Creep Analysis and Microstructural Evaluation of a Novel Additively Manufactured Nickel-Base Superalloy (ABD^®-900AM)

Additive manufacturing (AM) technology offers unique opportunities to improve efficiencies of gas turbines through utilization of novel designs. Current application of AM in large stationary power generation gas turbines is generally restricted to fuel delivery and combustion hardware, which are generally static component subjected to relatively modest conditions of temperature and stress. Increasing the temperature or improving the performance of hot section components, such as blades and vanes, provides the highest impact to increasing gas turbine efficiency and reducing overall carbon emissions. Recent studies have shown that stationary guide vanes can be redesigned with more detailed internal cooling features to reduce cooling air consumption and improve power output [1–4].

Hot section components often utilize nickel-base superalloys due to their exposure at extremely high temperatures and stresses. These alloys are specifically designed to have superior creep properties, and their microstructure can be complex. Typical materials used in turbine vanes include IN939, Udimet500, and IN738. These alloys are often produced using casting processes and are generally considered difficult to weld. Processing and weldability challenges stem in part from the rapid precipitation of the Ni₃(Al, Ti) gamma prime phase, which is the primary strengthening phase for this alloy class. Secondary strengthening mechanisms occur from solid solution strengthening, and the precipitation of other carbides, such as M₂₃C₆ and MX, mostly occur at dendrite and grain boundaries during heat treatment and high-temperature exposure. Since laser-based powder bed fusion (LBPBF) processing shares similarities with traditional welding, it is important to understand the challenges with welding this alloy class. The high-volume fraction of gamma prime phase (∼30 to 50%) in these alloys makes them susceptible to hot cracking and postweld heat treatment (PWHT) cracking, also known as strain age cracking, is a general concern [5]. Subsequent welding thermal cycles cause the strengthening phases (gamma prime and secondary carbides) to dissolve in the austenite matrix and grain growth occurs. The welding process may induce residual stresses. Thus, a PWHT is generally required to relax residual stresses, and a subsequent aging heat treatment is used to precipitate new gamma prime to restore strength. The overall PWHT process can also localize strains at grain boundaries, which may lead to strain age cracking depending on the competing mechanisms of relaxation of residual stresses and hardening due to precipitation of gamma prime.

Research on LBPBF of this alloy class based on traditional casting compositions has shown that these alloys can be successfully printed crack free and fully dense [2,6–10], but limited studies have shown significant debits in high temperature creep strength [4,11,12]. This study focuses on the evaluation of a novel nickel-base superalloy (ABD900-AM). An alloy-by-design process was used for the development of this material and chemistry was optimized to improve printability while maintaining superior high temperature strength [13,14]. The material was printed using an LBPBF process and was subjected to a subsolvus solution anneal and multistep aging heat treatment to produce a bi-modal distribution with ∼35% volume fraction of gamma prime without postprocessing hot isostatic pressing (HIP). A detailed metallurgical evaluation was carried out in the as-built and heat-treated structure to evaluate build density, grain size/structure, and carbide structure/morphology. A high temperature creep study was carried out to determine activation energies, stress exponents, rupture strength, and deformation resistance. These properties were compared to traditionally cast IN939 and IN738 materials. Post-test specimens were evaluated to determine location and features associated with creep damage. Several different creep mechanisms and comparisons between AM and traditionally cast materials are also discussed.

Small build blocks were manufactured for metallurgical evaluations, and larger round bars (84 mm length × 14.5 mm Ø) were printed for mechanical testing. A multistep heat treatment was carried out on several different build blocks and all the mechanical test bars. The first step was a solution anneal close to the gamma prime solvus temperature at 1060 °C for 2 h, followed by an air cool. The second step was at 850 °C for 4 h, followed by an air cool to allow for growth of the primary gamma prime phase. The last step was at 760 °C for 16 h, followed by another air cool to allow for secondary precipitation of gamma prime phase. This multistep solution and ageing heat treatment is commonly used for nickel-base superalloy castings [5]. Heat treatment optimization was not a major focus for this study.

Chemical composition of the material was measured using inductively coupled plasma optical emission spectrometry (ICP-OES) for Al, B, Co, Cr, Mo, Nb, Si, Ta, Ti, and W. Combustion methods were used for C and S. Inert gas fusion was used for O and N.

Samples from the as-built (AB) condition and at each heat treatment (HT) step were sectioned in both the transverse and longitudinal directions relative to the build direction. Creep rupture samples were also sectioned parallel to the gauge to evaluate microstructural features after testing. All samples were mounted using a hot isostatic pressing process with conductive bakelite. Optical microscopy was carried out using the Keyence VR-3200 white LED microscope to evaluate creep damage in post-test specimens. Scanning electron microscopy (SEM) was conducted using the FEI Tenero SEM to evaluate various microstructural features in both the as-built, heat-treated and post-test creep specimens. SEM electron backscattered diffraction (EBSD) techniques were used to determine differences in the grain structure due to build orientation and heat treatment condition.

High-temperature mechanical creep testing was conducted in the heat-treated samples in accordance with ASTM E139 [15]. Testing was carried out at temperatures ranging from 700 to 900 °C and stresses from 100 to 600 MPa for time up to ∼2000 h. Samples were machined with an overall length of 82.6 mm, gauge length of 31.75 mm, thread diameter of 12.7 mm, and gauge diameter of 6.35 mm. Temperature was kept constant to within ±0.1 °C.

Thermodynamic simulations were carried out using thermo-calc 2021a with the Ni-alloys package. The TCNI9 database was used for calculating the equilibrium mol fraction phase diagram and Scheil solidification models. Gamma prime nucleation, growth, and coarsening kinetics were evaluated using the TC-PRISMA package with the TCNI9 and MOBNI5 databases. Calculations at various temperatures from 700 to 900 °C were used to develop a Lifshitz–Slyozov–Wagner (LSW) coarsening model [16] for ABD900-AM.

Scanning electron microscopy images were captured in both the transverse and longitudinal directions relative to the build at several different length scales. The as-built (AB) microstructure showed fine grain structure with epitaxial grain growth parallel to the build direction. Fine substructure with circular/hexagonal subgrains ∼600 nm in diameter was noted transversely to the build direction. The heat-treated (HT) structure only showed subtle differences in microstructure in the SEM images. The MC precipitates in the as-built structure and heat-treated structure were ∼20 to 80 nm and were evenly dispersed throughout the matrix. The gamma prime precipitates in the as-built structure were very small (∼<5 nm) and were not quantitatively measured due to resolution limits in the SEM. In the heat-treated structure, gamma prime phase was in a bimodal distribution with primary gamma prime precipitates of ∼100 to 240 nm in size and secondary gamma prime precipitates of ∼20 to 40 nm in size. The total volume fraction of gamma prime phase was ∼35%. SEM images showing comparisons of the as-built and heat-treated structures in the transverse direction and longitudinal directions are shown in Figs. 2 and 3, respectively.

Electron backscattered diffraction data were collected for both the as-built and heat-treated microstructures in both orientations. Overall grain structure showed similar features, but the heat-treated sample exhibited less substructure due to the high temperature solution HT at 1060 °C. Pole figures also confirmed epitaxial growth and preference to the [001] build direction in both conditions. Figure 4 shows the inverse pole figure maps and pole figures for both the as-built and heat-treated microstructures in the transverse orientation. Figure 5 shows EBSD data in the longitudinal orientation.

Figure 7 shows the gamma prime coarsening rate as a function of time for various temperatures using thermo-calc simulations. Actual gamma prime precipitates were measured in the head (threaded) sections for creep samples exposed to 850 °C and 900 °C ranging from approximately 100 to 3000 h. Results show reasonable correlation to the LSW thermo-calc model. It should also be noted that the model predicts gamma prime coarsening only for primary precipitates (based on starting size of ∼170 nm). A comparison of gamma prime structure for each measured sample is shown in Fig. 8. Long-term exposure at these temperatures also changes the shape of primary precipitates from spherical to cuboidal and then appears, at the longest times, to show the start of rafting.

Chemical composition was measured to determine concentration of specified alloying elements as well as trace and tramp elements such as S, O, and N. A comparison of measured composition to nominal is shown in Table 1.

High temperature creep rupture testing was carried out for the fully heat-treated structures to study the resultant creep properties. Figure 9 shows the test results for reduction of area (RA) versus rupture time. The lowest ductility tests were also shorter term at high stresses and lower temperatures (more in Discussion section). The rupture time, minimum creep rate (MCR), % elongation, and % RA were reported for each test.

Test conditions were selected to allow for calculation of activation energy for creep and stress exponents. Tests at an isostress of 150 MPa were conducted at three different test temperatures of 850, 875, and 900 °C. The activation energy for creep was calculated to be ∼590,000 J/mol. Tests at an isothermal condition of 900 °C were conducted at three different stress levels of 100, 150, and 200 MPa. Some differences were noted between stress exponents for the two different build directions, but general results suggest a lower stress exponent (∼2.5 to 4) at testing stresses <150 MPa and a higher stress exponent (∼6 to 8) at testing stresses >150 MPa. Figure 10 shows plots for the isostress test conditions (MCR versus 1/T) and the isothermal conditions (MCR versus σ) in the natural logarithmic form.

The normalized strain rates were plotted against stress to better define stress exponents using all the creep data and compared to literature data on IN738 [17], shown in Fig. 11. The lower stress regions <150 MPa showed a stress exponent of ∼2.5 to 4 compared to a stress exponent of ∼4.5 to 6 for stresses in the 200 to 600 MPa range for ABD900-AM material. This is in comparison to stress exponents of ∼4.7 in the 100 to 200 MPa range and ∼10.1 in the 200 to 600 MPa range for IN738 material. Limited strain rate data exist in literature for other comparable alloys, such as IN939 or Udimet 500. However, it is expected that stress exponents in these alloys are comparable to IN738.

Here, $σR$ represents the normalizing stress value for rupture behavior and B is a constant. This final equation allows creep rupture to be predicted at various temperature and stress conditions without any additional parameters.

The calculated creep rupture parameters for ABD900 were compared to traditional superalloys IN939 and IN738, as shown in Table 2. The creep deformation (Eq. (10)) and rupture (Eq. (12)) correlations utilize standard parameters calculated from uniaxial creep data. The deformation constant (A) and rupture constant (B) were determining with a best fit approach using the built in excel solver. Figure 13 shows a comparison of creep rupture behavior for the different orientations tested in ABD900 compared to IN738. Figure 14 shows comparisons for deformation behavior in each material. There are clearly differences in both rupture and deformation behavior in the AM built ABD900 samples. These differences also suggest there may be different underlying creep mechanisms dependent on material processing including both orientation (transverse versus longitudinal) and fabrication method (AM versus casting).

The fracture surfaces in each creep rupture sample were evaluated after failure. Images for samples tested in both orientations are shown in Fig. 15 at various stress/temperature conditions. The transverse orientation failures at high stresses showed very low ductility. The highest stress test (600 MPa) failed in the threads of the sample. The stress axis was normal to the grain boundaries due to the sample orientation, and the low ductility at high stresses suggests there may be some “notch weakening” effects. Oxidation at one side of the fracture face also suggest crack initiation with preference to an edge and eventual failure occurring due to fracture with little oxidation (bluish tint) at the opposite side in high stress tests for both orientations. The lowest stress tests in the transverse direction exhibited larger ductility and movement of grains leading to ovality at the fracture face and across the gauge length. Observation of the fracture faces agree with creep rate and rupture data suggesting a change in creep failure mechanisms may be occurring with dependence on orientation, microstructure, and test conditions.

Creep rupture samples were also sectioned using electric discharge machining, mounted and polished to evaluate creep damage. Light optical images for both the specimens in the transverse and longitudinal direction are shown in Figs. 16 and 17, respectively. Creep damage was observed in all samples across the entire gauge length. The highest distributions of damage were in the lowest stress tests with damage concentrated at grain boundaries.

The initial microstructural results show that a higher strength nickel-base superalloy, ABD900-AM, can be printed both crack free and fully dense using standard machine printing parameters without the need for a postprocessing HIP step. This study continues to support the original alloy-by-design approach for printability and high-temperature strength [13,14]. Chemical composition was optimized to control key elements, such as Al and Ti to maximize gamma prime volume fractions but also stay in a range of weldability. The gamma prime in ABD900 was ∼35% volume fraction compared to 38% and 50% for IN939 and In738, respectively. There were also changes in refractory elements (increase of Nb, Mo, and W) that may impact solid solution strengthening and diffusion.

ABD900-AM in the as-built condition exhibits texture with epitaxial grain growth in the direction parallel (longitudinal) to the build. The grain structure in the heat-treated condition had comparable texture and grain size, but substructure was not present. The resulting microstructure had a clear influence on the overall creep behavior with the transverse samples having lower rupture lives (∼4–5×) compared to the longitudinal samples. It is also evident that creep damage is forming at higher misorientation grain boundaries in both orientations. Figure 18 shows EBSD data for a transversely tested specimen where damage was associated with grain boundaries, which were also perpendicular to the applied stress axis.

The stress exponents (n-values) measured were in the 2.5–4 range for the lower stress test data (<150 MPa) which is clearly below traditional power-law (climb/glide) values, which are expected to be around n = 5 [19]. Lower stress exponents may indicate diffusion-controlled mechanisms such as coble creep, but this study did not focus on evidence of diffusional flow of atoms along grain boundaries. Additionally, carbides at the grain boundaries are likely to influence diffusional creep behavior, so bulk or boundary diffusion is further complicated. The most plausible explanation based on the current evaluation is that grain boundary sliding (GBS) is occurring in this lower/intermediate stress range. This creep process was originally described by Langdon [21] who suggested that GBS takes place by a movement of dislocations along the grain boundary through a combination of both climb and glide. One characteristic of this mechanism is that the macroscopic strain due to GBS should be proportional to grain size, with increasing rate at smaller grain sizes. Evidence of this was observed in the fracture surface images and general observation of the gauge section, where there was ovality (see Fig. 14) which would require movement along grain boundaries. In longitudinal tested samples at lower stress, grains also showed deformation parallel to the build direction and strain was observed in the micrographs.

The higher stress test data (>150 MPa) had stress exponents in the 4.5 to 6 range. This suggests a dislocation glide or climb process is controlling creep deformation in this region. It is also important to note that the produced microstructure does not have MC carbides at the grain boundaries in comparison to IN738 or IN939 castings. The lack of carbides at the grain boundaries may prevent pinning of dislocation motion or GBS. This is one possible explanation for lower creep rupture lives and increased deformation rates at similar ε/D values (see Fig. 10). Furthermore, the traditional casting shows very high n-values at high stresses. Changes in n-values could be a consequence of microstructural differences in castings compared to AM printed material, such as grain size, and carbide distribution at boundaries.

Measured chemical composition was well within range for the nominal composition of all alloying elements. The oxygen was measured as 102 ppm, nitrogen to 83 ppm and sulfur to 20 ppm. These minor changes may have some influence on creep properties. Studies in Udimet 500 showed a factor of 2 in creep life when oxygen was reduced from 200 to <20 ppm [5], albeit at lower rupture lives (∼100 versus 200 h).

An additional factor is grain size is much smaller in the as built microstructure compared to similar castings. In the heat-treated condition, grain size of ABD900 was ∼30 μm in comparison in traditional castings ranging from ∼200 to 1000 μm grain sizes. Higher dislocation densities could also be present because the solution anneal temperature (∼1060 °C) was lower than typical practice of nickel-base superalloys in this class (>1200 °C), but TEM investigations would need to estimate such values. It should be noted that the gamma prime volume fraction was ∼35% in ABD900, but there were some denuded regions adjacent to grain boundaries where creep damage was occurring (see Fig. 19). Finally, grain orientation and relationships between high/low misorientations should be better characterized to understand how this may have an impact on both creep deformation and the location of damage.

They hyperbolic sinh creep rupture model was in good agreement with the generated data for ABD900. Creep rupture and minimum creep rate showed good correlation via a MG relationship. The two high stress tests in the transverse specimens were not used in MG relationship due to failures outside the gauge section (grip and shoulder). Although ductility in these samples were high (typically >10%), the creep deformation rates were increased by ∼2 orders of magnitude. This has significant consequences if AM is to be adopted for gas turbine components operating near the creep regime. Often these components are based on a design criterion of time to 1% creep. Thus, it is important to understand the underlying deformation rate processes in the AM built superalloys and not just rupture behavior.

Additional creep rupture studies are being carried out using a different heat treatment condition. This recent heat treatment was carried using a solution anneal temperature of ∼1240 °C, well above the gamma prime solvus temperature. Initial results have shown that the higher temperature heat treatment led to recrystallization and resulted in larger grain size. Early studies have also shown improved creep rupture lives and behavior more closely related to traditional cast superalloys. This study is being conducted to evaluate the mechanical properties of two different heat treatments of the same material with variation in grain size, texture, carbide structure, and orientation.

In this study, a novel nickel-base superalloy (ABD900) with ∼35% gamma prime volume fraction was printed crack free and fully dense using an LBPBF process. Microstructural characterization showed texture with epitaxial grain growth parallel to the build direction and very fine MC carbides dispersed in the matrix. A high temperature creep study was conducted in the heat-treated material to determine activation energies and stress exponents and a model was developed to estimate rupture strength and creep deformation. Evaluations of post-test specimens showed creep damage dominant at grain boundaries and changes in underlying creep mechanisms at low versus high stresses. Additional work is being conducted to evaluate creep properties in a higher solution anneal heat treatment (∼1240 °C) that exhibited recrystallization and larger grain size.

AM =

additive manufacturing

D =

diffusion coefficient (m²/s)

dε/dt =

creep rate (1/h)

EBSD =

electron backscattered diffraction

GBS =

grain boundary sliding

K =

coarsening rate (mm³/s)

LBPBF =

laser-based powder bed fusion

MCR =

minimum creep rate (1/h)

MG =

Monkman–Grant

PWHT =

postweld heat treatment

Q =

activation energy (J/mol)

SEM =

scanning electron microscope

γ′ =

gamma prime phase