Acta Materialia 55 (2007) 6284–6291 www.elsevier.com/locate/actamat

Experimental study of dislocation mobility in a Ti–6Al–4V alloy P. Castany *, F. Pettinari-Sturmel, J. Crestou, J. Douin, A. Coujou CEMES/CNRS, BP 94347, 31055 Toulouse Cedex 4, France Received 28 March 2007; received in revised form 12 July 2007; accepted 16 July 2007 Available online 14 September 2007

Abstract By in situ TEM deformation experiments, we studied in detail the deformation micromechanisms of a Ti–6Al–4V alloy at room temperature. All dislocations have an a-type Burgers vector and glide essentially in prismatic or basal planes. They are first emitted from a/b interfaces and take a preferential orientation along their screw direction. The motion of screw dislocations controls the strain rate. Our experiments allow the microscopic parameters of plasticity for this alloy to be determined for the first time. The results concerning the screw dislocation motion, its core structure and the influence of interfaces are then discussed in comparison with previously published results. Ó 2007 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Titanium alloys; Transmission electron microscopy (TEM); Dislocation dynamics; In situ; Tension test

1. Introduction Titanium alloys are widely used because of their excellent mechanical properties combined with low density. The most used of titanium alloys in the aerospace industry is Ti–6Al–4V [1]. It is a two-phase alloy with an a phase (hexagonal close-packed) and a b phase (body-centered cubic). Its microstructure and thus its mechanical properties are very dependent on the thermomechanical processes performed during its production. The microstructure can be fully nodular, fully lamellar or duplex with nodules and lamellar colonies. The deformation of pure a-titanium has already been widely studied [2–8]. Two types of Burgers vectors are possible for gliding dislocations: a-type ða=3h1 1  2 0iÞ, which can glide in prismatic, basal and first-order pyramidal planes, and c+a-type ða=3h1 1  2 3iÞ, which can glide in firstand second-order pyramidal planes. Some twinning systems can also be activated, but they are essentially observed at room temperature in low oxygen content alloys [8]. In a-

*

Corresponding author. Tel.: +33 5 62 25 78 70; fax: +33 5 62 25 79 99. E-mail address: [email protected] (P. Castany).

titanium, prismatic glide of a-type dislocations is the main deformation mode and the deformation is controlled by the motion of the screw segments [2–7]. Atomistic calculations show that the high lattice friction on screw segments is due to their three-dimensional core structure spread in several planes [9–14]. Indeed, screw dislocations have a stable and sessile configuration which must recombine in the glide plane into a glissile and metastable configuration to be able to glide. The deformation micromechanisms have been less studied in two-phase and polycrystalline alloys like Ti–6Al–4V principally because of their more complex microstructure. This complexity is due to the presence of two phases (eventually more) and the presence of different type of grains – nodules and lamellar colonies – in which the a phase can have different chemical composition and different dislocation behaviour. The contributions to the strength of the alloy are then more numerous, the main ones already studied being:  the presence of short range order, especially observed in nodular alloys [15–18];  the presence of interfaces, particularly in lamellar alloys or colonies [19–22].

1359-6454/$30.00 Ó 2007 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actamat.2007.07.032

P. Castany et al. / Acta Materialia 55 (2007) 6284–6291

6285

In the present paper, we present a detailed study of the deformation micromechanisms in a Ti–6Al–4V alloy. We focus especially on the origin of the strength in lamellar colonies using in situ transmission electron microscopy (TEM) deformation experiments at room temperature, which is the only available technique for studying the dislocation dynamics under stress and temperature. This technique has been previously performed only on Ti–6Al–4V alloy in cryogenic conditions at 20 K [20] or on pure a-titanium [6,7], but never on other titanium alloys. 2. Experimental The alloy under investigation is an industrial Ti–6Al–4V compound with 6% aluminium by weight and 4% vanadium by weight. Other elements are present as impurities (mainly oxygen, nitrogen, carbon and iron). It has a duplex microstructure (Fig. 1), with primary alpha nodules aP and lamellar colonies aS/b. The nodules and lamellar colonies have the same size of about 10 lm. Lamellar colonies consist of secondary alpha plates aS with an average thickness of about 500 nm separated by thin b plates (Fig. 2). The b plates between two aS plates are

Fig. 2. Detailed view of a lamellar colony with aS and b plates. In A and B, the b plates are discontinuous.

sometimes discontinuous (A and B in Fig. 2) and are a few tens of nanometres thick. Small b grains (about 100 nm size) are sometimes present at the grain boundaries between the nodules and lamellar colonies. Our observations indicate that the nodules and lamellar colonies have similar volume fractions and we estimate the total b volume fraction to be about 3%, in agreement with recent results by Bridier et al. for a similar alloy [23]. Thin foils were prepared for TEM in situ tensile tests. The samples were mechanically polished and thinned down by twin-jet electropolishing. However, for this alloy, the sample preparation produced some difficulties for in situ TEM experiments because of the b phase preferential milling compared with the a phase (either by electrolytic or by ion milling processes). Due to this preferential milling during the thinning process, hole edges were notched, favouring crack formation during tension tests. By adjusting the polishing parameters, this preferential milling was minimized and crack formation was avoided: the electropolishing was performed at 15 °C with a current of 500 mA. The A3 solution commercialized by STRUERS has been used. Note also that no hydride formation was observed, as reported sometimes for electropolished a/b titanium alloy samples [24]. In situ tensile tests were performed at room temperature with a Gatan straining holder. TEM observations were made with a JEOL 2010 equipped with a SIS CCD camera. Details of the TEM in situ deformation technique are specified elsewhere [25]. The in situ results were compared with TEM post-mortem observations from macroscopically tensile-deformed samples. 3. Results

Fig. 1. Microstructure of the investigated alloy with primary alpha nodules aP and lamellar colonies aS/b.

The main micromechanisms of the plastic deformation of this alloy were identified during in situ experiments. All observed gliding dislocations have an a-type Burgers vector. In lamellar colonies, as in nodules, we essentially found basal and prismatic glide and, less frequently, first-

6286

P. Castany et al. / Acta Materialia 55 (2007) 6284–6291

order pyramidal glide. In the three types of planes, dislocations generally had behaviour similar to that detailed here. The first sequence (Fig. 3) shows the emission and propagation of a dislocation loop in an aS plate. It is emitted from the interface between the aS and b phases at point A (Fig. 3a). The same dislocation expands in the following images (Fig. 3b and c); it is worth emphasizing that dislocations take a preferential orientation along the screw direction, which results in a large density of rectilinear screw segments. Only the part close to the edge segment is bowed, but departure from the screw orientation can occur because of pinning points, as seen in B (Fig. 3d). In this micrograph, the dislocation loop emerges at the sample surface at the point C and the slip plane trace is visible. This dislocation glides in the basal plane and has an a-type  Burgers vector: < ce : bold > b < =ce : bold >¼ a=3 1 1  2 0 (determined by the g Æ b = 0 invisibility criterion). All gliding dislocations in this lamellar colony have the same Burgers vector, are mainly rectilinear and aligned with their screw direction, and glide in the basal plane. At the early stage of deformation, such emission of dislocations from the aS/b interface is the main dislocation creation phenomenon and it has been observed to be the same at the aP/b interfaces [22]. Using such an experiment, the velocity of mobile dislocations can be measured. Screw segments have a jerky motion, with very fast jumps (the time is shorter than a video frame, which is a twenty-fifth of second). The average jump length is 35 nm. Between two jumps, dislocations are immobile during an average waiting time of 50 s. The average velocity of the screw segments is then about 2 nm s1. However, the waiting time distribution is very large: dislocations can be stopped for a few seconds or a few minutes before a jump. The waiting time is then the parameter which controls their velocity. The edge segments have rather a quasi-continuous movement and their average velocity is about 200 nm s1. Edge segments can be sometimes stopped but their waiting time does not exceed a few seconds. Some edge segments move so quickly that they emerge at the sample surface in the resolution time

of the camera; their velocity has not been considered because of the impossibility of a precise measurement. The edge segments therefore move at least 100 times faster than the screw ones and the dislocation motion is then essentially controlled by the screw segments. The dislocation motion can be inhibited by the formation of intrinsic obstacles. The jump of the screw segments generally does not occur for all the dislocation length and only one part jumps. This results in the creation of a macrokink with a size equal to the jump magnitude, which acts as an intrinsic obstacle. The sequence in Fig. 4 illustrates the typical motion of a screw segment from the left to the right in a basal plane. A macrokink, arrowed in Fig. 4a, has been created by a jump of part of the dislocation. It is next immobile for 13 s before a jump of the macrokink to another position (Fig. 4b). The same event occurs between Fig. 4b and c. Between Fig. 4c and d, the macrokink does not jump quickly but glides slowly along the dislocation line. The velocity of the kink is then relatively low (34 nm s1) and the movement is continuous along the screw direction. The dotted line in the figures is parallel to the screw direction and is at the same position in all the images. The macrokink then moves along the screw direction, allowing the displacement of the dislocation over a distance equal to the kink magnitude by the lateral propagation of non-screw parts. This lateral propagation is often very fast and cannot be recorded: only the jump between two positions of the screw segment is seen, leading to a jerky motion. The dislocations have the same kind of motion when they glide in basal, prismatic or first-order pyramidal planes [26]. Fig. 5 shows the creation of a dislocation loop: first, the dislocation is pinned and bowed (arrowed in Fig. 5a), and in the next image (Fig. 5b) it has slipped, returning to the screw character and leaving behind a dislocation loop. Next, this loop expands under stress (Fig. 5c and d). Finally, the initial dislocation moves away, leaving two new screw segments (arrowed in Fig. 5e). This loop is formed by cross-slip by the open loop mechanism [27]

Fig. 3. Emission of a dislocation loop from an aS/b interface during an in situ experiment. The edge segment velocity is much larger than the screw segment one.

P. Castany et al. / Acta Materialia 55 (2007) 6284–6291

6287

Fig. 4. Typical motion of a screw dislocation with the propagation of a macrokink (arrowed) along the same dislocation. The dotted line is at exactly the same position in every frame and is parallel to the screw direction.

Fig. 5. Creation and expansion of a dislocation loop by the open loop mechanism. At the end (e), the initial dislocation has moved away and has left two new mobile screw segments (arrowed).

and suggests frequent cross-slip. This loop expansion participates to the dislocation multiplication. This mechanism often occurs in lamellar colonies and it is the main dislocation multiplication mechanism. It occurs more rarely in aP nodules. As already shown in a-titanium [28], cross-slip is frequent in this alloy. An example of the trace left by a dislocation when it propagates is arrowed in Fig. 6. This trace is not rectilinear and is coherent with the existence of crossslip. The trace of the basal plane and a first-order pyramidal plane are reported in the figure: the dislocation glides mainly in the basal plane and it seems that cross-slip occurs in a first-order pyramidal plane. Cross-slips from a prismatic plane to a first-order pyramidal plane and from a first-order pyramidal plane to a prismatic plane have also

been observed during our experiments. Moreover, the occurrence of cross-slip is in good agreement with the open loop multiplication mechanism shown previously. The jerky motion of screw dislocations seen in Fig. 4 is due to the high lattice friction resulting from their threedimensional core structure. As will be discussed further, their core structure acts as an intrinsic obstacle controlling their motion. However, some extrinsic obstacles exist too. In Fig. 7, the dislocation denoted 1 is pinned (arrowed in Fig. 7a) and moves forward to the right, leaving probably a small dipole in Fig. 7b and c. Next (Fig. 7d), another dislocation, denoted 2 and gliding in the same plane, is pinned exactly at the same position, which indicates the extrinsic character of the obstacle; the first dislocation is still visible at the bottom of the image. Extrinsic obstacles exist but are

6288

P. Castany et al. / Acta Materialia 55 (2007) 6284–6291

Fig. 8. A post-mortem image of the same alloy deformed macroscopically to 0.3% plastic. The dislocation configuration is the same as in our in situ experiments: a dislocation loop, denoted 1, emerges from an aS/b interface and the dislocations are rectilinear and aligned with their screw direction, like the dislocations denoted 2 and 3.

Fig. 6. Observation of a slip plane trace of a dislocation gliding mainly in the basal plane. The occurrence of cross-slip is visible (arrow) and takes place probably in a first-order pyramidal plane. A first-order pyramidal plane trace (tr P1) and the basal plane trace (tr B) are reported.

less frequent compared to the intrinsic obstacles: for example, when the dislocation denoted 1 in Fig. 7 sweeps 0.3 lm2, it meets only one extrinsic obstacle. Fig. 8 is a post-mortem image of a aS plate in a lamellar colony of a sample macroscopically deformed by traction up to 0.3% of plastic deformation. Although the dislocation arrangement is more complex due to a higher strain rate, it is in agreement with in situ observations: a disloca-

tion loop (dislocation denoted 1 in Fig. 8) is emitted from the aS/b interface and many dislocations are rectilinear and aligned along their screw direction (e.g. dislocations numbered 2 and 3 in Fig. 8). These dislocations have an a-type Burgers vector and glide in the basal plane. 4. Discussion The TEM in situ technique allows information to be obtained on the dislocation dynamics. To avoid possible thin foil effects, it is useful to compare these results with post-mortem observations, exemplified in Fig. 8. We have observed several post-mortem samples and our in situ observations are in agreement with the bulk alloy deformation. Moreover, in situ experiments give results about the chronology of events and about the dynamics of disloca-

Fig. 7. Pinning of two dislocations on the same extrinsic obstacle: a first dislocation, denoted 1, is pinned at a point (arrowed) in (a) and has moved away in (b) and (c). Next (d), a second dislocation, denoted 2, is pinned at exactly the same point. In situ experiments can then determine unambiguously whether obstacles are extrinsic or intrinsic.

P. Castany et al. / Acta Materialia 55 (2007) 6284–6291

tions contrary to post-mortem observations, which give less information. For example, the nature of the pinning points is very difficult to determine from post-mortem observations, whereas the difference between intrinsic and extrinsic obstacles is easy to determine with TEM in situ experiments: if obstacles are extrinsic, dislocations slipping in the same plane are all pinned at the same point, which is not the case for intrinsic obstacles. 4.1. Screw segment motion The large anisotropy of velocity between edge and screw segments shows the strong lattice friction on screw segments. In this alloy, the deformation rate is then governed by the screw dislocation motion. In a-titanium, dislocations are rectilinear and elongated along their screw direction. They move by a locking–unlocking mechanism between two locking positions [6]. In this Ti–6Al–4V alloy, the jerky motion of screw dislocations is due to the same mechanism, but it is very often accompanied by the gliding of a macrokink along the dislocation line, which allows the propagation of the screw segments (Fig. 4). This behaviour with macrokinks looks like the glide of ordinary dislocations in c-TiAl [29] more than the glide in a-titanium, where one whole screw segment generally makes the jump [6,7]. In these two materials the screw dislocation motion is governed by their three-dimensional core structure [3,5,6,29], and this difference in behaviour illustrates that the screw segments are more rigid in a-titanium. During their motion, the dislocations in Ti–6Al–4V are globally rectilinear but are more flexible than in a-titanium, and the dislocations bow easily when they are pinned, particularly at extrinsic obstacles (Fig. 7). The comparison with the dislocation behaviour in various titanium alloys is impossible because in situ TEM experiments at room temperature have been performed only on pure a-titanium or on c-TiAl alloys: only postmortem results are available. In Fig. 4, the macrokink first moves very quickly and then moves much more slowly. This difference of velocity can be explained by the increase in the local stress close to the aS/b interfaces, as measured in a previous paper [30]: close to aS/b interfaces, there is an additional stress due to dislocations present at these interfaces before deformation. So, when a macrokink is close to aS/b interfaces, it moves faster because of a higher local stress. When it moves in the middle of the aS plate, the local stress is lower because of the lack of the additional stress due to interfaces and its velocity is lower. 4.2. Core structure It is now accepted for the prismatic glide in a-titanium that the lattice friction is due to the core structure of a-type screw dislocations [5,6]. As the prismatic glide is the main slip system in a-titanium, basal glide has not been investigated very much. Our in situ observations show that, on

6289

the whole, the dislocation motion is the same whether the glide is in prismatic, basal or first-order pyramidal planes [26]. In addition, dislocations gliding in the basal plane have a similar behaviour to ordinary dislocations in c-TiAl and to dislocations in a-titanium, where the motion of both is governed by their core structure [3,5,6,29]. Therefore an equivalent core spreading exists probably for screw dislocations in Ti–6Al–4V gliding in basal, prismatic or first-order pyramidal planes. Moreover, a core spreading is consistent with the large occurrence of cross-slip in our experiments. The slip plane trace analysis shows that dislocations cross-slip from the basal plane to a first-order pyramidal plane, which is not favourably oriented (a Schmid factor of 0.15) but return quickly to the basal plane by double cross-slip. Cross-slip from prismatic to first-order pyramidal planes and from first-order pyramidal to prismatic planes have also been observed. These results confirm that cross-slip occurs easily in this alloy as it has already been shown in titanium [3,28] and its alloys [23]. To explain the predominance of the prismatic glide in atitanium, the core structure analysis of screw dislocations generally considers a main spreading of the core into the prismatic plane with secondary spreading in basal planes [10–13]. Only Naka et al. [3] have proposed a model in which the core is spread principally in the prismatic plane and secondarily in the two first-order pyramidal planes, allowing cross-slip in the latter. From our observations of cross-slip and similar dislocation motion in basal, prismatic and first-order pyramidal planes, the core of screw dislocations can be assumed to be simultaneously spread in the prismatic, basal and first-order pyramidal planes with different spreading magnitude, depending on the orientation stress: the main spreading being in the slip plane and secondary spreadings lying in others planes to allow easy cross-slip. Previous calculations [10–13] have indeed only been performed on stress-free dislocations, and the core structure of dislocations is certainly modified by the applied stress and depends on its orientation. Moreover, the core structure is probably modified by the presence of aluminium, which could explain the differences between the screw dislocation behaviour in a-titanium and in Ti–6Al–4V. In the two alloys, dislocation motion is governed by the transition from a sessile state of their core to a glissile and metastable state. But there are some differences at room temperature:  the waiting time in the locked position is shorter in atitanium than in Ti–6Al–4V: respectively a few seconds at most in a-titanium [6], and an average value of 50 s in Ti–6Al–4V;  the jump length is longer in a-titanium than in Ti–6Al– 4V: respectively 100 nm in a-titanium [7] and 35 nm in Ti–6Al–4V;  macrokinks are present in Ti–6Al–4V, whereas in a-titanium the entire dislocation generally makes the jump [6,7].

6290

P. Castany et al. / Acta Materialia 55 (2007) 6284–6291

There is, then, a decrease in the dislocation mobility in the Ti–6Al–4V in comparison with a-titanium, which could be attributed to the aluminium content. Aluminium can indeed make the sessile–glissile transition more difficult through a chemical interaction, as assumed for interstitials atoms [3]. This core structure almost certainly governs the motion of screw dislocations and induces cross-slip, and thus the dislocation multiplication by the open loop mechanism (Fig. 5) that is due to double cross-slip. This mechanism was observed for the first time by Furubayashi [27] and later in different alloys [6,31]. Here, it is the main multiplication mechanism of dislocations in aS plates and the mobile dislocation density is quickly increased. 4.3. Extrinsic obstacles In this alloy, the density of extrinsic obstacle is low (Fig. 7). Moreover, in c-TiAl, there are many extrinsic obstacles and their strengthening contribution is much smaller than the intrinsic ones due to the non-planar core spreading contribution [32]. The similarity between the dislocation behaviour in this study and the ordinary dislocation behaviour in c-TiAl indicates that the dislocation motion is almost certainly controlled by the lattice friction due to the three-dimensional core structure of screw dislocations and that extrinsic obstacles play a negligible role in the strength of this alloy. The nature of these extrinsic obstacles is still unknown, but is probably related to impurities [33]. 4.4. Influence of the aS/b interfaces Plasticity begins with dislocations emitted from a/b interfaces (Fig. 3). Such emissions from a/b interfaces have been observed post-mortem in a nodular Ti–6Al–4V alloy at high temperature [34]. In lamellar colonies, it has been shown that a-type dislocations are present in these interfaces before deformation [35] and can also propagate because of the local stress. Emission of dislocations from the interfaces is not the only effect induced by the presence of b plates: it slightly favours the basal glide over the prismatic one. In a lamellar colony, we have indeed observed only basal glide (excepted double cross-slip), with a Schmid factor (SF) of 0.40, whereas for a prismatic glide system SF = 0.37. However, in such a situation, prismatic glide is expected as in aP nodules because a recent study has shown that only prismatic glide is activated in aP nodules when prismatic and basal glide have similar Schmid factors [23]. In aP nodules, we made the same observations during in situ TEM experiments: only prismatic glide is observed (SF = 0.38), whereas for basal glide SF = 0.39 [26]. When basal glide has a much higher SF than the prismatic ones, only basal glide is observed [23,26]. Therefore, it appears that in Ti– 6Al–4V alloys the basal glide is slightly favoured in lamellar colonies and the prismatic glide in aP nodules.

Ankem and Margolin [36] have shown that the elastic compatibility stress between the a and b phases in lamellar colonies of titanium alloys can inhibit or favour some slip systems depending on the applied stress orientation and that the basal glide is the more affected slip system. In the studied alloy, the b phase proportion in lamellar colonies is low, but it seems that the same effects are observed. The compatibility stress due to the presence of two phases could therefore favour the basal glide in lamellar colonies slightly in comparison with prismatic glide (which is always favoured in aP nodules, such as in a-titanium). Note that basal glide has also been observed preferentially in lamellar colonies of a Ti–6Al–4V alloy in cryogenic conditions [20]. The aS/b interfaces could be suspected of being important obstacles for dislocations in lamellar colonies, but this is not the case: we have shown in a previous study that dislocations can move simultaneously in two aS plates and the b plate separating them [22]. So the aS/b interfaces in lamellar colonies are not the obstacles controlling the dislocation motion but, rather, the lattice friction on screw dislocations. Only the interfaces between lamellar colonies, between nodules or between lamellar colonies and nodules, play a role in the strengthening. 5. Conclusion In situ TEM deformation experiments have allowed the observation of the dislocation dynamics and the detailed study of the deformation micromechanisms in a titanium alloy with a complex microstructure. All gliding dislocations have an a-type Burgers vector. Due to their core structure, they are preferentially aligned with their screw direction, resulting in a large density of rectilinear long screw dislocations. This core structure is almost certainly spread in the basal, prismatic and first-order pyramidal planes, and depends on the stress orientation. It is the cause of the jerky motion of screw dislocations, cross-slip and the formation of intrinsic obstacles inducing the creation of macrokinks and dislocation multiplication by the open loop mechanism. We have also shown that the high lattice friction resulting from the three-dimensional core structure of the screw dislocations is responsible for the strengthening in lamellar colonies. Dislocations are always first emitted from a/b interfaces. They glide in prismatic as well as in basal planes, but basal glide is slightly favoured in lamellar colonies because of the compatibility stress due to the aS/b interfaces. Numerous titanium alloys have a similar a phase composition and similar microstructure. Therefore, the present results on the elementary deformation micromechanisms can most likely be generalized to them. References [1] Eylon D, Seagle SR. In: Gorgnin IV, Ushkov SS, editors. Titanium ’99: Science and Technology, St. Petersburg, Russia, 2000. p. 37. [2] Akhtar A, Teghtsoonian E. Metall Trans A 1975;6:2201.

P. Castany et al. / Acta Materialia 55 (2007) 6284–6291 [3] Naka S, Lasalmonie A, Costa P, Kubin LP. Philos Mag A 1988;57:717. [4] Biget MP, Saada G. Philos Mag A 1989;59:747. [5] Naka S, Kubin LP, Perrier C. Philos Mag A 1991;63:1035. [6] Farenc S, Caillard D, Couret A. Acta Metall Mater 1993;41:2701. [7] Farenc S, Caillard D, Couret A. Acta Metall Mater 1995;43:3669. [8] Zaefferer S. Mater Sci Eng A 2003;344:20. [9] Legrand B. Philos Mag B 1984;49:171. [10] Legrand B. Philos Mag A 1985;52:83. [11] Vitek V, Igarashi M. Philos Mag A 1991;63:1059. [12] Girshick A, Pettifor DG, Vitek V. Philos Mag A 1998;77:999. [13] Bacon DJ, Vitek V. Metall Mater Trans A 2002;33:721. [14] Aoki M, Nguyen-Manh D, Pettifor DG, Vitek V. Prog Mater Sci 2007;52:154. [15] Odegard BC, Thompson AW. Metall Trans A 1974;5:1207. [16] Neeraj T, Hou D-H, Daehn GS, Mills MJ. Acta Mater 2000;48:1225. [17] Neeraj T, Mills MJ. Mater Sci Eng A 2001;319–321:415. [18] Xiao L, Umakoshi Y. J Mater Sci Lett 2002;21:517. [19] Suri S, Viswanathan GB, Neeraj T, Hou DH, Mills MJ. Acta Mater 1999;47:1019. [20] Ambard A, Gue´taz L, Louchet F, Guichard D. Mater Sci Eng A 2001;319–321:404.

6291

[21] Savage MF, Tatalovich J, Mills MJ. Philos Mag 2004;84:1127. [22] Castany P, Pettinari-Sturmel F, Douin J, Coujou A. Mater Sci Eng A 2007; in press, doi:10.1016/j.msea.2006.10.183. [23] Bridier F, Villechaise P, Mendez J. Acta Mater 2005;53:555. [24] Banerjee D, Shelton CG, Ralph B, Williams JC. Acta Metall 1988;36:125. [25] Couret A, Crestou J, Farenc S, Mole´nat G, Cle´ment N, Coujou A, et al. Microsc Microanal Microstruct 1993;4:153. [26] Castany P, Pettinari-Sturmel F, Douin J, Coujou A. to be submitted. [27] Furubayashi E. J Phys Soc Jpn 1969;27:130. [28] Naka S, Lasalmonie A. J Mater Sci 1983;18:2613. [29] Couret A. Intermetallics 2001;9:899. [30] Castany P, Pettinari-Sturmel F, Douin J, Coujou A. In: Science and technology Ti-2007: proceedings of the 11th world conference on titanium 2007, in press. [31] Couret A, Caillard D. Acta Metall 1985;33:1447. [32] Couret A. Philos Mag A 1999;79:1977. [33] Moskalenko VA, Puptsova VN. Mater Sci Eng 1974;16:269. [34] Kim JH, Semiatin SL, Lee CS. Mater Sci Eng A 2005;394:366. [35] Plichta MR, Aaronson HI. Acta Metall 1980;28:1041. [36] Ankem S, Margolin H. Metall Trans A 1980;11:963.