STRAIN INDUCED PHASE TRANSFORMATION IN Ti-15Mo β ALLOY

β-Ti alloys have been chosen for biomedical applications attributed to a combination of high strength, high fatigue resistance, good corrosion resistance and more importantly low modulus closer than other metallic materials for implants and osseointegrated prosthesis to the cortical elastic modulus (4-30 GPa). However, the phase constituents and phase transformation are still under dispute. A Ti-15Mo alloy after severe cold plastic deformation is studied to reveal the phases and phase transformation by TEM techniques. Athermal ω phase was observed in all samples evidencing the high stability of ω phase compared to β phase. However, the β to ω transformation does not proceed to a completion in heat treatment. Strain induced phase transformation happens in cold-draw wires through a coordinated shuffle of atoms along the {2 1 1}planes of β phase leading to the reduction on the <1 1 1> directions. The atomic level of shear causes the transformation of β to ω. The transformation of β to athermal ω under a train is not in a stable state but having a variable crystal structure between β and ω.

β-Ti alloys have been chosen for biomedical applica ons a�ributed to a combina on of high strength, high fa gue resistance, good corrosion resistance and more importantly low modulus closer than other metallic materials for implants and osseointegrated prosthesis to the cor cal elas c modulus   [1][2][3][4][5][6]. β-Ti alloys are also a group of important structural materials for aeronau cal applica ons due to lightweight, high strength and duc lity, and addi onally good workability [7][8][9][10]. The excellent corrosion resistance of β-Ti alloys makes the alloys suitable for applica ons in a corrosive environment, such as downhole oil drilling/explora on equipment [7,11,12]. The mechanical proper es of β-Ti alloys can be tailored in a broad range through microstructural control for different applica ons, i.e. UTS from 690 MPa to 1586 MPa, modulus from 55 GPa to 110 GPa, elonga on up to 17% and pseudoelas city [2,4,7,[13][14][15][16][17][18]. However, β-Ti phase has a bodycentered cubic (BCC) crystal structure and is not thermodynamic stable at room temperature. It is preserved by the addi on of alloying elements called β-phase stabilizers [7]. The metastable nature of the β-Ti phase has caused a lot of confusion about the phase transforma on of β-Ti alloys in processing, par cularly in cold working/annealing, and therefore about the strengthening mechanisms under different condi ons [19][20][21][22][23][24][25]. The phase transforma on of β-Ti to α-Ti through a series of β + β‫׳‬ and ω, α‫׳‬ or α‫״‬ etc. is further complicated by a recently reported O‫׳‬ metastable phase [26][27][28][29][30].
The β->α phase transforma on is realized through slow cooling from the β phase or aging below the β transus with a certain crystal orienta on rela onship (OR) between the hexagonal close packed (HCP) α and the parental BCC β phase to reach a thermodynamic stable α phase [31,32]. Meanwhile, metastable ω, α‫׳‬ or α‫״‬ and O‫׳‬ phases can form during the β to α transforma on [19,20,29,33,34].
Both athermal and isothermal ω phase with the same hexagonal crystal structure can precipitate from the parent β phase. The athermal ω phase forms during β quenching, whereas the isothermal ω phase forms during ageing at temperatures below 823 K [13]. It appears as small (nanoscaled) cuboidal or ellipsoidal par cles with a specific OR with the β matrix [20]. The two kinds of martensite, α‫׳‬ (hexagonal) and α‫״‬ (orthorhombic) phases with an acicular or a plate shape form during rapid cooling from a temperature above or near β transus. The solute-depleted β and the solute-enriched β‫׳‬ phases appear concomitantly with a rod-like morphology and are formed through spinodal decomposi on at the early stage of low-temperature ageing. The structure of the two β products are BCC, as same as the parent β phase, resul ng in no extra diffrac on spots other than those of β in a selected area electron diffrac on (SAED) pa�ern [35]. The phase transforma on is further complicated by the fact that α to ω and β to ω transforma on are enhanced by stress, resul ng in combined phase transforma on induced plas city (TRIP) and twinning induced plas city (TWIP) [21,22,[36][37][38].
This paper presents findings on the phase transforma on of a Ti-15Mo alloy a�er severe cold working in comparison with different heat treatment processes. An effort has been made on understanding the evolu on of phases in severe cold working β-Ti alloy in rela onship to its special proper es, pseudoelas city and a combina on of high strength and low modulus.

Materials and Experiments
A Ti-15Mo β-alloy has been selected for the study to understand the phase transformation in both heat treatment and cold working due to its simplicity in terms of compositional variables. Ti-15Mo rods with a 5 mm diameter that have been solutionized at 740 o C for 20 minutes and then quenched in water were chosen as the raw material. For heat treatment, the rods were solutionized at 900 o C for 2 hours and cooled to room temperature at different cooling rates from 1 o C/s to 1000 o C/s. The rods were also cold drawn to an 80 micrometers diameter under room temperature for microstructural analysis. The microstructure of the rods is presented in Fig. 1 for comparison with the heat-treated and cold drawn microstructures in the following Results section.
Both original and heat treated rods were cut and polished for microstructural analysis following a standard metallographic procedure. The SEM images were taken by using an FEI 3D Quanta Dual Beam SEM operated at 20 kV. A Philips XL-40 microscope with EBSD capability was operated at 30 kV acceleration voltage and a working distance of 10 mm for EBSD and SEM images. Discs of 3 mm diameter were also punched from thin discs perpendicular to the axis of the rods and electropolished to perforation in a Tenupol ElectropolisherTM at -35 o C and 70 volts, using the following reagent: 4% perchloric acid, 2% hydrochloric acid, 36% butyl alcohol and 58% methanol. Helios 600i dual-column focused ion beam (FIB)/field emission scanning electron microscope (FESEM) operated at 30 kV was used for taking thin films for transmission electron microcopy from the wires and rods. Transmission electron microscopy (TEM) was done on an FEI Talos 200 X operated at 200 kV.

Results and discussions
A typical TEM image of the original rods is presented in Fig. 1(a) with a selected area diffrac on (SAD) pa�ern as an insert. The typical microstructure is characterized by well-defined βphase grains of about 1 µm diameter. The grains are not perfect crystalline with defects as shown by the contrasts in the image. There are also smaller Mo-depleted grains (called Mo-depleted par cles in the following) forming along the normal β grain boundaries. An example of the par cles is shown in Fig. 1 (b) with a corresponding EDX mapping of Mo element (Fig. 1 (c)). The SAD pa�ern in Fig. 1 (a) revealed the existence of ω-phase in a β-phase matrix. There were no other phases iden fied by electron diffrac on. The matrix and the Mo-depleted par cles are likely β and β‫,׳‬ which are formed through a bimodal chemical decomposi on of the matrix β-phase [26][27][28]. The forma on of Moenriched β-phase and Mo-depleted β‫-׳‬phase were reported in β tanium. There are no informa on about the differences in mechanical proper es of the two β phases. Therefore, their effects on the follow on cold processing will not be discussed separately.

Fig. 1, (a) TEM bright field image of the original rods (Insert is a β -113 zone SAD pa�ern showing both β and ω phases), (b) Mo-depleted par cles in originall rods and (c) EDX elemental mapping of Mo
Severe plas c deforma on, cold drawing in this study, of the rods resulted in a microstructure that is shown in Figs. 2 (a)-(c). At a lower magnifica on, sharp contrasts appear due to highly localized strains, a common cold drawn feature. Traces of material flow can be seen resul ng in crystal imperfec ons and therefore high contrasts. The traces are parallel to the longitudinal direc on of the wire, i.e. the drawing direc on. Images that were taken perpendicular to the wire show irregular shape contours of severely deformed regions. The deforma on in cold drawing is not uniform at a micrometer level.
A high magnifica on image in Fig. 2 (c) revealed the existence of platelets of nanometers thickness parallel to the drawing direc on. The platelets are a few hundred nanometers in length and about 50 nanometers in width. Further analysis revealed the modula on nature of the platelets, i.e. there is not a clear crystal orienta on change across the platelet boundaries. This is different from the reported deforma on twins by others, where further deforma on is localized in small size twins [28]. The strengthening effects of nano twins have been well documented a�ributed to their effects on the emission and deflec on of disloca ons and back-stress thus generated [39]. In comparison, the platelets of less than 100 nm thickness in Fig. 2 (c) result in the aforemen oned special proper es of the β Ti-15Mo alloy. Deforma on heterogeneity is also clear from the image where a large size plate exists most likely due to its higher hardness than its surrounding β-phase grains, which could be an evidence of different hardness of the β and β‫׳‬ phases. Furthermore, no dynamic recrystalliza on happened in the cold severe plas c deforma on. The strengthening is therefore the synerge c effects of grain refinement and stress-induced phase transforma on from β to ω-phase.

Fig. 2, (a) TEM image of the wire viewed perpendicular to the wire and (b) parallel to the wire and (c) a magnified image showing platelets perpendicular to the wires, arrows indica ng platelet boundaries
High resolution TEM images of the severely deformed Ti-15Mo alloy wires revealed a different mode of phase transformation from β to ω under a high strain. From Fig. 3, it is clear that strain induced β to ω phase transformation happened in a planar manor, i.e. through atomic shear of the {2 1 1} planes, resulting in a layered transformation from β to ω in comparison with the athermal ω-phase in the un-deformed rods, which is generally no more than 20 nm in diameter. The transformation still belongs to the diffusion-less athermal phase transformation because no diffusion happens during the phase transformation. However, the additional driving force for the phase transformation is provided as external mechanical energy. The transformation therefore has a strong mark of mechanical deformation of the parental β-phase, i.e. slip on specific planes along specific directions related to the drawing direction. The slip and rotation of the original crystal lattice resulted in the strong planar feature of the ω-phase thus formed. Furthermore, unlike the diffusion controlled isothermal ω-phase, the strain-induced ω-phase does not necessarily have a coherent interface with the matrix β-phase lattice. The high mechanical energy of severe deformation effectively nucleates the ωphase particles through a cumulative mechanism in a layer-by-layer manner at an atomic level. As a result, ω-phase plates as thin as three atomic layers with a variable size can form in the β-phase matrix, which makes the observation of ω-phase particles in the β-phase matrix difficult. Meanwhile, isothermal ω-phase particles are generally easy to find due to its larger particle sizes [40]. Additionally, due to the near zero stacking fault energy, the transformation can stop or proceed to any intermediate stage between β-phase and α-phase. The low stacking fault energy of β-Ti alloys enables atomic shear in the {2 1 1} planes. Fig. 2 (c) showing atomic shear on the

{2 1 1} planes as indicated by arrows and the irregularity of the platelet boundaries (the region between the arrows), (b) an ω par cle formed as a result of accumula ve shear and (c) a different region of ω phase par cle with crystal imperfec on
Despite no other phases were observed, homogeniza on was done at 900 o C followed by different quenching rates as described in the Materials and Experiments sec on to understand the forma on of athermal ω-phase. Homogeniza on at 900 o C for 2 hours has eliminated the Modepleted par cles completely, resul ng in a microstructure in Fig. 4 regardless of the cooling rates. Fig. 4 (a) shows modula on and Fig. 4 (b) is a SAD pa�ern showing no extra crystal features from the modula on. The dark field image in Fig. 4 (c) shows ω-phase par cles from the circled ω-phase spots in Fig. 4 (b). These athermal ω-phase par cles are the evidence of high stability of the ω-phase in the β-matrix, which has been reported. However, the nature of the modula on is s ll not clear. It has been reported as twins, composi on varia on of the β-phase and ar facts, etc. [26,41] . From the SAD pa�ern in Fig. 4 (b), it is evidence that there are no other phases than β-and ω-phases. The uniform dispersion of ω-phase par cles in the β-matrix is clear. These par cles are athermal ω-phase. Both homogeniza on at 740 o C and 900 o C followed by quenching to room temperature have resulted in the forma on of athermal ω-phase par cles in a β-matrix regardless of the cooling rates.
The ω-phase par cles are dispersed uniformly in the β-matrix, evidencing a homogeneous nuclea on of the ω-phase and a thermodynamic driving force for the β to ω phase transforma on. However, the transforma on proceeded to a certain level before stopping, sugges ng a limita on for the thermodynamic driving force, i.e. addi onal driving force is required for further progress toward ω phase and even more stable α phase from the parental β phase. As reported by others, coarsening of the ω-phase par cles can happen in follow on isothermal heat treatment, evidence of the requirement of addi onal driving force for further β-to ω-phase transforma on. The effects of cold working is therefore interes ng to explore further.
Two kinds of ω-phases have been reported as isothermal and athermal according to the nature of the phase transforma on. Isothermal ω-phase is the product of isothermal annealing to allow sufficient diffusion of alloying elements to form a hexagonal ω-phase structure from a BCC crystal structure of the β-phase. Meanwhile, athermal ω-phase forms as a result of martensi c phase transforma on from β-phase, i.e. displaced shuffle, in which two neighboring {2 1 1} atom layers form a middle layer for each three layers [30]. The ω-phase in the rods is a result of athermal quenching from the β-phase at high temperature, as shown in Fig. 4 (a)-(c). The ω-phase exists as par cles dispersed uniformly in the β matrix. A dis nc ve feature a�er severe cold deforma on is slip bands at an atomic level as shown in Fig. 4 (a) and (b). The bands are, in fact, the results of the shuffling of {2 1 1} planes of the BCC β-phase to transform to ω phase. The phase transforma on is resulted from shearing on the {2 1 1} planes due mainly to the near zero stacking fault energy of the {2 1 1} planes. Therefore, the movement of the {2 1 1} planes is a free mo on without considering the restraints of the neighboring grains. As a result, a high strain is accumulated in the grain boundaries resul ng in the phase transforma on further towards stable α phase.

Conclusions
β to ω phase transforma on in a Ti15Mo alloy is characterized by an athermal transforma on in heat treatment, which is followed by an atomic strain induced transforma on in cold working. The transforma on is accomplished through the accumula ve shear of {2 1 1} planes. The β to ω phase transforma on is not a single step process directly from β to ω but experiencing a con nuous rota on of the {2 1 1} planes. The strain induced phase transforma on through shuffling of crystal planes of zero stacking fault energy inevitably results in the presence of intermediate crystal structures between the β and ω phases.