Characterization of the prior particle boundaries in a powder metallurgy Ti₂AlNb alloy

1. Introduction

Ti₂AlNb-based alloy found by Banerjee et al. [1] has been a potential material for gas turbine engine owing to a higher specific strength at elevated temperature than conventional common titanium alloys and has excellent room temperature ductility among intermetallics [2,3]. However, Ti₂AlNb-based alloy is prone to composition segregation during smelting, as well as shrinkage, porosity and other casting defects are easy to occur in solidification because the melting point and density of each element in the alloy vary greatly [4]. The inhomogeneous microstructure can result in obvious fluctuation of mechanical properties in the alloys. It is well known that powder metallurgy (PM) could avoid the casting defects above associated with the large ingots production. Particularly, PM is also attractive for Ti₂AlNb alloys taking into account that the mechanical properties of the alloys are very sensitive to composition variation [5,6].

Prior particle boundary (PPB) is a type of common defects, which is generally composed of a thin layer of second phase formed around prior particles during hot isostatic pressing (HIP) process in PM products [7,8]. The existence of PPB can damage the mechanical properties of the alloys, such as tensile and impact strength [9]. Therefore, special care has to be taken to eliminate the adverse effects of PPB during HIPping. It is usually thought that this problem is resulted from powder surface contamination, especially in superalloys [10,11]. Cai et al. [12] have observed PPB for simultaneous loading and heating during HIPping in Ti-6Al-4 V alloy, but there is no PPB observed for elevating temperature first before raising pressure. The metallurgical defects formed can cause crack to propagate along PPBs. Considering the carbide precipitates along PPBs can inhibit grain boundary migration and lead to interparticle fracture and low ductility, Chang et al. [13] has made the PM superalloys Inconel 718 free of PPBs. Our previous work showed that, PPB may stem from the precipitation of α₂ in Ti₂AlNb-based alloys, however there were no detailed characterization and analysis for the reason why the α₂ was present there, and the role of α₂ and other coexisting phases [5,14]. In this work, Ti-22Al-24Nb-0.5Mo pre-alloyed powders were filled in mild steel capsules, then degassed and HIPped, under a number of HIPping condition parameters (temperature 920-1080 °C, and pressure 120-140 MPa, held for 3 h). It is noted that PPB occurs mostly in the samples HIPped at relatively low temperature, e.g., 920 °C and 980 °C regardless the pressure of 120-140 MPa. According to the previous report, PPB is more sensitive to HIPping temperature than applied pressure, and the PPB precipitates can be dissolved through HIPping at a higher temperature for reducing the pinning effects of moving grain boundary [11].

Generally, there mainly exist three phases in the microstructure of Ti₂AlNb-based alloy including close-packed hexagonal α₂ phase, body-centered cubic B2/β phase [15] and orthorhombic Ti₂AlNb phase (O phase), respectively [16]. The O phase has a specific orientation relationship with B2 phase (B2-O OR here after) as {001}_O || {110}_B2, <110>_O || <1-11>_B2 [17]. The α₂ phase precipitating from B2 phase maintains the well-known Burgers orientation relationship (BOR) [18], {0001}α₂ || {110}_B2 and <11-20>α₂ || <1-11>_B2. For the α₂ phase precipitated from B2, the BOR would result in five axis/angle pairs for different types of α₂/α₂ variant boundary, and with misorientation of 10.53°, 60°, 60.83°, 63.26° and 90°, respectively [19]. Recently, Huang et al. [20] found that the growth of α₂ on grain boundaries in Ti-22Al-25Nb alloy is detrimental to mechanical properties of the alloy. In addition, some α₂ phase may precipitate from the O phase, and thus α₂ may also follow a certain orientation relationship with O phase (α₂-O OR here after), i.e., {0001}α₂ || {001}_O and <11-20>α₂ || <110>_O. These well-defined orientation relationships can give a guidance on the study of micro-texture of α₂ and O phase by variant selection around some large/low angle grain boundary of B2 in Ti₂AlNb-based alloy [21]. Recent phase field simulations show that the formation of micro-texture can be attributed to variant selection under external stresses [22], pre-existing defects in the matrix [23,24], and the effect of undercooling [25] etc. Besides simulations, Cao et al. [26] found that the texture of <0001 > _α || Rolling Direction (RD) can form after powder sintering and hot rolling in Ti-6Al-4 V alloy. The texture can also be observed in hot extruded magnesium of rapid solidified powder [27]. Additionally, Li et al. [28] proposed that the deformation of pre-alloyed powder during spark plasma sintering might be the reason for the appearance of the weak texture. However, it is not clear if micro-texture could occur in PM Ti₂AlNb-based alloy.

There has little research on detailed description for PPB in Ti₂AlNb-based alloys so far, so the main objectives of this study are to identify the origin of PPB, characterize the morphology, composition and distribution of the different phases, and analyze the micro-texture of α₂ and O phase via variant selection, further, study the inter-relationship between micro-texture and misorientation frequency in different phases respectively. In fact, there is no accurate phase diagram for this alloy after adding additional 0.5 Mo and 0.1 O [2,29,30]. In this study, the DTA test combined Pandat calculating phase diagram was used to determine the boundary between the phases. How to optimize the HIP process through reducing the appearance of PPB in Ti₂AlNb alloys from the perspective of phase diagram were discussed. In addition, the micro-texture in the alloys can be weakened by controlling the misorientation between B2 grains or in B2 grain.

2. Materials and methods

Pre-alloyed powder of Ti-22Al-24Nb-0.5Mo was prepared by the electrode induction melting gas atomization (EIGA). The chemical composition of pre-alloyed powder is shown in Table 1. The particle size of the powder ranges from 5 to 250 μm, with an average of ∼100 μm [31]. The gas atomized powders were canned in a mild steel container via powder filling and tapping process. In order to get rid of the adsorbed H₂O and O₂, the capsule was degassed through a typical process at elevated temperature under high vacuum, and then the capsule was sealed gas tight. The canned powder compacts were then HIPped for 3 h at 920-1080 °C at pressure of 120-140 MPa respectively. HIPping operations with pressurizing and heating simultaneously were conducted in a Mini-HIPer of QIH-15 type.

X-ray diffraction (XRD) analysis was made for the angular range of 2θ = 10°-80°. The microstructure of as-HIPped samples was observed using a scanning electron microscope (SEM). In order to understand the detailed orientation evolution of B2, α₂ and O phase during HIPping, electron backscatter diffraction (EBSD) was carried out. The EBSD samples were polished by Buehler VibroMet 2 vibratory polisher for 14 h. An Oxford -S-3400 N SEM, equipped with HKL channel 5 software was used for EBSD data acquisition and analysis. The step sizes of 0.3, 0.5 μm were used for analysis on areas of 158.1 μm × 118.5 μm and 159.5 μm × 119.5 μm in the middle of the samples HIPped under 920 and 980 °C, respectively. Electron probe micro analysis (EPMA) using JEOL JXA-8530 F was carried out to measure quantitatively the element distribution in different phases and phase boundaries. Thermal characteristics involved the phase transformations were measured using Setsys Evolution 18 thermogravimetric analysis-differential thermal analysis (TG-DTA) by heating and cooling Ti₂AlNb pre-alloyed powder from 600 to 1100 °C at a rate of 10 °C/min. The phase diagram of Ti-22Al-0.5Mo-0.1O-20∼28Nb was drawn using the Pandat software.

3. Results and discussion

3.1. Microstructure analysis by XRD and SEM

Fig. 1 shows the XRD patterns of the Ti₂AlNb-based alloys HIPped under 920 and 980 °C. Two profiles are almost the same except for the different intensity of two main peaks labeled as B2/β and O phase. It can be observed that O and B2/β phases are the main phases in the alloy, with small amount of α₂. Interestingly, for 920 °C sample, the strongest peak of B2/β is slightly higher than that of O phase; indeed the fraction of B2 is about 30% more than that of O phase, as computed with the RIR values for B2/β (8.68) and O (6.89) [32,33], using the RIR parameters corresponding to space group Im3m and Cmcm, respectively. From the EBSD test for 920 °C sample, as in Section 3.3, the fraction of O is shown also smaller than that of B2/β.

As shown in Fig. 2(a) and (b), the 920 °C HIPped microstructures consist of equiaxed α₂ phase, irregular shaped B2 phase around α₂, rimmed O phase between α₂ and B2, and B2 + O lamellae. When the HIPping temperature raised to 980 °C, it is difficult to find blocky B2 phase; B2 + O lamella with basketweave structure are the main component, however, the thickness of B2 + O lamellae is significantly thinner than that at 920 °C as shown in Fig. 2(c) and (d). This is because the HIPping at 920 °C is done in the three-phase field, i.e., α₂+B2/β+O, the O phase precipitates can grow sufficiently at this temperature.

The rimmed O and equiaxed α₂ phase also appear, with rimmed O phase getting thinner than those at 920 °C as well. From Fig. 2(a) and (c), it is noticed that the PPB is not obvious due to the increase of HIPping temperature, since the interphase fusion is easier among powder particles.

3.2. EPMA test for different alloy elements

The element partitioning with the diffusion of α stabilizing elements (Al, O) and β stabilizing elements (Nb, Mo) into corresponding phases is an important aspect of the microstructural evolution during thermal-mechanical processing [34,35]. As shown in Table 2, the concentrations of Ti, Al are enriched in α₂ phase, and depleted in O and B2 phases. The diffusion of Ti element is often neglected, due to its larger atomic radius, and it is worth noting that its diffusion into α₂ phase generally leads to a decrease in the lattice constant of the B2/β phase [36]. What is more, Nb and Mo are poor in α₂ phase, but enriched in B2 phase. The element partitioning is relatively weak in O phase, as compared to the former two phases. Here no apparent oxygen partitioning was detected, as seen from Fig. 3(f). The original intention of EPMA test is to check if the production of PPB is related to the presence of oxides or other impurities on the powder particle boundary. From the experimental results of element distribution maps above, no oxygen partitioning is found, probably because the oxide layer is extremely thin [37], the amount of oxygen on the surface layer is small and has already diffused into the neighboring lattice and the difference of concentration arround the PPB and other places is already below the resolution limit of the EPMA. Moreover, it is found in the previous study that the Al₂O₃ and TiO₂ surface layer on Ti₂AlNb-based alloy by X-ray Photoelectron Spectroscopy (XPS) are only several nanometer thick [38]. At the same time, as an α stabilizing element, the oxygen on the surface of powders can promote the precipitation of α₂ (Ti₃Al), and these compounds can hinder the formation of good metallurgical bonding between particles.

From the experiential phase diagram [2] and the one calculated using Pandat in Section 3.4, the α₂ phase comes out from B2 phase directly, forming B2 and α₂ region, then O and β form at lower temperature. However, the details of the precipitation process of the α₂ phase are currently unknown.

Since the oxygen is an α stabilizer, it diffuses into the surface layer forming thicker α layer compared with the α layer at a higher temperature. The thicker α can be a trap to the α stabilizer, so that Al goes into it while Nb and Mo escape form it. Forming more stable α compared with the layer α formed from the lattice. The α first attracts more α stabilizer and forms more α there, and then α₂ forms more inside α. With the increase of temperature, oxygen escapes further due to faster diffusion into the lattice, so that it is not detectable after HIPping. The details of the diffusion and phase transformation associated with α layer formation are dependent on the conditions of temperature and vacuum.

Fig. 4 shows the three different areas selected for quantitative composition measurement, from the center of the rod to the corner. Three test points for each phase are shown in Fig. 2(b) and (d). The spacial resolution is about 1.0 μm. The alloying element compositions in different phases are shown in Fig. 4(b), and the specific compositions are given in Table 2.

Under 980 °C HIPping, many acicular O phase precipitate from B2 phase, forming very fine mixture of the two phases, so the content detection in this area would consist of measuring B2/O phases together. The composition of the B2/O phases is average of the B2 and O phase obviously over the volume of the two phases present in the probe. As the HIPping temperature increases, the partitioning of alloying element is enhanced, e.g., the Al content in α₂ increased by 0.88 at.% (from 23.59 to 24.47) as the HIPping temperature changed from 920 °C to 980 °C, meanwhile in contrast, the Nb and Mo contents in α₂ decreased by 1.71 (from 16.61 to 14.90) and 0.06 at.% (from 0.14 to 0.08), respectively.

3.3. Orientation relationship between phases tested by EBSD

In order to measure the orientation and distribution of the α₂ phase inside and at the boundaries of particles, EBSD characterization was performed on the two samples of 920 and 980 °C. Fig. 5(a) and (b) shows the microstructure of 920 °C sample, consisting of the three phases, i.e., O, B2 and α₂. The fraction of α₂ phase is 8.15%, and O and B2 account for 32.85% and 40.77% (1:1.24), respectively, as estimated from the XRD test above, which can be concluded that the EBSD consists with the XRD results. The equiaxed α₂ phase precipitates along the boundaries of prior particle, and the discontinuous α₂ phase is arranged near the cell wall on the particle shown in Fig. 5(d), just like the precipitation of α_GB allotrimorphs [39]. Zhou et al. [40] have studied the hot deformation behavior and microstructure evolution in Ti₂AlNb-based alloy in dual phase field, no obvious dynamics recrystallization grains found at the boundaries of B2 phase below 990 °C whether with a low or high strain rate. Thus, to a large extent, the region among particles is filled by a number of equiaxed α₂ phases due to the deformation and fragmentation of smaller sized particles as seen in Fig. 5(d), called “severe deformation zone” here after.

Since B2 phase has no obvious micro-texture even after HIPping at 920 and 980 °C, it is not shown here. From Fig. 5(f), O phase lamellar precipitates mainly form inside of the cell. Fig. 5(e) and (g) shows the micro-texture of α₂ on {0001} and O on {100} pole, respectively, probably because the EBSD test area is small, and the proportion of cell crystals is larger than that of “severe deformation zone” in the test area. The orientations of B2 phase cells in the local area are close without the severe deformation during HIPping, and the local stress state in B2 cells is similar, which is a key factor in controlling variant selection of α variants and transformation texture [41].

Moreover, α₂ and B2 phases satisfy the BOR during the phase transformation, and then a certain micro-texture may occur due to variant selection [42,43]. However, in the “severe deformation zone”, this kind of orientation relationship is severely destroyed, the regions having deviation [44] larger than 10° from BOR are marked black, as seen in Figs. 5(c) and 6 (c). Ye et al. [45] found that stress concentration maybe occur inside powders, while strain concentration was in “severe deformation zone” through a finite element simulation. Therefore the OR among B2/α₂/O could be severely destroyed at the boundary of powders.

The volume fractions of α₂, O and B2 phases are 8.63%, 43.92% and 35.48%, respectively after 980 °C HIPping, as shown in Fig. 6(a) and (b), Fig. 6(d) shows the “severe deformation zone”. Compared with the micro-texture of α₂ and O phases at 920 °C, the micro-textures of α₂ and O are weaker in Fig. 6(e) and (g). The reason for this may be due to the change of misorientation of α₂ and B2 phase under a higher HIPping temperature. With the HIPping temperature raising, the driving force of transformation B2→α₂ during cooling process increases, and thus more different α₂ variants would be precipitated. Moreover, it is more favorable for deformation at high temperatures, so the orientation of α₂ can be changed. The orientation change of B2 phase will be explained in detail in the following discussion.

Fig. 7 shows the frequency distribution of the misorientation (point to point) for the three phases at 920 °C and 980 °C. It can be seen that the main frequencies of the three phases are concentrated on low misorientation, which are less than 10°. For α₂ phase, with the increase of HIPping temperature from 920 °C to 980 °C, the misorientation around 2° and 60° decreases while that of ∼30° increases, as seen in Fig. 7(a) and (d).

The orientation variation of α₂ phase from different locations at 920 °C is shown in Fig. 8. In Fig. 8(a) and (b), for α₂ phase at the cell wall, the misorientation of ∼60° observed using EBSD, while higher frequency of misorientation of ∼30° is present when α₂ phase precipitates around the boundary of particles, as seen from Fig. 8(c) and (d). Near the cell wall, the deformation is relatively small, the α₂ phases still satisfy the BOR with B2 matrix, so the misorientation of ∼60° due to different variants occurs [19]. However, the α₂ at the particle boundary may precipitate from different B2 grains, thus ∼30° is observed.

For O phase, the misorientation around 60° decreases while that around 90° increases, as seen in Fig. 7(c) and (f). And the frequency distributions of misorientation in B2 phase under 920 and 980 °C are similar. Interestingly, there is a larger orientation change of B2 phase in the cell crystal with the type of point to origin, seen from Fig. 9. Under each temperature, two cells with B2 phase are randomly selected respectively, as seen in Fig. 9(a, b) and (d, e). Then the results of line scan from EBSD data are drawn in Fig. 9(c) and (f). It is noticed that the orientation variation of B2 phase in the cell becomes larger from 920 to 980 °C: for 920 °C, the misorientation with point to origin ranges from ∼2° to ∼5°, while that ranges from ∼2° to ∼12° at 980 °C. The orientation change of B2 phase can also weaken the micro-texture of α₂ and O phases through phase transformation B2→α₂, α₂→O and B2→O indirectly. It is noticed that the boundaries between B2 phases in the cell crystal are all low angle grain boundary basically. Differently, those on a different cell wall are mainly large angle grain boundaries, seeing the following discussion from Fig. 10, Fig. 11.

From Fig. 10(a)-(f), it is concluded that the green and red variants of α₂ keep the strict BOR with the dark blue grain of B2 phase. And the red variant of α₂ maintains an approximate BOR with the pink grain of B2 while the green variant of α₂ does not. This is consistent with the work of Bhattacharyya et al. [46]. At the same time, the O phase not only keeps the specific B2-O OR with B2 phase, but also has a specific α₂-O OR with α₂. The light blue and pink O phases correspond to green and red α₂ respectively seen from Fig. 10(g)-(i).

In Fig. 11, it can also be seen that the red α₂ variant keeps a BOR with dark blue B2 phase, while the purple α₂ variant has also a BOR with green B2. Moreover, the red α₂ variant maintains an approximate BOR with green grain of B2 phase, while the purple α₂ variant does not satisfy the BOR with dark blue grain of B2. The red O phase precipitates along the red α₂ variant, it satisfies a specific α₂-O OR with red α₂ phase, and a B2-O OR with dark blue B2 phase, keeps an approximate B2-O OR with green grain of B2. The purple O phase has the similar OR with purple α₂ and green B2 phases. From (b) and (c) in Fig. 10, Fig. 11 above, it is noticeable that there is a large angle boundary (>15°) between the two adjacent B2 grains, i.e., cell grain, seen from Figs. 10(a) and 11 (a), respectively. For Fig. 9 shown above, some low angle boundaries are formed inside the cell of B2 phase.

When the HIPping temperature is 920 °C, the internal microstructure of the typical cell structure is interweaved, which mainly consists of the O phase and the B2 phase, and a small amount of the α₂ phase, as seen in Fig. 12. It can be found that α₂ phase and B2 phase maintain a strict BOR, and the O phase satisfies a certain orientation relationship α₂-O OR and B2-O OR with the α₂ phase and the B2 phase, respectively. From Fig. 13, it is found that the pink and light green O phase variants only maintain the specific α₂-O OR with the pink and green α₂ variants, respectively, and they deviated the B2-O OR with green B2 phase here. The phase transformation path of α₂→O appears obviously inside the cell under HIPping temperature of 980 °C. This transformation path not related to the OR of the surrounded B2 would be present when cooling from a higher temperature, it may be due to a larger transformation driving force during cooling process.

3.4. DTA thermogram of Ti₂AlNb pre-alloyed powder

To explore the boundary of phase fields of the Ti₂AlNb-based alloy, the heating differential thermal analyses from 600 to 1100 °C are done firstly, then cooling down in the same temperature range, as seen in Fig. 14. The rate for heating and cooling is 10 °C/min. The 1 st derivative of heat flow can be obtained from the DTA curves, and then some key points for distinguishing the phase field boundaries can be found, so the boundary of phase field could be drawn in Fig. 15(a) referring to the phase diagram information drawn using Pandat software in Fig. 15(b).

The exothermic peak at 1060 °C in Fig. 14(b) corresponds to the endothermic valley at 1057 °C seen in Fig. 14(a), while another exothermic peak at 845 °C during cooling corresponds to endothermic valley at 939 °C when heating sample. Considering that 1060 °C and 1057 °C are close to each other, 1071 °C can be selected as the transformation point of B2→B2+α₂ from the obvious minimum seen from 1 st derivative of heat flow in Fig. 14(d). When cooling sample from high temperature, the precipitation of O phase needs to overcome resistance including interfacial energy, strain energy etc., and therefore the transformation temperature of 862 °C for cooling should be lower than that of 959 °C during heating process. So 959 °C could be chose as the phase field boundary during B2+α₂→O₁+B2+α₂. Note that there are two distinctly ordered forms of the O phase, i.e., O₁ and O₂ [47]. Al atoms are all in the 4c₁ sites for O₁ and O₂, a random occupation of Ti and Nb atoms in sites 8g and 4c₂ is in O₁, however, Ti and Nb atoms are in the 8g and 4c₂ sites respectively for ordered O₂ phase.

There is no clear peak or valley from ∼600 to ∼800 °C under cooling condition in Fig. 14(b). However, there are one exothermic peak at 648 °C and an unobvious exothermic reaction at 750 °C during heating process in Fig. 14(a). It may be that the alloy has a larger modulus of elasticity at low temperature during cooling process, which causes the resistance of phase transformation to increase, and the corresponding endothermic valley has not been reflected in the current temperature range. In addition, the high temperature B2 phase is retained under high-speed cooling conditions while a part of the energy is stored during the preparation of powders. When it is heated to the low temperature phase field, this portion of energy would be released. Next, it can be inferred from the DTA curve under heating, 759 °C may be the transition point for the transformation from O₁ to O₂ (as a first-order phase transition [47]) by releasing a small amount of heat, while 700 °C is probably the transformation temperature from O₁+O₂+B2+α₂ to O₂+B2+α₂. Compared with Fig. 15(a), the corresponding part of the previous phase analysis can be found in the phase diagram of alloy using commercial Pandat software, shown in Fig. 15(b).

980 °C is located at the bottom of the two-phase field, the PPB gradually becomes weakened due to the reduction of α₂ precipitation. According to the previous work [5], PPB disappeared when HIPping temperature is 1030 °C. It can be inferred that increasing temperature and reducing the precipitation of α₂ phase during the HIPping process are beneficial to suppressing the occurrence of PPB. So the HIPping temperature can be raised to the top of the two-phase zone below the B2 single-phase zone, considering the pinning effect of the α₂ phase can inhibit the grain growth of B2 phase in the two-phase region, while in the single-phase field, the larger grain size of B2 can affect the tensile ductility for Ti₂AlNb-based alloy [48].

HIPping at high temperature is conducive to the compaction and fusion of the powders dominated by B2 phase, changing the α₂ orientations and increasing the orientation difference between B2 phases, therefore weakening the micro-texture of the α₂ phase and the O phase.

The processing path of simultaneous heating and boosting is adopted, and the temperature and pressure are maintained at the target values after about 3 h. What has to be discussed here is the potential effect of the heating rate during HIPping at a certain temperature (e.g., the top of two-phase field). When the heating rate is low, the deformation stage of HIP is mainly concentrated in the low temperature, at this time, O, B2/β and α₂/α coexist. When the temperature is raised to the target temperature, it may be only a heat preservation process at the stage of holding temperature and pressure. In the deformation stage during HIPping, O and α₂ are generally not conducive to plastic deformation [49,50], and the local stress may also cause variant selection. When the temperature rises to the target temperature quickly, the alloy is mainly composed of B2 phase, which is favorable for deformation [50], easy to be dense and fused, can increase the orientation difference between B2 phases, and it is also beneficial to inhibiting the formation of micro-texture.

4. Conclusions

(1) More α₂ phase is the characteristics of the PPB in Ti₂AlNb-based alloy when HIPped at relatively low temperature. Increasing HIPping temperature to the upper part of the two-phase region can effectively inhibit the formation of PPB, while care should be taken to keep the temperature below the single phase field of B2 to avoid rapid grain growth.

(2) When HIPping temperature raising from 920 to 980 °C, the deformation induced orientation change in B2 has an increase from ~5° to ~12° due to the deformation at a higher temperature. The HIPping at a higher temperature can weaken the micro-texture of α₂ and O phase with increase of misorientation between adjacent B2 phase.

(3) Some regions among particles are filled by a number of equiaxed α₂ phases due to the deformation and fragmentation of smaller sized particles, forming severe deformation zone, the orientation relationships there among B2/α₂/O are severely destroyed.

(4) The α₂ phase at the cell wall maintains the BOR with grain in one side and does not satisfy the BOR with the other. Some O variants inside the cell after 980 °C HIPping maintain only α₂-O OR with α₂ phase owing to the α₂→O transformation producing O, but no B2-O OR with B2 phase.

Acknowledgements

The supports from the National Key Research and Development Program of China (No. 2016YFB0701304), the CAS Informatization Project (No. XXH13506-304), and the Doctoral Scientific Research Foundation of Liaoning Province (No. 20180540133) are gratefully acknowledged. Simulations were performed at the CAS-Shenyang Supercomputing Center.

---