Available online at www.sciencedirect.com

www.elsevier.com/locate/actamat

ScienceDirect

Acta Materialia 66 (2014) 105–117

Precipitation kinetics in a severely plasticallydeformed 7075 aluminium alloy

A. Deschamps a,⇑, F. De Geuser a, Z. Horita b,c, S. Lee b,c, G. Renou a

a SIMAP, Grenoble INP–CNRS–UJF, BP 75, 38402 St Martin d’Heres Cedex, Franceb Department of Materials Science and Engineering, Faculty of Engineering, Kyushu University, Fukuoka 819-0395, Japan

c WPI, International Institute for Carbon-Neutral Energy Research (WPI-I2CNER), Kyushu University, Fukuoka 819-0395, Japan

Received 17 September 2013; received in revised form 28 November 2013; accepted 29 November 2013

Abstract

In this paper we report a quantitative study, using small-angle X-ray scattering, of the precipitation kinetics during ramp heating andisothermal ageing in an AA7075 aluminium alloy processed by high-pressure torsion. The precipitation behaviour has been comparedwith that of the same material processed in a conventional manner and observations are supplemented by transmission electronmicroscopy for precipitate and grain size characterization using automated crystal orientation mapping. After severe plastic deformationand natural ageing, the material is shown to contain a high density of GP zones. During ageing, the precipitate size distribution becomesbimodal, with small precipitates behaving similarly to those of the conventionally processed material and large ones associated with thecrystalline defects and reaching large sizes at considerably lower temperatures and shorter times as compared to the conventionallyprocessed material.� 2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Aluminium alloys; Severe plastic deformation; High pressure torsion; Precipitation; Small-angle X-ray scattering

1. Introduction

Severe plastic deformation (SPD) is now a well-estab-lished way to make materials with extremely small grainsizes and resulting high strength [1]. Using a great varietyof processes (the most studied being equal-channel angularpressing (ECAP) [2] and high-pressure torsion (HPT) [3],alongside many others, such as accumulative roll bonding,cryorolling and multiaxial channel compression), grain sizeis reduced to a range between 100 and 500 nm, whichresults in materials with very high strength.

In aluminium alloys, the conventional way of achievinghigh strength is through fine-scale precipitation. Thereforeit is not surprising that studying SPD in precipitation-strengthened aluminium alloys has attracted a largeamount of interest in the last 10 years [4–31], with the

1359-6454/$36.00 � 2013 Acta Materialia Inc. Published by Elsevier Ltd. All

http://dx.doi.org/10.1016/j.actamat.2013.11.071

⇑ Corresponding author.E-mail address: alexis.deschamps@grenoble-inp.fr (A. Deschamps).

aim of achieving combined strengthening between smallgrain sizes and precipitates (a yield strength of 1 GPa hasbeen achieved in an Al–Zn–Mg–Cu alloy processed byHPT [32]) and increased stability in the submicron grainsize by precipitate pinning [27].

Studying the combination of precipitation and SPD is acomplex topic given the number of parameters that can bechanged. Notwithstanding the variety of existing SPD pro-cesses, SPD can be carried out on a random solid solution(right after quenching from a solution treatment) or on amicrostructure already containing precipitates. It can becarried out at cryogenic temperatures (note that only alimited number of processes allow for this), at room tem-perature or at temperatures where classical precipitationtreatments are carried out (typically 100–200 �C). Notethat the low temperature processes can usually only be car-ried out on solution-treated materials, otherwise specimenfracture during SPD is difficult to avoid. SPD can be fol-lowed by a subsequent ageing treatment, with or without

rights reserved.

106 A. Deschamps et al. / Acta Materialia 66 (2014) 105–117

intermediate solution treatment. And of course the resultsare alloy-dependent, with studies carried out on five alloyfamilies: 7000 (Al–Zn–Mg–Cu) [4–9,15,24,27,29,31,32],6000 (Al–Mg–Si) [10,14,17,23,25], Al–Cu-based 2000[12,18,20,22,26], Al–Cu–Mg-based 2000 [28] and Al–Cu–Li-based 2000 [13,19,21].

Despite this complexity, some general rules can be sum-marized from the now relatively abundant literature:

– If precipitates are present before SPD, several phenom-ena can take place [13,14,16,18,20,22]: they can beprogressively fragmented and even dissolve during thedeformation process, resulting in a state close to asolid solution, or continue to precipitate while beingdeformed together with the matrix. These processesand the competition between them are highly tempera-ture- and strain-dependent.

– The microstructure resulting from room temperatureSPD carried out on solution-treated materials has notyet been fully characterized since the very fine grainstructure prevents the observation of extremely smallobjects with conventional tools such as electron micros-copy. Some papers invoke the absence of GP zones fol-lowing room temperature SPD [6] while others evidencethe presence of clusters or GP zones [28,29].

– If precipitates are not present initially and SPD is car-ried out at elevated temperature, accelerated dynamicprecipitation occurs during SPD at a rate one or twoorders of magnitude faster than corresponding precipi-tation in the absence of SPD [8,13,15,23–25,29].

– Precipitation following SPD occurs much faster thanprecipitation in the coarse-grained (CG) counterparts[7,10,19], and in many cases the intermediate metastablephases are skipped so that the equilibrium phase isformed at much lower temperatures than in conven-tional ageing treatments [12,18,22,24,26]. Among thereported cases, the extensive formation of h phase inSPD Al–Cu during a few months at room temperatureis particularly remarkable [12]. Many other reports existon the formation of h at the grain boundaries instead ofthe metastable h phase during medium temperatureageing [18,22,26]. Other systems may, however, behavedifferently. No extensive g0 phase formation has beenreported yet at room temperature following SPD onAl–Zn–Mg–Cu alloys, even though it is observed toform extensively at structural defects during elevatedtemperature ageing [29,31]. In AlCuLi alloys reports ofh0 and T1 formation exist, similar to what is found incoarse-grained materials [19,21].

– In parallel to all these phenomena, the mechanicalbehaviour is usually followed by hardness measure-ments. After moderately large deformation processes(e.g. cryorolling), materials present a strengtheningbehaviour similar to that of conventional ageing, butaccelerated to shorter times [7]. After SPD, however,the hardness is already very high (�200 HV for mostprecipitation-strengthened Al alloys) so that no strong

further increase can be obtained. Depending on thepost-SPD ageing temperature (from room temperatureupwards) and the alloy family, in some cases an addi-tional strengthening of �50 HV is found (Al–Zn–Mg–Cu[4,5] and Al–Cu–Li [19]); in other cases the hardnessremains stable and at higher temperatures hardnessdecreases continuously, because strengthening is reducedby precipitate coarsening, a concurrent reduction indefect density (dislocations, microstrains) and by graingrowth [7,9,17,19,26,29,33]. Several authors have pro-posed strategies to optimize the combination of strengthand ductility by acting on the SPD parameters andsubsequent ageing treatment [11,21].

Several critical parameters are invoked when discussingthe effect of SPD on precipitation, even though there hasbeen a lack of quantitative understanding until now. Diffu-sion rates many orders of magnitude above the equilibriumones need to be invoked to account for the rate of forma-tion of the precipitates. This has been attributed to extre-mely high vacancy concentrations due to the SPD process[34,35], and concentrations of 10�5–10�4 have been postu-lated [34]. However, such numbers are difficult to verify,since the positron annihilation technique, for instance, issensitive in such materials to all present structural defectsand proves to be difficult to interpret based on vacancy sol-ute interactions alone [36]. Additionally, since in manycases precipitation occurs together with some grain growth,it has been postulated that grain boundary motion, sweep-ing solute from the matrix, may play a strong role in accel-erating precipitation [12].

Until now, the study of precipitation phenomena inSPD materials has been almost exclusively carried outusing transmission electron microscopy (TEM), with asso-ciated indirect techniques such as differential scanningcalorimetry (DSC) and X-ray diffraction (DRX), and occa-sionally atom probe tomography (APT) observations.However, given the very small size of the crystallites pres-ent (�100 nm), the presence of high levels of microstrainsand a high density of dislocations, characterizing in detailmicrostructural states including very small precipitates,proves to be extremely difficult [10], and the interpretationof the evolution of such precipitates with time or tempera-ture remains necessarily qualitative, except for later stagesof ageing where the objects become sufficiently large to beeasily observed.

Small-angle X-ray scattering (SAXS) has been exten-sively used in precipitation-hardened Al alloys (particularlyin 7000 and 2000 series) to obtain a quantitative measure-ment of precipitation microstructures, in terms of both sizeand volume fraction, and allowing for in situ measure-ments along isothermal or non-isothermal heat treatments[37,38]. Since this technique is sensitive to spatial variationsof electron density (therefore of chemistry) within the sam-ple, the signal is dominated by the precipitate microstruc-ture, and the contribution of structural defects is usuallynegligible (except in low-contrast, low-volume fraction

A. Deschamps et al. / Acta Materialia 66 (2014) 105–117 107

situations; see e.g. Ref. [39]). This technique is particularlysuited to observing nanometre-scale precipitation, fromclusters containing a few atoms to precipitates of 10–50 nm.

The aim of the present paper is to clarify the precipita-tion process in SPD materials by a quantitative evaluationof the decomposition kinetics following the SPD process.For this purpose we have chosen to restrict ourselves towhat seems to be the simplest situation, namely:

– An alloy with precipitates of low aspect ratio with highcontrast in atomic number with respect to the matrix, sothat they are easily measured by SAXS and dominatethe signal as compared to other sources of small-anglescattering in the SPD material. Al–Zn–Mg–Cu alloysare best suited for this purpose as they have been exten-sively studied by SAXS before [40–44]. Among thesealloys, the alloy family AA7075 has been the subjectof extensive studies by SPD [29,31] and thus was ourchoice.

– An SPD process carried out on the material being insolid solution so that as little precipitation as possibleis present prior to further heat treatments.

– An SPD process that enables deformation at strictlyroom temperature to limit as much as possible the evo-lution of the precipitation microstructure during defor-mation. Among the different choices HPT is known torespect best this criterion.

– Precipitation is followed during isothermal and non-iso-thermal ageing after extensive natural ageing of severalmonths to minimize scatter between samples.

Measurement of precipitates by SAXS is carried outin situ during different ageing sequences. Hardness mea-surements are carried out during the same heat treatments.These measurements are supplemented on selected micro-structures by TEM observations. The evolution of grainmicrostructure during the ageing treatments was qualita-tively followed by automated crystal orientation mapping(ACOM) obtained from nanobeam scanning transmissionelectron microscopy (STEM) images using the ASTARsoftware developed at the Grenoble Institute of Technol-ogy, France.

2. Materials and experimental methods

The 7075 alloy used in this study is the same as the oneexamined in an earlier investigation [45]. The alloy containsthe major alloying elements 5.63% Zn, 2.56% Mg and1.68% Cu with minor impurity elements 0.21% Fe, 0.19%Cr, 0.14% Si, 0.05% Mn and 0.02% Ti (all in mass%).The alloy was received in the form of a plate of 10 mmthickness after T6 treatment. Cylinders with a diameterof 10 mm and a height of 10 mm were extracted from theplate by an electrical spark discharge machine and slicedto discs with 1 mm thickness. The discs were then solu-tion-treated at 490 �C in air for 5 h followed by quenching

in water. Some of these discs were kept at room tempera-ture for several months, resulting in a state that will be sub-sequently called solution-treated and naturally aged(ST + NA) and were then subjected to ageing treatments.After the discs were lightly polished on the surfaces, theother discs were processed, right after solution treatment,by HPT at room temperature under a pressure of 6 GPafor 3 revolutions with a rotation speed of 1 rpm. TheHPT-processed discs were kept at room temperature (i.e.,natural aging) for several months, resulting in a state thatwill be subsequently called HPT processed and naturallyaged (HPT + NA) and were then also subjected to ageingtreatments.

SAXS experiments were carried out on a laboratoryrotating anode working with a Cu Ka source. CCD cameradata were corrected for read-out noise, distortion, flat-fieldand background noise. Data was normalized using a refer-ence glassy carbon sample as a secondary standard forabsolute calibration [46]. In situ experiments were carriedout using a dedicated furnace in which the samples, thinnedto 70–100 lm, were placed. The diameter of the X-raybeam was 1 mm, and the signal was recorded from a loca-tion placed at the mid-distance between the centre and theedge of the HPT-processed disc in order to probe a repro-ducible microstructure from one sample to the other.

Samples for TEM were prepared at the same location asfor the SAXS measurements (namely at the mid-distancebetween the centre and the edge of the HPT-processeddiscs) by conventional mechanical polishing down to90 lm thickness, followed by double-jet electropolishingin a methanol + nitric acid solution working at �30 �C,20 V. Observations were carried out on a 300 kV JEOL3010 microscope. Orientation maps were acquired usingACOM in the TEM [47]. The TEM sample was scannedwith a beam size of �10 nm without precession, at a scanrate of 35 fps and a step size of 10 nm. The resulting mapshave a total size of 3 lm � 3 lm.

3. Precipitation kinetics in the alloy with conventional grain

size

Before studying the precipitation kinetics in the severelydeformed material, it is useful, as a reference, to evaluatethat of the material processed in a conventional manner.For this purpose, the ST + NA material was subjected tothree ageing treatments while recording in situ thesmall-angle X-ray scattering signal to monitor the stateof precipitation. In the initial state, the precipitates havea characteristic radius of �0.8 nm as determined by theSAXS signal using the methodology detailed in Ref. [48].Following the detailed work on a coarse-grained similaralloy by Sha and Cerezo [49], and the observations on asimilar, severely plastically deformed alloy [50], these smallprecipitates formed during room-temperature ageing willbe subsequently assumed to be GP zones. Their volumefraction is such that a pronounced interference maximumis observed in the small-angle scattering spectrum

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.2

0.3

0.4

0.5

0.6

20 30 40 50 60 70 80 90 100

Prec

ipita

te v

olum

e fra

ctio

n

Zn atom fraction in precipitates

Temperature (°C)

Zn content of precipitates

Volume fraction

(b)

(a)

Fig. 1. (a) Log–log plot of the SAXS intensity, along with the modelledsignal, in the initial state of the conventional grain size material(ST + NA) and at two stages of the ramp heat treatment defined withrespect to the evolution of volume fraction (b).

108 A. Deschamps et al. / Acta Materialia 66 (2014) 105–117

(Fig. 1a). This feature is characteristic of a volume fractionsufficient so that the distance between precipitates is of thesame magnitude as their size, so that the small-angle scat-tering signal is the convolution of the form factor of theparticles and the structure factor of the particle distribu-tion. In these conditions, it is possible using a model forthe SAXS signal to infer the particle composition andtherefore to calculate the precipitate volume fraction inabsolute value from the measured integrated intensity.Since the contrast in SAXS arises from the square of thedifference in electron density, for simplification Mg andAl (ZMg = 12 and ZAl = 13) on one side and Cu and Zn(ZCu = 29 and ZZn = 30) on the other side can be consid-ered to have the same contribution to the electron density.Under this approximation, the electronic contrast can beestimated as arising from a binary system Al–Zn. A lowq contribution with a q�4 Porod behaviour has beenobserved on the SAXS patterns, even without artificial age-ing. It originates from the Porod asymptotic behaviourfrom very large features (e.g. dispersoids) which are muchtoo large to be measured by our setup. Its contributionto the measured integrated intensity was always negligibleso that we have considered this contribution as a back-ground signal that could be fitted by a contribution pro-portional to q�4.

The SAXS intensity from the precipitates was consideredas arising from a log normal distribution of spheres. The

interference effect was introduced using hard sphere interac-tions in the Percus–Yevick approximations [51]. The radiusof the hard sphere is assumed to be proportional to theradius of the precipitates. To speed up the computation,the local monodisperse approximation introduced byPedersen [52] has been used. Within this approximation,we can write the scattered intensity at scattering vector q:

IðqÞ ¼ Dq2

Z 1

0

F ðq;RÞ2Sðq;RHS ;uÞNðRÞdR ð1Þ

where Dq is the difference in electron density between theprecipitates and the matrix, F is the form factor of the pre-cipitates, S is the structure factor of the precipitate distri-bution and N is the precipitate size distribution function.The form factor of a sphere is given by

F ðq;RÞ ¼ 4pR3 sinðqRÞ � qR cosðqRÞðqRÞ3

ð2Þ

The log normal size distribution is described by

NðRÞ ¼ nvf ðRÞ ¼ nv1ffiffiffiffiffiffi

2pp

Rse�

12s2 logðR=RmÞ ð3Þ

where nv is the number density of precipitates and f(R) isthe normalized lognormal distribution of radii (Rm is themedian radius and s is the dispersion parameter).

RHS is the radius of the hard sphere. We have assumedthat it is proportional to R, so that

RHS ¼ kR ð4Þwhere the proportionality factor k is between 1 and 2. Wefurther introduce fv, the volume fraction of precipitates:

fv ¼ nv4

3pR3 ¼ nv

4

3pR3

me92s2 ð5Þ

From Eq. (4), we deduce u, the volume fraction of the hardspheres:

u ¼ k3fv ð6ÞThe structure factor is written [53]:

Sðq;RHS ;uÞ ¼1

1þ 24u GðAÞA

ð7Þ

with A = 2qRHS.

GðAÞ¼aðsinA�AcosAÞ

A2þb½2AsinAþð2�A2ÞcosA�2�

A3

þcf�A4 cosAþ4½ð3A2�6ÞcosAþðA3�6AÞsinAþ6�g

A5

ð8Þ

a ¼ ð1þ 2uÞ2

ð1� uÞ4

b ¼�6u 1þ u

2

� �2

ð1� uÞ2

c ¼ ua2

ð9Þ

0

10

20

30

40

50

60

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

40 80 120 160 200

Prec

ipita

te ra

dius

(Å)

Precipitate volume fraction

Temperature (°C)

Radius

Volume fraction

Fig. 2. (a) Kratky plot of the SAXS signal during the heating ramp at0.5 K min�1 performed on the ST&NA material. (b) Evolution of volumefraction, average radius and precipitate composition calculated from thisdata along the same heat treatment (the dashed line corresponds to datawhere a significant fraction of the SAXS data is lost into the beamstop).

A. Deschamps et al. / Acta Materialia 66 (2014) 105–117 109

Within our assumptions on the composition and since weare far from any absorption edge, we have:

Dq2 ¼ C2Zn

fZn � fAl

X

� �2

ð10Þ

where fZn and fAl are, respectively, the scattering factors ofZn and Al, X is the atomic volume and CZn is the Zn atomfraction in the precipitates (neglecting that of the matrix).Finally, the intensity can be written:

IðqÞ ¼ fvC2Zn

fZn � fAl

X

� �2 Z 1

0

F ðq;RÞ2Sðq; kR; fvÞ

� f ðRÞ43pR3

dR ð11Þ

Eq. (11) shows that the volume fraction appears both in thepre-factor, which in a sense scales the intensity, and in thestructure factor from which originates the interferencepeak. Within the “scale factor” alone, fv and C2

Zn cannotbe separated so that it is only when enough informationabout the shape of the interference peak is contained inthe data that the composition of the precipitates can be ex-tracted. Therefore, we first studied the early stages of pre-cipitation when the interference peak is clearly resolvedin order to deduce a more general form of the evolutionof the precipitate composition that could be used for therest of the study.

The size distribution is described by a median radius Rm

and a dispersion factor s. In order to show a single valuethat is relevant to small-angle scattering, we have chosento represent the size evolution by the Guinier radius, whichcan be recalculated from the log normal distribution:

R2g ¼

3

5

R8

R6

Rg ¼ffiffiffi3

5

rRme7s2 ð12Þ

This Guinier radius has been shown to be close to the aver-age radius for values of s close to 20% [48].

In Fig. 1b, the result of the fitting process has been plot-ted during the ramp heating of the ST + NA initial state at0.5 K min�1. The volume fraction is shown to first decreaseat constant Zn composition at �36%. At some stage duringthe dissolution, the Zn content of the precipitates isobserved to grow, indicating a possible phase transitionfrom GP zones to more stable precipitates like g0 and g.This transition occurs simultaneously to an increase ofthe precipitate size. SAXS patterns corresponding to threedifferent times in the kinetics (highlighted by colour dots onFig. 1b) are shown in Fig. 1a, together with the fitted inten-sity. It can be deduced from these curves that the interfer-ence peak is well defined during the GP zones stage and thebeginning of the phase transition, so that the fitted Zn com-position can be determined with good confidence. In thesubsequent stages, when the volume fraction decreasessimultaneously with an increase in size, the amplitude of

the peak becomes too small to provide enough information(material at 90 �C). Furthermore, it will be shown that tointerpret the HPT-processed sample, a second populationof precipitates will be necessary, lowering even more theprobability of resolving the interference peak. For theremainder the study, and based on the results of Fig. 1,we have assumed that the Zn composition of the GP zonesis constant at 36% and we have fixed a maximal Zn compo-sition for the g0/g phase at 53%, corresponding to theresults from Marlaud et al. [42]. To allow for a smoothtransition from one phase to the other, we have assumedthat the composition is related to the size of the precipi-tates, evolving from 36% to 53% via a step function (erf)at a critical size of 0.9 nm. It should be noted that thespherical shape assumption may not be the best modelfor the g0/g phase; however, we do believe that thisassumption does not influence the trend in terms of size,or the estimation of the volume fraction (as the latter orig-inates essentially from an integration of the signal). Thismodel was then applied to all in situ data during the heattreatments.

Now that the SAXS intensity evolution has beenadequately modelled, we can turn back to the precipitateevolution during ramp heating at 0.5 K min�1. A few scat-tering curves are shown along this heat treatment in Fig. 2in a Kratky plot. In such a plot a peak corresponds to the

110 A. Deschamps et al. / Acta Materialia 66 (2014) 105–117

presence of a precipitate size distribution. The position ofthe peak (in value of q) is inversely proportional to themean precipitate size, and the area under the peak is pro-portional to the product of the precipitate volume fractionand the square of the contrast in electron density betweenthe precipitates and matrix [37]. At first (here between theinitial state and the signal at 100 �C), the signal is observedto decrease, corresponding to a decrease of volume fractionwithout much change in q-range and thus of precipitatesize. At higher temperature, the signal translates to muchsmaller scattering vectors, reflecting an increasing precipi-tate size. This can be evaluated more quantitatively byusing the intensity model introduced earlier. The evolutionof precipitate size and volume fraction during the rampheat treatment is shown in Fig. 2b. When the temperaturereaches �60 �C, the volume fraction starts to decrease. Theprecipitate radius is very stable almost until the end of theprecipitate reversion that happens at �110 �C. Subse-quently, the radius and volume fraction suddenly increasetogether. This corresponds to the increase in Zn contentdiscussed earlier and can therefore be associated with theformation of the more stable g0 phase. When the tempera-ture reaches �180 �C, the volume fraction saturates at�2.5% and starts to decrease again while the precipitatesize continues to increase regularly. This decrease in vol-ume fraction is an experimental artefact arising from thefact that an increasing fraction of the scattered intensityis lost into the beam-stop when precipitates become toolarge.

In the following we present the precipitation kinetics intwo isothermal situations at 70 and 100 �C. 70 �C is a tem-perature where, during ramp heating, reversion is still mod-erate and nucleation of g0 has not started yet. At 100 �Cduring ramp heating, the reversion is nearly complete andnucleation of g0 is merely starting. Fig. 3 shows the evolu-tion of the SAXS signal during isothermal heating for 40 hat the two temperatures. The evolution of precipitates isvery different at the two temperatures. At 70 �C, the micro-structural evolution is very gradual and the signal hardlyshifts to smaller scattering vectors, which means that theprecipitate size increase is moderate. At 100 �C, the signalevolves as soon as the temperature is reached and then

Fig. 3. Kratky plot of the SAXS signal during the isothermal heat treatm

much more strongly. The quantification of this evolutionis shown in Fig. 4. At 70 �C, a progressive reversion isobserved, to end up after 40–50 h with a volume fractionof precipitates half of the initial one. The precipitate radiusonly increases from 0.8 to 1.2 nm in this time interval. It isimpossible to ascertain if the phase transformation fromGP zones to g0 has occurred, but if it has, it is at most par-tial. At 100 �C, however, the reversion is very rapid (as wasobserved during ramp heating) and it is followed by asharp transition to the g0 phase with a rapid increase ofvolume fraction and precipitate size.

4. Precipitation kinetics in the severely deformed material

Fig. 5 shows the comparison of the SAXS signals for theST + NA and HPT + NA samples both in a log–log plotand in a Kratky plot. In both materials the signal is dom-inated by objects of very small size (radius less than 1 nm)that can be identified as GP zones. Their volume fraction(in a first approximation the area under the Kratky plot)is similar in the two materials and their size (in a firstapproximation inversely proportional to the position inthe scattering vector of the maximum of Iq2) is slightlysmaller in the HPT-processed material. The log–log plotshows that the signal at very small angles of the HPT-pro-cessed material contains a stronger contribution as com-pared to the ST + NA material, in the form of a powerlaw close to q�2.5. Such a contribution arises probably fromthe microstrain present after SPD [54]; however, its contri-bution to the integrated intensity of the Kratky plot is neg-ligible. Thus for all practical purposes the two initialmicrostructures are observed to be essentially similar.

Fig. 6 presents Kratky plots of the SAXS signals at char-acteristic stages during the three heat treatments investi-gated (namely, ramp heating at 0.5 K min�1, isothermalheat treatment at 70 �C and isothermal heat treatment at100 �C). In all three cases one observes a very different evo-lution of the signal as compared to that of the ST + NAmaterial (one SAXS curve from this material is representedfor comparison). In the conventionally processed material,the whole scattering curve was observed to shift progres-sively towards smaller scattering vectors, reflecting that

ent performed on the ST&NA material at (a) 70 �C and (b) 100 �C.

0

5

10

15

20

0 10 20 30 40 50Time at ageing temperature (h)

70°C

100°C

Precipitate radius (Å)0

0.01

0.02

0.03

0.04

0.05

100°C

70°C

Precipitate volume fraction

Fig. 4. Evolution of volume fraction and average precipitate radius duringisothermal heat treatments at 70 �C and 100 �C performed on the ST&NAmaterial.

A. Deschamps et al. / Acta Materialia 66 (2014) 105–117 111

the precipitate size distribution remained mostly self-simi-lar. In the HPT-processed material, however, a new signalat very small angles appears rapidly in addition to the stillexisting signal at large angles, reflecting the formation oflarge particles (of radius �3 nm) simultaneously to the exis-tence of small ones that are still similar to the GP zones ini-tially present (of radius smaller than 1 nm). With time, thepresence of the large particles becomes more prominentand the GP zones tend to disappear. This behaviour is morepronounced as the temperature increases. During rampheating the first indication of large precipitate formation isobserved at �80 �C. During the isothermal heat treatmentat 70 �C, the formation of these large precipitates occursin a few hours, although it was observed that phase transfor-mation to g0/g at 70 �C was very slow in the conventionallyprocessed material. At 100 �C during the ramp heatingexperiment, the process of formation of these large particlesis already well advanced, so that it is no surprise to observethat in less than 1 h at this temperature during isothermalheat treatment the large precipitates have practically over-come the small ones. Again, as exemplified by the SAXSsignal for similar heat treatment, in the conventionally pro-cessed material the evolution is much slower and there is noevidence of a bimodal precipitate size distribution.

1

10

100

0.02 0.04 0.06 0.1 0.3

Inte

nsity

(Å-3

)

q (Å-1)

ST+NA

HPT+NA

Initial states

Fig. 5. Log–log and Kratky plot of the SAXS sig

The next step is to quantify these two precipitate popu-lations so that it becomes possible to separate their respec-tive contributions. For that purpose the followinghypotheses are made:

The small particles are similar to those present in theconventionally processed material. Therefore the same ruleis followed for their composition in solute, namely 36% Znin the initial state and an evolving composition to 53% Znwhen their size is larger that 0.9 nm.

A second population of precipitates is also consideredwith the same rule governing their composition. However,as the results will show, their size is always above 1 nm sothat they appear directly with a composition of equilibriumparticles, which we assume to be 53% Zn.

Then the SAXS signal is adjusted to a model containingthese two precipitate families, which are both consideredspherical, with lognormal size distribution.

Fig. 7 shows the evolution of the size and volume frac-tion of both precipitate families during ramp heating at0.5 K min�1. The evolution of precipitates in the ST + NAmaterial is also shown for comparison. As noted qualita-tively in the Kratky plots of Fig. 6a, the large precipitatesthat have initially a negligible volume fraction (notice thelog scale) start to form significantly at 80 �C, and then theirvolume fraction, together with their size, increases rapidlyuntil 120 �C. The subsequent decrease of volume fractionis undoubtedly related to the loss of the SAXS signal intothe beam stop because of a too large precipitate size. Thebehaviour of the small particles (initially GP zones) is par-ticularly interesting. They are present in a somewhat lowerfraction as in the ST + NA sample, but then experience avery similar evolution, namely a reversion in the same tem-perature range, at a similar rate relative to the initial vol-ume fraction and with an almost identical evolution inradius. The evolution of the small precipitate family ofthe HPT-processed material deviates strongly from thatof the ST + NA material only at relatively high tempera-tures (above 140 �C), when the large precipitates havebecome dominant in the microstructure.

Now isothermal heat treatments will be investigated tosee if similar conclusions can be drawn. Fig. 8 showsthe evolution of the two precipitate families during the70 �C heat treatment. In agreement with the qualitative

nal of the ST&NA and HPT&NA materials.

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

I.q2

q (A-1)

70°C heat treatment on HPT+NA

ST+NA - 15h

0h2h4h

6h

10h

15h

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

I.q2

q (A-1)

100°C heat treatment on HPT+NA

ST+NA - 1h

HPT+NA85°C90°C

95°C100°C

30 min

1h

0

0.05

0.1

0.15

0.2

0.25

0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4

0 0.1 0.2 0.3 0.4

I.q2

q (A-1)

Ramp HPT+NA105°C

101°C

97°C

93°C90°C

86°C

ST&NA - 105°C

Fig. 6. Kratky plots of the SAXS signal during (a) the heating ramp at 0.5 K min�1, (b) isothermal ageing at 70 �C and (c) isothermal ageing at 100 �C forthe HPT&NA material. As a reference, one SAXS curve is shown for the ST + NA material for an identical heat treatment.

0

10

20

30

40

50

40 80 120 160 200Temperature (°C)

ST+NA

HPT (large)

Precipitate radius (Å)

HPT (small)

0.001

0.01

0.1

HPT (large)

ST+NARamp - precipitate volume fraction

HPT (small)

HPT (sum)

Fig. 7. Evolution of precipitate volume fraction and size in theHPT + NA material subjected to ramp heating, as compared to that ofthe ST + NA material. For the HPT + NA material, the contributions ofthe small and large precipitates are separated, and the sum of the volumefraction of the two families is also represented.

0102030405060

0 10 20 30 40 50Time at ageing temperature (h)

ST+NA

HPT (large)

70°C - Precipitate radius (Å)

HPT (small)

0.001

0.01

HPT (large)

ST+NA70°C - Precipitate volume fraction

HPT (small) HPT (sum)

Fig. 8. Evolution of precipitate volume fraction and size in theHPT + NA material subjected to an isothermal heat treatment at 70 �C,as compared to that of the ST + NA material. For the HPT + NAmaterial, the contributions of the small and large precipitates areseparated, and the sum of the volume fraction of the two families is alsorepresented.

112 A. Deschamps et al. / Acta Materialia 66 (2014) 105–117

conclusions drawn from the Kratky plots (Fig. 6b), thelarge precipitates grow significantly in a few hours; how-ever, in contrast with the ramp heating experiment, theirvolume fraction does not become larger than that of thesmall precipitates in the investigated time range (50 h).Their size is comparable to the size of the large precipitatesformed during ramp heating. It is representative of the sizeof precipitates that would form during heat treatment of

conventionally processed materials at much higher temper-ature: for similar alloys, it takes several hours at 160 �C toreach such sizes (see e.g. Ref. [42]). As for the small precip-itates (initially GP zones), their behaviour strikingly resem-bles that of the ST + NA sample, shifted to a slightlysmaller volume fraction. The reversion rate is remarkablysimilar in the two materials, and the radius evolution isnearly identical.

0102030405060

0 10 20 30 40 50Time at ageing temperature (h)

ST+NA

HPT (large)HPT (small)

100°C - Precipitate radius (Å)

0.001

0.01

HPT (large)

ST+NA

100°C - Precipitate volume fraction

HPT (small)

HPT (sum)

Fig. 9. Evolution of precipitate volume fraction and size in theHPT + NA material subjected to an isothermal heat treatment at100 �C, as compared to that of the ST + NA material. For the HPT + NAmaterial, the contributions of the small and large precipitates areseparated, and the sum of the volume fraction of the two families is alsorepresented.

(b)C

F

Fig. 10. (a) Bright-field TEM micrograph of the HPT&NA material, (b) dark-fi100 �C showing coarse (C) and fine (F) precipitates (note that the contrast hasof the electron diffraction patterns in the HPT&NA materials and after agein

A. Deschamps et al. / Acta Materialia 66 (2014) 105–117 113

The evolution during heat treatment at 100 �C is shownin Fig. 9. Consistently with the Kratky plots of Fig. 6c, theformation of large precipitates is very rapid. Their volumefraction becomes larger than that of the small precipitatesin �1 h, and then slowly decreases certainly because theirsize becomes too large for the SAXS measurement. Asfor the small precipitates, again their behaviour is very sim-ilar to that of the ST + NA material. The extent of rever-sion in the HPT material is larger, but happens with thesame kinetics, and the evolution of precipitate size is almostidentical in the two materials.

5. Electron microscopy observations of the microstructure in

the severely processed material

The quantitative evaluation of precipitation kineticsprovided by the in situ SAXS measurements presentedabove has been supplemented by observations with elec-tron microscopy. Fig. 10a shows a bright-field micrographof the microstructure in the material after HPT processing

C

(a)

(c)

(d)

F

eld and (c) bright-field TEM micrograph of the HPT material aged 48 h atbeen enhanced for better clarity in the framed area of (c)). (d) Comparisong of 48 h at 100 �C.

Fig. 11. Orientation maps from STEM of (a) the HPT material and after ageing at (b) 70 �C and (c) 100 �C.

114 A. Deschamps et al. / Acta Materialia 66 (2014) 105–117

and natural ageing. Classically for high-solute-content Alalloys processed in similar conditions, the microstructureshows a very small scale, although it is difficult to preciselyseparate the different origins of contrast (small grain size,high dislocation density, internal stresses). The magnitudeof the grain size can be better evaluated from the orienta-tion maps, as shown in Fig. 11a. It is of the order of100 nm. As for precipitates present in this initial state,given the high contrast due to crystal defects and the verysmall size of GP zones, these cannot be resolved in thebright-field micrograph.

Fig. 10b and Fig. 10c shows dark-field and bright-fieldmicrographs of the alloy aged 48 h at 100 �C. In this case,the contrast is quite lower, showing qualitatively a reduc-tion in the amount of crystalline defects present. Nowprecipitates can be readily observed. It can be seen veryclearly that the precipitate distribution is bimodal. Atthe core of the grains, extremely fine precipitates can beobserved. These correspond to the small precipitates mea-sured during the SAXS experiments. On the grain bound-aries and at other crystalline defects, much largerprecipitates are observed. At this ageing time, the radiusof the largest ones can reach 30 nm, which is clearly outof the measuring range of our SAXS measurements.However, these were certainly the precipitates measured

with a radius of 4–6 nm during the first hours at100 �C. In terms of grain size, Fig. 11c shows the grainmicrostructure after the same ageing treatment obtainedfrom the orientation map. Clearly the grain size has sig-nificantly increased in comparison to the HPT + NAstate. This is further confirmed by the comparison ofthe selected area electron diffraction patterns of theHPT + NA and 100 �C aged material (Fig. 10d); the dif-fraction pattern is initially characteristic of a very smallgrain size and a high density of defects with almost con-tinuous diffraction rings. After ageing at 100 �C the ringsbecome much more punctuated, reflecting the decrease ininternal microstrains and the increase in grain size. FromFig. 11b, showing an orientation map of the materialaged at 70 �C for 48 h, we can conclude that no signifi-cant grain growth has occurred.

6. Resulting hardness evolution

Finally, the evolution of the material’s microhardnesshas been evaluated after HPT and natural ageing, as wellas during isothermal ageing at 70 �C and 100 �C(Fig. 12). Data obtained on the conventionally processedmaterial were gathered in the same conditions forcomparison.

50

100

150

200

250

300

10 100 1000 104

Vick

ers

mic

roha

rdne

ss

Ageing time (min)

HPT 70°C

HPT 100°C

ST 70°C

ST 100°CAs-HPT

As-ST

15 days45 days

15 days

45 days

Natural ageing

Fig. 12. Evolution of microhardness during natural ageing and subse-quent artificial ageing at 70 �C and 100 �C of the ST and HPT materials.

A. Deschamps et al. / Acta Materialia 66 (2014) 105–117 115

In the conventionally processed material, natural ageingafter solution treatment increases the hardness to �150HV. Artificial ageing provides an additional hardening,which occurs faster at 100 �C as compared to 70 �C, consis-tent with the appearance of the g0 phase.

The HPT-processed material shows already a very highhardness of �210 HV. After natural ageing this hardnessreaches 240–250 HV, thanks to the combination of thesmall grain size, high density of crystalline defects and pres-ence of GP zones. When the naturally aged material is agedat 70 �C, the hardness is stable and even increases slightlyover 250 HV. At 100 �C, however, the hardness is initiallystable and then slowly decreases to reach 230 HV at the endof the heat treatment.

7. Discussion

The present study provides to our knowledge the firstquantitative evaluation of the kinetics of precipitate evolu-tion during ageing following SPD through HPT process-ing. It provides conclusive evidence on several points thatwill now be discussed.

The first interesting feature concerns the initial statesbefore artificial ageing, namely the naturally aged materialsafter either conventional processing (ST + NA) or HPTprocessing (HPT + NA). The SAXS results show that thetwo microstructures are very similar. The HPT-processedmaterial, after several months at room temperature, doesnot show any significant presence of coarse equilibriumprecipitates, in contrast to what has been observed in bin-ary Al–Cu [12] where the equilibrium phase h forms atroom temperature after SPD. However, this lack of coarseprecipitation at room temperature is consistent with formerpublished data on 7000 series Al alloys [29]. This strongdifference between the two alloy series is yet to beexplained, and may be due to the combination of differ-ences in vacancy–solute interactions, nucleation mecha-nisms and interfacial energies. Secondly, the two initialstates contain a similar GP zone microstructure, in termsof both size and volume fraction. Following the hardnessevolution during natural ageing, it is observed that right

after HPT processing the hardness is already very high(>200 HV). This strongly suggests that the formation ofmost of the GP zones has occurred extensively duringHPT processing by dynamic precipitation and that the sub-sequent evolution of the GP zone fraction is only moder-ate. The fraction of GP zones in our two initial statescertainly is simply the equilibrium fraction at room temper-ature (accounting for the Gibbs–Thomson effect associatedto the very small size of the particles).

When the conventionally processed material is aged, achange in precipitate composition is observed to occur pro-gressively, controlled by the precipitate size and thus indi-rectly by the sample temperature. This transition is veryslow at 70 �C, rapid at 100 �C and occurs from 80 �Conwards during ramp heating at 0.5 K min�1. It is corre-lated to the evolution of precipitate size, and it can be rea-sonably related to the transition between GP zones andmore stable phases (g0, g).

In the HPT-processed material, the precipitation kinet-ics is more complicated. Larger precipitates (nucleationradius of the order of 3 nm) are observed to appear at tem-peratures as low as 70 �C, where precipitates in the conven-tionally processed material evolve only very slowly andremain at a small size. These precipitates grow fast andat 100 �C or during ramp heating are observed to dominatethe microstructure. The accelerated formation of coarseprecipitates is consistent with the existing literature [6,29]and our contribution is mainly to assess the temperatureat which this precipitation can be activated. Although theydo not form at room temperature in the time scale investi-gated here (several months), we have shown that theirnucleation can be activated during a few hours at a temper-ature as low as 70 �C. From the electron microscopy obser-vations, it can be inferred that these precipitates form onthe structural defects created by the HPT process (grainboundaries and dislocations).

The really interesting feature is, however, the behaviourof the small precipitates initially present in the HPT-pro-cessed material (the GP zones). These small precipitatesare observed to evolve during artificial ageing in a mannerextremely similar to their counterparts in the convention-ally processed material. This similarity applies to the kinet-ics (like the time for the maximum extent of reversion) andto the quantitative evolution (like the growth rate of theprecipitates). The electron microscopy observations showthat these small precipitates are located in the regions,which are far from structural defects. It has been shown[55] that the kinetics of reversion is very sensitive to the dif-fusion rate of solutes independently of other parameters ofthe precipitation system. In fact the time for a givenamount of reversion is inversely proportional to the diffu-sion coefficient, all other parameters being equal. Thereforeit can be expected that the diffusion coefficient in the graininterior of the severely plastically deformed material is sim-ilar to that of the conventionally processed material. Aclose look at Fig. 7 reveals a temperature difference of only4 K for the minimum of volume fraction detected during

116 A. Deschamps et al. / Acta Materialia 66 (2014) 105–117

reversion between the ST + NA and HPT + NA materials.If we consider a range of activation energies for bulk diffu-sion of 1–1.5 eV, this 4 K temperature difference translatesinto a difference of 30 to 40% only in diffusion constant.This reveals that the vacancy concentration is almost thesame in the two materials (HPT + NA and ST + NA).We believe that this is the first (indirect) evidence of asimilar vacancy concentration in an SPD precipitation-hardened Al alloy as compared to its coarse-grained coun-terpart. It contrasts with some estimations that the vacancyconcentration in SPD materials may be many orders ofmagnitude larger than in the coarse-grained materials[12,34,36]. However, the materials evaluated in our studywere naturally aged for several months before the ageingtreatments were realized. It may be that high excessvacancy concentrations did exist during or immediatelyafter HPT processing, and that these excess vacancies didnot survive after natural ageing. However, the state ofGP zone development in the HPT + NA and ST + NAmaterials is similar, which means that possible differencesin vacancy concentrations did not profoundly change themicrostructure once a stable microstructure at room tem-perature is reached.

In preceding studies, the high vacancy concentrationshave been associated with the formation at low tempera-ture of large, equilibrium phases. Such particles have beenobserved in the present study, forming during artificial age-ing, where it has been shown that the vacancy concentra-tion was equal to that of the conventionally processedalloy. Therefore, an alternative mechanism needs to beinvoked to explain their accelerated formation. Actually,our observations of grain size evolution by ACOM showthat the grain size significantly evolves, at least at 100 �Cwhere the coarse precipitates are observed to form particu-larly rapidly. Since the coarse precipitates have been shownto be associated with the structural defects, it seems verylikely that the migration of these defects during heat treat-ment is closely associated to the formation of these precip-itates. Former studies have actually shown that precipitateformation and decrease of microstrain and dislocation den-sity were highly correlated during ageing of SPD processed7xxx alloys [29]. The solute collector mechanism hasalready been invoked before to account for the accelerationof precipitation kinetics in SPD materials [12]. And a recentstudy of mixing different materials by HPT processing [35]has shown that the diffusion rate necessary to account forthe kinetics of mixing was compatible with surface precip-itation, which further demonstrates the importance of dif-fusion at the interfaces and supports the hypothesis ofstructural defect migration to account for the acceleratedprecipitation kinetics in our study.

8. Summary and conclusions

In this paper we have reported a quantitative study,using SAXS, of the precipitation kinetics in an AA7075aluminium alloy processed by HPT at room temperature.

This precipitation kinetics has been compared with thatof the same material processed in a conventional manner.The SAXS measurements have been carried out in situ dur-ing a continuous heating experiment at 0.5 K min�1 andduring isothermal ageing at 70 �C and 100 �C. Addition-ally, TEM has been carried out to obtain a qualitative viewof the precipitate distribution after ageing and to character-ize the material’s grain size using ACOM. From theseresults the following conclusions are proposed:

– After HPT followed by several months of natural ageingthe material contains a similar GP zone microstructureas compared to the conventionally processed material.No sign of the presence of coarse precipitates is detectedat this stage.

– When temperature is increased in the HPT-processedmaterial, extensive coarse precipitation occurs at lowtemperature. The time necessary to obtain such coarseprecipitates is considerably shorter than in the conven-tionally processed material. These precipitates arerelated to the crystalline defects present initially, and itis likely that the presence of these defects, and possiblytheir migration, during ageing plays a key role in the sol-ute transport involved in their formation.

– Conversely, fine-scale precipitates remain at the core ofthe nano-grains during the first stages of the formationof these coarse precipitates. These fine-scale precipitatesrespond to temperature changes exactly in the same wayas in the conventionally processed materials. From thisresult, it can be inferred that the vacancy concentrationat the core of the nano-grains is similar to that of theconventionally processed material.

– From a more practical viewpoint, artificial ageing atlow temperatures (here 70 �C) makes it possible to fur-ther increase the strength from the SPD processedstate, due to the extensive precipitation with limitedcoarsening associated with the retention of the smallgrain size.

Acknowledgements

This research was supported in part by Japan Scienceand Technology Agency (JST) under CollaborativeResearch Based on Industrial Demand “HeterogeneousStructure Control: Towards Innovative Development ofMetallic Structural Materials”, in part by the Light MetalsEducational Foundation of Japan, in part by a Grant-in-Aid for Scientific Research from the MEXT, Japan, ininnovative areas “Bulk Nanostructured Metals”

(22102004).

References

[1] Valiev R, Islamgaliev R, Alexandrov I. Prog Mater Sci 2000;45:103.[2] Valiev RZ, Langdon TG. Prog Mater Sci 2006;51:881.[3] Zhilyaev AP, Langdon TG. Prog Mater Sci 2008;53:893.

A. Deschamps et al. / Acta Materialia 66 (2014) 105–117 117

[4] Chinh NQ, Gubicza J, Czeppe T, Lendvai J, Xu C, Valiev RZ, et al.Mater Sci Eng A 2009;516:248.

[5] Duan ZC, Liao XZ, Kawasaki M, Figueiredo RB, Langdon TG. JMater Sci 2010;45:4621.

[6] Ghosh KS, Gao N, Starink MJ. Mater Sci Eng A 2012;552:164.[7] Krishna KG, Sivaprasad K, Venkateswarlu K, Kumar KCH. Mater

Sci Eng A 2012;535:129.[8] Gubicza J, Schiller I, Chinh NQ, Illy J, Horita Z, Langdon TG.

Mater Sci Eng A 2007;460–461:77.[9] Han J-H. Mater Trans 2010;51:2109.

[10] Hockauf K, Meyer LW, Hockauf M, Halle T. J Mater Sci2010;45:4754.

[11] Horita Z, Ohashi K, Fujita T, Kaneko K, Langdon TG. Adv Mater2005;17:1599.

[12] Huang Y, Robson JD, Prangnell PB. Acta Mater 2010;58:1643.[13] Cabibbo M. Mater Charact 2012;68:7.[14] Cabibbo M, Evangelista E, Vedani M. Met Mater Trans A

2005;36A:1353.[15] Cardoso KR, Travessa DN, Botta WJ, Jorge AM. Mater Sci Eng A

2011;528:5804.[16] Cepeda-Jimenez CM, Garcia-Infanta JM, Zhilyaev AP, Ruano OA,

Carreno F. J Alloys Comp 2011;509:636.[17] Cerri E, Leo P. Mater Sci Eng A 2005;410:226.[18] Huang W, Liu Z, Lin M, Zhou X, Zhao L, Ning A, et al. Mater Sci

Eng A 2012;546:26.[19] Lee S, Horita Z, Hirosawa S, Matsuda K. Mater Sci Eng A

2012;546:82.[20] Liu Z, Bai S, Zhou X, Gu Y. Mater Sci Eng A 2011;528:2217.[21] Munoz-Morris MA, Morris DG. Mater Sci Eng A 2011;528:3445.[22] Murayama M, Horita Z, Hono K. Acta Mater 2001;49:21.[23] Roven HJ, Liu M, Werenskiold JC. Mater Sci Eng A 2008;483–84:54.[24] Sha G, Wang YB, Liao XZ, Duan ZC, Ringer SP, Langdon TG. Acta

Mater 2009;57:3123.[25] Tsai M-S, Sun P-L, Kao P-W, Chang C-P. Mater Trans 2009;50:771.[26] Tsuji N, Iwata T, Sato M, Fujimoto S, Minamino Y. Sci Technol Adv

Mater 2004;5:173.[27] Xu C, Furukawa M, Horita Z, Langdon TG. Acta Mater

2003;51:6139.[28] Zhang J, Gao N, Starink MJ. Mater Sci Eng A 2010;527:3472.[29] Zhao YH, Liao XZ, Jin Z, Valiev RZ, Zhu YT. Acta Mater

2004;52:4589.

[30] Straumal B, Baretzky B, Mazilkin A, Phillipp F, Kogtenkova O,Volkov M, et al. Acta Mater 2004;52:4469.

[31] Hu T, Ma K, Topping TD, Schoenung JM, Lavernia EJ. Acta Mater2013;61:2163.

[32] Liddicoat PV, Liao X-Z, Zhao Y, Zhu Y, Murashkin MY, LaverniaEJ, et al. Nat Comm 2010:1.

[33] Hirosawa S, Hamaoka T, Horita Z, Lee S, Matsuda K, Terada D.Met Mater Trans A 2013;44A:3921.

[34] Sauvage X, Wetscher F, Pareige P. Acta Mater 2005;53:2127.[35] Oh-ishi K, Edalati K, Kim HS, Hono K, Horita Z. Acta Mater

2013;61:3482.[36] Lechner W, Puff W, Mingler B, Zehetbauer MJ, Wuerschum R. Scr

Mater 2009;61:383.[37] De Geuser F, Deschamps A. C R Phys 2012;13:246.[38] Deschamps A, De Geuser F. Met Mater Trans A 2013;44A:77.[39] Malard B, Remy B, Scott C, Deschamps A, Chene J, Dieudonne T,

et al. Mater Sci Eng A 2012;536:110.[40] Nicolas M, Deschamps A. Acta Mater 2003;51:6077.[41] Dumont M, Steuwer A, Deschamps A, Peel M, Withers PJ. Acta

Mater 2006;54:4793.[42] Marlaud T, Deschamps A, Bley F, Lefebvre W, Baroux B. Acta

Mater 2010;58:248.[43] Bakavos D, Prangnell P, Bes B, Eberl F, Grossmann J. Mater Sci

Forum 2006;519–521:333.[44] Kenesei P, Horvath G, Bernstorff S, Ungar T, Lendvai J. Zeit Metallk

2006;97:315.[45] Horita Z, Fujinami T, Nemoto M, Langdon TG. Met Mater Trans A

2000;31:691.[46] Zhang F, Ilavsky J, Long GG, Quintana JPG, Allen AJ, Jemian PR.

Met Mater Trans A 2010;41A:1151.[47] Rauch EF, Veron M. Mater Werkst 2005;36:552.[48] Deschamps A, De Geuser F. J Appl Crystallogr 2011;44:343.[49] Sha G, Cerezo A. Acta Mater 2004;52:4503.[50] Zhao Y-H, Liao X-Z, Cheng S, Ma E, Zhu YT. Adv Mater

2006;18:2280.[51] Vrij A. J Chem Phys 1979;71:3267.[52] Pedersen JS. J Appl Crystallogr 1994;27:595.[53] Kinning D, Thomas E. Macromolecules 1984;17:1712.[54] Epperson J, Loomis B, Lin J. J Nucl Mater 1982;108:476.[55] Deschamps A, Perez M. C R Phys 2010;11:236.