Complexion-mediated martensitic phase transformation in Titanium

Abstract

The most efficient way to tune microstructures and mechanical properties of metallic alloys lies in designing and using athermal phase transformations. Examples are shape memory alloys and high strength steels, which together stand for 1,500 million tons annual production. In these materials, martensite formation and mechanical twinning are tuned via composition adjustment for realizing complex microstructures and beneficial mechanical properties. Here we report a new phase transformation that has the potential to widen the application window of Ti alloys, the most important structural material in aerospace design, by nanostructuring them via complexion-mediated transformation. This is a reversible martensitic transformation mechanism that leads to a final nanolaminate structure of [alpha''] (orthorhombic) martensite bounded with planar complexions of athermal ω (a-ω, hexagonal). Both phases are crystallographically related to the parent β (BCC) matrix. As expected from a planar complexion, the a-ω is stable only at the hetero-interface.

ARTICLE

Received 28 May 2016 | Accepted 29 Nov 2016 | Published 1 Feb 2017

The most efcient way to tune microstructures and mechanical properties of metallic alloys lies in designing and using athermal phase transformations. Examples are shape memory alloys and high strength steels, which together stand for 1,500 million tons annual production. In these materials, martensite formation and mechanical twinning are tuned via composition adjustment for realizing complex microstructures and benecial mechanical properties. Here we report a new phase transformation that has the potential to widen the application window of Ti alloys, the most important structural material in aerospace design, by nanostructuring them via complexion-mediated transformation. This is a reversible martensitic transformation mechanism that leads to a nal nanolaminate structure of a00 (orthorhombic) martensite bounded with planar complexions of athermal o (ao, hexagonal). Both phases are crystallographically related to the parent b (BCC) matrix. As expected from a planar complexion, the ao is stable only at the hetero-interface.

DOI: 10.1038/ncomms14210 OPEN

Complexion-mediated martensitic phase transformation in Titanium

J. Zhang1,2, C.C. Tasan3, M.J. Lai1, A.-C. Dippel4 & D. Raabe1

1 Max-Planck-Institut fr Eisenforschung, Max-Planck-Strasse 1, 40237 Dsseldorf, Germany. 2 State Key Laboratory for Mechanical Behavior of Materials, Xian Jiaotong University, Xian 710049, China. 3 Department of Materials Science and Engineering, Massachusetts Institute of Technology,77 Massachusetts Avenue, Cambridge, Massachusetts 02139, USA. 4 Deutsches Elektronen-Synchrotron DESY, Notkestrasse 85, D-22607 Hamburg, Germany. Correspondence and requests for materials should be addressed to J.Z. (email: mailto:[email protected]

Web End [email protected] ) or to C.C.T. (email: mailto:[email protected]

Web End [email protected] ) or to D.R. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 8:14210 | DOI: 10.1038/ncomms14210 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14210

Typical Ti alloys contain primarily phases with hexagonal lattice structure, the so-called a-Ti1. Alloying can be used to stabilize the high-temperature body-centred cubic

titanium phase also at ambient temperatures. These so-called b-titanium alloys have become a completely new material class due to their exceptional mechanical properties, namely ultralow elastic stiffness and gum-type deformation behaviour2. These are unique advantages, both for biomedical and aerospace applications, which no other alloy class can deliver. The underlying transformation mechanisms occurring in these alloys, which are held responsible for some of their key mechanical properties, are however not understood.

The new transformation phenomenon is observed in b-titanium alloys of Ti-23Nb-0.7Ta-2Zr, at.%. This system is a variation of the gum metal2, which has drawn signicant interest due to its unusual mechanical properties (low elastic stiffness and nearly hardening-free plasticity) and various deformation mechanisms that have been controversially reported as possible underlying reasons for such behaviour (bulk shearing36, stress-induced a00 martensitic phase transformation4,7,8, dislocation plasticity9,10, deformation twinning11,12, o-phase transformation13 and strain glass transition14). The b-phases instability, when exposed to mechanical loads, evidently plays a critical role in such gum alloys15,16; therefore, we aim at understanding the phase transformations in this system better. The observed mechanism involves a so far undiscovered coupling of several athermal phase transition steps at the interface, namely, b-a00 martensitic phase transformation and b-ao transition, creating planar complexions at the a00/b interfaces as well as nano-layered martensitic twinning (Fig. 1). It results in a nal nanolaminate composite microstructure throughout the bulk Ti gum metal.

A planar complexion1728 is referred to a metastable interfacial state conned and stabilized in the interfaces of the adjacent phases20. It can adopt states that have distinct interfacial structures and/or composition proles correlated with local minima in free energy21 and may undergo reversible structural transitions when the thermodynamic conditions vary19,27. The complexions known so far mostly appear at grain boundaries and

at planar hetero-interfaces, and originate from diffusional transitions27. In contrast, the new ao planar complexion observed in this study is induced by a diffusionless martensitic transformation (b-a00), that is, it forms through a diffusionless process. The connement and reversibility of this interface nanolaminate, consisting of adjacent orthorhombic a00, hexagonal o and twinned a00 nanolayers inside the host b-matrix realize a planar complexion state. A new transformation-induced planar complexion of o-structure is found. It possesses the athermal character and reversibility through the reversible b-ao transition. It forms from the b-matrix to accommodate the interfacial strain of a00/b-phase boundaries during b-a00 martensitic transition and further mediates the transition by inuencing the twinning mode, and results in a nal nanolaminate composite microstructure throughout the bulk Ti alloy on cooling.

In what follows, employing various experimental tools (for example, high-resolution transmission electron microscopy (HRTEM), in-situ synchrotron X-ray diffraction (SXRD), atom probe tomography (APT) and so on), we provide evidence of the structure, the reversibility and athermal nature of the individual interface features and transformation phenomena involved.

ResultsCharacterization of the phase transformation. We start the analysis with the characterization of the phase transformation occurring during quenching (Fig. 2). Figure 2a shows a one-dimensional (1D) SXRD pattern integrated over the two-dimensional (2D) SXRD pattern collected on an area detector from the as-quenched state. It indicates that the high-temperature b-phase partially transforms to a00 martensite and ao upon quenching. Phase fractions are estimated employing Rietveld renement of the diffractogram (Fig. 2a) to be 2.27, 76.55 and 21.18 vol.% for ao, a00 and retained b, respectively (tting results are presented in Supplementary Fig. 1). The lattice parameters of each structure are also calculated from the SXRD pattern (listed in Supplementary Table 1). Interestingly, the differential scanning calorimetry (DSC) curve (inset of Fig. 2a) shows no thermal peaks upon heating/cooling between 25 and 300 C, even at the ao and a00 reverse transformation nishing temperatures (200 and 245 C, determined by in-situ SXRD results presented later in Fig. 4). The absence of DSC peaks, despite the presence of large volume fractions of a00 and ao at room temperature, suggests that both transformation processes take place over a large temperature range, that is, they have a diffuse character.

Figure 2b shows a scanning electronic microscopybackscatter electrons (SEM-BSE) micrograph of a [011]b grain (the following lattice correspondences are used in the present paper: b-a00: [100]a00[100]b, [010]a00[011]b, [001]a00[011]b; b-ao: [1120]ao[011]b, [0001]ao[ 111]b). Two sets of edge-on

a00 twins are observed along { 220}a

00/{ 211}b, indicating that the

two edge-on { 211}b planes (that is, ( 211)b and (211)b) around

the [011]b zone axis can both serve as twinning plane for a00 martensite (that is, ( 220)a

00 and (220)a00). In Fig. 2c, TEMbright-eld image shows ne a00 twins with ( 220)a

00/

a

High T Room T

Step 3 Step 4

Heating

[afii9826]

a[afii9853]

[afii9825] T

a[afii9853]

Step 1

[afii9825] M

Step 2

Cooling

b

Parent phase [afii9826] (BCC)

Martensite[afii9825] (orthorhombic)

Complexion [afii9853] (hexagonal)

Figure 1 | Schematic illustration of the phase transformations and crystal structures of the phases involved. (a) Schematic illustration for complexion-mediated martensitic phase transformation in a b-Ti alloy matrix (from left to right): Step 1, formation of a00-matrix; Step 2, formation of rst ao planar complexion; Step 3, formation of a00-twin; and Step 4, formation of second ao planar complexion. (b) Crystal structures of b parent phase, a00 martensite and o complexion.

( 211)b twinning plane. For understanding the present twinning

mode, the selected area diffraction pattern (SADP) from the [011]b//[001]a00//[1120]ao zone axis is presented in Fig. 2d.

The {110}a00 faint spots (shufe spots), which are more clearly visible in the 2D SXRD patterns shown later, indicate the existence of the a00 martensite. Diffuse scatterings with intensity concentration at 1/3{211}b positions are the characteristics of the o structure. However, it is unlike the typical diffraction pattern of thermally induced ao (formed during quenching without the

2 NATURE COMMUNICATIONS | 8:14210 | DOI: 10.1038/ncomms14210 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14210 ARTICLE

a

0.10

b

SEM-BSE

c

TEM-BFI

[afii9826]

[afii9825]

[afii9853]

Cooling

Heating

300 400 500

Temperature (K)

Heat flow (W g)

DSC(10 K min1)

(220)/(211)

(211)

(211)

7k

6k

5k

4k

3k

2k

1k

Intensity (counts)

0.05

0.00

0.05

0.10

Zone [011][afii9826] Zone [011][afii9826]

TEM-DFI 2([afii9853])

f

0 3 4 5 6 7 8 9 10 11 12 13 14 15

2[afii9835] (degree)

d

e

TEM-DFI 1 ([afii9825] T + [afii9853])

[afii9825] M

200T + 2-201

2-201[afii9853]

020T + 1101

1101[afii9853]

DFI 1

220M,T

200M

020T + 1101

1101[afii9853]

SXRD25 C

(21.18 vol.%)

(76.55 vol.%)

(2.27 vol.%)

[afii9825] T

[afii9853]

(220)[afii9825] CT

Zone [011][afii9826]//[001][afii9825]//[1120]

[1120][afii9853]

Zone [011][afii9826]

Zone [011][afii9826]

Figure 2 | Characterization of the complexion-mediated martensitic phase transformation. (a) One-dimensional SXRD pattern, with the DSC inset showing no clear thermal peaks. (b) SEM-BSE micrograph (scale bar, 2 mm) from a [011]b grain shows two sets of {220}a CTs (bulk sample).

(c) TEMbright-eld image (BFI) shows a set of CT (scale bar, 100 nm). (d) Corresponding SADP from the area encircled in c. (e,f) DFIs (TEMDFI 1 and 2) from the overlapping spot (cyan and blue dotted circles in d) and from the ao spot (blue circle in d), which indicate that ao planar complexion only locate along the CT boundaries (both scale bars, 20 nm). CT, compound twin.

assistance of an athermal b-a00 transformation), where two sets of 1/3{211}b reections are present and of similar intensity29,30.

Only one set is present here and, moreover, half of the 1/3{211}b spots have much higher intensity than the others as indicated by cyan arrows (Fig. 2d). To understand the origin of the different intensities of the 1/3{211}b spots, two dark-eld images (DFIs), referred to as TEMDFI 1 and DFI 2, were taken by using one brighter and one weaker 1/3{211}b spot as shown in Fig. 2e,f, respectively. The TEMDFI 1 in Fig. 2e shows that one of two martensite laths is bright and the other is dark. The boundaries between them are also bright but in much lower intensity. This indicates that both the bright laths and the boundaries contribute to the brighter 1/3{211}b spots. More interestingly, in TEMDFI 2 (Fig. 2f), only the boundaries are bright. Therefore, the TEM

DFIs reveal that the brighter 1/3{211}b spots originate from both a00 twin (a00T) and ao, whereas the ones of lower intensity are formed solely from ao along the twin boundaries. With this, it can be concluded that the SADP in Fig. 2d reveals higher intensity 1/3{211}b spots and primary spots correspond to a { 220}a

00 compound twin (CT). It is noteworthy that the { 220}a

00 CT is the only twinning mode found in all TEM observations in this study. This is in contrast with binary Ti-Nb, where the major twinning mode of a00 is {111}a00 type I twinning and no CTs were found30,31. Moreover, the TEMDFIs reveal that 1/3{211}b spots originate from the preferentially formed ao phase along the pure shear direction { 211}o1114b of the b-

a00 transformation32. It indicates that the formation of ao is induced by the b-a00 transformation. As a result, a nanolaminate

microstructure is created, composed of sequentially arranged layer-by-layer a00 CT bounded with ao lms, which resembles a planar complexion state20,27.

Atomic conguration of a00 CT and ax planar complexion. Figure 3 shows HRTEM images from the [011]b zone axis, providing the atomic conguration of the a00 CT and ao planar complexion. Figure 3a shows a HRTEM where several successive nanolayers of a00 CT bounded by ao planar complexions were captured. The corresponding fast Fourier transform (FFT) pattern is shown in Fig. 3a (inset), which is almost identical to the SADP in Fig. 2d. An inverse FFT image of higher magnication (from the green zone in Fig. 3a) is shown in Fig. 3b,c, focusing on the internal structure of a00 CT and ao, respectively. Figure 3b shows the atomic conguration of a00 CT, where red/cyan lines along the (020)M,T planes reveal the deviation of the (020)M,T lattice planes across the ao layer, which also indicates the a00 strain accommodation direction by ao as denoted by thicker red/cyan arrows (with white background). The lattice spacings in the centre of a00M and a00T are d(020)M,T 0.24 nm and

d(200)M,T 0.16 nm in agreement with the values obtained from

SXRD (Supplementary Table 1). Furthermore, analysis of one ao zone has been performed, as shown in Fig. 3c. As plotted in the inset, the distance between two middle atoms of ao was measured and plotted with respect to the atomic layers along the purple lines. The following structures can be identied: b-IC_o (incommensurate o)-C_o (commensurate o)3335.

This sequence provides the rst clear perception revealing the

NATURE COMMUNICATIONS | 8:14210 | DOI: 10.1038/ncomms14210 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14210

a

FFT

b

c

(220)

[0001]

[1100]

200T

[afii9826]

020M

(211)[111]

(211)[111]

(211)[111]

(211)[111]

IC_[afii9853] C_[afii9853]

[afii9826]

0.16

[200]

[020] (220)

[200]

Distance (nm)

0.14

[afii9853]

0 4

Atomic layers

8 12 16

IFFT

0.18 [afii9826]

Figure 3 | HRTEM images of a00 CT and ao planar complexion. (a) Low-magnication HRTEM image; the inset shows the corresponding FFT diffraction pattern (scale bar, 10 nm). (b) Atomic conguration of a00 CT, where red and cyan arrows with white background represent the shear and accommodation directions of a00 matrix and its twin during transformation. (c) Analysis of atomic conguration within ao planar complexion: b-IC_o (incommensurate o)-C_o (commensurate o), where the inset is the plot of atom distance between the middle atoms along the purple lines. Scale bars, 2 nm (b,c).

a b

0001 a [afii9853]

[011][afii9826]//[001][afii9825]

600

4.0

110 [afii9825]

1000

0002 a [afii9853]

5,000

0003 a [afii9853]

222 [afii9826]

240 [afii9825]

204 [afii9825]

(220)[afii9825] CT + (220)[afii9825] CT

(220)[afii9825] CT + (220)[afii9825] CT

25 C

Intensity (counts)

Cool Heat

25 C 245 C

200 C 150 C 120 C 105 C 80 C 60 C 50 C 40 C 25 C

4.4 8.02[afii9835] (degree)

8.4 12.4 12.6

Heating

c

d

g

[011][afii9826]//[001][afii9825]

60 C 150 C

[afii9825]T

[afii9825]M

[afii9826]

e

f

Cooled

25 C

a[afii9853][afii9826]

200 C

[afii9826]

245 C

Figure 4 | Results of in-situ SXRD heating/cooling experiments. (a) Line proles integrated from 2D SXRD patterns. (bg) Representative 2D SXRD patterns at 25, 60, 150, 200, 245 C (during heating) and 25 C (after cooling), respectively. Scale bar, 5.

individual steps contributing to the b-ao transformation mechanism. Consequently, the C_o and IC_o zones can be outlined in the ao planar complexion regions as shown in Supplementary Fig. 2. On the boundary of the C_o zones, 1/2o1114{211} dislocations are identied. Moreover, one segment of a twin boundary without ao complexion is found between the two complexion regions outlined, where the

interfacial stress falls below a critical value. The histogram of the ao planar complexion-width distribution is shown in

Supplementary Fig. 3. It has been calculated based on

TEMDFI 2 (Fig. 2f) for better statistics.

This observation also provides another evidence that the ao planar complexion is induced by a shear stress exerted by the a00 transformation, as this shear stress along { 211}o1114b

4 NATURE COMMUNICATIONS | 8:14210 | DOI: 10.1038/ncomms14210 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14210 ARTICLE

(the pure shear of b-a00 transformation) has a perfect lattice correspondence for promoting the b-ao transformation, which will be explained further below. Furthermore, ao plays a role in accommodating a00 martensitic twins as manifested by a pair of remaining planar strain feature on both sides of the ao layers (Fig. 3b). Hence, a new complexion/second-phase accommodation mechanism is identied in martensitically transforming materials: the interfacial stress of b-a00 is accommodated by triggering the formation of planar complexion through b-ao transformation.

Reversibility of the a00 CT and ax planar complexion. The results of the in-situ heating/cooling SXRD experiments shown in Fig. 4 provide another direct evidence for the co-existence and the reversibility of a00 CT and ao complexion. Figure 4a shows three segments of the 1D SXRD patterns integrated from the 2D SXRD patterns. Some representative 2D SXRD patterns are shown in Fig. 4bg. In Fig. 4a, gradual disappearance of the o and a00 reections within 2y ranges of 3.8 to 4.6 and 7.8 to 8.6 during heating, and their restoration after cooling to 25 C prove the reversibility of b-ao and b-a00 transformations during heating/cooling. Another broad peak within the 2y range of 12.3 to 12.7 consists of (240)a00, (0003)o, (222)b and (204)a00 reections. During heating, the overlapping peaks become narrower at 200 C, indicating full reversion of ao to b around 200 C, as one ao peak is located at higher 2y. This is also revealed more clearly in the analysis of 2D SXRD patterns, presented in the following section. When the temperature is further increased, the peak becomes even smaller at 245 C, which is the reverse transformation nishing temperature of a00-b. Hence, at 245 C, a single b-phase is achieved, which gains even more solid proof through the 2D SXRD pattern captured at 245 C (Fig. 4e).

Next, the 2D SXRD patterns in Fig. 4bg are further analysed, to provide more insights on the transformation kinetics and overall sequence of a00T-b, ao-b and a00M-b transitions.

Based on the grain identication performed on the 2D SXRD pattern at 245 C (single b-phase), as presented in the

Supplementary Fig. 4, diffraction intensities in the current in-situ SXRD pattern originated mainly from three grains: two of them are [011]b and the other one is [012]b. In the following, we focus on the diffraction pattern of one [011]b grain with near on-pole condition, with its diffraction pattern outlined in green in

Supplementary Fig. 4.

In Fig. 4b, a 2D SXRD pattern obtained at 25 C (before heating) is shown and the primary diffraction spots are linked by red lines to indicate that the grain is mainly in a00 martensite state.

Additional to the primary spots, extra spots at 1/2{211}b and 1/3{211}b locations are also found. Based on the SADP analysis shown in Fig. 2d, two sets of { 220}a

00 CT diffraction spots from the [001]a00//[011]b zone axis are identied as marked in cyan and magenta, which is consistent with the SEM-BSE image in

Fig. 2b. The (020)a00 twin spots of each CT are indicated by cyan and magenta arrows (pointing to the o1114b shear direction of b-a00 transformation), respectively; in contrast, {110}a00 spots caused by {011}o0114b shufing associated with the a00 transformation are encircled (red). One noticeable difference between the diffraction patterns of TEM (Fig. 2d) and SXRD (Fig. 4b) is the absence of (1100)o and (2200)o spots in the

SXRD pattern. These signals are actually forbidden diffraction spots in the hexagonal lattice structure. Their appearance in TEM SADP (Fig. 2d) is due to the double diffraction effect36. In the 2D SXRD pattern, these spots disappear for the following two reasons. (1) The SXRD beamline used for this study has a much larger wavelength (l 0.020727 nm) as compared with TEM

operated at 200 kV (l 0.00251 nm) and hence a much smaller

radius of the Ewald sphere, which leads to lower double diffraction intensity values. (2) The thickness of the samples used in SXRD is B1 mm. In contrast, the TEM thin foils used in this work are below 200 nm in thickness. Therefore, in the SXRD experiments, double diffraction effects drop to near-zero intensity before penetrating the entire sample, different than in corresponding TEM diffraction experiments. For clarity, the key diagrams to this complex diffraction pattern are provided in Supplementary Fig. 5, which consists of two overlapping diffraction patterns from two possible edge-on a00 CT: ( 220)a

00 CT and ( 2 20)a

00 CT. During heating, the intensity of the extra spots due to a00 CT and ao decreases gradually (Fig. 4c), as can be seen more clearly in Supplementary Movie 1. In Fig. 4d, the disappearance of twin spots (that is, the intensity of arrow-pointed spots becomes the same as that of the(0001)o ao spots at the smallest 2y angle) at 150 C indicates that all a00T transforms reversely into b rst around 150 C. The 2D SXRD pattern taken at 200 C (Fig. 4e) shows the disappearance of o spots, indicating that the ao fully reverses back to b at around 200 C. When the sample is further heated up, the 2D SXRD pattern at 245 C (Fig. 4f) reveals that the (110)a00 shufe spot vanishes, indicating that a00 fully transforms back to b. After cooling to 25 C, the 2D SXRD pattern in Fig. 4g shows that the original diffraction pattern (Fig. 4b) is nearly fully restored. This in-situ heating/cooling SXRD analysis proves the complete reversibility of a00 CT and its associated ao complexion.

a

[afii9838]

b

500

30

20

10

20

10 20 30

O observed

O binomial Ti binomial Nb binomial Zr binomial Ta binomial

[afii9839](=[afii9851]

2 / N + [afii9851]

2)

APT

Element

O 0.0093 Ti

400

10

(211)[111]

Ti observed Nb observed Zr observed Ta observed

0.17100.07160.0447

[afii9839]=0 random [afii9839]=1 clustering

20

300

0.1443

Counts

30

40

Z

Nb

Zr Ta

200

100

Analysed volume : 75 72 71 nm34.6 Million ions

50

60

70

0

0

10 20 30 40 50 60 70 80 90 100

Concentration (at.%)

Figure 5 | APT analysis. (a) APT elemental map displays the atomic-scale local composition. (b) Frequency distribution curves for Ti, Nb, Ta,Zr and O atoms obtained from the APT data shown in a; binomial distributions for average solute contents are also shown for comparison, which stands for the ideal distribution in a homogenized alloy.

NATURE COMMUNICATIONS | 8:14210 | DOI: 10.1038/ncomms14210 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14210

The results mentioned above also suggest that the { 220}a

00

CT-induced b-ao transformation is martensitic, hence without composition change. To conrm this, APT analyses were carried out on tips that were focused ion beam (FIB) milled from a {111}b grain. The cluster analysis reveals that all alloying elements are distributed homogenously, indicating an athermal transformation mechanism (Fig. 5).

DiscussionTo elucidate the underlying mechanisms of the ao planar complexion mediated transformation, a schematic representation of the transformation process on cooling is presented in Fig. 6. When the b-a00 transformation starts (Step 1), the ( 211)

[ 1 1 1]b shear stress at the lower interface of the a00M and b

is built up as indicated by the red arrows. Simultaneously, conning back-stresses from the surrounding untransformed b matrix are built up, acting against the shear direction of a00M, the extent of which is indicated by the height of blue energy bands on both sides. These two stress components act in opposite directions, creating a compressive stress on the ( 211)b plane

along the effective axis [111]b of the ao formation35,37. Growth of a00 stops and ao forms when the stress reaches a critical value required for inducing b-ao (as illustrated in Supplementary Fig. 6a,b). This marks Step 2, the occurrence of which indicates that ao is thermodynamically not stable but can be stabilized under the assistance of stress. Formation of the ao planar complexion accommodates the strain (evident in Fig. 3b) and hence decreases the interfacial stress. When the local stress drops to values below the critical value, the growth of the ao layer stops, as the ao region is still thermodynamically unfavourable. This is also the reason why the occurrence of the b-ao transition is strictly conned to the interface region (forming a planar complexion of a size around 2 nm) and not viable outside of it. As the ao transformation produces shear along the ( 211)[ 1 1 1]b component, a00T of the { 220}a

00/

compound twinning character. It results in a nal nanolaminate composite microstructure throughout the bulk Ti gum metal.

In conclusion, a new reversible complexion-mediated marten-sitic phase transformation phenomenon has been discovered in titanium, which is mediated by a planar complexion state. It involves the joint formation of the { 220}a

00/{ 211}b

a00 martensitic CT and the accommodating ao planar complexion. TEM analysis shows that the two transformation steps are coupled, constituting a complexion interfacial state: the ao transformation is induced by a shear stress initiated by pure shear { 211}o1114b during b-a00, which has perfect lattice

correspondence for inducing the ao complexion. In-situ heating/cooling SXRD and APT prove that both a00 and ao transformations are reversible and martensitic. This coupled and complexion-mediated transformation mechanism enables novel nanostructuring and hence strengthening opportunities for Ti alloys.

Methods

Alloy synthesis and heat treatment. A reversible complexion-mediated martensitic phase transformation is observed in the oxygen-free gum metal (Ti-23Nb-0.7Ta-2Zr, at.%) upon quenching. The actual composition of the alloy was determined by chemical analysis as Ti-23Nb-0.67Ta-1.96Zr-0.26O (at.%). The chemical composition was tested by inductively coupled plasma optical emission spectrometry and infrared absorption spectroscopy measurements. The melt of the ingot was produced under argon atmosphere in an arc-melting furnace from pure elements, cast into a copper mould, homogenized at 1,200 C for 4 h and then furnace cooled. Smaller samples of the alloy were further annealed in vacuum quartz tubes for 1 h at 1,000 C and subsequently water quenched as the tube was broken simultaneously, to avoid iso-o formation39.

Synchrotron measurements. SXRD measurements were performed on the bulk samples with a thickness of B1 mm, at beam-line P02.1 at PETRA III (DESY Hamburg, Germany) with the wavelength l 0.020727 nm40. Beam size used was

500 500 mm2. The 2D SXRD patterns were collected on an area detector

(PerkinElmer XRD1621). Azimuthal integration was performed using the software FIT2D and the obtained 1D SXRD patterns were then analysed by Rietveld renement using MAUD41. The 1D SXRD pattern in Fig. 2a indicates the co-existence of a00 and ao phases, and a small volume of remaining b-phase. The ao reections are quite diffuse/broad and low in intensity. The patterns are rather complex, owing to the overlap with the a00 and b peaks, which will fail tting or

{ 211}b compound twinning character is formed (Step 3) due

to the requirement of interface energy minimization. The second

ao layer forms (Step 4) also because of a planar compressive stress accumulated by the b-a00T transition. However, the formation of the a00T involves very large shear stresses (due to 35% twinning shear strain) along ( 211)[ 1 1 1]b and

hence the critical stress of the transformation b-ao is reached faster (in fewer atomic layers) than that caused by a00M in Step 1.

Moreover, the formation of the second ao layer stops the further growth of the a00T layer. As a results, the width of the a00T layer should be much thinner than that of the a00M layer, which is consistent with the presented TEM observations (Fig. 2e). The accommodation potential of both ao layers are the same and therefore the two ao layers should have similar thickness (evident in Fig. 2f). These four individual structural steps are considered as one transformation event, mediated by a planar complexion state at the interface. With further cooling, the nal nanolaminate microstructures can be formed after repeating the four-step transformation event, as illustrated in Supplementary Fig. 6c,d. Two more SEM-BSE images in Supplementary Fig. 7 provide a larger eld of view of such nanolaminate microstructures from both [011]b and [111]b orientations.

Such a nanoscale successive transformation process, when progressing over a wider temperature range, shows no abrupt thermal signal as is evident from the DSC measurement (inset in Fig. 2a). This phenomenon is similar to the disappearance of the DSC peak observed during the strain glass transition38. Based on the above discussion, it is quite clear that the formation of ao planar complexions is to accommodate b - a00 interfacial strain and also mediates the b-a00 transition to possess { 220}a

00

Second

(211)[111] [011]

Shear stress

[200]

[200]

(211)[111]

[011]

Zone [011][afii9826] // [001][afii9825] // [1120][afii9853]

Step 1:

[afii9825] M

Step 2: a[afii9853]

Step 4: a[afii9853]

Step 3: [afii9825] T

Figure 6 | Schematic representation of atomic movements associated with the complexion-mediated transformation upon cooling. Open and solid circles represent the rst and second layers of atoms viewed from [011]b orientation. Colour in the gure: black to b-phase; blue to ao;

red to a00-matrix; green to a00-twin. For better clarity, only b-indexes are used here. The transformation starts from the top and propagates to the bottom during cooling as divided into four steps: Step 1 (formation of a00-matrix), Step 2 (formation of rst ao layer), Step 3 (formation of a00-twin) and Step 4 (formation of second ao layer).

6 NATURE COMMUNICATIONS | 8:14210 | DOI: 10.1038/ncomms14210 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14210 ARTICLE

lead to wrong tting results. Hence, the ao reections were not included during the rst-step Rietveld renement (shown in Supplementary Fig. 1a), through which a successful tting (Rw 9.0%) was reached and the volume fractions of a00 and

remaining b-phase were determined to be 82 and 18 vol.%, respectively. By using rst step renement as the starting point, second step Rietveld renement with a

o reections added, converged well (Rw 7.84%), as presented in Supplementary

Fig. 1b. From this analysis we suggest the room temperature phase fractions for a o, a00 and b as 2.27, 76.55 and 21.18 vol.%, respectively. In-situ heating/cooling

SXRD experiments were carried out under the same condition with a home-made heating stage as illustrated in Supplementary Fig. 8.

Physical and microstructural characterization. The DSC measurement was performed in a Mettler Toledo DSC1 at heating/cooling rate of 10 C min 1.

The sample was rst heated up from room temperature (25 C) to 300 C and subsequently cooled down to 25 C. TEM samples were electro-polished (Struers Tenupol 5) at 6 C using an electrolyte of A3. TEM observations were performed in a JEOL JEM-2200 FS at an acceleration voltage of 200 kV, through which bright-eld images, DFIs, SADPs and HRTEM images were recorded by a Gatan SDD Camera. FFT and inverse FFT analysis were applied to the HRTEM images. SEM-BSE and electron backscatter diffraction (EBSD) were carried out in a Zeiss Crossbeam XB 1540 FIB-SEM instrument (Carl Zeiss SMT AG, Germany). EBSD was used to determine the grain orientation where SEM-BSE micrographs were taken. To maintain consistency with the notation used for the TEM results, [011] and [111] are used to denote the plane normals of o1104 and o1114 grains determined by EBSD. Elemental distribution at atomic scale was studied using local electrode APT (LEAP 3000X HR, Cameca Inc.). APT tips were prepared from a [111]b grain (as shown in Supplementary Fig. 7b), following the standard procedure42 in a FEI Helios Nanolab 600i dual-beam FIB.

Data availability. The data that support the ndings of this study are available from the corresponding authors on request.

References

1. Leyens, C. & Peters, M. Titanium and Titanium Alloys: Fundamentals and Applications (Wiley-VCH, 2003).

2. Saito, T. et al. Multifunctional alloys obtained via a dislocation-free plastic deformation mechanism. Science 300, 464467 (2003).

3. Gutkin, M. Y., Ishizaki, T., Kuramoto, S., Ovidko, I. A. & Skiba, N. V. Giant faults in deformed Gum Metal. Int. J. Plasticity 24, 13331359 (2008).4. Obbard, E. G. et al. Mechanics of superelasticity in Ti-30Nb-(810)Ta-5Zr alloy. Acta Mater. 58, 35573567 (2010).

5. Withey, E. A., Minor, A. M., Chrzan, D. C., Morris, J. W. & Kuramoto, S. The deformation of Gum Metal through in situ compression of nanopillars. Acta Mater. 58, 26522665 (2010).

6. Plancher, E., Tasan, C. C., Sandloebes, S. & Raabe, D. On dislocation involvement in TiNb gum metal plasticity. Script. Mater. 68, 805808 (2013).

7. Tane, M. et al. Low Youngs modulus in Ti-Nb-Ta-Zr-O alloys: cold working and oxygen effects. Acta Mater. 59, 69756988 (2011).

8. Vorontsov, V. A., Jones, N. G., Rahman, K. M. & Dye, D. Superelastic load cycling of Gum Metal. Acta Mater. 88, 323333 (2015).

9. Yang, Y., Li, G. P., Cheng, G. M., Li, Y. L. & Yang, K. Multiple deformation mechanisms of Ti-22.4Nb-0.73Ta-2.0Zr-1.34O alloy. Appl. Phys. Lett. 94, 061901 (2009).

10. Besse, M., Castany, P. & Gloriant, T. Mechanisms of deformation in gum metal TNTZ-O and TNTZ titanium alloys: a comparative study on the oxygen inuence. Acta Mater. 59, 59825988 (2011).

11. Yang, Y. et al. Formation of zigzag-shaped {112}o1114 beta mechanical twins in Ti-24.5Nb-0.7Ta-2Zr-1.4O alloy. Script. Mater. 66, 211214 (2012).

12. Lai, M. J. et al. Origin of shear induced b to o transition in TiNb-based alloys. Acta Mater. 92, 5563 (2015).

13. Yano, T., Murakami, Y., Shindo, D., Hayasaka, Y. & Kuramoto, S. Transmission electron microscopy studies on nanometer-sized o phase produced in Gum

Metal. Script. Mater. 63, 536539 (2010).14. Wang, Y. et al. Strain glass transition in a multifunctional b-type Ti alloy. Sci. Rep. 4, 3995 (2014).

15. Raabe, D., Sander, B., Friak, M., Ma, D. & Neugebauer, J. Theory-guided bottom-up design of beta-titanium alloys as biomaterials based on rst principles calculations: theory and experiments. Acta Mater. 55, 44754487 (2007).

16. Banerjee, D. & Williams, J. C. Perspectives on titanium science and technology. Acta Mater. 61, 844879 (2013).

17. Tang, M., Carter, W. C. & Cannon, R. M. Diffuse interface model for structural transitions of grain boundaries. Phys. Rev. B 73, 024102 (2006).

18. Tang, M., Carter, W. C. & Cannon, R. M. Grain boundary transitions in binary alloys. Phys. Rev. Lett. 97, 075502 (2006).

19. Dillon, S. J. & Harmer, M. P. Multiple grain boundary transitions in ceramics: a case study of alumina. Acta Mater. 55, 52475254 (2007).

20. Dillon, S. J., Tang, M., Carter, W. C. & Harmer, M. P. Complexion: a new concept for kinetic engineering in materials science. Acta Mater. 55, 62086218 (2007).

21. Baram, M., Chatain, D. & Kaplan, W. D. Nanometer-thick equilibrium lms: the interface between thermodynamics and atomistics. Science 332, 206209 (2011).

22. Harmer, M. P. The phase behavior of interfaces. Science 332, 182183 (2011).23. Dillon, S. J., Harmer, M. P. & Rohrer, G. S. The relative energies of normally and abnormally growing grain boundaries in alumina displaying different complexions. J. Am. Ceram. Soc. 93, 17961802 (2010).

24. Frolov, T., Divinski, S. V., Asta, M. & Mishin, Y. Effect of interface phase transformations on diffusion and segregation in high-angle grain boundaries. Phys. Rev. Lett. 110, 255502 (2013).

25. Frolov, T., Olmsted, D. L., Asta, M. & Mishin, Y. Structural phase transformations in metallic grain boundaries. Nat. Commun. 4, 1899 (2013).

26. Kaplan, W., Chatain, D., Wynblatt, P. & Carter, W. C. A review of wetting versus adsorption, complexions, and related phenomena: the rosetta stone of wetting. J. Mater. Sci. 48, 56815717 (2013).

27. Cantwell, P. R. et al. Grain boundary complexions. Acta Mater. 62, 148 (2014).

28. Kuzmina, M., Herbig, M., Ponge, D., Sandlbes, S. & Raabe, D. Linear complexions: conned chemical and structural states at dislocations. Science 349, 10801083 (2015).

29. Kim, H. Y., Ikehara, Y., Kim, J. I., Hosoda, H. & Miyazaki, S. Martensitic transformation, shape memory effect and superelasticity of Ti-Nb binary alloys. Acta Mater. 54, 24192429 (2006).

30. Chai, Y. W., Kim, H. Y., Hosoda, H. & Miyazaki, S. Interfacial defects in Ti-Nb shape memory alloys. Acta Mater. 56, 30883097 (2008).

31. Inamura, T. et al. Composition dependent crystallography of a00-martensite in TiNb-based b-titanium alloy. Philos. Mag. 87, 33253350 (2007).

32. Tahara, M., Kim, H. Y., Inamura, T., Hosoda, H. & Miyazaki, S. Lattice modulation and superelasticity in oxygen-added b-Ti alloys. Acta Mater. 59, 62086218 (2011).

33. Sikka, S. K., Vohra, Y. K. & Chidambaram, R. Omega phase in materials. Prog. Mater. Sci. 27, 245310 (1982).

34. Schryvers, D. & Tanner, L. E. High-resolution electron microscopy observations of athermal omega phase in Ti-Mo alloys. Mater. Sci. Forum 56, 329334 (1990).

35. Banerjee, S. Omega phase transformation morphologies and mechanisms. Int.J. Mater. Res. 97, 963977 (2006).36. Williams, D. B. & Carter, C. B. Transmission Electron Microscopy: a Textbook for Materials Science (Springer Science & Business Media, 2009).

37. Banerjee, S. & Mukhopadhyay, P. Phase Transformations Examples from Titanium and Zirconium Alloys (Pergamon, 2007).

38. Zhang, J. et al. Stress-induced strain glass to martensite (R) transition in a Ti50Ni44.5Fe5.5 alloy. Phys. Rev. B 83, 174204 (2011).

39. Zhang, J., Rynko, R., Frenzel, J., Somsen, C. & Eggeler, G. Ingot metallurgy and microstructural characterization of Ti-Ta alloys. Int. J. Mater. Res. 105, 156167 (2014).

40. Dippel, A.-C. et al. Beamline P02.1 at PETRA III for high-resolution and high-energy powder diffraction. J. Synchrotron Radiat. 22, 675687 (2015).

41. Lutterotti, L. Total pattern tting for the combined sizestrainstresstexture determination in thin lm diffraction. Nucl. Instrum. Methods Phys. Res. B Beam Interact. Mater. Atoms 268, 334340 (2010).

42. Thompson, K. et al. In situ site-specic specimen preparation for atom probe tomography. Ultramicroscopy 107, 131139 (2007).

Acknowledgements

We acknowledge the funding by the European Research Council under the EUs 7th Framework Programme (FP7/20072013))/ERC Grant agreement 290998 SmartMet. J.Z. acknowledges support of Innovative Research Team in University (IRT13034), 973 Programs of China (2014CB644003), National Key Research and Development Program of China (2016YFB0701302) and NSFC (51501145, 51320105014and 51621063). Parts of this research were carried out at PETRA III at DESY,a member of the Helmholtz Association (HGF). The contributions of H. Springer, K.G. Pradeep, M. Nellessen, B. Grabowski and A. Kostka are also gratefully acknowledged.

Author contributions

D.R., C.C.T. and J.Z. designed the research. J.Z. was the lead experimental scientist of the study. J.Z., M.J.L. and A.-C.D. interpreted the data. J.Z., C.C.T. and D.R. wrote the paper. All authors discussed the results and commented on the manuscript.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

NATURE COMMUNICATIONS | 8:14210 | DOI: 10.1038/ncomms14210 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14210

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Zhang, J. et al. Complexion-mediated martensitic phase transformation in Titanium. Nat. Commun. 8, 14210 doi: 10.1038/ncomms14210 (2017).

Publishers note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional afliations.

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

r The Author(s) 2017

8 NATURE COMMUNICATIONS | 8:14210 | DOI: 10.1038/ncomms14210 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications