1. Introduction
Improvements in the fuel efficiency of a jet engine are needed in the aircraft industry for environmental protection and to counter the rising prices of crude oil. In jet engines, Ti alloys are used as blades and disks in compressors, and Ni-based superalloys are used in turbines where the operation temperature is much higher than in the compressors. The temperature of combustion gas is increased to improve the fuel efficiency of a jet engine. As a result, the operation temperature of the compressor is also increased, and high-pressure compressors have started to replace Ti alloys with Ni-based superalloys. An improvement in the heat resistance of Ti alloys is necessary because their replacement with Ni-based superalloys causes an increase in the jet engine weight. However, it is known that the oxidation resistance and creep properties of Ti alloys drastically deteriorate at above 600°C.
In α (hcp) + β (bcc) and near α alloys, three types of microstructures are obtained by changing the thermomechanical processing conditions, that is, lamellar structures consisting of plate-like α and β, an equiaxed α structure, and so-called bi-modal structures containing an equiaxed primary α phase and α+β lamellar structure. The mechanical properties are significantly different depending on these microstructures even for the same alloy. A fully lamellar structure has a better creep resistance than the other two microstructures. By contrast, the equiaxed α structure has a better fatigue resistance, and the bi-modal structure has a superior balance of both characteristics.1)
For application to compressor blades and disks used in a jet engine, a balance among the low cycle fatigue (LCF), high cycle fatigue (HCF), and creep properties are required. TIMETAL 834 alloy is one of the materials with the highest operation temperatures used in a jet engine. Considering the balance of these characteristics, a bi-modal structure with a volume fraction of the equiaxed α adjusted to 15–20% is used in TIMETAL 834 alloy.1,2) However, the bi-modal structures of Ti-6242 and Ti–6Al–4V alloys are used by adjusting the volume fraction of the equiaxed α phase to 35–40%. The mechanical properties of these alloys can be improved if the volume fraction of the equiaxed α phase decreases to 15–20%, although control of the microstructure remains difficult owing to a small annealing temperature window.1) This indicates that an understanding of the effect of thermomechanical processing is extremely important to control the microstructure of Ti alloy.
In our previous study, we focused on Ti–Al–Nb alloys because we found that Sn deteriorates the oxidation resistance, whereas Nb improves the oxidation resistance.3) The phase equilibrium, microstructure, oxidation resistance, and mechanical properties of Ti–Al–Nb alloy were investigated.3–8) The oxidation resistance of Ti–Nb–Al improved with an increase in the amount of Nb, but became saturated at 2 at% Nb.4) It was also found that α2-Ti3Al of approximately 30 nm precipitates in Ti–15Al–2Nb (at%).5) Next, the effects of the precipitation hardening by the α2-Ti3Al were investigated.8) By precipitating the α2 phase in Ti–15Al–2Nb, precipitation hardening was clearly observed within the moderate temperature range of 300°C to 450°C, and the precipitation hardening became weaker at 600°C or higher. This indicates that a strong solid–solution hardening is necessary to improve the strength of Ti alloy.
Because a Ti–Al–Nb ternary alloy is expected to have a weak solid–solution hardening, the addition of Zr to Ti–Al–Nb was attempted, and the effects of Zr on the oxidation behavior and mechanical properties were investigated.5–7) Like Nb, Zr was also found to improve the oxidation resistance.5) As a result, the simultaneous addition of Zr and Nb also improves the oxidation resistance. The addition of 2 at% of Zr into Ti–10Al–2Nb (at%) also improves the compression strength through solid–solution strengthening.7,8) This means that Zr is an effective alloying element, improving both the oxidation resistance and the mechanical properties.
In addition, creep tests were conducted on Ti–10Al–2Nb–2Zr and Ti–10Al–2Nb–2Zr–0.5Si (at%) with an equiaxed α structure, and the deformation mechanism of the creep was investigated.7) The deformation mechanism of the equiaxed α structure was a high-temperature power-law creep controlled through lattice diffusion, and no difference was shown between the two alloys, although the creep life became longer through the addition of Si.
Because the processing condition was constant in the previous study, in the present study, a microstructure evolution through various processing and heat-treatment conditions was investigated in Ti–13Al–2Nb–2Zr (at%) (Ti–7.5Al–4Nb–4Zr, mass%) alloy. An Al amount of 13 at% was chosen to form the α2-Ti3Al in a future study, and not 15 at% because an alloy with 15Al is too brittle and difficult to process. Three processing temperatures, 900°C and 1000°C in the α+β phase field, and 1080°C in the β phase field, and two rolling reduction ratios, 93% and 67%, were selected. The influence of the microstructure formed through various processing conditions on the creep properties was also investigated.
3. Results and Discussion
3.1 Microstructure
3.1.1 Microstructure of as-processed samples
Back-scattered electron images of Ti–13Al–2Nb–2Zr alloys after processing to a 93% rolling reduction ratio at 900°C, 1000°C, and 1080°C and processed to a 67% rolling reduction ratio at 1000°C are shown in Figs. 1(a), (b), (d), and (c), respectively. The phase with the dark contrast is the α phase, while that with the bright contrast is the β phase. A typical bi-modal structure consisting of an equiaxed α phase and the α+β lamellar structure, which is clear in the enlarged image, was formed through processing at 1000°C, as shown in Figs. 1(b) and (c). The equiaxed α and β phases were considered to form during processing in the α+β phase, and the equiaxed α phase remained at room temperature; in addition, the acicular α phase precipitated from the β phase during cooling after processing. The bi-modal structure was formed in the sample processed at 1000°C with a rolling reduction ratio of 67%, but the equiaxed α phase was not perfectly developed, and the α phase grew with an ellipsoid shape, as shown in Fig. 1(c). The microstructure of the sample processed at 900°C to a 93% rolling reduction ratio consisted of α and β phases; however, the structure was not clear due to heavy deformation, as shown in Fig. 1(a). By contrast, in Fig. 1(d), the α+β lamellar structure was formed in which the acicular α phase grew from the β phase during cooling after processing in the β phase.
The grain orientation spread (GOS) maps obtained by the EBSD of Ti–13Al–2Nb–2Zr alloys after being processed to a 93% rolling reduction ratio at 900°C, 1000°C, and 1080°C and processed to a 67% rolling reduction ratio at 1000°C are shown in Figs. 2(a), (b), (d), and (c), respectively. The GOS maps in Fig. 2 show the average strain found in the grain. Grains with a large strain of over 5% were observed in the samples processed in the α+β phase field, as shown in Figs. 2(a)–(c). By contrast, as shown in Fig. 2(d), the strain of the sample processed in the β phase field was less than 5%. The strain introduced in the samples processed to a 93% rolling reduction ratio at 900°C and 1000°C showed no large difference, as indicated in Figs. 2(a) and (b), but it was found that the rolling reduction ratio affected the strain introduced in the grains when processing was carried out at the same temperature of 1000°C; that is, the strain in the sample processed to a 93% rolling reduction ratio was over 5%, larger than the approximate 5% strain in the sample processed to a 67% rolling reduction ratio, as shown in Figs. 2(b) and (c). Fine grains of a few micrometers in diameter were also observed along the grain boundaries of the prior grains in samples processed at 1000°C regardless of the rolling reduction ratio. The strains in these fine grains were less than 5%. It is thought that these fine grains were formed through dynamic recrystallization during the processing owing to their small strain.
Back-scattered electron images of Ti–13Al–2Nb–2Zr alloys heat-treated at 950°C in the α+β phase field for 3 h followed by furnace cooling after processing to a 93% rolling reduction ratio at (a) 900°C, (b) 1000°C, and (d) 1080°C, and (c) processed to a 67% rolling reduction ratio at 1000°C, are shown in Fig. 3. The dark contrast phase surrounded by a bright contrast phase was observed in the sample heat-treated after processing at 900°C and 1000°C, as shown in Figs. 3(a)–(c). The composition of the constituent phase of the equiaxed structure obtained by the EDS of Ti–13Al–2Nb–2Zr alloy heat-treated at 950°C for 3 h followed by furnace cooling after processing to a 93% reduction ratio at 1000°C is shown in Table 1. The phase with a dark contrast contained more Al, which is an α-phase stabilizing element, and less Nb, which is a β-phase stabilizing element. By contrast, the phase with a bright contrast contained less Al and more Nb. Therefore, we identified that the phases with dark and bright contrasts are the α and β phases, respectively. It was also found that a higher amount of Zr was contained in the β phase, although Zr is known to behave as a neutral element.
Table 1 Composition of the constituent phase of Ti–13Al–2Nb–2Zr alloy of the equiaxed structure and bi-modal structure. The samples were processed at 1000°C to a 93% rolling reduction ratio and heat-treated by two different heat treatments. F. C. represents furnace cooling and A. C. represents air cooling.
Globularization progressed and formed an equiaxed α phase in the sample processed to a high rolling reduction ratio of 93%, as shown in Figs. 3(a) and (b). The ellipsoid morphology of the α phase remained after heat treatment, as shown in Fig. 3(c). It is thought that the amount of strain introduced during deformation was insufficient for globularization for processing to a 67% reduction ratio, as shown in Fig. 2. By contrast, a lamellar structure was formed in the sample heat-treated in the α+β phase field after processing in the β phase field, as shown in Fig. 3(d). It is thought that the martensitic α phase formed during cooling from the β phase grew and thickened during heat treatment.
3.1.3 Microstructure of heat-treated and air-cooled samples
The microstructure of the samples furnace-cooled after heat treatment at 950°C in the α and β phase field was investigated, as described in the previous section. In this section, the microstructure of the samples air-cooled after heat treatment at 1000°C in the α and β phase field was investigated. Back-scattered electron images of Ti–13Al–2Nb–2Zr alloy heat-treated at 1000°C for 3 h followed by air cooling after processing to a 93% rolling reduction ratio at 900°C, 1000°C, and 1080°C, and processed to a 67% rolling reduction ratio at 1000°C, are shown in Figs. 4(a), (b), (d), and (c), respectively. In Figs. 4(a)–(c), the bi-modal structure containing the equiaxed primary α phase and the α+β lamellar structure formed in the samples processed in the α+β phase are shown. The phase analysis results for the bi-modal structure processed at 1000°C to a 93% rolling reduction ratio by EDS are shown in Table 1. It is clear that the dark contrast phase is the α phase from the phase composition. The lamellar structure formed in the sample is processed at 1080°C in the β phase field, as shown in Fig. 4(d). Because the heat-treatment temperature of 1000°C is just below the starting temperature of transformation to the β phase, the volume fraction of the β phase is higher than that in the heat-treated sample at 950°C. To form a bi-modal structure, the α phase remained, and a lamellar structure was formed by precipitation of the plate α phase in the β phase during air cooling, as shown in Figs. 4(a)–(c). The rolling reduction ratio did not affect the formation of the bi-modal structure. In the case of a slow cooling rate, such as in furnace cooling, the volume fraction of the β phase reaches equilibrium at room temperature. The volume fraction of the β phase at room temperature was calculated as 2% by Thermo-Calc. An equiaxed α phase surrounded by a thin β phase is then formed through furnace cooling, as shown in Fig. 3.
The volume fraction of the equiaxed α phase in the bi-modal structure was measured through an EBSD analysis, as shown in Table 2. Although the samples were heat-treated at the same temperature of 1000°C, the volume fraction of the equiaxed α phase in the sample processed to a 93% rolling reduction ratio at 900°C was 30%, which is higher than that in the sample processed to a 93% rolling reduction ratio at 1000°C, namely, 17%. Compared with the sample processed to 93% and 67% rolling reduction ratios at 1000°C, the volume fraction of the equiaxed α phase in the sample processed to a 93% reduction ratio, namely, 17%, was higher than that in the sample processed to a 67% reduction ratio, namely, 3%.
Table 2 Volume fraction of the equiaxed α phase in a bi-modal structure.
Table 3 summarizes the microstructure of Ti–13Al–2Nb–2Zr formed by thermomechanical processing under the different processing conditions detailed in sections 3.1.1 to 3.1.3.
Table 3 Microstructure of Ti–13Al–2Nb–2Zr alloy produced by thermomechanical processing under different conditions. F. C. represents furnace cooling and A. C. represents air cooling. ( ) indicates the volume fraction of the equiaxed α phase.
To understand the difference in microstructure, a micro-Vickers hardness test was conducted for both the as-processed and heat-treated samples. For reference, the hardness of an ingot was also measured. The micro-Vickers hardness of the ingot, the as-processed, and heat-treated samples is shown in Fig. 5. As Fig. 5 shows, the as-processed samples were harder than the ingot, and the samples processed at low temperature were harder than those of the sample processed at high temperature. This indicates that a strain was introduced during the rolling processing and that rolling processing at low temperature introduced more strain than processing at high temperature. This result corresponds with the GOS maps shown in Figs. 2, respectively. For the samples processed to a different rolling reduction ratio at 1000°C, a higher strain was introduced in the samples processed to a higher rolling reduction ratio than in the samples processed to a lower rolling reduction ratio because the sample processed to a 93% rolling reduction ratio is harder than the sample processed to a 67% rolling reduction ratio. The influence of the microstructure difference on the hardness was also investigated. The microstructure was analyzed after the micro-Vickers hardness test, and the hardness values of the equiaxed α phase and lamellar phase were determined. It was found that there was no distinct difference in the hardness for the equiaxed α structure and lamellar structure. Through heat treatment, the hardness decreased and reached almost the same value as the samples processed under different conditions. This indicates that the dislocations introduced were recovered through the heat treatment. These findings reveal that the equiaxed α phase in the bi-modal structure formed through recrystallization during processing, and thus the driving force of the recrystallization, was larger owing to the high strain introduced in the sample processed at 900°C than in the sample processed at 1000°C.
In general, the creep resistance of Ti alloys depends on the microstructure. The creep resistance of a lamellar structure is higher than that of a bi-modal structure or equiaxed α structure.1) It is also known that the creep resistance of a bi-modal structure is higher than that of an equiaxed α structure.1) Considering the fatigue properties, the bi-modal structure is often used in aerospace applications. Hence, in this work, the creep behavior of the bi-modal structure processed under different conditions was investigated.
The creep curves at 600°C under 137 MPa of Ti–13Al–2Nb–2Zr alloys with a bi-modal structure obtained through heat treatment at 1000°C for 3 h, followed by air cooling after being processed to a 93% rolling reduction ratio at 900°C and 1000°C, are shown in Fig. 6. The creep life of the sample processed at 1000°C was twice that of the sample processed at 900°C. Although a bi-modal structure was formed in both samples, the volume fraction of the equiaxed α phase is smaller in the sample processed at 1000°C than in the sample processed at 900°C. This means the low volume fraction of the equiaxed α phase contributed to an extended creep life in the sample processed at 1000°C. In other words, when the α+β lamellar region is larger, the creep life becomes longer. This corresponds with a previous study indicating that the creep strain after 100 h of a bi-modal structure with a low volume fraction of the equiaxed α phase was smaller than that with a high-volume fraction of the equiaxed α phase in TIMETAL 834.1) The reason for this is explained through the alloying partitioning effect, and the strength of the lamellar structure continuously softens owing to a decrease in Al content with an increasing volume fraction of the equiaxed α phase. In another study, the creep behavior of TIMETAL 834 with an equiaxed α phase from 6% to 40% was investigated, and the creep life became shorter with an increase in the volume fraction of the equiaxed α phase.9) Some explanations for this have been introduced elsewhere,9) including the following:
-
(1)
The lower strength of the lamellar structure as a result of alloy element partitioning.10)
-
(2)
Banding and texture of the equiaxed α leading to accelerated creep.11,12)
-
(3)
A larger slip length of the equiaxed α than that of the α plate in the α+β lamellar structure.13)
-
(4)
Grain boundary sliding being the dominant creep mechanism.14)
It was concluded that alloy-element partitioning in the equiaxed α phase reduces the strength of the lamellar structure.
9)
In this study, the compositions of the equiaxed α phase and lamellar area in the bi-modal structure in the sample processed at 900 and 1000°C were investigated before and after the creep test. The compositions are summarized in Table 4. The compositions of the equiaxed α phase and lamellar area did not change during the creep test. Compared with the composition of the samples processed at different temperatures before the creep test, those of the equiaxed α phase and lamellar area of the samples processed at 900°C and 1000°C, respectively, were almost the same, regardless of the processing temperature. The composition of the microstructures was also the same after the creep test for the samples processed at 900°C and 1000°C. As there was no distinct difference between the compositions of the equiaxed α phase and lamellar area in the samples processed at 900°C and 1000°C, it is difficult to confirm that the difference in the alloy element partitioning in the lamellar area caused the different creep behavior.
Table 4 Composition of the equiaxed α phase and the lamellar area in the bi-modal structure processed at 900 and 1000°C before and after creep test.
To understand the creep behavior, the microstructure after the creep test was observed. Back-scattered electron images, the inverse pole figure (IPF), and the GOS, are shown in Fig. 7. The back-scattered electron images indicate the deformation of the equiaxed α phase in both samples, as shown in Figs. 7(a) and (b). The bright-contrast region was the prior lamellar structure shown in Figs. 7(a) and (b), although IPF maps indicate fine grains of a few micrometers in size, as indicated in Figs. 7(b) and (e). In the GOS maps, which indicate the average deformations in the grain, these fine grains showed a darker contrast area with a lower strain of less than 5%. The bright-contrast area indicates that the structure could not be identified as EBSD because the grains were too small or the strain was too severe. It can be concluded that the fine grains shown in Figs. 7(b) and (e) were formed in the prior lamellar structure through dynamic recrystallization during the creep test. It is thought that recrystallization in the lamellar structure causes an acceleration of the creep deformation. Thus, control of the creep life is considered to depend on the time required to accumulate dislocations in the lamellar structure for a dynamic recrystallization. As described above, the dislocation slip length of the equiaxed α is larger than that of α plate in the α+β lamellar structure.13) This suggests that the α/β interface in the lamellar structure resists the dislocation movement, and as a result, the equiaxed α phase deforms first and dislocations pass during the equiaxed α phase, but stop in the lamellar structure. Thus, a sample with a low volume fraction of the equiaxed α phase takes longer to multiply and accumulate sufficient dislocations for recrystallization in a lamellar structure than in a sample with a high-volume fraction of the equiaxed α phase. As a result, the creep life of the sample processed at 1000°C with a low volume fraction of the equiaxed α phase was longer than that processed at 900°C with a high volume fraction of the equiaxed α phase. However, after sufficient dislocations were accumulated in the lamellar structure, the deformation was accelerated through dynamic recrystallization for both samples.
To understand the creep deformation mechanism, an analysis using an Arrhenius equation was conducted. The stress exponent was evaluated using a single creep specimen heat-treated at 1000°C for 3 h followed by air cooling after processing at 1000°C to a 93% reduction ratio by changing the applied stress at 69, 104, 139, 174, 208, and 243 MPa at 600°C after the steady-state creep was reached. The creep curve is shown in Fig. 8. Under a constant temperature, the relationship between the steady-state creep rate $\dot{\varepsilon }_{\text{ss}}$ and the applied stress σ is represented through eq. (1).
| \begin{equation}
\dot{\varepsilon}_{\text{ss}} = \mathrm{B}\sigma^{n}
\end{equation}
| (1) |
Here, B is a material constant, and
n is a stress exponent.
15) Equation (1) with the logarithm of both sides is shown in
eq. (2).
| \begin{equation}
\ln \dot{\varepsilon}_{\text{ss}} = n\ln\sigma + \ln \mathrm{B}
\end{equation}
| (2) |
We estimated the steady-state creep rate for each applied stress from
Fig. 8, and the double logarithmic graph of the steady-state creep rate and applied stress is plotted in
Fig. 9. The stress exponent was evaluated by a slope of the graph shown in
Fig. 9. The stress exponent, n, of the bi-modal structure was 3 for low stress (σ ≦ 174 MPa) and 4 for high stress (σ ≧ 174 MPa). It is known there are two major deformation mechanisms occurring during creep. One is the dislocation creep (known as the power-law creep) in which the steady-state creep rate is proportional to the exponent of the stress, and the other is diffusion creep in which the steady-state creep rate is proportional to the stress.
15) The stress exponent of dislocation creep caused by the climbing and gliding of a dislocation is generally 3–8. By contrast, a lower applied stress than that triggering dislocation creep causes diffusion in the lattice or grain boundary because the applied stress operates as the mechanical driving force for the diffusion of atoms.
16) Under this condition, deformation occurs through lattice or grain boundary diffusion, which is called diffusion creep. The stress exponent of diffusion creep is 1. In our creep test, because the stress exponents were 3 and 4, the deformation mechanism of the bi-modal structure in Ti–13Al–2Nb–2Zr is identified as dislocation creep. The change in the stress exponent from 4 under a high applied stress to 3 under a low applied stress is considered to be the deformation mechanism that starts the change from dislocation creep to diffusion creep. The deformation mechanism is naturally considered to change continuously. Thus, under a low applied stress condition, although the main deformation mechanism is dislocation creep, it is affected by diffusion creep. In our previous study, the deformation mechanism of creep at a temperature of between 550°C and 650°C under a pressure of 137 to 240 MPa was investigated for Ti–10Al–2Nb–2Zr and Ti–10Al–2Nb–2Zr–0.5Si with an equiaxed α structure.
7) The stress exponent of the equiaxed α structure was 3.4 and 5.9, respectively, and the deformation mechanism was concluded as the power-law creep. This indicates that the deformation mechanism during the steady-state creep of the equiaxed α structure and bi-modal structure is the same even if considering the effect of solid-solution hardening due to the increase in Al addition in Ti–13Al–2Nb–2Zr from Ti–10Al–2Nb–2Zr. The creep life of the bi-modal structure, namely, 973 and 2,492 h for Ti–13Al–2Nb–2Zr processed at 900°C and 1000°C, was longer than that of the equiaxed α structure, i.e., 142 and 257 h, for Ti–10Al–2Nb–2Zr and Ti–10Al–2Nb–2Zr–0.5Si. It is considered that the time required to accelerate deformation was longer in Ti–13Al–2Nb–2Zr alloys than in Ti–10Al–2Nb–2Zr because of the bi-modal structure and stronger solid-solution hardening effect due to the increased Al content.
There have been numerous studies investigating the deformation mechanism. For example, the stress exponent of CP-Ti at 500°C was shown to be 4.5, and the deformation mechanism was concluded as high-temperature dislocation creep controlled through lattice diffusion.17) For the bi-modal structure, the stress exponent of Ti-64 at 600°C was 4.6, and the deformation mechanism was concluded as dislocation creep.18) Our results correspond with those of the previous study. Considering the previous reports and our own results, it can be stated that the deformation mechanism of creep in titanium or titanium alloy at 600°C is dislocation creep regardless of the microstructure and composition.
