The influence of the Nb:Ta ratio on the microstructural evolution in refractory metal superalloy systems

Gas turbines are an essential part of modern infrastructure, finding extensive use in the land based power generation and air transportation sectors. While there is global ambition to move to renewable energy sources, reducing the emissions associated with current technologies remains critically important. The efficiency of a gas turbine can be increased by operating at higher temperatures. However, the ability to increase this temperature is limited by the capabilities of the materials used in the turbine section. The currently employed materials, Ni-based superalloys, have been developed for ∼70 years and are approaching their operational limit.^1,2 Consequently, there remains a clear need to identify new materials that can enable higher temperature operation. Refractory metal superalloys (RSAs) are a recently developed class of materials that aim to fulfill this need by combining the high melting temperatures of the refractory metal elements with microstructures that exploit similar features to Ni-based superalloys.³ Both RSAs and Ni-based superalloys have fine-scale two phase microstructures comprising a cubic solid solution phase and an ordered superlattice phase. The key differences between these classes of materials are the crystal structures of their constituent phases and their configuration, with RSAs typically comprising bcc solid solution precipitates within an ordered B2 matrix.^4–6 

The exact pathway by which this fine-scale microstructure forms is still under debate, but it is commonly accepted that it involves a combination of a spinodal decomposition and an ordering transition.^3,7,8 A spinodal decomposition occurs during cooling and involves the separation of a homogeneous solid solution into two compositionally distinct phases, both of which have the same crystal structure as the parent phase (e.g., bcc→bcc₁ + bcc₂). This process can only occur within the central region of a miscibility gap, where the second derivative of the Gibbs energy curve is negative. Critically, within this region, there is no barrier to nucleation and the composition of each phase progressively evolves toward the equilibrium concentrations. Consequently, spinodal processes produce modulated microstructures that comprise fine-scale interpenetrating networks of the two product phases. Recent work in the Al–Ta–Ti–Zr system has shown that the morphology of the microstructure is similar, irrespective of whether the phases formed are bcc + bcc or bcc + B2.⁸ This observation suggests that the spinodal decomposition occurs prior to any ordering transformation, but unambiguous confirmation of this will require in situ data acquired during cooling.

Phase separation in these systems is thought to be a consequence of the positive enthalpy of mixing that exists between Zr and some of the refractory metal elements. Miscibility gaps exist in the Nb–Zr and Ta–Zr constituent binary systems as well as in the Mo–Ti–Zr ternary system. However, the compositional ranges and solvus temperatures over which these two phase regions extend vary quite dramatically between the different systems. For example, the peak solvus temperature in the Ta–Zr system is ∼1780 °C, whereas it is only ∼980 °C in the Nb–Zr system.^9–11 As such, the relative concentration of these elements is likely to have a large effect on the microstructural formation in RSAs.

Materials in commercial high temperature applications need to fulfill a demanding set of property requirements, necessitating careful control of multiple microstructural features. For example, the strength will be affected by the length scale, composition, and structure of the constituent phases, all of which will be influenced by the alloying species and concentrations.¹² However, key mechanical property targets will be defined at elevated temperatures, and so the shape of the miscibility gap, in addition to the solvus temperature, will also be critically important. Studies of Ni- and Co-based superalloys have shown that the crystallographic compatibility, or misfit, between the solid solution and superlattice phases also has a significant influence on in-service behavior.^13–15 Equally, it is critically important that the microstructure is stable under service conditions, both to the formation of competing intermetallic phases¹⁶ and to morphological changes through processes such as coarsening.¹⁷ Finally, these materials must also possess a high level of intrinsic resistance to environmental degradation.¹⁸ This list of examples is non-exhaustive but serves to illustrate the complexity of developing new high temperature materials, particularly as the desired operating conditions are continually becoming more hostile.

Many of the aspects in the above list are already under investigation in RSA systems, for example, the environmental resistance and microstructural stability of specific alloys.^19–25 Nevertheless, there remains a critical need to improve the fundamental understanding of compositional modifications. Several studies have reported how the microstructure and properties of RSA systems can vary through relatively coarse changes in composition, for example, the inclusion or removal of a specific element or halving/doubling its concentration.^26–28 However, despite only a few elements having been included in RSAs to date, no studies have sought to systematically understand the influence that the ratio of different refractory metal elements may have. Elucidating these effects will not only aid the optimization of existing alloy systems but will also provide key data that can be used to enhance the fidelity of related thermodynamic databases, improving future alloy design processes.

Here, we investigate the influence of the Nb:Ta ratio on the microstructural evolution in a model Ti–Zr–Nb–Ta quaternary system, a constituent system of the exemplar RSA AlMo_0.5NbTa_0.5TiZr. This simplified quaternary system was chosen to avoid any complications associated with the formation of Al–Zr based intermetallic phases reported in AlMoNbTaTiZr based alloys,^22,23 which might mask the effects being studied. A systematic series of alloys with compositions of 35Ti-30Zr-xNb-(35-x)Ta, where x = 0, 8.75, 17.5, 26.25, and 35 (at. %) were fabricated from elemental metals with purity ≥99.99%. These alloys will be referred to as 0Nb, 9Nb, 17Nb, 26Nb, and 35Nb from herein. Ingots, ∼12 × 8 × 70 mm³, were produced by vacuum arc melting under an inert Ar atmosphere that had been gettered with Ti. Macroscopic homogeneity was enhanced by the inversion and remelting of each alloy eight times. The cast ingots were homogenized at 1200 °C for 100 h and water quenched. Equilibrium conditions were achieved through long duration exposures on ∼8 mm long transverse sections (∼12 × 8 × 8 mm³) at temperatures of 1100, 900, 700, and 500 °C for 1000 h followed by water quenching. For all thermal treatments, samples were wrapped in Ta foil and sealed within quartz tubes under vacuum. To avoid oxidation upon contact with water, the quartz tubes remained intact during quenching. With this process, the sample temperature fell below 400 °C within one minute. Microstructural assessment was performed on polished samples using back scattered electron (BSE) imaging in a Zeiss GeminiSEM 300 operated at 20 kV. Elemental distribution data were acquired using an Oxford Instruments X-Max^N 50 energy dispersive x-ray (EDX) detector in the same instrument. Crystallographic information was obtained via x-ray diffraction (XRD) using a Bruker D8 diffractometer with Ni-filtered Cu radiation between 10° and 130° 2θ. Diffraction peaks were individually fitted using a Gaussian function, and lattice parameters were calculated using a nonlinear least squares refinement that accounted for the uncertainty in the position of each fitted reflection. Hardness measurements for each sample were obtained by averaging at least eight independent data points acquired using a Vickers indenter with a load of 5 kg. Thermodynamic modeling of phase equilibria was conducted in the ThermoCalc software using the TCTi2 database.

The bulk composition of each alloy was established from the solution heat treated condition material through large area (∼200 × 150 μm²) EDX analyses. The collated results are given in Table I, and the concentration of each constituent element was found to be in very good agreement with the nominal values.

Backscattered electron imaging of the alloys in the solution heat treated condition, Fig. 1, revealed a microstructure comprising bright and dark contrast nanoscale features in 0Nb, 9Nb, and 17Nb. The refined scale of this microstructure suggested that it had formed during cooling from the solution heat treatment temperature, and the morphology was very similar to those formed through a spinodal decomposition in similar systems.^29,30 Consequently, it was expected that the bright contrast features would be enriched in refractory metal elements, while the dark contrast features would be enriched in Zr.^8,29,30 The length scale of these features decreased as the Nb content, and thus Nb:Ta ratio, of the alloy increased, with no discernible features in either 26Nb or 35Nb. This suggests that the features are either absent, meaning 26Nb and 35Nb exist as single phase solid solutions at 1200 °C, or that the scale of the microstructure is beyond the resolution limit of the microscope (∼10 nm).

To clarify the phases present in these microstructures, x-ray diffraction data were obtained. As with the micrographs, a systematic trend was evident within these data, Fig. 2. The patterns obtained from 0Nb, 9Nb, and 17Nb contained two sets of reflections, corresponding to two phases with bcc crystal structures with slightly different lattice parameters. The two sets of reflections are most evident in 0Nb and become progressively less pronounced as the Nb content increases, consistent with the microstructures in Fig. 1 and the assertion that they formed through spinodal decomposition. The lattice parameters obtained from each phase are given in Table II.

In contrast, the diffraction patterns from 26Nb and 35Nb contained reflections from only one bcc phase with lattice parameters intermediate between the two observed in the other three alloys. Consequently, these data suggest that 26Nb and 35Nb were simply single solid solution phases. This postulate is also supported by the Vickers hardness data, which showed a bimodal distribution across the alloy series. The lower Nb content alloys with spinodal microstructures returned large Vickers hardness values, 440–465 HV. In contrast, 25Nb and 35Nb, which appeared to be single solid solution phases, exhibited much lower hardness values, 260–275 HV. As such, it appears that substitution of Nb for Ta in this series of alloys was sufficient to suppress the spinodal decomposition.

The equilibrium condition for each alloy was established by 1000 h exposures at temperatures of 1100, 900, 700, and 500 °C, with BSE images of the resulting microstructures shown in Fig. 3. Following exposure at 1100 °C, coarse bright contrast precipitates were observed in 0Nb, indicating that the solvus temperature for this alloy was above the heat treatment temperature. The coarser size of these precipitates, when compared to the spinodal structures, allowed their compositions to be assessed through SEM based EDX analyses (for example, Fig. S1). The bright contrast precipitates in 0Nb were found to be Ta-rich with a composition of 32Ti-16Zr-52Ta. In contrast, all of the other alloys exhibited uniform microstructures or fine scale features consistent with spinodal decomposition of this bcc phase upon cooling, akin to the features present in the solutioned condition. As such, it was concluded that these alloys remained in a single-phase field at the exposure temperature. Interestingly, fine scale spinodal features were also observed within the gray matrix phase (36Ti-36Zr-28Ta) of 0Nb. After exposure at 900 °C, coarse bright contrast precipitates were observed in both 0Nb and 9Nb, while the other alloys remained as solid solution phases. Elemental partitioning maps showed preferential segregation of the refractory metal elements to the bright contrast phase, which had compositions of 28Ti-5Zr-67Ta and 30Ti-9Zr-12Nb-49Ta, respectively. Spinodal structures were observed not only throughout 17Nb but also in the Zr-rich matrix phases of 0Nb and 9Nb, which had compositions of 40Ti-46Zr-14Ta and 37Ti-42Zr-7Nb-14Ta, respectively. A similar trend was observed following heat treatment at the lower temperatures with coarse bright contrast Ta-rich precipitates being observed in both 17Nb (34Ti-10Zr-24Nb-32Ta, 700 °C) and 26Nb (34Ti-13Zr-37Nb-17Ta, 700 °C), although only in the vicinity of existing grain boundaries for 26Nb at 700 °C. Interestingly, 35Nb remained as a single phase at all temperatures, indicating a high level of stability for this solid solution phase. X-ray diffraction data from the exposed samples, see Figs. S2–S5, indicated that at temperatures ≥700 °C all of the constituent phases were bcc. In contrast, the data for 0Nb-26Nb at 500 °C contained additional reflections indicative of a hexagonal phase. Evidence of an additional further Zr-enriched phase was found at the parent grain boundaries, which would be consistent with this observation. The hexagonal phase may also be present within the grain interiors of the alloys exposed at 500 °C, but the fine scale nature of these features is beyond the resolution of SEM based compositional analysis techniques.

These micrographs highlight one of the key effects of the Nb:Ta ratio within this series of alloys, which is the increased stability of the parent solid solution phase as the Nb content rises. The equilibrium microstructures show that the solvus temperature of the miscibility gap decreases by ∼20–25 °C per at. % of Nb that replaces Ta. Such a sharp depression of the solvus will likely mean that the Nb content would need to be tightly controlled should these alloys be produced commercially, as variations between or within different ingots could result in significantly different behavior. In addition, if designing a two-phase alloy for high temperature service then the concentration of Nb will need to be limited in order to keep the solvus above the operating temperature. Consequently, it is critical that such trends are accurately captured in the thermodynamic databases that underpin most modern approaches to alloy design.

A predicted isopleth corresponding to the studied alloy series, calculated using the CALPHAD method with the ThermoCalc TCTi2 database, is shown in Fig. 4. Overlaid are markers indicating the experimentally observed phase equilibria. Encouragingly, the prediction captures both the presence of the miscibility gap and the influence of Nb on the solvus temperature reasonably well. However, a slight discrepancy was identified for 35Nb, which was experimentally found to be stable as a single solid solution phase at all temperatures above 500 °C, while the thermodynamic modeling prediction suggested that the original bcc phase should have undergone a phase separation. Given the relatively low homologous temperature, it is possible that phase separation may be kinetically inhibited in this alloy. However, this would be inconsistent with recent assessment of diffusivity in the Ti–Nb, Ti–Ta, and Ti–Nb–Ta systems^31–33 and the response of the other alloys in the studied series, all of which decomposed despite being at even lower homologous temperatures. As such, the stability of the single solid solution phase in 35Nb is considered a true reflection, and therefore, the thermodynamic descriptions for the Nb–Ti–Zr ternary system require refinement.

The data also elucidates the influence of the Nb:Ta ratio on the occurrence of a spinodal decomposition. Spinodal microstructures were only observed in 0Nb, 9Nb, and 17Nb when quenching from 1200 °C, Fig. 1. Suppression of a spinodal decomposition can result from kinetic effects, significant lattice strain, or from changes in the shape of the parent phase's Gibbs energy curve. As mentioned above, kinetic based suppression seems unlikely given the greater diffusivity of Nb in this system when compared to Ta.³³ Local strain is well known to reduce the extent of a spinodal region, but this also seems unlikely given the extremely similar sizes of Nb and Ta.^37,38 Indeed, it is noted from the x-ray diffraction data in Table II that the alloys richer in Ta showed a larger difference in lattice parameters despite showing a greater prevalence for spinodal decomposition than alloys with higher Nb contents. As such, the stability of the parent phase at elevated Nb contents is most likely to arise from a modification of the Gibbs energy curve, altering its curvature such that the inflection points within the miscibility gap move toward each other. This hypothesis would also rationalize the decomposition observed in some of the Zr-rich phases shown in Fig. 3. Spinodal features were evident in the matrix phase of 0Nb at 1100 and 900 °C, as well as in 9Nb at 900 °C. However, no evidence of spinodal decomposition was observed in any of the Zr-rich matrix phases following exposure at temperatures of 700 °C. Consequently, this suggests that the composition of the matrix phases at 700 °C were outside of the spinodal region. Considering the compositions of these phases, it is notable that the Ta content of the Zr-rich matrix phases were markedly lower (≤11 at. %) than those at higher temperatures (≥14 at. %). Such values would also be consistent with the observations in Fig. 1, where 17Nb (17Ta) underwent spinodal decomposition, while 26Nb (9Ta) did not. Defining the shape of a multi-dimensional spinodal surface is complex and will also be influenced by other alloying additions. However, given the importance of this process in the formation of RSA microstructures, it is essential to gain an appreciation as to how different elements affect it for future alloy design.

In summary, the influence of the Nb:Ta ratio in a series of Ti–Zr–Nb–Ta alloys has been studied. Increasing the Nb content stabilized the parent solid solution phase, depressed the onset of phase separation, and above a certain concentration, suppressed spinodal decomposition. These trends were well described by current thermodynamic databases. However, for alloys within the miscibility gap, elevated Nb levels reduced the inter-phase misfit, which may be beneficial for some aspects of high temperature performance. Further studies on higher order RSA systems will be needed to confirm the fidelity of this trend, but this work demonstrates why careful balancing between these effects may be required when designing and optimizing future RSA.