Vous êtes sur la page 1sur 12

Available online at www.sciencedirect.

com

ScienceDirect
Acta Materialia 72 (2014) 125–136
www.elsevier.com/locate/actamat

Atomic-scale analysis of the segregation and precipitation


mechanisms in a severely deformed Al–Mg alloy
X. Sauvage a,⇑, N. Enikeev b,c, R. Valiev b,c, Y. Nasedkina a, M. Murashkin b,c
a
University of Rouen, CNRS UMR 6634, Groupe de Physique des Matériaux, Faculté des Sciences, BP12, 76801 Saint-Etienne du Rouvray, France
b
Institute for Physics of Advanced Materials, Ufa State Aviation Technical University, K. Marx 12, Ufa 450000, Russia
c
Saint Petersburg State University, Universitetsky Prospect, 28, 198504 Peterhof, St Petersburg, Russia

Received 31 January 2014; accepted 15 March 2014


Available online 19 April 2014

Abstract

Due to their interaction with crystalline defects, solute atoms play a critical role in the microstructure evolution of aluminum alloys
during deformation. In addition, deformed structures often exhibit a modified aging response. For a better understanding of these
mechanisms, we provide here a thorough study of deformation-induced segregation and precipitation mechanisms in an aluminum alloy
containing 5.8 wt.% Mg subjected to severe plastic deformation (SPD). The solutionized alloy was processed by high-pressure torsion at
room temperature and at 200 °C. The investigation of the microstructure and of the distribution of Mg after deformation by scanning
transmission electron microscopy and atom probe tomography revealed that clustering and segregations occurred during severe
deformation. Mg atoms agglomerate on grain boundaries (GBs), forming mostly nanoscaled clusters at room temperature and more uni-
form segregation along GBs at 200 °C. In any case, however, the equilibrium Al3Mg2 phase does not nucleate. Using post-deformation
annealing treatments, it was found that it can proceed only through a very specific orientation relationship with the face-centered-cubic
Al matrix. Both the contribution of dislocations and deformation-induced vacancies were considered to account for the enhanced
mobility of Mg atoms. From theoretical estimations it is, however, concluded that Mg atoms are dragged by the vacancy flux toward
GBs while dislocations should not play a significant role. These data provide new insights about mechanisms controlling dynamic pre-
cipitation and segregation during SPD of aluminum alloys. The segregation and formation of clusters that is revealed can additionally
contribute to the strengthening of these alloys, leading to a new understanding of dynamic ageing in non-age-hardenable alloys.
Ó 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Plastic deformation; Grain boundary; Segregation; Diffusion; Precipitation

1. Introduction distributed in solid solution in the face-centered cubic


(fcc) Al phase and strongly interact with dislocations
As they offer a good compromise of properties such as [1–6]. These interactions reduce the dynamic recovery
formability, mechanical strength, weldability and corrosion process compared to pure aluminum, even at room temper-
resistance, Al–Mg alloys with a typical Mg content varying ature. This leads to a higher strain hardening and a higher
in the range of 1–5 at.% are widely used for commercial dislocation density after deformation [7,8].
application. Compared to pure aluminum, their higher Using severe plastic deformation (SPD) techniques, it
yield stress and strain hardening is due to solid solution has been shown that ultrafine-grained (UFG) structures
strengthening, i.e. Mg atoms are homogeneously can be achieved in various metals and metallic alloys [9].
The grain size refinement mechanisms during SPD are con-
⇑ Corresponding author. Tel.: +33 2 32 95 51 42. trolled by the generation of dislocations and dynamic
E-mail address: xavier.sauvage@univ-rouen.fr (X. Sauvage). recovery processes, i.e. the way dislocations dynamically

http://dx.doi.org/10.1016/j.actamat.2014.03.033
1359-6454/Ó 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
126 X. Sauvage et al. / Acta Materialia 72 (2014) 125–136

reorganize to initially form low-angle boundaries (disloca- resulting from SPD were also investigated. Contrary to
tion cell walls) and finally, for larger strains, high-angle our previous work where the experimental approach was
boundaries [9,10]. It has been experimentally observed that mainly based on APT, here we have mostly used high-angle
a small amount of Mg may dramatically change the grain annular dark-field (HAADF) scanning transmission elec-
size achievable by SPD. Indeed, the typical grain size tron microscopy (STEM) to characterize the distribution
reported for pure Al processed by high-pressure torsion of Mg. This technique cannot provide quantitative infor-
(HPT) is 800 nm [11], going down to only 150 nm in an mation about local concentrations but, the field of view
Al–3% Mg alloy [12–14]. This trend has been confirmed being much larger, a much better statistic could be
by a large number of experimental investigations [15–18] obtained.
and is obviously linked to the strong interactions of Mg
atoms with dislocations mentioned above. Besides, it is also 2. Experimental
well known that Mg is prone to grain boundary (GB) seg-
regation. Both experimental evidence [19,20] and simula- The material investigated in the present study is a
tion work [21–23] have demonstrated this tendency. In hot-pressed commercial aluminum alloy containing 5.7%
our previous works [24,25] we have shown by atom probe Mg, 0.32% Sc and 0.4% Mn (wt.%), it will be named Al–
tomography (APT) analyses that the Mg distribution in a Mg in the following. Before grain refinement by SPD, the
homogenized Al–5.7 wt.% Mg alloy could be significantly material was homogenized at 380 °C for 2 h (above the sol-
affected by the SPD at room temperature. High local Mg ubility limit of Mg in fcc Al). After this treatment, the grain
concentrations were indeed observed along defects that size was 60 lm.
were attributed to “non-equilibrium” GBs, and the signifi- Discs (20 mm in diameter and 2 mm in thickness) were
cant deviation of the yield stress to the Hall–Petch law machined for SPD by HPT at room temperature or at
recorded for this UFG structure was attributed to these 200 °C. During processing, a constant pressure of 6 GPa
segregations [24]. Thus, the redistribution of alloying ele- was applied and discs were deformed up to 20 revolutions.
ments can significantly affect the strength, and understand- This process being intrinsically heterogeneous (larger
ing the related mechanisms becomes an important problem deformation in the outer part of the disc), samples for
of modern materials science. the microstructure characterization were always carefully
The aim of the present work is to propose a comprehen- extracted always in the same region, i.e. at a distance of
sive study of the redistribution of the solute element in a 6 mm from the center (approximately half the radius from
similar supersaturated solid solution of Mg in Al. At first, the center).
beyond fundamental interest, the present study is also Details about X-ray diffraction (XRD) experiments were
motivated by a better understanding and a better control described in our previous work [18]. Transmission electron
of microstructural features in UFG materials processed microscopy (TEM) samples were prepared by standard
by SPD. Indeed, GB segregations may, for example, signif- electropolishing methods and then examined using a JEOL
icantly affect properties [25] such as strength [24,26–28] or ARM-200F probe corrected microscope operating at
thermal stability [29–31], which is a key issue for numerous 200 kV. Observations were performed both in conventional
possible applications of UFG materials. Second, the scien- transmission with a parallel beam and in STEM with a
tific issue raised by this paper deals with the stability of probe size of 0.2 nm and a convergence angle of 34 mrad.
supersaturated solid solutions under plastic deformation. STEM-HAADF images were recorded with a collection
It is indeed interesting to note that 5.7 wt.% Mg is much angle in the range of 50–180 mrad and dark field (DF)
higher than the equilibrium solubility limit of Mg in Al images with a collection angle in the range of 20–50 mrad.
at room temperature (below 1 at.% [32]). Thus, as reported Elemental mapping was performed using energy dispersive
for the Al–Zn system [33], phase separation may occur dur- X-ray spectroscopy (EDS) with a JED2300 detector. APT
ing SPD. However, Al–Mg supersaturated solid solutions samples were also prepared by standard electropolishing
with only a few at.% Mg usually decompose very slowly techniques [38]. Analyses were performed in ultra-high vac-
because of the high nucleation barrier of the b or b0 phases uum conditions, using an energy-compensated atom probe
(Al3Mg2) [34]. GP zones or metastable phases with a L10 or equipped with an ADLD detector [39] and a reflectron.
L12 structure have also been reported [35–37] but are Samples were field-evaporated using electric pulses
unstable at room temperature or higher if the Mg content (30 kHz pulse repetition rate and 20% pulse fraction). As
of the alloy is below 9 at.%. In any case, GBs play a key SPD samples are extremely sensitive and prone to early
role by promoting heterogeneous precipitation [34]. Thus specimen failure during analysis, it was impossible to ana-
specific features are expected in UFG structures obtained lyze them at a temperature below 80 K. However, at a tem-
by SPD where typically a large amount of GBs are created. perature higher than 40 K, a significant amount of Mg
Therefore, a similar material (Al–5.7 wt.% Mg) was pro- might be evaporated between pulses (preferential evapora-
cessed by HPT at room temperature and 200 °C after solu- tion), leading to non-quantitative measurements. Thus,
tion treatment to clarify the influence of the processing samples were analyzed at 80 K to avoid specimen failure
temperature. Then, the thermal stability and the redistribu- but a systematic correction was applied on the Mg content
tion of Mg atoms upon annealing of the UFG structure to account for the preferential evaporation. This correction
X. Sauvage et al. / Acta Materialia 72 (2014) 125–136 127

is based on a calibration measurement that was performed


on the non-deformed material that was analyzed by APT
both at 80 K and at 40 K (7.0 ± 0.2 at.% Mg measured at
40 K vs. 4.4 ± 0.2 at.% at 80 K).

3. Results

3.1. Redistribution of Mg atoms during SPD

As already reported in our previous work [24,25], the


homogenization treatment performed before HPT (2 h,
380 °C), leads to a very homogeneous distribution of Mg
atoms within the fcc Al matrix. The average concentration
measured by APT was 7.0 ± 0.2 at.%, i.e. very close to
the nominal composition and consistent with the fcc Al lat-
tice parameter estimated from XRD measurements
(Table 1). Besides, since the alloy contains some Sc and
Mn, it was not surprising to find some Al3Sc nanoscaled
particles with a size in the range of 5–10 nm and also signif-
icantly larger Al–Mn intermetallic particles. During HPT
at room temperature, significant modification of this origi-
nal microstructure occurred. STEM observations were car-
ried out using two detectors: (i) a DF detector that
provides images with diffraction contrast to visualize GBs
and (ii) a HAADF detector that provides Z contrast to
exhibit local composition changes (Sc- and Mn-rich clus-
ters are brightly imaged and Mg-rich clusters appear dark).
As shown in the STEM-DF image (Fig. 1b), after HPT at
room temperature, the grain size is in the range of 100–
200 nm (consistent with previous data [24,25]). Mn- and
Sc-rich precipitates also clearly appear in the STEM-
HAADF image (bright in Fig. 1a), but their size, shape
and distribution were not significantly affected by the
deformation. Besides, a large number density of nano-
scaled dark areas with various shapes (spherical, ellipsoid, Fig. 1. STEM images of the AlMg alloy severely deformed at room
plates) is exhibited in the same image. Due to their dark temperature. (a) HAADF image (Z contrast) showing bright Al3Sc
contrast, they are believed to be Mg-rich clusters or segre- spherical nanoparticles and dark zones that could be attributed to Mg-rich
clusters. (b) Dark field image where grain boundaries are clearly exhibited
gations. Their typical size or thickness is in the range of 5–
thanks to diffraction contrast.
20 nm. Many of them seem located in the vicinity or along
GBs, but others could be observed at triple junctions or
inside grains. To confirm that these dark clusters are Mg- Fig. 3a clearly shows indeed that the Mg distribution after
rich, EDS mapping was performed at a higher magnifica- HPT at room temperature is not homogeneous. The local
tion (Fig. 2b). The line scan (Fig. 2c) performed along fluctuations of the Mg concentration were quantified on
the arrow plotted in Fig. 2a confirms that bright (respec- a two-dimensional (2-D) mapping (Fig. 3b), where it is also
tively dark) particles are Sc- (respectively Mg-) rich. demonstrated that nanoscaled clusters with a size in the
Besides, spot analyses revealed local concentrations in a range of 5–10 nm and with a local composition ranging
typical range of 10–20 at.% in these dark areas against only from 10 to 20 at.% could be detected. Additionally, con-
few at.% nearby. Such nanoscaled Mg-rich clusters were ventional TEM observations were carried out and it is
also revealed by APT. The analyzed volume displayed in important to note that only Al3Sc and Al–Mn intermetallic

Table 1
Mg concentrations in the aluminum matrix in the initial material and after HPT at room temperature and at 200 °C. Data obtained by local APT
measurements are compared to estimations based on the measurement of the fcc Al lattice parameter (XRD).
at.% Mg in sol sol Initial material HPT at room temperature HPT at 200 °C
APT 7.0 ± 0.2 6.4 ± 0.5 6.1 ± 0.5
XRD 7.0 ± 0.15 6.4 ± 0.3 6.0 ± 0.1
128 X. Sauvage et al. / Acta Materialia 72 (2014) 125–136

Fig. 3. APT data collected in the AlMg alloy severely deformed at 200 °C:
(a) 3-D reconstruction showing nanoscaled Mg-rich clusters (Mg atoms
are plotted in red and Al atoms in blue); (b) 2-D Mg chemical map
computed across these clusters showing that their Mg content is 10 at.%
and locally up to 20 at.% (sampling volume 2  2  2 nm3). (For
interpretation of the references to color in this figure legend, the reader
is referred to the web version of this article.)

particles gave rise to diffraction contrast in bright field, or


lattice contrast in high resolution. Besides, it was impossi-
ble to detect in diffraction patterns any reflection
corresponding to the b or b0 phases (Al3Mg2) or any other
Mg-rich metastable phases reported in the literature
[40,41].
After HPT at 200 °C, the average grain size is in the
range of 200–350 nm (Fig. 4b), significantly larger than
after room temperature processing. Mg-rich layers cover
GBs (darkly imaged on HAADF images), or are located
at triple junctions (Fig. 4a). However, a significant fraction
(in the range of 30–50%) of GBs is not covered by such seg-
regations. As reported in the literature, this could be
related to various GBs structures or misorientations [42–
44,20,19,21,22]. Few Mg-rich clusters or segregations were
also observed along some interfaces between the fcc Al
matrix and Al–Mn intermetallic particles. Beside, like in
Fig. 2. Al3Sc spherical nanoscaled precipitate and Mg-rich zones in the the material processed at room temperature, it was impos-
AlMg alloy severely deformed at room temperature: (a) STEM-HAADF sible to reveal b or b0 phases (Al3Mg2) or any other Mg-rich
image, (b) corresponding EDS map (Al–K blue, Sc–K green and Mg–K metastable phases using diffraction in conventional TEM.
red); (c) line profile along the direction indicated by the arrow in (a),
EDS analyses (Fig. 5) was carried out and it was found that
clearly indicating that bright zones in HAADF images are Sc-rich
precipitates while dark zones are Al-poor and Mg-rich. (For interpretation the Mg concentration along GBs and at triple junctions is
of the references to color in this figure legend, the reader is referred to the typically in the range of 10–15 at.%. This is slightly lower
web version of this article.) than the values reported for the material processed at room
temperature. The typical width of these segregations along
X. Sauvage et al. / Acta Materialia 72 (2014) 125–136 129

Fig. 4. STEM images of the AlMg alloy severely deformed at 200 °C: (a) Fig. 5. Al3Sc precipitate and Mg-rich zones in the AlMg alloy severely
HAADF image (Z contrast) showing bright Al3Sc and Al6Mn particles deformed at 200 °C: (a) STEM-HAADF image; (b) corresponding EDS
and few dark zones, especially along triple junctions and GBs that could map (Al–K blue, Sc–K green and Mg–K red). (For interpretation of the
be attributed to Mg segregations; (b) dark-field image where grain references to color in this figure legend, the reader is referred to the web
boundaries are clearly exhibited thanks to diffraction contrast. version of this article.)

GBs is 10 nm but there is a large scatter (Fig. 5). To con- Since Mg-rich clusters and segregations are formed dur-
firm these observations, the material was analyzed by APT. ing SPD both at room temperature and at 200 °C, the
The three-dimensional (3-D) reconstruction of an analyzed amount of Mg in solid solution in the matrix decreases.
volume containing a triple junction is shown in Fig. 6 Local measurements by APT and macroscopic measure-
where GBs are arrowed. The grain located in the bottom ments by XRD measurements (based on the relationship
of the volume was orientated with a h1 1 1i direction close between the lattice constant and the Mg concentration)
to the analysis direction so that (1 1 1) fcc Al planes are confirm this point and are reported in Table 1. The
clearly exhibited (Fig. 6a). In this volume, the Mg distribu- relatively large scatter of APT measurements indicates that
tion looks fairly homogeneous (Fig. 6b), but concentration some significant variations were observed from one
profiles computed across two GBs show that some small analyzed volume to another. Anyway, from these measure-
local fluctuations may arise. One of the grain boundaries ments and assuming that these clusters/segregations have
is slightly depleted (Fig. 6c), while the other appears locally the same molar volume than the matrix, it is possible to
significantly enriched in Mg (Fig. 6d). In general, from all estimate the average volume fraction Fv of Mg-rich clusters
the collected data, it is confirmed that the length scale of or segregations as:
Mg concentration fluctuations is larger compared to the
F v ¼ ðX o  X m Þ=ðX c  X m Þ ð1Þ
material processed at room temperature.
130 X. Sauvage et al. / Acta Materialia 72 (2014) 125–136

microstructure, even if many of them are located in the


vicinity of GBs. Thus, if it is assumed that these clusters
are spherical with a mean radius Rc and homogeneously
distributed, then the mean distance between these clusters
Lc can be written as:
Lc ¼ Rc ð4p=ð3F v ÞÞ1=3 ð2Þ
With Xo = 7%, Xm = 6%, Xc = 20% and Rc in the range of
5–10 nm, it leads to Lc in the range of 20–40 nm, in
relatively good agreement with our observations by
STEM-HAADF (Figs. 1 and 2).
In the material processed by HPT at 200 °C, Mg-rich
zones appear mostly located along GBs as segregations.
Then, if it assumed that grains with a mean diameter d have
boundaries covered by a homogeneous Mg-rich layer, the
thickness t of this covering layer can be written as:
t ¼ dF v =3 ð3Þ
With Xo = 7%, Xm = 6%, Xc = 15%, d = 300 nm, this leads
to a segregation layer thickness in the range of 15–22 nm
(corresponding to a fraction of covered boundaries in the
range of 70–50%), in relatively good agreement with our
observations by STEM-HAADF (Figs. 4 and 5).

3.2. Redistribution of Mg atoms during annealing of the


severely deformed alloy

For a better understanding of the formation of Mg-rich


clusters and segregations during SPD, it was decided to
compare with the nucleation and growth of the b phase
or other metastable Al–Mg intermetallic phases in the
UFG structure. This could be easily achieved by a low tem-
perature annealing, typically in the range of 100–200 °C
[34,23,45]. However, to avoid any interference between
the nucleation processes and recovery or grain growth,
the UFG structure was first stabilized in the single phase
domain, at 380 °C for 2 h. As shown in Fig. 7, such treat-
ment gives rise to a significant grain growth. However,
nanoscaled Al3Sc and Al–Mn intermetallic particles that
clearly appear in the HAADF image (Fig. 7b), obviously
pin GBs, constraining the grain size in the range of 1–
2 lm. Some grayscale fluctuations on a large length scale
(a few lm) are clearly noticeable on the HAADF image.
They are typical of sample thickness modulations and can-
Fig. 6. 3-D reconstruction showing a triple junction in a volume analyzed not be attributed to Mg concentration fluctuations. This
by APT in the AlMg alloy severely deformed at 200 °C: (a) Al map where point was confirmed by EDS analyses. Besides, higher
(1 1 1)fcc Al atomic planes are clearly exhibited; (b) Mg map with arrows magnification images did not reveal any dark clusters, thus
corresponding to the location of concentration profiles; (c) Mg concen-
tration profile showing a Mg depletion along the first GB; (d) Mg
it is believed that nanoscaled Mg-rich domains formed dur-
concentration profile showing locally strong Mg segregation along the ing HPT were dissolved during annealing. These observa-
second GB (sampling volume thickness, 1 nm). tions are consistent with the Al–Mg binary phase
diagram [32]. Indeed, at this temperature the nominal Mg
concentration (5.7 wt.%) is below the solubility limit.
where Xc is the mean Mg concentration of the clusters/seg- Then, to trigger the nucleation of the b or b0 phases, the
regations, Xm the mean Mg concentration of the matrix material was further aged at 175 °C for 15 days. As shown
and Xo the nominal Mg concentration of the alloy. in the HAADF image (Fig. 8a), it does not lead to a signif-
In the material processed by HPT at room temperature, icant grain growth but copious nucleation of Mg-rich pre-
Mg-rich clusters appear rather well distributed within the cipitates occurred. Most of them are located along GBs,
X. Sauvage et al. / Acta Materialia 72 (2014) 125–136 131

Fig. 8. AlMg alloy severely deformed at room temperature, subsequently


annealed for 2 h at 380 °C, water-quenched and further annealed at 175 °C
for 15 days: (a) STEM-HAADF image showing extensive b’ phase
precipitation along GBs; (b) EDS map (Al blue, Mg red and Mn green)
showing an Al6Mn precipitate at a triple junction and b’ phase that has
nucleated along GBs; (c) HRTEM image showing the interface between
the fcc Al matrix and b’ phase; (d) corresponding SAED pattern where
typical reflexions of the b’ and fcc Al phases are indexed, showing their
orientation relationship. (For interpretation of the references to color in
this figure legend, the reader is referred to the web version of this article.)

Fig. 7. AlMg alloy severely deformed at room temperature and subse- ð1 1 1Þb0 ==ð0 0 2ÞAl and ½1 3 1b0 ==½1 1 1Al
quently annealed for 2 h at 380 °C: (a) low magnification bright field
STEM image showing grain boundaries; (b) corresponding HAADF Probably only nuclei exhibiting such a specific OR may
image showing particles in bright (Al3Sc and Al6Mn) – no Mg-rich clusters overcome the nucleation barrier and grow. Thus GBs free
or segregations can be identified.
of precipitates are probably locations where this OR is
difficult or impossible to match.
but also along Al/Al6Mn interfaces and occasionally inside
grains. It is interesting to note that few GBs are not 4. Discussion
covered by these precipitates, and that there is a large dis-
persion of the layer covering GBs (from 10 to 100 nm). The 4.1. Specificities of the decomposition of the supersaturated
Mg concentration in these precipitates was measured using solid solution during HPT
EDS analyses (Fig. 8b) and was found to lie in the range of
35–40 at.%, close to the value expected for the b or b0 Obviously, Mg-rich clusters and segregations formed
phases (Al3Mg2). Besides, the amount of Mg in the sur- during SPD are very different in nature compared to the
rounding matrix was measured in the range of 1–2 at.%, b0 precipitates that nucleate during the thermal annealing
corresponding to a volume fraction of b phase in the range treatment at 175 °C. It is difficult to compare volume
of 10–15% (estimated from Eq. (1)). High resolution TEM fractions and sizes of precipitates or clusters because these
(Fig. 8c) and selected area electron diffraction (Fig. 8d) features are typically time-dependent in any precipitation
were carried out to check the crystallographic structure process. However, Mg-rich clusters in the material
of these precipitates and to determine their orientation processed by HPT at room temperature: (i) are more
relationship (OR) with the matrix. The metastable hexago- homogeneously distributed (not only along GBs); (ii) do
nal b0 phase was clearly identified and the OR with the fcc not exhibit a crystallographic structure different from the
Al matrix is: fcc Al matrix with a specific OR; and (iii) exhibit a large
132 X. Sauvage et al. / Acta Materialia 72 (2014) 125–136

scatter in Al/Mg ratio with local Mg concentrations much the typical grain size exhibited in STEM-BF images
lower than the b0 phase. Besides, Mg-rich layers along GBs (Fig. 9b) is significantly larger than in the Al–Mg alloy
in the material processed by HPT at 200 °C: (i0 ) are signif- (Fig. 1b); it is typically in the range of 150–300 nm. But
icantly thicker than the GB segregations observed in the it is very interesting to note that Mg-rich clusters (darkly
material annealed at 380 °C; (ii0 ) are much more numerous; imaged in Fig. 9a) were also formed during SPD in the
(iii0 ) exhibit a higher Mg content (for some of them, vicinity of GBs. Thus, even with a much lower driving force
locally); and (iv0 ) do not exhibit a crystallographic structure for the decomposition of the supersaturated solid solution
different from the fcc Al matrix with a specific OR. (because of the lower Mg concentration), these atoms dif-
All these differences clearly indicate that the mechanisms fuse and agglomerate during SPD.
leading to the decomposition of the supersaturated solid
solution during SPD are different from those operating 4.2. Mg enhanced diffusion during SPD
during a conventional thermally activated process. To
check that these observed phenomena are not specific to The decomposition of the Al–Mg supersaturated solid
the Al–Mg commercial alloy of the present study, a model solution, even if it is only partial, requires a significant
alloy containing only 2 at.% Mg was solutionized and pro- mobility of Mg atoms. From the STEM-HAADF images
cessed by HPT at room temperature in exactly the same (Figs. 1a and 4a) it seems realistic to assume that in average
conditions. As expected, due to the lower Mg content, Mg atoms have to diffuse over a typical distance of about
k  d/4 where d is the mean grain size. Within the random
walk theory of diffusion, the effective diffusion distance can
be written as [46]:
keff ¼ ð6DtÞ1=2 ð4Þ
where D is the diffusion coefficient and t the time. If one
writes k = keff in Eq. (4), then it is possible to estimate
the equivalent diffusion coefficient Deq during the SPD pro-
cess. For HPT at 200 °C where the grain size d is 300 nm
and the total deformation time t = 1200 s, then it yields Deq
19
200°C  8  10 m2 s1. This is fully consistent with the
typical diffusion coefficient of Mg in fcc Al at 200 °C
reported in the literature [47]. However, for HPT at room
temperature where the grain size d is 150 nm, it yields
Deq RT  2  1019 m2 s1. Such a value of the thermal dif-
fusion coefficient of Mg in Al is typically reached for a tem-
perature of 180 °C [47].
Even if a significant part of the work required for the
HPT process is dissipated into heat by the plastic deforma-
tion of the material, it was shown both experimentally and
numerically that the temperature increase is very limited
due to the high thermal conductivity of anvils [48]. Thus,
even for a large number of revolutions, the temperature
increase of aluminum processed at a speed of 1 rev min–1
should not exceed 20 °C (i.e. from room temperature to
40 °C) [48]. Such a small change in temperature can obvi-
ously not account for the enhanced atomic mobility of Mg
atoms. However, as demonstrated in earlier works, a large
density of vacancies is often created in metallic alloys pro-
cessed by SPD at low homologous temperature, with con-
centrations up to CSPD  105 [49–55]. Since Mg atoms are
substitutional atoms in the fcc Al matrix, these SPD-
induced vacancies could significantly affect the atomic
mobility, and the diffusion coefficient should rather be
written as:
D ¼ ðC 0 þ C SPD ÞD=C 0 ð5Þ
Fig. 9. Al–2 Mg alloy homogenized and severely deformed at room
temperature: (a) STEM HAADF image showing Mg-rich clusters (darkly
where C0 is the equilibrium vacancy concentration and D
imaged); (b) corresponding STEM dark-field image where diffraction the diffusion coefficient of Mg in fcc Al. From Ref. [56],
contrast reveals grains boundaries. the vacancy concentration in fcc Al can be written as
X. Sauvage et al. / Acta Materialia 72 (2014) 125–136 133

C 0 ¼ 11:47 expðEf =k B T Þ
where Ef is the formation enthalpy of vacancies in
aluminum (0.66 eV), kB the Boltzmann’s constant and T
the temperature. From Ref. [47], the diffusion coefficient
of Mg in fcc Al can be written as
D ¼ D0 expðQ=RT Þ
where D0 = 1.49  105 m2 s1, Q = 120.5 kJ mol1,
1 1
R = 8.314 J mol K and T is the temperature. Both D
and D* are plotted in Fig. 10 and compared with the equiv-
alent diffusion coefficient estimated from Eq. (4). At low Fig. 11. Critical dislocation velocity V for Mg solute drag as a function of
temperature, the equilibrium vacancy concentration is the temperature, estimated from Refs. [73]; see text for details. The critical
extremely low (typically 1011); thus the diffusion coeffi- dislocation density V* is shifted to the left when the SPD enhanced
cient is dramatically affected by SPD-induced extra vacan- diffusion coefficient is taken into account. The vertical dashed lines
correspond to the two different processing temperatures (RT and 200 °C).
cies while at higher temperatures D* converges toward the The horizontal dashed line at 104 m s–1 is an estimated lower bond of the
conventional diffusion coefficient. Anyway, it is interesting mean dislocation velocity during HPT (see text for details), showing that
to note that both Deq RT and Deq 200°C are rather close to D* solute drag was negligible during the process.
estimated from Eq. (5), confirming that SPD-induced
vacancies could play a significant role in the Mg atoms’
mobility.
dc=dt ¼ ð2pr=60hÞ ð7Þ
Dislocations that are continuously nucleated and that
propagate during deformation are another media that where r is the distance from the disc center (r  5 mm) and
could promote solute drag [57,58]. The critical velocity V, h is the sample thickness under loading (h  1 mm). Then,
above which moving dislocations do not drag Mg atoms, using Orowan’s equation one may estimate the mean
can be estimated as follows [57,58]: dislocation density Vm:
V ¼ ð3 þ 220:5 ÞAD=ð2b2 k B T Þ ð6Þ dc=dt ¼ /bqV m ð8Þ

where b the Burgers vector (b  2  1010 m), A the elastic where / is the Schmid factor (to get a lower bound of Vm, it
energy interaction that can be written as 3lbXe/p, with l is assumed to be equal to 0.5) and q the density of mobile
the shear modulus of fcc Al (l  27 GPa), X the atomic dislocations. The typical dislocation density reported for
volume in the fcc Al X  1.6  1029 m3), e the misfit (an SPD metallic alloys is usually in the range of 1012–
estimation based on the lattice expansion resulting from 1014 m2; using this later value to estimate a lower bound
Mg in solid solution leads to e  10%). This critical dislo- of the dislocation mean velocity during HPT leads to
cation velocity is plotted as a function of the temperature Vm > 104 m s–1. As shown in Fig. 11, both at room tem-
in Fig. 11, taking both D and D* estimated from Eq. (5) perature and at 200 °C, the critical dislocation velocity
to visualize the influence of the enhanced atomic mobility for Mg drag lies well below this lower bond value, even if
induced by SPD. If the mean dislocation velocity is well the enhanced diffusion estimated by Eq. (5) is taken into
above the curve, then it means that dislocations cannot effi- account (V*). Thus, one may conclude that the Mg
ciently drag Mg atoms. The HPT process, both at RT and enhanced atomic mobility during HPT at room tempera-
200 °C, was performed with a rotation speed of 1 rpm, ture is predominantly mediated by SPD-induced vacancies,
leading to a strain rate in shear dc/dt that can be written as: while at 200 °C little effect, if any, of these extra vacancies
and of dislocations is expected.

4.3. Mg clustering against nucleation of b phase

In both cases, during SPD at room temperature or at


200 °C, it seems clear that the atomic mobility of Mg is suf-
ficiently high to promote clustering, segregation or precip-
itation of Al–Mg intermetallic phases. In some equilibrium
conditions, this latter situation should be expected because
at these two temperatures, the Mg concentration in solid
solution is significantly higher than the solubility limit
[32]. However, experimental data (TEM, HAADF, STEM
Fig. 10. Diffusion coefficient D of Mg in Al plotted as a function of
and APT) did not reveal such phases directly in the as-
temperature (from Ref. [63]) and diffusion coefficient D* estimated from
Eq. (6), taking into account the contribution of SPD-induced vacancies. deformed alloy, but only after a post-deformation aging
Deq RT and Deq 200°C are the equivalent diffusion coefficients during SPD at 175 °C (Fig. 8). Nevertheless, it is well known that the
estimated from Eq. (5) and experimental data. nucleation barrier of the Al3Mg2 phase is rather high and
134 X. Sauvage et al. / Acta Materialia 72 (2014) 125–136

that is why it usually nucleates almost exclusively along


GBs or along Al6Mn precipitates [34,45,59]. Thus, it could
be that the deformation time of only 20 min was not long
enough to overcome massively the nucleation barrier neces-
sary for a structural transformation. Besides, as shown in
the material aged for 15 days at 175 °C, a special OR
between the Al3Mg2 precipitates and the fcc Al matrix is
required at GBs for nucleation and growth. During SPD,
some dislocations are continuously generated and most of
them dynamically recover at the newly formed GBs, lead-
ing to a progressively change in their misorientation
through grain rotation [60]. In such a situation, if not Fig. 12. Schematic representation of the diffusion of Mg toward GBs
assisted by the vacancy flux (Jv). This flux of vacancies is created by the
impossible, it is probably very unlikely that a steady state large level of plastic deformation, while GBs act as a sink. As a result, a
could be reached and that some GBs could allow the nucle- vacancy gradient and thus a vacancy flux is created from the grain interior
ation and growth of the Al3Mg2 with the required OR. toward GBs. Since Mg atoms exhibit a positive binding energy with
Moreover, it has been demonstrated that the Al3Mg2 phase vacancies, they are strongly coupled to these defects and a collective
could be quite easily decomposed by SPD, both during ball motion of Mg atoms occurs. At the end, when vacancies recombine and
annihilate at the GBs, Mg atoms are left, forming clusters or segregations.
milling [61,62] or in shear bands [63]. Then, Mg-rich clus-
ters or segregations observed in the SPD processed Al–
Mg alloy could be also intermetallic particles that have irradiation [25,65]. As represented in Fig. 12, Mg atoms
nucleated along GBs in the early stage of deformation being strongly coupled to vacancies, the flux of vacancies
and later sheared and subsequently decomposed during toward GBs drags Mg atoms during the plastic deforma-
the HPT process. One should note that similar deforma- tion, leading to strong local enrichments. Using high preci-
tion-driven cyclic phenomena of nucleation and decompo- sion dilatometry measurements, Würschum et al. have
sition have been reported for various metallic systems recently estimated that the typical free volumes in metals
processed by ball milling [64]. processed by SPD are 103 [66]. If it is assumed that these
free volumes are exclusively created by the agglomeration
4.4. The role of SPD induced vacancies on Mg clustering and of vacancies coming from the grain interiors and if each
segregations of those vacancies successfully drags one Mg atom, then
a fraction of 103 Mg (0.1 at.%) could be extracted from
It is, however, interesting to note that in any case, the matrix. Our measurements show that it is in fact 5 or
Mg-rich clusters and/or segregations have been observed 10 times higher (see Table 1) but this difference can obvi-
mainly along boundaries or in their vicinities. It seems ously be explained by the massive recombination and anni-
therefore that there is a non-isotropic flux of Mg atoms hilation of point defects at GBs. However, if we still
and that these solute atoms are dragged toward GBs in assume that each vacancy diffusing to a GB manages to
the course of SPD. During deformation of fcc aluminum, drag one Mg atom, it yields that at least about Nv = 102
a large number of dislocations is created and their motion vacancies were created during the HPT process (leading
is the medium for plastic deformation. Most of them are to a decrease of the mean Mg concentration in the matrix
absorbed along boundaries where they dynamically of 1 at.%). The SPD vacancy production rate Vp being
recover, so there is a constant flow of dislocations toward the number of vacancies created per second, it can be writ-
GBs. In principle, these dislocations may drag Mg atoms ten as:
[5,57,58]; however, as demonstrated earlier, the strain rate V p ¼ N v =t ð9Þ
during HPT is obviously too high and most of the disloca-
tions (if not all) move free of solute atmospheres. On the where t is the duration of the HPT process (t = 1200 s),
other hand, the large density of vacancies created during leading to Vp  105 s1. This estimate is consistent with
SPD [49–55] may affect also the distribution of solute previous estimations performed from SPD-induced
atoms. Like in irradiated materials, GBs are sinks for chemical mixing in the Cu–Fe system deformed in similar
vacancies (even if some of them are also probably annihi- conditions by HPT [67].
lated along dislocations), which leads to an anisotropic Recently, using Small angle X-ray scattering measure-
vacancy flux toward GBs. As discussed above, the diffusion ments, it has been shown that quenched vacancies in pure
of Mg atoms is driven by a vacancy mechanism but one aluminium form nanoscaled voids with a typical radius of
should note that Mg atoms in solid solution in fcc Al are 2 nm upon low temperature annealing (60–90 °C)
strongly coupled to vacancies, with a positive binding [68,69]. It is interesting to note that this size is consistent
energy in the range of 0.2–0.4 eV [8,40,41]. Thus, the with the size of the Mg clusters found in the present Al–
atomic scale mechanism of the SPD-induced segregation Mg alloy processed by SPD at room temperature
observed in the present study is probably very similar to (Fig. 2a). This may indicate that these Mg-rich clusters
that of non-equilibrium segregation resulting from could be directly linked to the agglomeration of vacancies.
X. Sauvage et al. / Acta Materialia 72 (2014) 125–136 135

Besides, the same authors have also shown that vacancy (vi) It was demonstrated that the equilibrium Al3Mg2
voids in pure Al dissolve at a temperature higher than phase typically nucleates and grows at GBs with a
145 °C [69]. This feature could explain why the Mg very specific orientation relationship with the fcc Al
distribution in the material processed by SPD at 200 °C matrix. Thus it is emphasized that the continuous
is significantly different with mostly grain boundary segre- evolution of GBs created during the SPD process
gations (Fig. 4a). Moreover, it is well known also that in a may not provide these conditions.
temperature range of 80–120 °C, vacancies can free them- (vii) It was demonstrated that the mobility of Mg atoms is
selves from traps [41], leaving more freedom to Mg atoms probably enhanced by vacancies created by SPD,
that could reorganize in the vicinity of the GBs. Of course, while solute drag by dislocations appeared very
dislocations and dislocation debris stored along boundaries unlikely.
should play an important role in stabilizing locally Mg (viii) SPD-induced vacancies could annihilate at GBs,
atoms. Indeed, as reported for aged dilute Al–Mg alloys, leading to a vacancy gradient in grains and a vacancy
magnesium could easily segregate along dislocation loops flux toward GBs. Since Mg atoms do exhibit a posi-
[8]. Besides, GBs created during SPD processes are often tive binding energy with these vacancies, they are
called “non-equilibrium grain boundaries” because they dragged by this flux toward GBs.
exhibit large local distortions (due to extrinsic dislocations) (ix) It may be concluded that SPD leads to new type of
and enhanced free volumes [9,70]. Mg atoms being larger dynamic aging when the Al–Mg alloy exhibits some
than Al, some elastic strain could be relaxed by Mg features of age-hardenable alloy behavior as the for-
segregations, giving rise to local compressive stresses and mation of deformation-induced segregations and
stabilizing GBs by reducing their energy [71,72]. clusters of alloying elements which can contribute
Summarizing the results of investigation on the alloy to the alloy’s hardening.
subjected to SPD, it should be mentioned that the observed
phenomena may be interpreted as new mechanisms for
dynamic aging of Al–Mg alloys, making their behavior Acknowledgements
similar to age-hardenable alloys. Indeed, the shown results
allow us to state that Al–Mg alloys deformed below 200 °C The authors acknowledge the financial support of the
may be hardened not only at the expense of work-harden- ERA-NET.RUS Namastreco STP #122 project. Prof. D.
ing (as a non-age-hardenable alloy) but also by forming Embury is also gratefully acknowledged by X.S. for fruitful
segregations [24] and clusters of alloying elements similar discussions. R.Z., N.E. and M.M. acknowledge support by
to that on the precipitation stage in age-hardenable alloys. the Russian Ministry for Education and Science through
The most striking difference is that the products of solid Contract No. 14.B25.31.0017 (28 June 2013).
solution decomposition in the Al–Mg alloy in the course
of SPD tend to inhibit grain boundary area. This References
observation testifies that SPD may be used to produce
UFG materials of a new structural design, changing our [1] Waldron GWJ. Acta Metall 1965;13:897.
understanding of dynamic aging of non-hardenable alloys. [2] Hughes DA. Acta Metall Materi 1993;41:1421.
[3] Atodiresei M, Gremaud G, Schaller R. Mater Sci Eng A
2006;442:160.
5. Conclusions [4] Zolotorevsky NYu, Solonin AN, Churyumov AYu, Zolotorevsky VS.
Mater Sci Eng A 2009;502:111.
[5] Van den Beukel A. Phys Status Solidi A 1975;30:197.
[6] Picu RC, Zhang D. Acta Mater 2004;52:161.
(i) The distribution of Mg atoms in solid solution in a [7] Takamura Jin-Ichi, Takahashi Isao, Amano Muneyuki. Trans ISIJ
homogenized 1570 aluminum alloy was found to be 1969;9:216.
strongly affected by SPD. [8] Embury JD, Nicholson RB. Acta Metall 1963;11:347.
[9] Valiev RZ, Islamgaliev RK, Alexandrov IV. Prog Mater Sci
(ii) After HPT at room temperature, Mg-rich clusters 2000;45:103.
(concentration in the range of 10–20 at.%) with a size [10] Ito Y, Horita Z. Mater Sci Eng A 2009;503:32.
of 5–10 nm were observed near GBs. [11] Zhilyaev AP, Oh-ishi K, Langdon TG, McNelley TR. Mater Sci Eng
(iii) If the process is carried out at 200 °C, then Mg segre- A 2005;410–411:277.
gations (concentration in the range of 10–15 at.%) [12] Horita Z, Langdon TG. Mater Sci Eng A 2005;410–411:422.
[13] Sakai G, Horita Z, Langdon TG. Mater Sci Eng A 2005;393:344.
with a thickness of 10 nm were observed along a [14] Morris DG, Munoz-Morris MA. Acta Mater 2002;50:4047.
significant fraction of GBs and triple junctions. [15] Morishige T, Hirata Y, Uesugi T, Takigawa Y, Tsujikawa M, Higashi
(iv) These clusters and segregations do not exhibit the K. Scripta Mater 2011;64:355.
composition nor the cyrstallographic structure of [16] Iwahashi Y, Horita Z, Nemoto M, Langdon TG. Metall Mater Trans
the equilibrium Al3Mg2 phase. A 1998;29:2503.
[17] Prangnell PB, Bowen JR, Apps PJ. Mater Sci Eng A 2004;375–
(v) Preliminary data also show that in an alloy contain- 377:178.
ing significantly less Mg in solid solution (2 at.%) [18] Murashkin MYu, Kilmametov AR, Valiev RZ. Phys Metal Metall
and no other elements, similar phenomenon occur. 2008;106:90.
136 X. Sauvage et al. / Acta Materialia 72 (2014) 125–136

[19] Malis T, Chaturvedi MC. J Mater Sci 1982;17:1479. [46] Zhu Y. Metall Mater Trans A 2012;43:4933.
[20] Paine DC, Weatherly GC, Aust KT. J Mater Sci 1986;21:4257. [47] Cahn RW, Hasen P. Physical metallurgy. 3rd ed. Amsterdam: North-
[21] Liu X-Y, Adams JB. Acta Mater 1998;46:3467. Holland; 1983. p. 385.
[22] Liu X, Wang X, Wang J, Zhang H. J Phys: Condens Matter [48] Yong Dua, Chang YA, Baiyun Huang, Weiping Gong, Zhanpeng Jin,
2005;17:4301. Honghui Xu, et al. Mater Sci Eng A 2003;363:140.
[23] Razumovskiy VI, Ruban AV, Razumovdkii IM, Lozovoi AY, Butrim [49] Edalati K, Miresmaeili R, Horita Z, Kanayama H, Pippan R. Mater
VN, Vekilov YK. Scripta Mater 2011;65:926. Sci Eng A 2011;528:7301.
[24] Valiev RZ, Enikeev NA, Murashkin MY, Kazykhanov VU, Sauvage [50] Quelennec X, Menand A, Le Breton JM, Pippan R, Sauvage X.
X. Scripta Mater 2010;63:949. Philos Mag 2010;90:1179.
[25] Lejcek P. Grain boundary segregations in metals. Berlin: Springer- [51] Setman D, Schafler E, Korznikova E, Zehetbauer M. Mater Sci Eng
Verlag; 2010. A 2008;493:116.
[26] Sauvage X, Ganeev A, Ivanisenko Y, Enikeev N, Murashkin M, [52] Van Petegem S, Dalla Torre F, Segers D, Van Swygenhoven H.
Valiev R. Adv Eng Mater 2012;14:968. Scripta Mater 2003;48:17.
[27] Vo NQ, Schäfer J, Averback RS, Albe K, Ashkenazy Y, Bellon P. [53] Würschum R, Greiner W, Valiev RZ, Rapp M, Sigle W, Schneeweiss
Scripta Mater 2011;65:660. O, et al. Scripta Metall 1991;25:2451.
[28] Zhang XF, Fujita T, Pan D, Yu JS, Sakurai T, Chen MW. Mater Sci [54] Saada G. Physica 1961;27:657.
Eng A 2010;527:2297. [55] Ruoff AL, Balluffi RW. J Appl Phys 1963;34:2862.
[29] Schäfer J, Stukowski A, Albe K. Acta Mater 2011;59:2957. [56] Mecking H, Estrin Y. Scripta Metall 1980;14:815.
[30] Zhang HW, Huang X, Pippan R, Hansen N. Acta Mater [57] Simmons RO, Balluffi RW. Phys Rev 1960;117:52.
2010;58:1698. [58] Kazantzis AV, Chen ZG, De Hosson JThM. J Mater Sci
[31] Hegedus Z, Gubicza J, Kawasaki M, Chinh NQ, Fogarassy Z, 2013;48:7399.
Langdon TG. Mater Sci Eng A 2001;528:8694. [59] Soer WA, Chezan AR, De Hosson JThM. Acta Mater 2006;54:3827.
[32] Rajgarhia RK, Saxena A, Spearot DE, Hartwig KT, More KL, [60] Goswami R. Metall Mater Trans A 2013;44:1279.
Kenik EA, et al. J Mater Sci 2010;45:6707. [61] Tóth LS, Beausir B, Gu CF, Estrin Y, Scheerbaum N, Davies CHJ.
[33] Straumal BB, Baretzky B, Mazilkin AA, Phillipp F, Kogtenkova OA, Acta Mater 2010;58:6706.
Volkov MN, et al. Acta Mater 2004;52:4469. [62] Scudino S, Sakaliyska M, Surreddi KB, Eckert J. J Phys: Conf Ser
[34] Straumal BB, Sauvage X, Baretzky B, Mazilkin AA, Valiev RZ. 2009;144:012019.
Scripta Mater 2014;70:59. [63] Morris DG. Acta 2002;50:4047.
[35] Goswami R, Spanos G, Pao PS, Holtz RL. Mater Sci Eng A [64] Komiya Y. J Jpn Inst Light Metals 1980;30:271.
2010;527:1089. [65] Suryanarayama C. Prog Mater Sci 2001;46:1.
[36] Starink MJ, Zahra A-M. Philos Mag A 1997;73:701. [66] Etienne A, Radiguet B, Cunningham NJ, Odette GR, Pareige P. J
[37] Osamura K, Ogura T. Metall Trans A 1984;15:835. Nucl Mater 2010;406:244.
[38] Sato T, Kojima Y, Takahashi T. Metall Trans A 1982;13:1373. [67] Würschum R, Oberdorfer B, Steyskal E-M, Sprengel W, Puff W,
[39] Sauvage X, Dédé A, Cabello Muñoz A, Huneau B. Mater Sci Eng A Pikart P, et al. Phys B: Condens Matter 2012;407:2670.
2008;491:364. [68] Sauvage X, Wetscher F, Pareige P. Acta Mater 2005;53:2127.
[40] Da Costa G, Vurpillot F, Bostel A, Bouet M, Deconihout B. Rev Sci [69] Westfall L, Diak BJ, Singh MA, Saimoto S. J Eng Mater Technol
Instrum 2005;76:013304. 2008;130:021011.
[41] Beatrice CRS, Garlipp W, Cilense M, Adorno AT. Scripta Metall [70] Chaudhuri A, Singh MA, Diak BJ, Cuoppolo C, Woll AR. Philos
Mater 1995;32:23. Mag 2013;93:4392.
[42] Panseri C, Gatto F, Federighi T. Acta Metall 1958;6:198. [71] Sauvage X, Wilde G, Divinsky S, Horita Z, Valiev RZ. Mater Sci Eng
[43] Searles JL. Metall Mater Trans A 2001;32:2859. A 2012;540:1.
[44] Sha G. Ultramicroscopy 2011;111:500. [72] Kirchheim R. Acta Mater 2007;55:5129.
[45] Itoh G. Mater Trans JIM 1990;31:1041. [73] Kirchheim R. Acta Mater 2007;55:5139.

Vous aimerez peut-être aussi