Acta Materialia 53 (2005) 4019–4028

www.actamat-journals.com

Transition of tensile deformation behaviors

in ultrafine-grained aluminum
C.Y. Yu 1, P.W. Kao *, C.P. Chang
Institute of Materials Science and Engineering, National Sun Yat-Sen University, # 70 Lian-Hi Road, Kaohsiung 804, Taiwan

Received 2 February 2005; received in revised form 23 April 2005; accepted 3 May 2005
Available online 5 July 2005

Abstract

An evident transition of tensile deformation behaviors appeared in commercial purity aluminum as the grain size reduced from
micrometer to submicrometer range. The transition grain size decreased as the deformation temperature decreased from room tem-
perature to 77 K. A model was proposed to explain the deformation behaviors in ultrafine-grained (UFG) aluminum. As grain size
decreased below a critical value, which may correspond to the cell size present at the applied strain rate and temperature, the mean
free path of dislocation was no longer determined by the dislocation structure, but limited by the grain boundaries. This causes an
enhancement of dynamic recovery rate, and reduction of hardening rate in UFG metals, leading to the presence of yielding peak,
Lüders deformation, and reduced tensile elongation.

2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Ultrafine-grained structure; Mechanical properties; Yielding; Lüders band

1. Introduction boundary sources [11,13] or grain boundary shear pro-

moted by dislocation pile-ups [12]; (c) grain boundary
Nano- and submicrocrystalline materials have processes only [11,12]. The transition grain size from
attracted considerable interest among researchers, be- one regime to another depends on the strain rate and
cause the presence of a large amount of grain boundary temperature. In the present work, we focused on the
area in the materials results in unusual and extraordi- change in tensile behaviors as the grain size reduced
nary changes in both mechanical and physical properties from the traditional to the ultrafine-grain (UFG) regime.
[1–9]. Based upon the dominant deformation mecha- In general, the relationship between yield stress and
nism operating in different grain size (d) ranges, three grain size can be expressed by the well-known Hall–
grain-size regimes may be identified [10,11]: (a) tradi- Petch relation (ry = ro + kd1/2), where ry is the yield
tional (d >
106 m), (b) ultrafine-grain (d
108 to stress, ro is the frictional stress, and k is a constant

106 m), and (c) nanocrystalline (d <
108 m). The [14]. This relationship predicts that the yield stress of
following mechanisms were suggested to be rate-control- materials with submicrometer or nanometer grain size
ling in each regime, respectively: (a) dislocation intersec- can be extrapolated from the Hall–Petch relation ob-
tions [11,12]; (b) dislocation emission from grain tained in the traditional grain size range. But deviation
of the Hall–Petch slope in the submicrometer range
has been reported in various metals. Several studies
*
showed that the Hall–Petch slopes in nanocrystalline
Corresponding author. Tel.: +886 7525 2000; fax: +886 7525 4099.
E-mail address: [email protected] (P.W. Kao).
materials are essentially negative [15–19]. A decreasing
1
Present address: Department of Materials Science and Engineering, slope in Hall–Petch relation was also observed in an
National Tsing Hua University, Hsinchu 300, Taiwan. Al–3%Mg alloy with a grain size smaller than 0.15 lm

1359-6454/$30.00
2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2005.05.005
4020 C.Y. Yu et al. / Acta Materialia 53 (2005) 4019–4028

[20]. However, a weak grain size dependence of the our previous study [47], AA1050 aluminum of grain
Hall–Petch relation, i.e., the Hall–Petch slope ap- size ranging from 0.35 to 45 lm with a small variation
proaches zero, was reported in the submicrometer re- in HAB fraction and texture can be produced by prop-
gion of electroplated Ni [21]. On the other hand, a er combination of ECAE and annealing treatments.
positive deviation of the Hall–Petch slope at small grain Therefore, it is possible to investigate the grain size ef-
size was also observed in several reports [22–25]. fect on the mechanical properties of aluminum over a
In most studies of the Hall–Petch relation, the grain wide size range without the interference of the variation
size is controlled by the use of different annealing pro- of other factors associated with grain growth. In this
cesses, and the boundary structure is assumed to be paper, the change of tensile deformation behaviors in
independent of grain size. However, reports have indi- aluminum as the grain size decreases from the tradi-
cated that the Hall–Petch slope may change significantly tional to the UFG range will be presented and the
as a result of the different grain boundary characters in- deformation mechanisms in the UFG regime will be
volved [20,26,27]. Recently, Sun et al. [28] reported that discussed.
different boundary characters had a strong influence on
the tensile deformation behaviors in UFG aluminum.
Therefore, it is necessary to minimize the variation of 2. Experimental
grain boundary characters with a different grain size
range in the study of the grain size effect on tensile Commercial purity aluminum (AA1050) was stud-
properties. ied in this study. It was homogenized at 873 K for
Several interesting tensile behaviors exhibited by 12 h and air cooled to room temperature. The initial
UFG alloys have also been noted [29–38], such as the grain size of the material was about 330 lm. Speci-
yield-drop phenomenon, Lüders deformation, reduced mens of 12 · 12 · 80 mm were subjected to ECAE at
tensile elongation, and shear banding. The poor ductility room temperature. Before each pass, the specimen
of UFG alloys can be attributed to the lower work hard- was coated with a lubricant containing MoS2. The
ening rate [28,31–34]. The onset of plastic instability die for ECAE consisted of two channels of identical
(necking) during tensile test can be explained based on cross-section intersecting at a die angle U of 90.
the Considère criteria, which can be expressed by The angle defining the outer arc of the two channels
  was 20. Specimen was passed through the die repeti-
or
6 r; ð1Þ tively for eight passages to achieve an equivalent
oe e_
strain
8. The specimen was rotated counter-clock-
where r and e are true flow stress and true strain, respec- wise about the extrusion direction (ED) by 90 be-
tively. The flow stress increases significantly by refine- tween each passage. This process route was
ment of grain size to the submicrometer range, designated as route Bc [48]. After ECAE, the speci-
especially at an early stage of plastic deformation. How- mens were subjected to various annealing treatment
ever, the work hardening rate decreases with decreasing to obtain the desired grain size. The grain size (d) is
grain size. As a result, plastic instability (necking) occurs defined by the equivalent circle diameter, which is cal-
at the very early stage of tensile deformation in the UFG culated by setting the projected area of each grain (A)
materials, which results in limited uniform elongation. A equal to an equivalent area of a circle (A = pd2/4). For
large tension–compression asymmetry in the strength each specimen, more than 500 grains were measured.
was noted in UFG materials [35–37]. Shear bands were The specimens for transmission electron microscopic
observed in UFG Al [35], and Fe [38,39]. According to (TEM) observations were mechanically thinned down to
TEM observations, the microstructures inside the shear about 150 lm thick and finally polished by a standard
bands indicated dislocation-based mechanisms in these twin jet polishing method using an electrolyte of 25% ni-
UFG materials. tric acid and 75% methanol at 243 K and 15 V. The
Ultrafine-grained structure can be introduced into TEM work was carried out in a Philips CM200 micro-
metals through the use of severe plastic deformation scope operated at 200 kV.
(SPD) [40]. One of the popular SPD techniques used re- For materials of fine grain size, the misorientations of
cently is equal channel angular extrusion (ECAE) [41– neighboring grains were determined by the use of TEM
43]. Chang et al. [44] showed that grain boundaries measurement with a computer program, which was
could be induced in UFG aluminum by ECAE. The developed by Zaefferer [49]. The method can provide
UFG structure consisting of a large fraction (>65%) orientation determination with an accuracy better than
of high angle boundaries (HABs) can be developed by 0.1 by the use of Kikuchi patterns. In this work, the
ECAE [45,46]. Humphreys et al. [46] suggested that a Kikuchi patterns was recorded by a CCD camera, and
cellular structure should be stable against discontinuous indexed by a commercial computer program (TOCA,
recrystallization and coarsen uniformly, provided that TSL). The misorientation between neighboring grains
it contains greater than about 65% HABs. Based upon was calculated as ‘‘angle/axis pair’’, where the minimum
 C.Y. Yu et al. / Acta Materialia 53 (2005) 4019–4028 4021

misorientation angle was selected from the 24 equivalent longitudinal and transverse directions, which are al-
solutions. For each specimen, about 150 grains were most free of dislocations, typical of well-annealed
analyzed from five different areas, which contribute to grain structures. Typical microstructures of specimens
about 300 boundaries. For materials of larger grain with submicrometer grain size and traditional grain
sizes, the grain boundary character distribution was size are shown in Fig. 1. Since the grain size distribu-
determined by a field emission scanning electron micro- tion may have a strong effect on the stress–strain
scope (SEM, Jeol JSM-6330TF) equipped with an elec- curve [16], the annealing conditions which resulted in
tron backscatter diffraction (EBSD) measuring system a wide grain size distribution were excluded in this
(Opal, Oxford Instruments). study. For the annealing conditions selected here,
The specimens were machined from the ECAE mate- the size distributions approximate to the typical log–
rials with the tensile axis parallel to the extrusion direc- normal distribution and the largest grains are smaller
tion. The gauge length of the specimen was 14 mm, and than 3· (average grain size). The distribution of
the diameter was 3 mm. All tensile specimens were elec- boundary misorientation shows little change as the
tropolished with an electrolyte of 800 ml methanol, grain size increased from 0.35 to 0.78 lm, which has
140 ml distilled water and 60 ml perchloric acid at 65% high angle boundaries (HABs); here, HABs are
298 K and 25 V. Tensile tests were conducted on an defined as the boundaries that have misorientations
Instron model 5582 universal testing machine using con- higher than 15. As the grain size increases to
stant crosshead speed, which corresponds to an initial 20 lm, the fraction of HABs increases to 80%. The
strain rate of 7.1 · 104 s1. distributions of boundary misorientation for selected
specimens of different grain sizes are shown in Fig. 2.

3. Results
3.2. Tensile behaviors at room temperature
3.1. Microstructure
The tensile stress–strain (r–e) curves depend strongly
For the as-ECAE condition, there are dislocation on the grain size and testing temperature. In general, the
structures in some grains even though most of the r–e curves obtained in this study could be categorized
grains are free of dislocations in their interior, and into four different characteristic types.
the average grain size is 0.35 lm. The grains have
an equiaxed shape in the transverse section; however, Type I: The curve shows a limited uniform elongation
they are mixture of both elongated and equiaxed followed by strain-softening.
shape in the longitudinal section. By the use of proper Type II: The curve exhibits a distinct yielding peak fol-
annealing treatment, grain sizes in the range of 0.4– lowed by strain-softening.
45 lm were obtained from ECAE processed alumi- Type III: The curve shows Lüders strain after the onset
num. For specimens with a grain size of 0.4 lm, which of yielding, the plastic flow proceeds at nearly
were obtained by annealing at 423 K, the microstruc- constant stress, then exhibits strain-hardening.
ture is similar to that of the as-ECAE condition. For Type IV: The curve shows continuous strain-hardening
specimens of larger grain sizes (>0.47 lm), the micro- normally observed in coarse-grained
structure consists of grains of equiaxed shape in both aluminum.

Fig. 1. Microstructures of specimens with average grain size of (a) 0.78 lm (TEM), and (b) 20 lm (SEM, BEI).
4022 C.Y. Yu et al. / Acta Materialia 53 (2005) 4019–4028

15 size range is localized deformation in the region of

0.78 µm
neck and shear failure, similar to that observed in
20 µm
smaller grains with Type I behavior.
10 (3) Type III: 1 lm < d < 4 lm
Proportion (%)

The tensile r–e curves show Type III characteris-

tics. In this grain size range, the uniform elongation
5 increases significantly with increasing grain size.
The total tensile elongation reaches
30%. After
tensile tests, the specimen surface show parallel
0 microscopic shear bands homogeneously distrib-
uted across the gauge length, which incline at angle
0 10 20 30 40 50 60 of
40 to the tensile axis. These microscopic shear
Misorientation (θ) bands were the traces left by the propagation of
Lüders bands. In comparison with the shear band
Fig. 2. Distribution of boundary misorientation angles for specimens
exhibited in Type II specimens (0.4 lm < d < 1 lm),
with average grain size of 0.78 lm (TEM), and 20 lm (SEM, EBSD).
the microscopic shear bands are more homoge-
neously distributed on the surface of Type III spec-
The tensile behaviors of various grain sizes at RT are imens. Another noticeable feature for Type III
shown in Fig. 3, which can be summarized as follows: specimens is that the gauge cross-section changes
to an elliptical shape, whose minor axis is parallel
(1) Type I: d < 0.4 lm to the projection of the shear direction. From the
For grain sizes smaller than 0.4 lm, the tensile r–e geometry of the cross-section, it can be estimated
curves exhibit the Type I behavior. The strain of that about 37% of the measured tensile strain can
uniform elongation is below 0.03. The failure mode be attributed to shear deformation. This indicates
is necking accompanied by localized shear, and the that in this grain size range, shear deformation
specimen surface outside the necked section still makes a significant contribution to the extension
shows the original electro-polished finish without of the specimen.
any sign of deformation. (4) Type IV: d > 4 lm
(2) Type II: 0.4 lm < d < 1 lm For grain sizes larger than 4 lm, the tensile r–e
The tensile r–e curves for grain sizes within the curves exhibit the Type IV behavior, in which plas-
range of 0.4–1 lm show Type II characteristics. tic deformation proceeds homogeneously in the
They are characterized by an obvious ‘‘yield-drop’’. specimen. The elongation to failure is larger than
At a strain of 0.015, a clear shear offset already 40%. The specimen surface after tensile testing has
appeared on the specimen surface (Fig. 4(a)), which a very rough appearance, typical for coarse grained
might imply that the occurrence of shear banding specimens after large deformation.
was associated with the onset of yielding. Increasing Yield stress can be expressed as a function of
the strain to 0.035, necking became quite evident as inverse square root of grain size (Fig. 5(a)). The
shown in Fig. 4(b). The failure mode for this grain results of pure aluminum given by Thompson et al.
[50] are also included in this figure. In this study,
200 0.2%-offset stress was taken as the yield stress for
0.35µ
specimens without yield-drop, while the lower yield
stress was chosen for specimens that showed yield-
Engineering Stress (MPa)

150
drop or Lüders extension. The specimens of grain
0.59µ
size below 0.4 lm were not taken into account
because of the presence of un-recovered microstruc-
1.03µ
ture. By fitting the data to the Hall–Petch relation,
100
1.72µ this results in ro = 13 MPa and k = 74 MPa lm1/2
for grain sizes greater than 1 lm, which are in rea-
2.5µ
sonable agreement with those obtained for pure alu-
50
minum with coarse-grained structures [18].
12µ
However, the yield stresses of the grain sizes below
45µ
1 lm do not follow the above Hall–Petch relation-
0 ship, which may be related to the change of r–e
0 0.1 0.2 0.3 0.4
behavior from Type III to Type II. Philips and
Engineering Strain
Armstrong [51] suggested that the discontinuous
Fig. 3. Tensile stress–strain curves for selected specimens at RT. change of Hall–Petch relation in the ultrafine-grain
 C.Y. Yu et al. / Acta Materialia 53 (2005) 4019–4028 4023

Fig. 4. Optical micrographs showing shear bands on the surface of a specimen (d = 0.78 lm) deformed at RT to (a) e = 0.015 (right after the yield
peak), and (b) e = 0.035 (onset of necking).

(a) 200 0.1

IV III II
77 K
RT
0.08
150
Luders strain 0.06
Stress (MPa)

100
0.04

50 0.02

0
0 0 0.5 1 1.5 2 2.5
0 0.5 1 1.5
-1 -1
d (µm )
d-1/2 (µm)-1/2
Fig. 6. Grain size dependence of Lüders strain obtained at RT and
(b) 50 77 K.
Uniform elongation
Total elongation elongation decrease abruptly as the grain size
40 reduces to the submicrometer range. Ductile dim-
ples were the main feature of the fracture surfaces
Elongation (%)

30 of all the tensile specimens deformed at RT, even

in samples of submicron grain size, which possessed
poor tensile elongations.
20 Lüders strain appeared in specimens of grain size
smaller than 3.6 lm, and increases with decreasing
10
grain size. However, as the grain size drops below
0.78 lm, necking began soon after yielding so that
Lüders deformation was not observed. The effect
0 of grain size on Lüders strain at room temperature
0 0.5 1 1.5 2
is shown in Fig. 6. As an approximation, the Lüders
d-1/2 (µm)-1/2 strain is proportional to d1, as suggested by Lloyd
Fig. 5. Grain size dependence of tensile properties obtained at RT: and Morris [52].
(a) yield stress, in which the dashed line is from the data of Thompson
et al. [50], and (b) uniform elongation and total elongation to failure. The microstructure in the uniformly deformed region
of tensile tested specimens was examined by TEM. For
size range might be due to the occurrence of discon- specimens with a grain size smaller than 1 lm, the uniform
tinuous yielding. elongation was no more than 1%. The deformation struc-
The grain size dependence of ductility is shown in ture is not uniformly distributed in the specimens so that
Fig. 5(b). The uniform elongation and the total most grains are free of dislocation as shown in Fig. 7(a),
4024 C.Y. Yu et al. / Acta Materialia 53 (2005) 4019–4028

Fig. 7. TEM micrographs showing (a) dislocation-free grains, and (b) grain boundaries with ‘‘spotty’’ contrast in the specimen (d = 0.78 lm) after
tensile failure at RT (eu
0.006).

while some grains do have dislocation tangles in the grain (1) Type II: d = 0.35 lm
interior. For those grains showing dislocations, most of As shown in Fig. 9, this exhibits a yielding peak
the dislocations are concentrated near the grain bound- followed by plastic flow at nearly constant stress to
aries. By careful examination of those grains that are free an elongation about 11% before the onset of local-
of dislocation, it was noted that the grain boundaries of ized necking. The r–e curve has the characteristics
some of them had lost their sharpness and exhibited a of the Type II behavior at RT. Compared with the
‘‘spotty’’ contrast as shown in Fig. 7(b). This might result tensile behavior of the same grain size at RT,
from dislocations trapped at the grain boundaries. Be- deformation at 77 K shows higher values of both
cause the dislocations are trapped at boundaries, an en- uniform elongation (
11% vs.
3%) and total
hanced rate of dynamic recovery due to higher atomic elongation to failure (
29% vs. 16%). The fracture
mobility in grain boundaries is anticipated. This may ex- surfaces of the tensile tested specimens are flat and
plain the lower work-hardening rate in submicron- at an angle about 45 from the tensile axis, that
grained specimens. For specimens of grain size larger than might be due to significant shear contribution to
1 lm, the deformed structure is quite uniform throughout the final failure in the necked region. However,
the specimen, which consists of dislocation tangles and there is no apparent evidence of shear bands on
cell structures formed in grain interior as shown in Fig. 8. the specimen surface.
(2) Type III: 0.4 lm < d < 1 lm
3.3. Tensile behaviors at 77 K The tensile deformation proceeded by the propa-
gation of Lüders band initially and followed by
The r–e curves at 77 K are shown in Fig. 9. In the fol- strain hardening, the Type III behavior. In this
lowing, the tensile behavior will be described separately grain size range, both strength and tensile ductility
for different grain size ranges.
300

0.35µ
250
Engineering Stress (MPa)

0.47µ
0.59µ
200
0.78µ

150 1.03µ

100 1.72µ

12µ
50 45µ

0
0 0.2 0.4 0.6 0.8
Fig. 8. TEM micrograph showing tangled dislocations in grain Engineering Strain
interior in the specimen (d = 1.72 lm) after tensile failure at RT
(eu
0.15). Fig. 9. Tensile stress–strain curves for selected specimens at 77 K.
 C.Y. Yu et al. / Acta Materialia 53 (2005) 4019–4028 4025

vary significantly with grain size. The surface of a (a) 300

specimen of d = 0.59 lm, which was tensile IV III
deformed to a strain within Lüders extension, 250
was examined using an optical microscope (OM). extraploated yield stress
experimental yield stress
As shown in Fig. 10, the Lüders band, which prop-
200
agates partly through the gauge length, actually

Stress (MPa)
consists of uniformly distributed microscopic
shear bands. Similar to those specimens exhibiting 150
the Type III behavior at RT, fine marks of micro-
scopic shear bands are distributed uniformly on 100
the specimen surface. The cross-section of uni-
formly deformed gauge section has changed from 50
the initial circular shape into an elliptical shape
after tensile deformation.
(3) Type IV: d > 1 lm 0
0 0.5 1 1.5
For grain sizes larger than 1 lm, the r–e curves
d-1/2 (µm)-1/2
show the normal strain-hardening behavior (Type
IV behavior). It is worth noting that specimens in
this grain size range show excellent ductility (total (b) 80
elongation of 60–70%) at 77 K. No sign of shear 70
bands were observed on the surface of tensile
deformed specimens. 60
Elongation (%)
50
It is found that decreasing the testing temperature
from RT to 77 K results in a significant increase in both 40
strength and ductility. Lüders extension appears in the
size range of 1–3.6 lm at RT, while at 77 K, it appears 30
in specimens with smaller grain sizes (d < 1 lm). The 20
yield stresses at 77 K were related to the grain sizes by total elongation
the Hall–Petch relationship as shown in Fig. 11(a). 10 uniform elongation

The best-fitted line of the data of grains greater than

0
1 lm cannot extrapolate to grains of submicrometer 0 0.5 1 1.5 2
size, which has a different fitting line. The transition be- d-1/2 (µm)-1/2
tween these two Hall–Petch relations occurs at a grain
size of about 1–2 lm. The Hall–Petch constants for Fig. 11. Grain size dependence of tensile properties obtained at 77 K:
grain sizes larger than 1 lm are ro = 37 MPa and (a) yield stress, in which the solid symbols are the experimental data
and the open symbols are extrapolated values of 0.2% offset yield stress
k = 68 MPa lm1/2. The constant k is almost identical (refer to the text), and (b) uniform elongation and total elongation to
with that obtained at RT for the same grain size range. failure.
As the grain size decreases to the submicrometer range
(d < 1 lm), the Hall–Petch slope, k, increases to
200 MPa lm1/2. The grain size dependence of ductility
at 77 K is shown in Fig. 11(b). Both ductility indexes,
uniform elongation and total elongation, show a similar
trend of continuous decrease with decreasing grain size.

Fig. 10. Optical micrographs showing the propagation of Lüders band Fig. 12. TEM micrograph showing the dislocation cell structures
on the surface of the specimen (d = 0.59 lm) deformed at 77 K to formed in grain interior of the specimen (d = 0.47 lm) after tensile
e = 0.07. testing to failure at 77 K (eu
0.24).
4026 C.Y. Yu et al. / Acta Materialia 53 (2005) 4019–4028

Ductile dimples were observed on the fracture surfaces effect in strengthening the material. If this is not the
for all samples. The Lüders strain at 77 K is also inver- case, the storage or pile-up of dislocations at the grain
sely proportional to the grain size as shown in Fig. 7. boundaries will result in hardening. The influence of
For specimens with a grain size smaller than 1 lm, the boundary characters on the strain-hardening behavior
deformation structures after tensile test at 77 K are quite has been shown in UFG aluminum with a different pro-
uniform throughout the specimen. It exhibits a significant portion of high angle boundaries [28].
number of dislocations tangled in grain interior and dis-
location cell structure develops as shown in Fig. 12. 4.2. Inhomogeneous yielding in UFG structure

Apart from the present case, the presence of yield

4. Discussions peak or Lüders deformation has been observed in var-
ious alloys with UFG structure, including high purity
4.1. Grain size dependent deformation behaviors (99.99%) aluminum [53], 1100 Al [29], Al–3%Mg [34],
steel [29]. Wyrzykowski and Grabski [53] indicated
A continuous change of tensile deformation behavior that the inhomogeneous flow in UFG aluminum was
was observed as the grain size decreased from microme- not a result of the existence of solute atoms but
ter to submicrometer range. Materials with traditional may arise from the lack of mobile dislocations. Be-
grain sizes exhibit continuous hardening (Type IV cause the inhomogeneous flow appears in different al-
behavior), which is the normal behavior of annealed loys with UFG structure, it is believed that the lack of
polycrystalline aluminum with coarse grain size. As the mobile dislocations is associated with the UFG
grain size reduces below a critical value Lc, inhomoge- structure.
neous yielding occurs and necking follows soon after For specimens with a grain size below 0.4 lm, the dis-
yielding (Type II behavior). The early onset of necking locations introduced in the ECAE process cannot be re-
may be related to the reduced hardening rate with moved completely by the low annealing temperatures.
decreasing grain size. The appearance of Type I behav- During tensile deformation, the tested specimen has en-
ior is simply due to the presence of residual dislocations ough mobile dislocations to keep up with the applied
in materials, which are not fully annealed. The four strain rate. Consequently, the tensile r–e curves of grain
types of tensile behaviors classified here may correspond size smaller than 0.4 lm do not exhibit the yield-drop
to the different grain-size regimes suggested in the liter- behavior at RT. However, a yield peak does appear in
ature [10,11]: (a) Type IV belongs to the traditional re- the 77 K tensile r–e curve of the as-ECAE condition
gime, (b) Types I, and II belong to the UFG regime, (d = 0.35 lm). It seems to imply that the number of dis-
and (c) Type III may be a phenomenon occurring in locations that can be activated at 77 K is lower than that
the transition region between traditional and UFG re- at room temperature. Therefore, larger stress is needed
gimes. As the deformation temperature decreases from at 77 K to activate enough dislocations to fulfill the ap-
RT to 77 K, the value of critical grain size, Lc, shifts plied strain rate. It needs to be pointed out that the
to a smaller size. change of tensile behavior from Type I to Type II is
These grain-size dependent behaviors observed in the actually affected by grain boundary sources and mobile
present study may be explained qualitatively by the fol- dislocation density, which is reduced concurrently with
lowing model. The model was extended from the sugges- grain growth upon annealing.
tions given in the literature [11,13,54]. The critical value As proposed by Johnston and Gilman [55], yield-
Lc may relate to the cell size present at the applied strain drop of pure metals can be attributed to a shortage of
rate and temperature. As grain size decreases below Lc, mobile dislocations in the specimens to fulfill the applied
the mean free path of dislocation is no longer deter- strain rate. Shear strain c can be expressed as c = bNA,
mined by the dislocation structure, but is limited by where b is Burgers vector, N is the number of mobile dis-
the grain boundaries [54]. This is the size range of the location per unit volume, and A is the area swept by the
UFG regime, in which materials may exhibit Type I, moving dislocation. For grain sizes in the UFG regime,
or II behaviors. In the UFG regime, grain boundaries the deformation is controlled by dislocations emitted
would serve as both source and sink of dislocations from grain boundary sources instead of intragranular
[11,13]. Fewer dislocations than those in coarse grained sources [11,13], and then N would be proportional to
materials are stored in the grain interior and more dislo- the source density on the grain boundaries. As the grain
cations are stored at grain boundaries. Therefore, the size is reduced, the surface area of grain boundaries in-
hardening behavior becomes strongly dependent on creases, so that provides more dislocation sinks, which
grain boundary characters. If dislocations are easily in turn reduces A, thus a higher value of N is required
annihilated at grain boundaries through thermally acti- to fulfill the applied strain rate. If the mobile dislocation
vated processes of structural rearrangement like migra- density is lower than the required value of N, the dislo-
tion of ledges, the grain boundaries may have little cations must increase their velocity to maintain the
 C.Y. Yu et al. / Acta Materialia 53 (2005) 4019–4028 4027

strain rate; consequently a high stress results from the hardening capacity at 77 K, which is a result of a lower dy-
increased velocity and a yield peak appears. It should namic recovery rate at lower deformation temperature.
be emphasized that it is the dislocation sink effect of
grain boundaries that causes the yielding phenomenon 4.4. Deviation of Hall–Petch relationship in UFG
in UFG materials. Since the dislocation annihilation at aluminum
grain boundaries is a thermally activated process, the
occurrence of yielding phenomena would shift to a smal- As shown in Figs. 5 and 11, there is a positive devia-
ler grain size at lower testing temperatures. tion of the Hall–Petch relationship for aluminum with
In polycrystalline materials, there are grains that are UFG structure. In the present work, the materials of
easy to yield distributed in a matrix of grains that are grain size below about 1 lm exhibit inhomogeneous
relatively difficult to yield, which results in pre-yielding yielding, which is absent at larger grain sizes. For defor-
microstrain as the stress increases. Finally, those yielded mation proceeding by inhomogeneous yielding, there
grains percolate and general yield occurs. The transition must be difficulty in spreading the deformation to cause
of Type II to Type III behavior depends on the strain general yielding of the specimen. In other words, higher
hardening rate following yielding. If the hardening rate stress is required to initiate the plastic flow. Another
is enough to sustain the spreading of deformation to consequence of inhomogeneous yielding is flow localiza-
the surrounding undeformed matrix, the Type III tion in the specimen such that the measured average
behavior appears. Otherwise, the general yielding is strain across the gauge length does not represent the ac-
quickly followed by local necking (the Type II behav- tual strain experienced in the deformed zone. The flow
ior). The strain hardening rate in the UFG regime is behavior can be assessed without complication of inho-
strongly dependent upon grain size, grain boundary mogeneous yielding by extrapolating the work harden-
characters, and deformation temperature. A smaller ing region of the flow curves back to the elastic line
grain size will reduce the probability of dislocation trap- [18]. The extrapolation was performed by the use of
ping in the grain interior, therefore reducing the harden- Ludwik work hardening equation
ing rate. A lower deformation temperature, which rðeÞ ¼ ri þ Cen ; ð2Þ
reduces the dynamic recovery rate and promotes the
storage of dislocations in grain interior, would shift where r(e) is the true flow stress at e, ri is the propor-
the transition of Type II to Type III behavior to a smal- tional limit, e is true plastic strain, and C and n are work
ler grain size. hardening parameters. The work hardening part of the
r–e curves was fitted to Eq. (2) and the intercept of this
4.3. Low tensile ductility in UFG aluminum curve with the elastic loading line was taken as the pro-
portional limit ri. This extrapolation was performed for
The poor tensile ductility in materials of submicron the r–e curves at 77 K. By adopting the extrapolated
grain sizes at RT (Fig. 5) might be attributed to the values of 0.2% offset yield stress for the submicron grain
lower work hardening rate. This argument could be sup- size range, a near linear relationship in the Hall–Petch
ported by the TEM observations of dislocation struc- plot can be realized at 77 K as shown in Fig. 11(a). Sim-
tures in tensile deformed specimens. There was little ilar extrapolation cannot be applied to the results ob-
dislocation interaction within the grain interior for tained at RT, because softening follows the yielding
materials of submicron grain size. The dislocations peak. This finding suggests that the positive deviation
trapped at grain boundaries (Fig. 7(b)) resulted in an of the Hall–Petch plot in the submicrometer size range
increasing rate of dynamic recovery; therefore, UFG may be attributed to the initial inhomogeneous flow.
aluminum has a lower work-hardening rate compared
to traditional grain size. The onset of plastic instability
(necking) during tensile test can be explained based on 5. Summary and conclusions
the Considère criterion. The flow stress greatly increases
by the refinement of grain size to the submicrometer Commercial purity aluminum (AA1050) of grain size
range. However, the work-hardening rate is not en- ranging from 0.35 to
45 lm was prepared by the com-
hanced but rather suppressed by UFG structure. As a bination of ECAE and annealing treatment. The tensile
result, plastic instability (necking) occurs at a very early deformation behavior was found to change continu-
stage of tensile deformation in the UFG materials, ously as the grain size decreased from the micrometer
which results in limited uniform elongation. to submicrometer range. It was also noted that the ten-
The above argument can be further justified by com- sile behaviors of this material showed a strong depen-
paring the tensile behaviors observed at RT and 77 K. dence on testing temperature as the grain size reduced
The total elongation of submicron-grained aluminum de- to the submicrometer range. Materials with traditional
formed at 77 K is greater than 30%, which is much higher grain sizes exhibit continuous hardening, which is the
than that at RT. It can be attributed to an improved work normal behavior of annealed polycrystalline aluminum
4028 C.Y. Yu et al. / Acta Materialia 53 (2005) 4019–4028

with coarse grain size. As the grain size reduces to the [15] Fougere GE, Weertman JR, Segal RW, Kim S. Scripta Metall
UFG regime, the deformation starts with a distinct 1992;26:1879.
[16] Masumura RA, Hazzledine PM, Pande CS. Acta Mater
yielding peak followed by localized necking. For grain 1998;46:4527.
size in the transition region between traditional and [17] Sanders PG, Eastman JA, Weertman JR. Acta Metall
UFG regimes, tensile deformation proceeds by the prop- 1997;45:4019.
agation of Lüders band. As the deformation tempera- [18] Chokshi AH, Rosen A, Karch J, Gleiter H. Scripta Metall
ture decreases from RT to 77 K, the transition region 1989;23:1679.
[19] Neiman GW, Weertman JR, Segal RW. Scripta Metall Mater
shifts to a smaller grain size. 1989;23:2013.
A model was proposed to explain the deformation [20] Furukawa M, Horita Z, Nemoto M, Valiev RZ, Langdon TG.
behaviors in UFG aluminum. As grain size decreases be- Acta Mater 1996;44:4619.
low a critical value, which relates to the cell size that [21] Thompson AW. Acta Metall 1975;23:1337.
would be formed in a coarse-grained material, at the ap- [22] Kashyap BP, Tangri K. Acta Metall 1995;43:3971.
[23] Lloyd DJ. Metal Sci 1980;14:193.
plied strain rate and temperature, the mean free path of [24] Sangal S, Tangri K. Scripta Metall 1989;23:2079.
dislocation is no longer determined by the dislocation [25] Wyrzykowski JW, Grabski MW. Metal Sci 1983;17:445.
structure, but is limited by the grain boundaries. The [26] Matsumoto K, Shibayanagi T, Umakoshi Y. Scripta Metall
presence of a yielding peak in UFG aluminum is attrib- 1995;33:1321.
uted to the lack of enough mobile dislocations due to the [27] Wyrzykowski JW, Grabski MW. Philos Mag 1986;53:505.
[28] Sun PL, Yu CY, Kao PW, Chang CP. Scripta Mater
large dislocation sink area provided by grain bound- 2005;52:265.
aries. The consequence of grain boundaries as sinks of [29] Tsuji N, Ito Y, Saito Y, Minamino Y. Scripta Mater
dislocations is an enhanced dynamic recovery rate and 2002;47:893.
reduced hardening rate in UFG metals. The early onset [30] Miller RL. Metall Trans 1972;3:905.
of necking in UFG aluminum can be related to the re- [31] Wang YM, Ma E, Chen MW. Appl Phys Lett 2002;80:2395.
[32] Wang YM, Chen MW, Zhou FH, Ma E. Nature 2002;419:912.
duced hardening rate with decreasing grain size. [33] Wang YM, Ma E. Acta Mater 2004;52:1699.
[34] Hayes JS, Keyte R, Prangnell PB. Mater Sci Tech 2000;16:1259.
[35] Yu CY, Sun PL, Kao PW, Chang CP. Scripta Mater 2005;52:359.
Acknowledgment [36] Carsley JE, Fisher A, Milligan WW, Aifantis EC. Metall Mater
Trans 1998;29A:2261.
[37] Han BQ, Lavernia EJ, Mohamed FA. Metall Mater Trans A
This work is supported by the National Science Coun- 2003;34A:71.
cil of ROC under contract NSC-92-2216-E-110-006. [38] Wei Q, Jia D, Ramesh KT, Ma E. Appl Phys Lett 2002;81:1240.
[39] Jia D, Ramesh MaE. Acta Mater 2003;51:3495.
[40] Valiev RZ, Islamgaliev RK, Alexandrov IK. Prog Mater Sci
References 2000;45:103.
[41] Segal VM. Mater Sci Eng 1995;A197:157.
[1] Benson DJ, Fu HH, Meyers MA. Mater Sci Eng 2001;A319– [42] Furukawa M, Iwahashi Y, Horita Z, Nemoto M, Langdon TG.
321:854. Mater Sci Eng A 1998;257:328.
[2] Koch CC, Morris DG, Lu K, Inoue A. MRS Bull [43] Iwahashi Y, Horita Z, Nemoto M, Langdon TG. Acta Mater
1999;February:54. 1997;45:4733.
[3] Muhai T, Kawazoe M, Higashi K. Nanostructure 1998;10:755. [44] Chang CP, Sun PL, Kao PW. Acta Mater 2000;48:3377.
[4] Tsai TJ, Sun PL, Kao PW, Chang PW. Mater Sci Eng [45] Sun PL, Kao PW, Chang CP. Metall Mater Trans A
2003;A342:144. 2004;35A:1359.
[5] Valiev RZ, Alexandrov IV, Zhu YT, Lowe TC. J Mater Res [46] Humphreys FJ, Prangnell PB, Bowen JR. Philos Trans Roy Soc
2002;17:5. Lond 1999;A357:1663.
[6] Wang YM, Ma E, Valiev RZ, Zhu YT. Adv Mater 2004;16:328. [47] Yu CY, Sun PL, Kao PW, Chang CP. Mater Sci Eng
[7] Zhu YT, Liao XZ. Nat Mater 2004;3:351. 2004;A366:310.
[8] Youssef KM, Scattergood RO, Murty KL, Koch CC. Appl Phys [48] Iwahashi Y, Horita Z, Nemoto M, Langdon TG. Acta Mater
Lett 2004;85:929. 1998;46:3317.
[9] Zhang X, Wang H, Scattergood RO, Narayan J, Koch CC, [49] Zaefferer S. J Appl Crystallogr 2000;33:10.
Sergueeva AV, et al. Appl Phys Lett 2002;81:823. [50] Thompson AW, Baskes MI, Flangan WF. Acta Metall
[10] Conrad H. Metall Mater Trans A 2004;35A:2681. 1973;21:1017.
[11] Cheng S, Spencer JA, Milligan WW. Acta Mater 2003;51:4505. [51] Philips WL, Armstrong RW. Metall Trans 1972;3:2571.
[12] Conrad H. Mater Sci Eng 2003;A341:216. [52] Lloyd DJ, Morris LR. Acta Metall 1977;25:857.
[13] Hayes RW, Witkin D, Zhou F, Lavernia EJ. Acta Mater [53] Wyrzykowski JW, Grabski MW. Mater Sci Eng 1982;56:197.
2004;52:4259. [54] Li YJ, Zeng XH, Blum W. Acta Mater 2004;52:5009.
[14] Li JCM, Chou YT. Metall Trans 1970;1:1145. [55] Johnston WG, Gilman JJ. J Appl Phys 1959;30:129.