is a concentration of impurity and porosity compared to
small grains. Therefore we might expect a difference as the mode of fracture
changes from inter-grain to trans-grain patterns. Thus in materials where
 REFRACTORY OXIDES 263
Al,O,+ 5% SiO, ; \

400 |
!
| |
| |
Fracture | \
stress | \

10° Nm@ | \
|
|
|
\ |
100 Region A | Region C
|
Region B |

1050

Test temperature °C

Figure 5.6 Strength v. temperature for Al,O3.

there is such a transition with increasing grain size, y, should increase, but
where the fracture remains intergranular it should decrease with grain size.
From equation (5.16) the strength should then vary appropriately, and for
increased strength in use, a transgranular fracture pattern should be sought.
In order to achieve this, some form of grain boundary strengthening or strain
energy relief mechanism must be sought. In the case of Al,O3, this is often
done by the incorporation of up to 5 &% silica to form a glassy binding phase
in the grain boundaries which can act via viscous flow to relieve stress, to
blunt cracks, and through strong glass-ceramic bonding (3.5.2) encourage
transgranular cracking. This is effective at temperatures up to 900—950°C
(Fig. 5.6).
If Al,O; is to be used as a refractory, however, this solution to the strength
problem will not do.
(c) Strength and flaw size. Although y; can only be determined by direct
measurement, there is a very close relationship between the value of c to be
used in equation (5.16) and microstructure. There seems little doubt that the
important microstructural features are grain size and pore size with the flaw
size approximating to the largest microstructural feature. For alumina and
other fully-dense ceramics, this will be the largest grain size; for ceramics with
few small pores it is still the largest grain; for structures with pores and grains
approximately equal in size it is one grain plus one pore long; and for a very
fine-grained material with a few relatively large pores it is equal to the largest
pore size. Figure 5.7 shows clearly that the fracture origin in an Al,O, sample
is large crystals.
In general, the state of the specimen surface must be considered, for normal
machining tends to produce grain-size cracks in the surface region and as a
264 CERAMIC SCIENCE

|-——| 10 pm
Figure 5.7 Fracture initiation at large grain sizes in dense Al,O, rod. From H. P. Kirchner and
R. M. Grover, Phil. Mag., 27, 1433 (1973).

result specimens with polished surfaces tend to be stronger. Where a material

can be shown to have y, independent of grain size, then linear relationships
between strength and (grain size) ‘/* do arise as will be seen in the case of
MgO (5.6). .
Alumina is extremely brittle since the dislocations are immobile, grain-
boundary sliding does not occur, and the material exhibits anisotropic
behaviour for both elasticity** and thermal expansion. As a result of this,
even the densest materials when subjected to tensile stresses fail at the grain
boundaries. Even the transparent, fully-dense Lucalox material fails inter-
granularly, but when large grains are present some degree of transgranular
fracture can be achieved with a resultant increase in the measured ); and hence
fracture strength. Figure 5.6 shows the average values for the strength of
alumina and indicates a decrease in strength with temperature rise. Two
opposed features produce this relationship. First, plastic flow stress, the
energy needed to move dislocations on active systems, decreases with
increasing temperature; secondly, the stress required to extend inherent flaws
increases with temperature from the fact that the surface energy term in
 REFRACTORY OXIDES 265
equation (5.16) increases with temperature. The resultant curve can be
divided into three sections.
Region A. This is common to all ceramics at low temperatures and is the region where fracture
is brittle. Behavioural interpretation is satisfactory here in terms of Griffith brittle fracture theory
(3.4). All the foregoing discussion is relevant to region A where the important considerations are
surface energy and inherent flaw size. Control and strength improvement here lies in extrinsic
variables relating to microstructure, preparation, and handling.
Region B. This is a higher-temperature region of more interest in refractory science. Failure is
analysed to be semi-brittle and the strength is found to be less than that predicted by equation
(5.16). The explanation now lies in intrinsic properties of the material, for the plastic flow stress
drops below the strength required to extend inherent flaws after some critical temperature is
reached. Dislocation flow without complete plasticity then causes generation of some more
damaging flaw in or near a grain boundary. Other possible modes of critical flaw generation in
this region are grain-boundary sliding, slip within a grain, or twinning. The characteristic stress
necessary for the relevant mechanism to operate defines the fracture stress in this region, and
improvements in refractory strength must focus on identification of the failure mechanism before
prescribing means of overcoming the problem.
Region C. At temperatures near the melting point, the ceramic fails in a ductile manner with
considerable plastic flow taking place in this region due to macroscopic yielding of the material.

Arising from the strong ionic bonding and the distortion in the structure
due to polarization, Region A type behaviour is found in Al,O; up to 1050°C,
and above this there is evidence that the mechanisms of Region B are
important. No Region C has been seen for alumina. It is perhaps because of
the need to identify the failure mechanism that considerable interest has
centred on slip systems in Al,O3. Basal slip has the lowest critical resolved
shear stress with possible slip systems: (0001) 4 <1120), (0001) <1010), but
controversy exists as to other possible slip systems.*° Those slip systems
reported to date are given in Table 5.9.
The slip planes in Table 5.9 are also the most commonly observed crystal
faces in corundum, but because naturally-occurring crystals usually have
faces that are most densely packed, and slip usually occurs on the densest
packed planes, this is not unexpected. The common approximation that the
self-energy of a dislocation increases as the square of the Burgers vector does
not necessarily hold in anisotropic materials, so it is not easy to rank the
probable slip systems. With so many reported slip systems it might be
surprising to find that polycrystalline alumina does not have behaviour in
accord with Region C on the strength versus temperature behaviour curve.

Table 5.9 Dislocation slip systems in Al,O,

(0001) $<1120, (1070

{1010} 4<i2T0) | te 3
{1120} 41101), 4<2201), <1100),4 <1102)
{0112} 4<2T10),4<0111),4<2021)
mt

{1011} 4<1210),4<1101)
{2113} wl41101), <0110)
{2112} <0110), 41012)
{4233} <0110)
266 CERAMIC SCIENCE

For a polycrystalline body to deform plastically by dislocation glide, each

grain must be capable of undergoing a perfectly general strain in order to
match shape changes in its neighbours. A general state of strain requires five
independent components of the strain tensor. Von Mises recognized that one
independent system in a crystal yields a single component of the strain tensor,
as a result of which five independent systems comprising a slip plane and a
slip direction are necessary for a polycrystalline body to undergo general
strain. A slip system becomes independent if the strain produced by its
operation cannot be arrived at using a suitable combination of glide on other
systems.
It has been shown that slip on {1120} prism faces in ¢1100> or <1101)
direction in combination with basal slip gives rise to only four independent
systems. Other systems involving basal slip and pyramidal slip on (0001)
{1120} and {1102} {1101}, or {1011} (1011), or {0112} ¢2021)}, will all give
rise to five independent deformation modes, so why do we not observe
Region C type behaviour even at temperatures up to 1800°C? The answer
seems to lie in the fact that the critical resolved shear stress in the pyramidal
systems is so much higher than in the basal system at all temperatures, with
the result that basal slip will always, in a polycrystalline sample, be initiated
first and then give rise to stress concentrations at obstacles to glide, resulting
in crack nucleation and intergranular failure. One can conclude from this
that only under high hydrostatic pressures that are able to prevent inter-
granular crack development will alumina be able to exhibit high temperature
plasticity and a Region C in its strength v. temperature behaviour.
To summarize, for «-Al,O, the strength is controlled by pre-existing flaws
up to 1050°C and by basal slip-generated flaws interacting with grain
boundaries above 1050°C. Failure is in the main intergranular. Improvements
can be achieved by improving intergranular cohesion by using glass-forming
impurities. This, however, decreases the strength at refractory temperatures
by causing extensive grain boundary sliding. For very dense bodies the
strength may be improved by treating the surface, which usually contains the,
critical flaws (3.4.2). This can be achieved by building a chemical compound
on the surface that has a more open structure than Al,O,.”” It will have a
larger volume and so compress cracks (3.4.8). Alumina is reacted with
powders at 1250—1350°C according to the equations below:

CaO +2Al,0, > CaAl,O, AV =31%

CaO + 6Al1,0, a CaAl,,O;5 AV = 15.4%

2Si0, + 3Al,0;3 > mullite (1.5.1) AV= 9.7%.

The AV percentage is the relative change in unit cell volume of the product
compared to the alumina.
Alumina rods so treated can be strengthened by as much as 65 %. However
 REFRACTORY OXIDES 267

Tensile
stress
Distance from
rod axis

Compressive
stress

Figure 5.8 Stress distribution within a surface-treated Al,O, rod.

it must be emphasized that such an improvement does depend on the

uniformity of the microstructure, because (as Fig. 5.7 demonstrates) unlike
glass, strength-weakening flaws can exist inside the ceramic body, and if we
set up a surface compression then there must be an internal tension which can
cause flaws to become critical. Figure 5.8 gives an indication of the magnitude
of the stress profile in surface-treated alumina rods.

5.4.5 The ruby laser—a new use for an old material

Here is a use for Al,O, that arises directly from the distorted structure when
it is made to contain a few per cent of the transition ion Cr?*. In the distorted
octahedra of O7~ ions, the Cr3* d orbitals cannot have equal energy because
of their directional properties (2.2), the anion arrangements and the anion

4P or as
4
7; 2
7
TA |
/ Ty 2 ‘

4 / ee)

4E fe ae OS.
4 fs
nee
aN

+ +
(a) Isolated cr ions (b) cr? in Al,0,
in gas phase

Figure 5.9 Schematic representation of the Cr>* energy levels (a) in isolated state, (b) in Al,O3.
268 CERAMIC SCIENCE

charge. Thus the simple energy level diagram (Fig. 5.92) becomes the more
degenerate one (Fig. 5.9b). Two parallel energy level diagrams are shown
(Fig. 5.9a,b) because Cr>* ions can have two possible total spin states. If the
three d electrons all spin parallel, +A,, or if one spin is opposed, 7. Transfer
between +A, and 7E is very difficult, requiring spin reversal, and is termed
“forbidden” although it does occur occasionally. Transition by energy
absorption from +A, to *T, requires 3.5 x 10° *° J, equivalent to 2 = 556nm
in the green region, making ruby appear red.
From the excited state, Cr?* ions can return to the ground state either by
spontaneous emission or stimulated emission, the former greatly exceeding
the latter. It is the electromagnetic field of the incident light that promotes the
stimulated emission and to have an amplification of light one must have more
ions in the excited state than in the ground state, which is normally an
unrealistic expectation. However, in the Al,O, structure a mechanism exists
whereby interaction of Cr** ions with the lattice through the nearer three
oxygen ions in the coordination sphere absorbs energy with spin change to
become the 7E state. Because the 7E to *A, transition is “forbidden”, a build-
up of excited ions occurs relative to the ground state *A, that represents a
population inversion. The energy gap *E > *A, corresponds to a wavelength
694.3nm in the red part of the spectrum, so that radiation with this
wavelength promotes the stimulated process and a burst of 7E Cr?* ions
return to the +A, state, providing amplification of the stimulating wavelength
of 694.3 nm. This is laser activity (light amplification by stimulated emission).

5.4.6 fb-Alumina
(a) Structural details and preparation. At one time these materials were
thought to be variants of the a-Al,O, structure with a variety of transforma-
tion temperatures at which there was a change to the «-form. Some degree of
metastability was indicated by the fact that the transformation was not
reversible. In due course, through experiments such as heating f-Al,O, in
silica tubes whereby Na,O could be removed easily, it became apparent that ,
a series of ternary oxides were encompassed by the term f-Al,O,. Bragg in
1931 determined the main features of the structure, and in 1937 Beevers and
Ross elucidated more fully the nature of the all-important mirror plane.
The ideal formula is M,O-11A1,0,, and the most common member
encountered is when M is sodium. Some thermodynamic data exist for the
reaction
Na,O iat 11Al1,0, =< Na,O i 11A1,0,

(AG° = —22.0 x 10*—0.57T +5.5 x 10° J).

Structural refinement was achieved by using a crystal containing 29 % excess

sodium over the ideal formulation, which demonstrates the non-stoichio-
metric nature of these materials. Electrical neutrality is arrived at either by
Al°* vacancies or extra O*~ ions in the mirror plane of the structure. All
 REFRACTORY OXIDES 269
Mirror plane

eee 0:476nm

Sue 128m
oe Seta
(a)

Oxygen in
spinel block _T-.

yy $8
®
O
sae Senet” Ss.” Oxygen in
mirror plane
@
0a GZ02062
(b) (c)

Figure 5.10 Structure of B-Al,O3, showing (a) a section perpendicular to the mirror plane,
(b) Beevers-Ross sites (O) and anti Beevers-Ross sites (]), and (c) a section in the mirror plane.

other compositions can be made easily by ion exchange of the Na* to get
lithium, silver, potassium etc., B-alumina, in which the c-axis expands or
contracts slightly and the a-axis changes very little. Furthermore, by refluxing
in concentrated H,SO, for 2 days, Na*-BA1,O, can be changed to H;O0*-
BAI,O3.
The structure is basically hexagonal P63;/mmce. Each unit cell has perpen-
dicular to the c-axis two mirror planes of O?~ and M"* ions separated by a
spinel-like block formed by four layers of close-packed O*~ ions (Fig. 5.10
and section 5.9).
Since electrical conductivity does not occur through these layers, attention
is best focused on the mirror plane structure. The O?~ in the mirror planes
which are bonded to Al°* ions in the spinel blocks above and below link
them together and form a non-close-packed layer. In the ideal composition,
they form an extended hexagonal network in two dimensions within which
two types of site for the M”* ions are identified and themselves form
hexagonal networks. These sites are known as Beevers-Ross and anti-
Beevers-Ross sites, with the latter lying immediately above an O?~ in the
spinel layer and the former lying over an interstice in that layer: they differ in
energy by some 100kJ mol” *. The unit cell is 2.26 mm long in the c direction
and involves two spinel blocks related by a two-fold screw axis.
270 CERAMIC SCIENCE

Solid state mixing of sodium carbonate and a-alumina followed by firing

at 1100°C is the standard way of preparing sodium $-alumina, and this is a
precursor for preparation of other f-aluminas by ion exchange. The instrinsic
problem with this method is the volatility of Na,O, which makes the whole
process variable with respect to composition control. New methods of
preparation and fabrication*® are being sought to overcome these prob-
lems!7:3° and these are mainly based on organic procedures. For example,
starting from AICI,, alcoholates can be prepared that are subject to hydrolysis:
Al(OC3Hg)+C3H,OH =slurry (stir continually)
Add dropwise
NaHCO, (soln.)
or
KHCO, (soln.) > thick gel dry powder
or
Mg(OOCCH;), (soln.)
After milling, the powder is heated to the lower temperature of 950°C. On
route, at 350°C NaAlO, appears, then y-Al,O3, and at 950°C ”-Al,O, forms
depending upon the hydrolysis agent used. This method does ensure good
mixing and lower alkali volatility losses.
To facilitate shaping*® and membrane formation the powders from the
above preparations can be crushed under organic liquids and electro-
phoretically formed :**
p 200000" B- + Na? in N-amyl alcohol with no stirring.g
(b) Polymorphism of B-Al,03. Viewing once again the structure sections
shown in Fig. 5.10 we can see that variants of this basic unit are easily arrived
at by the following routes:
(i) Change in the number of O?~ layers in the spinel blocks.
(ii) Increases in the number of blocks along the c-axis.
(iii) Change in the screw axis relationship between neighbouring blocks.
(iv) Replace Al,O; in the blocks by other M,O, oxides.

Clearly many of these variations will be found when sought, but some already
exist and are adding to the confusing nomenclature f’, 6’, p’” etc. Examples
are 6” which is M,O-xAl,O, where x = 5 to 7. It has three spinel blocks
related by a three-fold screw axis. Investigation is showing that samples of
B-Al,O3 usually consist of mixtures containing f’-Al,O3. When M?* is
available, the 6” structure appears to be stabilized and since Li,O also seems
to favour the f” structure, this may reflect an ionic size effect. The p” has a
greater ionic conductivity but lower mechanical strength.
B’-Al,O, has a mirror plane identical to f but separated by spinel blocks
containing six oxygen layers to give Na,O -4MgO- 15A1,03.° The formula
indicates the mixed cation nature of this phase. A recent report of B”” has
the formula 0.8Na,0-2.4MgO-7A1,03, and, with the same six oxygen
 REFRACTORY OXIDES Dae
layer spinel structure, demonstrates the non-stoichiometry of these phases.
Analogous to f and $”, but exemplifying variation (iv) in the above list, are
K,0-11Fe,0, and K,0-5Fe,0; - PbO - 6Fe,O, is similar to B except that
the mirror plane is fully packed with adverse consequences on the electrical
conductivity.
Finally, this idea of simple structural variation leads to a connection with
the technologically important ferrites (5.11), because Braun’s structure “W”
for BaO-FeO-Fe,O;, consists of two six-O?~-layer spinel blocks separated
by a fully-packed mirror plane.
(c) Electrical conductivity.*° The structure shows that the conductivity is
ionic, with mobility perpendicular to the c-axis and confined to the mirror
planes. The rather wide separation of the mirror planes and constraint to 2-D
ion motion reduces the volume that transports the M"* ions, but nevertheless
good characteristics can be achieved with 399° = 4ohmcm perpendicular
to the c-axis, and measured activation energies for mobility around
15kJmole~*. The high ionic conductivity with unusually low activation
energy for cation migration has attracted interest.?° In an obviously ionic com-
pound, theoretical approaches have utilized the concepts of the ionic bond
model. In such calculations all ions except the monovalent mirror plane ions
are assumed to be fixed in equilibrium positions determined by X-ray analyses,
and the M* ions are constrained to move along straight lines joining neighbour-
ing Beevers-Ross and anti-Beevers-Ross sites, i.e. along the edges of a hexagonal
network. Ideal composition M,O-11A1,O0, is assumed, with the M* ions
completely occupying the Beevers-Ross sites. Then the total crystal potential
energy is calculated as a function of the position of 1 M™* as it moves along
the line to the nearest anti-Beevers-Ross site. This procedure gives a total
potential energy increase of about 192kJ mole‘ for all B-aluminas which is
clearly much larger than the measured activation energy. A repeat calculation
can be made after including an extra M* ion at an anti-Beevers-Ross site,
with the M* ion nearest this extra one being allowed to move while other
M* move simultaneously to keep the total potential energy at a minimum.
To obtain a minimum, the ion at the anti-Beevers-Ross site has to be paired
with another M* and the ion pair moved in phase. The potential energy
barrier is then found to be close to the 15kJ mole‘ found by experiment.
The implication of these calculations is that in the stoichiometric structure
M"*, ion diffusion is expected to be very slow due to the large potential
energy barrier. Even when there is a deficiency in M””, the diffusion will be
slow as the barriers for vacancy transport will be similar. However, when
there is a slight excess in M”* concentration, then diffusion will be easy due
to the paired mechanism and electrical conductivity will be proportional to
the concentration of excess M”*. Some incidental findings in such calculations
indicate that high polarizability of the charge carrier ion is beneficial to
conductivity, but large polarizability of fixed oxygen ions in the mirror plane
is detrimental.
Die. CERAMIC SCIENCE

5.4.7 y-Alumina
Reasonably crystalline samples of y-Al,O, can be made by adding NH,
solution to stirred Al(NO;); solution, filtering the precipitate, washing, and
drying at 80°C before firing at 915°C for one hour.
y-Alumina is the name given to a number of phases that arise during the
decomposition of gelatinous Al(OH); and boehmite, AlO(OH). When sintered
they transform from y— a, producing a microstructure that consists of
micron size a-grains with a high degree of interconnected porosity. This
suggests that the reported rapid sintering rate of fine y-aluminas,’ although
attractive, may not lead to improvements in the sintering route to alumina
products.
The y to « transformation is however of continuing interest, for y-Al,O, is
widely believed to be a defect spinel in which protons play a stabilizing role,
and intermediates with closely related structures appear in the commercially
important reaction whereby kaolinite is changed to mullite (15:7):

5.5 Zirconia, ZrO,

5.5.1 Sources and manufacture

In the free form as baddeleyite it occurs in substantial deposits in Brazil,
where an ore consisting of 80% ZrO, contaminated by Fe,O3, SiO, and
TiO, occurs as water-worn pebbles and boulders in stream gravels. This is
often used with little further dressing.
The commonest source is zircon ore, ZrSiO,, widespread in all granites
(1.5.1). It is usually one of the first minerals to crystallize out of the magma,
but zircon (being very resistant to weathering and heavier than most of the
associated minerals) becomes concentrated in parts of the alluvial deposits
with other heavy minerals such as monazite, ilmenite and rutile (5.8). The
most productive source of zircon mineral is the black sands at Byron Bay,
Australia, where it occurs with rutile and ilmenite minerals.
Occasionally large zircon crystals of gemstone quality can be found which
exhibit wide variations in their specific weight. One of the reasons for this
variation demonstrates a commonly-encountered impurity in ZrO,. The next
group IV element below zirconium is hafnium, which should be about 14%
larger than Zr. However, intervening between these elements is the first
f-transition series as the 4f energy levels are occupied from lanthanum
through to lutetium. Because of the high angular momentum of f orbitals, the
shielding of the outer electrons from the increased nuclear charge is poor,
thus as the atomic number increases, the atomic size decreases along the
lanthanide series. This effect is known as the lanthanide contraction and
amounts to about 14%. Hence by chance Hf is equal in size to Zr, and HfO,
or HfSiO, are then considerably denser than their zirconium counterparts;
 REFRACTORY OXIDES DS
zirconia Or zircon mixtures containing varying amounts of hafnia or HfSiO,
then show wide density variations.
The great similarity in size and chemical behaviour between HfO, and
ZrO, lead to considerable problems in separation of these materials, and in
general no attempt is made, so that zirconia often contains hafnia. Another
reason for zircon density variations lies in the fact that they often contain
amorphous silica and zirconia and need to be heated to recombine these
materials into zircon.
Extraction of the oxide from these sources follows routes similar to those
used for alumina production and can be divided into methods based on
zircon and on baddeleyite.
(a) Zircon is powdered with coke sufficient to reduce half the oxygen
content and pig-iron is added. The whole charge is fused in an electric
furnace. The process is energy-expensive, and the inclusion of cheap iron
brings down the cost by providing a liquid medium and an enriched
by-product in the form of very saleable ferrosilicon pig.
Z1SiO, +C +Fe > ZrO, + Fe-Si.
(b) A purer product is obtained at lower temperatures by incorporating a
strongly basic chemical; zircon ore is ground with caustic soda and heated to
600°C. Reaction produces sodium zirconate and sodium silicate, which
attests the amphoteric nature of ZrO,. The charge is leached with water to
take the silicate into solution and hydrolyse the zirconate to a gel which, after
filtration, is dissolved in hydrochloric acid as the oxychloride. Further
separation and purification can be achieved by dissolution in dilute HCl
followed by crystallization of the oxychloride on concentrating. The oxy-
chloride is either directly calcined at 1400°C to produce ZrO,, or dissolved
in water, zirconia gel precipitated with ammonia, and then calcined at 800-
1200°C. The two routes produce material with substantially different particle
sizes.
(c) From the oxide ore a purer product can be obtained by one of the
following methods.
(i) Powdered ore is leached with concentrated sulphuric acid to take
zirconia into solution as the sulphate, while leaving SiO, and zircon
unaffected. After dilution and filtration, a basic sulphate is deposited
when the solution is made just neutral with ammonia. This solid is
heated to 1200°C to convert it to ZrO).
(ii) Products of the highest purity are obtained by a lower temperature
method which utilizes the exothermic nature of the reaction of chlorine
with the ore and coke. The ore is heated to red heat ina silica-lined
vessel with coke sufficient to change all the combined oxygen to
CO+CO,, while at the same time chlorine gas is passed through. By
this process the Zr, Fe and Ti are changed to volatile chlorides which
sublime from other impurities in the charge. The chlorides are dissolved
274 CERAMIC SCIENCE |

and then concentrated until zirconyl chloride precipitates. Several re-

solutions and crystallizations of ZrOCI, result in a very pure product
that on heating to 1200°C gives a heavy-grained pure zirconia.
All methods except the last one produce materials containing from 2 to 22 a
HfO,.
All the methods eventually produce zirconia by the calcination of the
hydrous oxide gel precipitated by the hydrolysis of zirconyl salt solutions.
The texture of the gels produced is quite dependent on the pH of precipita-
tion.!3-45 At pH 4 a microporous gel of low surface area is produced, and as
the pH of the precipitating solution is raised, the apparent surface area, pore
volume and average pore size increase. Crystallization occurs in the range
390—420° to give the metastable tetragonal phase from a pH4precipitate, the
monoclinic phase from a pH 10 gel, and a mixture of the two forms from pH
values between 4 and 10.

5.5.2 Bonding and structure

Although the Zr** ion is small and highly charged, zirconia is best regarded
as a wholly ionic brittle ceramic. The radius ratio, 0.57, is close to a boundary
value suggesting cubic coordination of eight oxygens round each Zr** and
four tetrahedrally disposed Zr** around each O?~ ; for an ideal radius ratio
this corresponds to the fluorite, CaF, structure. The coordination poly-
hedron in the actual zirconia structure shows distortion until the temperature
is raised above 2200°C, when the ideal 8 :4 arrangement with no distortion is
achieved.
As many as ten forms of zirconia were once believed to exist, but for the
pure material the following series of polymorphs are now known to give the
complete picture. At room temperature to around 1000°C the stable phase
has the baddeleyite structure, with a monoclinic unit cell a = 0.5174nm,
b = 0.5266 nm, c = 0.5308 nm and B = 80.8°. The closeness of this cell to cubic
dimensions is noteworthy. At about 1000°C the monoclinic structure becomes
tetragonal with a = 0.507nm, c = 0.516nm and all angles 90°. This trans-
formation is accompanied by a sudden volume contraction of around 9%
that proves to be catastrophic for shapes madé from pure zirconia. The
coordination polyhedron of the tetragonal phase is shown in Fig. 5.11 where

Figure 5.11 Coordination polyhedron of tetragonal ZrO).

 REFRACTORY OXIDES Dal
the closer approaches of four of the eight ions are seen in a tetrahedral
arrangement. At 2280°C the coordination polyhedron becomes quite regular
with a transformation to a cubic unit cell of the fluorite type. This structure
persists up to the melting point at 2850°C.

5.5.3, Mechanical properties

When the effect of the damaging phase change is excluded (see below), the
behaviour of ZrO, parallels that of «-Al,O3, with an extensive brittle range
A where it behaves as a Griffith brittle solid. It has the range B where
damaging critical-sized flaws are generated by the mechanisms listed for
a-Al,O, (5.4). It has a cubic structure at very high temperatures where it is
possible that a ductile failure mode can occur. This is not certain, and the
possibility is best left for a discussion of the mechanical behaviour of the
cubic ceramic MgO (5.6).

5.5.4 Problem of the monoclinic to tetragonal phase transition

Phase transformations in ceramics can involve changes in composition
and/or symmetry, but it is customary to call the transformations that involve
no change in composition phase transitions. The transference of atoms to the
growing product is a useful means of classification which emphasizes the
great diversity of phase transitions in ceramic systems.
A first step in the classification is to decide whether the transition is
heterogeneous or homogeneous. A heterogeneous transition is one in which
the sample can be theoretically separated into microscopically distinct
regions containing transformed or untransformed material. The implication
of this coexistence is that the transition must be first order, since conditions
exist where both product and starting phase can coexist. Furthermore, it is
implied that the products form at discrete nuclei. Homogeneous transitions
are of second or higher order, with product phase forming everywhere within
the parent phase, giving rise to a diffuse interphase between them. It needs to
be stressed, however, that thermodynamic data rather than structural data
are really needed to prove the order of a transition unambiguously (3.1.1).
The second step is to consider the nature of the interface between the
un-transformed and the transformed phase; is it glissile or non-glissile? A
glissile boundary is very stress-sensitive and can migrate even at low
temperatures under a suitable stress requiring no thermal activation; once
nucleated, transformation is inevitable and rapid. There is no diffusive mixing
of atoms, only small shifts in position in response to the boundary movement.
There is always a shape change in the specimen undergoing transition.
Because stress can promote the change, we have a convenient experimental
method for recognizing this form of transition, namely microscopy, where the
characteristic microstructure of martensitic lathes is easily discernible. Some-
times, as in the ortho — clino inversion in wollastonite, CaSiO, (1.5.2), a
specific driving force in the form of shear stress is needed, but it is more
276 CERAMIC SCIENCE

common for either temperature or pressure changes to promote such a

first-order, heterogeneous, martensitic change.
A non-glissile boundary can only move by passing through transition states
of higher free energy which require nucleation, and so the boundary move-
ment is thermally activated. Only at high temperatures will such boundaries
move easily, and as the temperature falls they may become immobile; hence
once nucleated it should be possible to quench the reaction.
The final step in the classification comes from considering how the addition
of atoms at the boundary is controlled. Is it controlled by the rate of their
arrival or by the rate at which heat is conducted away from the boundary? It
seems that transitions of this type can be interface-controlled or heat
transport-controlled; « > 8 quartz (1.5.4) is an example of heat transport
control.
On the basis of the above scheme the transition from monoclinic ZrO, to
tetragonal ZrO, on heating is first-order heterogeneous, involving the
movement of a glissile boundary between phases. As far as ceramic usage is
concerned, it is accompanied by a drastic 9 % volume contraction, which is
catastrophic for zirconia-ware. This has made the transition the centre of
numerous investigations. Wolten first suggested that the change which occurs
between 1000°-1200°C was martensitic in nature because of the athermal
behaviour and thermal hysteresis, when he noted that the transition occurred
over a 30—40°C range between 1160-1190°C on heating. Within this range
the strain energy arising from the coexistence of two phases brings the
reaction to a halt at a given temperature, and therefore to change the volume
transformed, the temperature must be altered. To account for this Wolten
introduced the term “athermal martensite”.+* The hysteresis is quite marked
with the tetragonal to monoclinic transition occurring in the range 1100-
1700°C, i.e. some 90—100°C lower than the monoclinic to tetragonal transition.
Differences in the strain involved in producing a monoclinic nucleus within a
tetragonal matrix, and vice versa, are given as the reason for this hysteresis.
Conventional material with a grain size 50-100 um is monoclinic at room
temperature. High purity ZrO, powder of about 50nm particle size is cubic
at 25°C, and undergoes transitions in the sequence cubic — tetragonal>
monoclinic on heating to 400°C. It has been suggested that because of surface
excess free energy differences (2.1) between the monoclinic and tetragonal
forms, the monoclinic becomes unstable with respect to the tetragonal form
at 25°C when the grain size is below 30 nm.
Until recently,® because of the relatively high temperature of the transition,
and because single crystals are broken down into polycrystals by the severity
of the stress on transition, it was difficult to establish orientation ratios
between the two phases that might provide definite insight into the nature of
the transition; topotaxy (1.5.7) has been difficult. Now, however, it is clear
that during heating the monoclinic phase transforms to the tetragonal form
by the motion of an interface parallel to the (100),, plane. Simultaneously,
 REFRACTORY OXIDES DAG

Figure 5.12 Projection on (001) of (i) monoclinic ZrO, and (ii) tetragonal ZrO).

twinning occurs behind the advancing interface to produce twins on the (112)
or (112) planes in the tetragonal phase. On cooling, a twinned tetragonal
crystal produces twins on the {110},, and {001},, planes. The resultant
orientation relationships are (100), ||(110),¢, [010], ||[001],¢ or,
because of the twinning that occurs, (100)m||(110)po, [001]m ||[001 ]oct-
These topotaxial relationships are maintained on cooling from high
temperatures.
The distorted polyhedral arrangement around the Zr** 4+ ion is shown in
Fig. 5.11, while the projection on to (001) given in Fig. 5.12 shows how a
small coordinated rotation of the triangular coordinated O?~ in the (100)
planes will give rise to the tetragonal phase from the monoclinic form with no
O?~ needing to move more than 0.12 nm. In this way the non-diffusive, that
is martensitic coordinated, nature of the change can be seen.
Observations that metastable tetragonal and even cubic ZrO, can be
produced by suitable chemical routes or by protracted milling? are of some
interest when it is remembered that the transformation from tetragonal back
to monoclinic can be effected by stress, and furthermore, because of the 9 %
volume increase accompanying the change, a compressive stress field can be
generated. Thus if a crack-tip is acting as a stress concentrator projecting a
tensile stress field ahead of itself then this can produce the transformation of
a tetragonal particle with its resultant stress field acting to stop the crack and
toughen the material. Therefore either mixed polymorph ZrO, or other
ceramics doped with tetragonal ZrO, may be stronger and tougher than the
pure material. This exciting possibility is being researched and will be referred
to again below. At the moment, results suggest that below acritical particle
size there is a positive driving force for transformation to the tetragonal
278 CERAMIC SCIENCE

phase. Clearly the tetragonal phase will be stable only if

Grt+Sryrt+ Vr < Ga PS mlin te (5.17)

where G is the free energy of ZrO, as large single crystals, V is the strain
energy, 7 is the free surface energy and S is the specific surface area. Assuming
the particles to be spherical, S = p/ID where p is the density and D is the
particle diameter. Substitution into (5.17) shows that for stabilization to
occur, the tetragonal particles must be smaller than the monoclinic particles.
Rearrangement of equation (5.17) provides an expression whereby the
critical particle diameter can be estimated if the strain energy term is
neglected:
Dis p Ym aeYT
(5.18)
: (G7 -=Gi) (Vee Vis) be lr

Substitution of available data into (5.18) gives the order of magnitude of D,

as 10nm.

5.5.5 Stabilized zirconia

In summary zirconia undergoes a heterogeneous glissile phase transition at
modest temperatures, and the catastrophic volume change which also occurs
makes it impossible to use the material widely in the pure form. This raises
the question of how this drawback may be overcome. Three possibilities arise
from an understanding of the type of change:

(i) Prevent nucleation by using some form of doping.

(ii) Use the material in the form of single domain particles below the critical particle size,
when the tetragonal form will be stable.
(iii) It is observed that the radius ratio is close to that where cubic coordination would be
expected. Thus if the mean cation size can be increased slightly, then a suitable ratio would
lead to a cubic coordination which would preclude the monoclinic and the tetragonal
phases.

The first suggestion has not been pursued to any degree. The second route
has been shown to work when organic-based syntheses are used to produce
ZrO, with extremely fine particle sizes. For example, if zirconium iso-
propoxide, Zr(OC3H,),, is dissolved in benzene and then water is added
dropwise to the solution, a rapid two-stage hydrolysis occurs:
Zr(OC3H,),+H,O — ZrO(OC3H,), +2C,;H,OH
ZrO(OC,H,), +H,0 > ZrO, +2C,H,OH.
Because the solubility of water in benzene is limited, the reaction is carefully
controlled to give a precipitate in which the average particle size is 5.0nm,
which is below that estimated as the critical value. Hence stabilization has
been effected by this method. Unfortunately, such a fine powder has a large
excess surface energy, which makes sintering easier, and in artefacts pressed
 REFRACTORY OXIDES 279

from such powders the particle size increases on heating until tetragonal and
eventually monoclinic zirconia results. The use of grain growth inhibitors
must be considered if such material is to be of eventual use, but the principle
has been clearly demonstrated.
The main technological solution to the problem is based on adjustment of
the mean cation radius to produce solid solutions with favourable radius
ratios. In order to achieve this, the mean cation radius must be increased;
therefore the additive must be a larger cation. The additive must itself be in
the form of a cubic oxide and the oxide must not be too obviously basic in
nature, or it will react strongly with amphoteric zirconia. We need only
consider a few oxides in periodic groups II and III.
In group II, calcium has a cubic oxide, Ca?* is larger than 0.09 nm and
therefore CaO has potential. However, it is a very basic oxide which readily
forms a compound calcium zirconate, CaZrO,, with ZrO,. The addition of
CaO reduces the refractory nature of ZrO,, but because of the ready
availability of CaO and its cheapness, the system has been studied in detail
up to the eutectic at 30% CaO with the result that this system now has
technological importance.*’4* The sketch shown in Fig. 5.13 does indicate
that cubic solid solution formation is possible between 15 and 28 % CaO.
Also in group II, there is the readily-available cubic oxide MgO, which
should not have prospects for stabilization according to our criteria because
Mg?* is too small. Figure 5.14 confirms this prediction because only the
monoclinic phase is stable down to room temperature. Magnesia is only
slightly basic, and since no compounds are formed, this has encouraged
further research. There is a cubic stabilized form existing above 1400°C with
MgO additions in the 15-30 mole % range. This can be quenched to room
temperature, since the ZrO. ,.) transforms to the tetragonal form only very

Cubic 2105 (ss

+

Cazro,

ZrO. 15 28 30
mole% CaO

Figure 5.13 Cubic zirconia: stability range.

K
280 CERAMIC SCIENCE

Te ae Cubic 210,iss]
2550 a= a
!
1
Tet \ ' Tet. \
etragonal ;
i788) Cubicic ZrO
Zr >(ss]

nC 21) iss] ‘cubic 5

MgO
Zr Oo159) 9

1000 -%-. 5 Tetragonal ZrO + MgO

2(ss}

Y Monoclinic ZrO + MgO

2\ss)

10 20 30 40

mole % MgO

Figure 5.14 Sketch of the ZrO, region of the ZrO,-MgO system.

slowly, requiring the precipitation of MgO (Fig. 5.14). Thus one can obtain
magnesia cubic stabilized zirconia. However, this material is metastable and
must not be used for any length of time at temperatures below 1400°C. If the
material is allowed to reach equilibrium below the critical temperature, the
situation is not too desperate since the very fine MgO crystallites, 0.5—1.0 um,
are able to soften the effect of the tetragonal > monoclinic transformation by
plastically deforming under the stress (5.6). Furthermore, the original cubic
stabilized form can be regained by heating above 1400°C followed by rapid
quenching. Because of this fortunate kinetic parameter, commercial magnesia-
stabilized zirconia is encountered.
In periodic group III only a few of the rare earth oxides are cubic and have
suitable sized cations; of these La,O, and Y,O, have been shown to achieve
stabilization. Both these oxides have simple eutectic diagrams, which is in
their favour, and both show a cubic solid solution phase stable to room
temperature when between 30 and 40 mole % M,O; is dissolved in ZrO).
No other periodic group contains suitable sized cations which themselves
have oxides with the cubic structure.
Too sharp a focus on the mean cation size does divert attention from a
recently explored possibility, namely decreasing the mean anion size.'’
Carbides, oxycarbides, nitrides and oxynitrides of zirconium exist (6.1). In
such materials some models (6.1), propose that carbide and nitride anions
exist, which, depending on the size of the localized charge, might be smaller
than O7?~. Alternatively, the presence of (say) N*~ ions in the lattice
necessitates vacancies in the anion sub-lattice that produce the overall
decrease in the mean anion size.
Early results show that this route to stabilization is possible, with AIN
 REFRACTORY OXIDES 281

10-30 mole %, Mg3N, 1.5mol %, Si;N, 10mol % and ZrN 10-30 mol %,
all achieving between 17 and 100% stabilization of the cubic phase. While
this is scientifically interesting in the sense that it supports the hypotheses put
forward above for achieving stabilization, it does not offer a technologically
important route since the use of such stabilized zirconia in air would
eventually lead to oxidation (5.2) and disintegration of samples.
The discussion above leads to the expectation that commercially-stabilized
zirconia must contain substantial concentrations of magnesium or calcium
ions in the lattice to produce a cubic solid solution. This is reflected in the
commercial preparations outlined below.
(i) ZrO, +MgF,+CaF, — wet milled > bricked — fired at 1500° >
further grinding + HCl wash+refined > 1900° > cubic stabilized zirconia.
This product contains both Mg?* and Ca?* for the purpose of offsetting the
relative size difference, but utilizes the kinetic factor of stabilization to some
degree. To obtain complete cubic conversion, a minimum of 15 mol % of this
mixture would be required, but no more than 10 % is used in order to ensure
that some monoclinic solid solution is left. A carefully-controlled amount of
monoclinic solid used in this way can offset the normal thermal expansion on
heating above the monoclinic transition temperature, giving rise to a finished
product with zero expansion coefficient and therefore good thermal shock
characteristics and improved creep behaviour.
(ii) Cream-coloured opaque materials unsuitable for vacuum work.
ZrO, + MgO = bricks — 1600°C — cubic,,,) > ground 6-8 ym > HCI wash
mS
(13 mol %)
— slip at pH 3—4 for casting > 1750°C — cubic stabilized zirconia.
(iii) Impure yellow coloured products.
ZrSiO, + CaO + coke — arc furnace > ZrO(.,) + Fe + Si+ Ti.
Two liquid layers result with ferro-silicon pig iron as the bottom layer. In this
case, sufficient carbon in the form of coke must be added to just reduce the
oxides of iron, titanium and silicon in the ore. The high operating costs in
terms of energy are partly offset by the valuable by-product and the minimal
pregrinding and separation stages.
Further development work with new preparative routes has led, in the case
of yttria-stabilized zirconia, to a reduction in the percentage of Y,O, required
and the production of almost transparent articles in the Zyttrite*? process.
This process utilises one of the above-mentioned organic routes to obtain
perfect mixing of the reagents and submicron-sized grains by simultaneously
hydrolysing zirconium and yttrium alkoxides. A 6mol % Y,O3, fully-
stabilized zirconia with a mean grain size of less than 1 wm can be achieved
by this process. The mixed oxide can be sintered to shape at 1000°C in 30
minutes, which indicates the tremendous increase in solid state reactivity that
282 CERAMIC SCIENCE

brings about such energy savings as well as savings in the amount of additive
needed. These figures i.e. 6 °% from the organic route or 30% Y,O3 from the
conventional grinding route serve to emphasize that the quoted percentages
of additives necessary to bring about stabilization are centainly not equi-
librium values; these are probably lower, but reflect rather the mixing
methods used and the firing temperatures achieved.
In all the above discussion we have concentrated on adjusting the mean
cation or anion size as a route to cubic stabilized zirconia by addition of
larger cations or smaller anions, but it should be noted that all the
satisfactory additives have a valency different to either Zr** or O7~, with
the result that one is introducing defects into the structure in the form of
anion vacancies greatly in excess of the thermodynamic concentration. One
should for example describe these materials as Zr(y_,,Ca,O2-.» Lx to
emphasize the defect nature of stabilized zirconias. Recently, much use of this
has been made by using stabilized zirconias for electrical conductivity cells
that have e.m.f’s sensitive to oxygen partial pressure. Stabilized zirconia
containing CaO has a high conductivity for O?~ ions, the conduction by
positive holes and defect electrons being of little importance (2.6.2).

5.5.6 Ceramic steels: a use for the zirconia phase change

It has been known for some time that total stabilization in the cubic form was
not altogether a good goal to aim for, since significant increases in strength
of partially stabilized zirconia have been recorded, e.g. mean transverse
rupture strengths around 650 MPa for partially stabilized zirconia (PSZ) have
been recorded compared to 250 MPa for the monoclinic form. The work of
fracture is about 500Jm~? in PSZ, which is approximately five times the
value for the unstabilized or fully stabilized materials with the same grain size
and porosity. Therefore the increased strength is not coming from a reduced
defect size but rather from this increased work of fracture.
Where does the increased work of fracture come from that leads to
the stronger material? The two most likely explanations come from our
knowledge of the structure:
(i) absorption of energy as the metastable tetragonal particles transform to the monoclinic in
the stress field.
(ii) In the process of sample preparation, any surface grinding*® induces the tetragonal>
monoclinic transformation? in the outer layers, which leads to a biaxial compression at the
surface which can reduce the effects of Griffith type flaws in the way described for Al,O,
(5.4.5) and glass (3.4).

A value for the compressive stress can be obtained from equation (5.19) for
materials with similar E and v and for low volume fractions of zirconia:
AV f 2E 5.19
oo. = — jf,——
© 3V 1404, il)
where E and Vv are the mean values of Young’s modulus and Poisson’s ratio,
 REFRACTORY OXIDES 283

f, is the volume fraction of transformed material, and AV/V is the volume

expansion of the phase change.
Both explanations we see make the tetragonal > monoclinic phase change
work to improve the system. In fact, having some tetragonal material in a
matrix is analogous to having a multiphase alloy system based on ZrO,. This
being true, we can see that there is an analogy between ZrO,-based alloys
and iron alloys:
(a) Both have three crystalline modifications.
(b) Both systems have a martensitic change occurring between two of the allotropes or
polymorphs.
(c) Both systems can produce metastable phases depending upon the microstructure or when
other alloying materials are added.
A more detailed study of alloy systems based on zirconia may enable these
similarities to be pushed further, with the development of a range of new
ceramics with much improved strength and toughness characteristics. How-
ever, the transformation behaviour of zirconia will have to be further studied,
since another variation of its effect is possible for particles embedded in a
solid matrix (Fig. 5.15). It is apparent that the tetragonal — monoclinic
transition produces both volumetric and shear strains; for it to occur, it has
to work against a matrix pressure, and so either the transition temperature is
lowered or the critical particle size is increased.'° In this way considerable
percentages of tetragonal material can be held in a composite compared to
that expected in pure zirconia. This can contribute to the toughness, but may
decrease the strength, in that the large grains will transform on thermal
cycling with the result that a series of microcracks will be generated in the
matrix around such grains. External stress can be absorbed by strain energy
release on propagation of such microcracks but ultimate strength is reduced.
It follows that well-dispersed ceramic composites containing fine particle
size zirconia as a toughening (perhaps strengthening) agent are worth
developing. For example:
2Z1SiO, + 3Al,0, > 3Al,0, -2SiO, +2ZrO,.
Monoclinic ZrO,

Tetragonal zr0,

Figure 5.15 Transformation of metastable tetragonal ZrO, particle to the monoclinic form.
284 CERAMIC SCIENCE

Kye -3 8;
MNm 2
2-85 um 125ym

Flexural
strength
MNm2

Volume fraction zr0, in AIO, matrix

Figure 5.16 Effect of particle size and volume fraction of ZrO, on toughness and strength of
alumina.

Here zircon and alumina form mullite (1.5.1) and tetragonal zirconia precipi-
tates for strengthening. The reaction is carried out at 1450°C before the
temperature is raised above 1500°C to produce a 98 % dense body containing
around 30% of the ZrO, in the tetragonal form. Surface grinding will
decrease the tetragonal phase to less than half a percent with a resultant
surface compression and considerable improvements in the strength as well
as toughness.
When composites are made by normal mixing, grinding and sintering
methods, the results are not as satisfactory, since while toughness can be
achieved it is frequently at the expense of strength unless the ZrO, particle
sizes and dispersion can be carefully controlled’? (Fig. 5.16).
Much work clearly has to be done before “ceramic steels” are developed
for use; however the discussion so far will serve to show how an apparently
damaging feature in the structure of a material can, with study, be turned to:
a technological advantage.

5.6 Magnesia, MgO

5.6.1 Sources and manufacture

Magnesia has been, and will continue to be, a refractory of great importance
because there are vast well-distributed resources of magnesite (MgCO3)
in the world. There are even more abundant resources of dolomite
(MgCO, -CaCO;) which are cheaper to obtain than magnesite, and have in
the past been frequently used in the dead-burned forms as cheap refractory
for open-hearth steel furnaces. To add to these abundant sources, the USSR,
 REFRACTORY OXIDES 285

USA, Great Britain, Japan, Mexico and Eire have all built plants capable of
extracting a total of over 1.6 million tonnes of magnesia from sea water. This
material is thus commonly encountered in refractory use although it
does not meet the ideal criterion that we established at the beginning of the
chapter, i.e. its radius ratio produces a cubic structure and its r/n value leads
not to an amphoteric material, but rather to a basic oxide which is then
chemically active. Magnesite occurs in two commercially important forms,
one compact or “amorphous” and the other crystalline or spathic. The
amorphous form is usually found associated with serpentine, (1.5.7), formed
in most cases as the result of carbonated waters reacting with the magnesia-
rich serpentine.
Crystalline deposits have usually resulted from calcium replacement in
dolomite or limestone by magnesia-bearing solutions ascending from an
igneous magma. Deposits of this type are usually lenticular in shape in beds
of enormous extent, containing 10° tonnes or more, in the Urals, Manchuria,
Veitsch in Austria, Czechoslovakia, Quebec and Washington in the U.S.A.
Dolomite is far more abundant in nature: most beds have formed from the
interactions of magnesium salts in sea water which have converted the
CaCO, into CaMg(CO;),. Most industrial countries have abundant supplies
of dolomite.
Magnesia is easily prepared by calcination, as a result of which the thermal
decontfposition of dolomite and magnesite is perhaps one of the most closely
investigated problems in thermal analysis. Although the mechanism of the
decomposition is simple, the literature gives highly contradictory data on
kinetics. Some impurities are known to promote the decomposition, for
example NaCl, while others such as H,O hinder the reaction. In the case of
magnesite, decomposition is outlined in the scheme

MgCo, Cal 3MgO -4MgCO, — 3MgO : MgCO,; +CO, (stable at 510°C)

oes (when poo, = 1 atmosphere) MgO.

When prepared at this low temperature the crystallites are extremely small
and therefore chemically very reactive, so that CO, and H,O are easily
re-absorbed. Dead-burned material, i.e. with all the CO, removed, has to be
prepared by prolonged heating above 1500°C in order to increase the grain
size and reduce the reactivity.47 Even when decomposed at 1000°C it still
retains a small percentage of CO,. At that stage it is known as caustic
magnesite and it can be employed in making strong quick-setting cements
(1.5.6) by adding a solution of magnesium chloride to produce needle-like
magnesium oxychloride crystals.
To obtain MgO from dolomite, two common processes are used:

(i) MgCO, -CaCO,+CO,+H,0 + pressure — insoluble MgCO; +

soluble Ca(HCO;),.
286 CERAMIC SCIENCE

In this way the soluble bicarbonate is removed and MgCO; is calcined in the
normal way. The second process differentiates on the basis of the strongly
basic nature of CaO, compared to MgO, after dolomite calcination by using
the weak acid NH,Cl.
(ii) CaCO; MgCO, G0"e, CaO : MgO + NH, Cl > MgO + soluble CaCl).

5.6.2 Structure and strength

Magnesia fits one of our criteria excellently, namely it has only one structure
type and one composition at all temperatures up to the melting point at
2450°C. It crystallizes with the rock salt fcc structure with four formula
units per unit cell. This structure cleaves easily and perfectly along (100)
planes as a result of which the common name for the material —periclase —is
derived. Slip is easy on {110} planes, even though it would appear that glide
on {100} in <110> directions should be more favourable on the basis that
{100} are more densely packed and have a wider spacing. However, at 4
translation in the {100} <110) system, cations are close enough to develop
strong repulsions which require the energy associated with high temperatures
to overcome them. Known slip systems are:*8
primary = {110} <110>;
secondary, activated at high temperatures = {100} (110);
above 1600°C {110} <110).
In all cases the direction is <110), which connects nearest ions of the same
sign and is therefore the shortest translation vector. Glide in ¢100> would
result in juxtaposition of like ions and therefore would produce large
electrostatic repulsive forces; consequently there is a large energy barrier
compared to <110) glide.
Above 1600° all three slip systems act, and this leads to complete ductility
if stress is not applied until temperatures in excess of 1600° are reached.
(a) Problem of ductility. The dislocation movement mechanism is responsible
for mechanical failure in all temperature regions for fully dense MgO° with a
polished surface. Machined MgO has a Griffith crack failure mechanism up
to 600°C (Region A behaviour, Fig. 5.6), then above this temperature
dislocations pile up at grain boundaries to produce cracks which propagate
to bring about failure.

Table 5.10 Measured flow stresses in magnesia

Go Gy
dislocation flow stress (primary) general flow stress by dislocation
(MN m~?) motion in all systems (MN m~*)

20°C = 1000°C 20°C ~—-1000°C

200 50 2x 103 200
 REFRACTORY OXIDES 287

__twradiated_and_
Irradiated and chemically
chemically pons
polished

Fracture %I
stress

MN m=

50

011 1 0:3

a IIT)
VGrain diameter

Figure 5.17 Strength v. grain size for magnesia.

Porous magnesia shows a rapid decrease in strength above 1200°C due to

grain boundary sliding induced by grain boundary porosity.’® The fracture
mechanisms in pure polycrystalline magnesia can be clearly identified from
the grain size versus strength data (Fig. 5.17).
These data in 5.17 show that in unpolished ceramic, for grain size > 10 um
fracture is consistent with the extension of the machinery flaws, so MgO in
this state has only Region A behaviour (Fig. 5.6). Chemically-polished
materials have fracture stresses increased to values in excess of the dislocation
flow stress ¢9, confirming that fracture is dislocation-initiated, giving Region
B behaviour (Fig. 5.6). When dislocations pile up against a stable barrier,
a large stress concentration is generated in the vicinity of the lead dislocations.
The applied stress that can initiate cracks by this mechanism depends upon
whether nucleation or growth is the critical stage. If nucleation is critical,
then:
t= tothe, (5.20)
where 1, is the shear stress along the slip plane, ty is the dislocation flow
stress, K, depends on the effective surface energy for localized crack initiation,
and L is the length of the pile-up.
Considering Fig. 5.17, there is agreement between observed strengths and
equation (5.20) when o9 = 2to. This is the main evidence that fracture occurs
at the stress needed to initiate cracks by a grain boundary pile up of
dislocations. The value of o, obtained is equal to that stress needed to form
a slip band in MgO.
Since the cracks are nucleated by blocked slip bands at grain boundaries
after even very small deformations, failure can occur by either intergranular
separation, or by trans-granular cleavage. Which of these two mechanisms
occurs depends upon the relative cohesive energies across the (100) cleavage
‘plane and across the grain boundary. Where, as in cubic ceramics, the
cleavage energy is low, MgO = 1.2J m”, and the grain size is large, the
288 CERAMIC SCIENCE

450

Stress o

MN m2
300

150

Strain €

Figure 5.18 Strength of iron-doped magnesia.

intergranular cracks on nucleation turn into cleavage cracks with not too
much difficulty. y, for fracture must be close to cleavage surface energy, about
5 Ja
(b) Mechanism for strengthening. Figure. 5.17 shows that irradiation on
chemically polished samples increases the strength, and the discussion ties the
observed strength to dislocation movement. These suggest a method of
strengthening that has something in common with the zirconia strengthening
methods, namely production of coherent precipitates.”*
The strengthening shownin Fig. 5.18 is achieved by the following mechanism:

MgO + Fe,0,;—-"S MgO

+ MgFe,0,.
Iron is soluble in MgO to about 4 % above 1500°C, but when cooled cubic
crystals of the spinel magnesioferrite, (5.9), are precipitated as coherent
octahedra with {111} and <110> edges, so that when magnesia is cleaved in
its (100) plane the precipitates appear as square particles, while in the slip
planes, {110}, they appear with hexagonal habit.
When a dislocation in MgO enters or leaves a particle, a {110} particle
matrix interface is produced, and since MgO has twice as many octahedral
cations on {110} surfaces as MgFe,O,, the new particle-matrix interface
must have an energy higher than that of the magnesia stacking fault. Thus the
precipitate increases the overall flow stress and so makes it necessary to have
a larger stress to initiate cracks.

5.6.3 Role of impurities in the use of magnesia

Magnesia illustrates that chemical quality is the most relevant refractory
property. A possible exception to this statement concerns cost. Chemical
quality is directly linked to refractoriness, volume stability at high temperature,
to texture, and high-temperature mechanical properties. It is also directly
 REFRACTORY OXIDES 289

Table 5.11 Typical impurity

compositions in magnesia

Compound ie

SiO, 2-0.2
CaO 3-1.0
Fe,0; 0.7-0.2
Al,O, 0.2
B,O, 0.01

linked to slag resistance. Therefore a few paragraphs are relevant here to

describe the care taken to control impurity effects.
In magnesia brick obtained from magnesite, typical impurity compositions
are as given in Table 5.11. These are taken as being always present in even
the best commercial magnesite refractories and we will briefly consider the
role they play.
(a) Boric oxide must be kept low, since it combines with the silica and
silicate phases to produce low-melting glassy phases in the grain boundaries
(3.1). These are obviously detrimental to high-temperature strength since
they cause high-temperature creep through viscous flow.
(b) The CaO and SiO, components can react in use to produce either
CaSiO, or Ca,SiO, (1.5.6). It is the aim of the materials engineer to have the
ratio adjusted to give Ca,SiO, because this is more refractory. Below 2:1
ratio it is also possible to obtain MgCaSiO,, (1.5.1), which is less refractory
and produces a marked fall-off in strength between 1400—1600°C. However,
it has been noted that slag resistance drops off rapidly with total silicate
content, which must therefore be kept low; one possible way that this might
be achieved is through pitch impregnation. This is discussed further below.
(c) It is significant that Fe,03 is always present, and following the
arguments in 5.6.2 this could be increased with benefits to both strength and
hardness.
In order to overcome the effects of inherent impurities, it has been common
practice to add other impurities, often in large amounts.
(d) Chrome-magnesite.* When the open-hearth steel process was widely
used, chromite bricks made from a naturally-occurring spinel FeCr,O,,
mixed with varying amounts of magnesia, were a common refractory. These
were used essentially to lessen the effect of the basic nature of the magnesia
and bestow slag resistance cheaply. Consumption of chromite in refractories
has halved in the U.K. in the past two years, as purer magnesia bricks are
required for basic oxygen steel furnaces. However, since chromite is already
in the refractory oxide form, it requires no pre-calcining and therefore less
energy is required in brick manufacture, with the result that it may stage a
recovery as an additive.
(e) Escolaite additions. This technique has developed to win some of the
290 CERAMIC SCIENCE

y| oy Y, eT
=2 \ OSG

Figure 5.19 Dihedral angle between magnesia grains.

slag resistance of chromite brick whilst still using very high magnesia
contents. Between 5-35 % of escolaite, Cr,O3, is added to magnesia for two
main reasons. First, at high temperatures, the presence of chromium oxide
dissolved in the magnesia increases the dihedral angle @ shown in Fig. 5.19.
This results in more solid—solid bonding and less slag-wetting, as it is unable
to permeate the grain boundaries.
This advantage is achieved after small escolaite additions, but when larger
amounts of Cr,O, are used, then the spinel MgCr,O, results from solid-state
reaction and this phase bonds well to periclase as a grain boundary material,
so allowing a reduction in the total silicate content.
(f) The role of pitch. We should in fact consider the role of the residual
carbon left after the bonding tar has burned out in manufacture. Whether
carbon confers advantages on magnesia refractories in use is not clear, as
conflicting data are obtained from different users. This is particularly true
when considering results from BOF (basic oxygen furnace) and AOD (argon
oxygen decarburization) steel producers. In BOF furnaces, carbon-bonded
bricks have a marked improving effect by decreasing the penetration of slag
into the brick matrix and so preventing MgO dissolution. In AOD vessels, no
such improvement is noted, and the extra cost of pitch impregnation is not
recovered. A satisfactory explanation for this apparently contradictory
evidence can be found in brick chemistry at a very localized level.* In the
BOF process, the liquid slag contains FeO and so the oxygen partial pressure
at the brick face as determined by equation (5.21) is approximately 107° atm
at 1600°C using the methods of (5.2).
FeO = Fe +40). (5.21)
In the AOD process, the slag has no FeO and only a little Cr,O3, in which
case the oxygen partial pressure is determined by either (5.22) or (5.23):
Cr,0, = 2Cr+3/20, (5.22)
SiO, = Si+0,. (5.23)
Hence at 1600°C the po, will be in the range 107 *°-10" *° atm. Now around
 REFRACTORY OXIDES 291

each carbon particle in the magnesia brick there will be a reaction giving rise
to a local oxygen partial pressure:
MgO+C=Mgg)+CO- - (5.24)
CO=C+40,. (5.25)
From (5.24)+(5.25), the local po, within the brick at 1600°C will be
approximately 10- ‘8 atm, and the following scheme can be envisaged under
the influence of the oxygen gradient:
Brick at 1600° Hot Face
(i) BOF: Mg0+C=Mgqg)+CO —Mg,y) +30, MgO+FeO/Fe slag + metal
(Po, = 107
*8atm.) J (po, = 10-8 atm.)
dense periclase
layer
(ii) AOD: Mg0+C=Mgg) +CO MgO +slag
+ metal
(po, = 10~*8 atm.) (Po, = 10°
*°atm.)

Thus in the BOF refractory a dense secondary periclase layer will form just
behind the hot face and prevent slag penetration, but not in the AOD case.
Other, less important, improvements arising from the presence of carbon
in the magnesia are an improvement in the hot strength, particularly if the
brick contains higher levels of iron oxide impurity. This is due to carbon
reduction of the iron oxide, making it unavailable for low melting silicate
formation. There is also the fact that the carbon will increase the thermal
conductivity and so improve the spalling resistance.
(g) Solid solution formation—modern application to furnace control. Following
stricter environmental pollution control regulations, new materials have been
sought for monitoring and controlling the combustion processes in internal
combustion engines and furnaces. Monitoring combustion by means of
oxygen sensing is a possibility, because there is a direct relationship between
the equilibrium oxygen partial pressure in the exhaust from an engine or
furnace and the air-to-fuel ratio of the burned mixture. In order to achieve
control resulting in more efficient burning by controlling the air-fuel mixture,
necessary sensor characteristics are fast response time (about 1s), the ability
to operate at high temperatures, and the ability to withstand corrosive
environments. It can be seen that these last two requirements meet our
definition of a refractory ceramic. However, MgO is so stable that few defects
exist even at temperatures just below the melting point and so ionic
conductivity is always too low.
Cobalt monoxide, CoO, when stoichiometric is a wide-gap semiconductor
(6.1.4), with very high resistivity but the material is readily made non-
stoichiometric by the presence of Co** vacancies and resultant charge
balance by oxidation of some Co?* ions to higher valencies. Thus in this
material electronic conductivity is strongly influenced by composition fluctua-
tions, which are controlled by the ambient oxygen pressure at high tempera-
tures. At 1000°C the resistivity of CoO changes from around 40 ohm-cm at
292 CERAMIC SCIENCE

Po, of 10° ** atm, to 10-1 ohm-cm at 107! atm, and since this is a useful
range for monitoring air-to-fuel ratios from 14 downwards, it is potentially a
very good material for control purposes. However, CoO on its own does not
possess the necessary stability, being reduced to metallic Co at po, ®
10-!2atm, and is also oxidized on prolonged heating in oxygen-rich
atmospheres.
Can the MgO and CoO successfully be married to form a stable device?
There are promising possibilities, because both compounds have rock salt
structures and Mg?* and Co?* have a close size relationship. When fine
powders of MgO and CoO are sintered in the range 1400-1600°C, a solid
solution Co,,_,)Mg,O results that is single phase with a rock salt structure
for 0 S x $0.85. Such phases have been shown to be stable over the range
1-10~2° atm oxygen pressure between 700-1100°C.** They contain sufficient
singly and doubly ionized cobalt vacancies to give rise to p-type conductivity
(2.6) with the resistivity being proportional to (po,)*/* at high po, and
(Po,) */° at lower pressures. The composition Mgp.7Coo,30 has a resistivity
change of 104-10! for a po, change 10° **-1atm at 1000°C, and as it is
chemically stable, is an exciting prospect for future pollution control device
technology.

5.7 Urania, UO,

So far in this chapter we have finished each section with a modern application
for old materials: lasers with Al,O, (5.4.6), ceramic steel with ZrO, (5.5.6)
and pollution control with MgO (5.6.3). In contrast, UO, is a modern
ceramic finding its use in the new atomic energy industry. Without the advent
of the nuclear power industry, it is unlikely that UO, would have found many
users, suffering as it does from the serious drawback of not being the only
stable oxide of uranium. The result of this is that, unless the atmosphere in
which it is used is controlled, it can undergo changes in composition. Starting
from uranium the following scheme shows the number of oxides:

U3 uO Sos UO 2
U0, 925-04
Oo

rock salt structure fluorite structure

cubic cubic cubic
(basic oxide) (amphoteric oxide) (acidic oxide)

There are three cubic oxides which are expected when the radius ratios are
calculated for U?*, U** and U®* to O?~. From this it can be seen that for
UO,, reducing conditions will lead to the monoxide and oxidizing conditions
to the higher oxides. The position is even more complicated, because UO, is
an ideal composition for what is a widely non-stoichiometric compound with
both hyper- and hypostoichiometric compositions that have substantially
different mechanical properties.
 REFRACTORY OXIDES 293

5.7.1 Source and preparation

As uraninite, uranium dioxide occurs naturally as the major ore of uranium
in granites where it is associated with rare earth minerals. It crystallizes as
combinations of octahedra and cubes often in botryoidal masses, when it is
known as pitch blende. It is brittle, and black to brownish in colour
depending on the degree of oxidation, since the composition is always
between UO, and U3Ox,. The production routes reflect the amphoteric
nature of the oxide. The ore is crushed and treated with either nitric acid or
sodium carbonate to take uranium into solution as either uranyl nitrate or
sodium uranate. From UO,(NO;), solution in dilute nitric acid, the uranium
is separated from most impurities by solvent extraction with tributyl phos-
phate in odourless kerosene solution. The UO,(NO;), is regained by back
extraction with dilute nitric acid, then treated with ammonia to precipitate
ammonium diuranate (NH,),U,0,3. The precipitate is dried and heated to
450°C to produce a U;0, powder with moderate surface area to minimize
reoxidation. Finally, the U,O, is heated to 700°C in hydrogen to obtain the
near-stoichiometric UO,. Control of the solution pH and use of tributyl-
phosphate extraction leaves most impurities behind in the aqueous extract.

5.7.2 Structure
In this ceramic we encounter the structure that zirconia just misses and has
to be doped to achieve: the cubic fluorite structure, in which there is 8:4
coordination. Each U** has eight O?~ at the corners of a cube and each O77
has four U** ions at the apices of a regular tetrahedron.
The fluorite structure is an open one, having a low coordination number
component, which when coupled with the possibility of variable valency in
the cation, leads to substantial numbers of O7~ ion sites being vacant and
hence the hypostoichiometric range. The association of U** vacancies and
oxygen ion interstitials leads to the hyperstoichiometric compositions. The
generation of defects, the numbers of defects, and the interaction of defects in
the fluorite structure is a greatly studied field because of the interest in UO,
as a high-temperature nuclear fuel element.

5.7.3 Strength
Since UO, has a cubic structure, what has been said in 5.6.2 concerning the
strength of MgO can also in the main. be applied here. Dislocation activity
has been observed down to room temperature in a variety of tests, including
microhardness indentations. The primary slip system is {100} <110> and the
secondary is {100} <110), which shows that although the planes are different
than those in MgO because of the different structures, the directions are the
same in both systems. At 1000°C the flow stress oy is 40 MNm_?, compared
to 50 MN m~? for MgO and 100 MN m ? for Al,O3, which gives some idea
of the relative ease of deformation of these materials.
There is little doubt that plastic flow occurs above 800°C, which is a
294 CERAMIC SCIENCE

500 1000 1500 2000

Temperature ‘C

Figure 5.20 Temperature range of controlling strength mechanisms.

considerable disadvantage to using the material at high temperatures in a

reactor since it must be protected from any appreciable stress. The onset of
plasticity is considerably affected by stoichiometry.
In the stoichiometric material of small grain size around 6 pm, the inherent
flaws control the strength up to 800°C; these appear to be large pores about
50 um in diameter and lead to fracture stresses around 160 MN m~ 2. Thus in
this temperature range we have, as discussed for Al,O3, Region A behaviour.
Between 800°-1200°C, fracture initiation is by dislocation motion and
strength decreases as the temperature rises. In this region, the fracture stress
has values equal to the compressive flow stress; the reason for such a
coincidence is not clear since, as we have seen, dislocation movement is
recorded at temperatures down to ambient. Above 1200°C, gross plastic flow
precedes fracture, giving rise to a Region C and ductile behaviour. Figure 5.20
shows the relative sizes of the A, B and C ranges for some of the materials
discussed to date.’®
It might be expected that, as in the case of MgO, impurities could be used
to increase the strength of UO,. In general, second-phase precipitates such as
Si from impurities in the starting material, together with Al,O, from
processing, do exist as impurities. Both are quite strongly bonded to the
matrix UO,, but have a deleterious effect on strength, lowering it by about
100 °%%. On the other hand, metallic precipitates such as U or Fe seem to have
little effect. The identification of impurity particles such as Al,O, as fracture
nuclei has importance because purification and thermal processing may
usefully improve fracture strength.
Non-stoichiometry causes variations in properties, and even small excesses
of oxygen e.g. O/U ratios of 2.06, cause marked differences in mechanical
behaviour, with lower yield stress and cross slip to {111} planes.*? Activation
of two {001} systems causes two systems of dislocations to interact on {111}
to form dislocation networks. Thus polycrystalline UO,.,,, samples are
notably softer due to the lowering of yield stress in the lattice, and to the
availability of additional slip systems which permit general shape changes in
 REFRACTORY OXIDES 295

suitably aligned grains. It is postulated that such results are the consequence
of the effect of lattice defects on glide on {001} planes. An acceptable model
for the defect cluster consists of one or two oxygen atoms at interstitial sites
near the centre of cubes of anions, two oxygen ions displaced slightly from
their normal sites by interstitials, and two anion vacancies. When a dislocation
approaches such a defect cluster, the elastic stress fields surrounding the
dislocation and the cluster will impinge causing changes in the size of the
defect cluster. It is likely that in all compositions glide begins on the principal
{001} <110> slip system, but in hyperstoichiometric material it rapidly
encounters the defect cluster, which because it contains two interstitial atoms
splits into two single interstitial defects. The dislocation continues its glide by
depositing an anion with the smaller cluster. The repositioning of interstitials
and deposition of glide anions is in effect a motion of anions on a near (111)
plane causing cross-slip from (100) to (111) planes.

5.8 Titania, TiO,

Titanium dioxide is widely used in the paint, plastics, paper, ceramic and
rubber industries, mainly as a white pigment or opacifier. In the ceramics
industry its main applications are in enamelling (3.5), and as a catalyst (4.6).
Total production in 1980 was 2.75 x 10° tonnes, which makes it an important
material, and at approximately £500 per tonne it is financially an important
industry. Fortunately titanium is an abundant element, with large mineral
deposits in Canada, the U.S.A., Norway and India. The main minerals
exploited are the non-silicates ilmenite, FeTiO, (5.11), and rutile, TiO).

5.8.1 Manufacture
Whilst one process route, the sulphate process, still dominates, a new route is
becoming important—the chloride process. The raw material requirements
for these two processes are considerably different. Both have economic
problems in that they use expensive chemicals, are energy-intensive, and lead
to large amounts of economically low-grade waste products.
(a) Sulphate process. Here reliance is placed on the digestion of suitable ores
with sulphuric acid, noting that H,SO, does not attack rutile or high-grade
ores. Furthermore, high chromium-content ores are unsuitable because of
their undesirable effect on the final white colour of the product.
Ground ilmenite is digested with 98% concentrated sulphuric acid at
elevated temperatures to convert the titanium and iron to sulphates, which
can be leached from the mass with water or dilute acid. Ferric salts at this
stage are reduced to Fe(II) by addition of scrap iron. On cooling, FeSO,
7H,O, known as copperas, separates out and can be removed. The concen-
trated liquor is treated with steam to hydrolyse the titanyl sulphate, TiOSO,,
by a strongly exothermic process that thickens the slurry and then produces
296 CERAMIC SCIENCE

a colloidal hydrous oxide known as sulphate cake. Prior to hydrolysis, seed

agents are added to determine the final crystal form of the TiO,: if the more
durable and whiter rutile form is required, small quantities of zinc or
aluminium salts are added, and if anatase is required phosphates are added.
Finally, the cake is calcined to temperatures in the range 800—900°C with
control of +5°C needed to deter sintering and discoloration.”°
(b) Chloride process. In this process the raw materials are used to prepare
TiCl,, which is subsequently purified and oxidized to TiO, :
ore +C+Cl, > TiCl, +CO,+CO
TiCl, + O,+N, > TiO, +2Cl, + No.
Both rutile and ilmenite are used in the process as raw material, but rutile has
the advantage of requiring only 0.77 tons of chlorine per ton of TiCl,
produced, of which 0.73 tons is recovered, whereas ilmenite requires 1.1 tons
of which only 0.73 tons are returned at the second stage. As a result, it is
common practice to upgrade ilmenite to synthetic rutile by a leaching process
following oxidation in a furnace.
The reactions involved in the leaching route can be represented as follows:
2FeTiO, +40, > Fe, TiO, + TiO,
Fe, TiO, + TiO, +4C > 2FeO + 2TiO, +4CO,
FeO +2HCI > H,O0 + FeCl.
Reaction between raw material, coke, and chlorine usually takes place
around 950°C and reaction between TiCl, and oxygen-enriched air around
1000°C.

5.8.2 Structure
Titanium dioxide has three polymorphs, anatase, brookite and rutile. There
are several ways in which the structure of rutile can be viewed in order to
emphasize the distortion it possesses by virtue of containing Ti** ions in a
close-packed array of O?~ ions. We cannot realistically expect a wholly ionic
structure for a system containing M** ions, and a degree of covalency is
implied, as in the case of Al,O, (5.4.3). In order to achieve satisfactory
covalency with oxygen in sp? hydridization, (2.2), considerable distortion
from a simple octahedral site occupancy of a close-packed anion structure is
required.
A picture of the structure can be built up by starting with a close-packed
hexagonal array of oxide anions. In such an arrangement, the octahedral
sites, one per oxide ion, lie above each other in lines parallel to the c axis. The
composition TiO, allows for alternate occupation of the octahedral sites by
Ti**, since there are twice as many octahedral sites as cations, and in this way
the strong repulsion of Ti**—Ti** is minimized. Further stabilization can be
achieved if the close-packed oxide ion layers are now distorted so that
 REFRACTORY OXIDES 297

(c)

Figure 5.21 (a) Idealized precursor of rutile, TiO,. (b) Projection of TiO, structure, emphasizing
the off-set Ti** ions and puckered O?~ layers. (c) View of TiO, structure, emphasizing edge-
shared octahedra producing chains of TiO, linked by apex sharing to paraliel TiO, chains.

buckling occurs, allowing three oxygens to become coplanar with Ti**, and
resulting in an sp? bonding configuration that allows a degree of covalency
to occur and stabilize the structure. Movement of the oxide anions in this way
obviously introduces a distortion in the octahedra occupied by each Ti** ion.
Thus rutile can be seen to be a distorted version of the idealized precursor
shown as Fig. 5.21a. Offset Ti** and puckered O77 layers are illustrated in
Fig. 5.21b.
Since the structure has two identifiable polyhedra, slightly distorted TiO,
octahedra and OTi, nearly equilateral triangles, then the structure can be
described by the polyhedral sharing model (5.3). Rutile has, on this model,
chains of edge-sharing octahedra with overall composition TiO,, and these
chains are linked by the apices not already shared in the chains as sketched
in Fig. 5.21c.
It is appropriate at this point to outline the fascinating way in which this
structure, as recently shown, accommodates what appear to be wide vari-
ations in non-stoichiometry by concentrating vacancies in selected planes
and changing the octahedral linkage patterns in those planes. Vacancies in
the anion sub-lattice can be eliminated by a change in anion coordination along
a specific crystallographic plane whilst cation coordination remains octa-
hedral. The principle can be illustrated by considering corner-shared octa-
hedra shown in Fig. 5.22. Removal of oxygen, that is defect creation, at oxide
ion sites 4 and 5 can be accommodated by a movement left to right and
vertically as shown. In such a process atoms 1, 3, 8 and 9 have been changed
with respect to their coordination, for example 1 was linear two-coordinate
298 CERAMIC SCIENCE

Figure 5.22 Schematic diagram of four corner sharing octahedra undergoing a C.S. to eliminate
oxygen vacancies at 4 and 5.

but is now 90° two coordinate, whereas 3 has become three-coordinate from
two coordinate. One part of the structure (i.e. atoms 1, 2, 3, 7, 6) have moved
left to right in a shear manner, but at the same time have moved 90° to that
shear direction. The movement by 90° to the shear plane eliminates the anion
defects at 4 and 5 but also introduces a composition change, since octahedral
corner-sharing becomes edge-sharing. Because the envisaged movement
involves shear combined with a 90° movement to the shear plane, the total
process is denoted as crystallographic or C.S. shear. This formalism is recent
(Wadesely, 1969), and has recently been extensively investigated.
When they occur, C.S. planes run right across entire crystals. Remarkably,
CS. planes are found to be regularly spaced and unperturbed slabs of original
structure are separated by narrow CS. planes. Each slab has the same width,
and the composition of the overall crystal depends upon the width of the
unperturbed slabs and the particular crystallographic plane in which the CS:
occurs. In rutile, unlike the example in Fig. 5.22, TiO, octahedra link by
sharing faces along the C.S. planes.
The mechanism by which vacancies are cooperatively ordered in this way
is as yet unknown, but it is speculated that the process starts at a crystal edge
or dislocation and grows inwards. Catalytic activity may be associated with
them.

5.8.3 Bonding
By focusing attention on the planar arrangement of Ti* *O%~ units that are
such a feature of the structure, an instructive bond model has been developed
using the molecular orbital principles (2.2.7). A o-bonded framework arises
from overlap of sp? hybrid orbitals with d*sp* hybrids on the metal ion; the
remaining p orbital on each oxygen, at 90° to the plane of the sp?-hybrids,
forms x bond molecular orbitals with two of the remaining 3d orbitals with
suitable symmetry. The remaining d-orbital from the ft, set, with lobes
 REFRACTORY OXIDES 299

directed along the c-direction, is available for Ti**—Ti** bond formation if

the separation permits. The discrete levels of the 2Ti**, 407~ unit are
broadened into bands in the reality of a crystal and these can accommodate
twenty-four electrons in the Ti-O o-bonding level plus eight in the Ti-O
m-bonding level. For a Ti,O, unit, since Ti** has no valency electrons
and 40?~ have thirty-two valency electrons, all the bonding levels are filled and
TiO, is stable and electrically insulating. Clearly other oxides with the rutile
structure might have interesting properties on this model: VO, has a single
d electron per V**, which must utilize the overlapping M—M o bond formed
from the d,2 orbitals and M—O n* band, which then, being only partly filled,
bestows metallic conductivity. VO, is known to have metallic electrical
conductivity and as might be expected from this model a decreased stability
compared to TiO).

5.8.4 Mechanical properties

Although not used as a structural ceramic to any degree, some measurements
have been made to determine the effects of quenching and various chemical
treatments on the strength.*) Average strength of TiO, bars of 9.3 x 6.35 x
76 mm tested in 4-point loading is reported to be 128.9 MN m ”, and this can
be improved to 171 MNm ” by quenching from 1200°C or to 179 MNm ?
when the surface is glazed. Further improvements result when the samples
are glazed and quenched, because the glass then has a compressive surface
layer induced on it (3.4.7).

5.9 Spinels
It was stated in 5.1 that, from the twenty-two refractory oxides, over 9000
binaries can be produced, but only 1000 are likely to be refractories. We have
now seen that many of these are naturally-occurring and fall in the class
known as spinels, with the general formula AB,O,. The family takes its name
from the 50:50 reaction product of two refractories we have so far considered,
Al,O, and MgO, for these produce MgA1,O,, known as the mineral spinel,
which (with a melting point of 2135°C) is more refractory than Al,O3. It has
high hardness, good thermal shock resistance and high mechanical strength
and chemical inertness to alkalis, but despite these points it, and most other
naturally occurring and available spinels such as gahinite, ZnAl,O,,
hercynite, FeAl,O,, magnetite, FeFe,O, and hausmannite, MnMn,O,, are
little used as refractories for reasons to be made clear in the ensuing sections.
Nevertheless they are tremendously important because of the sensitive
dependence of the electrical and magnetic properties on temperature and
composition, and so they are vital ceramics in a modern sense. They are
excellent examples of how the theories of the ionic bond can be used to
interpret behaviour and predict properties.
300 CERAMIC SCIENCE

The only spinel which is used as a refractory is one naturally occurring

member, FeCr,O,, chromite (5.6.3) which is the only commercial source of
chromium. For refractory purposes chromite is embodied in bricks, cements
and plasters in the construction of furnaces. Usually the chromite is employed
as a chemically neutral insulating layer between a basic hearth of magnesite
brick and an acid roof of silica or firebrick. Many steel furnaces use, or used,
chrome-magnesite bricks which are direct mixtures of FeCr,O, and MgO,
which improved the MgO behaviour as explained in 5.6. Such refractory use
has until recently consumed between 15-20 % of a world chromite production
of around 11 million tonnes. However, for some years the consumption of
chromite in refractories has been falling, because open-hearth furnaces
require no chromite, with the result that consumption has halved in two
years. This trend may not continue, because energy costs rise, and the energy
needed to produce a chromite refractory is lower than most others (the ore is
in the refractory oxide form and therefore no pre-calcining is required). To a
certain extent this will depend on supply, for rich deposits exist either in the
USSR or less politically stable areas of the West like South Africa, Cuba and
the Philippines.

5.9.1 Structure
Frequently the composition of spinels is expressed as AO -B,O3, which is
misleading since X-ray analysis does not reveal any difference of oxygen ion
arrangement in the unit cell: all oxygens are equivalent and form a close-
packed cubic structure. The unit cell contains thirty-two oxygen ions in a
close packed fcc array, giving rise to the formula AgB,,O32. We can
immediately see great potential for variation because such a close-packed
array generates sixty-four tetrahedral and thirty-two octahedral sites for the
twenty-four cations to occupy. Then, since per formula unit there are eight
negative charges to neturalize, this can be done in three ways:
A?*+2B°* (these are 2,3 spinels)
A** +2B2* ~ (4.2 spinels)
and A®*++2B* (6,1 spinels.)
Of these the 2,3 spinels are most common, constituting 80% of the class,
while the 6,1 group are frequently regarded as acid salt derivatives because of
the great polarizing power of a 6* ion, which leads to covalency. Thus one
would write Na,WO,, sodium tungstate, or Ag,MoO,, silver molybdate,
rather than WNa,O, spinels etc.
Within the 2,3 and 4,2 spinels the varying sizes of ions would be expected to
influence which of the ninety-six sites they use, but this is only one factor.
However, whenever the tetrahedral sites are occupied, only a very few cations
will fit exactly and not distort the oxygen ion packing. Since the degree of
distortion of the oxygen ion structure is an important variable in spinel
behaviour, it is quoted and has the name oxygen parameter, u. It is defined
 REFRACTORY OXIDES 301

A site

Figure 5.23 The spinel structure.

as {+6 where 6is the ratio of distance from the centre to the corner of a
cation-filled tetrahedron to the distance of the body diagonal of the unit cell.
If the tetrahedral cations fit perfectly, the O?~ are in perfect cubic
close-packing and yp = 0.375, but in most spinels u > 0.375. A normal spinel
can have p values greater than 0.375 but has all the eight A ions in tetrahedral
sites and the sixteen B ions in octahedral sites. This produces a structure
containing eight cubelets (Fig. 5.23) linked alternately as shown. If all the
cubelets were linked A—A...the zinc blende structure would be generated,
and if all were B—B-B...the rock salt structure would be produced. When
linked A-B—A-B... eight times, the spinel structure emerges. This serves to
emphasize, first, the more complex nature of the structure, and secondly,
because it contains elements of the low coordination ZnS structure, the
inherent covalency with deviations from a wholly ionic situation.
Inverse spinels are defined in terms of disorder amongst the B cations, when
3B cations occupy the tetrahedral sites and 4B+A cations are found in
octahedral sites. Normal and inverse are limiting cases, and degrees of
inversion are common. The variation in inversion or disorder is noted by the
A parameter which is the fraction of B cations in tetrahedral sites, thus:

Normal A =0 = (A),.B,],0,4
Inverse A= 3 = (B),[ABI,0,
1,04.
Complete random 4 = 4 = (Ao.5Bo.5):LAo.sBi.5

Ais not constant but varies with temperature, which is one reason why spinels
are not so useful as refractories. The equilibrium value of 4 is known to be
given by the Néel migration law:

aan = exp[Ey)/kT] (5.26)

where E,,) is the molar energy required to move a divalent ion from a
tetrahedral to an octahedral site. The energy E,,) consists of several terms.
302 CERAMIC SCIENCE

5.9.2 Factors determining degree of inversion

Five factors can be identified which in combination determine 4. The first
three are closely linked to the ionic nature: geometrical size factor, electro-
static Madelung factor, and crystal field energy. The fourth factor is the
physical one of temperature, and the fifth arises from polarization and the
possible partial covalent nature of these materials.
(a) Size. The close packing produces octahedral and tetrahedral sites, but
individual cations have coordination environments determined by the radius
ratio (5.3.7). Octahedral coordination is favoured when the range is 0.414—
0.732 and tetrahedral when in the range 0.225-0.414. In general, cations of
higher charge are smaller and so there is a driving force for B?* to occupy
tetrahedral sites and A2* octahedral. Thus, if this was the only factor, all
spinels with rgs+ <r42+ would adopt the inverse configuration.
(b) Madelung factor. Since inversion changes the coordination of the higher
charged ions, the Madelung constant (5.3.4) for each arrangement will be
significantly different. The difference must also be a function of the pu
parameter, because this changes the O?~-cation distances. For 2,3 spinels
the calculation shows that above y = 0.376 the normal form has a greater
Madelung constant, hence this factor outweighs the geometric one in most
cases. Some examples of the Madelung constant are given in Table 5.12.
2,4 spinels by contrast have the opposite tendencies, as the A** ions seek
the position of higher coordination by O?~-. A further point, leading to
Coulombic gain, is that A** is ordered on the octahedral sites. For values
above 0.380 the inverse form has a Madelung constant that is larger than that
of the normal form.
(c) Crystal field energy. Clearly this factor only applies to spinels containing
d-transition element ions. For d> or d'° ions, there is no net stabilization in
either octahedral or tetrahedral sites. For d' to d* ions, the crystal field gives
the ion an octahedral site preference, which means that spinels containing
Ti?*, V3*, Cr?* and Mn°* will have this tendency working with the
Madelung factor to produce normal 2,3 spinels.
Such divalent ions as Fe?+ =d®° have an additional octahedral field
stabilization for the A?* cation that can overcome the Madelung energy,
with the result that divalent ions occupy the octahedral sites to produce the
inverse structure (B**),[A?*B?*],O,4.

Table 5.12 Madelung constants

for 2,3 spinels as a function of u

LB A(N) A(I)
0.375 128.6 130.7
0.380 131.4 131.0
0.385 134.0 131.2
0.390 136.5 131.4
 REFRACTORY OXIDES 303

Table 5.13 Octahedral site preference energies

Ion P (kJ mole~*) Ion P (kJ mole~*) Ion P (kJ mole~'*)

Li — 15.0 Mg?* — 20.9 Ales — 10.5

Gua — 35.9 Gaz* 128.3 Tiss —91.5
Ag* —81.9 Mn* —61.4 Weer — 48.5
Fe?* —41.3 Creu +69.4
Core — 43.9 Mn?+ + 13.0
Ni?* + 37.6
ug —0.4 Fe3* —55.6
nen — 132.1 Gat — 64.4
Cd — 121.6 nea — 167.2

However, it must be stressed that the electron distribution for cations

containing 4, 5, 6 or 7 electrons in an octahedral field depends on the relative
magnitudes of the crystal field splitting and the energy needed to pair two
electrons in the same orbital. Fortunately O?~ is a strong field-producing ion
and most of the oxides of the first row transition element occur as high spin
configurations.
Differences between the energies of ions in octahedral sites or tetrahedral
sites can be estimated and expressed as octahedral site preference energies
some of which are given in Table 5.13. A cation with a larger P value will
displace a cation with a smaller P value from an octahedral site, and in this
way normal or inverse configuration can be predicted. When the P value
difference is less than 12.5kJ mol~', spinels containing these cations have 4
values around + and are described as random. However, each case must be
considered more fully than this generalization suggests, since degrees of
ordering within the inverted form can alter the Madelung calculation. For
example, detailed lattice energy calculations for NiAl,O, show that the
normal form (Ni?*),[Al3*],O, is 104kJmole~* more stable than the inverse
configuration (Al>*),[Al? * Ni? *),O,. However, the crystal field stabilization
energy of Ni2* on octahedral sites is almost 96kJ mole”*,which makes the
normal and inverse configuration almost equal in energy, with the result that
a more random, temperature-dependent structure is adopted.
(d) Temperature. So far we have discussed the equilibrium distribution; this
must then be temperature-dependent, and depending upon cooling rates,
metastable equilibria can be frozen into spinels.'’ Thus the thermal life can
alter the cation distribution and hence the spinel properties in a drastic
manner. This is the main reason why they have limited possibilities as
refractories. On the basis of simple theory, one can predict the form of the
temperature dependence in the following manner.
For an inverse spinel, the ideal site distribution is Bit [B** A? *] 04 but
with increased temperature, the site exchange equilibrium can be written:
3+
Adct + Bit = Ate + Beet
304 CERAMIC SCIENCE

and

_ FAbt Bee‘oct

An2+ ap; +
tet

expressing the activities as site fractions, this becomes:

x(1+ x)
K=
(=x
and from equation (5.4),
x(1+x) AG
—_ = —2,303 —.
log (ix) RT

This has a form similar to the Néel migration law equation (5.26).
(e) Covalent bonding. This is an important consideration, and where it occurs
it counts more heavily than the preceding four factors. In the close-packed
structure, the extension of the filled 2p orbitals of the anion causes overlap
into the space defined by the orbitals on the cations, which (if they are empty)
can accept a share of the O?~ electron pairs. Each O7~ ion has three
octahedral and one tetrahedral cation near neighbours, with the tetrahedral
site lying [111] to the quadrant of the octahedral sites.
Since the oxygen p orbitals are mutually perpendicular, if a p orbital is
directed towards the tetrahedral site then others cannot overlap any octa-
hedral sites, and vice versa, and one has a preference depending upon the
hybridization possibilities (2.2), of the A and B cations. For example,
tetrahedral hybrid orbitals can arise with cations containing full d shells and
empty s and p shells (Zn?* and Cd?* are excellent examples). These hybrid
sp? orbitals produce enhanced overlap with O?~—2p orbitals and make a
strong driving force for the Zn?* and Cd** ions to occupy tetrahedral sites
and produce normal spinels. d° ions with a symmetrical half-filled shell also
produce good sp? hybrids, and drive ions such as Fe** or Mn?* to occupy
tetrahedral sites to benefit from the enhanced orbital overlap and electron
sharing. Hence Fe** in spinels moves towards inverse configurations, except
when Zn?* or Cd?* are present.
Octahedral hybrid orbitals are preferred when d orbitals as well as s and p
are available to give the combination d’sp*. These orbitals have directions
towards the faces of a cube and can form strong overlap with oxygen 2p
orbitals. Ions such as Cr>*, V°* have the configuration able to produce the
empty d?sp? orbitals and benefit from the covalent electron sharing with O**
whilst occupying octahedral sites, therefore these ions are strongly driven to
octahedral sites and normal spinel formation. When ions have (say) the d°
configuration, as in Fe?*, Co**, Rh°*, the spin of two electrons must be
reversed, which gives a changed magnetic behaviour for these ions when they
are behaving covalently. This can be detected experimentally. Coplanar
 REFRACTORY OXIDES 305

square orbitals arise from dsp? combination and enhance octahedral site
occupancy for d* ions.
The final degree of inversion in any spinel is an interplay of these five
factors,’* but in particular cases one can dominate and cause strong normal
or inverse tendencies. Most of the important uses of spinels arise from the
degree of inversion. This is illustrated by magnetite, FeFe,O,, ie. Fe;O,,
which contains Fe** ions (= d>) and therefore no crystal field stabilization
energy. Also, the half-filled d shell gives rise to an S state with preference for
sp hybridization and tetrahedral site occupancy. On the other hand, Fe?*
(=d°) has a strong octahedral site preference, all of which produces
(Fe** )te (Fe? *Fe?*),.¢0.4, and gives rise to a random array of Fe?*+ and
Fe** ions on equivalent octahedral sites, which explains why this is a
hopping charge semiconductor. Addition of Zn with its dominant tetrahedral
site occupancy can alter these characteristics, until at (Zn?*),..(Fe?* 2404
the conducting mechanism is completely lost.

5.9.3. Preparation and reactivity

(a) Preparation. The most direct method of making spinel itself is to fuse a
mixture of Al,O,+MgoO in an electric arc furnace. This produces a dense
refractory grog material, but is expensive. The spinel will form at relatively
low temperatures, in the range 1450—1600°C, from magnesite and Al(OH),.'®
Mineralizers such as V,O,, NaCl and MgCl, are reported to lower these
temperatures, but the spinel so obtained has a low reactivity, and high firing
temperatures for fabrication are necessary. Spinel does start to form around
1050°C, but complete formation needs prolonged firing above 1450°C. More
complete conversion at low temperature is brought about by addition of
1.5°% AIF, to the green mix. AIF, accelerates the y—« transition in Al,O3.
During calcination, AIF, pyrohydrolyses to gaseous HF and Al,O3, and
therefore it does not contaminate the product. In the presence of AIF3,
conversion to spinel is almost complete at 900°C, with resultant small particle
sizes in the range 0.5—10.0 1m bringing enhanced reactivity. The enhanced
reactivity leads to high-fired densities ~ 3520kgm~* after only 1 hour
firings at 1600°C, and’a modulus of rupture in the range 172.3-193.0 MN m~ a
Another route to reactive powders is the coprecipitation of hydroxides,
oxalates or sulphates and calcining at 1000°C.
A method of obtaining spinel from waste slags containing Al,O,;-CaO—
MgO-SiO, has been described. The slag is fused and rapidly cooled, when
MgAI,O, and calcium disilicate are formed, but the rapid cooling causes the
calcium disilicate to form as fine dust, easily separated.
(b) Reactivity. Spinel is more refractory and chemically inert than a-Al,O,
and it has, also arising from its cubic structure, an isotropic expansion and a
high mechanical strength. The structure has, in general, the opportunity to
exhibit ranges in stoichiometry, because only a fraction of the available
cation sites are used. This latter feature plays an important part in the ability
306 CERAMIC SCIENCE

to sinter the material to translucency, which can be achieved by sintering in

the presence of small excesses of MgO above the stoichiometric value.
Translucent envelopes for sodium vapour lamps have now been fabricated
and shown to be excellent high temperature materials, whose life is limited by
eventual attack and formation of f and f’—Al,O, (5.4.7) at grain boundaries.
The reactivities of other spinels depend on the cations present in the
structure, and to a degree on the » parameter. At the limit of the » parameter
some combinations can be very unstable, e.g. monocalcium aluminate.’
(c) Valency disordering. By considering the possibilities for non-stoichiometry
in the structure (in an extreme sense) allied to the inversion possibilities, the
spinel structure leads to other important structure types. For example, the
eight Mg?* ions can be replaced by four M* and four M°?* on the octahedral
sites, and when M* is Lit and M3* is Al** the unit cell composition is
Li,Al,9O3, = LiAl;Og, which is a garnet.

5.10 ABO,-type oxides

Most of the ABO3-type oxides crystallize in the relatively simple structure of
the mineral perovskite, CaTiO3, or in a closely related structure. Although
not encountered as refractories in the accepted sense, they do form an
important class, since they are useful in devices that can be utilized at high
temperatures in hostile environments. For example, LaCoO; is used as a
catalyst in the oxidation of CO in car exhaust systems. Many ABO, oxides
are used in transducer devices such as strain gauges and in ultrasonic
equipment; YAIO, is a laser host material; thus these ionic oxides are
modern materials of growing importance.

5.10.1 Structure
Most have the perovskite structure which, when ideal, is cubic with a unit cell
size about 0.39 nm containing one formula unit in the space group Pm3m.
The B ions have octahedral environments of six oxygen anions while the A
cations have 12-fold coordination. The structure can be visualized in several
ways, but the most useful is probably to pursue the idea of polyhedral sharing
outlined in (5.3.7). In this manner, it is considered to be based on ReO, with
corner-sharing BO, octahedra arranged along the cube axes, and the A ions
occupying interstices in this network structure. For the cubic structure to be
stable, the A and B ions should have the correct radius ratios with O7~ to
form 12- and 6-fold coordination, hence rg > 0.09nm and rg > 0.051 nm. As
a result the ideal cubic structure is found in only a few cases—even CaTiO,
itself is distorted. Assuming that the ions all just touch, the relationship (5.27)
must be fulfilled for the perfect perovskite structure to form:

ratty =4/2(rp+7o)- (5.27)

 REFRACTORY OXIDES 307
Depending on the degree of interaction between A and B with O?~, there is
a degree of tolerance in the above relationship given by the tolerance factor
t, where

i=
(ra +10)
J/2rg tro)
The perovskite type structure is usually obtained when t is in the range
0.75 < t < 1.0. These limiting values may vary depending upon the set of
ionic radii employed, but comparisons and predictions can be made as long
as one consistent set is used.
The severity of deviations from the ideal t = 1, cubic structure, can be
discussed in terms of this tolerance factor.**

5.10.2 Properties
These were the first reported oxide ferroelectric materials. Ferroelectricity is
generaliy found in perovskite oxides with small B ions (Ti**, Ta** ) and large
A ions (K*, Pb?*). The size of the octahedral interstices formed by AO,
layers with large ions is also large, and therefore small B ions can be displaced
at low temperature from the centre of symmetry. When these displacements
are cooperative, with all the dipoles oriented in one direction, spontaneous
polarization occurs. The tetragonal form of BaTiO, has this type of displace-
ment in the <001>, and application of an electric field along this direction
pushes the Ti** to the equivalent position across the centre of the octahedron.
Reversals in this manner give rise to hysteresis and ferroelectricity. Some,
like PbZrO, have displacements of the B ions such that no residual dipole
occurs and strong electric fields are needed to align them; thus PbZrO, is an
anti-ferroelectric. Ferroelectric perovskites are piezoelectric and are used to
convert electrical pulses to mechanical oscillations and vice versa.
When the B cations are transition metals with d electrons, a wide variety
of magnetic properties arises. LaMnO, exhibits ferromagnetism by coupling
between spins of Mn>* ions by 180° superexchange involving the O?~ ions
which link them. In general, the magnitude of the interaction depends not on
size as much as the electronic configuration of the B cations. When both B
cations have the configuration 13, e° or partly filled 3, e; the interaction
produces anti-ferromagnetism. B cations with different electron configura-
tions, one ¢§, e, and one t3, ef, produce ferromagnetism. Some, like
SrVO,, are Pauli paramagnets.
Like the magnetic behaviour, a wide range of electrical properties is
encountered: BaZrO, has no d-electrons at B sites and is an insulator,
LaFeO, has a high resistivity, SrRuO; is a metallic conductor like SrVO3.
Most ABO, oxides are, however, semiconductors. These behaviour patterns
are predicted by considering only the covalent interactions between orbitals
on the B cations and the six surrounding oxygen anions, the A cations
308 CERAMIC SCIENCE

presenting a secondary disturbing influence depending on their size. When

the overlap between B orbitals and O?~ orbitals is small, the orbitals in the
electronic energy level scheme are localized and semiconduction results.
When the overlap integral passes a critical value the orbitals become
delocalized and the behaviour then depends on the extent to which they are
filled. Partial filling results in metallic conductivity. The nature of the band
electrons depends upon whether m-bonding between B and O?~ is large or
whether o-bonding is large; in the former case (LaTiO3) the band is a
collective T3,—n* band. Thus in these ceramics careful materials engineering
can be applied to the occupancy of the B sites and to the average size of
cations on A sites in order to produce materials with desired electrical and
magnetic properties, which make them an exciting and valuable range of
modern ceramics.*®

5.11 Ferrites
Ferrite is a technical name covering oxide ceramics containing Fe?* and/or
Fe** ions.*”7 Many of them fall in the two classes we have already discussed
(the cubic spinels and the perovskites) but some important ones belong to a
related family of structures. The ferrites are classified according to their
magnetic properties.
(a) Magnetically soft. These ferrites all belong to the cubic spinel family,
MFe,O, or FeO, itself. They can be engineered, depending on the selection
of M, on composition, purity, preparation or magnetic or thermal after-
treatment, to conform to sought-after properties. For use in high frequency
transformers or deflection coils, they need high permeability with low loss at
high frequency. For use in computers and memory devices, they need
square-loop B-H curve characteristics. For accelerometers and ultrasonic
generators, they can be made with low permeability and large magneto-
striction, while for phase shifters and modulators they can be fabricated with
good gyromagnetic resonance characteristics.
(b) Magnetically hard. These are hexagonal ferrites related in the following
way to the spinels. The oxygen ions are hexagonally close-packed rather than
cubically close-packed. The stoichiometry is then ill-defined, because in
selected layers one of the O?~ ions is replaced by Ba**. When it is the fifth
layer, the composition is ideally BaFe,,O, 9, with the Fe** and Fe**
occupying interstitial sites of which there are now five types instead of the two
in spinels.
When the Ba?* replacement is in the seventh layer, the composition is
BaFe,,O,,, and in this way a family of ferrites can be seen to exist. Having
an hexagonal structure, they are magnetically highly anisotropic, finding uses
in loudspeakers, television, telephones, generators and motors.
(c) Orthoferrites. These have the ideal composition MFeO, and usually
 REFRACTORY OXIDES 309

possess high coercivity allied with ferroelectricity as described in 5.10.

Recently a method of preparation has been developed for rare earth ortho-
ferrites that removes one uncertainty in their behaviour, i.e. the composition.
Well-defined crystals of LnFe(CN), -5H,O can be precipitated from solutions
of LaCl, and K;Fe(CN),. Calcination of the crystals in air leads to LaFeO,
of guaranteed Ln: Fe ratio.

References and further reading

lt. Badkar, P. A. and J. E. Bailey, The mechanism of simultaneous sintering and phase trans-
formation in alumina. J. Mater. Sci., 11, 1794 (1976).
2: Bailey, J. E., D. Lewis, Z. M. Librant and L. J. Porter. Phase transformations in milled
zirconia. Trans. J. Brit. Ceram. Soc., 71, 25 (1972).
3. Baker, B. H., B. Brezny and R. L. Shultz. The role of carbon in MgO refractories. Am. Ceram.
Soc. Bull., 54, 665 (1975).
. Baptista, J. L. and J. White. Phase equilibrium relations of the spinel phase in the system
MgO-Al,O,—CaMgSiO,, MgO-Cr,0,—-CaMgSiO, and MgO-(Aly 5Cro.5),03-
CaMgSiO, and their bearing on spinel bonding in magnesite-chrome refractories. Trans.
J. Brit. Ceram. Soc., 78, 122 (1979).
. Banerjee, M. and D. W. Budworth. The preparation of transparent magnesia bodies. Trans.
J. Brit. Ceram. Soc., 71, 51 (1972).
. Bettman, M. and L. I. Terner. On the structure of Na,O:4MgO-15Al1,03, a variant of
B-alumina. Inorg. Chem., 10, 1442 (1971).
. Bradbury, C., P. M. Callaway and D. D. Double. The conversion of high alumina cement/
concrete. Mat. Sci. Eng., 23, 43 (1976).
. Buljan, S. T., H. A. McKinstry and V. S. Stubican. Optical and X-ray single crystal studies
of the monoclinic = tetragonal transition in ZrO,. J. Am. Ceram. Soc., 59, 351 (1976).
. Carswell, G. P., M. P. Crosby and D. R. F. Spencer. Development of high performance
ladle linings. Refract. J., 55, 9 (1980).
10. Clausen, N. and J. Jahn. Umwandlungsverhalten von ZrO,-Teilchen in einer Keramischen
Matrix. Ber. Dt. Keram. Ges., 55, 487 (1978).
sl Clausen, N., R. Wagner, L. J. Gauckler and G. Petzow. Nitride-stabilised cubic zirconia.
J. Am. Ceram. Soc., 61, 369 (1978).
12. Clausen, N. Stress induced transformation of ZrO, particles in ceramic matrices. J. Am.
Ceram. Soc., 61, 85 (1978).
13. Crucean, E. and B. Rand. Calcination of zirconia gels. Trans. J. Brit. Ceram. Soc., 78,
58 (1979).
14. Cooley, R. F. and J. S. Reed. Equilibrium cation distribution in NiAl,O,4, CuAl,O, and
ZnAl,O, spinels. J. Am. Ceram. Soc., 55, 395 (1972).
LS: Datta, R. K. and R.,Roy. Equilibrium order-disorder in spinels. J. Am. Ceram. Soc., 50,
578 (1967).
. Davidge, R. W. and A. G. Evans. The strength of ceramics. Mater. Sci. Eng., 6, 281 (1970).
. Gottardi, V., M. Guglielmi, A. Tiziani and G. Carturan. AZSC refractories prepared from
gels. J. Non-Cryst. Solids, 43, 105 (1981).
. Hing, P. Fabrication of translucent magnesium aluminate spinel and its compatibility with
sodium vapour. J. Mater. Sci., 11, 1919 (1976).
Lo} Hiibner, H. and W. Jillek. Sub-critical crack extension and crack resistance in poly-
crystalline alumina. J. Mater. Sci., 12, 117 (1977).
20. Jenkins, H. D. B. and T. C. Waddington. Lattice energies, charge distributions and thermo-
chemical data for salts containing complex ions. Nature, 232, 5 (1971).
21. Kirchner, H. P. Strengthening of Ceramics, Marcel Dekker, New York (1979).
ie Kirchner, H. P., R. M. Grover and R. E. Walker. The strengthening of polycrystalline
alumina by crystalline surface layers. Trans. J. Brit. Ceram. Soc., 70, 215 (1971).
23. Kruse, E. W. and M. E. Fine. Precipitation strengthening of MgO by MgFe,O,. J. Am.
Ceram. Soc., 55, 32 (1972).
310 CERAMIC SCIENCE
Chem.
24. Holm, J. L. The lattice energy and thermochemical properties of NaAIF,. Acta.
Scand., 27, 140 (1973).
Acta, 8,
25; Lister, M. W. Some aspects of the calculation of lattice energies. Thermochimica
341 (1974).
Ceram.
26. Mackenzie, K. J. D. and P. J. Melling. The calcination of titania. Trans. J. Brit.
Soc., 73, 23 (1974).
and other
Dil Maczura, G., T. L. Francis and R. E. Roesel. Special aluminas for ceramics
industrial applications. Inter. Ceram., 3, 206 (1976).
28. McGeehin, P. and A. Hooper. Fast ion conduction materials. J. Mater. Sci, 12, 1,977):
t
29: Mazdiyasni, K. S., C. T. Lynch and J. S. Smith. Cubic phase stabilisation of translucen
yttria-zirconia at very low temperatures. J. Am. Ceram. Soc., 50, 532 (1967).
30. Morgan, P. E. D. Low temperature synthetic studies of B-Al,O3. Mat. Res. Bull., 11, 233
(1976).
Sil Park, K. and E. M. Logothetis. Oxygen sensing with Co, _,.Mg,O ceramics. J. Electrochem.
Soc., 124, 1443 (1977).
32. Paul, A. and S. Basu. Effect of ligand field stabilisation on the thermodynamic properties of
solid solutions with spinel-type structure. Trans. J. Brit. Ceram. Soc., 73, 167 (1974).
33: Pauling, L. Nature of the Chemical Bond, Cornell Univ. Press (1948).
34. Powers, R. W. The electrophoretic forming of f-alumina ceramic. J. Electrochem. Soc., 122,
490 (1975).
35, Ramadass, N. ABO3-type oxides—their structures and properties. Mater. Sci. and Eng., 36,
231 (1978).
36. Reed, J. S. and A. M. Lejus. Effect of grinding and polishing on near-surface phase trans-
formations in zirconia. Mat. Res. Bull., 12, 949 (1977).
37: Reed, J. S. and R. M. Fulrath. Characterisation and sintering behaviour of Ba and Sr
ferrites. J. Am. Ceram. Soc., 56, 207 (1973).
38. Rickerby, D. G. Observations on the hardness anisotropy in MgO and LiF. J. Am. Ceram.
Soc., 62, 222 (1979).
39: Rivier, M. and A. D. Pelton. A new slip-casting technique for laboratory fabrication of
B-Al1,O, and other ceramics. Am. Ceram. Bull., 57, 183 (1978).
40. Snow, J. D. and A. H. Heuer. Slip systems in Al,O3. J. Am. Ceram. Soc., 56, 153 (1973).
41. St. Jacques, R. G. and R. Angers. Creep of CaO-stabilised ZrO,. J. Am. Ceram. Soc., 55,
571 (1972).
42. Stubican, V. S. and S. P. Ray. Phase equilibria and ordering in the system ZrO,—CaO. J. Am.
Ceram. Soc., 60, 534 (1977).
43. Subba Rao, E. C., H. S. Maiti and K. K. Sirivastava. Martensitic transformation in
zirconia. Phys. Stat. Sol., 21, 9 (1974).
44. Swain, M. V., R. M. Latanision and A. R. C. Westwood. Further studies on environment-
sensitive hardness and machinability of Al,O,. J. Am. Ceram. Soc., 58, 372 (1975).
45. Veiga, M. L., M. Vallet, A. Jerez and C. Pico. Sintering of ZrO, gels. Ann. Chim. Fr., 6,
345 (1981).
46. Wang, J. C., M. Gaffari and S. Choi. On the ionic conduction in f-Al,O;: potential
energy curves and conduction mechanism. J. Chem. Phys., 63, 772 (1975).
47. Webster, R. and B. Jackson. The property requirements for fired magnesite refractories used
in oxygen-steel-making vessels. Trans. J. Brit. Ceram. Soc., 74, 233 (1975).
48. Yan, M. F. Microstructural control in the processing of electronic ceramics. Mat. Sci. Eng.,
48, 53 (1981).
49. Yust, C. S. and C. J. McHargue. Model for the deformation of hyperstoichiometric UO). J.
Am. Ceram. Soc., 56, 161 (1973).
6 Special ceramics—refractory
hard metals |

This chapter will consider some of the properties and uses of metal carbides,
nitrides, borides and silicides. To do this, it is convenient to introduce a
classification scheme. First, from relative sizes we can consider the small
atoms carbon and nitrogen as one sub-set, with boron and silicon as a
separate set, because their relative large size leads to non-metal orbital
overlap and non-metal networks dominate their structures. Then for each of
the two sub-sets we can can identify four classes depending on relative
electronegativities of the metals and non-metals. Taking carbon (with an
electronegativity of 2.5) as example, a continuous and systematic change
occurs across the periodic classification.
Class 1 includes periodic groups IV to VIII, where metal electronegativities
are in the range 1.3—2.2. Across this class there is a steady change from stable
to unstable compounds, and structures change from cubic to austenite. This
class is often termed the interstitial group, and in practical terms is the most
important, containing at one end most of the hardest and refractory materials
we possess’? and at the other end some of the more unstable materials that
give to steels their important properties.’® Class 2 contains periodic group
Ill, the lanthanides and actinides. Electronegativity values range from
1.1-1.81 and so some overlap with Class 1 is expected. It smooths out the
transition from the next class, Class 3, across to the interstitials. Class 3
contains periodic groups I and II with electronegativities in the range 0.7-1.2.
Such a large electronegativity difference from carbon leads to a very salt-like
class which have almost no value in ceramic terms because of their extreme
chemical sensitivity. Class 4 has already been considered, (2.3), since it
contains some elements from periodic group III and group IV. It is dominated
by covalent bond formation e.g. SiC, graphite, B,C (2.10), (2.5) and (2.7).
This classification is suitable also for the nitrides'’ and shows that we need
investigate only Classes 1 and 2, that is the interstitial carbides and nitrides.
For the borides!? and silicides'* the above classification is not so useful; it is
in fact more convenient to develop a structural classification based on an
L 311
Si CERAMIC SCIENCE
silicate
increasing complexity in non-metal interaction rather in the way the
classification was developed in 1.5.
form a
Nevertheless, by whatever means these materials are classified, they
ss
large group characterized by high melting points, brittleness, high hardne
al inertne ss
and strength at high temperatures, and for most a chemic
and
bestowed by protective oxide layers. They mostly have metallic lustre
electrical and thermal conductivities of the same order as metals. This
in
admixture of properties indicates that ionic (5.3), covalent (2.2), and
the
particular, delocalized molecular orbitals, (2.2.7), must be involved in
t discuss ions of
cohesive forces in these materials. Accordingly the relevan
covalent (chapter 2), and ionic (chapter 5), bonding need to be understood
in order to see how models have been developed to rationalize their
.
properties and predict differences between, for example, carbides and nitrides

6.1 Interstitial carbides and nitrides of classes 1 and 2

6.1.1 Preparation
The non-metals nitrogen and carbon show considerable chemical inertness,
for kinetic rather than thermodynamic reasons. Difficulties are associated
with strong binding of the atoms in the elemental form (2.2.6). Hence the
methods of preparation reflect this in the temperature used or, in the case of
carbides, by utilizing unstable, volatile carbon species capable of producing
carbon in a reactive form. Equations 6.1-6.12 represent technologically
established processes.
C+M=MC (6.1)
C+MO=MC+CO (6.2)
C+MO+M’> MC+M’O (6.3)
2CO+M > MC+CO, (6.4)
H,C, +H, +M > MC+H,+H,C, (6.5)
CaC, +2MO > MC+CaO (6.6)
H,C, +H, + MX (gaseous halide) > MC + HX (6.7)
M(CO), +H, > MC+H,C, (6.8)
M’'CO;+ Maan MC+CO, + M'O (6.9)
2M +N,=2MN (6.10)
M+NH,=MN+3/2H, (6.11)
2MO+2C+N,=2MN+2CO (6.12)
 SPECIAL CERAMICS BS

The direct methods (6.1) and (6.10) require very high temperatures, in the
case of carbides in excess of 1800°C, which usually involves arc melting in a
carbon electrode furnace. Nitride preparation can occur at lower temperatures
around 1000°C.'’ They have another disadvantage as well as the need
for high temperatures—they suffer from reversibility, dependent upon the
temperature, pressure and cooling rates. Solid as opposed to liquid state
carburization (6.3, 6.4) has the attraction not only of operating at lower
temperatures but also of being more controllable, as CO evolution reduces
the reversibility problem. Furthermore, the method yields a smaller grain size
which is of some practical importance.
Methods based on carbiding from the gaseous phase, equations (6.4), (6.5),
(6.11), where the non-metal element is deposited in a reactive state, are
frequently used—these have the advantage that an article can first be
fabricated from the metal, and can then be inductively heated in a hydro-
carbon atmosphere for carbide production,’ or in ammonia for nitride
manufacture.’
Method (6.6) can be used to make only the most inert carbides, otherwise
the removal of CaO becomes a major problem. Equation (6.9) represents the
Andrieux—Weiss process and involves mainly the carbonates of periodic
groups I and II.

6.1.2. Compositions and non-stoichiometry

In the two classes of carbide and nitride being considered, the compounds
have exceedingly high cohesive energy and their structures are typically based
on six-fold coordination of metal atoms around carbon and vice-versa (6.1.3).
Some portion of these strong cohesive forces must be from delocalized
electron-deficient bonding, consequently the compositions of the phases are
not determined by the simple valency considerations that apply to the
majority of ceramics.
(a) Class I carbides.'*:?° The carbides in this class exist as three distinct phases
at high temperature, each of which is non-stoichiometric. First, M takes up
carbon in solid solution (from 5 to 10%) to form the «-phase, which then
produces a eutectic with the f-phase when the carbon content exceeds the
solubility limit. The B-phase is hexagonal, with a mean composition M,C
and is not widely non-stoichiometric at low temperatures but becomes so at
high temperatures. The non-stoichiometry in the B-phase results from carbon
deficiency, not metal excess.
The y-phase arises from increased carbon content, or by peritectic decom-
position of B-phase at high temperatures. It is a widely non-stoichiometric
phase, MC, with the rock salt structure. Once again the non-stoichiometry
is caused by carbon vacancies and not excess metal. For all systems examined
in detail, it appears that the »-phase never attains a metal/carbon ratio of 1,
and any composition of MC,9 is a mixture containing graphite. The reason
for not attaining a 1:1 ratio of metal to carbon is probably intrinsic and must
314 CERAMIC SCIENCE

be interesting. Some doubt exists as to the intrinsic nature of the non-

stoichiometry because nitrogen and oxygen impurities may be present.
However, the general trend of the results outlined below, and the way some
of the bonding theories predict a degree of intrinsic non-stoichiometry,**:*°
does support the view that the stoichiometric composition does not exist.
Because not all of the octahedral sites in the close-packed metal array are
occupied by carbon even when the composition approaches MC, the carbon
atoms can have ordered arrangements that may require unit cells which are
multiples of the simple structures.° Some carbides, for example those in the
V-C system, depending on temperature and carbon content show a variety of
ordered structures!* with some of the phase changes occurring at very high
temperatures.
The structures are often described as close-packed: in these, five different
stacking variations of the metal atoms can be distinguished. A convenient
method of presenting the stacking sequence is to describe all the structures
with hexagonal unit cells and to show atom arrangements in the (1 120) plane
therein. The number of octahedral interstices in any close-packed structure is
equal to the number of atoms and is independent of the way the hexagonal
metal layers are stacked. Thus the composition MC is expected for all
sequences but in practice this is only found when the metal layers adopt cubic
(c) stacking. For h, hcc, hhc and hhcc, not all interstices are occupied,
probably because in h stacking the octahedral sites are located directly above
each other in the (1120) plane which leads to occupation of only one site.
Using this criterion, and assuming complete occupation of sites adjacent
to ac layer, the following limiting compositions for close-packed carbides
emerges: M,C for h, M,C, for hcc, M3C, for hhc, M,C, for hhee and MC
for c. Thus the f and y phases are at the ends of an ordered series that is found
in some systems, for example molybdenum:

a-Mo,C
= p-Mo,C
B-Mo,C+C= Mo,3C,
Mo,C,+C=MoC, _,.

The intermediate Mo,C, phase appears from neutron diffraction studies to

have a random vacancy arrangement amongst the carbon sites.
Figure 6.1 shows the known close-packed stacking sequences in this class
of carbide and how the Jagodzinski-Wycoff symbols are derived for each
lattice: a layer is denoted by h if the two adjacent layers are of the same type,
by cif they are of a different type.
Table 6.1 shows typical composition ranges for some of the f and y phases;
in general the stoichiometric range decreases in crossing from periodic group
IV to VI and increases in descending any periodic group.
(b) Class 2 carbides. Metal-rich and carbon-rich phases are known to exist in
 SPECIAL CERAMICS BulS

aa0dna

Figure 6.1 Stacking sequences of hexagonal metal atom layers in close packed class 1 type
carbides and nitrides. (a) = M,C (h), (b) = M3C; (hcc), (c) = M3C, (hhc), (d) = M4C; (hhcc),
(e) = MC (c). Large circles are metal atoms, black circles are possible ordered carbon positions.

this class. The lanthanides have a non-stoichiometric phase with composition

in the range M,C—M.,C based on partial octahedral filling in a close-packed
metal array. It is a relatively low temperature phase that at high temperatures,
or when heated with more carbon, produces the sesquicarbide phase M,C,
which has little or no stoichiometric variation from the stated formula.
Sesquicarbide phases are cubic and contain the catenated unit C}~ which is
an energetically favourable species.” as shown in (2.2.7). Even though the
sesquicarbide is a stable phase, it gives way at high temperatures or when
heated with more carbon, to the MC, phase which has a tetragonally
distorted rock salt structure caused by the C3~ anions aligning in the [001]
and occupying the Cl” positions in the parent structure. Both M,C, and
MC,, although possessing the C3~ cations, still exhibit a considerable
variety of metallic properties, not least of which is appearance.
The actinide members of this class form M,C, and MC, phases identical
to those just described, but also exhibit a close-packed rock salt structure

Table 6.1 Composition ranges of Class 1 carbides

Phase Range Temperature °C

B-VC VCo.59-VCo.42 Room temperature

VCo.50-VOo.36 1630
B-NbC NbCo 59-NbCo 45 1000
NbCp 59-NbCp.39 2320
B-TaC TaCp 59—-LaCo.38 2900
B-MoC MoCo,59>—MoCo,44 1400
p-WC WC .52-WC,36 2710
y-TiC TiC. 97-TiCo.28 2900
TiG hos Lio 56 1645
y-VC Vg gAVEa 2160
316 CERAMIC SCIENCE
different in one
phase centred on the composition MC. This y-like phase is
a range where
major respect to Class 1 y phases, in that not only does it have
phase where
carbon vacancies occur, MC, _,, but passes through MC to a
atures the
metal vacancies occur: MC,,,- Furthermore at high temper
nt to metal
MC, .., phase merges into the MC, phase as carbon atoms adjace
vacancies catenate to become C3” species.
and the
(c) Class 1 nitrides. There are great similarities with the carbides,
phases do
stacking sequences in Fig. 6.1 apply. However some extra ordered
occur.
the
Periodic groups IV, V and VI all have an a-phase, amongst which
y solid solubil ity of
metals Ti and Zr stand out by having such a high primar
ed. For
nitrogen in the hexagonal metal lattices that their stability is increas
2350°C conside r-
example, the peritectic temperature attained for a-(Ti—N) at
packing
ably exceeds the melting point of the metal. Movement to cubic close
of this
in the y-phase occurs via the hexagonal B-phase, as in the carbides
group
class, but the position is not now so clear because the heavier metals of
form.
IV, Nb and Ta, do not achieve the cubic packing but exist in hexagonal
metal stackin g
The information in Table 6.2 shows the greater variety of
sequences encountered. Wide homogeneity ranges occur around the «, B and
y compositions but as stated some systems show structural variation within
the ranges, a feature unlike that of carbides.
It should be noted that, as in the carbides, the homogeneity range is a result
of nitrogen vacancies and that the M:N ratio of 1:1 is not achieved. One
higher nitride, WN), does exist.
(d) Class 2 nitrides. Very refractory nitrides of the actinides exist, and these
are more stable than the carbides. In general, less complicated phases and
fewer structures exist than in Class 1. The stable nitride in all cases has the

Table 6.2 Some stability and structural details of Class 1 nitrides

Compound Structure Composition range and stability

Ti,N tetragonal peritectic decomposition at 1050°C

B-Ti,N antirutile
y-TiN cubic Range TiNo.4s-liNo.95 2950°C
y-ZrN cubic ZINp.5-ZtNo,92 2980°C
y-Hf{N cubic HfNoy s-HfNo,.g7 3310°C
B-VN hexagonal V,N-V,N
y-VN cubic VNo.72-VNo.95
B-NbN hexagonal NbN, 3-NbNo,; 2400°C
y-NbN cubic NbN o.s—NbNo.9
hexagonal NbNo >—NbNo.98 2050°C
WC-type
a-TaN hexagonal Ta-TaNo 05
B-TaN hexagonal TaNo.4—-1aNo.45
y-TaN simple TaNo.g—-1aNo.9
hexagonal (WC)
hexagonal TaNo.9—TaNo.93 3090°C
 SPECIAL CERAMICS 317

rock salt structure and MN composition. An M,N; hexagonal or cubic phase

exists!* and a cubic UN, with the fluorite CaF, structure is known.
The lanthanides have only the «-solid solution phase and very refractory
rock salt MN phase.

6.1.3 Properties and uses

(a) Thermal properties. The Class 1 carbides have the highest melting
temperatures known, and carbon may be sublimed from crucibles made from
sintered TaC. An alloy consisting of 80% TaCy93+20% HfCo.o3 has the
highest melting point yet recorded, 4050°C. Generally, for any one metal
system the melting point decreases in the order MC > MN > MO, for
example TiC (3150°C), TiN (2950°C), TiO (1750°C), but the order may
appear to change if stoichiometry is not taken into account.
Where the f-phase exists it is only stable at much lower temperatures and
decomposes peritectically to the y-phase plus metal. There is a gradual
decrease in the melting or decomposition temperature in passing from
periodic group IV to VI for both carbides and nitrides in this class.
Accompanying the high melting behaviour is a composition-dependent
vaporization rate which shows that minimum rates are achieved for alloy
compositions, for example 80/20 TaC/HfC has a minimum value around
10-°kgm~?s~! at 2425°C. In the case of nitrides, the high melting points
are only realized under nitrogen atmospheres because they do not vaporize
congruently but as the metal plus N,, and so preferentially lose nitrogen.
Linear thermal expansion coefficients are very composition-dependent, as
Table 6.3 shows.
It is notable that one non-stoichiometric composition for TiN, around
x = 0.8 exhibits the minimum expansion, which must be associated with the
greatest bonding forces being present at that composition. It is true that for
any given composition the expansion of the carbide is less than the nitride.
For Class 2 carbides and nitrides, the MC, phase is refractory, melting in
the range 2200—2700°C but possessing vaporization characteristics such
that metal is removed at high temperature and the structure then becomes
two-phase, graph+ MC). All the dicarbides undergo a brittle martensitic
ite
transformation (5.5.4) when the tetragonal structure changes to cubic in the

Table 6.3 Linear thermal expansion

Expansion (%)

Phase Z>-E 2502C 500°C

TING 2 0 0.170 0.373

GaN oe 0 0.144 0.350
TiNo 95 0 0.169 0.374
TiC ee 0 0.155 0.341
318 CERAMIC SCIENCE

range 1050°-1820°C. Suitable alloying additions can depress this transition

to below ambient temperature, enabling the cubic modification to exist at
room temperature.?” The sesquicarbides all decompose peritectically below
1800°C.
The «-nitride phase is invariably more stable than the carbides in contrast
to Class 1 for example; whereas it is doubtful if lanthanide monocarbides
exist,? the LnN phase is very refractory. Furthermore, UN (2650—2850°C)
has a melting point which is composition-dependent but exceeds UC at
2560°C throughout the range observed.
(b) Mechanical behaviour. The notable feature of all Class 1 and Class 2
nitrides and carbides is their extreme hardness and brittleness.** Here again
the property shows an extreme composition sensitivity; for example, the
room temperature microhardness of TiC, and TaC, phases show contrasting
behaviour: TaCy g, to TaCy.97 shows a linear decrease in hardness from
2750kgmm~? to 1400kgmm~', whereas TiC), to TiCo47, shows a
marked linear increase from 2000 to 2750kg mm ~~”.
The nitrides are less hard than the corresponding carbides. Some example
are: TiN,, 1700-1900kg mm? and VN,, 1520-1900 kg mm ?.
General mechanical behaviour of the cubic carbides® and nitrides is similar
to that of diamond and silicon—the slip systems, elastic behaviour at room
temperature and plastic behaviour at elevated temperatures are identical.
Ductile-brittle transition occurs over wide ranges: 1725—1980°C for TaC,,
and 800—1100°C for TiC,,.
Below the ductile-brittle transition temperature the strength is composition-
dependent, but not always in a simple manner, as Fig. 6.2 indicates.

194

138

- m2
MN
bending
Median
stress
Ductile
brittle
zone 82

0-8 0-9 1:0

Carbon content x in Ta Cc,

Figure 6.2 Median bending stress in the ductile-brittle zone v. carbon content for tantalum
carbide.
 SPECIAL CERAMICS 319

(c) Chemical stability. So far, these ceramics appear exceptionally stable with
respect to melting, evaporation and decomposition, but to be ultimately
useful they must have some degree of stability in a range of atmospheres at
high temperatures. The thermodynamics of possible reactions of these
materials with surrounding atmospheres indicate how extensive such re-
actions can be. Therefore the data in Table 5.2 and the relevant equations in
(5.2) must be utilized. However, as always, the extent of any possible
reactions will depend on the detailed mechanisms and their governing
kinetics. For example, for the reaction

VC@) + 2H 2g) > Vis) + CHagg)

the equilibrium constant k,, defined as k, = pcu,/ Ph, by neglecting the
activities of the solid phases, can be calculated using the equations given in
(5.2). A mean value for k, at temperatures around 2000°C is 6 x 10~’, and so
use of VC in one atmosphere of hydrogen indicates low partial pressures of
CH, (around 6 x 10-7 atm.), indicating a high stability. A similar calculation
for MgOw) + Hog) =H20¢)+Mgg yields k,~1x10~°, and if py, = 1
this equals pyg X Py,o Which gives a magnesium vapour pressure around
5x 10° 3 atm., an appreciable partial pressure, and so reaction will proceed at
a significant rate. This indicates that Class 1 carbides are in general useful
refractories in reducing atmospheres compared to most oxides discussed in
chapter 5.
For Class 2, reaction with hydrogen is even more inhibited as long as no
lower carbide exists. For example, MC, + H, > Mg) +C,H, has a k, value
around 10737, and for MC,+2H, > Mg)+CHy, a value around 107*?.
However, if a stable lower carbide exists, as in the case of uranium, it appears
to facilitate the reaction
UC, +2H, =UC+CH,.
For this equation, k, ~ Pcu, |Pit,» and for H, = 1 atmosphere at 1000°C,
Pcu,is 10-7 atm. This reaction can be used to ensure UC phase purity,
when prepared by carbon reduction of UO,, by removing any UC,. Class 1
and 2 nitrides are more stable than carbides in hydrogen atmospheres mainly
because NH, is comparatively less stable than CH,, as data in Table 5.2
shows.
The situation in oxidizing atmospheres is totally reversed because all
classes of carbide and nitride are thermodynamically unstable. Equations
such as
TiC) Se Or) => T1045) ai CO)

have equilibrium constants requiring CO, pressures in excess of 107° atm to

prevent the reaction proceeding to the right at temperatures in the refractory
range. Despite this, some are still useful because they form protective oxide
coatings that inhibit further reaction.”® It is difficult to identify these in a
320 CERAMIC SCIENCE

general way, since their usefulness depends upon the coherence of the oxide
coating and upon its volatility. Class 1 carbides are the most stable and TiC,,
is the most stable of these, but even it shows an appreciable linear oxidation
rate with respect to both temperature and oxygen pressure at 435°C.
Oxidation is mainly dependent upon oxide ion diffusion through the pro-
tective layer, and as a result non-stoichiometric oxides such as WO3_, and
UO,_, allow easier oxidation. For example, at 25°C it takes 30 years to
produce 0.05 % oxidation in UC, but only 4.8 seconds at 200°C, rates which
are orders of magnitude higher than TiC. There is little apparent difference
between nitride and carbide behaviour but both are more easily oxidized in
the presence of moisture, particularly the Class 2 materials with higher
non-metal/metal ratios: LnC, and UC), which are easily attacked by water
vapour at room temperature.” This reaction is called hydrolysis, and as a
general rule it is easy for carbides and nitrides of metals whose oxides are
basic and difficult for metals whose oxides are acidic. Again the inertness is

Table 6.4 Some uses of refractory hard metals

Phases Principal properties Application

TiC,, Resistance to molten alkali metals Pump parts for transferring molten
TiC,—W cermet up to 1100°C and 860kN m~? Na in fast reactors
TiN,,, ZrN,, HEN,, Resistance to molten metals Crucibles and boats
ZrC,,, NbC,.
WUGs Thee High melting point and thermal Cores of fuel elements of nuclear
conductivity, absence of phase reactors
transformations
CrN,, High hardness, high corrosion Protection of moving surfaces when
resistance, high thermal stability, plated-on.
low sinterability
TiN,—Co cermet Very hard Cutting tools
Cre, Chemically inert and resistant to Filters in the chemical industry.
oxidation Electrodes for use in electro-
chemical processes. Pump
bearings. Nozzles for corrosive
liquids and gases.
mC wz. High melting point, high electrical Electrodes for underwater oxy-
conductivity, low evaporation electric cutting of steel. Thermionic
rate cathodes for equipment working
under low vacuum.
ic x
High melting point, hardness and Protective coatings for rocket
heat resistance components. Gas turbine blades
LnC, High melting, low work function Emitters for thermionic power
transformers
TaC High melting point, low vapour Heating elements in electric furnaces
x

pressure, metallic conduction

Mo,C High melting point Solder for high temperature soldering
Cry34Ce Metallic character of conduction. Low ohmic resistances 5—50 ohms at
Small temperature coefficient 400°C
WC. TCs, VC. All constituents of refractory hard
VN, metals
 SPECIAL CERAMICS 321

kinetic rather than thermodynamic, and as a result some Class 1 materials are
not attacked by aqua regia when refluxed.

6.1.4 Uses of cemented hard-metals

The properties of these materials allow their use in a range of sophisticated
applications, some of which are listed in Table 6.4. At present from amongst
the wide range of applications, one dominates and is worthy of more detailed
discussion, namely cemented hard-metal cutting tools.”°
In cemented hard metals, the most obvious properties of hardness and
thermal stability are used to enable tools made from them to be used at high
cutting speeds and to resist tool wear.'* It also follows from the properties so
far described that their essential drawbacks are brittleness and the ability to
self-sinter. Thus, to be ultimately useful they must acquire a bonding matrix
and a degree of toughness. In 2.13.3 it was shown how composite manufacture
automatically bestows improved toughness, and so the way forward was long
ago seen to be the manufacture of a composite in which the matrix is a ductile
metal and in which the refractory phase formed the major (> 94 %) part.
Suitable metal matrices will be few because they have to fulfil as many as
possible of the following criteria.
(a) The metal must form strong bonds between the carbide particles and so must wet (3.5.3) the
hard-metal grains when molten in order to eventually cement them (hence the terminology,
cemented hard metals).
(b) It must not be a carbide or nitride former.
(c) It must have partial solubility for the carbide or nitride at elevated temperatures, but on
cooling the solubility should decrease to zero to preserve the purity of the hard-metal phase.
(d) It should not form a ternary phase.
(e) It should have good oxidation resistance.
(f) It should be as ductile as possible.
(g) To ensure good thermal shock resistance of the composite, a close match of thermal expan-
sion between metal and hard-metal is necessary.
(h) Good self-sintering characteristics of the powdered metal are often required, since melting to
produce shapes is rarely feasible.

Many of these requirements are in conflict, and only an optimum solution

can be found. Most advantage is found in cobalt.
With cobalt as a binder the commonest base is tungsten carbide WC, _,.
A simplified ternary section at 1400°C is shown in Fig. 6.3, which also shows
the nature of the compromise, because brittle ternary carbides are formed if
insufficient carbon is present, or from a peritectic reaction below 1350°C. It
is also implied in the diagram that W,C might form as a result of carbon
extraction into the ternary and this brings with it embrittlement and
oxidizability. Early attempts to avoid this used excess graphite but this has a
deleterious effect on wetting and subsequent metal to hard-metal bonding.
One of the early drawbacks of the WC-—Co composites, or cermets, was
local welding of the WC grains to hard steel work pieces when they were used
as cutting tools. This of course caused tool wear and arose from oxidation of
WC to the volatile WO}. To overcome this problem, substantial amounts of
327 CERAMIC SCIENCE

Figure 6.3 Simplified section at 1400°C of the Co-W-C system. 7 = Co,;W3C, 0 = Co,;W,.C),
k = Co,W,oCa.

TiC, were added to the compositions to provide protection and self-lubrication

at high temperatures. The protection is thought to arise from a tenacious
coating of rutile, TiO, (5.8): first TiC, is oxidized to a TiC,,_,O, oxycarbide
with a good crystal structure match with TiC,,; on further oxidation the outer
layer becomes the cubic phase TiO,, and retains the structural coherence, and
finally TiO, becomes TiO, through a series of non-stoichiometric oxides
(5.8). This mechanism leading to coherence is not available for WC, because
no WO phase exists and the highest oxide is quite volatile, unlike TiO).
Within the system Co—WC,-TiC,, there have been wide composition
variations used by cermet hard-metal suppliers, depending on the properties
sought. TiC lowers the melting point but bestows protection, Co lowers the
melting point but makes the composite tougher, TiC, and WC, increase the
hardness and strength. Some idea of the way the properties of the cermets'
depend on composition can be obtained from Fig. 6.4. In general, Co is
limited to 6-13 vol. %. Fe, Cr and stainless steel matrices have been used in
place of Co but not as extensively.
The basic system has been experimented with in attempts to improve the
product. Addition of tantalum carbide (TaC,,) has brought about useful
improvements through formation of TiC,—TaC,, solid solutions which can
increase the hardness and melting point. It has already been noted that
Tip gl ap..C, is the most refractory material available to technologists.
Improved oxidation resistance is also bestowed by adding TaC,.. Tantalum is
approximately equal to W in size and so there is an increased solubility of
WC into the ceramic phase. Useful compositions usually lie in the two-phase
field (Ta,Ti,W)C,+ WC, shown in Fig. 6.5. Compositions close to the
 SPECIAL CERAMICS 323

- =Young's Modulus

---- =Relative bend strength

+ +++ =Hardness

Figure 6.4 Property values v. composition for the Co-TiC—WC system.

two-phase boundary are preferred because this is the region where the cubic
and hexagonal structures of the carbides have similar stability with maximum
facility, therefore, for stacking fault generation and dislocation locking. NbC,
acts similarly to TaC,. In order to increase ductility at the expense of
refractoriness VC, or Cr,C ; have been added. For extra highly refractory
cermets, HfC,-WC,—Co compositions are used. More exotic compositions
involving MgO to improve ductility have also been reported.

6.1.5 Bonding
Naturally the properties discussed above for these materials have excited
much interest, not only with respect to their potential uses, but in presenting
an overall view of the cohesive forces that is in accord with such complex

Two phase region used

in cermet manufacture

TaC WC

Figure 6.5 Sketch of the ternary system TiC-TaC-WC.

324 CERAMIC: SCIENCE
consider
behaviour. Many approaches have been explored, but here we will
only some highlights in the theoretical development.
,
(a) Hume-Rothery model. For the Class 1 and 2 carbides and nitrides
that
Hume-Rothery noted that hexagonal close packing was not favoured, so
cubic. In the
many hexagonal metals on absorption of the non-metals become
six
fec rock salt structure common to these phases, each metal atom has
interstitial non-metals arranged on three mutually perpendicular axes, whilst
at
in hexagonal close-packed arrangements only some are at 90°, others are
70° to the metal atoms. This suggested that covalent bonding was of major
importance and the C-M bonds must be as near to 90° as possible. Hence
d?sp> hybridization (2.2.6) of the metal before bonding was the most
important feature. In a modified valence bond approach (2.2.6) each carbon
or nitrogen is regarded as forming electron pair bonds using the 2s and 2p
orbitals overlapping with the d?sp* hybrids. Since carbon has four, it gets
2/3e~ from each of the six metal neighbours, producing a type of resonating
covalent bond. Metallic conduction arises because some d’sp® orbitals are
unoccupied, are of low-lying energy, and are therefore available for electron
promotion.
In nitrides, because nitrogen has five valency electrons, each d’sp° orbital
supplies only $e” and therefore bonding in nitrides should be weaker than
carbides. This concurs with the observation that for a given metal the carbide
is harder, stronger and more refractory than the nitride which in turn is more
stable than the monoxide, if one exists. Nitrogen is smaller than carbon, so
the smaller unit cells are expected. Because the cohesive force is covalent, all
these materials are brittle. These then are the reasonably satisfactory pre-
dictions from what is a modified valence bond explanation.
(b) Dempsey model.'° This is an empirical approach outlined here because it
emphasizes the metallic nature of these ceramics. Dempsey noted that the
metals forming stable carbides and nitrides in these classes have in their
free atoms d-shells (2.2) that contain relatively low numbers of electrons.
Secondly, the free metals themselves gain in stability as the number of d
electrons increases, as is illustrated by the melting points: Sc(3d') 1539°C,
Ti(3d2) 1668°C, V(3d3) 1900°C, Cr(3d*) 1875°C. Thus it might be expected
that the early transition metals would accept electrons into their d-bands.
Section 2.2.7 shows that the 3d band cannot remain separated from the 4s
band; a mixed band structure must be assumed. The model assumes that all
the free atom d electrons and all but 0.5 s electrons populate the band. The
next assumption is that all the p electrons from the non-metal find themselves
in this band, leaving positive non-metal cores stabilized by the delocalized
band electrons. Directional aspects are completely ignored.
A plot of assumed d electron number against melting point as a measure of
bond strength is shown in Fig. 6.6, and has a shape from which some general
predictions about these ceramics can be made without recourse to structural
considerations.
 SPECIAL CERAMICS 325

9-5

d-electron
number

5:5

15

3000

Melting point °C

Figure 6.6 Plot of assumed d-electron number against melting point for metals and hard-metals.

(i) The majority of stable ceramics in these classes, that is with melting
points in excess of 2600°C, lie on levels corresponding to 5.5 or 6.5 d-electrons
per metal atom.
(ii) Up to 6 d electrons per metal atom, the bonding increases; beyond,
it decreases as first a bonding band is filled and then the corresponding
anti-bonding level must be used.
(iii) A useful aspect of this approach is that the ceramics can be con-
sidered without detailed knowledge of the structure. For example, the series of
transition metals hafnium, tantalum and vanadium have the outer electronic
structure 5d26s?, 5d°6s? and Sd*6s* respectively. Considering these in com-
bination with carbon (2s? 2p”) (2.2.7), the number of d electrons per metal
atom for the MC,, phase is 5.5, 6.5 and 7.5 respectively. This is in accord
with stable HfC and TaC phases, and a mixed (Ta,Hf)C phase as the most
refractory material available. When the d electrons per metal are calculated
for the M,C phase they are found to be 4.5, 5.5 and 6.5 respectively, which
leads to the view that W,C might be more stable than WC. In general the
expectation is for lower carbides to be more stable on passing across the
transition metal series, which is certainly the case.
(c) Anionic model. X-ray spectroscopic data and electron diffraction measure-
ments do not confirm the transfer of charge from non-metal to metal that is
postulated in the Dempsey model. In Mo,C, for example, all the super-
structure lines in the electron diffraction pattern are weak and arise from
carbon atoms, so that intensity calculations, involving them only, reveal
something about the state of the carbon atoms. When refined, the results
show that carbons have approximately a single negative charge localized on
them. Similarly, N~ is believed to be the state of the non-metals in Fe,N.’°
326 CERAMIC SCIENCE

These negative charge localizations can arise from dp hybridized orbitals

directed to centred non-metal atoms from face-centred metal atoms. Covalent
bonds so formed are highly polarized, (5.3), so that the nitrogen has ionic
character. The greater the ionic character the greater is the chemical
reactivity of the compound. Remaining metal s, p, d electrons achieve
bonding between metal atoms so that these interstitial ceramics have metallic
character. The covalent character produces brittleness, the degree of ionicity
produces some chemical reactivity and the metal-metal bonding gives rise to
the metallic characteristics.
(d) Combined model—Denker approach."' This is a useful semi-quantitative
approach that can be used prior to a full application of energy band
calculations. In effect, it begins where (c) above leaves off, i.e. it is a model
involving an admixture of ionic bonding and covalent bonding that leads to
metallic properties. The following simplifying assumptions and steps are
taken:

(i) Assume the metal atoms are ionized by the “loss” of two d electrons e.g.
Ti(Ar) 3d24s? > Ti2* Ar 3d°4s?, or Mo(Kr)4d°5s! + Mo?* (Kr) 4d°5s*

and that these electrons are accepted into low-lying orbitals of correct
symmetry on the non-metal.
(ii) Extract from the cubic structure one common polyhedral building unit
which in this case is an M?* cation surrounded by 6 non-metal anions X*~
at the apices of an octahedron.
(iii) It is implied that calculations for this octahedron, ignoring interactions
with other MX, units or XM, units that constitute the structure by edge
sharing, will show the behavioural trends in these materials.
(iv) For overall comparison purposes, a property known as the valency
electron concentration, VEC, is defined?! as («x the total number of outer
electrons on the cation) plus (8 x the total number of outer electrons on the
anion), where a and f are the fraction of M and X sites occupied in M,Xz.
For example, TiC,, would have a VEC of 1x 2+1x6 = 8, and MoC, o, one
of 4+0.9 x6 = 9.4.
(v) A molecular orbital diagram, like those described for the systems in
2.2.7, is constructed after the degree of orbital overlap for the M?*X2~ unit
is found. In order to do this, hybridization on the cation is ignored and only
simple s, p and d orbitals are used, whilst on the six anions sp hybridization
producing equivalent sp orbitals directed along the axes towards the M?* ion
is assumed. This last feature implies that since a combination of an s with a p
orbital to give hybrids produces two such orbitals (2.2.7), one directed at 180°
from the other, then the outward pointing sp orbitals are non-bonding. In a
wider picture they may be used to bond neighbouring M?*X2~ units as in
(2.4.1), but this is ignored here.
Within the framework of these wide-ranging assumptions it is possible to
 SPECIAL CERAMICS 327

Table 6.5 Overlap integrals for the M**X2~ unit and Valency
Electron Concentrations for some hard-metals with rock salt
structure

Compound

Orbital pair TiC TiN VC VN

Cation—non-metal
4s—2sp 0.420 0.420 0.420 0.420
4p,—2sp 0.617 0.618 0.619 0.619
4p,—2p, 0.189 0.195 0.203 0.204
3d,—2sp 0.026 0.030 0.034 0.035
3d,,—2p, 0.006 0.006 0.008 0.008
Cation—cation
4s—4s 0.622 0.632 0.644 0.645
3d,—3d, 0.006 0.008 0.009 0.010
Anion—anion
2sp—2sp 0.0003 0.0004 0.0006 0.0006
2p,—2D, 0.007 0.008 0.009 0.009
VEC 8 9 9 10

calculate from the model, using orbital equations and measured lattice
parameters, which of the anion orbitals will overlap with cation orbitals to
produce molecular orbitals of useful energy. The degree of spatial overlap of
the orbitals is indicated by the Slater Overlap Integral, some values of which
are given in Table 6.5 for some carbides and nitrides. These show that
metal—metal overlaps are important although neglected in this model.
Emphasis is given to the orbitals within the complex that produce strong

Figure 6.7 Regular octahedron of 6 anions around the central cation with reference axes for sp
and p, anion orbitals. Numbers refer to anions | to 6.

M
328 CERAMIC SCIENCE

Table 6.6 Molecular orbitals in MX hard-metals

Cation
Representation orbitals + Anion orbitals

Aig 4s [1/,/6(sp(1) + sp(2) + sp(3) + sp(4) + sp(5)+ sp(6))]

E, ie [1/2,/3(2sp(5) + 2sp(6) — sp(1) —sp(2) —sp(3)—sp(4))]
3dyx_y» —-$(sp(L)+ sp(2) — sp(3) — sp(4))
Thu 4p, [1/./2(sp(1)—sp(2)) and 3(m,(3) + <(4)+25(5) +74(6))
4p, [1/./2(sp(3)—sp(4)) and 4(m,(1)+7,(2)+7,(5)+ 2,(6))]
4p, [1/./2(sp(5)—sp(6)) and 3(m,(1)+7,(2)+7,(3) + 2.(4))]
Tog ade 3(x,(1)—72,(2) + 2,(3) —2,(4))
Say 4(m,(1)—1,(2) +7,(5)—7,.(6))
3d), 3(1,(3) —7,(4)+ 2,(5) —7,(6))
Tou none = 4(,(3) + 2,(4)— (5)—,(6))
3(n,(1)+7,(2)—7,(5) —2,(6))
3(1,(1)+ 2,(2) —1,(3)—7,(4))
Wig none 4(n,(1)—2,(2) —2,(3) + 2,(4))
3(n,(1) — 2,(2) —7x(5) + 2x(6))
3(2,(3) —2,(4) —7,(5) +7,(6))

overlap and good covalent bonding; these are the first five combinations in
Table 6.5 and in total are approximately twice as important as the metal-metal
overlap, which emphasizes the covalent part of the bonding forces.
The next stage is to construct a molecular orbital energy ladder in the
manner of (2.2.7) by considering the octahedral arrangement around the
cation and the directional and symmetry properties of the orbitals on the six
anions. Figure 6.7 assists in this process.
From the direction and symmetry of the metal orbitals and the geometry
in Fig. 6.7, the molecular orbitals in Table 6.6 arise. They are identified by a

Aig

4p _tiy

4s ig

3d _°g ag
Bou Eu
tay
——t

2+ , 2-
M orbitals Molecular orbitals X orbitals from 6 anions
2+ 2-
M Xe

Figure 6.8 Molecular orbital energy level diagram for o and x cation-anion bonding in MX
hard-metals.
 SPECIAL CERAMICS 329

representation (2.2.5) in the table so that they can be identified in the

molecular orbital energy diagram assembled from the overlaps and sketched
in Fig. 6.8. The molecular orbitals in Table 6.6 give rise to twelve bonding
levels (and therefore twelve antibonding levels) and six non-bonding levels;
only the bonding and non-bonding levels are shown in Fig. 6.8.
Armed with the energy diagram of Fig. 6.8, it is possible to correlate some
of the behaviour patterns already described for these materials.
First consider a material with VEC = 8, namely TiC. On the model, the
Ti2* has two valency electrons and each C*~ has six electrons, making a
total of thirty-eight electrons in all. These electrons can be accommodated in
the diagram so that thirty-six are in the bonding and non-bonding levels and
two must be assigned to the lowest anti-bonding level, T3,. Clearly greater
stability will be achieved if there were no electrons to put into the anti-
bonding level, which may account for the failure to achieve a composition
TiC,>. Maximum compositions around TiC), have been reported—
because of the anion vacancies therein electrons need not enter the anti-
bonding levels as long as the energy for vacancy formation is not too great.
A balance of the energy needed to form vacancies and some energy level
overlap to form bands from these isolated octahedra make it possible to just
fill to the top of the non-bonding level in TiC,, at the noted upper composition
limit of 0.97. In the absence of the suggested band overlap, TiC 97 would be
a semiconductor instead of the good metallic conductor it is, whilst TiN with
2 +42 electrons to accommodate has the electrons across the energy gap and
in the antibonding levels.
Materials with higher VEC’s, formed by metals with more d electrons or
having nitrogen or oxygen in the lattice, will inevitably have electrons in the
anti-bonding levels after first exhibiting a greater range of non-stoichiometry
as the system attempts to minimize the anti-bonding level occupation. The
general predictions from this are (as reported) lower non-metal/metal ratios
at the limiting compositions, lower melting points and decreased hardness for
nitrides compared to carbides and on moving across the periodic table from
group IV.
Compositions with VEC’s such that the bonding and non-bonding levels
are just filled will be more resistant to reaction with nitrogen or oxygen, since
all extra electrons to the system must be accommodated in the anti-bonding
levels.
It has been shown already that TiCy.,-TiCo,97 has a linear increase in
hardness as the carbon content rises and VEC changes from 6.8-7.8 at
which concentration the bonding levels and non-bonding levels are filled.
Conversely, TaCy g—1aCo,96 Shows a steady decrease in hardness, because at
TaC,., the VEC is already 7.8 and extra electrons from carbon are consigned
to the anti-bonding levels. Similarly, it is noteworthy that Tap gHfy .Co.93 is
the most refractory material known and hasa total of 36.3 electrons per
M?2*+X21, unit, which just fills the levels in Fig. 6.8. It is interesting
330 CERAMIC SCIENCE

correlations such as these that make this model, crude though it may be, very
useful in a semiquantitative way.
Monocarbides in Class 2, like UC, Gd,C and Sc,,C,5, can be accom-
modated in the discussion so far by invoking the artificial concept of f— d
electron transition.’®
The higher carbides, La,C3, UC,, etc., present a special case that uses the
methods of (2.2.7). X-ray and neutron diffraction experiments, chemical
attack by protonic solvents, and the shape and size of the unit cell all establish
the existence of C, anionic units? that are generally described as C3”. Since
these carbides have pronounced metallic properties, the bonding cannot be
wholly ionic, rather the metal is present as M"* where n > 2 and the extra
electrons are delocalized in a band involving the metal orbitals and the C37
anion. The molecular orbital energy level diagram for C, is derived and
shown in 2.2.7, from which it is clear that two electrons can be accommodated
in the bonding 2p, level, making the unit more strongly bonded. This leaves
all the C3~ anion anti-bonding orbitals available for further molecular
orbital formation with suitable metal orbitals. For Y and Sc, the metal
orbitals involved are 5s and 4d, for the lanthanides 6s and 5d, and for the
actinides 7s and 6d. In the case of lanthanum it is suggested that
La(Xe) 5d'6s? +C, > La** (Xe) +C3> +1le™
and that the le~ per lanthanum occupies a level in an orbital composed of
6s, 5d (metal) +2p* (non-metal). Thus there will appear to be some electron
density in the anti-bonding levels of the C, molecular orbital energy diagram
(2.2.7) which will have the effect of weakening the bonding and increasing the
C-C separation. In the UC, case, the ionization to U** places extra electrons
in the p* level. In this way the increased C-C distance (0.134nm compared
to 0.130nm in LaC,) is expected, and the short C-C bond at 0.119nm in
CaC, is understandable. Increased electron occupation of the s, d—pz levels
implies a stronger covalent character in the M-—C bond which helps to
rationalize the increased stability on moving from CaC, to LaC, to UC.

6.2 Borides
It has already been stated that a separate classification method must be
employed because of the larger size of boron compared to carbon and
nitrogen. Because the non-metal atom size depends upon the coordination
number and the metal into which they are put, only a comparison of sizes can
be given, but whereas C ~ 0.077 nm and N ~ 0.071 nm, boron is 0.091 nm.
This does mean that undistorted octahedral site occupation is very limited
and the B—B bonds of various types are present in many borides.
Mixed bonding is still encountered”? because these materials have all the
characteristics of covalent and metallic materials. Generally speaking, borides
 SPECIAL CERAMICS Soil

Table 6.7 Metal boride phases

Metal M;B M3;B M,B, M,B M3B, MB M3B, MB, M,B, MB, MB, MB,,

ate
Na

Ca
’ e
Sr
Ba
Al
Rare
earths
Actinides SoG=
Ti
ZI a
H;
V
Nb
Ta
Cr
Mo
W
Mn
Te
Re oo —— pee
Fe
Ru
Os ee
eee CO
Ni
Pd oes
PE OS
ea
eeeSe
Se
ee,
een
Ag
Au y

are very stable, more so than carbides and nitrides. All are refractory, mostly
chemically inert and several have unusual and useful physical properties.
A great range of structures and compositions are observed within the limits
M,;B-MB,, and occasionally MB. Several metals form as many as SIX OF
seven distinct boride phases; some idea of the commonness of borides and
their distribution relative to the periodic table can be gained from Table 6.7.
The following points emerge from Table 6.7.
(a) Aclassification by periodic element is not very meaningful.
(b) No one metal will form all types of boride.
(c) Only MB,-is common to the three main periodic groupings, “A” elements, d-transition
metals, and f-transition metals.
(d) In general main group-A metals and f-transition elements form higher borides, whilst d-
transition metals form lower borides. There are some obvious exceptions to this.
(e) There appears to be a disregard for normal valency laws.

At first sight Table 6.7 may present a completely haphazard arrangement,

but it can be systematized by the degree of interconnection of boron atoms as
permitted by the number and size of the accompanying metal atoms. In
332 CERAMIC SCIENCE

Table 6.8 Metal borides

Atomic B-B Boron

Unit ratio Examples (nm) units

Isolated B atom M,B Mn,B <0.21

M3B Co;3B,Ni;B <0.21 ite cen oe
M,B Be,B 0.330
Pairs of B atoms M,B, V3B, 0.179 I
Single chains MB FeB 0.177 eA AA
Double chains M,B, Ta,B, — Soke
Cr3B,
Layer networks MB, TiB, 0.175 meee
YB, 0.190
Three-dimensional MB, UB, 0.170 a
frameworks MB, CaB, 0.174 B, octahedron
MB,, ZrB,, 0.175 24B atoms in a cubo-
octahedral cage

metal-rich borides, the borons are often so diluted that they are too far apart
to interconnect and are then considered as isolated atoms. As the boron
concentration increases or the size of metal atom decreases, the borons unite
to form pairs, then single zig-zag chains, then double chains, then two-
dimensional networks, and finally three-dimensional frameworks. It is worth
comparing this to the development of silicate structures from isolated SiO,
tetrahedra (1.5). The main stations of this structural development are
summed up in terms of composition in Table 6.8.
When the metal structure has to accommodate isolated or paired boron
atoms, then interstitial occupation of octahedral holes is not a viable
hypothesis because of the size problem. When these structures are considered
in terms of B atom occupation at the centre of trigonal prisms of metal atoms,
the structures referred to in Table 6.8 can be divided into two classes for
further consideration: trigonal prismatic borides and rigid lattice borides.

6.2.1 Structures
(a) Trigonal prismatic borides (MB > MB,). Borides at the metal-rich end of
the composition range have structures which are determined essentially by
the metal lattice. The metal atoms usually have, or can be approximated to,
a hexagonal arrangement built up in several ways from arrangements of more
or less distorted trigonal prisms, some of which contain boron atoms.
(i) M,B. Here the structure is closely related to cementite (Fe,C) and can
be viewed as metal prisms like the one shown in Fig. 6.9, which are arranged
so that there is a metal atom near to the centre of each rectangular face. Each
of the metal atoms beyond the rectangular faces is part of another boron-
filled trigonal prism. By this offsetting of the metal prisms, each boron atom
has a nine-fold metal coordination, three in a planar triangle above, three
around it at the apices of a triangle, and three below in triangular array.
 SPECIAL CERAMICS 333

Figure 6.9 (a) The basic trigonal prism of metal atoms forming the building unit of metal-rich
borides. Boron, the large atom, occupies a central position in this unit. (b) Projection of the M3B
structural unit: metal atoms 3,3’,3” are coplanar with the boron.

Furthermore, each boron atom is effectively isolated in distance and in terms

of orbital angles from all the others.
(ii) M,B,—M,B. In the M;B structure, not all the trigonal prisms can be
filled with boron if the composition is to be achieved. These compositions
represent an increased trigonal prism occupation by boron as the composition
changes, whilst maintaining the basic essentials of the MB structure.
(iii) M3B,. The metal atoms still have a quasi-hexagonal packing but the
increased boron content demands that some metals are replaced by boron.
Replacement of a metal atom outside the trigonal prism faces, atom 3in Fig.
6.10b, by a boron brings two borons on the same plane at an angle favouring
sp? overlap and resultant strong bonding to contribute to the overall
stability. As Fig. 6.10a shows, this arrangement leads to pairs of boron atoms
in the structure. In fact, the change from M,B to M3B, can be thought of as
a change from an arrangement where each boron-filled trigonal prism was
fused at the three rectangular faces to metal-filled trigonal prisms, to an
arrangement where one in three of the face-fused prisms is centred by a
boron atom and not a metal atom.
(iv) MB. In the prismatic metal structure, more of the metal-filled prisms
are replaced by boron-filled prisms, so that long zig-zag chains of boron
atoms can be identified in the structure. In other words, around the
individual units in Fig. 6.10b, a second metal atom is replaced by a boron
atom. Occasionally some cross-linking of the chains can occur in some
systems at compositions which cannot lead to full double chain structures;
Ru, ,B, is an example.
334 CERAMIC SCIENCE

PoOK,
es Jans

— 7 are ve!

Figure 6.10 (a) Projected view of two face-sharing boron-filled trigonal prisms present in the
MB, structure. (b) Further replacement of metal atoms at prism faces to give chains of B-B
atoms in MB ceramics.

(v) M3B,. In the extended structure discerned as face-sharing trigonal

prisms in the hexagonal packing arrangement, as more metals at the prism
centres are replaced by boron, a cross-linked double chain structure emerges
at the M,B, composition as shown in Fig. 6.11a.
(vi) MB,. The limit is reached when the extra boron added replaces the
third metal atom at the prism face, and an extended 2-D net of boron atoms
is produced as shown in Fig. 6.11b. The diborides are strictly hexagonal,
comprising planar networks of alternate metal and boron atoms. They have
the ultimate structure obtainable by the simple process we have described
(condensation of trigonal prisms) and are also the first class of boride in
which an extended boron lattice occurs. Together with M,B,, in which
alternate boron layers in the MB, structure are puckered rather than planar
and some of the prism metal atoms are replaced by boron, these structures
form a link between the two structural classes of borides.
(b) Rigid lattice borides (MB,—MB, 5). Structure and properties are dominated
by the rigid covalent boron framework. Simple units can be discerned and the
structures are in the main formed from cubic packing of these units with the
metal atoms forming a second, interpenetrating cubic lattice.
MB, as the link has already been described.
 SPECIAL CERAMICS 339

KEXEXE ’

ere
S

‘Ooo
%LGweye\YY

Figure 6.11 (a) Projection of M3B, structure showing cross-linked boron double chains as metal
in 2 of each 3 fused trigonal prisms is replaced by boron. (b) Hexagonal network of planar borons
developed in MB, when all trigonal prisms are centred by boron atoms.

(i) MB,. Derived from the MB, planar structure, the view of structural
evolution with boron addition continues. The necessary distortions in the
boron network arise by considering the B, octahedron of Fig. 6.12a, and
imagining these to be periodically inserted into the hexagonal network boron
layer of MB, so that the square plane of the octahedron, atoms 1, 2, 3, 4, is in
a boron plane of MB,. The result of this is that hexagonal rings become
heptagons, and the boron layer consists of 4- and 7-membered rings as shown
in Fig. 6.12b. Octagon apex borons, atoms 5 and 6 in Fig. 6.12a, are part of
the boron layers above and below, and so form the 3-D covalent linked
structure. The inserted octahedra intrude into the hexagonal metal layers and
move them into nearly cubic positions so that they are above and below the
centres of heptagons in neighbouring layers.
(ii) MB. All the boron atoms are contained in B, octahedra, the bonding
of which is discussed below. Each B, unit is centred on a Cl” site in the bec
structure of CsCl. The metal atoms are found at the Cs* sites. The octahedra
are linked together at all six boron atoms of the octahedra to give a rigid but
open framework, able to maintain the phase even when up to one-quarter of
the metal atoms are removed by volatilization.
336 CERAMIC SCIENCE

(a) (b)

Figure 6.12 (a) Bs octahedron with atoms numbered as referred to in the text. (b) The boron
layer in MB, formed by inserting B, octahedra into the hexagonal boron layers of the MB,
structure; atoms 1, 2, 3, 4 refer to the planar atoms in the octahedron of (a).

Since every metal atom is coordinated by eight of the B, units, in effect

each metal is coordinated by twenty-four boron atoms from a triangular
octahedral face of each of the eight units, and so the phase is restricted to
large metal atoms.
(iii) MB,,. The boron unit in this structure has already been described in
2.4.1, where the structure of boron itself was discussed. As above, the B,,
units and metal atoms are located on Cl” and Cs” sites in the CsCl structure.

6.2.2 Non-stoichiometry
From the discussion above it is clear that there will be considerable scope for
compositional variation; this is found to depend on the structure.
(a) Trigonal prismatic borides M,B > M,B. In principle, from our view of
increased filling of metal trigonal prism centres by boron and replacement of
metal atoms in the centres of neighbouring prisms by boron, complete
variation is possible. Such variation and ease of boron replacement with
other atoms leads to restricted uses of these materials. In the case of the
monoborides, non-stoichiometry still arises from variable boron content, e.g.
MoBo 95-1.062 WBo.92-1.02: Lhis is achieved by adding or subtracting atoms
to or from the chains, and non-stoichiometry by this mechanism suggests that
impurity pick-up is less likely because few elements will form covalently-
bonded chains.
The diborides can exist, with substantial vacancies in either metal or boron
layers, but evidence to date all suggests that the composition ranges, e.g.
TaB, 73-TaB, 57’ arise from boron lattice vacancies or extra borons included
at the centres of hexagonal rings, and not from vacancies, other than the few
expected thermodynamically, in the metal sub-lattice.
 SPECIAL CERAMICS 337

(b) Framework borides. Because of the strong covalent bonding between

boron atoms in the higher borides, non-stoichiometry invariably arises from
metal sub-lattice vacancies. Already it has been stated that LaB, can tolerate
metal deficiency until a composition Lag g,B, is arrived at. Furthermore, it
should be noted that metal vacancies in such an open network structure lead
to very high metal atom mobility and confers some useful and interesting
properties (6.2.5).
In summary, we can write:
M intact metal lattice boron lattices intact eb ay
4B boron vacancies MB, metal lattice vacancies MB, MB. MB, 2

deficiency in both
sub-lattices but boron
network mostly affected

6.2.3 Bonding and valency

Much interest has centred on whether boron atoms gain or lose electrons in
these special ceramics. It has frequently been suggested that electron transfer
must be in the same direction for all types of boride because of the
electronegativity differences between boron and metal atoms. Application of
the simple methods outlined in (2.2) show, however, that a variety of
situations can be expected. Nevertheless some generalizations can be made:
in the metal-rich prismatic borides, electrons are transferred from boron to
metal, whereas in the boron-rich rigid-lattice borides, electrons are transferred
from the metal atoms to the boron lattice.
(a) Trigonal prismatic borides. Some weight has been given to the facts that
CuB does not exist, that the maximum B content decreases as the atomic
number of the metallic elements increases, and that for non-transition
elements lower borides do not exist, to suggest that electrons are transferred
from boron to metal atoms.2? However, more direct evidence for this is
provided from magnetic measurements on CrB, MnB, CoB, NiB, Fe,B and
Co,B. The magnetic behaviour can be interpreted in terms of three energy
bands: first, a dense narrow 3d band in which electrons are virtually localized
on the metal atoms; second, a broad shallow 3s band which gives rise to the
conduction properties; and thirdly, a 2s2p bonding band between B atoms in
the non-metal chains. Distribution of electrons between these bands depends
on their relative energies and the energy of the Fermi level. Variations in the
Bohr magneton number for the ferromagnetic borides is linear, with total
electron concentration suggesting that one spin band is full and that the Bohr
magneton number is the number of vacancies in the sub-band. This situation
is illustrated in Fig. 6.13.
From the magnetic measurements the number of 3d electrons per metal
atom can be estimated (Table 6.9) from which it can be seen that, with the
exception of nickel, this number exceeds the total number of valency electrons
per metal atom and supports the view of electron transfer from boron to
metal.
338 CERAMIC SCIENCE

Fermi levels.

MnB CoB

sp E

Figure 6.13 Postulated energy bands in magnetic borides.

Mossbauer measurements on FeB, Fe,B and iron metal are consistent with
the electron gain at metal sites. Hence the bonding consists of M—M metallic
bonds, B-B covalent bonds, and some B—M bonds produced when excess
electrons are transferred from boron to metal. This strengthens the metallic
bonding and the metallic nature, whilst the presence of BM and B-B bonds
produces the brittle behaviour because of the directed nature of such bonds.
The overall model is similar to the Dempsey model for hard-metals (6.1.5).
(b) Boron lattice borides —MB,, MB,, and MB,. Molecular orbital energy
level diagrams, (2.2.7), have been calculated for an isolated B, octahedron
using atomic 2s and 2p orbitals on each boron.”? It is seen that the molecular
orbitals can be divided into two sets: those with large amplitudes external to
the octahedron with energy levels approximately in a non-bonding range
(there are six such orbitals) and a second set of eighteen molecular orbitals
which all have large amplitudes directed within the octahedron. On the
energy ladder, seven of the internal MO’s are bonding and eleven anti-
bonding. This situation can be sketched as:

2p =
a od
6B — Be 4444
octahedron
2s ee
att eect

outward inward
directed directed

This arrangement of molecular orbitals is available for the eighteen

electrons in the system, and it can be supposed that since the B, octahedra
are all interconnected that each of the outward directed orbitals is singly
occupied. Such single-electron occupation will allow for an electron pair
covalent bond to be formed with neighbouring octahedra in the manner of
the valency bond theory (2.2.6). There are then twelve electrons for internal
bonding to hold the B, unit as an octahedron, and clearly all these can be
 SPECIAL CERAMICS 339

Table 6.9 Number of electrons per metal atom

Compound electrons per metal __d electrons per metal Total s+d electrons per metal

MnB 8.1 5 iT
FeB 8.9 6 8
CoB 7 7 9
NiB 10.0 8 10
Fe,B 8.2 6 8
CoB 9:2 7 9

accommodated in the seven bonding molecular orbitals, still leaving two

holes in the bonding levels. To fill these and further stabilize the structure, a
metal atom can donate two electrons if the ionization energy (5.3) is not too
excessive. In this process a further contribution to bonding is now provided
from the ionic system M?*B2~, but the metallic component of the total
bonding is reduced. However, this may not decrease the metallic properties
of the system as a whole if the band gap between inward-pointing B, bonding
and antibonding orbitals is small, or band overlap occurs as in the real
situation if an array of B, octahedra within the crystal and not just a single
B, unit is considered. Indeed, when an array of B, octahedra in their cubic
environment is considered, a set of energy bands is obtained which are
considerably overlapped but in which a band gap of approximately 1.0eV
exists between the bonding and anti-bonding bands. When calculated like
this, the outward-pointing molecular orbitals form a band that isdistinctly
stabilized and so would be occupied.
Without considering the extended calculation and band system, some
interesting predictions emerge from the isolated B, model. For example, Na
is relatively easily converted to Na* and NaB, can be formed; the one
electron provided to the B, unit does not fill the bonding level, and so NaB,
is weakly metallic in nature, whilst being refractory from the ionic and
covalent contributions as well as brittle and hard. For the periodic group II
elements, ionization to M’* is relatively easy, but the two electrons will fill
the bonding levels in the B,. Thus CaB, should be refractory, hard and brittle
whilst having no metallic properties. This is almost the case, for CaB, is a
weak semiconductor due to the electrons transferred across the narrow 0.4eV
band gap.
For group III elements, Al, La, etc., it is still possible to obtain M3t
without excessive energy expenditure relative to the final bonding arrange-
ment. Now, however, one electron must be accommodated in the anti-
bonding levels, making the B, octahedron less stable and larger in size.
However, the increased size of the B, octahedra causes overlap with other B,
units and stronger covalent bonds, to maintain, with the ionic component,
very strong overall bonding. The presence of one electron in the anti-bonding
level restores easy metallic conductivity and appearance. Work functions, or
340 CERAMIC SCIENCE

the energy required to expel electrons from such a system, will be lowered.”’
For elements like Zr the energy required to obtain Zr** appears to be
prohibitive, and MB, phases for group IV, V, VI elements are not formed.
From the energy band scale, it has been shown that for MB, phases to form
there must be a limit on the first and second ionization potential of the metal
component. A suitable Born-Haber cycle (5.3) enables this to be done. The
result is that the first ionization potential <6.7eV and the second <12.0eV,
which excludes d transition elements and Be, the notable exceptions in Table
6.7 for forming this phase.
Thus this mixed ionic, covalent, metallic bond model anticipates most of
the observed facts.
A semiquantitative treatment of bonding in the dodecaborides considers the
molecular orbitals formed from an isolated B,, cubo-octahedral boron unit
in much the same way as was done above for MB,. Isolated cubo-octahedra
are considered to be joined by localized valency bonds arising from overlap
of outward-pointing molecular orbitals. Thirty-six molecular orbitals are
inwardly directed so that thirteen are bonding, three are non-bonding and
twenty are anti-bonding. Thirty-eight electrons are needed to complete all the
levels before using the anti-bonding levels and twelve borons can supply
thirty-six of these. Thus M?*(B,,)*~ is likely. MgB,, is non-conducting, all
levels below the band-gap are full, whilst ZrB,, is metallic as two electrons
per metal occupy the anti-bonding levels.
NMR measurements on diborides (MB,) suggest that the B layers do not
contain only o bonds but a z-bonding component. This could be achieved by
transfer of electrons from the hexagonal metal layers to give M?*(B™ ),, when
the boron layer would become isoelectronic with graphite and be bonded in
the same way (2.5.1). Certainly YB, has a conductivity consistent with one
free electron per metal atom, and a model Y** (B~ ), would apply.
In conclusion, it appears that although most borides possess metallic
conductivity the principal binding forces are ionic and covalent, and no
amount of alloying will cause them to lose the often undesirable property of
brittleness. .

6.2.4 Preparation
With very few exceptions, metal borides are prepared by high-temperature
reactions, and although there are seven general methods the common
manufacturing route can be summarized in the flow chart:

B,C
 SPECIAL CERAMICS 341

The starting material is boric acid and frequently a metal oxide and
carbon. The process is known as the carbothermic reduction process; high
temperatures (around 2000°C) and an inert atmosphere are required. Vari-
ations are reported when zircon (5.5) or Ti;(PO,), are used. Volatilization of
B,O, is a problem, and for this reason B,C is often used as a source of boron.
If the product can be used in an impure form, the cheap aluminothermic
process can be used:
MO+B,0,+Al— MB, +Al,03.
Chromium borides are made this way since they can stand the alkali leaching
needed to remove the alumina.
Tofacilitate the electrolysis route, the metal oxide must be dissolved in a
molten salt which is either a halide or a fluoroborate salt. The operation is at
700-—1000°C in a mild steel container maintained as a cathode. Graphite is the
usual anode material. Higher borides such as CaB, are made this way, and
frequently the product is a mixture of boride phases and elemental boron.
Very pure samples or vapour plated deposits are made by simultaneous
decomposition of a volatile boron compound and a volatile metal compound
on a hot substrate in the presence of hydrogen:*
ZrCl, + 2BBr, + 5H, > ZrB, + 4HCI+ 6HBr.

6.2.5 Properties
The characteristic boride properties of very high melting points, high
hardness and electrical conductivity are quite comparable to those of the
carbides and nitrides, but their improved inertness to non-oxidizing reagents
and resistance to oxidation at red-heat and above shows advantages over the
carbides and nitrides. Chemical inertness increases with boron content, but
comparisons are subject to preparative route—different methods produce a

Table 6.10 Representative physical properties for some boride

phases

M. pt. Density Resistivity microhardness

Boride (°C) (kgm~?) (ohm. cm.) (kg mm~*)

CaB, 2230 2400 = 1x 108 2750

LaB, 2230 4700 15 2770
TiB, 2980 4500 9-15 3350
ZB, 3040 6100 7-10 2250
Z1B,> 2680 3600 60-80 re
TaB, 3100 12600 14-68 2500
CrB, 1950 5200 21-56 1800
MoB, 2250 7100 20-40 1200
Mo,B, fe 7500 18-45 2350
Fe,B 1390 a ee
UB, 2100 9400 2 a
342 CERAMIC SCIENCE .

variety of microstructures and purities, both of which critically affect the

chemical and physical properties. Some typical physical property values are
summarized in Table 6.10.
(a) Thermal properties. Melting points of all except the iron borides are high.
The most stable appear to be group IV elements. It can be noted that where
the structure is dominated by the boron lattice, e.g. CaB,, LaB,, the melting
points are identical. Thermal expansion values are of a ceramic rather than
metallic order, around 5—8 x 107° K~'. They are usually anisotropic in this
property. Thermal stabilities are generally good up to the melting points, but
there are some exceptions, notably:
TiB > TiB, + Ti,B at 2060°C
ZrB — ZrB, + Zr below 800° and above 1250°C
B,Si > B,Si+ Si above 1370°C
LaB, — La, _,.B, + La above 1400°C.
(b) Electrical and magnetic properties. In the table above, CaB, is notable for
its high resistivity, but the ionic model already discussed for this boride
anticipates this behaviour as it does the sharp drop to metal-like values at
LaB,. For comparison purposes Cu, Al and Fe have values of 1.7, 2.8 and
10.0 microohm cm. Thus TiB, has a comparable conductivity to that of iron
metal.
Hexaborides have low work functions (2.2eV for YB,, for example) and
LaB, has the highest electron emissivity of any substance, a fact that finds
use in electronic applications?’ and arises from the model discussed in 6.2.3.
All are mainly weak paramagnetic substances except ZrB,, VB and CrB,
which are strongly paramagnetic along with the lanthanide hexaborides, and
Fe,B, FeB and MnB, which are ferromagnetic.
(c) Chemical properties. Diborides, and in particular TiB, and ZrB,, can be
heated in air to 1200—1400°C without excessive oxidation. The controlling
factor is the surface layer of oxide, formed initially at low temperatures, which
prevents further oxidation, unless mechanically removed, up to temperatures
at which B,O, volatility becomes appreciable. In this respect MoB, is less
satisfactory, because MoO, is volatile at low temperatures. A feature of
borides not encountered in the carbides and nitrides is the possibility that the
metal oxide skin may be a glass former with B,O;, and hence able to assist
in further oxidation inhibition. Borides of those metals whose oxides are
good adherence promoters in enamelling (3.5) are notably the most easily
oxidized, whilst elements that form stable borates, such as chromium, are the
most resistant to oxidation. Thus CoB, NiB and FeB can only be used up to
700°C whilst CrB, can withstand 1700°C.
Glass-forming ability clearly assists in the process of protection, because all
borides are attacked by sulphur.
Resistance to nitrogen increases with increasing boron content but, since it
is generally true that borides are more stable than nitrides, all borides are
 SPECIAL CERAMICS 343

stable to high temperatures in the presence of nitrogen. For example, CaB,

can be held indefinitely at 2000°C in a nitrogen atmosphere.
In the case of molten salts and metals, borides are resistant to attack by
metals which themselves are not ready boride formers. Eutectic formation is
often the basis for attack, and as a result TiB,, ZrB, and CrB, are all soluble
in molten titanium, but are very resistant to molten Al, Cu, Mg, Sn, Bi, Zn
and Pb. This is true even though they are often wetted by them, as in the case
of aluminium in TiB, crucibles.
All borides are resistant to non-oxidizing salts, such as molten fluorides
and metallurgical slags, but do react with oxidizing melts like hydroxides,
nitrates, carbonates and peroxides. Reaction is always accompanied by
complete disruption of the boron lattice.

6.2.6 Uses
The increased availability and lower cost of borides makes them the most
widely used amongst the special ceramics. Commercially the most important
are CrB, CrB,, TiB, and ZrB,. Applications are naturally based on the
properties outlined above. Because of their high melting points, high oxida-
tion resistance and high strengths, the diborides are used as (or as coatings
on) turbine blades, combustion chamber liners, rocket nozzles, and rocket
nose cones; all applications where heat and abrasion have to be resisted.
Composites (2.13) of TiB,+BN are increasingly used to improve electrical
resistance and facilitate electrical heating when necessary.
An area of growing importance is in pump parts able to resist molten zinc,
copper, aluminium and magnesium at 600°C. Frequently such parts are made
from CrB, or CrB,—Fe composites. TiB,, because of its good electrical
resistivity and resistance to molten aluminium, is being used as cathode
material in processing plants, and for sliding electrical contacts where high
hardness helps to resist wear. However, the full potential of the high strengths
of borides has not yet been exploited, as has been done with carbides by
introducing a degree of toughness when sintered hard-metal-cobalt com-
posites are made (6.1.4). The main reason for this lies in the failure to find a
matrix metal that does not form a boride itself or become part of a ternary
boride.
Hexaborides, and in particular LaB,, are now used as electron sources for
electron and scanning electron microscopes. However, a more common use
for hexaborides and a variety of borides with varying compositions is as
additives in both the ferrous and non-ferrous metal industries. Some examples
are the addition of CaB, +La,O; for grain control in steel, preparation of a
50:50 composite of Al+TiB, as a high-strength low expansion alloy,
addition of up to 1.5 wt. % TiB to titanium which improves Young’s modulus
by 18% without worsening the ductile properties of the metal, and CaB,
which when added to copper increases its conductivity.
344 CERAMIC SCIENCE

6.3 Silicides
Carbon and silicon have striking structural similarities in elemental form
(2.5.1), (2.6.1), but because of the large size of silicon, the decreased electro-
negativity, and the availability in silicon of 3d orbitals for bonding purposes,
there is no similarity in the ceramics they form.”° Silicon behaviour and types
of silicide can be quite easily inferred from all the discussions in (6.2) on
borides. The most notable difference is that silicides where the M: Si ratio is
greater than 1 : 2 are rare, being limited to OsSi and IrSi;. With this difference
noted, it is still true to say that metal silicides are, like borides, characterized
by their wide occurrence and the large variety of compositions they exhibit.
Stable silicides are formed by the same elements as those that form borides,
except that while a copper silicide exists, there is no aluminium silicide.
With increasing silicon content, silicides are formed containing the follow-
ing features.
(i) Isolated Si atoms bonded directly to metal atoms in Cr,Si, U,Si and in the monosilicides
CrSi, FeSi, CoSi (this last class is in contrast to the boride classification).
(ii) Pairs of Si atoms*° that remain isolated, as in U,Si, and FeSi,. Again the ratio of metal:
non-metal that contains this feature is not so precise as for the borides.
(iii) Zig-zag chains of Si atoms. Co,Si, Mn,Si,, MosSi, are examples demonstrating that this
structural feature occurs over a range of atom/silicon ratios.’
(iv) Two-dimensional close-packed layers of silicon and metal atoms. Most compositions MSi,
have this structure —MoSi,, NbSi,, TiSi, are good examples.
(v) Three-dimensional framework of silicon atoms. This feature is found in ThSi, and
lanthanide disilicides.

This pattern of behaviour broadly resembles the information given in Table

6.7, and much of the discussion following that table regarding structures can
be applied here.
The general properties outlined for borides are present in the silicides,
because the chemical bonding present. is closely related to the models
discussed so far in this chapter. As a result, most of the applications for these
special ceramics are in the field of thermal protection. Oxidation resistance is
generally superior to the borides and lasts longer, and ZrSi, can be used at
1100°C, TaSi, at 1400°C, CrSi, 1550°C and MoSi,, WSi, above 1700°C for
many hours in the air. The mechanism of protection is as seen in the case of
MoOSi,, arising from the glassy film of SiO, that is left on the surface after
MoO; has volatilized. With time, such a film becomes more viscous and
adherent to the underlying MoSi,. It is worth noting that such protection can
be lost if the oxygen pressure is too low, as the calculation below will show.
This mechanism is clearly not universal, for many other well-protected
silicides do not have volatile metal oxides. Protection must then result from
a combination of factors, one of the most important of which appears to be
the ability of the metal oxide to form a glassy phase with silica. Such a useful
property in the case of MoSi,, allied to its metallic electrical conductivity, has
led to its development as a good high-temperature heating element. The
development has followed along composite lines: SiO, and Al,O, have been
 SPECIAL CERAMICS 345
added to a MoSi, matrix to increase the electrical resistance and the heating
effect, and give stronger, more adhesive surface glassy layers.
A final application of the predictive principles set out in 5.2 to the use of
MoSi, does show that such furnace windings must be used in the air, not
vacuum, to obtain protection from oxidation! First, a series of equations
expressing the processes involved must be written:
5/7MoSi, +O, = SiO.)+ 1/7Mo. Si, (6.13)
10/7MoSi, +O, = 2SiOg@)+2/7Mo,Si; (6.14)
5/7MOSi, + SiO, = 2SiO,) +1/7Mo,
Si (6.15)
2SiOq) +O, = 2Si0,. (6.16)
Equation (6.13) is the normal reaction in ambient conditions and equations
(6.14), (6.15) express two ways in which the SiO, coating can be lost when the
Po, drops too low.
When (6.14) is favoured, then AG¢6
14) —AGi6.13) = AGie.15) < 0. The con-
dition for (6.15) to occur, and the oxidation behaviour of SiO as expressed in
(6.16), define the po, below which the oxidation process for MoSi, due to
(6.13) will stop.
The AG values for equation (6.15), expressed in half quantities as
5/14MoSi, + 3810, = SiO) + 1/14Mo,Si,
become a convenient indicator of reactivity through equation (5.4), since
K, = Dsio for the above equation. When AG 6.15) is put equal to zero, then
the thermodynamic driving forces for (6.13) and (6.14) are equal and the psio
can be found; this is given below for three temperatures.

Tox log, 9K,

1600 — 3.406
1800 — 2.083
2000 — 1.033

Now since pgio is related to po, through (6.16), use can be made of this.
The critical values of oxygen partial pressure below which (6.13) cannot
occur and only (6.14) occurs can be estimated for the three temperatures in
the table to be 0.15, 3.14 and 35mm of Hg for 1600, 1800 and 2000°K
respectively. Thus if the temperature exceeds 2000°K at an oxygen pressure
of 35mm Hg, or the oxygen pressure does not reach 35mm Hg at 2000°K,
then SiO evaporation would be appreciable and protection by SiO, would
not be achieved.

References and further reading

1. Agcour, L., E. Fitzer, E. Ignotowitz and M. Sahebkar. Chemical vapour deposition of pyro-
carbon, SiC, TiC, TiN, Si and Ta on different types of carbon fibre. Carbon, 12, 358
(1974).
346 CERAMIC SCIENCE

_ Anderson, J. S., N. J. Clark and I. J. McColm. The hydrolysis of rare earth carbides. J.
Inorg. Nucl. Chem., 30, 105 (1968).
. Anderson, J. S., N. J.Clark and I. J. McColm. On the existence of cerium monocarbide:
cerium nitrite carbides. J. Inorg Nucl. Chem., 31, 1621 (1969).
_ Besmann, T. M. and K. E. Spear. Morphology of chemical vapour deposited titanium
diboride. J. Cryst. Growth., 31, 60 (1975).
_ Besmann, T. M. and K. E. Spear. Analysis of the chemical vapour deposition of TiB,.
J. Electrochem. Soc., 124, 786 (1977).
. Bowman, A. L. Crystal chemistry of refractory carbides. NBS Special Publ. 364, Solid State
Chemistry, Proc. 5th Mat. Res. Symp. (1972), p. 561.
_ Button, T. W. and I. J. McColm. Non-metal absorption by rare earth—silicon systems. Proc.
Int. Symp. on Metal Hydrogen Systems, Miami Beach, Florida, 1981, ed. T. N. Veziroglu,
p. 415.
. Chang, Y. A., L. E. Toth and Y. S. Tyan. On the elastic and thermodynamic properties of
transition metal carbides. Met. Trans. 2, 315 (1971).
. Clark, N. J., R. Mountford and I. J. McColm. The coexistence of C, and C, species in
carbides. J. Inorg. and Nucl. Chem., 34, 2729 (1972).
10. Dempsey, E. Bonding in refractory hard metal. Phil. Mag., 8, 285 (1963).
ie Denker, S. P. Cohesion in cubic refractory monocarbides, mononitrides and monoxides.
J. Less-Common Metals, 14, 1 (1968).
Ps Fujimo, T. and H. Tagawa. Non-stoichiometry in U,N3. J. Phys. Chem. Solids, 34, 1611
(1973).
13 Goldschmidt, H. J. Interstitial Alloys, Butterworth & Co., London (1967).
14. Hiraga, K. Vacancy ordering in vanadium carbides. Phil. Mag., 27, 1301 (1973).
. Honig, J. M. On the band structure of rare earth chalcogenides and pnictides. J. Solid
State Chem., 1, 19 (1969).
16. Jack, D. H. and K. H. Jack. Carbides and nitrides in steel. Mat. Sci. and Eng., 11, 1
(1973).
ie Juza, R. Nitrides of metals of the first transition series. Adv. in Inorg. and Radiochem., 9, 81
(1966).
18. Lye, R. G. A simple model for the stability of transition metal carbides. Nat. Bur. Stand.
Publ. 364, Sth Mat. Res. Symp., 567 (1972).
. Nagakur, S. Electronic structure from electron diffraction. J. Phys. Soc., Japan, 25, 488
(1968).
20. Nowotny, H., H. Boller and G. Zwilling. Carbides and silicides. NBS Publication 364, Proc.
Symp. Mat. Res., 487 (1972).
Ae McColm, I. J. and C. Dimbylow. Wetting of refractory titanium hard metals by sodium
disilicate glass. J. Mat. Sci., 9, 1320 (1974).
om McColm, I. J. The cubic-tetragonal transformation in metal dicarbides IV: f-orbital
participation in bonding and its effect on hardness and transition temperatures. J.
Less-Comm. Metals, 78, 287 (1981). :
Pip Perkins, P. G. and A. V. J. Sweeney. An investigation of the electronic band structures o
NaB,, KB,, TiB, and CrB. J. Less-Comm. Metals, 47, 165 (1976).
24. Ramgvist, L. Variation of hardness, resistivity and lattice parameter with carbon content of
5b metal carbides. Jernkontorets Annal., 152, 465 (1968).
25. Ramgvist, L. Electronic structures of cubic refractory carbides. J. Appl. Phys., 42, 2113
(1971).
26. Rogl, P., H. Nowotny and F. Benesovsky. Neue K-Boride und verwandte Phasen (Re3B-
Typ, aufgefullt). Monatsh. fiir Chem., 104, 182 (1973).
Dis Shimizu, R., Y. Kataoka, T. Tanaka and S. Kawai. Application of LaB, in electron micro-
scopy. Japan J. Appl. Phys., 14, 1089 (1975); Appl. Phys. Letters, 27, 113 (1975).
28. Voitovich, R. F. and E. I. Golovko. Oxidation of titanium nitride. Russian J. Phys. Chem.,
54, 817 (1980).
29: Warren, R. Effects of the carbide composition on the micro-structure of cemented binary
carbides. Plansee Ber. f. Pulumet., 20, 299 (1972).
30. Yozhihiko, S. and M. Uda. Structure of USi,. J. Solid State Chem., 18, 63 (1976).
. Yvon, K. and E. Parthé. On the crystal chemistry of close packed transition metal carbides.
Acta Cryst., B26, 149 (1970).
Index

abortive bond rupture 170 andalusite 11,236

acceptor levels 96 andradite 10
acid phosphate 60 Andrieux—Weiss process 313
acidic oxides 237 angular wave functions 74-6
actinolite 23 anioniccarbon 86
activationenergy 4 model of bonding 325
for creep in glass-ceramics 232 anions, in glass 149
for viscous flow 141, 155 production of 244
of crystal growth 154 anisotropy 27, 38, 92—4, 102, 109
of processes 42 anorthite 8,39
adherence promoting oxides 190 anorthoclase 39
akermanite 13,121 anthophyllite 23
albite 8, 39, 141 antibonding orbitals 84-7, 329, 339
alite 45 antigorites 29
alkali polymetaphosphate 59 apatite 121
alkoxide preparation of ZrO, 278, 281 apophyllite 25
almandite 10 arsenic for bubble removal 148
alumina 257 ff, 88, 114-15, 120, 235-7 asbestos 22-5, 67
aluminium composite strength 129 ASTM, classification of mullite 17
effect on glass durability 195 method E399 for Kjc 167
modulus 133 athermal martensite 276
from kaolinite 30, 54 attapulgite 33
hot strength 58 atmosphere, and porcelain colour 5-6
hypothetical defect form 114 in cement phase development 46
inglass 133 in ceramic production 54
in glass-ceramics 212, 225 in mullitization 54
Madelung constant 247 atomic orbitals 71 ff
metal seals 187 aufbau principle 76, 82
nickel composite strength 129 augite solid solutions 20
reaction welding 187 axinite 8,13
solution in Si,;N, 114
whisker strength of 129 B-alumina 268
aluminate incement 44, 46 B-sialon 114-15, 117, 120
aluminium, acid phosphate 59 baddeleyite 272-4
bond formation 77 basic oxide 237
fluoride from cryolite 251 bauxite 39
in sialon preparation 120 incement 50
sulphate as silicate precipitant 44 source of Al,O3 236, 257-8
aluminium nitride 118, 120 Bayer process 257-8
aluminosilicate 9, 16, 40-41, 46, 236 Beevers-Ross sites 268-9
aluminothermic method 341 and conductivity 271
amosite 23-5 belite 45-6
amphiboles 7-8, 19, 21-3, 31, 67 benitoite 8, 13,15
amorphous, carbon 92, 96 berlinite 59
silica in zircon 273 beryl 8, 14-15
amphoteric oxides 59, 237, 273 beryllium 77, 88, 122

347
348 INDEX

beryllium —continued boroncarbide 88, 100, 311, 346

hybridization 81 boron nitride 88, 101 ff
beryllium oxide 102, 143, 187 boroxole groups 104
inglass 133 Braun’s structure 271
radiusratio 254 bridging oxygens, angle 12, 36, 81, 145
binodal 215,219 destruction byH,O 107, 174
biotite 31 in B-eucryptite ceramics 225
boehmite 272 inB,O,; 81
in grain boundaries 262 in B,O, glass 104
Boltzmann distribution, inberyl 15
of composition in spinodal 215 inCaO-SiO, 219
ofelectrons in silicon 98 incrystobalite 37
bonding, changes in graphite todiamond 195 infaujasite 42
inAl,O, 260 in quartz 36
inamphiboles 21 in silica 145
in boric oxide 82 in silicates 81
in borides 337 in zeolites 42
in boroncarbide 100 role in thermal expansion 36
in boron icosahedra 89, 100 brittle martensite 275
inBN 101 brittle micas 28, 30
inCO 86 brucite 26-7, 29, 31-2
indiamond 91
inglass 132 cadmium oxide nucleant 227
in glassy B,O, 103-4, 149 calcium aluminate hydrate 48, 50-51
ingraphite 94 calcium boride 339-41
inkaolinite 26 calcium oxide 237
inOjZ 86 stabilized zirconia 279
inoxygen 85 calcium silicate 44-5
in pyroxenes 21 asbestos tiles 23
in serpentine 26 incements 44-51
insilicates 81 inMgO 289
phosphate 59 carbides 312-20
mina—B quartz 36 bondingin 328 ff
semi-polar 36 boron 100
bond lengths, AI-Nin sialon 117 C-C distances 330
inAl,O; 261 carbon 77, 88, 91 ff, 153
in B-spodumene 226 anions 77,315
in TiO, 296,298 electronegativity 310
Na-O 218 fibres 92, 95-6
O-O in oxygen,O} 86 fibre moduli 96, 100
Si-O insialon 117 fibre strength 67, 96, 129
Si-O in silicate 218 glassy 92,153
boracite 89 size 330
borax 89 whiskers strength 129
boric acid 103 carbon monoxide, bonding 86
to prepare borides 340 donor-acceptor 87
boric oxide 82, 88, 103-4, 149-50 carbonization 92
activation energy viscous flow 155 castables 58
effect on glass durability 195 containing Al,O; 59
modulus 133 catalytic activity, control in zeolites 42
inMgO 289 of F~ in mullitization 54
borides 330ff sitesfor 42
Born Repulsion Term 246 cations, cavity type 256
exponent, table of 248-9 coordination changes in kaolinite
-Haber cycle 250-51 decomposition 53
boron 88 ff distribution in nepheline 40
aluminium composite 89, 127 incarbides 326
coordination 104 in chain silicates 21
covalency 77 infeldspars 39
fibres 67, 89, 129 inframeworks 256
hybridization 238 in nitrides 326
inBN 101 in sheet silicates 26-7
mineralizer in Al,O, 259 production of 239
quadrivalency 150 size, table of values 225
size 330 cemented hard metals 321
 INDEX 349
cement 10,44 ff chromium boride 337
activatorsfor 50 chromium carbide 320
asbestos 23 clay 1, 8, 29-33, 51
autoclaved 50 and pottery microstructure 52
expansive 50 from albite 39
heatevolution 47 incement manufacture 44,50
highalumina 50 minerals 32, 43
notation 44 cleavage,inamphiboles 22
phosphate 58 diamond 93
Portland 9, 44ff graphite 94
pozzolanic 50 melilites 13
silicate 42,50 micas 31
slag 42,50 talc 30
ceramic, catalysts 40 three layer silicates 27
claysin 33 two layer silicates 27
cordieritein 15 close packing, in two layer silicates 27
corundumin 1 limiting compositions 481
ease offormation 1 of icosahedra in boron 90
electrically resistant 15 of icosahedra and C, chainsinB,C 100
energy use 2-3 of OH” in brucite 26
environmental stability 238 of octahedral layers 27
eucryptitein 39-40 of oxygen in metakaolinite 55
for combustion control 291 of silicate chains 21
industry 1 of SiO{~ in forsterite 57
kaolinite in 29-30 stacking variations 481
kilning atmosphere 54 cobalt boride magnetism 337
kilning energy use 3 cobalt oxide, conductivity 291
manufacture 51,54, 58 promoter 186
maturing 3 coesite 35
mechanical behaviour 261 cold sets 58
metal bonding 186 colemanite 89
nephelinein 39 composites 96, 123 ff
porosity 1 asbestos 23
products 1, 33,54 compressive stress in, glass 179
reflectance 4-5 TiO, 299
restrictions 24, 58 Al,O, 267
special classification 311 condensation, in layer silicates 26
stability calculations 18, 238 ff of SiO, from metakaolinite 53
steels 229, 282 oftetrahedra 6,19, 149
traditional 2,58 polymerization in glass 149
transmittance 4 conduction band, levelin SiC 106
ceria, electrostatic bond strength 255 levelin Si 98
in glassceramics 212 conductivity in B-Al,0; 271
radiusratio 254 Congleton—Petch equation 165
cermets 321 contact angle 193-4
chain silicates 7, 19 ff cooling rate
chalazite 256 to prevent spinodal decomposition 217
chamottes 236 prediction of glass formation 156
charge coordination, numbers 254
neutralization in amphiboles 21 polyhedra 254, 274
layer silicates 26 coprecipitation, incements 49
transfer 5, 6, 22 for mullite preparation 19
Charles—Hillig theory 174 for sialon preparation 120
chemical strengthening 179 of polysilicic acids 44
of Al,O, 266 cordierite 15
Chemico process 258 in glassceramics 227
china 1, 5-6, 29 cordsin glass 148
chiolite 251-2 Corning 9608-glass ceramic 225
chloride process 296 corrosion of glass 194 ff
chlorine 77 at crack tips 174
in TiO, production 296 Coulomb, energy 68, 149, 245
chlorites 28, 31-3 forces 174
chopped strand fibre 128 potential 268
chrysotile 23 ff stabilization 149
chromite bricks 289, 300 covalent bonding 68 ff
350 INDEX

covalent bonding —continued E-glass 104, 133

in boric oxide 104, 150 efficiency of reinforcement 126
incarbides 324 elastic, properties of composites 125
indicarbides 330 energy stored 161
inspinels 304 electrically resistant ceramics 1, 15, 227
Cox theory of glass strength 169 electron
crack, blunting 177 affinities, table of 244
extension force 166 degeneracy 75
propagation in multiphase ceramics 123-4 energy levels 75
crack velocity 108-9, 122, 124 momentum 69
and stressintensity 175 pairing 78
equation 162 probability 72
creep resistance in glass ceramics 231 electrostatic bond, energy 218
cristobalite 35-8 strength 255
in glassceramics 220 electrostatic field in zeolites 42
inSiz;N, 113 electrolytic surface roughening 189
structure in phosphates 59 embryos 209
type sialon 122 emery 257
critical cooling rate 156-8 enamels 31, 186
critical length 128 energy, bands 87, 98, 338
critical volume fraction 127 change on nucleus forming 208
crocidolite 23-4 crystal field stabilization 249
cryolite decomposition 250 demandcurve 124
crystal field stabilization energy 249, 302 free, of mullite decomposition 18, 53
crystal growth equation 154, 222 Griffith balance 123, 160-61
crystal orientation in ceramics 58 internal of glass 43
crystallization control in glass ceramics 224 levels of Cr>* in Al,O; 267
crystallite model for glass structure 147 maximum of covalent bonds 76
cubo-octahedra in zeolites 41 of graphite-diamond transition 92
ofhole bands 98
De Broglie relationship 71 of levels in He, Li 76-7
degeneracy 73,75 of orbitals 72 ff
dehydration polymerization to glass 134 totalelectronic 68
dehydroxylation, heterogeneous 57 enstatite 22,57, 157,228
homogeneous 53 ferrosilite solid solution 20
of kaolinite 52, 56 from nitrogen glasses 122
Dempsey model for hard metals 324 from serpentine 29
dendritic crystal growth 223 enthalpy 238
dendrite theory 188 of mixing 214
Denker model for bonding 326 entropy 55, 120, 238
densification aid in sintering 112 loss on cooling glass 136
devitrification, of B,0, 104 ; of mixing 209
of glass 133 value for metakaolinite transformation 55
diamagnetism of oxygen 78, 86 epidote silicates 13
diamond 91 ff equivalent representation of sialons 116
strength 67 escolaite 289
diaspor 257 ettringite 47, 50
dicalcium silicate 45 eucryptite 40, 179, 225
dichromate type silicates 12 Eyring equation 141, 146
dickite 30
diffusion, limited crystal growth 222 Failure prediction in glass 182
of Na* in glass 170 faujasite 41
Dinas refractory 236 fayalite 9,191
diopside 8, 20, 22 feldspar 1, 8, 34, 39-40
discreteionconcept 148 feldspathoid group 34, 40-41
dislocation activity in Al,O; 264-5 ferrite,incement 44
dodecahedral tunnels in SiO, 150 reaction withH, 46
dolomite 22, 30, 285 ferrites 308
donor covalency 81, 87 ferrosilite 20
donor levels in bands 98 fibre, aspect ratio 129
drying agent 44 composite modulus 125
ductile-brittle transition in TaC, TiC 318 criticallength 128
durability 132, 194-5 critical volume fraction 127
of glass ceramics 205 developmentincement 46ff
of Li,Si,0, 225 glass manufacture 139
 INDEX 351

fibre —continued glass—continued

polyacrylonitrile 92,95 fibre composite 125, 129
surface flaws 96 flame polishing 179
fibre glass, Al,O,in 259 flaw model 266
composite 133 fracture surface 163 ff
fibrous silicates 22 ff, 33, 46-7 from albite 39
fictive temperature 135 from Gossage process 42
field assisted bonding 187 hardness of 168
firebrick 29, 33,35 inhomogeneitiesin 148
fireclay 235-6, 259 liquid phase separation 213
five-fold symmetry model 150 -metal dovetails 188
flame polishing 179, 201 -metal seals 184
flaws, self generating 277 neutral 43
size and strength 262 nitrogen 122
float glass process 139 number 158
fluoride, affinity 251 phases in phosphates 59
glasses 148 S-type strength 67
mineralizer 54 soda 42,132
fluorine 77 strength theories 158
in glassceramics 228 glass ceramics 22, 35, 122, 204 ff
fluorite structure 274, 293 base composition 212
fluorophlogopite 227 crystallization controls 224
foaming 112 mean free path 230
forbidden transitions 268 residual glass phase 229
forsterite 8, 9, 56—7, 122, 157 glass strengthening methods 284—6
ceramic bonding 187 glass transition temperature 135-7, 141
fracture initiation in Al,O;, 264 glassy aluminosilicates 49-50, 122
fracture 165-6 glassycarbon 92
framework silicates 33 ff, 50 glassy matrix, of ceramics 58, 115
cationsin 256 phosphate 59
freeenergy 238 glazes 39
free volume concept 142, 153 glissile, phase boundary 276
fritforenamelling 186, 297 Goodman’s structure model 151
Fuller’s earth 33 graphite 91-7
furnace, atmosphere 5, 6, 46, 54 oxidation rate 97, 103
energy use 3 to BeO bonding 187
forcement manufacture 44 white 101
mullitein 17 Griffith, crack length 161
criteria 161
gahnite 226,299 flaw characterization 162
garnet 10, 306 flaw hypothesis 160
from akermanite 13 group relationship in silicates 28
hydro-,incement 50 gypsum 44-5, 48, 50
series 10
structure 20 H-componentincement 49
gehlenite 13,121 H-sialon 118
gel 44 hackle zone 165
oncement particles 48 hafnia in zirconia 272-4
texture in ZrO, preparation 274 halloysite 30,33
germanium oxide 254 hausmannite 299 .
gibbsite 26-7, 257-8 heat, capacity of ceramics 3
glass 132ff capacity change 238
alkaline 43 cycle 3,54
as model material 158 of sublimation 65
B,O,;in 104 radiationin glass 146
B,0,onBN_ 102 treatment diagrams 207
blooming 194 work 3,51
bondinceramics 58 hedenbergite 20
bubble removal 148 Heisenberg principle 69
bullet-proof 181 hemimorphite 9, 12
ceramic seals 184 Hercuvit cooker top 225
chemical treatment 179 hercynite 299
cordsin 148 heterogeneities, in fluoride glass 148
crack velocityin 161-2, 164 in Vycor 148
E-type strength 67 heterogeneous nucleant 210
352 INDEX

hexagonal layers incarbides 314, 315 kanemite 43

hole, conductivity mobility 98 kaolinite 28-32, 52ff
homogeneous nucleation 153, 207 mullite transformation 52 ff
house brick 33 refractory 236
hot pressing of SisN, 111 source of Al,O3; 257-8
Hund’srule 76, 82 to prepare cordierite 15
hybridization 79, 81 to prepare sialons 120
inBN 80, 101 Kapustinskii equation 249, 251
incarbides 324 keatite 35, 226
inSiC 105 kenyaite 43
in spinels 304 kernite 89
inTiO, 298 Kic_ 109, 166-7
hydrargillite 257 kimberlite 91
hydration ofcement 48-50 Kovar-glass sealing 185
hydrogel 44 kyanite 236
hydrogen, bondingin 76, 83
bond in SiO, corrosion 197 L-component ofcement 49
like orbitals 75 lanthanide, carbides 330
hypersthene 20, 24 contraction 272
nitrides 332
icosahedron 89-90, 100 lattice, energy 247-9
ideal cooling criteria 156 fringes 119
illite 30 trapping 176
ilmenite 272,295 layer structures 2,6, 13
Inconel-ceramic bonding 187 boat and chair forms in tridymite 37
Inglis equation 160, 181 close packed OH in brucite 26
instantaneous shear modulus 135 decomposition 51
intercalationcompounds 96 disorder in montmorillonites 32
interfacial bonding 124 inberyl 14
internal energy, glass 145 inchlorites 31
interstitial sites in glass 169 incordierite 15
inversion, of orbitals 73 in dickite 30
spinels 301-2 in halloysite 30
ion diffusion, in serpentine 57 in melilites 13
glasscorrosion 196-7 in three-layer silicates 27
ion distribution calculation 55 in wadeite 13
ion exchange, in glass 162 of oxide in sheet silicates 27
in glass corrosion 197 pseudo-layer description 118
strengthening 179 relationship SiO,-SiC 37
ion leaching 48 relationship serpentine-forsterite 56
ion stuffing 179 spacing in R-type sialons 118
ionic bond 239 ff lepidolite 31
and geometry 395 leucite 1
insilicates 13, 21 ff lime, in pozzolanas 50
partialin quartz 36 inslagcement 49
ionic radii, table of 219 linear combination of atomic orbitals 82
ionization energy, potential 243 liquid phase separation 148
isomorphous replacement 6, 8, 13 liquidus shape for immiscibility 213
inamphiboles 21 lithia, in glass ceramics 212
incement. 45 silica glass ceramics 224
in layer silicates 27-8 lithium 77, 87
inmicas 28, 30 aluminosilicate glass 179
inmontmorillonites 32 fluoride and B,O, viscosity 150
in pyrophyllites 32 sialons 122
insialons 120 silicon nitride 121
intalcs 30,32 load duration theory 181
of NinSi,N, 114 localized molecular orbitals 75
of Si by Alin glassceramics 225 Londonterm 249
toform feldspars 34, 39 Lucalox 262, 264
to form feldspathoids 34, 39
Italian asbestos 22 M-glass 133
machinable ceramic 227-8
Jagodzinski-Wycoff symbols 314 Madelungconstant 247, 302
magnesia 284—92
kalsilite 40 cold set 60
 INDEX 353

magnesia —continued mullite—continued

fromasbestos 24 on Al,O; surface 266
incermets 323 relationship to sillimanite 52
in densification of Siz3N, 113 secondary 55
in glass-ceramics 226 multiplicity of anatom 76
in Portlandcement 45 muscovite 28, 31
latticeenergy 249
reaction welding 187 nacrite 30,33
reaction with H,, calculation 319 Néel migration law 301, 304
refractory 44,54, 237 neon electron structure 77
sodium polyphosphate bonded 58 nepheline 39-40
vapour above 246 in glassceramics 227
magnesio ferrite 288 network formers, modifiers 145
magnesite 284-5 Newtonian cooling criteria 156
bonding with polyphosphate 60 viscosity 138
inglass 133 nickel sulphide in glass 178-9
magnesium, electronic configuration 77 niobium carbide 313
nitride 120 niobium nitride 316
oxychloride cement 285 niobium oxide nucleant 227
silicon nitride 120 nitrides 311 ff
magnetite 299 nitrogen, bondingin 78
maketite 43 energy levels 77
margarite 28 size 330
matter waves 71 nodes in orbitals 73,75, 83
Maxwell relaxation time 135 non-bridging oxygens in CaO-SiO, 219
mean free path 230 non-stoichiometry, in borides 336
mechanical bond in seals 188 incarbides 313
mechanical properties, and chemical bonds 67 insialon 114
specific of composites 129 inSi,;N, 110-11
melilites 9, 13,121 norbergite 228
mesophase 92 normalization of orbitals 79
metakaolinite 52-3, 55—6 nucleant characteristics 212
methylchlorosilane treatment of glass 201-2 nucleation 48, 57, 151, 206-7, 210
mica 8, 28, 30-32 nucleus critical radius 208
fluorine containing 227
microcline 39 O-O separation, in brucite 29
mineralization, of Al,O;, 259 in layer silicates 26
of mullitization 54 O} bond length 86
of spinel 305 O’ phaseinsialons 118
mirror zone 164-5 octahedral site preference energies 303
mist zone 165 olivine 9-10, 29, 256
molecular cagesinB,O; 150 opal 5,35
orbital theory 82 optical transparency 232-3
sieve 41 orbital 71-4
molecular orbitals, delocalized 87 angular momentum 244
energy diagram forH, 83 combinations 79, 82
energy levelsinB, 336 energy of 83
inB,C 100 hybrid 78-81
inboron 90 nodes 72-3, 83
incarbides 328 orientation effect in SiO, corrosion 197
oandz 84 orthogonality 80
tulesfor forming 84 n—o designation 78
moly-manganese method of bonding 187 Slater 75
molybdenum, carbides 314 tricentricin boron 91
glass sealing 184 organosilicons, hydrolysis of 19
monticellite 13 orthoclase 39
montmorillonites 28, 32 orthosilicates 8ff, 15 ff, 56
mullite 15 ff, 236 oxidation, of borides 342
from kaolinite 52ff of BN calculation 103
from metakaolinite 55 of MoSi, calculation 345
from sillimanite 16, 236 of SiC 106
from subsaturates 11 of SizN, 109, 113
in traditional ceramics 1 of silicides 344
mineralization 52-3 requirement in glass-metal seals 185
nitrogenformof 118 oxide classification 237
354 INDEX

oxygen 77-9, 81, 86 quartz —continued

deficiency in mullite 54 B-form in glass ceramics 225, 226, 227
electron affinity 244
molecular orbitals 85 radial electron density 54
parameter in spinels 300 radiative-conductivity formula 146
partial pressure in furnace bricks 290-91 radiusratio 142, 253
polyhedra 254-5 and polyhedra, table of 253
Ramsdell symbols 106, 119
partial molar volume of glass 146 random network model 143, 149
Pauli principle 76, 82 raw materials for refractories 236
Pauling rules 255 rectifier, silicon 99
Peniakoff method 258 redmud 258
periclase 58, 286 teflectance 4-6
perovskites 227, 256, 306-7 refractory, Al,O;-SiO, 16
perthite 39 cordierite 15
phase diagram, cobalt-WC 322 definition 235
for CaO stabilized ZrO, 279 glass-ceramics 122
for MgO stabilized ZrO, 280 hot strength 58
NaF-AIF; system 252 mortar 59
phase separation inCaO-SiO, 219 mullite 16
phase transitions 275 productionin U.S.A. 235
phenacite 122 quartz 35
phlogopite 31,227 relaxation time 135
phosphate bonded refractories 58 ff resistivity of graphite 95
phosphate glass corrosion 195 rib marks 163, 165
phosphate in bonechina 6 riebeckite 23
phosphoric acid 59-60 ring silicates 13-14, 21
phosphoric oxide as phase separation promoter rocksalt structure, of B,C 100
213, 225 ofcarbides 313
phosphorus 77 overlap integralsin 327
pi(z) bonding 36, 78, 81-2, 84-5, 94 R-sialons 117-19
plastic flow in glass 168 ruby laser 267
plasticity ofclay 32-3, 51 tule of mixtures 125
polarization 80, 104, 253 ruthenium in zeolites 42
and oxide properties 237 tutile 35, 151, 256, 295 ff
compensation in ionic model 253 Madelung constant 247
in glass networks 145 protective layer ofcement 322
polyhedra and radius ratio 254 radius ratio 254
polyhedral families 6, 33, 255-6
polymorphism, and glass formation 153 sanidine 39
inZrO, 274 Schrodinger equation 71-2
mechanism in B-Al,O0, 270 screw dislocation crystal growth 222
porcelain 1, 5—6, 11, 15-16 semiconductors 97-9
enamelling 186 sepiolite 33
pores, formationincement 49 serpentine 27-9, 56
incarbons 93 shear modulus of glass 135
inSi,N, 112 shear plane development 297-8
size ofin gels 44 sheet silicates 25 ff
Portland cement 44ff sialons 114 ff
potassium 28, 34, 39-40, 148, 243 silane 97
probability, surfaces 70 polycarbo fibres 105
functions 71 silica,amorphous 19, 38, 56, 57
pull-out work 130 attack byH,O 195
pyralspite series 10 condensation from metakaolinite 52 ff
pyrex 132, 168, 195 critical cooling rate 157
pyrogenic process 258 fromasbestos 24
pyrope 10 from kaolinite 52 ff
pyrophyllite 27-8, 31 from silicon alkoxide 134
pyrosilicate 8, 12ff gel 203
edge sharing 25 glass structure 145
pyroxenes 8, 19, 21-2, 24 group 34, 40ff
inceramics 1, 236
quantum numbers 72 in sialoncompositions 122
quartz 35ff polymorphs 256
«-form in glass ceramics 227 polymorphs in phosphate binders 59
 INDEX 355

silica—continued spinel—continued
protective layer 106 layersin 53, 56
radius ratio 254 nitrogen containing 121
strength 173 precursor to mullite 52
thermal expansion 58 silicon content 54
viscosity 157 spinodal 215-19
viscous flow activation energy 141 spodumene 20, 22, 31, 226
silicates 6ff volume contraction 228
calcium 45 stabilized zirconia 278, 281-2
chain 19ff stacking sequence in sialons 118,119
framework 33 ff static fatigue 172, 230
nitrogen analogues 121-2 steatite 30, 187
ortho- 9ff stishovite 35
polymers 149 strain, definition 66
pyro- 8,12 in fibre composites 125
ring 13 rate 139
sheet - 7, 25ff strain energy, from Na* inglass 170
Si-O-Siangles 81 release rate, critical 124
thermochemical radius 250 equation 166
silicide classification 344 stored 161
silicon 77, 88, 97 ff, 111 strained mixed cluster model 151
silicon carbide 37, 88, 104 ff, 118, 311 strength
fibres 105 and density 65
oxidation 107 and grain size 230
semiconduction 106 and microstructure 263
whisker strength of 67 and molecular weight 65
silicon monoxide, vapour pressure over mullite, molecular 169
calculation 18-19 -temperature data for glass 173
silicon nitride 88, 107 ff tensile, tables, 24, 67, 95
densification 102 tensile of SiC 67, 105
reaction bonded 101 tensile of Si;N, 108, 112
strength v temperature 294 theoretical 63, 65, 169
silicon—oxygen bridging angle 81-2, 110 theories of, for glass 158
silicon oxynitride 110-11, 116, 118 versus temperature inceramics 265
from nitrogen glasses 122 versus timeincement 45
sillimanite 11,52 stress, compressive in ZrO, 282
refractory 236 concentration equation 160
relationship to mullite 17, 52-3 corrosion 174
sintering energy 3 definition 66
size effect in glass fracture 183 development in glass seals 184
slow crack growth 175,177 distribution in Al,O; 267
smectites 32 distribution in glass 178
soapstone 30 flow in oxides 293
soda based glassceramics 227 formof 159
-silica glass failure 182 fracture of Al,O, 263
sodalite 41,50 intensity factor 108-9, 166, 175
units in zeolites 42 localized incement 49
sodium, electron configuration 77 restoring in solids 159
ferrite 258 wave velocity 163
polyphosphate 60 striations 163, 165
rhenanite 60 structural families 34, 256
silicate 42 ff, 258 subsaturate group 11, 45
tetrafluoroaluminate 250 sulphate process 295
sols 43 sulphur 77
spark plug, Al,O3in 259 surface area ofcarbons 93
specific heat of asbestos 24 surface compression values 178
specific stiffness, strength 129 sulphoaluminate clinker 50
spectroscopic notation 76 sulphur dioxide treatment of glass 201
spherulitic crystal growth 223 surface energy, and glass strength 160, 175
spessarite 10 and radius of curvature 174
spinel 121-2. 299 ff composite expression 166
blocks in B-Al,0, 269, 271 definition and units 64
critical cooling rate 157 effective 261
family framework 256 excess and strength 63, 160, 262
in glassceramics 227 incement 49
356 INDEX

surface energy —continued tobermorite —continued

in glass sealing 193 in autoclaved cement 50
thermodynamic 166 tolerance factor 307
surface tension, and surface energy 64, 160 topaz 11,35
in the float glass process 139 topotaxy 2, 16, S51 ff
symmetry of orbitals 72 in ZrO, transformation 276-7
synergism in glass corrosion 195 toughness 123-4
ofcordierite 15
Tabor relationship 168 of fibre composites 130
talc 8, 15, 27-8, 30, 32, 57 transistors 99
tantalum boride 341 transition layers 187,191
tantalum carbide 315, 317-18, 322-3 tremolite 22-4
tantalum nitride 316 tricalcium aluminate 45
tetrahedra, AIN, 120 tricalcium silicate 45
crumpling of 39, 107 tridymite 35, 37-40
distortion in beryl 14 family 256
edge length in sialons 119 type in phosphate binders 59
from Si, Al,O,N 114, 118 tungsten expansion coefficient 185
inglasses 144-5 tungsten carbide 315, 321
rotation of 9, 34, 38-9, 57, 107 tungsten nitride 316
SiN,O 110
spiralling 35 ugrandite series 10
tilting of 9, 34, 39, 107 uncertainty principle 69
tetrahedral angle 145 universal fatigue curve 173
tetrahedral boron 104 uranium boride properties 341
tetrahedral diagrams 57 uranium carbide 319-20
tetrahedral layers 27 ff uranium dioxide 237, 292 ff
tetrahedral linkage patterns 8, 9, 27, 34, 39, 41, uranium nitride 318
104, 145 uvrovite 10
tetrahedral representation of sialons 115
tetrahedral sites 55 vacancies in Si,;N, 111
tetrasilicic-mica ceramics 227 valence bond theory 77 ff
terminal fracture velocity 162 applied toSiC 105
thermal conductivity, of BN 102 valence disordering in spinels 306
of graphite 95 valence electron concentration 326
thermal decomposition calculation 108 valence orbitals 77, 101, 104
thermal density fluctuations 148 vanadium carbide 314-15, 319
thermal expansion, anisotropy in Al,O, 264 vanadium nitride 316
anisotropyin BN 102 vanadium oxide, bondingin 299
anisotropy in Si,N, 109 Van der Waal’s bonds
anisotropyinSi,N,O 111 in layer silicates 27, 29-30
as function of Si-O-Si bond 36 inB,O, 150
match in glass seals 184-5 vapour deposition, of BN 102
thermal insulation 33, 112 of SiC 105
thermal matching, of glass and crystal 58 of SizN, 111
of glass and metal 185 vermiculite 28, 32-3
thermal shock criteria 14 Vickers Hardness 168, 318, 341
resistance 230 and flow stress 168
thermal strengthening 178 indentation fracture pattern 180
tempering 178 viscosity of glass 138, 140, 142
thermal window material 31 anomalies 146, 150
thermodynamic, barrier to nucleation 153 volcanic ash to prepare sialon 120
surface energy 166 Von Mises criterion 266
thorite 10 Vycor 148, 195-6
thortveitite 10
tiles 1,23, 33 Wallner lines 163-4
tin oxide 227, 254 warm moulding 111
titania 151, 186, 237, 254, 295 ff water inclays 32-3
catalyst in glass ceramics 212, 225, 226, 227 wave, amplitude 70
deposition on glass 201 equation 71
titanium carbide 315, 318-20, 322, 329 function 70-1
titanium monoxide 317 nature of light 70
titanium nitride 316-18 weathering 32
titanyl sulphate 295 Weibull distribution 183
tobermorite 25, 50 wetting 193-4
 INDEX 357

wetting —continued yttrium aluminate 121

in glass-ceramic sealing 185 yttrium sialon 121
white graphite 101
willemite 12 Zachariasen 142 ff
wollastonite 13, 20,275 glass structure and strength 170
nitrogen containing 121 zeolites 8, 40 ff, 256
work, ofadhesion 192 zinc blende 37, 92, 105, 301
offracture 261 zinc oxide nucleant 226
of fracture of ZrO, 282 zircon 10,272,341
wurtzite structure 105,118 zirconia 59, 237, 272 ff
calcia stabilized 279
xenon fluoroplatinate 243 comparison toiron 283
X-phase in sialons 118 in glass strengthening 201
X-ray diagrams for glass 147 in SisN, hot pressing 113
magnesia stabilized 280
Young—Dupré equation 194 nucleant 225-7
Young’s modulus particle size and stability 278
and porosity in B,C 100 zirconium, carbide 280, 315
andtoughness 123 isopropoxide 278
definition 60 nitride 280, 315-16
of glass as function of composition . 133 oxychloride 273-4
primary and secondary 126 oxynitride 280, 315
yttria 113, 280 zyttrite 281
 cond 0 Os > -O
a f = Oo rariie api ne > O- ;

» / Aiden + at =——§e@ 6s: ' ‘

9 i> ae ae ae
ous ‘PAs ite
el ee otea
: — wean

* = , > .
wee Bi

7 :

wee

eh

$
ROeC10O4 17853
wp 24 fs Aeeera eb)
Y Ate ar a
RoThmentee:
rar ae
Aeapanneent:
FOOT AS: eaakon eat
ROAR POL AEiar
POMONA SDe
eres
PET ny ree ee
ROSESCen an
SS f geek ; Secsnah esas 5

RAE
eh McDee IS A
ecarcanedsAgo)

eeaaama
PRENSA Satan ee aed Asanee

cae
lotr ereemiran: WAR StasSinan dnd

FA
Bai ha ae TAA Ara enone naRA ak

ae
anne

aN
Sa SERN oe SADINBADDDoak tote taceearare SA aes Oconee
Seas ea eae DAS :
ce : SRP cM Aichenern thar
pee a e . . y acide saxcnsieye.
a—s does eae
BORDAahk!Arena hepato
TONRena
ener aPLC antes taPau dae ape Ninen da BOC i nan andeeVANE NS
eee fat esta crs aga
..CBee ants Poeun Roe ah) Pert RRA:
tenth Dying hn ab a PRD NOEL Retaal ON oe eiorete
Rae
SES PNP rea Ra ete Or aaa a AIE AeeGACLs dxesdseaterbo
eteRenee ehee
hen aad trees
Pera neeSenet ese by UA
Bare Rahat tonereehatCRPAL ROARS Sa tists tatenet Reeverete Kaien Spat ae 1gOb
dash eae Ake aeeyedeestes
Pr eheeBeMey AOreAL RRM ACNE
grenergienye temo RENEE Syne EY LORY
5
. DA as
aayee veone
Oeore tebe teases
PAT EPRI ale Da
HSIN ekDhSIN CAMA RAPETh he
ay eedeBeeee eesRees Peon rnrhmate pies
eae SASS PERERA ENS CLIN Ae Msnahed¥
¥ rf
elarescrasereretans
teneneb chasm Pam ena ee> NCA SOCAN Be Wasec covets
tea DIO TGy. es 043 ethonea8 eisaenaatt
SS Tacha achet he earPho
ORR TORO hak aera
Fae eho allyhha ei CEN aati ates
AOESED AP tata Ser heen?
Nehaaab anna: ean a Seana eats anewAt
ROA Ses HPA ERA tyes on eae70 Wan at aaah teak byliotate
ts taco
late ots
ANSE CREE SLs Rehan met een os WD ah Ne MutT. Seer!
eben tae
eran SRG ANANS a OAL Pid POS LING EELead ek gla ete
teat eee
SARS AREAS ta URS A
he PHS sere ole erase eAotele IG
Ca sat
INS dete cast i eae i
DAS tea aeae B apaat ena tee
Ut PAAsaad a SEAtate8,
“= pereest OP tie
weet eet Teeadele
Aba Ryd oRB NAYS ACCS Sue ERENT ke thie en!mS DMCSeRACR inRG
FasNeaes OG che Dent Et oRae sh Dad F.Lpease
eee aaapeie
MADESKS er OATS had hookant a
ohBoot
MehraRoeents tA Shy ere sained
rrpseaENG
SRT thheh recAp hehin ece nbete
Glyroyaeree larheen aon eer‘Tanspee na eA Re PSI GESON Len : deaa aaanetrea f
See) . eo bh eal SN Danes NY rarer Lon ange enna Pesta Be eaticns shares
Satins x a
raga ‘ AE AIO OER ats aon 3 Defences aise liea mete at iaateers ee
INL HORE G Otte EAT ANEee
Sere UE Di Rha ah toe
ANS A TaeAmeo ahah cada ds od hHt4LAG GCA6 HERS
1sdea Nan Ue mahal Hs od Sah)
Pe nnepran aaai Pree ek
paaean
SP ehonennh) SOR ea iaita ta haathSayal RAPP vehtansoknarn Bayar f=x
ah ’ emcee Pal acaeetned
Rakhes
Peet aan eter A Cine etal y uncial eG tote
RESP RCRRR
ettahsALN Ps snes
Speha TEU
Sora ee aes r
ee
RATES oust
Sho- Ae PaeaonARN
tharaaa!
eaeh altaREN oincen Anion onsite
Alnbhen Ohcnrentaansnieg CRenceaensn; nd
cateene > ron
as ans be e e oie
SORA
ORAS
SSTGt Ts EVV sae
AnrepSh PERS
BSA
teva rete S
. tate
AV
ebrtaac
khethpheath
ASOT Yat
Sancta aie ees ean Taek sina ee
aso sure
net aan
ard ea eaenee
CAESAR AAR ATDESO ARREARS WSeapaeersz UCN Nawe Zoaitn On ey we vvereetan NE i eae
OIeKGd
ORR DORR AR Wed + areaRa Cee Ab eceoar rept Radeon cam cAdhs - ie
AY ttae tyes eSaye; a Ree eo oy re
SSA ia RR CeRRSA AR Rhy Pines way aein
ELE the
s) =
ASnE h Sty).
ehh
BAA A Core Ane aA aya ae acta Nahar pede 54k
Deve acha an he
Senn h uM
; Nhat
ealeTe Ey ts
oa
a Swe Wey PULe
f
Tatas
ake
ie)
ny Seue
AE
atu iahsic PES One
aeieitaenaed
OPIS akenhebenee
eG as
ONG a haat
nroinshv ek Prrearnes O dil ean
Kran
arcane tar
ten andaos bones
SteeEien ayaaa
Sgtacancl
ium
; et
i SES a
ETL: spaenteal Heeananeariee
Sistah eky eee Cote ines mnering tan 54 t nee
Mae Si Sats amienteeiey
24 estesiain cece
Shae 3 Ros easaan nn pene
eae has anan
aleNE cnc
ene nets Naa?
ahaha orate
Bia ate
Sieh nana :
a K ; ta Ibe
ta EIAs£1005ts
ace Taye
ia
RabLS RRL
SRAei
SRR
poewenth octca Tare
RUR Sree
TEN each
tee foe ihre
Rbah thona
ates een RATT TEN
ah eanae oacen Soe : ; : berrccer sot L ore tonalotytetgta ata
teeDn eae
Sea a eeNIAAA Maa et eee ees sa
: Mra eS
Wane Re Dae ton OES yale a pak
bpenivare tener std
ae 1 Gs yeyea
“4 cS SEW aN e
EAM Naar ean Naat areaseta iatoo 2ateh atid a atesbe
ai AAScaaeA
Sapa ates ie ks
OSA ane as eens raneanoh R A Net at ee aie ees seat
ate elestatta
a
yretdarts ef i
aes ie au ee carne i mi eagte:
aig itetetatohel Cree +
r a bay
SeeMU EOE Un IES
thas tine
behes Ca aaranay
att sae Saat a aaa H,;
,
te uu! SoM
Panay
Re
Set aE siti
re
aN es i ca GAN ie On Be aged asad
serel antares
Nes a Bras aeAtaroR bavacate cate Aah Paaaae plglits
het toleee
Sahay eh RENE n Ms eee)
On on
DURES Cantatas
ca thSanna
EARSADANaC hay 4
me
an asta
oh
bak hee
tt
s
sahPR ups at
a =e earnata
s UNIS ess apes
pte
SASH AN eaHne Nan Canatnt> eatin
SARA SuS snnRUAAT
te een ea
cues .
cag ant Gh
amen
caer
ech
ae
AsPi Maasai
Mean Lal
ei
Perretan Satan
CARRE ANTE apt SEM
chbweatiathea Ans Re: mop hie BAputomeelhs Het atts ae
iia
ceRnra Ran omen: iss eaten:
t*
Raatuncer am
AM gona ahah
ee Pea men
Pe N ate
NubevesonG@ente e AEN,
wil ed gt
nen
Hiatt eels
mCi oa Rn a igis aurasbt
RIURAAAAPaSs SRNL AN
on
MRR DR i ieee aig eit aaahaed
ce
x its
Bit
ha NanSa
ath ; west
PA ily
Seautthy) Ritacaigy eae
si Nee seithapn
he
pasts
Hi f
Ahi anes i Aa
A ssAn Se ea aa yt
sy f
‘ eee
bee fa : bel pe
aie. KG
i
fu : Me
aaa

pr
Pie
Ntoh MET RU ‘
Se NOR
nee i eis iout ches oe
EAR
,
ns rotate eee i Peraatie
ere Rouge Meer coma aL at citotis Pesae accaotee spas ee
Siete tinNM PMU MMe
bi Cacia ater He ies His Sec Mora aN sD
CHA GH ea
EM
U iUMM AHofHate MSRM Tiae
Beet sare CO anPare cua iknie; i
reste Rt
iy
arta) 1 iG
rata Mate
nay i
ia
Gv
Met
ean Ae
Ha i eh
i
G Sen an
Ca AA RR
tie
} eee
oe,
CSaicaNn aa
Nes
8 ae iHrea
if ef if is
i
aa
i

ee
i }

ea