Appl.Magn.Reson.

(2007) 32, 663689

DOI 10.1007/s00723-007-0043-y Applied
Printed in The Netherlands Magnetic Resonance

29
Si, Al, 1H and 23Na MAS NMR Study of the Bonding
27

Character in Aluminosilicate Inorganic Polymers

M.R.Rowles1,&, J.V.Hanna2, K.J.Pike2,&&, M.E.Smith3, and

B.H.OConnor1

Centre for Materials Research, Department of Imaging and Applied Physics,

1

Curtin University of Technology, Bentley, Western Australia, Australia

2
NMR Facility, Institute of Materials Engineering Science, Australian Nuclear Science and
Technology Organisation, Sydney, New South Wales, Australia
3
Department of Physics, University of Warwick, Coventry, United Kingdom

Received 2 May 2007; revised 12 May 2007

Springer-Verlag 2007

Abstract. 29Si, 27Al, 1H and 23Na solid-state magic-angle spinning (MAS) nuclear magnetic resonance
(NMR) has been used to relate nominal composition, bonding character and compressive strength prop-
erties in aluminosilicate inorganic polymers (AIPs). The 29Si chemical shift varies systematically with
Si-to-Al ratio, indicating that the immediate structural environment of Si is altering with nominal com-
position. Fast 1H MAS and 29Si TSiH/T1r relaxation measurements demonstrated that occluded pore H2O
mobility within the disordered cavities is slow in comparison with H2O mobility characteristics observed
within the ordered channel structures of zeolites. The 27Al MAS NMR data show that the Al coordina-
tion remains predominantly 4-coordinate. In comparison with the 29Si MAS data, the corresponding 27Al
MAS line shapes are relatively narrow, suggesting that the AlO4 tetrahedral geometry is largely unper-
turbed and the dominant source of structural disorder is propagated by large distributions of SiO bond
angles and bond lengths. Corresponding 23Na MAS and multiple-quantum MAS NMR data indicate
that Na speciation is dominated by distributions of hydration states; however, more highly resolved 23Na
resonances observed in some preparations supported the existence of short-range order. New structural
elements are proposed to account for the existence of these Na resonances and an improved model for
the structure of AIPs has also been proposed.

1 Introduction

Aluminosilicate inorganic polymers (AIPs), also known as geopolymers [1], have

received considerable attention [24] in recent years, showing promise in appli-
cations such as cements and concretes [5, 6], toxic waste storage [7, 8] and fi-

&
Present address: Division of Minerals, CSIRO, Clayton South, Victoria, Australia
&&
Present address: Department of Physics, University of Warwick, Coventry, UK
664 M.R.Rowles et al.

ber-reinforced composites [9]. AIPs consist of an amorphous structure resulting

from the polymerization of aluminosilicate monomers in an alkaline solution [1].
The amorphous nature of these AIPs makes characterization by techniques for
ordered materials, notably X-ray diffraction (XRD), difficult to interpret and elu-
cidate any structural information. To this end, nuclear magnetic resonance (NMR)
has been used to elucidate short-range order information regarding the bonding
configuration of atoms present in the material. Recent work by Rowles and
OConnor [10] described the influence that chemical composition exercises on
the compressive strength of these materials. However, further investigation is
needed to establish the link between chemical bonding character and specia-
tion, and compressive strength, to more formally understand the structureprop-
erty relationship characterizing AIPs.
One potential use of AIP-type materials is as a replacement for Portland
cement. Portland cement concretes are the most extensively used construction
material, with approximately 6 billion tonnes being produced annually [11]. Ce-
ment manufacturing is one of the leading contributors to world CO2 production,
accounting for 5% of the worlds anthropogenic CO2 output[12]. It is estimated
that the use of geopolymer technology could reduce CO2 emissions from the
cement and associated aggregate industries by a factor of five [13]. Another
potential application of this class of materials has arisen through investigations
into their suitability as immobilization matrices of high-level and low-level nuclear
waste from spent nuclear fuel reprocessing and/or weapons production [1416].
Initial product consistency test (PCT)-type leach testing on Cs and Sr extraction
from these systems has demonstrated, at least to a primary suitability stage, that
such an application is feasible [17, 18]. To date, little work has been reported
on concretes based solely on geopolymers [2, 1921], as the majority of con-
crete research has been done on using Portland cements in combination with
different geopolymeric ingredients including flyash [2224], slag [25] and silica
fume [26].
Rahier et al.[2729] proposed the following schematic polymerization reac-
tion for the formation of AIPs by sodium silicate activation of metakaolinite:

wNa2O.xSiO2.yH2O + Al2O3.2SiO2
<100 C
wNa2O.Al2O3.(2 + x)SiO2.zH2O

The values of w, x and y depend on the composition of the alkaline activa-

tion solution and z is the degree of hydration of the resultant polymer. While
the precise reaction paths are still unknown, Davidovits[1] has proposed reac-
tion pathways involving the polycondensation of hypothetical orthosialate ions,
as shown below for two Si-to-Al ratios.
Pathway for AIPs with Si-to-Al molar ratio of 1:1:

NaOH
(Al2O3.2SiO2)n/2 + 3nH2O n(OH)3SiOAl()(OH)3,
 Bonding Character in Aluminosilicate Inorganic Polymers 665

NaOH
()
n(OH)3SiOAl (OH)3 (
Nan OSiOAl() )n + 3nH2O.
O O

Pathway for AIPs with a Si-to-Al molar ratio of 2:1:

NaOH
(Al2O3.2SiO2)n/2 + nSiO2 + 4nH2O n(OH)3SiOAl()OSi(OH)3,
(OH)2

NaOH
()
n(OH)3SiOAl OSi(OH)3 (
Nan OSiOAl()OSi )n + 4nH2O.
(OH)2 O O O

The first reaction pathway shows the production of the polysialate inorganic
polymer, with a Si-to-Al ratio of 1, and the second reaction pathway shows the
polysialate-siloxo polymer. As additional SiO2 is added to the system, the siloxo
(SiOSi) chain between the sialate (SiOAl) groups becomes longer.
Solid-state magic-angle spinning (MAS) NMR has been demonstrated to be
a very effective tool for characterizing AIPs [3034]. The technique is able to
determine the different coordination environments for every Si, Al and Na posi-
tion for any given nominal stoichiometric preparation. Furthermore, provided that
the preparation is free from paramagnetic species, this information can be quan-
titative and representative of the bulk sample. Given the typical ranges of coor-
dination of silicon and aluminium in such materials, i.e., silicon is predominately
tetrahedral and aluminium can vary between AlO4, AlO5 and AlO6, different fac-
tors characterize the change in shift within these materials. In true geopolymers,
silicon is present as SiO4 units that are completely connected (i.e., Q4; see ref.
25) and it is the number of next-nearest-neighbor aluminium (i.e., Q4(mAl)) that
varies, with m"04. Since Al is more electropositive than Si, an increase in
the number of Al atoms surrounding the SiO4 unit results in a down-field shift
of the 29Si resonance to a less negative position within well-defined chemical
shift regions [30, 32, 3537]. In contrast, an Al coordination number varies more
strongly by changing the number of oxygen atoms immediately bonded to each
Al position; this number can be 4, 5 or 6. Due to the Loewenstein avoidance
principle [38], which asserts that AlOAl bonds cannot exist, it is therefore re-
dundant to refer to the number of SiO4 tetrahedra that can connect to an AlOn
moiety as only silicon can be a next-nearest neighbor. The 27Al chemical shift
ranges defining AlIV, AlV and AlVI speciation have been extensively discussed[30,
39]. The ranges are well separated and allow the coordination number of AlOn
units to be unambiguously determined. Unfortunately, the same degree of cer-
tainty cannot be obtained from 23Na chemical shifts as the strongly ionic nature
of Na!, and its inability to form bonds with significant covalent character, re-
sults in a small overall chemical shift range with ill-defined 23Na shift ranges
describing Na speciation. Koller et al.[40] have attempted to map these shift
666 M.R.Rowles et al.

ranges, but knowledge of the 23Na isotropic chemical shift provides little cer-
tainty towards the identification of the Na local environment.
The principal aim of the work presented in this paper is to elucidate changes
in the bonding network and speciation as the chemical composition of the AIPs
is varied. To address this objective, 29Si, 27Al, 1H and 23Na MAS NMR studies
have been undertaken on suites of samples representing a range of composi-
tions in order to give a broad view of the structural variations in the polymer
network.

2 Experimental

Three suites of samples with differing chemical compositions were chosen (Table
1 and Fig. 1). Suite A, defined by a fixed Si-to-Na molar ratio (Si:Na"2), was
chosen to give an indication of the change in bonding over the complete SiAl
range as it encompasses both the weakest and strongest samples as measured by
compressive strength[10]. Suites B and C consisted of samples that were chosen
from the extremities of the composition range to give an indication of the influ-
ence of Na-to-Al ratio on the Si-bonding character.

2.1 Raw Materials and Processing

The starting materials were a crystalline kaolinite (Kingwhite 65, Unimin Aus-
tralia Ltd.) and an amorphous silica fume (Australian Fused Materials Pty Ltd.).
Metakaolinite was obtained by heating the kaolinite at 750 C in air for 24 h,
in order to obtain an X-ray amorphous aluminosilicate material with enhanced
reactivityto alkaline media [41]. The detailed chemical compositions of the pre-
cursor materials are given in ref. 10. All sodium silicate activating solutions were

Table 1. Sample designation, compressive strength and molar composition. The errors given in the
compressive strength represent one standard deviation calculated from three samples.

Sample Suite Compressive strength (MPa) Si:Al:Na

Kaolinite # # #
Metakaolinite # # #
Silica fume # # #
1.1/0.6 A 0.40(2) 1.08:1:0.57
1.5/0.8 A 6.2(5) 1.5:1:0.75
2.0/1.0 A 51.3(13) 2.0:1:1.0
2.5/1.3 A 64(3) 2.5:1:1.26
3.0/1.5 A 2.6(2) 3.0:1:1.53
1.1/0.8 B 2.2(3) 1.08:1:0.75
3.0/2.0 B 19.9(7) 3.0:1:2.0
1.1/1.0 C 4.4(7) 1.08:1:1.0
2.0/2.0 C 11.8(16) 2.0:1:2.0
 Bonding Character in Aluminosilicate Inorganic Polymers 667

Fig. 1. Contour plot of compressive strength versus compositional Si-to-Al and Na-to-Al molar ra-
tios for the samples under study, from ref. 10. Composition suite A provides a sequence of samples
across the composition range, which includes the maximum in compressive strength. Suites B and
C provide examples for the extremes of the compositional range.

prepared by mixing the required amounts of sodium hydroxide (Sigma Chemi-

cals Pty Ltd.) and silica fume with deionized water (Si-to-Na molar ratios,1.0,
1.5 and 2.0; H2O-to-SiO2 molar ratios,2.8) and heating at 75 C until a clear
solution was obtained. No precautions were taken to exclude CO2 from the so-
lutions. The insoluble zirconia contained in the silica fume was separated from
the activating solution by decanting. Samples were prepared with nominal Si-to-
Al molar ratios between 1.083.0, and with Na-to-Al molar ratios between 0.57
2.0. Water was added to give the final AIP paste the same consistency for all
samples. The nominal compositional ratios given in this study are molar ratios
unless otherwise explicitly stated, and all AIP samples are identified by their two
characteristic ratios of Si to Al and Na to Al (Si:Al/Na:Al).

2.2 MAS NMR Measurements

High-resolution solid-state 29Si and 27Al MAS NMR spectra were acquired at am-
bient temperatures on an MSL-400 NMR spectrometer (B0"9.4 T) operating at
668 M.R.Rowles et al.

the 29Si and 27Al frequencies of 79.48 and 104.23 MHz, respectively. 29Si MAS
NMR data were acquired using a Bruker 7 mm double-air-bearing probe with cross-
polarization (CP)MAS and single-pulse (Bloch decay) methods, both of which uti-
lized high-power 1H-decoupling during data acquisition. The MAS frequencies
implemented for these measurements were about 5 kHz. For the 29Si CPMAS
experiments, a recycle delay of 5 s, a 1H-29Si HartmannHahn contact period of 5
ms and an initial 1H p/2-pulse width of 5 s were common to all CPMAS data.
For the corresponding 29Si MAS single-pulse high-power 1H-decoupling measure-
ments, a single 29Si p/4-pulse width of 2.5 s and a preacquisition delay of 10 s
were used in conjunction with recycle delays of 3060 s for quantitative 29Si
measurements. All 29Si MAS and CPMAS chemical shifts were externally refer-
enced to tetramethylsilane (TMS) at 0 ppm via a high-purity sample of kaolinite
(#91.2 ppm) which was also used to establish the 1H-29Si HartmannHahn condi-
tion. All 27Al experiments were conducted using a Bruker 4 mm double-air-bear-
ing probe from which MAS frequencies of about 1516 kHz were implemented
for line narrowing, and all measurements utilized single-pulse (Bloch decay) ex-
periments without 1H-decoupling. For quantitative estimates of the central transi-
tion intensities to be made, the quadrupolar nature of the 27Al (I"5/2) nucleus
necessitated that flip angles are close to the condition [30, 37, 39]

(I!1/2)wrf tp7p/6. (1)

For these 27Al measurements, nonselective p/2-pulse times of 4 s were cali-

brated on a 1 M Al(NO3)3 solution from which selective pulse times of 0.6s
were employed for data acquisition on all solid samples. Preacquisition delays
of 3 s and relaxation delays of 5 s duration were typical for these measure-
ments; however, checks for abnormally long T1s were undertaken with recycle
delays of up to 30 s being implemented. 1 M Al(NO3)3 solution was also used
as a 27Al chemical shift reference at 0.0 ppm.
High-resolution, variable-temperature, solid-state 29Si MAS data were ac-
quired at 8.45 T on a Varian/Chemagnetics Infinity 360 spectrometer to inves-
tigate the magnetization buildup and decay (TSiH/T1r) characteristics within each
geopolymer preparation. All data were acquired using a Varian 6 mm double-
air-bearing probe with 1H-excitation pulse and 1H-decoupling conditions simi-
lar to those chosen for data acquisition at 9.40 T, with the HartmannHahn
contact period being varied typically from 200 s to 7 ms. For these measure-
ments shorter recycle delays of 1 s were used. Each variable TSiH/T1r experi-
ment was performed at 21 and #40 C to investigate the temperature depen-
dence of the 29Si magnetization dynamics.
High-resolution one-dimensional (1-D) solid-state 1H MAS NMR data were
acquired at 14.1 T using a homebuilt, Samoson design, 1.8 mm fast MAS probe.
Each 1H spectrum was acquired with a single-pulse-acquire sequence in conjunc-
tion with a fast MAS frequency of 40 kHz, a 1H p/2-pulse time of 2.5 s, a
preacquisition delay of 8 s and a recycle delay of 120 s. All 1H data are ex-
ternally referenced to TMS (0.0 ppm) via doubly deionized and distilled H2O at
4.7 ppm.
 Bonding Character in Aluminosilicate Inorganic Polymers 669

High-resolution 1-D solid-state 23Na MAS NMR data were acquired at mag-
netic field strengths of 8.45, 9.4 and 14.1 T on Varian/Chemagnetics Infinity 360,
Bruker MSL-400 and Varian/Chemagnetics Infinity 600 spectrometers operating at
characteristic 23Na frequencies of 95.25, 105.10 and 158.74 MHz, respectively.
Single-pulse (Bloch decay) 23Na experiments without 1H-decoupling were con-
ducted using Bruker 4 mm double-air-bearing probes at 8.45 and 9.4 T, and a
Varian/Chemagnetics 4 mm T3 design probe at 14.1 T, with MAS frequencies
in the range of 1516 kHz. To satisfy the quantitative acquisition conditions
described by Eq. (1) for the quadrupolar 23Na (I"3/2) nucleus, nonselective
p/2-pulse times of 3 s were calibrated on a 1 M NaCl solution from which
selective pulse times of 0.6 s were implemented for data acquisition. The pre-
acquisition delay at each field was 3 s and the relaxation delays were typically
of 8 s duration, with checks for abnormally long T1s also being undertaken (re-
cycle delays up to 60 s). The 1 M NaCl solution also served as a 23Na chemi-
cal shift reference at 0.0 ppm.
23
Na 2-D multiple-quantum (MQ)MAS NMR studies [4245] were performed
on a range of AIPs at magnetic field strengths of B0"8.45 and 14.1 T on Varian/
Chemagnetics Infinity 360 and Infinity 600 spectrometers. An amplitude-modu-
lated triple-quantum (3Q)MAS NMR experiment [42, 46] was used which pro-
vided high-resolution hypercomplex 2-D spectra whose indirectly detected dimen-
sions (F1) have sign discrimination and are free from the second-order quadru-
polar broadening exhibited by the central transition of the spin I"3/2 23Na iso-
tope. At each B0 field two 2-D MQMAS data sets and a conventional 1-D MAS
NMR spectrum using a single-pulse-acquire sequence were acquired for each
sample, the latter being used as the F2 projection for the presented 2-D MQMAS
data. All 14.1 T experiments were performed using a Varian 4-mm T3 probe.
Each 1-D projection spectrum was acquired with a MAS rate of 15 kHz, and a
nonselective p/2-pulse time of 3.12 s that was calibrated on solid NaCl which
corresponded to a B1 radiofrequency (rf) field strength of about 80 kHz. A pulse
width of 1.0 s was used for data acquisition (which corresponded to a flip angle
of about p/6), in conjunction with a preacquisition delay of 3 s and a relax-
ation delay of 8 s. The 2-D MQMAS studies utilized a MAS rate of 12.5 kHz,
with the 3Q-excitation, 3Q-conversion and Z-filter pulses having widths of 4.0,
1.6 and 5.0 s, respectively. The B1 field strengths were about 80 kHz for the
3Q-filter pulses and about 20 kHz for the Z-filter pulses. For each sample, re-
laxation delays of both 6 and 1 s duration were used for the acquisition of all
2-D MQMAS data. The latter experiments exhibited less F1 resolution, thus fo-
cussing on the broader spectral components with faster relaxation rates (as dis-
cussed below). Each 2-D MQMAS data set typically required about 4 h for ac-
quisition. Similar experiments were performed at 8.45 T using a Bruker 4 mm
rotor probe and a MAS rate of 12.5 kHz. 1-D spectra were acquired with single
pulses of 1.0 s duration with an overall B1 rf field strength of about 80 kHz
again implemented. For the 2-D MQMAS experiments the 3Q-excitation and 3Q-
conversion pulses were of 5.0 and 1.9 s duration, respectively, generated from
a B1 field of about 80 kHz, while the Z-filter pulses were of 4.5 s duration
670 M.R.Rowles et al.

from a B1 field of about 25 kHz. The 8.45 T experiments utilized relaxation de-
lays similar to those performed at 14.1 T (6 and 1 s), but each experiment was
typically averaged for twice as many acquisitions.

2.2 Data Analysis

All 29Si and 27Al chemical shifts, integrated intensities and line widths from the
data presented in Figs. 2 and 3 were measured by the WINFIT profile fitting
software [47]. Chemical shifts (d) and line widths were extracted from these
simulations; however, the reported 27Al shift positions do not represent isotropic
chemical shifts as they remain uncorrected for second-order quadrupolar effects
[30, 37, 39], unless otherwise stated.
Each 29Si and 27Al spectrum was simulated five times, with the average of
these data reported in Table 2 as chemical shifts and line widths. An example
of the curve fitting procedure used to calculate the peak positions and line
widths is given in Fig. 4. For the 29Si and 27Al spectra of the inorganic poly-

a b c d

Fig. 2. Measured 29Si MAS data (a), 29Si CPMAS data (b), 27Al MAS data (c) and 1H fast MAS
data (d) for the range of AIP compositions under study. In the 1H MAS data, the resonance at 18.1
ppm is assigned to HCO3# resulting from small amounts of sample hydrolysis and carbonation, while
the less resolved resonance at about 1.0 ppm (asterisk) is a background artifact due to the organic
polymer rotor cap material.
 Bonding Character in Aluminosilicate Inorganic Polymers 671

29Si MAS 21 C
40 C

100 100

50 50
1.1/0.6

0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
100 100

50 50
1.5/0.8

0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
100 100

50 50
2.0/1.0

0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
100 100

50 50
2.5/1.3

0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
100 100

50 50
1.1/0.8

0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
100 100

50 50
1.1/1.0

0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
100 100

50 50
2.0/2.0

0 0
0
50
100
150 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
ppm contact time (ms) contact time (ms)

Fig. 3. 29Si TSiH/T1r variable contact time experiments as performed for the entire suite of AIP prepa-
rations. The 29Si shift positions at which this behavior is measured is indicated for the AIP
(C) and zeolite (x) components.

mers, the values for the chemical shift and line width for metakaolinite were
fixed, and only the resonance area was refined. The change to relative specia-
tion of each sample was elucidated from the change in these fitted spectral
regions. All errors quoted are two standard deviations of the mean, calculated
672 M.R.Rowles et al.

Table 2. Reported 29Si and 27Al chemical shifts, line widths and 29Si CP period time constants of the
nominal compositions composing suites A, B and C and associated precursor materials.

Sample Suite 29
Si 27
Al

diso Line width TSiH T1r TSiH T1r d Line width

)0.1 )0.02 )0.02 )0.02 )0.02 )0.02 )0.1 )0.01
(ppm) (kHz) 21 C 21 C #40 C #40 C (ppm) (kHz)
(ms) (ms) (ms) (ms)

Kaolinite # #91.7 0.11a 0.0 1.05

1.6 5.38
Metakaolinite # #104.8 1.64 29.1 3.56
53.6 2.17
Silica fume # #111.4 1.11 # #
1.1/0.6 A #85.8 0.84 0.27 6.03 0.40 4.34 58.3 1.04
#89.7b 0.11 1.47b #c 1.00b 6.95b
1.5/0.8 A #88.9 1.18 0.28 4.67 0.26 3.43 56.5 1.60
2.0/1.0 A #92.1 1.26 0.28 4.85 0.32 4.12 55.3 1.38
2.5/1.3 A #94.0 1.26 0.35 4.50 0.37 4.20 54.6 1.40
3.0/1.5 A #92.8 1.32 56.2 1.13
1.1/0.8 B #84.4 0.78 0.29 5.59 0.26 5.02 58.2 1.04
#89.4b 0.11 2.11b #c 0.26b #c
3.0/2.0 B #92.9 1.27 55.5 1.06
1.1/1.0 C #86.5 0.84 0.30 5.26 0.40 3.13 57.4 0.55
#89.5b 0.12 1.58b # c 0.34b 16.80b
2.0/2.0 C #85.9 1.17 0.35 2.71 0.24 3.29 58.0 1.12

a
This represents a total line width which is composed of two crystallographically inequivalent posi-
tions; each resolved resonance (observed under appropriate 1H-decoupling conditions) contributes a
width of about 5055 Hz to the total line width at 9.4 T.
b
Measurement of zeolite A component within this preparation.
c
Not measured.

from the fivefold simulation of each spectrum. The coordination numbers for
the different Si and Al structural moieties composing each AIP were ascertained
by directly comparing all measured 29Si and 27Al chemical shifts with well-
established 29Si and 27Al chemical shift ranges that have been reported for alu-
minosilicate, cementitious and siliceous materials [30, 37, 39]. The notation
describing Si and Al coordination environments is consistent with that employed
by Lippmaa et al. [35, 37] and Engelhardt et al. [32].

3 Results and Discussion

The 29Si, 27Al and 1H MAS NMR data are shown in Figs. 2 and 3, and the cor-
responding variable B0 23Na data are given in Figs. 6 and 8 (see below). Table 2
provides the corresponding 29Si and 27Al chemical shifts, line widths, and the mea-
sured TSiH and T1r time constants characterizing the competing processes respon-
sible for the buildup and decay of the 29Si magnetization within these systems.
 Bonding Character in Aluminosilicate Inorganic Polymers 673

Fig. 4. Examples of the spectral simulation and deconvolution of 29Si and 27Al MAS data used to
calculate chemical shifts, line widths and relative intensities for sample 1.1/0.6, 29Si MAS (a), sample
2.0/1.0, 29Si MAS (b), sample 2.0/1.0, 29Si CPMAS (c), sample 2.0/2.0, 29Si MAS (d), sample
2.0/2.0, 29Si CPMAS (e), and metakaolinite, 27Al MAS (f).
674 M.R.Rowles et al.

3.1 29
Si and 1H MAS NMR Studies

The kaolinite spectrum in Fig. 2a exhibits a characteristic single narrow 29Si reso-
nance positioned at #91.5 ppm, indicating a classic Q3 Si species in a highly
ordered matrix with no chemical shift dispersion[30, 32, 4850]. An expansion
of this spectrum (not shown) indicates that very highly crystalline kaolinite
samples devoid of paramagnetic impurities can exhibit two partially resolved
resonances with very similar chemical shifts if the high-power 1H-decoupling field
strength is of sufficient intensity. In contrast, metakaolinite and silica fume show
very broad resonances at #104.8 ppm (Q4(1Al)Q4(0Al)) and #111.5 ppm (Q4),
respectively, which are indicative of highly disordered structural networks. Since
the nearest-neighbor interactions influence the chemical shift of a particular Si
species, the shift of the 29Si resonance from kaolinite to that of metakaolinite
suggests that the Si network is becoming detached from the Al network; i.e., Al
is exerting less influence on the Si chemical shift and the Si sublattice in meta-
kaolinite is tending towards that observed for silica fume.
Upon accurate spectral simulation of the quantitative 29Si MAS NMR data
in Fig. 2a, the AIPs exhibit chemical shifts ranging between #85.5 and #94.0
ppm. In terms of the next-nearest-neighbor aluminium coordination number (m)
in Q4(mAl), the 29Si chemical shifts correlate to a variation in m of about 3.7
2.2, as estimated from an interpolation of established literature values [30, 32,
35, 36]. A plot of m versus sample composition for suite A is shown in Fig. 7a,
and it reflects a reduction in m from about 3.7 for sample 1.1/0.6 (Si:Al/Na:Al)
(#85.8 ppm) to a value of about 2.2 for sample 2.5/1.3 (#94.0 ppm). (The esti-
mated errors in the silicon-to-aluminium coordination number m in Fig. 7a take
account of the 29Si chemical shift spread associated with each coordination num-

27Al MAS 23Na MAS

a 29Si MAS b c

Metakaolinite

6 h, 1 M NaOH

24 h, 1 M NaOH

6 h, 4 M NaOH

24 h, 4 M NaOH

24 h, 12 M NaOH

50 100 150 200 100 0 100 100 50 0 50 100

29Si (ppm) 27Al(ppm) 23Na(ppm)

Fig. 5. 29Si MAS (a), 27Al MAS (b) and 23Na MAS data (c) for the reaction of metakaolinite precur-
sor with varying-strength NaOH for different time periods, without stoichiometric adjustment from
Na silicate solutions.
 Bonding Character in Aluminosilicate Inorganic Polymers 675

8.46 T a 9.4 T b 14.1 T c

1.1/0.6

1.5/0.8

2.0/1.0

2.5/1.3

3.0/1.5

1.1/0.8

3.0/2.0

1.1/1.0

2.0/2.0

Na2CO3

40 20 0 20 40 60 40 20 0 20 40 60 20 0 20 40

ppm ppm ppm

Fig. 6. 23Na MAS NMR data for the entire AIP suite acquired at 8.45 T (a), 9.4 T (b), and 14.1 T (c).
Data for Na2CO3 is included for comparative purposes to demonstrate its influence on some data.

ber and were calculated as two standard deviations of the average of the chemi-
cal shift values reported for sodium aluminosilicate systems[30, 32, 35, 36].) Fur-
thermore, this shift in the resonance position, and the concomitant change in Si
coordination, is mirrored by the physical properties of the material. The com-
pressive strength of the AIPs shown in Fig. 7b increases to a maximum at sample
2.5/1.3 and then decreases rapidly and monotonically on either side of the com-
positional range, directly correlating with the amount of residual unreacted meta-
kaolinite coexisting in each final product. From Fig. 7b it can be observed that
maximum compressive strength is achieved when undetectable levels of meta-
kaolinite are observed in the highest strength 2.5/1.3 preparation. Residual un-
reacted metakaolinite is readily observed in the 29Si MAS spectra of Fig. 2 as a
semiresolved higher-field tail situated at about #105 ppm, which is an unam-
biguous match with the spectrum of the metakaolinite precursor also shown in
Fig. 2. Some typical examples of the spectral deconvolution and simulation
methods employed to ascertain the amounts of the metakaolinite residues within
these data are shown in Fig. 4a and b. Metakaolinite and other unreacted pre-
cursor residues within AIP preparations have also been reported by other au-
thors[17, 18, 28]. This additional inhomogeneity can influence the strength of
the AIP in two ways: firstly, it can act as flaws in the bulk polymer structure
thus promoting cracks and fractures to propagate through the bulk material, and
secondly, the presence of substantial quantities of metakaolinite alters the nomi-
676 M.R.Rowles et al.

Fig. 7. a Variation of the parameter m describing the Si next-nearest Al neighbor coordination

(Q4(mAl)) versus sample composition. The upper plot shows the variation in m coordination calcu-
lated from the measured chemical shift. The lower plot shows the theoretical coordination number
assuming the polymer network is made of solely SiO4 and AlO4 tetrahedra, with no residual un-
reacted metakaolinite accounted for. b Relative atomic percent concentrations of residual unreacted
metakaolinite present in the AIP compositions of suite A as calculated from the 29Si MAS spectra,
correlated with their compressive strength characteristics.

nal composition of the AIP relative to the bulk stoichiometry, which may inhibit
the complete development of the polymer network.
From Fig. 7a it can be observed that the Si coordination to next-nearest Al
neighbors m derived from the 29Si chemical shifts differs from the theoretically
derived m values calculated from the nominal Si-to-Al ratio which assumes the
complete dissolution of metakaolinite. The theoretical m values derived from these
nominal compositions assume that the polymer network is composed entirely of
SiO4 and AlO4 tetrahedra. The deviation in m value between the observed and
theoretical trends of Fig. 7a implies that the Al content of the AIP is higher than
 Bonding Character in Aluminosilicate Inorganic Polymers 677

what is possible according to the nominal sample composition, suggesting that

other factors such as the Na content of the AIP may be affecting the depen-
dence of 29Si chemical shift on the calculated coordination number. One prospec-
tive mechanism for this effect is that some Na is being incorporated into the
AIP network as a SiO#(Na!)Si moiety, thus altering (reducing) the number
of SiO4 tetrahedra directly bonded to the reference SiO4 tetrahedron. The conse-
quence of Na creating an altered Q4 Si species of this type may invoke a down-
field shift of the 29Si resonance, in similar fashion to having an electropositive
next-nearest Al neighbor, subsequently inflating the apparent Al content. This
behavior mirrors the previously reported growth of calcium aluminosilicate poly-
mer networks in glasses through the formation of SiOCa linkages[51].
For samples 1.1/0.6, 1.1/0.8 and 1.1/1.0 a narrow resonance situated at about
#89.4 to #89.7 ppm is observed in both the 29Si MAS and CPMAS spectra,
indicating the presence of an ordered second phase through zeolite formation.
Previously reported XRD analyses[52] revealed that zeolite A is the preferen-
tially formed ordered phase occurring from these nominal compositions (Linde
type A zeolite [53]). The extent of this zeolite formation was measured from
the deconvolution and simulation of the 29Si MAS data which were acquired
under quantitative conditions (see Fig. 4a). Within these preparations this addi-
tional component is estimated as 1.1/0.6 (6.8 w/w%), 1.1/0.8 (7.1 w/w%) and
1.1/1.0 (45 w/w%). It is interesting to observe that no evidence of zeolite for-
mation was detected in sample 2.0/2.0 by either 29Si MAS or CPMAS NMR,
even though small amounts of poorly formed zeolite A were detected by XRD
studies [52]. This notable absence suggests that under the formation conditions
of this zeolite component within sample 2.0/2.0, the resolution and intensity af-
forded by this phase is insufficient for detection despite the fact that this sample
composition resides on the edge of the zeolite formation region. The 29Si MAS
NMR line widths of these AIPs range between 0.78 and 1.3 kHz, (i.e., about
ten times the width of crystalline kaolinite), indicating that the Si environment
in these preparations is amorphous. These line widths are distributed into two
distinct groups; the AIP formulations that support zeolite formation exhibit line
widths in the range of 0.780.84 kHz, whereas samples not bearing any zeolite
content exhibit broader line widths of 1.21.3 kHz. This bimodal distribution sug-
gests that any zeolite present in the material modifies the development of the
AIP network. Depending on the distribution of the zeolite crystals, the zeolite
could act as a pinning center, thus assisting the AIP network formation from
becoming increasingly disordered. A series of well-distributed crystallites would
continuously interrupt the development of the AIP structure and could act as a
template which would serve to modify the AIP structure over several bond lengths
as the aluminosilicate changes from a zeolite to an AIP.
As noted in previous 29Si MAS and CPMAS studies of AIPs[14], the corre-
sponding data of Figs. 2a, 2b and 3 demonstrate that under typical 1H-29Si CP
experimental settings (i.e., contact period about 25 ms, recycle delay about 2
10 s) a significant signal decrease is observed with the implementation of the
CPMAS technique in comparison with single-pulse methods. Even though typi-
678 M.R.Rowles et al.

cal H2O contents of the as-synthesized AIPs reside within a 2035 w% range
[8, 14], the resulting high 1H-concentration alone is insufficient to produce a
sensitivity enhancement. From Fig. 2a and b it is observed that the AIP 29Si signal
intensity is more severely affected than the zeolite component present in some
of the AIP preparations; in such cases the influence of thermal motion on the
heteronuclear 1H-29Si spin decoupling process has to be considered. Theoretical
treatments of this problem predict that insufficient thermal averaging coupled with
the (typically) inefficient heteronuclear decoupling field strengths implemented
on systems exhibiting motional correlation times (tc) in these regimes render the
line broadening to be unmanageable and high-resolution data of the rare (29Si)
spins cannot be acquired [32, 54]. In an analogous manner, the fast 1H MAS
spectra of Fig. 2d demonstrate that the 1H line widths of all the AIPs are far
broader than those of any zeolite component; although chemical shift dispersion
is a small contribution to this observation, this phenomenon is dominated by the
very different T2s characterizing these systems, thus indicating that the H2O mo-
bility in the ordered zeolite channels is far more rapid than that displayed by
the occluded H2O trapped in the disordered pores of the AIP network. In this
case, the fast 1H MAS experiment is much more efficient in removing any re-
sidual homonuclear 1H1H dipolar interactions in situations where rapid H2O
mobility is already modulating a large portion of this interaction (i.e., in the
zeolite cases), and it is less effective as a line-narrowing technique when it is
dealing with much larger residual 1H1H dipolar interactions in the AIP systems
where the H2O mobility is slow.
The variable contact time TSiH/T1r data of Fig. 3 and Table 2 are consistent
with these observations. Irrespective of AIP nominal composition, the measured
TSiH values fall in a narrow range of about 0.270.35 ms, which is approximately
an order of magnitude shorter than the corresponding TSiH values of about 1.47
2.11 ms measured for the zeolite A components of samples 1.1/0.6, 1.1/0.8 and
1.1/1.0. Furthermore, from this short TSiH behavior, the optimum contact period
that should be adopted for CP signal enhancement appears to be about 1 ms,
and this is very consistent across the entire AIP range demonstrating that the
heteronuclear 1H29Si dipolar interaction facilitating the HartmannHahn contact
in the CP experiment is much stronger in the AIPs where the restricted H2O
mobility is present. Additionally, the corresponding AIP T1r values of about 2.71
6.03 ms are similarly short, indicating that an efficient spinlattice relaxation
mechanism exists. This contrasts markedly from the zeolite cases where, over
the contact time range employed in this study, the T1r measurements were un-
able to be determined as the very slow magnetization increase via the TSiH mecha-
nism is dominating an even slower T1r relaxation mechanism. A further contrast
in H2O mobility characteristics, and subsequent heteronuclear 1H-29Si dipolar
coupling strengths, between the AIP and zeolite systems is emphasized when
the temperature of the variable contact experiment is reduced from ambient tem-
peratures (21 C) to #40 C. For the AIP case, the TSiH/T1r characteristics re-
main largely invariant over this temperature range, whereas the zeolite TSiHs and
T1rs reduce markedly to the ranges of about 0.241.00 ms and about 6.95 and
 Bonding Character in Aluminosilicate Inorganic Polymers 679

16.8 ms, respectively, thus demonstrating an increasing strength of the 1H-29Si

dipolar characteristics through reduced mobility. The general relaxation behavior
of these AIP systems appears to be dominated by a dipolar mechanism with the
characteristic motional tc values clustering around the intermediate T1 minimum/T2
plateau regime where both relaxation processes are fast and T1<T2[32, 54]. Al-
though the 1H T1s become very short, this property cannot be fully exploited as
this motional regime is incompatible with CP dynamics and heteronuclear dipo-
lar decoupling processes as mentioned above. Quick checks of the 1H T1 relax-
ation at 8.45 T (21 and #40 C) and at 14.1 T (both not shown) demonstrate
that these T1s are under 0.2 s.
Rahier et al. [28] reported that their 29Si MAS NMR data were invariant with
changes in the Si-to-Al ratio, and that samples synthesized with Na-to-Al ratios
exceeding stoichiometric values (Na:Al"1) showed silicate resonances in addi-
tion to the AIP resonance. These resonances disappeared after washing, indicat-
ing that they were due to excess Na silicate in the activating solution. In con-
trast to the Rahier study [28], the present study has shown that the 29Si spectra
change systematically when the Si-to-Al ratio is altered, and that no Na silicate
peaks were present in any of the AIP formulations. The method of sample prepa-
ration via creation of a sodium silicate solution adopted by Rahier et al.[28]
was closely followed in this study; however, there were differences in the cur-
ing methods that were undertaken. Their curing practice employed the use of a
closed mould at room temperature for at least two days, while the sample cur-
ing method adopted in this study [10] involved the use of similar closed moulds
at an elevated temperature of 75 C. The fundamental differences in the 29Si MAS
NMR data are ascribed to the different curing methods employed, thus empha-
sizing that the production method can profoundly affect the chemical, microstruc-
tural and physical properties exhibited by these materials.

3.2 27
Al MAS NMR Study

In Fig. 2c the 27Al MAS spectrum of crystalline kaolinite shows a single reso-
nance characterized by an uncorrected 27Al shift of #0.01 ppm (at 9.4 T) and a
line width of 1.05 kHz, indicating the presence of an ordered AlVI species[39,
55]. A detailed variable B0 field analysis of kaolinite shows that there is a con-
siderable second-order quadrupolar contribution to this line width and that the
apparent 27Al shifts are very field-dependent, with an isotropic chemical shift (diso)
of 8.0 ppm being reported [39]. In contrast, the highly disordered metakaolinite
precursor material exhibits three broader resonances at 53.6, 29.1 and 1.6 ppm
(elucidated via spectral simulation, see Fig. 4f), which are assigned to AlIV, AlV
and AlVI coordination environments, respectively [56]. These resonances exhibit
larger line widths in the range of 2.25.4 kHz. From this deconvolution the
relative percentages of these AlIV, AlV and AlVI coordination environments com-
posing the metakaolinite precursor are estimated to be about 10%, 34% and about
56%, respectively. The multiple coordination environments and amorphous na-
680 M.R.Rowles et al.

ture of metakaolinite enhances the reactivity of the material and initiates the
polymerization reaction[57].
The 27Al MAS NMR data for the entire suite of AIPs is dominated by a
single resonance exhibiting an apparent uncorrected shift in the 54.558.5 ppm
range suggesting a complete (or near complete) transformation under high pH
conditions from the Al distributions encountered in the metakaolinite bulk mate-
rial to only AlIV speciation. A vertical expansion of the baseline of these data
(not shown) indicates that very minor contributions of AlV and AlVI are evident
which correspond to the presence of residual unreacted metakaolinite, as sup-
ported by the 29Si MAS data. This is most apparent in samples 1.1/0.6, 1.5/0.8
and 3.0/1.5. From initial observations of the 29Si and 27Al MAS data in Figs.
2a, c and 4b there appears to be some discrepancy in the amount of unreacted
metakaolinite detected by these experiments. As reported above, the deconvolution
of the 29Si MAS data accurately measures the unreacted metakaolinite content
to be the order of about 2040% (see Fig. 7b), whereas similar deconvolution
of the corresponding 27Al MAS data (at best) reliably estimates a reduced pres-
ence in only some of these samples (see Table 3). For example, the 29Si MAS
spectrum of sample 1.5/0.8 predicts that about 30% of the measured signal in-
tensity is due to residual metakaolinite; however, the AlV and AlVI resonances in
the corresponding 27Al MAS spectrum only account for about 5% of that mea-
sured total intensity. To investigate these observations further, NaOH-activated
metakaolinite has been studied without any further stoichiometric adjustment with
sodium silicate solutions. In these cases, the effects of metakaolinite exposure
to variable-strength NaOH (i.e., changing pH) for different time periods are ob-
served, as an analogue for incompletely mixed or inhomogeneous portions of an
AIP preparation. The results of this simplified system are shown in Fig. 5, and
two conclusions can be drawn. Firstly, for certain combinations of time and pH,
the evolution of Al speciation to AlIV appears to be more rapid and detectable
than the concomitant transformation in the 29Si network to the sodalite-type struc-
tures. This is particularly evident during the initial stages of the metakaolinite
activation and is clearly highlighted by the case of metakaolinite exposure to 4 M
NaOH for 6 h within Fig. 5. From this observation it could be inferred that all
residual metakaolinite in the total suite of AIPs has altered from its initial state
as a precursor. Secondly, any speciation transformation from highly distorted,

Table 3. Estimated levels of Al present in samples according to coordination number.

Sample Al(IV))2 (%) Al(V))2 (%) Al(VI))2 (%)

Metakaolinite 10 34 56
1.1/0.6 69 06 25
1.5/0.8 95 00 05
2.0/1.0 98 00 02
2.5/1.3 98 00 02
3.0/1.5 59 12 29
 Bonding Character in Aluminosilicate Inorganic Polymers 681

metastable AlV and AlVI environments can produce 27Al resonances with exces-
sively large line widths [30] which are characterized by quadrupole coupling
constants (CQs) of over 15 MHz. These very broad resonances will not be ad-
equately narrowed and sampled with MAS rates of about 1516 kHz at 9.4 T and
may be acutely underestimated. These systems require more extensive investiga-
tion with fast MAS at higher fields, coupled with a more rigorous treatment of
the spectral simulation and associated satellite contributions, to reliably quantify
Al speciation characterized by a very large range of CQ values[30, 39, 58].
An important observation of these 27Al MAS data is that the measured line
widths of these AIP spectra at 9.4 T lie within the range of about 0.61.6 kHz.
From Table 2 it is clear that the narrower 27Al line widths of about 0.61.0
belong to samples 1.1/0.6, 1.1/0.8 and 1.1/1.0 possessing large zeolite A con-
tributions. Hence, the majority of the AIP suite without additional zeolite com-
ponents exhibit line widths in a more representative (reduced) range of about
1.01.6 kHz, which is very comparable to that observed for crystalline kaolin-
ite. From this result it can be inferred that some degree of short-range order
exists within the AlIV positions of the AIP network. In similar fashion to the
kaolinite resonance, the 27Al AIP resonances have a second-order quadrupolar
contribution to their line widths as evidenced by the very small extended tail-
ing on the upfield side of each line shape, which indicates that only minor
distributions within diso and CQ are influencing the width and shape of these
resonances [5962], and that the AlO4 tetrahedral geometry is largely unper-
turbed. This contrasts markedly with the 29Si MAS data which describes these
systems as being highly disordered, where there is over an order of magnitude
difference between the line widths observed for kaolinite and the AIPs (see
Table 2). The disparity between the structural descriptions reflected by the 27Al
and 29Si MAS data implies that the short- and long-range structural disorder
throughout the AIP network is mainly propagated through large distributions
of bond angles and bond lengths about the Si positions only. A structural de-
scription of this nature is consistent with the pathway for AIP formation (Si-
to-Al molar ratio, 2:1) proposed in Sect. 1, where the formation of a polysialate-
siloxo polymer possessing additional siloxo (SiOSi) linkages between the
sialate (SiOAl) moieties could promote this type of partitioned disorder.

3.3 23
Na MAS NMR Study

The conventional single-pulse 23Na MAS NMR data that were acquired at mag-
netic field strengths of 8.45, 9.4 and 14.1 T are shown in Fig. 6. The corre-
sponding 2-D MQMAS data acquired at 8.45 and 14.1 T are given in Fig. 8.
From Figs. 6 and 8 it is quite evident that straightforward homogeneous Na
speciation does not exist throughout a large proportion of the nominal composi-
tions studied. A close inspection of the variable field spectra of Fig. 6 reveals
that the line widths decrease with increasing B0, and that the apparent shifts move
to lower field, thus converging on the true 23Na isotropic chemical shift diso,
682 M.R.Rowles et al.

indicating that a substantial second-order quadrupolar contribution to the line

width of these 23Na resonances exists.
The dominant 23Na resonance observed across the suite of AIP preparations
is a broad featureless line shape that occurs at an apparent shift of about 5 ppm.
Very little information describing the nature the Na speciation can be ascertained
from these 1-D data except that the majority of Na in each sample resides in
disordered environments similar to those experienced in glass systems. The 2-D
MQMAS data are more informative, showing contours that appear skewed in
specific directions depending on the applied magnetic field strength. The 2-D
MQMAS data acquired at 14.1 T exhibits contours that are predominantly
skewed at about 60 to the conventional isotropic chemical shift axis, i.e., along
the chemical shift distribution axis [43, 63]. Hence, the MQMAS data at this
higher field indicate that a significant distribution of isotropic chemical shifts (diso)
contribute to this broad inhomogeneous resonance. In contrast, the 2-D MQMAS
data acquired at the lower field of 8.45 T emphasize more prominently the
quadrupolar contribution to these 23Na line widths in comparison with that ex-
perienced at 14.1 T. Figure 8 illustrates that these 8.45 T contours have altered
to reflect a predominant skewing at about #50 along the quadrupolar distribu-
tion axis[43, 63], thus indicating that a substantial distribution in 23Na quadru-
polar parameters is also induced by this structural disorder. Both 8.45 and 14.1
T MQMAS data sets are represented with more clarity when shorter relaxation
delays are employed, implying that the Na species represented by these broad
inhomogeneous resonances have shorter T1 relaxation characteristics in compari-
son with other spectral components that may also exist. Although the apparent
shifts for these broad resonances occur at about 510 ppm (depending on B0),
the true isotropic chemical shift (diso) will be located further downfield at about
0 ppm. This evidence suggests that the broad resonance represents hydrated Na!
ions in the pore water and void spaces within the AIP network, and the distri-
butions described above represent actual variations in the hydration state of each
solvated Na! ion and size distributions of the occupied void spaces where these
hydrated species reside.
For samples with nominal compositions reflecting an excess Na content (i.e.,
2.0/2.0), Figs. 6 and 8 show that a prominent field-dependent resonance at about
#5 ppm is observed. In these preparations the Na speciation is dominated by
extraneous Na2CO3 and NaHCO3 which is not formally incorporated into the AIP
network. Unambiguous evidence of this assignment can be elucidated from both
the 1-D and 2-D data of Figs. 6 and 8, and from the prominent HCO3# reso-
nance at 18.1 ppm in the 1H fast MAS data of Fig. 2d. A comparison of the
23
Na MAS spectra of anhydrous Na2CO3 at each field with that from the 2.0/2.0
sample (see Fig. 6a, b and c) shows that an excellent correlation with the down-
field resonance at about #5 ppm is observed. Furthermore, this comparison shows
that a large portion of the much broader 23Na resonance centered about 5 ppm,
which is associated with the AIP network, is also part of the total Na2CO3 spec-
trum. These second-order-broadened quadrupolar resonances attributed to Na2CO3
narrow markedly with increasing B0; however, their line shapes are not well-de-
fined. Since the dominant room temperature form (the g polymorph) shows an
 Bonding Character in Aluminosilicate Inorganic Polymers 683

8.45 T 14.09 T 8.45 T 14.09 T

a 1.1/0.6 e 1.1/0.8
20 20
resolution

20 20
0 0
ppm

0 0

20 20 20 20
20 20
20 20
0
signal

0
ppm

0 0

20 20 20 20
20 0 20 40 60 0 20 20 0 204060 0 20

b 2.0/1.0 f 3.0/2.0

20 20 expt.

expt.
20 sim. sim.
resolution

0 0
ppm

0 expt.
expt.
20 20 sim. 20 sim.
20 20
20 20
0 0
signal

ppm

0 0

20 20 20 20
20 0 20 40 60 0 20 20 0 20 40 60 0 20

c 2.5 / 1.3 g 1.1/1.0

expt. 20 expt. 20

sim. 20
resolution

sim.
0 0
expt. 0
expt.
sim. 20 sim. 20 20
20 20
20 20
0
signal

0
ppm

0 0

20 20 20 20
20 0 204060 0 20 20 0 20 40 60 0 20

d 3.0/1.5 h 2.0/2.0

20 20
resolution

20
expt. 0 expt. 0
0
sim. sim.
20 20 20

20 20
20 20
0 0
ppm
signal

0 0
20 20 20 20
20 0 20 40 60 0 20 20 0 20 40 60 0 20
ppm ppm ppm ppm

Fig. 8. 23Na 2-D MQMAS data for the entire AIP suite acquired at 8.45 and 14.09 T. At each field,
long recycle delay (6 s) experiments focussing on more highly resolved components with slower T1
relaxation, and short recycle (1 s) experiments focussing on broader components with faster T1 re-
laxation, were acquired.
684 M.R.Rowles et al.

incommensurate structure[64], some degree of structural disorder is inherent; in

addition, there is the possibility that some hydrated variants of Na2CO3 may
coexist with NaHCO3, thus creating a mixture of phases. From Fig. 8h the 2-D
MQMAS data from this sample clearly distinguishes the 23Na narrower down-
field resonance at each field and in both the long- and short-relaxation forms of
the experiment. In addition, the long-relaxation experiment at 8.45 T also resolves
(via T1 discrimination) the broader (i.e., larger CQ) site at the expense of obser-
vation of the remainder of the total AIP Na speciation. These observations are
in agreement with previous studies utilizing 23Na MQMAS detection of Na2CO3
impurities in NaHCO3 systems upon hydrolysis and carbonation[65].
In addition to the broad dominant 23Na resonance described above, Figs. 6
and 8 illustrate that additional Na species can also be detected. These are
characterized by narrower resonances superimposed upon the broad dominant
AIP bulk 23Na resonance, with apparent shifts in the range of about 05 ppm.
Furthermore, these species exhibit distinct second-order quadrupolar structure and
much longer T1 characteristics. In particular, the 1-D 23Na MAS data reported in
Fig. 6 for the high Si content AIP preparations 2.5/1.3, 3.0/1.5 and 3.0/2.0 show
that these additional components are quite prominent. A study of the correspond-
ing 2-D MQMAS data from Fig. 8 further resolves these species from that of
the AIP bulk. For samples 2.5/1.3 and 3.0/2.0 the MQMAS data reveals that two
additional Na moieties exist, both of which appear to exhibit second-order quadru-
polar structure and diso values close to 0 ppm. The extraction of the slices par-
allel with the F2 (or anisotropic MAS) dimension shown on Fig. 8c and f con-
firms that clearly defined second-order quadrupolar broadened resonances are
observed at both fields. From these 23Na subspectra, and the pronounced B0 field
dependence exhibited by them, line shape simulation yields values of diso, CQ
and the asymmetry parameter (h) with reasonable accuracy. One resonance is
characterized by the parameters diso"0.0)0.5 ppm, CQ"1.35)0.05 MHz and
h"0.22)0.1, while the other exhibits the following characteristics:
diso"2.0)0.5 ppm, CQ"1.6)0.1 MHz and h"0.9)0.1. Figure 8d shows
that only the latter type of Na species is evident in sample 3.0/1.5.
The appearance of such line shapes further verifies that some degree of short-
range order is present within certain structural components within the AIP net-
work corroborating the findings reflected by the 27Al MAS NMR studies. It is
important to note that these high Si content AIPs are far removed from the
nominal composition regime that defines the low Si content preparations such
as 1.1/1.0, 1.1/0.8 and 1.1/0.6, which are prone to concomitant zeolite A crys-
tallization, as evidenced by XRD studies [52] and by the absence of crystalline
phases in the 29Si MAS NMR. Hence, these 23Na resonances describing some
degree of short-range order cannot be ascribed to zeolite formation.

3.4 Proposed Structural Models

The AIP structural models presented by Barbosa et al. [31], Davidovits [66] and
Li[67] refer to the polymer network as being constructed of SiO4 and AlO4 tet-
 Bonding Character in Aluminosilicate Inorganic Polymers 685

rahedra, and that the Na! charge-balancing cation is present in cagelike struc-
tures in the polymer. Davidovits [66] and Li [67] further associate the Na cat-
ion directly with AlO4 tetrahedra, although no mention is made of how the Na
is structurally incorporated. The cavities shown in the structural model of Barbosa
et al. [31] suggest that Na is present solely in a hydrated form. The data pre-
sented in this current study shows that Na is present in a range of hydrated states
and sometimes shows short-range order, which suggests that the role of Na solely
as a charge-balancing cation is understating its purpose and that it may be ac-
tive in forming and modifying the polymer network.
As mentioned above, the major component of network disorder in these AIPs
is generated by large distributions of bonding characteristics associated with the
Si positions, while the AlO4 tetrahedra remain largely unperturbed. While there
is no question that the bulk of the Na speciation resides within various hydra-
tion states of Na! ions within the pore water, the notion of short-range order
can be introduced through different modes of Na interaction with specific frame-
work segments. As demonstrated by the 23Na MAS NMR data, this appears most
prevalent with high Si content preparations. In situations where sufficient OH#
is present and able to interact with the cavity structure, surface-bound Na alu-
minate moieties such as the structure proposed in Fig. 9a are possible and would
be consistent with the lower h line shape at diso"0.0)0.5 ppm (CQ"1.35)0.05
MHz, h"0.22)0.1) observed in Fig. 8c and f. Similar Na aluminate-type struc-
tures have been observed to form as quasistable intermediates in the acid hydrolysis
of high pH aluminate solutions during the polymerization and eventual condensa-
tion reactions to form gibbsite [68, 69]. For situations where OH# is not directly
involved or accessible, Na may associate with siliceous parts of the framework to
form SiO#(Na!)Si moieties, as depicted in Fig. 9b. In this case it is proposed
that the net negative charge from AlO4 tetrahedral units would become localized
on the bridging O atoms thus forming SiO#Si intermediate, which would
subsequently stimulate an ionic association with Na! ions to stabilize this type
of framework structure. The consistent appearance of a higher h line shape at
diso"2.0)0.5 ppm (CQ"1.6)0.1 MHz, h"0.9)0.1) in high Si content
AIPs (see Fig. 8c, d and f) could be rationalized with this type of short-range
order. The formation of both of these proposed short-range order species poten-
tially accounts for the discrepancy between the observed and theoretical behav-
ior of m (see Fig. 7a) characterizing the degree of next-nearest-neighbor Al oc-
cupancy (i.e., Q4(mAl)) for the Si speciation. In addition, these proposed struc-
tures exist without inducing substantial network termination and concomitant
sample degradation. It is interesting to note that for this suite of AIPs, the su-
perior compressive strength performance was displayed by sample 2.5/1.3, which
exhibits both forms of these proposed short-range order species. This observa-
tion is consistent with improved strength and hardness properties in materials such
as glass ceramics which are achieved by the promotion of ordered microdomains
formed during the quenching and cooling stages of preparation [7073].
In light of these propositions, an improved structural model (as depicted
in Fig. 9c) is proposed for AIP polymerization and formation from an NaOH-
activated metakaolinite precursor approach. It incorporates all of the structural
686 M.R.Rowles et al.

a b

Fig. 9. a Proposed model for the location of the charge-balancing Na! cation which assumes the
form of a semiattached Na aluminate species. b Proposed model for the location of the charge-bal-
ancing Na! cation which assumes the form of a modified bridging network. c Proposed schematic
model for AIP structure based on an original model by Barbosa et al. [31] but which attempts to
incorporate the new elements mentioned above.

elements discussed above and it acknowledges that aspects of previously pro-

posed models are still valid. From this study it is evident that the initial nominal
composition of each AIP is critical in dictating the structural speciation and
the physical properties displayed by these materials, and that any proposed
model needs to be specific to that type of preparation.

4 Conclusion

The use of the multinuclear MAS NMR technique for the systematic study of a
suite of AIP compositions has provided valuable insights into the physical and
structural aspects of AIP chemistry and mechanical performance. Each of the nu-
clei involved (29Si, 27Al, 1H and 23Na) allows quite different aspects of the AIP
network to be investigated. Of primary importance is the observation that, for a
systematic suite of AIP preparations of known composition, the 29Si chemical shift
is sensitive to the changing nature of the bulk network, irrespective of whether
other phases such as zeolites and metakaolinite are present. The systematic varia-
tion of the 29Si chemical shift is able to be directly correlated with compressive
strength properties and the overall Q4(mAl) coordination of the Si. Fast 1H MAS
and 29Si TSiH/T1r relaxation measurements have demonstrated that occluded pore
 Bonding Character in Aluminosilicate Inorganic Polymers 687

H2O mobility within the disordered cavities is slow in comparison to H2O mo-
bility characteristics observed within the ordered channel structures of zeolites.
This leads to much stronger homonuclear 1H1H and heternonuclear 1H29Si di-
polar bound systems as evidenced by the larger residual line widths in the 1H
MAS data and the very short TSiH and T1r relaxation times characterizing the
AIP systems, in comparison with corresponding measurements of ordered zeo-
lite frameworks. Furthermore, the insensitivity of the relaxation measurements
to temperatures in the range of 21 to #40 C supports these observations. In
contrast to the 29Si MAS studies, the corresponding 27Al data suggest that less
disorder exists about the AlO4 structural units in comparison to what is propa-
gated through the AIP network by disorder in SiO bond lengths and angles.
This is a fundamental feature of the AIP network. The different estimations of
unreacted metakaolinite as measured by the 29Si MAS and 27Al MAS experiments
suggests that Al transformation from AlV and AlVI to AlIV species is more rapid
than overall Si transformation, at least for the early stages of metakaolinite ac-
tivation. Finally, the 23Na 1-D and 2-D MAS data clearly show that different types
of Na species can be observed, ranging from highly disordered hydration states
to proposed elements of short-range order involving structural aspects of the
framework. The presence of these newly proposed Na species can be rational-
ized with the measured disparity between the observed and theoretical behavior
of the parameter m characterizing the degree of next-nearest-neighbor Al occu-
pancy describing Q4(mAl) Si speciation. Most importantly, they are not associ-
ated with any form of a network termination process that would facilitate the
degradation of the AIP physical properties; these newly proposed Na species are
most readily observed in preparations exhibiting superior compressive strength
performance. From these many facets of structural information, an improved
model of a typical AIP network is proposed.

Acknowledgments

M.R.R. acknowledges receipt of an Australian Postgraduate Award scholarship

and an Australian Institute of Nuclear Science and Engineering (AINSE) Post-
Graduate Research Award. J.V.H. acknowledges the funding of AINSE project
02/189. M.E.S. thanks Engineering and Physical Sciences Research Council and
the University of Warwick for funding NMR equipment.

References

1. Davidovits, J.: J. Therm. Anal. 35, 429441 (1989)

2. Palomo, A., Grutzeck, M.W., Blanco, M.T.: Cem. Concr. Res. 29, 13231329 (1999)
3. van Jaarsveld, J.G.S., van Deventer, J.S.J., Lorenzen, L.: Miner. Eng. 10, 659669 (1997)
4. Allahverdi, A., Skvra, F.: Ceram. Silik. 45, 8188 (2001)
5. Roy, D.M.: Cem. Concr. Res. 29, 249254 (1999)
6. Shi, C., Stegemann, J.A.: Cem. Concr. Res. 30, 803808 (2000)
7. van Jaarsveld, J.G.S., van Deventer, J.S.J., Schwartzman, A.: Miner. Eng. 12, 7591 (1999)
688 M.R.Rowles et al.

8. Perera, D.S., Blackford, M.G., Vance, E.R., Hanna, J.V., Finnie, K.S., Nicholson, C.L., in:
Hanchar, J.M., Stroes-Gascoyne, S., Browning, L. (eds.) MRS Proceedings Scientific Basis
for Nuclear Waste Management XXVIII, vol. 824, pp. 607612. Materials Research Society,
Warrendale (2004)
9. Lyon, R.E., Balaguru, P.N., Foden, A., Sorathia, U., Davidovits, J., Davidovics, M.: Fire Mater.
21, 6773 (1997)
10. Rowles, M., OConnor, B.: J. Mater. Chem. 13, 11611165 (2003)
11. Sabir, B.B., Wild, S., Bai, J.: Cem. Concr. Compos. 23, 441454 (2001)
12. Worrell, E., Price, L., Martin, N., Hendriks, C., Meida, L.O.: Annu. Rev. Energy Environ. 26,
303329 (2001)
13. Davidovits, J.: World Resour. Rev. 6, 263278 (1994)
14. Blackford, M.G., Hanna, J.V., Pike, K.J., Vance, E.R., Perera, D.S.: J. Am. Ceram. Soc. 90, 1193
1199 (2007)
15. Zosin, A.P., Priimak, T.I., Avsaragov, K.B.: Atom. Energy 85, 510514 (1998)
16. Khalil, M.Y., Merz, E.: J. Nucl. Mater. 211, 141148 (1994)
17. Perera, D.S., Vance, E.R., Aly, Z., Davis, J., Nicholson, C.L.: Ceram. Trans. 176, 9196 (2006)
18. ASTM Committee C26, Standard Test Methods for Determining Chemical Durability of Nuclear,
Hazardous, and Mixed Waste Glasses and Multiphase Glass Ceramics: The Product Consistency
Test. ASTM, Philadelphia (2000)
19. Alonso, S., Palomo, A.: Mater. Lett. 47, 5562 (2001)
20. Moulin, E., Blanc, P., Sorrentino, D.: Cem. Concr. Compos. 23, 463469 (2001)
21. Davidovits, J., Davidovits, M., Davidovits, N.: Process for Obtaining a Geopolymeric Alumino-
Silicate and Products Thus Obtained. US patent 5342595, 1994.
22. Wong, Y.L., Lam, L., Poon, C.S., Zhou, F.P.: Cem. Concr. Res. 29, 19051913 (1999)
23. Shehata, M.H., Thomas, M.D.A.: Cem. Concr. Res. 30, 10631072 (2000)
24. Fan, Y., Yin, S., Wen, Z., Zhong, J.: Cem. Concr. Res. 29, 467472 (1999)
25. Pereira, D.A., Aguiar, B.D., Castro, F., Almeida, M.F., Labrincha, J.A.: Cem. Concr. Res. 30, 1131
1138 (2000)
26. Lam, L., Wong, Y.L., Poon, C.S.: Cem. Concr. Res. 28, 271283 (1998)
27. Rahier, H., Simons, W., van Mele, B., Biesemans, M.: J. Mater. Sci. 32, 22372247 (1997)
28. Rahier, H., van Mele, B., Biesemans, M., Wastiels, J., Wu, X.: J. Mater. Sci. 31, 7179 (1996)
29. Rahier H., van Mele, B., Wastiels, J.: J. Mater. Sci. 31, 8085 (1996)
30. MacKenzie, K.J.D., Smith, M.E.: Multinuclear Solid-State Nuclear Magnetic Resonance of Inor-
ganic Materials. Pergamon, London (2002)
31. Barbosa, V.F.F., MacKenzie, K.J.D., Thaumaturgo, C.: Int. J. Inorg. Mater. 2, 309317 (2000)
32. Engelhardt, G., Michel, D.: High-Resolution Solid-State NMR of Silicates and Zeolites. Wiley,
Chichester (1987)
33. Duxon, P., Fernanez-Jiminez, A., Provis, J.L., Luckey, G.C., Palomo, A., van Deventer, S.J.: J.
Mater. Sci. 42, 29172973 (2007)
34. Brew, D.R.M., MacKenzie, K.J.D.: J. Mater. Sci. 42, 39903993 (2007)
35. Lippmaa, E., Mgi, M., Samoson, A., Tarmak, M., Engelhardt, G.: J. Am. Chem. Soc. 103, 4992
4996 (1981)
36. Nofz, M., Engelhardt, G., Wihsmann, F.-G., Forkel, K., Mgi, M., Lippmaa, E.: Z. Chem. 26,
221222 (1986)
37. Lippmaa, E., Samoson, A., Mgi, M.: J. Am. Chem. Soc. 108, 17301735 (1986)
38. Loewenstein, W.: Am. Miner. 39, 9296 (1954)
39. Smith, M.E.: Appl. Magn. Reson. 4, 164 (1993)
40. Koller, H., Engelhardt, G., Kentgens, A.P.M., Sauer, J.: J. Phys. Chem. 98, 15441551 (1994)
41. Murat, M.: Cem. Concr. Res. 13, 511518 (1983)
42. Brown, S.P., Wimperis, S.: J. Magn. Reson. 128, 4261 (1997)
43. Fernandez, C., Amoureux, J.P.: Chem. Phys. Lett. 242, 449454 (1995)
44. Medek, A., Harwood, J.S., Frydman, L.: J. Am. Chem. Soc. 117, 1277912787 (1995)
45. Frydman, L., Harwood, J.S.: J. Am. Chem. Soc. 117, 53675368 (1995)
46. Amoureux, J.-P., Fernandez, C., Steuernagel, S.: J. Magn. Reson. 123, 116118 (1996)
47. Krumm, S.: Winfit, ver: 1.2.1, 1997. www.geol.uni-erlangen.de
48. Sherriff, B.L., Grundy, H.D., Hartmann, J.S., Hawthorne, F.C., Cerny, P.: Can. Miner. 29, 271
285 (1991)
 Bonding Character in Aluminosilicate Inorganic Polymers 689

49. Sherriff, B.L., Grundy, H.D., Hartmann, J.S.: Eur. J. Miner. 3, 751768 (1991)
50. Mgi, M., Lippma, E., Samoson, A.: J. Phys. Chem. 88, 1518 (1984)
51. Petkov, V., Billinge, S.J.L., Shastri, S.D., Himmel, B.: Phys. Rev. Lett. 85, 34363439 (2000)
52. Rowles, M.R.: Ph.D. thesis, Curtin University of Technology, Curtin, Australia, 2004.
53. IZA Structure Commission: Database of Zeolite Structures, 2004. http://www.iza-structure.org/da-
tabases/
54. Ernst, H., Fenzke, D., Pfiefer, H.: Ann. Phys. 38, 257270 (1981)
55. MacKenzie, K.J.D., Brown, I.W.M., Meinhold, R.H., Bowden, M.E.: J. Am. Ceram. Soc. 68, 293
297 (1985)
56. Rocha, J.: J. Phys. Chem. B 103, 98019804 (1999)
57. Granizo, M.L., Blanco-Varela, M.T., Palomo, A.: J. Mater. Sci. 35, 63096315 (2000)
58. Massiot, D., Bessada, C., Coutres, J.P., Taulelle, F.: J. Magn. Reson. 90, 231242 (1990)
59. Kunath-Fandrei, G., Bastow, T.J., Hall, J.S., Jger, C., Smith, M.E.: J. Phys. Chem. 99, 15138
15141 (1995)
60. Kunath, G., Losso, P., Schneider, H., Steuernagel, S., Jger, C.: Solid State Nucl. Magn. Reson.
1, 261266 (1992)
61. Kohn, S.C., Dupree, R., Mortuza, M.G., Henderson, C.M.B.: Am. Miner. 76, 309312 (1991)
62. Toplis, M.J., Kohn, S.C., Smith, M.E., Poplett, I.J.F.: Am. Miner. 85, 15561560 (2000)
63. Angeli, F., Charpentier, T., Faucon, P., Petit, J.-C.: J. Phys. Chem. B 103, 1035610364 (1999)
64. van Aalst, W., den Holander, J., Peterse, W.J.A.M., de Wolff, P.M.: Acta Crystallogr. B 32, 47
58 (1976)
65. Hanna, J.V., Smith, M.E., Whitfield, H.J.: J. Am. Chem. Soc. 118, 57725777 (1996)
66. Davidovits, J., in: Proceedings of the First International Conference on Alkaline Cements and Con-
cretes, pp. 131149. Kiev State Technical University, Kiev (1994)
67. Li, Z., Ding, Z., Zhang, Y., in: Wang, K. (ed.) Proceedings of the International Workshop on
Sustainable Development and Concrete Technology, pp. 5576. Iowa State University, Ames, Iowa
(2004)
68. Bradley, S.M., Hanna, J.V.: J. Chem. Soc., Chem. Commun. 1993: 12491251 (1993)
69. Bradley, S.M., Hanna, J.V.: J. Am. Chem. Soc. 47, 77717783 (1994)
70. MacMillan, P.W.: Glass Ceramics. Academic Press, London (1964)
71. Meneghini, C., Mobilio, S., Lusvarghi, L., Bondioli, F., Ferrari, A.M., Manfredini, T., Siligardi,
C.: J. Appl. Crystallogr. 37, 890900 (2004)
72. Mandal, S., Chakrabarti, S., Ghatak, S., Das, S.K.: Bull. Mater. Sci. 28, 437443 (2005)
73. Xia, F., Zhang, X., Ren, J., Chen, G., Ma, H., Adam, J.L.: J. Am. Ceram. Soc. 89, 21542157
(2006)

Authors address: John V. Hanna, NMR Facility, Institute of Materials and Engineering Science, Lucas
Heights Research Laboratories, Australian Nuclear Science and Technology Organisation, Private Mail
Bag 1, Menai NSW 2234, Australia
E-mail: [email protected]