2018 Volume 59 Issue 4 Pages 538-545
The multi-component alloy Nb-20Si-23Ti-6Al-3Cr-4Hf was produced by powder injection molding or hot isostatic pressing of pre-alloyed, gas-atomized powder. The resulting microstructure comprises the Nb solid solution as well as the α- and γ-modifications of Nb5Si3. Creep is evaluated in constant true stress tests at 1000 and 1100 ℃. The analysis of the creep behavior regarding its dependence on microstructural and testing parameters such as grain size, stress, and temperature reveals grain boundary sliding as the prevalent deformation mechanism. This is backed up by SEM/EBSD and TEM observations in the undeformed and deformed state. This creep mechanism was found to be a direct result of the small grain/phase sizes after powder metallurgical processing and led to a creep resistance even lower than that of a single-phase niobium-based alloy.
In high temperature applications such as gas turbines, today nickel-base superalloys are commonly employed to fulfill the requirements on high temperature creep strength as well as oxidation resistance and toughness. A further improvement in efficiency of those combustion engines is possible by increasing the turbine inlet temperature or by reducing the density1). Niobium-based silicide composites have been researched for some while now to deliver on both of these prospects. Alloys of this class of materials usually contain a niobium solid solution for ductility and toughness at lower temperatures, or more specifically at room temperature, and an Nb5Si3 intermetallic phase for high temperature strength. The density of those alloys is typically around 7 g·cm−3 which compares favorably with 9 g·cm−3 for state of the art nickel-base alloys. The binary system, however, while offering sufficient creep strength2,3) suffers from insufficient oxidation resistance4). Alloying elements such as titanium, aluminum, and chromium have been added to improve on that5–8); hafnium is also oftentimes added for solid solution hardening of the niobium phase7). A prominent example of this alloying scheme was the MASC alloy (metal and silicide composite, Nb-16Si-25Ti-2Al-2Cr-8Hf, compositions are given in atomic percent throughout this paper) introduced by Bewlay et al. in 19969). Those alloys, typically produced via induction skull melting (ISM), arc-melting, or directional solidification (DS), show a good balance of creep and oxidation properties. As segregations are frequently observed during cast metallurgy of this kind of materials6), castability is often poor, and since near net-shape component production is desirable, powder metallurgy (PM) offers the potential to target all three of these issues. Besides, PM also allows for composition variations that are inaccessible by cast metallurgy38). Therefore, in this work, a MASC-derived composition (Nb-20Si-23Ti-6Al-3Cr-4Hf) was assessed regarding its creep properties and underlying deformation mechanisms after being produced via advanced powder metallurgical techniques.
Rods of the nominal composition Nb-20Si-23Ti-6Al-3Cr-4Hf were produced by plasma-melting of pure elements. The rods were directly atomized by electrode induction-melting gas atomization (EIGA). The resulting powders were sieved into different size classes of <25, 25–45, 45–106, 106–225, and >225 µm10). In this work, it will only be reported on properties of the powder fraction <25 µm. Compaction of the powders was done via powder injection molding (PIM) or hot isostatic pressing (HIP).
For PIM, the powders were mixed with a polymer wax-based binder to produce a feedstock (powder load of 70 %) and subsequently injection molded into small cylinders. After solvent-based debinding with hexane and thermal debinding at 600 ℃, the samples were sintered at 1500 ℃ for 3 hours. For further details on PIM processing, the reader is referred to Ref. 10).
HIP was done at 1230 ℃ for 4 hours at a pressure of 150 MPa, after the powder was filled in a mild steel can on a vibrating table and evacuated over night before crimping and welding the can shut.
Additionally, monolithic intermetallic phases of α-Nb5Si3 and γ-Nb5Si3 were produced by traditional non-consumable tungsten electrode arc-melting with compositions of Nb-36.5Si-13Ti-1Al-4Hf and Nb-36Si-23.5Ti-2.5Al-0.5Cr-7.5Hf, respectively. These compositions had been determined by EDS on the separate phases in the compacted composite. To insure homogeneity, buttons were flipped and remelted at least 5 times.
Cuboid samples (4 × 4 × 8 mm) were extracted via electro discharge machining (EDM) from HIP cylinders and arc-melted buttons.
Heat treatments were performed in a Gero tube furnace under high purity argon flow at 1300 ℃ for 100 hours, or 1500 ℃ for 20 or 100 hours.
The resulting sample conditions will be denoted by the compaction process (HIP or PIM) and the heat treatment state (AC for as-consolidated or HT for heat treatment followed by the temperature in centigrade and the duration in hours).
Compressive creep testing was done in a Zwick Z100 electromechanical testing machine with attached Maytec vacuum radiation furnace. Load-deformation feedback control was set up to ensure constant true stress during deformation. Additionally, compression tests on select samples were performed at constant initial strain rates between 10−4 s−1 to 10−2 s−1.
Punch faces of compression samples were grinded plan parallel to a finish of grit P2500 and covered with a thin layer of h-BN to reduce friction between the SiC punches and the samples.
Microstructural analysis was done with a Zeiss Evo 50 scanning electron microscope (SEM) equipped with tungsten filament with energy dispersive X-ray spectrometer (EDS) and for EBSD mappings a Zeiss Auriga SEM with field emission gun was used. X-ray diffraction patterns in this work were recorded with a Bruker D2 Phaser diffractometer with Cu-Kα radiation (λ = 0.15406 nm) and an attached Lynxeye line detector at room temperature to identify the phases present.
In all powder metallurgical cases, the microstructure reveals up to four phases: niobium solid solution (Nbss), α-Nb5Si3, γ-Nb5Si3, and HfO2. They show up in bright grey, dark grey, medium gray and white contrast in SEM backscatter electron micrographs, respectively. One exception is the HT1300-100 condition, were γ-Nb5Si3 appears brighter than Nbss due to increased solubility of hafnium in this silicide at this temperature. Figure 2 shows the evolution of microstructure with increase in heat treatment duration and temperature. Exemplarily, the diffraction patterns recorded for HIP are given in Fig. 1, left.
XRD pattern for HIP AC and heat-treated conditions (left) and arc-melted monolithic intermetallics α-Nb5Si3 and γ-Nb5Si3 (right).
Backscatter electron micrographs (SEM) of AC and heat-treated conditions of HIP (top) and PIM (bottom) samples.
The arc-melted material is indeed monolithic intermetallic except for a small Nbss-peak detectable for γ-Nb5Si3 (Fig. 1 right).
While area fractions of the silicide phases significantly vary depending on heat treatment (α-Nb5Si3: 27–37 %; γ-Nb5Si3: 17–27 %), the Nbss fraction stays constant at around 46 % and HfO2 is below 1 %. The grain size of HIP and PIM was extracted from EBSD measurements for the HT1300-100 and HT1500-100, and HT1300-100, HT1500-20, and HT1500-100 conditions, respectively (Table 1). Note here that even after a 1500 ℃ heat treatment, coarsening of the microstructure is rather weak, giving grain/phase sizes of less than 10 µm on average. This indicates sluggish diffusion being a typical signature of RM-based silicide alloys11,12).
Usually, the steady-state strain rate $\dot{\varepsilon}_{\rm s}$ of a creep experiment in dependence of testing parameters stress $\sigma$ and temperature $T$, and the microstructural parameter grain size $d_{\rm g}$, is given by the following relationship13,14)
| \[ \dot{\varepsilon}_{\rm s} (\sigma, T, d_{\rm g}) = \frac{A \cdot D}{d_{\rm g}^{p}} \cdot \sigma^{n} \cdot \exp \left( \frac{-Q_{\rm c}}{RT} \right) \] | (1) |
However, we will start with the creep curve itself (Fig. 3), as it does not show the expected steady-state, i.e. constant, strain rate over a marked amount of plastic strain but rather a minimum creep rate $\dot{\varepsilon}_{\rm m}$. This can stem from microstructural changes very early during deformation as it is known e.g. for Ni-base superalloys14), leading to the commencing of the tertiary creep before a steady-state strain rate is reached: in the particular case of Ni-base superalloys the microstructural instability is often called “rafting”15). In the present case, the creep rate increase is not linked to microstructural changes as shown later, but rather to damaging. During deformation, phase boundaries fail and cracks propagate along them and into adjacent grains as shown in Fig. 3 (right). Equation (1) applies accordingly for the minimum strain rate $\dot{\varepsilon}_{\rm m}$.
Typical creep curve for the investigated material (left), TEM micrographs (scanning) showing phase boundary cracking after 15 % of plastic deformation (right).
Using isothermal creep tests at varying stresses, the stress exponent $n$ can be determined for each sample condition as the slope of a double-logarithmic plot of minimum strain rate over stress (Fig. 4 and Fig. 5). In Fig. 4, also the goal for creep resistance is given as dashed horizontal line (corresponding to <1 % deformation in 100–125 h) as proposed by several researchers familiar with the requirements for aircraft turbines5,6,16,17). However, this goal was postulated for temperatures of more than 1200 ℃ and stresses on the order of 150 MPa. For the present material, this goal is not met even for 1000 ℃ and 50 MPa. Another pre-requisite, a plastic deformation of less than 0.5 % during primary creep5) is not met either.
Plot of minimum strain rate over applied true stress at testing temperature of 1000 ℃ for material produced by HIP or PIM from gas-atomized powders in the AC and heat-treated conditions; the dashed line indicates the creep goal for 1200 ℃ and stresses of 150 MPa.
Plot of minimum strain rate over applied true stress at testing temperature of 1100 ℃ for material produced by HIP or PIM from gas-atomized powders in the AC and heat-treated conditions, respectively.
Independent of sample condition or testing temperature, a stress exponent $n$ of about 2 is apparent. This is in a range often observed when grain boundary sliding is the predominant creep mechanism, e.g. in fine-grained (<10 µm) multi-phase materials14).
3.4 Temperature dependenceBy changing the testing temperature at constant stress, the activation energy $Q_{\rm c}$ is determined as the slope of a $\ln (\dot{\varepsilon}_{\rm m})$ over $1/RT$ plot according to the transformation of eq. (1) for constant stress and constant grain size (Fig. 6).
| \[ \begin{split} & \ln (\dot{\varepsilon}_{\rm m}) = \ln (A_{2}) - Q_{\rm c} \cdot \frac{1}{RT}\\ & A_{2}:{\rm constant} \end{split} \] | (2) |
Temperature dependence of strain rate for select sample conditions to determine activation energy for creep.
During “pure” dislocation creep, the deformation is controlled by a dynamic equilibrium between the formation and annihilation of dislocations in the sub-grains, which limit their mean free path. Even in small grains, dislocation interaction is assumed to mainly take place in the three-dimensional dislocation network. Usually, the observed increase in strain rate for decreasing grain size is a result of grain boundary sliding becoming non-negligible anymore13).
Another mechanism that will lead to a grain size dependence of creep is the presence of diffusional creep. Nabarro and Herring18,19), and Coble20) proposed dislocation independent high temperature deformation based on diffusional flow of vacancies through the volume or along the grain boundaries, respectively. This would lead to a grain size exponent $p$ of 2 or 3, respectively, combined with a stress exponent $n$ close to unity (compare eq. (1)).
Similar to the so called Norton plots, Fig. 4 and Fig. 5, the logarithmic strain rate can be plotted over the grain size of different heat treatment conditions to obtain $p$ as the slope (Fig. 7). For different testing temperatures and stresses a mean value of the grain size exponent of 4.4 is observed.
Double logarithmic plot of creep rate over grain size for PIM ; grain sizes of 4.7, 6.0, and 8.9 µm correspond to the heat treatments HT1300-100, HT1500-20, and HT1500-100, respectively.
To fully rationalize the deformation mechanisms operating in these multicomponent multiphase materials, changes in microstructural features during deformation such as grain morphology and texture have to be analyzed as well. For that purpose, a heat-treated sample (PIM HT1300-100) was deformed in compression to a true plastic strain of $\varepsilon_{\rm t} = 1$ (1100 ℃, 50 MPa, corresponding to a height reduction by 63 %). The aspect ratio $S$ of the grains is in this case defined as the phase size perpendicular to the loading direction divided by the one parallel to it, as measured by the linear intersect method on optical micrographs. The increase in $S$ due to the deformation compared to the undeformed state is from unity to 1.23 and, thus, hardly significant (Table 2). This is also depicted in EBSD phase maps before and after the deformation (Fig. 8).
EBSD phase map of PIM HT1300-100 before (left) and after (right) deformation to $\varepsilon_{t} = 1$, Nbss in red, α-Nb5Si3 in blue, γ-Nb5Si3 in green; low angle grain boundaries (2–15°) in grey, high angle grain boundaries (>15°) in black; image width 80 µm, each.
For previously equiaxed grains the expected aspect ratio can be easily estimated for pure dislocation-based deformation (only change in grain shape without rearrangement of the grains). Under the boundary conditions of constant volume and isotropic strain perpendicular to the loading axis the resulting aspect ratio is given by eq. (3), where $\varepsilon_{\rm t}$ is the final true strain. In case of $\varepsilon_{\rm t} = 1$ a value for $S$ of 4.5 is expected.
| \[ S = \exp \left( - \frac{3}{2} \varepsilon_{\rm t} \right) \] | (3) |
At the same time, the bulk intermetallic phases (α-Nb5Si3 and γ-Nb5Si3 produced by arc melting) yield creep rates more than three orders of magnitude lower – at higher stresses – than the composite (Table 4). This is well in line with results obtained for binary and multi-component Nb-based silicides, where γ-Nb5Si3 demonstrated to have lower creep strength than α-Nb5Si325–28). This suggests that deformation of the intermetallic compound itself cannot be the rate-controlling factor in the composite's high creep rates.
As the grains do not deform, strain has to be obtained by rearrangement of grains, i.e. grain boundary sliding. Another feature of grain boundary sliding, the coalescence of phases during deformation29,30) is found as well (Fig. 8). The sliding will induce misfit stresses in the grain boundary triple junctions that have to be accommodated by dislocation or diffusional creep. When those accommodation processes cannot keep up, void formation and eventually cracking of the grain boundaries as shown earlier will occur. The activation energy for creep in this study was found to be 400 ± 26 kJ·mol−1, which reflects the activation energy for the rate controlling accommodation mechanism. This value is in good agreement with the activation energy for niobium self diffusion of 349–440 kJ·mol−131–34) and that for diffusion of titanium in niobium 364 kJ·mol−135). Diffusion in Nb5Si3, however, requires significantly lower activation energies (201–271 kJ·mol−136–38)). TEM investigations have shown the presence of dislocations in the Nbss after deformation (Fig. 9) and the lack thereof before it. Hence, most likely, dislocation movement in the niobium solid solution is the dominant accommodation mechanism. This explains also the observed coalescence of phases, as Nbss has to participate in the grain displacement with the silicides not being able to accommodate triple junction stresses at the given temperatures. Similar behavior was found by Jéhanno et al.11,39) in molybdenum-based silicide composites. In this case, the grain size exponent $p$ was close to unity, though, confirming accommodation by dislocations40,41). The grain size exponent found in the present work (4.4) is out of range even for grain boundary diffusion as main accommodation mechanism ($p \approx 3$), which can be excluded based on the activation energy.
TEM micrographs (dark field) of solid solution grains showing dislocations after 15 % of deformation.
Grain boundary sliding will become the dominant creep mechanism, when the stress dependent sub-grain size $d_{\rm sg}$ that forms during the primary stage of creep becomes larger than the grain size $d_{\rm g}$13) and the necessary dislocation interactions cannot take place anymore in the sub-grain boundaries.
| \[ \begin{split} & d_{\rm sg} \sim \frac{Gb}{\sigma}\\ & G:\ {\rm shear\ modulus}\\ & b:\ {\rm Burgers\ vector} \end{split} \] | (4) |
Creep data from Fig. 4 with constant strain rate results (dotted symbols) for a testing temperature of 1000 ℃.
While the grain size dependence does not completely disappear for high stresses where dislocation creep appears to be rate controlling ($n = 5$), it is much less pronounced than in the GBS regime. Even though constant strain rate tests were performed only on AC and HT1500-100 material, transition stresses can be estimated as 230, 190, and 145 MPa for HT1300-100, HT1500-20, and HT1500-100 heat treatment conditions, respectively. If plotted over the inverse grain size, the linear relationship seen in Fig. 11 is expected. For the above-mentioned application stress of 150 MPa some coarsening would be necessary to lie well within the dislocation creep regime. A further increase in temperature, e.g. to the postulated 1200 ℃ and above would require even more pronounced coarsening. Given that a heat treatment of 1500 ℃ for 100 h barely doubled the grain size and that coarsening is proportional to the square root of time, this would likely be not feasible.
Inverse grain size over transition stress for grain boundary sliding.
A multi-component Nb-Si composite (Nb-20Si-23Ti-6Al-3Cr-4Hf) has been produced by powder metallurgical means, i.e. powder injection molding or hot isostatic pressing of pre-alloyed gas-atomized powders. The processing led to a fine-grained microstructure that could be coarsened by heat treatments as high as 1500 ℃ to a grain/phase size of ≈10 µm.
In general, creep performance was not sufficient to be a viable alternative for current turbine solutions. The main reason being that the creep behavior is characterized by grain boundary sliding over the whole range of stresses and temperatures applied. Based on the activation energy of 400 kJ·mol−1 and the observation of dislocations inside the solid solution grains after high temperature deformation, deformation accommodation in grain boundary triple junctions by dislocations is likely to be dominant (compared to accommodation by diffusional creep). Only for very high strain rates (≥10−3 s−1) a transition to a dislocation-controlled creep deformation could be observed ($n = 5$). By determining a transition stress for different grain sizes, a minimum grain size could be deduced, where dislocation creep would be active at 1000 ℃ and in a stress regime which is closer to application ≈150 MPa. This critical grain size should exceed 20 µm, which is, however, not achievable economically by heat treatment coarsening of the present material.
This work was financed by the European Union within the HYSOP project (Framework Programme 7, grant no.: 266214). Particular thanks are dedicated to Dr. Nicolas Adkins and Mick Wickins from the University of Birmingham, and Dr. Thomas Hartwig und Marco Mulser from Fraunhofer Institute for Manufacturing Technology and Advanced Materials, Bremen, for providing the sample material.
We acknowledge the Karlsruhe Nano Micro Facility (KNMF, www.kit.edu/knmf) of the Forschungszentrum Karlsruhe for provision of access to instruments at their laboratories and we would like to thank Dr. Alexander Kauffmann for assistance in using the Laboratory for Microscopy and Spectroscopy.
