Near-Net-Shape Production of Hollow Titanium Alloy Components via Electrochemical Reduction of Metal Oxide Precursors in Molten Salts

The plaster mold was made in two segments, each of which has a hemi-spherical cavity in the center. A 16-mm-diameter spherical cavity can be formed inside the plaster mold when combining these two parts together. Two small holes, which served as the channels for slip injection and gas releasing, were drilled on the top part of the mold. The mold was sized so that it could be separated after the solid slip-casting process without damaging the fragile slip-cast metal oxide precursor. The slip was prepared by mixing the ball-milled metal oxide mixture with distilled water to a mass ratio of 1:1.5, e.g., 3 g of metal oxide mixture with 4.5 ml water. As the volume of the spherical cavity inside the mold was ca. 2 ml, while that of the prepared slip was ca. 4.5 ml, a syringe was used to slowly and continuously inject the slip into the mold, giving enough time for water absorption by the plaster. Inevitably during this process, some spillage of the slip was observed at the gas release hole, as well as minor leakage at the connection between the top and bottom parts of the mold. The mass of the final slip-cast metal oxide precursor was approximately 2 g. At the final stages of slip injection, the gas-releasing hole was sealed quickly by the solidifying slip due to the small quantity of slip permitting rapid water drainage. The syringe needle was kept in the injection hole during the process to create a hole on the cast, which would facilitate post-reduction washing. Figures 1(a) through (c) illustrates schematically this solid slip-casting method. Once the slip was injected into the mold, the water was removed by capillary action (Figure 1(a)), leaving a close-packed deposit of metal oxide particles against the inner wall. As this draining process proceeded continuously, a cavity was generated in the upper central part of the slip due to the influence of gravity (Figure 1(b)); this would result in an uneven wall thickness in the produced precursor. To insure uniform thickness, the plaster mold was rotated several times during the process (Figure 1(c)). Following casting, the plaster mold was dried in a fume cupboard for 48 h, after which the syringe needle was removed, the mold opened, and the precursor exposed as shown in Figure 1(d). The slip-cast metal oxide precursor was spherical in shape with a slightly reduced diameter of approximately 15.6 mm; this allowed it to be easily released from the plaster mold due to contraction during the drying process. The hollow structure of the precursor can be seen in Figure 1(e), where one such precursor was intentionally fractured.

Other precursor shapes were also prepared, e.g., a pellet and cup, for electro-deoxidation under the same conditions to further demonstrate the feasibility of the FFC-Cambridge Process for near-net-shape production of Ti-6Al-4V components. Figures 2(a) through (f) depicts the solid slip-casting process for producing metal oxide precursors in either of these shapes.

In the slip-casting process, different mass ratios of the metal oxide mixture to water were investigated, including 1:1, 1:1.25, 1:1.85, and 1:2.5. The ratio was found to be a crucial parameter for the slip-casting of cup-shaped precursors. The generated slips were poured into plaster molds (ca. 13 mm in diameter and 13 mm in depth) as shown in Figures 2(d) through (f). To examine the wall thickness of the cast piece, the precursors were first immersed in epoxy and after solidifying for 24 hours, the samples were polished using alumina paste to explore the cross section. Figure 3 presents cross sections of four slip-cast metal oxide precursors made from slips with different mass ratios of metal oxide mixture to water.

The cross-sectional images presented in Figure 3 indicate that the mass ratio of metal oxide mixture to water should be controlled in a specific range in order to obtain a uniform wall thickness across the part. Based on these observations, when the mass ratio was at or above 1:1.25, the slip-cast metal oxide precursor tended to form a tapered wall shape, lacking in thickness control (Figures 3(a) and (b)). This was thought to be due to the rapidity of casting resulting from the small quantity of water in the slip.[20] Conversely, when the mass ratio was slightly decreased to either 1:1.85 or 1:2.5, a relatively uniform wall thickness could be achieved (Figures 3(c) and (d)). It should be noted that if the prepared slip is too dilute, variations in thickness and density across the part may occur due to settling. Therefore, the ideal ratio for slip-casting of these parts was found to be in the range of 1:1.85 to 1:2.5. Additionally, the porosity of slip-cast metal oxide precursors was in the range of 55 to 60 pct, which would facilitate the movement of electrolyte in the precursor during electrolysis. Figure 4(a) shows a cup-shaped metal oxide precursor (ca. 1.5 g) which was cast from a slip with a mass ratio 1:1.85; the top surface of the slip-cast part was polished to make it flatter, facilitating better contact with the cathode current collector assembly. This precursor was 10.4 mm high (after polishing the top surface), with a 12.8 mm OD, and a wall thickness of ca. 1.72 mm. SEM analysis of this part (Figure 4(b)) shows that the ball-milled metal oxide mixture exhibited a sub-micron particle size uniformly distributed in the range of 0.2 to 0.7 μm.

The metal oxide precursors investigated in this study were relatively fragile and could not be attached to the cathode current collector using metal meshes and wires as reported before.[11,13] Placing the precursor at the bottom of the graphite or metal crucible used as the current collector[11,21] was also infeasible as the metal oxide precursor could float owing to the buoyancy provided by the hollow cavity. Therefore, a basket cathode design was developed to offer both mechanical protection and electric contact to the precursor.

The oxide precursor was placed between two molybdenum (Mo) meshes and contained inside a titanium (Ti) basket holder (0.5 mm thickness) which had several drilled holes to facilitate molten salt movement during electrolysis. Stainless steel wire was gently wrapped around the holder to secure the Mo mesh-sandwiched precursor. The cathode assembly was then attached to a threaded stainless steel rod using screw nuts as illustrated in Figure 5. Based on experimental observations, both the Mo meshes and Ti basket holder could be reused for more than 3 times with careful washing and polishing after each electrolysis experiment.

In the FFC-Cambridge Process, sintering of the green precursors is a standard pretreatment to ensure sufficient mechanical strength is attained for handling. However, the decreased porosity as a result of the sintering process impedes the electrolyte infiltration and movement in the pores, hindering the electro-deoxidation process. Although fugitive agents have sometimes been employed to generate high porosity in the sintered precursors,[14,22] the energy consumption of the sintering process remains a concern. Fortunately, by taking advantage of this cathode design, no additional sintering was required for successful electrolysis.

Electrolysis of the oxide precursor was performed at 3.2 V and 1173 K (900 °C) in molten CaCl₂ for about 24 hours under argon. Based on previous studies, these conditions are sufficient to achieve complete reduction of either pure TiO₂[22,23] or low-grade TiO₂ precursors[13] with similar feedstock masses. Considering voltage losses due to the resistance of electrodes, electrolyte, and other materials utilized in the cell (the “iR drop”) at 1173 K (900 °C), the 3.2 V applied cell voltage should have been below the decomposition voltage of CaCl₂ (i.e., 3.23 V), but still sufficient to drive the reduction of each metal oxide component in the precursor as shown in Table I.[24]

Using the cell voltage of TiO₂ to Ti₃O₅ as the reference, the potential for reducing the oxide to the respective product can be derived from Table 1. For example, the potentials are −0.011 V for V₂O₃ to VO and −0.155 V for Ti₃O₅ to Ti₂O₃ with reference to the reduction of TiO₂ to Ti₃O₅. These are of course thermodynamic predictions, but can help understand the cathode reactions that can be revealed by cyclic voltammetry. Figure 6(a) presents a typical CV of pure TiO₂ (solid line in Figure 6(a)), showing three main reduction peaks at −0.37 (C1), −0.52 V (C2), and −1.23 V (C3) vs the Mo wire pseudo-reference. C1 and C2 can be attributed to the reactions of TiO₂ → Ti₃O₅ (C1) → Ti₂O₃ (C2), as their potential difference of 0.150 V agrees with the value of 0.155 V in Table I. It is worth noting that, between TiO₂ and Ti₂O₃, there are many intermediates known as the Magneli phases, the stoichiometry of which varies according to the formula of Ti _n O_2n−1 (n ≥ 2).[11,25,26] Thus, Ti₃O₅ is only representative of the reduction of TiO₂ to the Magneli phases.

Peak C3 is negatively shifted from C1 by 0.86 V and may be linked to TiO → Ti which should occur at −0.896 V according to Table I. The small discrepancy, less than 0.04 V, could be related to the fact that the CV of TiO₂ in Figure 6(a) does not shown clearly the reduction of Ti₂O₃ to TiO. A possible cause is the disproportionation of Ti₂O₃ to TiO and TiO₂[25] (or one of the Magneli phases). More likely, it is the in situ formation of perovskites as exemplified by the following reactions associated with C1 and C2:[26‐28] where α is 1/3 for Ti₃O₅ and 1/2 for Ti₂O₃, and can be 0 for TiO₂ in Eqs. [2] and [3]. Clearly, Eq. [3] is a chemical reaction (i.e., it does not contribute to the current on the CV) and only occurs after Eq. [1] which provides O²⁻, while Ca²⁺ comes from the molten salt.[26,28] Further reduction of these perovskite phases may proceed directly to TiO[27] which can explain the absence of the Ti₂O₃ to TiO reduction on the CV in Figure 6(a).

Interestingly, when compared with C1 and C2, the reduction peaks C1′ and C2′ on the CV of the metal oxide mixture (dashed line in Figure 6(a), 88.7 wt pct TiO₂, 7.4 wt pct Al₂O₃, and 3.9 wt pct V₂O₃) shifted positively from −0.37 to −0.28 V and −0.52 to −0.48 V, respectively. However, C3 and C3′ still share almost the same potential (−1.23 V vs −1.24 V). These voltammetric changes can be regarded as the reduction of the oxide mixture being easier than that of pure TiO₂ and warrant an analysis on the causes.

According to Table I, the reduction of Al₂O₃ to Al alone occurs at a more negative potential than TiO to Ti. Also, V₂O₃ to VO occurs at almost the same potential as that for the reduction of TiO₂ to Ti₃O₅, and VO to V occurs at a potential close to that of Ti₂O₃ to TiO. Thus, the reduction of Al₂O₃ or V₂O₃ alone is not expected to contribute to the positive potential shift of C2 and C3 in Figure 6(a). Incidentally, positive shifts of C1 and C2 were also observed when CaTiO₃ was added to TiO₂ to prevent in situ formation of the perovskites as represented by Eqs. [2] and [3].[27] The consequence of Eqs. [2] and [3] is the inclusion of Ca²⁺ which causes volume expansion and hence partial or complete blockage of the pores in the cathode. This in turn slows down or even stops the removal and/or supply of ions (O²⁻ and Ca²⁺) via diffusion through the molten salt (electrolyte) in the pores, making the electro-reduction more difficult.

Obviously, the occurrence of Eqs. [2] and [3] requires the presence of “free” TiO₂ or one of the Magneli phases. In other words, if free TiO₂ may be partially or fully replaced in the precursor by “compounded” TiO₂ with another metal oxide, e.g., V₂O₃ or Al₂O₃, in situ formation of perovskites can be difficult or avoided, resulting in easier reduction. To confirm this hypothesis, XRD patterns of ball-milled metal oxide mixture before and after sintering at 1173 K (900 °C) in air are displayed in Figures 7(a) and (b), respectively. It can be seen that, before and after sintering, almost no change occurred in the Al₂O₃ peaks, but all the V₂O₃ peaks disappeared. This is evidence that Al₂O₃ is non-reactive toward TiO₂ and V₂O₃ under the applied sintering conditions. The absence of V₂O₃ peaks together with shifted rutile peaks observed in the post-sintering sample indicates the formation of metal co-oxides (or compounded oxide). The positions of shifted rutile peaks were found to be closer to that of TiVO₄, indicating the possibility of conversion of anatase TiO₂ and V₂O₃ into the pseudo-rutile phase, i.e., TiO₂/TiVO₄, at high temperatures.

Formation of TiVO₄ in the oxide precursor is expected as the result of oxidation of V(III) to V(IV) during sintering in air. This oxidation would have brought oxygen into the precursor and caused volume expansion as VO_1.5 (15.39 cm³/mol) is smaller than VO₂ (17.84 cm³/mol) in molar volume. However, the added amount of V₂O₃ (<4 wt pct) was small in the mixed oxide precursor, and the V(IV) formed was included in the lattice of the resulting pseudo-rutile phase of TiO₂/TiVO₄. Thus, the oxidation of V(III) to V(IV) should have an insignificant effect on the actual overall volume of the mixed oxide precursors, particularly considering the volume shrinkage due to sintering which was always observed in this work.

To further investigate this mechanism, a partially reduced sample was subjected to XRD analysis. It was found that instead of the formation of perovskite (i.e., CaTiO₃) which was found to be a reduction inhibitor,[14,27] peaks of the metallic pseudo-oxide Ti₂O and partially reduced metal co-oxide were observed (Figure 7(c)). Therefore, it can be concluded here that the formation of metal co-oxides or the pseudo-rutile phase (i.e., TiO₂/TiVO₄) during sintering of the oxide mixture could help diminish the formation of perovskite to some extent, bringing about easier electro-reduction.

In contrast to the positive shifts of C1 and C2, a slightly negative shift from −1.23 to −1.24 V was observed on C3. However, the shift is clearly smaller than the difference between reduction of TiO to Ti and Al₂O₃ to Al. It indicates that the Al₂O₃ reduction occurred at a more positive potential in the mixed oxide electrode than that in Table I. This is evidence that the energy released from alloy formation, which is almost wasted in the conventional melting method, was utilized to lower the polarization for reduction and save energy. Further, at 1123 K (850 °C) (for recording the CV), Al would be a liquid and, if formed, may flow away from the electrode. This obviously did not occur since analysis of the electrolytic alloys confirmed the designated Al concentration (see Section III–D). Thus, liquid Al was likely never formed, and the energy required to melt it for alloying is saved in electro-reduction.

A typical current–time curve recorded during constant-voltage electrolysis of the metal oxide precursor is presented in Figure 6(b). An initial current peak was observed (Figure 6(b), insert), in agreement with the three-phase interline (3PI) propagation mechanism featured in the FFC-Cambridge Process,[21] with increased conductivity among the outer surface of the metal oxide precursor.[11] However, when compared with previous observations, this initial current peak was less pronounced,[13,21] possibly arising from the large contact area of the Ti basket coupled with a higher extra background current. The remainder of the current–time curve in Figure 6(b) matches well with that obtained from the reduction of pure TiO₂ or different TiO₂ precursors.[13,18] This was expected as TiO₂ accounted for ca. 88.7 pct of the slip-cast metal oxide precursor.

An intriguing question may be asked about the declining nature of the current in Figure 6(b). Is it because the “contact area” between the oxide and the metal formed decreases with time as electro-reduction approaches the core of the oxide precursor? In other words, if the current is normalized against the “contact area,” could the “current density” remain constant? The answer is yes if diffusion is disregarded in the pores of the cathode according to the recently proposed three-phase interline (3PI) models.[21,29,30] However, oxide ion diffusion through the pores in the formed metal layer would eventually become the control step in determining the overall speed of the reduction process when the metal layer becomes thicker with electrolysis. This effect of diffusion will lead to a declining current density. Of course, there are also other factors affecting the overall electrolysis current, such as temperature, voltage, molten salt composition, porosity, and particle sizes in the oxide precursor, formation of intermediate phases in the cathode, and the electronic conductivity in each part of the cell.

Figure 8 shows the photos of electrolytic products obtained from the molten salt electrolysis. After the solidified salt was removed by washing with water, the electrolytic products exhibited a dull gray color as shown in Figures 8(a) and (d). As expected, Figures 8(b), (e), and (h) confirm the metallic luster on all the metallized samples after a minor polish. As shown in Figure 7(d), XRD analysis proved that the obtained products were the (α + β) phase alloy. It is worth mentioning here that both α and β phase peaks were slightly shifted which is common on commercial Ti-6Al-4V alloys[31] owing to the given positions of α and β phase being based on pure Ti rather than titanium alloys.

Actually, the observation of both the α and β phases in the produced samples is expected because the electrolysis temperature [1173 K (900 °C)] falls in the α-to-β phase equilibrium region of the Ti-6Al-4V alloy. Further increasing the temperature above the α-to-β phase transition temperature [ca. 1253 k (980 °C) for the Ti-6A1-4V alloy][3] would be interesting as this may allow quenching the product to possess a higher proportion of the β phase. Previous attempts to quench the Ti-6Al-4V alloy from 1323 K (1050 °C) resulted in the transformation to the martensite phase[3] which could be decomposed to separated α and β phases by slow annealing at lower temperatures.[32] A recent study showed that it was possible to produce the TiZr alloy in the metastable β phase or mixed α and β phases by electro-reduction in the absence of any added β stabiliser. This unusual finding was attributed to the effects of quenching, but more importantly an appropriate amount of oxygen remaining in the alloy might have functioned as an inhibitor to the phase transition at low temperatures.[33]

It can be seen in Figures 8(a) through (e) that all electrolytic products retained the original shape of their metal oxide precursors (Figures 1(d) and 4(a)), while some shrinkage happened. The diameter of the electrolytically reduced hollow sphere was about 40 pct less than its metal oxide precursor, i.e., reduced from ca. 15.6 mm to ca. 9.4 mm. The shrinkage of the electrolytic cup-shaped sample was approximately 23 pct in height and 13 pct in diameter. The shrinkage was mainly caused by the combined effects of solid-state sintering, change of density, and atomic rearrangement[22,34] which occurred during electrolysis at 1173 K (900 °C). The shrinkage rates of the hollow sphere and cup-shaped samples were relatively consistent. Therefore, a desired size of the electrolytic product could be obtained by giving the metal oxide precursor a predefined allowance. Moreover, when compared with cup-shaped samples, the hollow spheres exhibited more regular contour shrinkage. This phenomenon could be ascribed to both the spherical shape which evenly distributes stress to avoid cracking and the inner cavity which functions as a cushion to bear more inward shrinkage. Figure 8(c) presents an electrolytically reduced hollow sphere which was sectioned with uniform wall thickness of ca. 1 mm. The small hole generated during the slip-casting process permitted water access during post-electrolysis washing for salt removal. In addition, the electrolytic Ti-6Al-4V sample was broken by a pair of pliers to produce a fracture surface which was used to examine the interior structure under SEM. As shown in Figure 8(f), when compared with its metal oxide precursor (Figure 4(b)), the electrolytic Ti-6Al-4V exhibited a much less porous structure of heavily sintered nodular particles than in similarly prepared Ti metal.[11,35] The average porosity of the electrolytic products was in the range of 26 to 30 pct, implying an enhanced mechanical strength, which will be discussed below. Furthermore, by simply changing the plaster mold to a golf club head shape, a miniature hollow golf club head-shaped metal oxide precursor was fabricated (Figure 8(g)). After 24 h of electrolysis [3.2 V, 1173 K (900 °C)], a miniature hollow Ti-6Al-4V golf club head was obtained (Figures 8(h) and (i)) which matches very well to the shape of a golf club driver head (see insert in Figure 8(h)).

Elemental composition and homogeneity are essential for titanium alloy properties.[36] Typically, the oxygen content of the electrolytic products was approximately 1970 ppm as measured by the LECO RO-416DR, which was significantly lower than that of titanium components produced by powder metallurgy.[16] A lower oxygen content can improve ductility.

For analyses of the other elements, the electrolysis products were sectioned, polished, and investigated under BSE-SEM, while EDX maps were recorded as shown in Figure 9(a). The black dots observed in the BSE-SEM images were due to the voids in the electrolytic products. The distribution of the alloying elements, i.e., Al and V, was relatively uniform. A typical BSE-SEM image and the associated EDX spectrum of the electrolytic samples are shown in Figure 9(b), together with elemental composition, confirming the samples to be Ti-6Al-4V. Moreover, both the SEM image in Figure 8(f) and the BSE-SEM image in Figure 9(b) revealed grain sizes of the produced alloy samples to be in the range between a few and a few tens of micrometers, much larger than the sizes of the particles in the mixed oxide precursor. This is interesting because the removal of oxygen from the metal oxide particles should have produced smaller metal particles. However, at the electrolysis temperature of 1173 K (900 °C), sintering of small metal particles to a bigger grain size must have happened.

Nevertheless, when compared with that of a commercial Ti-6Al-4V sample (annealed, purchased from Rolls-Royce^®) as shown in Figure 9(c), a finer distribution of the α (dark) and β (bright) phases is present across the electrolytic product. This was thought to have arisen from the solid-state reduction of the ball-milled metal oxide mixture with fine particle sizes and well-proportioned distribution of alloying elements.

All as-produced electrolytic Ti-6Al-4V alloy samples were subjected to mechanical tests. A typical compressive stress–strain curve, which was obtained from a cup-shaped sample, is presented in Figure 6(c). It should be stated that the conditions, such as shape, dimension, and porosity, of the as-produced near-net-shape samples were not ideal for compression tests, which could have contributed to the noticeable fluctuations in Figure 6(c). However, there was always an approximate linear region on the stress–strain curve in the low stress region for all samples tested, which was used to derive the Young’s modulus. The observed maximum compressive stress was ca. 243 MPa and the corresponding modulus was ca. 14 GPa, which are relatively comparable with those of natural bones (2 to 200 MPa, and 0.1 to 20 GPa)[37] and porous Ti-6Al-4V alloy produced by powder metallurgical techniques (225 MPa, and 14 GPa with 50 pct porosity).[38] Therefore, it could be concluded that the electrolytic Ti-6Al-4V products exhibit enough strength to be directly used in some devices without further downstream processes such as smelting or pressing.

It is acknowledged that the samples produced in this work still contained too many voids as revealed by Figures 9(a) and (b) and obviously need further densification for high stress applications. While post-electrolysis improvement of the products may be achieved by high-pressure sintering (e.g., hot isotropic pressing), our follow-on research will focus on changing the electrolysis from one step to two steps.[14] The first step will be same as that for electro-reduction with the second step being a prolonged period of in situ sintering. In the second step, a lower cell voltage may be applied to prevent reoxidation of the product without too much energy input. The temperature of the cell may also be increased in the second step to promote sintering, although this may be challenging in large-scale operation. Last, but not the least, precursor fabrication should be further investigated in relation to shape and property control in the near-net-shape product. Nevertheless, highly dense oxide precursor is not recommended unless the product is, for example, a thin-walled vessel. This is because, for reduction of a bulky precursor (e.g.,>2 mm in the thinnest dimension), a sufficiently porous structure is needed to enable the removal of oxygen ions at an acceptable speed through the molten salt in the pores, instead of the solid phase.