3.1.2. Influence of Milling Time on Microstructure of the MA Powders
The properties of MA powders strongly depend on the control of the MA process, including rotation speed, milling time, BPR, large/small milling ball mass ratio, process control agent (PCA), milling atmosphere, and temperature of milling [
42]. Considering that the powders used in this study were very fine, the BPR was chosen as 15:1 and the large/small milling ball mass ratio was chosen as 1:5. As the dissolution of W was regarded as a key signal of effective ball milling [
54], the rotation speed was chosen as 300 rpm, to promote the dissolution of elemental W [
55]. To avoid contamination from PCA and oxygen, PCA was not added, and the milling process was carried out in a high-purity argon atmosphere. In addition, the intermittent milling method was applied, to avoid overheating the powders. The MA powders were ball milled for 5 min, 16 h, 32 h, 48 h, and 64 h, to study the effect of milling time on them.
Figure 6 shows the XRD patterns of the alloyed powders with different milling times. With the increase in milling time, the (Fe, Cr)
(110) peak width increased, due to the grain refinement and internal stress, while the W peak intensity decreased. When the milling time extended to 48 h, the W peak disappeared, which means that W was completely dissolved into the matrix. Other substances (such as Y
4Zr
3O
12) were undetectable by XRD, due to their low contents.
Figure 7 mainly presents the morphology of the powders after different milling times, investigated with SEM. W powders were attached around (Fe, Cr) powders at the initial stage of ball milling, while many powders still maintained the initial spherical shape. A small amount of the particles were deformed, and coldly welded, under the impact of the milling balls and milling container (
Figure 7a). It is well known that particles are repeatedly plastically deformed, coldly welded, and fractured during MA, and finally reach a dynamic stable state [
42,
47,
55]. As the ball-milling time increased to 16 h, almost all the particles were irregularly shaped (
Figure 7b), and the particles tended to be welded together to form large particles. Thus, the average particle size increased. The lamellar powder morphology caused by plastic deformation was observed after 32 h of milling (
Figure 7c). When the milling time extended to 32–48 h, the effect of cold welding between particles continued to increase, and it could be clearly seen that several small particles made up a large particle (
Figure 7c,d). After that, the particles were crushed by a violent collision of grinding balls and the average particle size decreased (
Figure 7d,e). It could be estimated that the coarsening caused by welding would be balanced by the refining caused by crushing, and the particle size would tend to be stable after a long period of ball milling.
On balance, the ball-milling time was chosen as 48 h, to avoid potential contamination from an excessive ball-milling time [
54], and the rotation speed of 300 rpm was sufficient to facilitate the complete dissolution of W. The powder size distribution, counted by a laser diffraction particle size analyzer, after 48 h of ball milling, is shown in
Figure 7f. The value of D
50 of the alloyed powders after 48 h of ball milling, was only 5.33 μm, with the distribution mainly ranging from 3 μm to 8 μm, which indicates that the alloyed powders used in this study were extremely fine [
47,
55,
56,
57]. The element distributions in the powders, observed with SEM-EDS after 48 h of ball milling, are shown in
Figure 8. These revealed that the elements were evenly distributed on the surface of the powders. In conclusion, milling for 48 h was required to obtain refined powders with a homogeneous element distribution, which was favorable for the SPS process.
3.1.3. Influence of SPS Temperature Combination on Microstructure of the Sintered Samples
The alloyed powders, after 48 h of ball milling, were consolidated by SPS, and the temperature and pressure cycles are schematically illustrated in
Figure 2. Based on the obvious displacement of the punch at about 750 °C, the two temperature platforms were adopted, to complete the sintering process. The specific sintering conditions and relative densities of the specimens are listed in
Table 2. The two temperature platforms, together affected the relative densities of the sintered specimens, which was not only determined by the maximum temperature. At the temperature of platform 1, the powders became relatively soft and sintering necks tended to be formed, while the higher temperature at platform 1 helped to facilitate the short-distance movement of the particles, and to ensure a better fit between particles and particles. Considering that a temperature of 1150 °C would cause the samples to melt, sample A4, with the highest relative density, was considered as a “relatively better sample” and was further studied in some respects.
The XRD patterns of specimens A1–A4 are shown in
Figure 9. They indicate that the sintered specimens were all α-γ dual-phase, while γ was the main phase. However, according to the Fe–Cr phase diagram [
11], Fe–13.5Cr should be fully ferritic throughout the heating process, and W should further shrink the austenitic phase zone. Thus, Fe–13.5Cr–2W would not undergo austenite transformation at high temperature [
12]. The contents of some common austenitizing elements, of the as-milled powders, are listed in
Table 1. The results show that the contents of C and N were much higher than expected, which was the origin of the austenite transformation. There were also some M
23C
6 precipitates in the specimens. Furthermore, the contents of C and N in the raw powders and the as-sintered steel were detected, and the results are displayed in
Table 3. This shows that the C and N were almost exclusively derived from the raw Fe powders prepared by atomization, which were claimed to be qualified. In addition, no new C or N contamination was added into the steels during the sintering process. To eliminate the interference of accidental factors, a new batch of Fe powders, with a particle size of 5 μm, produced by another company, using an electrolytic method, were prepared, and the contents of C and N in the two kinds of Fe powders are shown in
Table 4. The contents of C and N are seen to be similar between the two kinds of Fe powders, which was unexpected, and seemed to be close to the maximum naturally dissolved amounts of C and N. A possible explanation for this result, is the different relative contents of C and N in raw Fe powders with different sizes, during the Fe powder production process. It is generally recognized that powders with a wide size distribution are prepared simultaneously during the production process, and then particles of various sizes are sieved. In general, although companies should check the composition of the powders, which might be tested qualified, they might not check the compositions of powders with different sizes. The activity of the very fine Fe powders tended to be higher than the coarse powders, and the specific surface area tended to be larger as well. Therefore, the relative contents of C and N of fine Fe powders, were higher than those of coarser Fe powders. It is important to note that powder production companies should be concerned about this potential engineering issue.
However, the appearance of austenite at RT, found in the Ni-free and Mn-free steels, still seemed somewhat unusual. To further explain the dual-phase of the sintered specimens at RT, the following two exploratory experiments were carried out, including the effects of the cooling rate and the absence of Y
4Zr
3O
12 or W. Firstly, in order to simulate the maximum temperature during the sintering process, the specimens were heated to 1100 °C and held for 30 min, in air, and then cooled down to RT with different cooling rates. The cooling conditions are listed in
Table 5. The oxide layers on the surface of the specimens after heat treatment were ground with SiC abrasive paper and then polished for XRD analysis, and the results are shown in
Figure 10. Based on the cooling rate, the specimens were ranked as H2 < H1 < H3 < H4. Only specimen H2, with the lowest cooling rate, was fully ferritic at RT, while the other three specimens were all α-γ dual-phase. This revealed that the phase structure was related to the cooling rate. On the other hand, the effect of some elements on austenite stability also needed to be further investigated. Fe–13.5Cr–2W–0.6C–0.45N (named as M1, without Y
4Zr
3O
12) and Fe–13.5Cr–0.6C–0.45N (named as M2, without Y
4Zr
3O
12 and W), were prepared with the same MA and SPS conditions as sample A4, to study the effect of Y
4Zr
3O
12 and W on austenite stability, during the cooling stage. The XRD patterns of specimens M1 and M2 are shown in
Figure 11. Combined with the results in
Figure 9, it can be concluded that the phase structure at RT was hardly affected by the absence of Y
4Zr
3O
12. However, the proportion of α-Fe became significantly higher with the absence of W, which demonstrates the stabilizing effect of W on the austenite. The mechanism of W’s effect on the phase transformation in Ni-free and Mn-free Fe–Cr steels containing high contents of C and N, is not clear yet.
The stabilizing effect of W on austenite was surprising, since W is a ferrite stabilizer, thermodynamically. Similarly, Nb is also known to be a ferrite stabilizer in thermodynamics, and will raise the A
e3 temperature [
58]. However, kinetically, the addition of a small amount of Nb will greatly delay the γ→α phase transformation. There is a large misfit between the Nb atoms and Fe lattice [
59,
60], thus Nb tends to segregate to the grain boundaries and reduce the energy of grain boundaries. There is also a drag effect of Nb on phase interface migration in the phase interfaces. In addition, the strong interaction between Nb and C will inhibit the diffusion of C. The inhibitory effect of Nb on γ→α transformation is mainly attributed to the solute drag effect [
58,
61,
62]. The effect of W on γ→α transformation is rarely analyzed from the perspective of kinetics. There is a strong interaction between W and C as well. The explanation of austenite stabilization will be further studied from the perspective of kinetics in the future. The phase transformation during the process of SPS can be described as follows. (i) Heating stage: the ferrite transformed to austenite gradually, and there was no ferrite or a fraction of untransformed ferrite at the highest temperature [
63,
64]. (ii) Cooling stage: the austenite at high temperature began to change back to ferrite. However, when the cooling rate was fast enough, there was not enough time for austenite to transform completely, resulting in its partial retention at RT. Pure ferrite could be obtained with a sufficiently low cooling rate. In other words, the sintered samples should be completely ferritic at RT, the unexpected contamination of C and N made these samples become dual-phase. In the meantime, the excessive cooling rate and the presence of W, also contributed to the results described above.
Microstructural observations indicated a significant influence of the sintering temperature combinations on the microstructure of the steels.
Figure 12 shows SEM observation micrographs of the sintered specimens. Considering the particle size of the very fine as-milled powders, and the as-milled powders’ morphology, shown in
Figure 7, the as-sintered specimens preserved some morphological and dimensional features of the alloyed powders, to a certain extent.
Figure 12a shows several large and dark areas, which are identified as Cr-rich phases in
Figure 13a. A relatively small number of Cr-rich phases can also be seen in
Figure 12b,d, while the Cr-rich phase is barely visible in
Figure 12c, due to the higher temperature at platform 2. During the SPS process, the following mechanisms began to operate: surface activation, powder and element diffusions, surface melting, the formation of necks between powders, and plastic flow, which combined to influence the microstructural evolution in the sintered alloys [
65,
66].
Figure 12e,f show the precipitates of specimen A4, distributed both at grain boundaries and within grains. A high number density of nano-scale particles, ranging from ~20 nm to ~400 nm, were arranged in circles and chains, which are marked by a white rectangle and an irregular closed curve, respectively, indicating that they interacted with the extended defects, such as grain boundaries and dislocations. The EDS point scanning mode was used to measure the elemental content of the matrix, and the results are shown in
Figure 13. The element compositions of specimens A3 and A4 were closer to the nominal compositions, and more stable due to the higher sintering temperatures, which facilitated the “flow” of the elements. Since specimen A1 contained more Cr-rich phases, the Cr content in the matrix was lower than expected. The distribution of Cr in specimen A2 was not quite even, due to the lower temperature at platform 1. As the measurement of light elements with EDS is not accurate, EPMA, equipped with a wavelength dispersive spectrometer, was adopted for further analysis.
Taking specimen A4 as an example, a typical area was chosen for EPMA mapping analysis, and the results are displayed in
Figure 14. The distribution characteristics of the elements within the matrix and at grain boundaries, were different. As specimen A4 was well prepared by SPS, all the elements were evenly distributed within the matrix, except for C. C was evenly distributed regionally within the matrix where ferritic and austenitic phases were both present, while the solubility of C in the ferrite and austenite was different. The evenly distributed W, also impeded the diffusion of C in the Fe matrix. Cr was continuously distributed at grain boundaries. Cr was particularly locally enriched at the triple junctions, which suggests the formation of Cr-rich precipitates. The Cr-rich precipitates would result in Cr depletion at grain boundaries, which would decrease the stability of the grain boundaries. The Cr-rich regions correspond to the dark “hole-like” areas in the morphology image (
Figure 14a). Compared with the ordinary grain boundaries, new phase nucleation and void creation are more likely to occur at triple junctions; and triple junctions can be favorable channels for the diffusion of solute atoms, due to the potential for larger space, a looser structure, more severe stress concentration, a more chaotic atomic arrangement, and more vacancies, dislocations, and other defects [
67,
68,
69,
70,
71]. As a result, Cr-rich precipitates, and other small-scale compounds, were more likely to precipitate at triple junctions. Moreover, many precipitates were also present at the ordinary grain boundaries. The high dislocation density at grain boundaries provided energy for the nucleation of precipitates, while the high grain boundary density, resulting from the fine grains, also provided more sites for the nucleation of precipitates. W was slightly enriched at some of the triple junctions as well. The distribution of C at the triple junctions was partly overlapped with that of Cr and W, which indicated the formation of carbides such as M
23C
6 (M = Fe, Cr, W), M
7C
3 (M = Fe, Cr, W), and WC. There was also a certain enrichment behavior of N at the triple junctions, which correlated with that of C, implying that compounds containing C and N might have formed.
Y, Zr, and O were highly coincident, indicating that Y
4Zr
3O
12 particles were stable in composition during the preparation process. Liu et al. demonstrated that pre-prepared Y
2Ti
2O
7 powders became amorphous during the MA process, while the powders remained stable in composition and did not dissolve in the Fe matrix [
38]. Y
4Zr
3O
12 with a higher binding energy than Y
2Ti
2O
7 would also remain stable in composition and not dissolve in the Fe matrix during the MA process. A typical Y
4Zr
3O
12 particle, with a diameter of 8 nm, was observed in Fe–15Cr–2W–0.35Ti–0.6Y
4Zr
3O
12 steel, which was prepared by a similar sol-gel method, as used in this study, MA, and HIP [
39]. As a result, Y
4Zr
3O
12 would remain stable during the SPS process. The Y
4Zr
3O
12 particles were continuously distributed at the grain boundaries and dispersed within the matrix, which showed that there were many Y
4Zr
3O
12 particles distributed in the steels. Besides, the distributions of Cr and O were also overlapped in some areas, which revealed the formation of some Cr
2O
3 particles. The MA process was beneficial to the uniform distribution of carbides, nitrides, and carbonitrides. However, the oxides were enriched in some areas, where defects were regionally distributed. The high-density defects impeded the diffusion of elemental O, and promoted the combination of O with oxyphilic elements.
Figure 15 shows the specific compositions of some typical precipitates, distributed at grain boundaries and triple junctions, in specimen A4, which were measured with EPMA. The results of the typical particles show that the precipitates mainly consisted of M
23C
6, M
7C
3, Cr
2O
3, M
2(C,N), WC, etc. The state of Y
4Zr
3O
12 particles distributed at grain boundaries was not clear, they could be present alone at grain boundaries or dissolved in large precipitates. Overall, the point scan results shown in
Figure 15 are consistent with the element distribution results displayed in
Figure 14.
Figure 16 displays the EBSD analysis results of specimens A1–A4, with noise reduction. The unresolved areas were reasonably eliminated, which might include holes, precipitates, or ultrafine grains affected by high lattice distortion. The red areas correspond to austenite, while the blue areas correspond to ferrite. The white lines represent low angle grain boundaries (LAGB, 2–15°), and the black lines represent high angle grain boundaries (HAGB, > 15°). Similar bimodal grain size distributions have been observed for other ODS alloys prepared by the SPS method [
72,
73,
74,
75]. The coarse grains helped to improve the plasticity of the steel, while the fine grains were beneficial to the increase in strength. The orientation imaging maps show that the grain orientations of specimens A1–A4 were essentially random, which is considered to be a typical feature of alloys sintered by SPS [
75,
76]. The average grain sizes of the specimens were 0.48 μm, 0.65 μm, 0.82 μm and 0.64 μm, respectively, which were obtained by counting more than 3000, 2000, 1400, and 2000 grains, respectively. The average sizes of austenitic grains of the specimens were 0.54 μm, 0.73 μm, 0.96 μm, and 0.72 μm, respectively, while the average sizes of ferritic grains were 0.47 μm, 0.54 μm, 0.59 μm, and 0.57 μm, respectively. Combined with the α/γ grain sizes, the inverse pole figure (IPF) maps, and the phase distribution maps for observation, it can be concluded that most of the ferritic grains were smaller than the austenitic grains. With the increase in sintering temperature, the grains grew and thus the grain size increased. The proportion of LAGB for all specimens was much higher than that of HAGB. Specimen A3 showed the highest proportion of LAGB, which was beneficial for the mechanical properties of the steel.
The simulated equilibrium phase compositions of the steels with different compositions, at different temperatures, calculated with the Thermo Calc software, are shown in
Figure 17. The following conclusions can be drawn: (i) the matrix of Fe–13.5Cr–2W was fully ferritic at any temperature, under the ideal conditions; (ii) the presence of O would not affect the phase transformation of the matrix at high temperature, but would result in the formation of oxides; (iii) C and N would lead to α→γ transformation at high temperature; (iv) C, N, and O would lead to the formation of M
23C
6, Cr
2O
3, M
7C
3, M
2(C,N), and other precipitates. It should be noted, that the simulation results were in equilibrium, but the actual preparation process of the steels was a non-equilibrium transformation. A great deal of austenite was not able to transform back to ferrite, due to the rapid cooling rate and the γ-Fe stabilization effect of W. Therefore, the simulation results were approximately consistent with the experimental results, for the phase structure at RT. Additionally, as for the composition of precipitates, the simulation results were also generally consistent with the experimental observations.