1. Introduction
Ti alloys have been utilized for aerospace applications such as compressor disks and blades due to their lightweight and excellent specific strength, fracture resistance, and creep resistance up to 600°C.1) The Ti alloy products have been primarily fabricated by the forging and machined to form desirable microstructure and shape. However, the forging process caused a local inhomogeneity in the microstructure,2) and the machining of Ti alloys has been intricate due to their low thermal conductivity.3)
Then, recently, the additive manufacturing (AM) process has gained attention as an innovative processing method. The layer-by-layer melting by rapid heating and following rapid solidification enables the fabrication of complex three-dimensional structures while minimizing the material wastes.4,5) Numerous investigations of AM processing have been performed for the most widely used commercial Ti–6Al–4V, which consists of the hcp-α phase and bcc-β phase. It was revealed that a martensitic α phase was formed by rapid cooling above 410°C s−1 by phase transformation from the β phase, often forming in Ti alloys fabricated by AM.6) The martensitic α phase changes to a Widmanstätten microstructure by heat treatment due to coarsening of martensitic α phase.7)
While the effect of AM processing method on microstructure has been widely investigated, the creep behavior of AM Ti–6Al–4V have also been studied. Cardon et al. reported that the martensitic α structure transformed into the Widmanstätten structure during creep test at temperatures greater than 600°C.7) Later, Spigarelli revealed that the transition of deformation mechanisms was verified in both martensitic α phase and the Widmanstätten structures.8) They also pointed out that the difference in the morphology of the α phase did not change creep controlling mechanisms significantly.8) In both martensitic α and the Widmanstätten structures, the creep controlling mechanisms transferred from dislocation glide to diffusion creep as the testing temperature increased from 450 and 500°C to 650°C.9,10) Despite the similarity of their deformation mechanisms, the minimum creep strain was relatively lower in the Widmanstätten structure than in the martensitic structure.8)
We developed Ti–6Al–4Nb–4Zr (mass%) alloy that exhibited superior oxidation resistance at temperatures greater than 600°C by adding Nb and Zr to near-α alloys.11–15) Previously, we fabricated this alloy by PBF-LB process, one of the AM processing, and investigated the microstructure change by processing condition and heat treatment in the α + β two phase regime.16) The effect of these microstructures on creep behavior was also investigated. It was found that creep life depended on the melting pool or grain size, and longer life was obtained in the Widmanstätten structure formed in 300 µm grains. It was also found that creep life was elongated by HIP indicating elimination of micropores improves creep life. In order to further investigate the effect of heat treatment conditions on creep behavior, in this study, we fabricated this alloy by PBF-LB and applied various heat treatments with different heat treatment temperatures, durations, and cooling rates. The effect of different heat treatment conditions on the microstructure evolution of the martensitic α phase was investigated. The heat treatment at β regime was also performed to compare the obtained microstructure. The correlation between microstructures obtained by the different heat treatment conditions and creep rupture life was explored to determine the activated creep deformation mechanisms.
2. Experimental Procedure
Specimens with dimensions of 14 × 14 × 50 mm3 were fabricated by PBF-LB (EOSINT M290, EOS) with a Yb fiber laser (oscillation wavelength: 1030–1070 nm) using powdered Ti–6Al–4Nb–4Zr (mass%) alloy made by Taniobis with an average particle size of 30 µm. The scan condition is shown in Table 1. The laser power, the hatch spacing, and the layer thickness were fixed at 300 W, 0.1 mm, and 60 µm, respectively. Two different scanning speeds, 1000 mm·s−1 (Condition A) and 1400 mm·s−1 (Condition C) were used as we utilized in our previous study16). Samples A-1 to A-3 were fabricated by utilizing condition A while samples C-1 to C-6 were fabricated by utilizing condition C. The laser was scanned in a zigzag direction for each layer by rotating the direction 90° for all specimens in Ar gas atmosphere to prevent oxidation of the specimens. Sample C-1 to C-6 were used for the investigation of the microstructure evolution while Sample A-1 to A-3 were used for the creep tests.
Table 1 PBF-LB processing conditions.
Various heat treatment was performed for samples fabricated in condition C to investigate the effect of heat treatment temperature, time, and cooling rate. The heat treatment conditions are listed in Table 2. To investigate the effect of the heat treatment duration and cooling rate on microstructure, the heat treatment was performed at 970°C in the α + β two phase region for durations between 5 mins to 168 hours, followed by Air cooling (A. C.). In order to investigate the cooling effect on microstructure, A. C. and furnace cooling (F. C.) were applied after heat treatment at 970°C (Samples C-3, C-4). The effect of the heat treatment at β phase regime (1100°C) was also investigated to compare the microstructure to the samples heat treated at α + β regime (970°C). The selected heat treatment was performed for Samples A-1 to A-3 for creep tests to compare the effect of microstructure to creep properties.
Table 2 Heat treatment (HT) conditions.
Specimens with 7 × 7 × 2 mm3 were cut from the heat-treated samples for microstructure observation. These specimens were mounted and polished using SiC polishing paper and diamond pastes with particle sizes of 9, 6, and 1 µm, followed by SiO2 as the final polishing. The microstructures of the heat-treated samples were observed by field-emission scanning electron microscopy (FE-SEM, JEOL JSM-7200F). The phase composition of each phase was measured using energy-dispersive X-ray spectrometry (EDS) with an acceleration voltage of 20 kV.
Three specimens with gauge diameter and gauge length of 3 and 13.5 mm were prepared from Samples A-1 to A-3 for the tensile creep tests. Tensile creep tests were conducted at 600°C and under a loading stress of 137 MPa in air until fracture occurred. The elongation was measured using a linear gauge, and the testing temperature was measured using R-type thermocouples attached to the specimens. To investigate the transition of the deformation mechanism during the creep test, the step creep test was also conducted at 600°C. In the step creep test, the stress was first applied at 69 MPa until it reached steady-state creep. The applied stress was increased to 104 MPa and the creep test was conducted to reach the steady state. The applied stresses were then increased to 139, 194, 208, and 243 MPa. The strain rate and the applied stress were analyzed using an Arrhenius plot, and the activated deformation mechanism was estimated.
Microstructures of samples A-1 to A-3 were also investigated in SEM to compare the microstructures of samples C.
3. Results and Discussion
3.1 Microstructure evolution of α phase
3.1.1 Effect of heat treatment duration
Microstructure change for different heat treatment time at 970°C in the α and β phase region was investigated. Backscattered-electron images of the cross section parallel to the build direction of the sample C heat treated for 5 mins, 10 mins, 2 hours, and 168 hours are shown in Fig. 1. Fish-scale patterns of the prior-β grains with the grain size of 100 µm were observed in the low-magnification images (Figs. 1(a), (c), (e) and (g)). These unique morphologies were related to the melting pool morphology.16) In the corresponding high-magnification images (Figs. 1(b), (d), (f), and (h)), the Widmanstätten structure formed by coarsening of the martensitic structure was observed. The equiaxed α phase was not observed with the duration of 5 mins and 10 mins whereas they formed along the melting pool boundaries with a duration of 2 hours (Fig. 1(e)). The stress-induced equiaxed α phase with the size of 10 µm was observed. The equiaxed α phase formed because of the residual stress induced during PBF-LB process due to the high thermal gradient.16) After heat treatment for 168 hours, the volume fraction of the equiaxed α phase formed along the melting pool boundaries increased as well as their size exceeded 10 µm due to coarsening (Fig. 1(g) and (h)). The thickness of some α plate inside the melting pool also coarsened. The microstructure evolution was quite similar to those observed in Ti–6Al–4V.17)
The cooling effect on the Widmanstätten structure was also investigated (Fig. 2). The volume fraction of the α phase was investigated using Image J software. The volume fractions of the α phase were 76 and 86% in air-cooled (C-3) and furnace-cooled (C-4) samples, respectively. Most of the β phase remained in air-cooled sample because it partially transformed to α phase during air cooling. While the β phase perfectly transformed to α phase during furnace cooling, resulting in the remaining thin β.
The phase composition investigated by EDS is shown in Table 3. The composition of the equiaxed and plate α in the Widmanstätten structure was almost the same in the samples regardless of the cooling rates while the phase composition of the β phase was differed by furnace cooling. The Nb content was 3.1 at% in the air-cooled sample, while that is 6.2 at% in the furnace-cooled sample. Since Nb is a β stabilized element, Nb diffused from α phase and concentrated into the β phase during the transformation from β to α.
Table 3 Phase composition of sample C after heat treated at 970°C followed by air cooling (A. C.) and furnace cooling (F. C.).
The microstructure after the heat treatment at 1100°C is shown in Fig. 3. The α/β lamellar structure was observed in the grain size of approximately 100 µm. The α/β lamellar structure grew by formation of the α plate during cooling from the single β phase regime following the Burgers orientation relationship.18) When forged samples are heat treated in the β phase regime, drastic grain growth of the β phase occurs. For example, when the heat treatment of the forged sample at 1080°C for 2 hours, the grain size was approximately 550 µm.19) On the other hand, in PBF-LB sample, the melting pool boundaries were not affected by heat treatment at β phase regime, thereby the melting pool width was maintained at approximately 100 µm. During heat treatment, several segments formed inside the melting pool with 100 µm and the α plate formed in the segment. Each segment size was also approximately 100 µm. It indicated that the melting pool boundaries were varied by the fabrication process regime and did not move by heat treatment even in the β phase region. This is a large difference in microstructure in PBF-LB sample and the forged sample where grain growth of β phase was promoted.
Microstructures of Sample A for the creep test were also investigated. Figure 4 represents the cooling effect on the Widmanstätten structure. The β phase area of Sample A-1 decreased compared with Sample C-3 in Fig. 2(a). However, the volume fraction of the α phase was still higher, 90%, in a furnace-cooled sample (A-2) than in an air-cooled sample (A-1), 83%.
The phase composition investigated by EDS is shown in Table 4. The Nb content was 4.0 at% in the air-cooled sample while that is 7.1 at% in the furnace-cooled sample. High Nb concentration in the furnace-cooled samples was observed in Sample A similar to Sample C.
Table 4 Phase composition of sample A after heat treated at 970°C followed by air cooling (A. C.) and furnace cooling (F. C.).
The microstructure of Sample A-3 heat treated at 1100°C is shown in Fig. 5. The lamellar structure clearly appeared similar to Sample C (Fig. 3).
3.2 Creep properties
3.2.1 Effect of microstructure on creep
To investigate the effect of the cooling rate of the Widmanstätten structure on creep properties, creep tests were conducted for samples heat treated at 970°C cooled by air cooling (A-1) or furnace cooling (A-2) at a test temperature of 600°C under loading stress of 137 MPa. The creep curves of samples A-1 and A-2 are shown in Fig. 6(a). The creep test of the air-cooled sample (A-1) failed during the tertiary creep, and the fracture occurred without enough elongation. However, a small difference in creep strain was observed between the air-cooled (A-1) and furnace-cooled (A-2) samples. The minimum creep strain rate at the steady-state regime of the furnace-cooled sample (A-2) was slightly smaller than that of the air-cooled sample.
In Fig. 6(b), the creep curve of the sample heat-treated at 1100°C (A-3) during the creep test at 600°C under 137 MPa is shown together with creep curves of the PBF-LB (Samples B and D) and the forged Ti–6Al–4Nb–4Zr in our previous study.16) The creep life of Sample A-3 was close to Sample D-HIP.16) In the previous investigation,16) it was found that the creep life strongly depends on melting pool size or grain size;16) that is, the shortest creep life was obtained in the forged sample with bi-modal (10 µm), then, it increased as grain size increased as B-HT (100 µm), D-HT (300 µm). The longest creep life was obtained in the forged sample with a lamellar structure (550 µm). Here, HT indicates heat treatment at 950°C for 2 hours followed by water quenching. Creep deformation of near-α Ti alloys was dominated by the dislocation glide and grain boundary sliding;20) thereby the samples with larger grains generally resulted in prolonged creep life.
Creep response in primary and steady-state regimes of samples heat treated at 970°C (A-1 and A-2) were similar in that of B-HIP in Fig. 4(b). The thickness of the α and β phase of the furnace-cooled sample (Fig. 5) was similar to that of B-HIP, then similar creep life was obtained.
3.2.2 Deformation mechanism
To investigate the deformation mechanism at 600°C, step creep tests were performed for Sample A-1 and A-2 utilizing different cooling rates. The strain rate $\dot{\varepsilon }$ is defined as eq. (1) if it is assumed that the strain rate follows an Arrhenius-type equation.21)
| \begin{equation}
\dot{\varepsilon} = \dot{\varepsilon}_{0}\frac{G\varOmega}{k_{0}T}\left( \frac{b}{d_{g}} \right)^{p}\left( \frac{\sigma}{G} \right)^{n}\frac{D}{b^{2}}
\end{equation}
| (1) |
where
$\dot{\varepsilon }_{0}$,
G,
Ω,
k0,
T,
b,
dg, σ,
D,
p, and
n are the strain rate that does not depend on stress and temperature, the shear modulus, the atomic volume, the Boltzmann constant, the absolute temperature, the burgers vector, the grain size, the stress, the diffusion constant, the grain size exponent, and the stress exponent, respectively. In the present study, the testing temperature and the grain size (melting pool size) are constant, then the strain rate
$\dot{\varepsilon }$ is simply the function of the applied stress σ. Therefore,
eq. (1) is reintroduced as:
| \begin{equation}
\textit{ln}\,\dot{\varepsilon} = \textit{ln}\,A + n\,\textit{ln}\,\sigma
\end{equation}
| (2) |
| \begin{equation*}
\text{where}\quad A = \dot{\varepsilon}_{0}\frac{G\varOmega}{k_{0}T}\left( \frac{b}{d_{g}} \right)^{p}\frac{D}{b^{2}}
\end{equation*}
|
The stress exponent n was estimated as a slope using a double logarithm plot of
$\textit{ln}\,\dot{\varepsilon }$ and
ln σ.
The creep curves of the step creep test are shown in Fig. 7 and 8. The minimum creep rate was taken from the steady-state regime where the creep curve exhibited a distinct plateau. The double logarithm plot of $\textit{ln}\,\dot{\varepsilon }$ and ln σ was plotted as shown in Fig. 9 where σ is the applied stress and $\dot{\varepsilon }$ is the steady-state creep rate. In Samples A-1 and A-2, the values aligned along two segments depending on the different stress levels. The stress exponent rapidly increased at the applied stress of 150 MPa. The stress exponent in each segment was calculated and found to be 2.4 and 6.1, respectively. Although the utilized cooling rate was different after heat treatment, Samples A-1 and A-2 exhibited the same propensity in the change in the deformation mechanisms as the applied stress was increased. It is well known that dislocation slip is the dominant deformation mechanism when the stress exponent is above 3 while grain boundary sliding is the dominant deformation mechanism when the stress exponent is around 2.22) The present creep test condition at 600°C under 137 MPa is located in the area of grain boundary sliding regime, and our samples with larger melting pool sizes or grain sizes exhibited longer creep life. This confirmed that the dominant creep deformation mechanisms were grain boundary sliding. In the Samples A-1 and A-2, the melting pool with fish scale boundaries formed, and on these boundaries, the equiaxed α phase was found. They were also found in other melting pool boundaries. The size of the equiaxed α phase was approximately 10 µm. When grain sliding was the dominant deformation mechanism, the grain boundary sliding first occurred at the boundaries in the small equiaxed α phase, and this triggered the grain boundary sliding in the large grains and melting pool boundaries; thereby, the small equiaxed α phase will accelerate creep deformation. In Sample A-3 heat-treated at 1100°C, the equiaxed α phase did not form. The longer creep life of the A-3 sample compared with A-1 and A-2 may be realized by the absence of the equiaxed α phase.
Although the main deformation mechanism was determined as the grain boundary sliding, dislocation slip was also expected during creep deformation because the test condition was close to the point that the stress exponent changed.15,19) If the dislocation slip occurred, the microstructure in grains affected creep life. Comparing the microstructure of Samples A-1 and A-2 with different cooling rate, the thick α phase with high volume fraction and thin β phase with high Nb contents were obtained in the furnace cooled Sample A-2 as shown in Table 3. The α phase was hcp structure and the creep resistance was superior compared to bcc-β phase. Hence, in Sample A-2, deformation of β phase dominantly occurred during the creep test, but high Nb contents restricted the rapid increase of strain rate due to its lower diffusion rate in β-Ti.23,24) Thus, dislocation motion at α/β interface was effectively impeded, leading to the slightly longer creep life in the furnace cooled Sample A-2.
Regarding the creep deformation behavior of forged Ti–6Al–4Nb–4Zr with the equiaxed α phase, in our previous study, the stress exponent of creep deformation at 550°C in the equiaxed α structure was 5.9, suggesting that it was dislocation creep.15) We also calculated stress exponent at 600°C for the bi-modal structure and the lamellar structures with different grain sizes of 110 and 550 µm in Ti–7.5Al–4Nb–4Zr with slightly higher Al content.19) Their stress exponent was approximately 4 at stress higher than 150 MPa, but decreased to 3 at stress lower than 150 MPa. Our previous works indicated the deformation mechanism was almost equivalent regardless of the difference of processing technics such as PBF-LB or forging or microstructudifferences such as equiaxed, bi-modal, or lamellar structures.
Kim et al. investigated creep deformation mechanisms of commercial Ti–6Al–4V fabricated by PBF-LB.10) They found that the stress exponent of the sample with the martensitic structure was 3 at 500°C under stresses ranging from 80 to 140 MPa, indicating dislocation creep was the dominant deformation mechanism.10) In addition, Viespoli et al. exmined creep properties at 450, 550, and 650°C for PBF-LB samples of Ti–6Al–4V with the martensitic structure.9) The stress exponent ranged from 3.7 at 650°C to 5.6 at 450°C. They found that the stress exponent decreased with lower applied stress. It suggested that the deformation mechanism was also dislocation slip at lower temperatures and high applied stress, but as the testing temperatures increases with the decrease of applied stress, the effect of diffusion-mediated deformation also became significant. Our present study also showed that the stress exponent becomes smaller in the lower stress at 600°C, indicating other creep deformation mechanisms such as grain boundary slip or diffusion creep became dominant mechanisms rather than dislocation slip. Our findings in the creep deformation mechanism were consistent with other investigation.
The stress exponent of commercial forged Ti alloy in the temperature ranges from 400 to 600°C were generally 3 to 6, even though the alloy composition and microstructure difference, indicating that the deformation mechanism was dislocation creep.25–30) Since the microstructures of forged and PBF-LB samples was similar, there should be no process-dependent differences when the microstructure provided resistance to dislocation motion.
