Precipitates and Particles Coarsening of 9Cr–1.7W–0.4Mo–Co Ferritic Heat-Resistant Steel after Isothermal Aging

The precipitates obtained by EPE technology from the 9Cr-1.7W-0.4Mo-Co ferritic heat-resistant steel subject to isothermal aging were investigated using SEM, TEM and XRD. The particle size distribution and the coarsening kinetics of M23C6 with duration of isothermal aging were also analyzed with or without consideration of Laves phase. The results show that the isolated dislocations were detected in delta ferrite interior, and the precipitates on delta ferrite and martensite boundaries are obviously larger than other locations. Fe2W-Laves phase can only be found as duration of aging time to 2000 h, and is preferential to form adjacent to M23C6 particles. The small M23C6 particles firstly coarsen, but the large M23C6 are relatively stable during short aging. The total coarsening rate of M23C6 precipitates is 9.75 × 10−28m3s−1, and the coarsening of M23C6 depends on the formation of Laves phase.

Whereas, Laves phase with small particle size also can increase the creep strength by precipitation hardening at early stage of aging/creep [18][19][20] . In addition, compared to M 23 C 6 and Laves phase, MX carbonitrides (i.e., M = metals, X = C/N) exhibit the smallest size, and mainly distribute at subgrain boundaries and within martensite laths 21,22 . Chen et al. 23 showed that martensite laths with the finely dispersed nanosized MX in 9 wt.% Cr ferritic steel would contribute to the improvement of high-temperature mechanical performance. Panait et al. 22 displayed that MX carbonitrides are very stable during thermal and creep exposure of P91 steel at 873 K up to 10 5 h.
According to Zener pinning equation, the pinning effort of the particles is related to particle size and distance between each other 24,25 . Based on this, the pinning forces of MX carbonitrides should be higher than that of M 23 C 6 carbides. Furthermore, Some references 26,27 also proved that MX carbonitrides in 9 wt.% Cr ferritic steel were more advantageous than M 23 C 6 carbides because MX precipitated finely and densely, and did not coarsen easily at high temperatures. Factually, MX carbonitrides effectively strengthen boundaries because they can act as obstacles to migrating boundaries 28 , and are very stable during short aging/creeping. Thus, in the case of without considering the evolution of MX carbonitrides, the coarsening of M 23 C 6 particles and the formation of Laves phase, both as the key factors that will affect the microstructural stability of 9~12 wt.% Cr ferritic creep steels during aging/creep, especially 9 wt.% Cr creep strengthening steel with addition of Co, should be given more enough data to find ways for improving the aging/creep strength of the steels.
The content of delta ferrite in 9~12 wt.% Cr ferritic creep steels can be influenced by the alloy elements as W, Mo, Co, etc., especially as C 29 . The diagram of transformed phase (delta ferrite) as a function of net chromium equivalent (NCE) has been characterized in some references 29,30 . For example, there is no delta ferrite in T91 ferritic steel, but it is up to 6~8% in modified ferritic steel with Co addition and elements adjustment 31,32 . With the situation of high content, delta ferrite plays important roles on solid solution strengthening of alloying elements and precipitation behavior of secondary phases, such as M 23 C 6 and Laves phase. The dissolved alloying elements (such as Cr, C, W, Mo, and so on)contents in delta ferrite matrix are theoretically higher than that in martensite for the tempered samples, which will promote the formation of the precipitates [33][34][35] . Whereas, the solidifications of other alloying elements, such as Co, Mn and Ni, are crucial in hindering the formation of delta ferrite 33,36 . The works in our previous researches also proved that the content Redjaimia et al. 37 showed that the precipitation of M 23 C 6 phase priors to the other precipitates on delta ferrite boundaries, which is always attributed to the high mobility of carbon towards the boundaries enriched in M 23 C 6 forming alloying elements. In addition, accompanying with precipitation of secondary phases, the structure of delta ferrite has been changed, and some dislocations should be found in the interior of delta ferrite, which is rarely reported and explained.
The microstructural evolution and the distribution of alloying elements of the new modified 9Cr-1.7W-0.4Mo-Co ferritic heat-resistant steel during isothermal aging were described in our previous article 35 . This article presents a detailed description of the precipitation behavior in the same aged steel. The powder of the precipitates is obtained by electrolytic phase extraction (EPE) technology, and analyzed using scanning electron microscope (SEM), transmission electron microscopy (TEM) and X-ray diffraction (XRD). M 23 C 6 and Laves phase were identified by the selected area diffraction (SAD). The particle size distribution and the coarsening kinetics of M 23 C 6 precipitates were also analyzed.

Results
Microstructural evolutions on delta ferrite boundaries. TEM micrographs of 9Cr-1.7W-0.4Mo-Co ferritic steel aged for 500 h are displayed in Fig. 1. Martensite consists of laths and delta ferrite can be distinguished, and some M 23 C 6 precipitates with large particles distributed at martensite lath boundaries and the interface of delta ferrite and martensite. In addition, nano-sized MX precipitates were homogeneously distributed in martensite lath, which would impede the migration of high-density dislocations and insist stable microstructure during isothermal aging (seen in Fig. 1b).
As we discussed in our previous works 35,38 , there were many precipitates particles located at the interface of delta ferrite and martensite in ferritic heat-resistant steels. The delta ferrite, enriched in BCC stabilising alloying elements, notably Cr, Mo and W, is a favourable site for some precipitates, such as M 23 C 6 , Laves phase, comparing to austenite and martensite. Therefore, it means that the alloying elements, which are contributed to the formation of precipitates located at the boundaries of delta ferrite grain, are provided by delta ferrite phase. During aging, the migrations of alloying elements from the interior to the grain boundaries of delta ferrite phase lead to the formation of precipitates with a large number, and new dislocations located in the interior of delta ferrite phase might be emerged. Figure 2 shows TEM images of delta ferrite containing in 9Cr-1.7W-0.4Mo-Co ferritic steel after isothermal aging for different time. There are three obvious characters can be distinguished as follows: (i) Some isolated dislocations (as marked by the light arrows in Fig. 2) were detected in the delta ferrite grains during isothermal aging, and the dislocation density kept to be increased with the increase in aging time.
The atoms of dissolved alloying elements migrate from the interior to the boundary of delta ferrite, and form the precipitates located at the grain boundaries with the duration of isothermal aging, which means that many lattice vacancies are remained in delta ferrite grains. (ii) The density of the precipitates located on the interface of martensite and delta ferrite was obviously higher, comparing to which located at the boundaries of martensite laths. This also can be attributed to that there is a relatively high density of dissolved alloying elements in delta ferrite, which is beneficial to the formation of precipitates, and also indicating that the particle size of precipitates at the interface should be larger. (iii) Most of the particles of precipitates grew into martensite laths with the duration of isothermal aging, only partially small particles would inversely intrude into delta ferrite grains.
Compositions of the precipitates. The powder of the precipitates in the aged 9Cr-1.7W-0.4Mo-Co ferritic steel obtained by EPE technology was analyzed using routine X-ray diffraction, as shown in Fig. 3. In the all XRD patterns, the remarkable peaks were clearly visible, and which could be identified as the Cr-rich carbide of an M 23 C 6 type, i.e., Cr 23 C 6 phase. Additionally, the peaks around 2-theta angles 35° and 71° represented an MX type precipitate enriched in Nb (NbC phase), which has also been detected from the XRD patterns. Fe 2 W Laves phase was not identified until the time of isothermal aging was extended to about 2000 h. In our previous research, the increase in W element content in some areas, such as grain boundaries and around M 23 C 6 particles, is beneficial to match the formation of Fe 2 W Laves phase with the duration of isothermal aging 35 . Therefore, the migration and enrichment of W element on the grain boundaries or around the M 23 C 6 particles during aging result in the formation of Fe 2 W phase. The energy disperse spectroscopy (EDS) analyses of alloying elements distributions in the microstructure of the material subjected to isothermal aging for 2000 h are displayed in Fig. 4. Some subgrains can be observed with the duration of aging. The large and small precipitates particles were located on the subgrain boundaries and in the interior of subgrains. It can be seen that the alloying elements distributions among different particles are obviously heterogeneous. The high densities of Cr, C and W alloying elements distribute in the large particles with ellipsoidal and spherical shapes located at the subgrain boundaries, which indicates that the large particles can be identified as M 23 C 6 precipitates enriched in Cr element. Several ellipsoidal particles enriched in W element and depleted in Cr element should be considered as Fe 2 W Laves phase (as shown by yellow arrows in Fig. 4). The content of Fe element in the Fe 2 W Laves phase was lower comparing to the Fe-matrix phase (martensite), which also should be noted. Figure 5a and b represent TEM micrograph and EDS results of the precipitates powder extracted from 9Cr-1.7W-0.4Mo-Co ferritic steel after 100 h aged. The ellipsoidally shaped particles with 200 nm and 65 nm in long axis and short axis, respectively, and a plate-shaped particle with 400 nm in length, can be obviously distinguished. The chemical compositions of these particles are similar to each other, containing high Cr and Fe contents with traces of W and Mo as shown in Fig. 5b. The selected area diffraction (SAD) patterns of the particles marked in Fig. 5a are displayed in Fig. 5c and d, identifying that both the particles with the two shapes are face-centered cubic structure (FCC), Cr 23 C 6 precipitates. For the case of 9Cr-1Mo ferritic steel, it has already been established that the composition, size, and distribution of M 23 C 6 precipitates are mainly dependent on ageing time and temperature: a longer time and higher ageing temperature are considered to induce equilibrium and hence a higher volume fraction of the precipitates 39 . Whereas, a long duration such as after 1000 h may mark the onset of carbide particle coarsening and the appearance of new brittle phases such as Laves and Z-phase 11,20 . Figure 6 shows TEM analysis of Laves phase formed in the 9Cr-1.7W-0.4Mo-Co ferritic steel after isothermal aging for 2000 h. It should be pointed out that there is no Laves phase being found in other samples, such as aging for 1000 h, which means that the microstructure of the ferritic steel is still very stable during long-termed aging. The particle size of laves phase is measured to be ~ 200 nm, which is consistent with the literatures 40,41 . Fe 2 W Laves phase can be identified using SAD and EDS methods. The EDS result indicates that Fe and W are dominant in the Laves phase with traces of Mo and Cr. One mechanism is that the nucleation of Laves phase precipitates at the Cr-depleted regions with relatively high W and Mo contents adjacent to M 23 C 6 particles. The M 23 C 6 particles are identified as a composition of (Fe, Cr, Mo, W) 23 C 6 as we discussed above. Therefore, the alloying elements distributions in M 23 C 6 particles with the durations of isothermal aging are analyzed by EDS, which can reflect the rearrangement of alloying elements and indicate the formed location of Laves phase, and the results are displayed in Table 1. The content of Cr increases from 53.87 wt.% to 63.31 wt.% in M 23 C 6 precipitates with the increase in aging time from 100 h to 2000 h, W content inversely decreases from 19.25 wt.% to 11.86 wt.%, both of them show a rearrangement of alloying elements during isothermal aging. It means that the dissolved W in M 23 C 6 phase will precipitate, and the precipitated W elements in the vicinity of M 23 C 6 phase are beneficial to the formation of Fe 2 W Laves phase, and thus Laves phase is preferential to precipitate at the locations adjacent to M 23 C 6 particles. In other words, for M 23 C 6 phase during aging, the chemical compositions will be changed due to the formation of Laves phase, except the coarsening of the particle size during aging as reported in the literatures 42, 43 . Particle size of M 23 C 6 . Quantification of the particle size distributions of M 23 C 6 phase was carried out based on TEM images of the EPE samples, and thus it is difficult to distinguish the location of M 23 C 6 phase. To obtain sufficiently accurate statistics for the particle size distributions, the number of particles counted using computer-aided software was above 300. Employing Banerjee' s method 44 , it is assumed that A represents the particle area, and the equivalent area diameter (EQAD), d A , can be given as: (1)

A
The particle diameter was obtained by using the measured particle area in Eq. (1). Hence, the obtained particle size distributions as a function of particle diameter are displayed in Fig. 7.
The changes in particle sizes of M 23 C 6 precipitates can be considered to obey the Gaussian distribution in total. However, the distributions of particle sizes with ranges of 50~180 nm and 50~220 nm for the samples subjected to aging for 100 h and 500 h, respectively, are more uniform, comparing to the samples with a longer duration of ageing time (1000 h and 2000 h). With the increase in aging time to 1000 h, most of particle sizes increased to a range of 100~240 nm, indicating that M 23 C 6 precipitates with small size are preferential to coarsen firstly, and the large M 23 C 6 are relatively stable during short aging. The reason can be attributed to the distributions of alloying elements adjacent to M 23 C 6 particles. The alloying elements, such as Cr and C around small M 23 C 6 particles, can be considered to be homogeneous, which means that there are enough elements to migrate into M 23 C 6 particles and promote the coarsening of precipitates. For large M 23 C 6 particles, whereas, the existence of Cr depleted area will partly decrease the coarsening progress of M 23 C 6 phase.

Discussion
Theoretically, the alloying elements, formed the precipitates located on the interface of martensite and delta ferrite, are from the interior of delta ferrite, and thus the precipitate particles should be inclined to grow into delta   ferrite. In fact, the original interface of martensite and delta ferrite has migrated into the delta ferrite with the diffusion of atoms, and advances ahead of the precipitates particles, which will leave the precipitates as isolated particles in the martensite. The migration of delta ferrite interface depends not only on the chemical composition, but also to some extent on the temperature which affect the diffusion rate of atoms as reported in the literatures [45][46][47] . When the aging temperature is to be above 873 K, it is considered that the migration of δ/α' interface is obviously ahead of the growth of the precipitates 45 . Hence, the precipitates were retained in martensite due to the migration of δ/α' interface during aging at 973 K which can be seen in Fig. 2.
In our previous works, the alloying elements comparisons between different phases have been accomplished, and the results showed that the contents of Fe in martensite, delta ferrite and precipitates were about 85.9 wt.%, 89.6 wt.% and 36.5 wt.%, respectively 35 . The highest content of Fe in matrix phase is easy to be understood. Besides, the MX-type precipitates enriched in Nb located on the subgrain boundaries and in the interior of subgrains also can be recognized, respectively (as shown by red arrows in Fig. 4). During aging/creep, Laves phase precipitated on grain or subgrain (martensite lath) boundaries has been proved that it will reduce the efficiency of grain boundary as a sink of dislocations, and suppresses the recovery of dislocations at grain boundaries 41,48 . Nevertheless, the small particle size of Laves phase improves the pinning force of the migrations of dislocations and grain boundaries, and increases the stability of microstructure, which means that the role of Laves phase with this situation is similar to M 23 C 6 and MX 49,50 . As seen in Fig. 4, the average particle size of Laves phase is about  200~500 nm, which is same to M 23 C 6 particles, indicating that the microstructure of the materials aged for 2000 h should be still stable due to precipitation strengthening of the precipitates. M 23 C 6 particle coarsening as a function of aging time has been studied in ferritic heat-resistant steels and other alloys 42, 51-53 . The particle coarsening rate for M 23 C 6 phase so far follows the equations based on the classical Lifshitz-Slyozov-Wagner (LSW) theory which assumes supersaturated solid solution, misfit-free spherical particles, and one-rate controlling solute 51,54,55 . According to LSW theory, the coarsening kinetics of the particles for volume diffusion-controlled growth processes follows the law: where d and d 0 correspond to the average particle diameter at aging time t and the initial average particle diameter, respectively, and k is the coarsening rate constant with a formula as 56 : with C e as the equilibrium concentration of atom in the matrix, V m as the molar volume of the precipitates, D as the diffusion coefficient of atom in the matrix, Ω as the interfacial energy of precipitates/matrix, R as the gas constant, and T as the aging temperature in K. The average particle diameters of M 23 C 6 phase aged at 100 h, 500 h, 1000 h and 2000 h are 143 nm, 162 nm, 196 nm and 212 nm as obtained from Fig. 7, respectively. Hence, the cubic of average particle diameter as a function of aging time in second (s) is shown in Fig. 8. The Laves phase is consisted with the same alloying elements, such as Fe, W, Mo and Cr, as M 23 C 6 phase except for C. According to Gustafson 57 , the M 23 C 6 coarsening should be higher if the Laves phase is not allowed to form, comparing to the situation that when the amount of Laves phase has an equilibrium amount. The other researchers inversely showed that the formation of Laves phase does not affect much the coarsening of M 23 C 6 58 . There is no Laves phase can be found until increasing aging time to 2000 h in our work, it means that the influence of Laves phase on the M 23 C 6 coarsening can be ignored under shorter aging time. A linear relationship between the cubic of average particle diameter and the aging time was determined for aging at 973 K, which approximately satisfies the predictions of LSW theory. The thus coarsening rate constant of M 23 C 6 phase during aging was determined as k 0 = 9.75 × 10 −28 m 3 s −1 from the slope (green fitted line). Whereas, the R-square value is bout 0.9081, indicating that the fitting result is not accurate. In fact, Laves phase is found in the sample aging for 2000 h, the initial nucleation/formation stage of Laves phase is still unknown. In addressing, two stages of M 23 C 6 coarsening without and with the effect of Laves Phase are considered. Therefore, the first stage is under the assumption that the Laves phase was not allowed to form at all, i.e., from the aging beginning to about 1000 h, and the coarsening rate of M 23 C 6 is about k 1 = 1.21 × 10 −27 m 3 s −1 with a higher R-square of 0.9823. The second stage in our work should be considered to be from 1000 h to 2000 h with the coarsening rate of k 2 = 4.52 × 10 −28 m 3 s −1 . The results indicate that the formation of Laves phase will affect the coarsening of M 23 C 6 phase.
The coarsening rates of M 23 C 6 particle at the range from 873 K to 1023 K deduced from data in references and this work are listed in Table 2. As we know, the coarsening rate of M 23 C 6 particle is mainly dependent on aging/ creep temperature, and theoretically increases with the increase in temperature for the same materials, which is obviously identified in Table 2. The lowest coarsening rate is only 0.088 × 10 −29 m 3 s −1 in modified 9%Cr Steel with 1% Co at 873 K. Considering the existence of Laves phase at exposing temperature 973 K, the M 23 C 6 coarsening is relatively low in 9Cr-1.7W-0.4Mo-Co ferritic steel, comparing to 12%Cr steel and H13 steel, indicating that M 23 C 6 phase in our materials is stable at 973 K.

Methods
Material used in the present work is a new modified ferritic heat-resistant steel, 9Cr-1.7W-0.4Mo-Co steel subject to being hot-rolled to 12 mm thick at 1373 K, and then tempered at 1023 K for 2 h. The chemical composition of the steel was determined by Inductive Coupled Plasma-Optical Emission Spectrometry (ICP-OES), and the thus obtained results are listed in Table 3. Further, the tempered steel was cut to several small pieces which were isothermally aged at 973 K holding for 100 h, 500 h, 1000 h and 2000 h, respectively. The aging tests were conducted in a vacuum tube furnace with a temperature tolerance of ± 5 °C, and a vacuum degree of −0.09 MPa in the furnace was controlled to avoid oxidation.
Characterization of microstructure was realized by SEM (ZEISS, SUPRA-55) and TEM (JEOL, JEM-2010) at an accelerating voltage of 15 kV and 200 kV, respectively. The specimens for SEM were prepared by mechanical polishing, and then etching in a solution composed by iron trichloride (5 g), hydrochloric acid (20 mL) and distilled water (100 mL). Thin foils with 130~150 micrometers for TEM were prepared by a twin-jet electropolisher using a solution of 10% perchloric acid and 90% ethanol.
Electrolytic phase extraction (EPE) technology was employed to analyze the composition of the precipitates. Using EPE, the matrix phase was etched and dissolved in the solution, and only the precipitate particles were extracted. EPE of the samples at a current density of 120 mA/cm 2 operated in a solution of 25% hydrochloric acid and 75% alcohol, and then the precipitates powder extracted from the samples were collected on the bottom of the flask after washing 3~5 times with alcohol. The compositions of the powder were identified by XRD analysis using Cu Kα radiation at room temperature, with a step size of 0.02° and observed by TEM.

Conclusions
The precipitates in the 9Cr-1.7W-0.4Mo-Co ferritic heat-resistant steel subjected to isothermal aging at 973 K were investigated using SEM, TEM and XRD. The M 23 C 6 particle coarsening as a function of aging time is also described. The conclusions are summarized as follows: (1) The isolated dislocations were detected in delta ferrite interior, and increase with the increase in aging time. Both the density and the particle size of the precipitates on δ/α' boundaries are obviously larger than other locations, and the migration of δ/α' boundaries leads to the retained precipitates in martensite. (2) Cr 23 C 6 , NbC and Fe 2 W precipitates are recognized based on XRD, and Fe 2 W-Laves phase can only be found as duration of aging time to 2000 h. The precipitated W elements in the vicinity of M 23 C 6 phase are beneficial to the formation of Fe 2 W Laves phase, which shows that Laves phase is preferential to form at the locations adjacent to M 23 C 6 particles. (3) M 23 C 6 precipitates with small particle size are preferential to coarsen firstly, and the large M 23 C 6 are relatively stable during short aging.   Table 3. Chemical compositions of the modified high Cr ferritic heat-resistant steel (mass%).
(4) The total coarsening rate of M 23 C 6 precipitates is 9.75 × 10 −28 m 3 s −1 , but it changes to 1.21 × 10 −27 m 3 s −1 without the formation of Laves phase, and 4.52 × 10 −28 m 3 s −1 with consideration of Laves phase, respectively, indicating that the coarsening of M 23 C 6 precipitates depends on the formation of Laves phase.