Temperature-Dependent Deformation Behavior of Al­Mg­Sc Alloys Fabricated byMulti-Directional Forging at Room Temperature+1

Shunsuke Yamazaki1,+2, Syutaro Sawa1,+2, Chihiro Watanabe2,+3, Ryoichi Monzen2, Tomoya Aoba3,+4

and Hiromi Miura3

1Division of Mechanical Science and Engineering, Graduate School of Natural Science and Technology, Kanazawa University,Kanazawa 920-1192, Japan2Faculty of Mechanical Engineering, Institute of Science and Engineering, Kanazawa University, Kanazawa 920-1192, Japan3Department of Mechanical Engineering, Toyohashi University of Technology, Toyohashi 441-8580, Japan

Ultrafine-grained Al­Mg­Sc alloys were fabricated by multi-directional forging (MDF) with different number of forging passes of 3, 9 and15, i.e., to cumulative strains of ­¦¾ = 1.2, 3.6 and 6.0, at room temperature. The achieved average grain sizes were 950, 680 and 360 nm at 3, 9and 15 passes, respectively. Peak-aging treatments at 473K for 172.8 ks were adopted for a portion of specimens after MDF in order to obtainfinely dispersed Al3Sc precipitates. Grain coarsening did not take place in all the specimens during the aging. The activation volume for plasticdeformation was estimated from the strain-rate jump tensile tests before and after the aging. Aging-free 3- and 9-pass specimens showed positivetemperature dependence of the activation volume, while that of the aging-free 15-pass one bearing the smallest grain size exhibited a negativetemperature dependence. Contrary to these results, values of the activation volume in the peak-aged specimens were approximately identicalregardless of grain size or deformation temperature. These results strongly suggested that, due to the precipitation of Al3Sc, the rate-controllingprocess of deformation was changed from interaction between forest dislocations and mobile dislocations for the aging-free 3- and 9-passspecimens to interaction between mobile dislocations and Al3Sc precipitates, or from bowing-out of dislocations from grain boundaries for theaging-free 15-pass specimen to the interaction between the mobile dislocations and precipitates. [doi:10.2320/matertrans.MT-M2020287]

(Received September 1, 2020; Accepted November 27, 2020; Published January 12, 2021)

Keywords: aluminum­magnesium­scandium alloy, ultrafine-grained structure, multi-directional forging, strain-rate jump tests, activationvolume

1. Introduction

Bulky ultrafine-grained (UFGed) metals and alloys withnanometer-scale grain sizes have been recently fabricatedvia severe plastic deformation (SPD) methods, which includeaccumulative roll-bonding (ARB),1) equal-channel angularpressing (ECAP),2) high-pressure torsion (HPT)3) and multi-directional forging (MDF).4) UFGed metals and alloysfabricated by means of SPD methods have been activelyinvestigated in the last two decades5­7) and many researchershave reported the excellent mechanical properties. In thisway, the unique mechanical properties of UFGed materialsare being come to light, which are largely different from thoseof conventional coarse-grained ones. Thus, UFGed materialsare highly attractive from both the practical and academicperspectives; for example, the activation volume of plasticdeformation has a positive temperature dependence in coarse-grained pure face-centered cubic (fcc) metals, but it hasexhibited a negative temperature dependence in submicron-grained materials.8­11) Kato et al.10,11) quantitatively ex-plained the grain-size and temperature dependencies of theactivation volume of the UFGed pure fcc metals with adepinning model considering the dislocation bowing-outfrom grain boundaries. This model is based on the decreasingin-grain dislocation sources with decreasing grain size.According to the model, the transmission of dislocationstransitions from in-grain sources to grain boundaries, and

then, the temperature dependence of activation volumechanges. However, the effect of second-phase particlesdispersed in UFGs on the temperature dependence ofactivation volume has not been clarified yet. Whenprecipitates are dispersed in the UFGs, it can be assumedthat the rate-controlling process of the plastic deformationwould no longer be the dislocation bowing-out from theboundaries but the process of overcoming the precipitates.Therefore, dispersoids in UFGs would also change thetemperature dependence of activation volume.

In the present study, UFGed Al­Mg­Sc alloys with havingdifferent grain sizes were fabricated by means of MDF. Somespecimens were followed by aging to have finely dispersedAl3Sc precipitates in grain interiors and on boundaries. Thedifferences in the temperature dependence of mechanicalproperties of these specimens were systematically inves-tigated.

2. Experimental Procedure

A hot-rolled Al­Mg­Sc alloy plates, provided by UACJCorporation, were used in this study, and the chemicalcomposition is listed in Table 1. The plates were fabricatedby the following procedure. Ingots with dimensions of

Table 1 Chemical composition of an Al­Mg­Sc alloy used in this study(mass%).

+1This Paper was Originally Published in Japanese in J. Japan Inst. Met.Mater. 84 (2020) 208­215.

+2Graduate Student, Kanazawa University+3Corresponding author, E-mail: chihiro@se.kanazawa-u.ac.jp+4Present address: Department of Mechanical Engineering, NationalInstitute of Technology, Kisarazu College, Kisarazu 292-0041, Japan

Materials Transactions, Vol. 62, No. 2 (2021) pp. 213 to 220©2021 The Japan Institute of Metals and Materials

170mm length, 200mm width, and 60mm thickness wereprepared by book mold casting and homogenized at 813K for14.4 ks. Then, their surfaces were milled down by 5mm andhot rolled at a starting temperature of 788K to a thicknessof approximately 20mm. The hot-rolled plate was cut intorectangular-shaped pieces followed by solid-solution treat-ment at 863K for 7.2 ks and MDF at room temperature (RT).In the MDF processing, constant forging-pass strains arerepeatedly applied from three directions (the axes perpendic-ular to the three surfaces of the rectangular-shaped pieces).4)

During MDF, the dimensions of the pieces are unchanged byemploying an appropriate ratio of the specimen dimensionsand the fitting forging pass strain, ¦¾. Thus, the forging canbe repeated by rotating its direction by 90°, and unlimitedstrain can be theoretically given to the specimen. In thisstudy, ¦¾ was set as 0.4 in true strain and dimensions of therectangular-shaped pieces as 15.0, 22.4 and 18.3mm whichaxes were parallel to the rolling direction (RD), transversedirection (TD) and normal direction (ND), respectively. MDFprocessing was conducted on an Amsler-type universalmechanical testing machine with an initial strain rate _¾ of1.0 © 10¹3 s¹1 up to forging passes of 3, 9, and 15, whichcorrespond to cumulative strains of ­¦¾ = 1.2, 3.6 and 6.0,respectively. Some specimens were aged at 473K for variousperiods of time up to 302.4 ks at maximum. Hereafter, thespecimens are labeled using characters of ST (solution-treated), PA (peak-aged) and X (the number of MDF passes);for example, “the ST-9” indicates the specimen solutiontreated and followed by 9 passes of MDF, and “the PA-9”peak aged of the ST-9. The specimens used in all the testswere cut out from the center of the rectangular-shaped pieces.

The microstructure of the specimens was investigatedusing an optical microscopy (OM) and a transmissionelectron microscopy (TEM). For the OM observation, thespecimens were mechanically polished using SiC papers first,an alumina suspension with an abrasive size of 0.05 µm next,and finally a colloidal silica suspension to have mirror-likesurface. They were finished by etching with a hydrofluoricacid solution (distilled water:hydrofluoric acid (46%) = 10:1in volume) at RT for 30 s. The TEM observations wereconducted using an FEI TECNAI G2 microscopy underan accelerating voltage of 200 kV. Thin foils for TEMobservations were prepared by twin-jet electro-polishingwith a perchloric acid/methanol solution (1:9 in volume) at243K and 28V.

Measurements of hardness and electric resistance, andtensile tests were also performed. The hardness was evaluatedwith a micro-Vickers hardness tester (Akashi, HM-102). Themeasurement was repeated 10 times for each specimen undera loading of 4.9N for 10 s, and then the average values weretaken. The electrical resistance was assessed through thefour-terminal method by using a resistance meter (HiokiElectric, RM3545-01) with a current of 1A at RT. Themeasured values were converted into the specific resistancebased on the cross-sectional area of the specimens and thedistance between the terminals. For the tensile tests, dog-bone-shaped specimens with a gage section of 5mm length,1.5mm width and 0.6mm thickness were employed. Theywere cut from the rectangular-shaped pieces to have thetensile axis perpendicular to the final forging axis of MDF.

The tests were conducted on an Instron-type universalmechanical testing machine (Shimadzu AG-10kNX Plus) atboth 77K and RT at _¾ ¼ 1:0� 10�3 s¹1. Furthermore, strain-rate jump tests were performed at RT, 200K and 77K byrepeatedly changing the strain rate between 1.0 © 10¹4 and1.0 © 10¹3 s¹1. The refrigerants used for the tests at 200Kand 77K were methanol chilled to near its melting pointand liquid nitrogen, respectively. In this case, the specimenswere preliminarily soaked in each refrigerant for 600 s andkept soaking during the measurements. For the test at 200K,the refrigerant temperature was often measured, and furtherliquid nitrogen was appropriately added to maintain thedesired temperature (approximately «2K).

3. Results

3.1 Microstructure and mechanical propertiesFigure 1 shows the optical micrographs of a hot-rolled

and solutionized Al­Mg­Sc specimen with grains elongatedalong RD. The average grain-boundary spacings perpendic-ular to ND, RD and TD were about 30, 153 and 54 µm beforethe solution treatment and approximately 96, 225 and 102 µmafterward, respectively. This indicates grain growth duringthe solution treatment.

The TEM observation of the hot-rolled specimen revealedthe presence of Al3Sc precipitates with an average radiusof around 40 nm, while no precipitate was confirmed aftersolution treatment, which became single-phase fcc. Figure 2displays a bright-field TEM micrograph of the ST-15specimen, indicating that the initial coarse grains weredramatically fragmented after MDFing. The corresponding

Fig. 1 Optical micrographs of an Al­Mg­Sc specimen after hot-rolling andthen solution treatment at 863K for 7.2 ks. The micrographs were takenfrom (a) ND and (b) TD, respectively.

S. Yamazaki et al.214

selected-area-diffraction pattern (SADP) (inset of Fig. 2),taken using a selected-area aperture of 6 µm in diameter,revealed nearly continuous diffraction rings which suggestsrandom-orientation distribution of UFGs surrounded byhigh-angle grain boundaries. The grain size decreased withincreasing the number of MDF passes. The average(sub)grain diameter d of the ST-3, ST-9 and ST-15 specimenswas approximately 950, 680 and 360 nm, respectively. Thegrain size was measured regardless of the grain-boundarycharacter, i.e., whether they were low-angle or high-angleones.

Figure 3 presents the age-hardening curves of specimensaged at 473K. The hardness before aging increased withthe number of MDF passes. In each specimen, the hardnessreached its maximum at around 172.8 ks of aging and thendecreased. The specimens forged by 3 and 9 passes exhibitedrelatively large softening. The microstructural observationsrevealed that they underwent partial recrystallization afterthe aging for 302.4 ks. Hereafter, the PA label indicates thespecimens aged at 473K for 172.8 ks.

Figure 4(a) displays a bright-field TEM micrograph andthe SADP of the PA-15 specimen, showing that the UFGedstructure is still maintained after the peak aging. The averagevalue of d was about 360 nm, which is almost the samewith that of the ST-15 specimen. This indicates that graincoarsening hardly occurred during aging at 473K. Detailed

TEM observations revealed that the precipitates often formedon the boundaries and pinned the grain-boundary migration(Fig. 4(b)). These precipitates were identified as the Al3Scphase based on the SADP analyses. The Al3Sc precipitates,therefore, effectively stabilize the UFGed structure duringaging. The fine spherical Al3Sc precipitates were dispersedalso in grain interior as examplified in Fig. 4(c). Therefore,the hardness increase by aging can be attributed to theformation of these Al3Sc precipitates. Table 2 summarizesaverage grain size of d and average radius r of the in-grainprecipitates for both the ST and PA specimens. The value ofd appears identical among the PA and the correspondingST specimens regardless of the MDF passes. Moreover, thevalue of r among the three PA specimens looks almost thesame.

Figure 5 displays the stress-strain curves attained bytensile tests. All the curves exhibited serrations as shown in

Fig. 2 TEM micrograph of a ST-15 specimen. The inset is selected-areadiffraction pattern taken using a selected-area aperture with 6 µm indiameter.

Fig. 3 Age-hardening curves of Al­Mg­Sc specimens processed by MDFof 3, 9 and 15-passes at room temperature. The aging treatments wereconducted at 473K.

Fig. 4 (a) TEM micrograph of a PA-15 specimen. (b) High-magnificationTEM micrograph showing an Al3Sc precipitate pinning a grain boundary.(c) High-resolution TEM micrograph showing an Al3Sc precipitate withingrain-interior. The inset in (a) is selected-area diffraction pattern takenusing a selected-area aperture with 6µm in diameter.

Temperature-Dependent Deformation Behavior of Al­Mg­Sc Alloys Fabricated by Multi-Directional Forging at Room Temperature 215

the insets in Fig. 5. Solute Mg atoms in an Al matrix formCottrell atmosphere around the dislocations.12­14) Theoccurrence of serrations could be, therefore, due to theimpediment and separation of mobile dislocations by andfrom the Cottrell atmosphere.14­17) Table 3 summarizes themechanical properties, and Fig. 6 illustrates the relationshipbetween d and the 0.2% proof stress (·0.2) of the specimens.All the specimens exhibited a linear relationship between ·0.2and d¹1/2. Thus, it is evident that the increase in ·0.2 by grainrefinement follows the Hall-Petch relation. Moreover, dueto the strengthening by the in-grain Al3Sc precipitates, thePA specimens showed higher ·0.2 values compared to theST ones.

Figure 7 shows the results of tensile tests at 77K. For allthe specimens, the flow stress levels and elongationsincreased compared to those observed at RT. Besides, no

serration appeared. Since the diffusion rate of solute Mgatoms exponentially decreases along with decreasing temper-ature, the Cottrell atmosphere hardly formed at cryogenictemperatures.18)

3.2 Strain-rate jump testsFor the strain-rate jump tests, _¾ was first set to 1.0 ©

10¹4 s¹1 (_¾1), and then the crosshead speed was suddenlychanged up to 1.0 © 10¹3 s¹1 (_¾2) during the tests. Afterstraining to about 0.5%, the crosshead speed was suddenlyreturned down to 1.0 © 10¹4 s¹1 and then another straining of0.5% was added again. The cycle of strain-rate change wasrepeated until the flow stress reached the tensile strength.When the strain rate is increased, the flow stress generallyincreases as well.19) Actually, the increase in flow stress at200 and 77K corresponded to the increase in strain ratealso in the present tests. Nevertheless, at RT, the flow stressdecreased as the strain rate increased and vice versa.Miyajima et al. have reported that the deformation of aUFGed Al­Mg alloy at RT is mainly controlled by thedynamic strain aging effect from the solute Mg atoms20) andthe flow stress decreases with increasing strain rate.Therefore, this reversed behavior of flow stress observed atRT can be ascribed to the dynamic strain aging by the Mgatoms. Thus, only the results of the tests at 77 and 200K aredescribed from now on.

Table 2 Average grain size d and radius r of Al3Sc precipitates in PA-specimens. Average grain sizes of ST specimens are shown forcomparison. Also shown are data for an OA-15 specimen (15-passspecimen over-aged at 473K for 302.4 ks).

Fig. 5 Stress-strain curves of (a) ST-specimens and (b) PA-specimens.Tensile tests were conducted under an initial strain rate of 1.0 © 10¹3 s¹1

at room temperature.

Table 3 Fracture strain ¾f , 0.2% proof stress ·0.2 and tensile stress ·UTS ofST- and PA-specimens. The data were obtained from tensile tests at aninitial strain rate of 1.0 © 10¹3 s¹1 at room temperature.

Fig. 6 Grain size d dependence of 0.2% proof stress ·0.2 of ST- and PA-specimens. The data were obtained from tensile tests at an initial strainrate of 1.0 © 10¹3 s¹1 at room temperature.

S. Yamazaki et al.216

The results of strain-rate jump tests for the ST and PAspecimens are shown in Fig. 8 and Fig. 9, respectively. Whenthe strain rate was changed, the flow stress variation wasrelatively more significant at 77K than at 200K. The curvesat 77K exhibited a sawtooth-like shape. In contrast, at 200K,a sharp increase in flow stress was then followed by gradualdecrease, as described in the circle in Fig. 8. This behavioris typically observed when dislocations detach from theCottrell atmosphere.21) Therefore, the characteristic changein the flow stress occurred at 200K can be attributed to theCottrell atmosphere of the Mg atoms.

The activation volume V� was calculated using the resultsof the first strain-rate jump in Figs. 8 and 9, immediately afterthe macroscopic yielding (at a plastic strain of approximately0.5%), as follows:11)

V� � MTkT ð@ ln _¾=@·ÞT ¼ MTkT ½lnð_¾2=_¾1Þ=�·�T ð1Þwhere ¦· is the flow stress change caused by the changein strain-rate from _¾1 to _¾2, k is the Boltzmann constant, Tis the deformation temperature, and MT is the Taylor factor,whose general value for fcc metals is 3.06.22) Figure 10summarizes the temperature dependence of V� for eachspecimen, derived from eq. (1). Here the V� was normalizedusing the magnitude of the Burger’s vector (b = 0.287 nm),whose value was obtained from the lattice constant (a =0.4063 nm) experimentally determined via X-ray diffractom-etry. As shown in Fig. 10(a), the V�=b3 of the ST-3 and ST-9specimens exhibited positive dependence on the temperature,while that of the ST-15 one demonstrates a negativetemperature dependence. On the other hand, the V� values

Fig. 7 Stress-strain curves of (a) ST- and (b) PA-specimens. Tensile testswere conducted at an initial strain rate of 1.0 © 10¹3 s¹1 at 77K.

Fig. 8 Stress-strain responses of ST-specimens during strain-rate jumptests. The strain-rate jump tests were conducted between _¾ ¼ 1:0� 10�4

and 1.0 © 10¹3 s¹1 at (a) 77K and (b) 200K.

Fig. 9 Stress-strain responses of PA-specimens during strain-rate jumptests. The strain-rate jump tests were conducted between _¾ ¼ 1:0� 10�4

and 1.0 © 10¹3 s¹1 at (a) 77K and (b) 200K.

Temperature-Dependent Deformation Behavior of Al­Mg­Sc Alloys Fabricated by Multi-Directional Forging at Room Temperature 217

for the PA specimens were all about 300b3 regardless of thenumber of MDF passes and temperature (Fig. 10(b)).

4. Discussion

The V� of pure fcc metals strongly depends on the grainsize, and its temperature dependency varies when the grainsare fragmented to submicron order (50­500 nm).4­8) Moreconcretely, the temperature dependence of V� changes frompositive to negative with grain refinement.8­11) With adepinning model of dislocations bowing-out from the grainboundaries, Kato et al. quantitatively explained the grain-sizeand temperature dependencies of V� in UFG fcc metals.10,11)

According to this model, the interactions between mobileand forest dislocations control the deformation rate when thedislocation sources are within the grains, but the dislocationbowing-out from grain boundaries becomes the rate-controlling process when the grain refinement stronglyreduces such in-grain sources. As shown in Fig. 10(a), forthe ST specimens, the temperature dependence of V� changedfrom positive to negative with increasing the number of MDFpasses, i.e., with decreasing grain size. This is in goodaccordance with the model proposed by Kato et al. That is,with decreasing grain size of the ST specimens, dominantdislocation source transitions from grain interior to grainboundary and, therefore, the deformation-rate-control processchanges from the interaction between mobile and forestdislocations within grains to bowing-out of dislocations fromgrain boundaries.

In contrast, the V� of the PA specimens, which had fineAl3Sc precipitates within the grains, exhibited almost no

temperature dependence (Fig. 10(b)). Thus, the presence ofprecipitates should also cause a change in the rate-controllingprocess of deformation.23) This is because the dislocationsare interacted with and pinned by the in-grain precipitatesirrespective of dislocation sources and, therefore, thedislocation overcoming of precipitates becomes the deforma-tion rate-controlling process. In this case, the V� shoulddepend on the precipitate radius r and the interprecipitatespacing ­. Here, the discussion will be proceeded assumingthat the Al3Sc precipitates are distributed at the points in aregular hexagonal shape on a certain plane for simplicityschematically described Fig. 11. The value of ­ can beestimated as follows,24) based on the r and volume fractionf of the Al3Sc precipitates:

­ ¼ r½ð8³=3ffiffiffi

3p

fÞ1=2 � 2�: ð2ÞBy considering the effect of Sc addition to Al on theresistivity (¦μSc = 34 n³m/at%25)), the amount of Scconsumed to form the Al3Sc phase was derived from theresistivity variation ¦μ before and after aging. Then, thevalue of f could be evaluated using the molar volume ofthe Al3Sc phase (1.0425 © 10¹5m3mol¹1 26)). The ¦μ valuesof the specimens forged to 3, 9 and 15 passes wereapproximately 4, 5, and 5 n³m, and the correspondingf values were estimated to be 0.30%, 0.37% and 0.37%,respectively. Finally, the ­ values were derived using eq. (2)with the r values listed in Table 2. The results aresummarized in Table 4. For the PA specimens, the ­ valueswere nearly identical regardless of grain size d.

Fig. 10 Temperature dependence of normalized activation volume V�=b3

of (a) ST- and (b) PA-specimens. Also shown in (b) is data for an OA-15specimen (15-pass specimen over-aged at 473K for 302.4 ks).

Fig. 11 Schematic illustration of the distribution model of precipitateparticles.

Table 4 Volume fraction of Al3Sc precipitates f, average precipitate radiusr, and inter-precipitate spacing ­ for PA-specimens. Also shown aredata for an OA-15 specimen (15-pass specimen over-aged at 473K for302.4 ks).

S. Yamazaki et al.218

In general, the precipitate-shearing mechanism and theOrowan mechanism can be pointed out as the majorinteractions between mobile dislocations and precipitates atrelatively low temperatures.27) When the shearing mechanismis dominant (Fig. 12(a)), the V� value can be expressed as

V� � S�b ¼ 2r­ � ð­ 2=4Þ½ð2ª � sin 2ªÞ= sin2 ª� ð3Þwhere, S� is the activation area schematically indicated by thegray hatch in Fig. 12(a) and ª is the bowing-out angle of thedislocation. The V� values for the all PA specimens wereexperimentally estimated by the strain-rate jump tests to beabout 300b3 (see section 3.2 and Fig. 10(b)). In this case,based on eq. (3) and the values in Table 4, ª must be 8­9°.On the other hand, it is reported that the Orowan mechanismis dominant when the radius of Al3Sc precipitates exceeds2.4 nm in an Al­2.0mass%Mg­0.2mass%Sc alloy, whichcontained the same volume fraction of precipitates as in thepresent alloy.28) In all the PA specimens, the r values werelarger than the critical radius (µ2.4 nm) (Table 4). Hence, theOrowan mechanism should be dominant in the PA specimensand the ª must be about 90°, which is much larger than theabove estimated value of 8­9°. Consequently, it is understoodthat the precipitate shearing mechanism did not occur in thePA specimens. Moreover, if the dislocations overcame theprecipitates from the state of bowing-out with ª µ 90°(Fig. 12(b)), the V� can be expressed as

V� � S�b ¼ ð³=8Þ½ð­ þ 2rÞ2 � ­ 2�b: ð4ÞBased on the ­ and r values in Table 4, the resulting V� isapproximately 9800b3, which is over one order of magnitudelarger than the experimental values (Fig. 10). Thus, themodel illustrated in Fig. 12(b) cannot reasonably explain ourexperimental results.

On the other hand, if the Orowan mechanism is thedeformation rate-controlling process, the dislocation of thecritical state probably bypasses the precipitates with a slightincrease in ª (¦ª) from the semi-circular shape (Fig. 13).In this case, the S� can be expressed as

S� ¼ V�=b ¼ ð³­ 2=8Þ½ð1= cos2 �ªÞ þ ð�ª=180�Þ � 1�þ ð­ 2 tan�ªÞ=4: ð5Þ

Using eq. (5) and the ­ values in Table 4, the V� values wereconsidered again. In this case, the ¦ª required by thedislocations to bypass the precipitates via the Orowanmechanism should be significantly small; thus, the value ofV� was calculated in the range of ¦ª = 0.1­0.5°. The valueswere almost the same for all PA specimens, ranging from150b3 (¦ª = 0.1°) to 800b3 (¦ª = 0.5°). Moreover, thesevalues were in the same order as the experimental ones(Fig. 10(b)). Therefore, the V� values experimentallyobtained for the PA specimens are reasonably understoodfrom dislocation bypass of the Al3Sc precipitates by theOrowan mechanism as illustrated in Fig. 13. The modeldiscussed above is qualitatively consistent with the study byMarquis et al., in which the Orowan mechanism controlledthe yield stress of an Al­Mg­Sc alloy when the radius ofAl3Sc precipitates is larger than 2.4 nm.28) If the precipitatesize is relatively large and cannot be shear deformed bymobile dislocations, the precipitates can be considered as“strong long-range obstacles” against the dislocationmotion.23) Since such obstacles are generally consideredathermal,23) the strain rate and temperature dependenciesof the flow stress become extremely small.29) And so, in thePA specimens with almost constant r and ­ (Table 4), allthe V� values were almost identical (Fig. 10(b)) and notemperature dependence appeared. To verify this conclusion,a portion of the specimen forged with 15 passes wassubjected to over aging at 473K for 302.4 ks; this specimenwas referred as OA-15. Both the r and ­ of the OA-15specimen increased by the over aging compared with thoseof the PA-15 specimen (Table 4). The OA-15 specimen wasalso subjected to strain-rate jump tests to investigate thetemperature dependence of its V�. The results are displayedin Fig. 10(b). As for the PA specimens, the V� showed notemperature dependence. Furthermore, its values wererelatively larger than those of the PA specimens. This isbecause the value of V� monotonically increases along with­ according to eq. (5). These results strongly support theconclusion stated above.

5. Conclusions

The temperature dependence of the deformation behaviorof ultrafine-grained Al­Mg­Sc alloys, having three different

Fig. 12 Schematic illustration of activation volume in the (a) shearing and(b) overcoming process of precipitates.

Fig. 13 Schematic illustration of activation volume in the bypassingprocess of precipitates.

Temperature-Dependent Deformation Behavior of Al­Mg­Sc Alloys Fabricated by Multi-Directional Forging at Room Temperature 219

grain sizes fabricated by various numbers of multi-directionalforging (MDF) passes at room temperature, was systemati-cally investigated. The results yielded are summarized asfollows.(1) In the specimens forged within 3 and 9 passes, which

had relatively large grain sizes of 950 and 680 nm, theactivation volume of plastic deformation showed apositive temperature dependence. In contrast, that ofthe specimen prepared via 15 MDF passes, which hadthe smallest grain size of 360 nm, exhibited a negativetemperature dependence. These results can be reason-ably explained by the depinning model in which therate-controlling process of deformation transitions withdecreasing grain size from the interaction betweenmobile and forest dislocations to the dislocationbowing-out from boundaries.

(2) The activation volume of plastic deformation in thepeak-aged specimens with dispersed fine Al3Scprecipitates showed no temperature dependence andwas nearly identical regardless of the number of MDFpasses, i.e., grain size. The Al3Sc precipitates acted asstrong obstacles against mobile dislocations, and thus,the deformation rate was controlled by the interactionbetween mobile dislocations and precipitates. There-fore, dislocation bypass of the Al3Sc precipitates by theOrowan mechanism should work as a dominant rate-control process of deformation independent of grainsize.

Acknowledgments

Some of the authors, C. Watanabe and H. Miura,appreciate the financial support given by Japan Science andTechnology Agency under the Industry-Academia Collabo-rative R&D Program “Heterogeneous Structure Control:Towards Innovative Development of Metallic StructuralMaterials” (Grant#: JPMJSK1413).

REFERENCES

1) Y. Saito, H. Utsunomiya, N. Tsuji and T. Sakai: Acta Mater. 47 (1999)579­583.

2) V.M. Segal, V.I. Reznikov, A.E. Drobyshevskiy and V.I. Kopylov:Russ. Metall. 1 (1981) 99­105.

3) H. Iwaoka, Y. Fuzioka, Y. Harai and Y. Horii: J. Japan Inst. Metals 75(2011) 412­418.

4) T. Sakai and H. Miura: Tetsu-to-Hagané 94 (2008) 590­598.5) N. Tsuji: Tetsu-to-Hagané 88 (2002) 359­369.6) K.T. Park, S.Y. Han, D.H. Shin, K.J. Lee and K.S. Lee: ISIJ Int. 44

(2004) 1057.7) H. Miura, Y. Nakao and T. Sakai: Mater. Trans. 48 (2007) 2539­2541.8) T. Kunimine, N. Takata, N. Tsuji, T. Fujii, M. Kato and S. Onaka:

Mater. Trans. 50 (2009) 64­69.9) S. Okubo, Y. Miyajima, S. Onaka and M. Kato: Mater. Trans. 55 (2014)

1525­1530.10) M. Kato: Mater. Sci. Eng. A 516 (2009) 276­282.11) M. Kato, T. Fujii and S. Onaka: Mater. Trans. 49 (2008) 1278­1283.12) R. Horiuchi, H. Yoshinaga and S. Hama: J. Japan Inst. Metals 29

(1965) 85­92.13) J. Balik and P. Lukac: Acta Metall. Mater. 41 (1993) 1447­1454.14) H. Yoshida and T. Fukui: J. JILM 38 (1988) 496­512.15) J.M. Robinson and M.P. Shaw: Int. Mater. Rev. 39 (1994) 113­122.16) I.S. Kim and M.C. Chaturvedi: Mater. Sci. Eng. 37 (1979) 165­172.17) K. Matsuura, Z. Nishiyama and S. Koda: J. Japan Inst. Metals 31

(1967) 1042­1048.18) J. Japan Inst. Metals: Zairyokyoudonogensiron, (J. Japan Inst. Metals,

Sendai, 1985) pp. 263­265.19) D. Caillard and J.L. Martin: Thermally Activated Mechanisms in

Crystal Plasticity, (Elsevier, Amsterdam, 2003) pp. 13­54.20) Y. Miyajima, R. Kaizumi, S. Onaka and M. Kato: Collected Abstracts

of the 2016 Autumn Meeting of the Japan Inst. Metals, (2016) J43.21) S. Zhao, C. Meng, F. Mao, W. Hu and G. Gottstein: Acta Mater. 76

(2014) 54­67.22) G.I. Taylor: J. Inst. Met. 62 (1938) 307­324.23) M. Kato: Nyumonteniron, (Shokabo, Tokyo, 2010) pp. 139­154.24) J. Miyake and M.E. Fine: Acta Metall. Mater. 40 (1992) 733­741.25) T. Fujikawa: J. JILM 49 (1999) 128­144.26) C. Watanabe, T. Kondo and R. Monzen: Metall. Mater. Trans. A 35

(2004) 3003­3008.27) J. Japan Inst. Metals: Zairyokyoudonogensiron, (J. Japan Inst. Metals,

Sendai, 1985) pp. 132­153.28) E.A. Marquis, D.N. Seidman and D.C. Dunand: Acta Mater. 51 (2003)

4751­4760.29) H. Yoshinaga: J. JILM 29 (1979) 528­537.

S. Yamazaki et al.220