Study into the Role of Nickel Vapor on Surface Modification of a Third-Generation Single-Crystal Superalloy
A third-generation single-crystal (SX) superalloy was heat treated at 1350 °C for 24 hours while being exposed to nickel vapor. It is observed that a layer of 60-µm thickness deposited onto the alloy. This prevented the formation of TCP phases and the occurrence of melting, which were reported in previous studies. The interdiffusion between the alloy and deposited layer shows strong indications of cross-diffusion effects, as shown from the composition profile measurements and simulations.
Single-crystal (SX) nickel-based superalloys, due to their superior resistance in creep and fatigue properties, have undoubtedly contributed to the impressive evolution of energy efficient modern jet engines.[1, 2, 3, 4] Industrial processing involves casting followed by solution heat treatment, during which pronounced surface modification has been reported for the Re-rich third-generation Ni-based superalloys. It was suggested that this could be the result of sublimation—the direct transition from solid into the gas state—of alloying elements at high temperature and high vacuum. The idea that a Ni source can act as an inhibitor to the reaction of sublimation was first suggested by Wang et al. The current work aims to explore this idea further and study the effect on the microstructure beneath the surface of a third-generation SX superalloy after it has been exposed to Ni vapor at high temperature. Simulations are used to rationalize the observations.
2 Experimental and Simulation Methods
2.1 Heat Treatment Experiments
Composition of the Studied Alloy CMSX-10K (Weight Percent)
The cross section of the sample was then examined by cutting a section of 5-mm length followed by mounting, grinding, polishing, and fine polishing. The layer beneath the surface was studied using scanning electron microscopy (SEM) and backscattered electron imaging (BEI). Composition profiles were measured using energy-dispersive X-ray (EDX), starting from the surface and extending about 530 μm into the sample, with a point measurement every 5 μm. Additional measurements took place parallel to the surface at various locations. Finally, electron backscatter diffraction (EBSD) mapping was carried out. SEM, BEI, EDX, and EBSD were performed with the field emission gun–scanning electron microscope JEOL2 7000F.
2.2 Modeling the Interaction Between the Substrate and the Deposited Layer
DICTRA simulations were used to simulate the interaction between the deposited Ni layer and the alloy at 1350 °C. The domain was selected to be of 700 μm, and 700 grid points were entered linearly across the domain. The thermodynamic and kinetic database TTNI8 and MOBNI1 were used, respectively. A labyrinth factor was used to include the γ′ phase, as discussed in a previous work. At the chosen temperature, however, very little, if any, formation of γ′ phase is expected. As the deposition of Ni cannot be simulated in DICTRA, the experiment was approximated in two ways: by (a) single-layer and (b) multilayer simulations. In the first case, it is assumed that a layer of 60-µm thickness of pure Ni formed on the surface of the superalloy sample. The interdiffusion between the Ni layer of 60-μm thickness and a layer of CMSX-10K of 640-μm thickness was then simulated. The dimension of the Ni layer was chosen in line with the experimentally observed layer thickness. In the case of the multilayer approach, the simulation was refined by gradually building up the Ni layer during the simulation. It was assumed that 1 µm of Ni was deposited every 24 minutes, in total 60 layers within 24 hours of heat treatment. The DICTRA grid was, accordingly, extended by additional grid points containing pure Ni. This layer would then interdiffuse with the existing material before an extra layer would be added, and so on. Adding new grid points is not possible inside DICTRA, however. This limitation was overcome by stopping simulations, writing composition profiles to files, and altering the profiles as described previously, before restarting the simulation. A small, dedicated software tool was written for this purpose.
3 Results and Discussion
When compared to the measured composition profiles, the simulation results show higher compositions in the deposited layer for the elements Al, Cr, Co, and Ni, while Re, Ta, and W show lower ones. From the surface of the sample toward the bulk, the elements Cr, Co, Re, and W show a steady increase in composition, while Ni shows a drop. Al also shows a drop, but the simulations show a local minimum, while measured values stay below the nominal composition. Ta shows a decrease in the experiments but an increase in the simulations, with a local maximum. The measured Ta composition at the surface even lies above the nominal composition. The multilayer simulation approach gives a Ta composition closer to the measurement, except for the point 50 µm below the surface. Generally, both simulation approaches reproduce the trends seen in the measurement for all elements, except for Al and Ta, but the absolute compositions are not always correct. This clearly indicates that an important effect was not taken into account in the simulations. This will be discussed in detail in Section IV.
The results clearly show that Ni sublimated from the foil and deposited on the alloy. This can be explained by looking at the thermodynamics of the problem. The activity of Ni in the foil is equal to unity and considerably higher than that in the alloy. For the pure Ni to be in equilibrium with the gas phase, the Ni partial pressure in the gas has to be higher than it is in the gas over the alloy. Hence, Ni will sublimate from the Ni foil and deposit on the alloy, providing a source and a sink of Ni.
As the added Ni layer interdiffuses with the alloy, its composition evolves toward the alloy composition, while the alloy gets diluted by the additional Ni. In both cases, the refractory element content is more diluted, lowering the probability of forming TCP phases, while reducing the Al content mitigates the risk of melting. It is, thus, interesting to consider if this technique can be used to mitigate risks of surface melting and TCP formation during solution heat treatments. It is often observed that the outermost turbine blades in a heat treatment furnace load show the most severe damage. This indicates that sublimation from heat-treated parts creates a Ni-rich atmosphere that is almost enough to prevent damage. Adding a sacrificial Ni part, at the outer parts, has the potential to reduce or event prevent sublimation-induced damage. Furnace operators might be concerned about the additional Ni vapor, which will eventually deposit elsewhere in the furnace. This vapor is there even without sacrificial Ni, sublimating from heat-treated alloys and, hence, damaging them.
The composition profiles observed in the experiments look surprising at first glance, especially when compared to the simulation results. It is clear that the high temperature (1350 °C) sufficiently accelerates diffusion to enable noticeable composition change in the deposited layer. As the layer is relatively thin (60 µm), the entire deposited layer is affected by interdiffusion. Cross-diffusion effects are very noticeable in the composition profiles of most elements, either as local minima/maxima (Al, Ta, Cr) or as overshoots in composition (Re, W). These occur when the gradient in one element drives the diffusion in another and are quite common in multicomponent alloys, especially when the composition difference is rather large. In the case of Al, cross-diffusion is also the reason for the composition profile observed in both experiments and simulation. In both cases, high concentrations are observed on the surface of the sample. Al is driven toward the surface by gradients of other elements; it is, however, not possible to determine without further analysis which elements are the main driving forces.
Another important observation is that the measured composition of some elements at the surface is higher than calculated (Re, W, Ta, Ni), while for other elements, the reverse is true (Al, Co, Cr). The elements Al, Co, Cr, and Ni have significantly higher vapor pressures than the other elements. In a previous study, it was suggested that these elements sublime during heat treatments. The vapor above the sample is rich in Ni, due to the presence of a Ni source, but poor in Al, Cr, and Co. The latter elements, therefore, experience a driving force to sublimate and increase their partial pressure in the vapor. Consequently, the alloy will lose some of these elements. Even Ni will sublime from the alloy during the early stages of the experiment, while the Ni partial pressure builds up. In the case of Al, cross-diffusion drives Al to the surface, as gradients in composition of other elements drive Al diffusion. Al is the fastest diffusion element in the alloy and will, therefore, reach the surface first. This might also explain why the measured Al composition does not recover to the nominal Al composition, as sublimation started from the earliest stages of the experiment. As Al, Cr, and Co are lost, the relative compositions of Re, W, and Ta will increase, as they are left behind. This explains why the measured compositions of these elements are higher than the simulated ones. The difference between experimental and simulated results for Ta is quite significant, though. The phenomenon explained previously, plus the cross-diffusion effect seen in the simulation, might explain most of the difference, but it should also be considered that the thermodynamic and kinetic data are not as accurate as they should be. Further investigations are needed to clarify this. However, most of the difference between the experiment and simulation is due to the fact that the model assumes a closed system, while the alloy in the experiment is an open system with exchange of matter between gas and solid phases.
The assumptions made in the simulations with regard to the kinetics of Ni deposition are quite simple: all the Ni is deposited either instantly or at a constant deposition rate. The reality, of course, is more complex, as the deposition rate depends on the actual partial pressure of Ni in the gas phase and the actual activity of Ni on the surface of the alloy. As the Ni layers deposit and the alloying elements interdiffuse, the Ni activity changes and, hence, the deposition rate changes. The assumption of constant deposition, therefore, is not accurate, but it does, however, allow a qualitative estimate of the effect of noninstantaneous deposition, which is an increase of diffusion into the deposited layer, resulting in higher compositions of all alloying elements.
Finally, the rather large composition differences between the alloy and deposited layer can lead to nonbalanced diffusion fluxes. In the present case, more alloying elements diffuse into the deposited layer that Ni diffuses into the alloy, and the difference in volume shows up as porosity, commonly referred to as Kirkendall porosity. However, this does not explain the enrichment of some alloying elements around the pores. Further investigations are needed to confirm this in the present case.
Heat treatment of CMSX-10K in the presence of pure Ni was performed and DICTRA simulations were used to rationalize the observed composition profiles.
It was found that a considerable amount of Ni was deposited onto the CMSX-10K sample. The deposited material was polycrystalline and, interestingly, the composition was not pure Ni but contained considerable amounts of all alloying elements of CMSX-10K. Furthermore, no TCP phases or melting of the alloy were observed. The compositional change of the deposited layer was explained by the interdiffusion between the alloy and deposited layer and simultaneous sublimation of Al, Cr, and Co.
The results show that by controlling the gas atmosphere, the reaction between the gas/vapor and alloy can be reversed from the loss of Ni to the deposition of Ni. The observed structures are fundamentally different from structures seen without the presence of pure Ni. This also supports the sublimation hypothesis set up in previous publications.
The authors thank Dr. Bo Pang for helping with the EBSD mapping and Mr. Peter Cranmer for carrying out the heat treatment at the high vacuum furnace. SEM, EDX, and EBSD studies took place at the Electron Microscopy Group Facility, University of Birmingham. The financial support of this work was provided by Rolls-Royce plc and Innovate UK.
- 2.C.T. Sims: Superalloys II, C.T. Sims, S.N. Stoloff, and C.W. Hagel, eds., Wiley, New York, NY, 1987, pp. 3–26.Google Scholar
- 6.F. Cosentino, J.-C. Gebelin, N. Warnken, and R.C. Reed: 9th Int. Conf. on CFD in the Minerals and Process Industries, 2012, C.B. Solnordal, P. Liovic, G.W. Delaney, and P.J. Witt, eds., CSIRO, Melbourne, http://www.cfd.com.au/cfd_conf12/PDFs/031COS.pdf, accessed March 14, 2018.
- 7.N. D’Souza and H.B. Dong: Superalloys 2008, R.C. Reed, K.A. Green, P. Caron, T. Gabb, M.G. Fahrmann, E.S. Huron, and S.A. Woodard, eds., TMS, Warrendale, PA, 2008, pp. 261–69.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.