2024 Volume 64 Issue 2 Pages 295-302
In order to understand microstructural changes in 9Cr-1Mo-V-Nb weld metal after long term use, microstructure and precipitates distribution before and after aging at 1013 K were investigated. In the weld metal, regions with coarse or fine prior austenite grains were observed due to thermal cycle during welding. In the coarse grain region, precipitate particles inferred to M23C6 were densely located on grain boundaries, however, in the fine grain regions, they were sparsely observed not only on grain boundaries but also inside grains. Post weld heat treatment (1013 K/7.7 h) followed by aging (1013 K/100 h) led to ferrite grains formation in the fine grain region. EBSD analysis implied that dislocation density in ferrite grains was low. After the aging, mean diameter of particles became coarser and interparticle spacing became sparser in the fine grain region than in the coarse grain region. On the other hand, dislocation density calculated by hardness in martensite structure was almost no deference between these regions before and after the aging. Therefore, it was suggested that ferrite grains were formed because pinning energy by precipitate particles locally reduced in the fine grain region.
Thermal power generation provides more than half of the electrical power supplied in Japan1) and plays a crucially important role in its adjustment as a back-up power source.2,3) Although power supplied by renewable energy generation is recently increasing to reduce CO2 emissions, fluctuations in power output depending on weather conditions are an important shortcoming. For stable power supply, thermal power generation is necessary.4) Accordingly, existing Ultra-Super-Critical (USC) plants are expected to be operated in the future ahead. Because the operating times of some USC plants will be reached at 200000 to 300000 h, it is important to elucidate the microstructural changes of materials constituting USC plants during long-term use.
For main steam pipes and heat transfer tubes in USC plants, 9Cr-1Mo-V-Nb steel is commonly used. Creep strength of the steel weld joints is well known to be degraded compared to that of the base metal. Moreover, the creep fracture mode of the weld joint often exhibits Type IV failure (fractured at fine grain region in heat affected zone (HAZ)).5,6,7) For this reason, earlier studies examining 9Cr-1Mo-V-Nb steel weld joints have specifically emphasized long-term creep strength,8,9,10,11) remaining creep life estimation12,13) and microstructural changes at HAZ.14,15,16) Reportedly, Type I failure (fractured at the weld metal) occurred after long-term creep tests exceeding 10000 h,17,18,19) or in the case of test materials after long-term use.20,21) Therefore, expanding our knowledge of microstructural changes in the weld metal is necessary. Furthermore, it has been reported that ferrite grains were observed in the weld metal after long-term use.22,23) Because most of the initial microstructure in 9Cr-1Mo-V-Nb steel weld metal is martensite structure,24) ferrite would be formed through microstructural changes during long-term use. Unfortunately, few researchers have reported detailed examinations of microstructural changes of the weld metal and ferrite formation mechanisms have not been clarified. Reportedly, precipitates play an important role in recovery of martensite structure and microstructural changes.25,26) Therefore, considering the relation between microstructure and precipitate distribution is necessary to understand microstructural changes that take place in the weld metal.
This study was conducted to clarify microstructural changes in 9Cr-1Mo-V-Nb steel weld metal after long-term use. For that purpose, we examined changes of microstructure and precipitate distributions in the weld metal after high-temperature and short-time aging instead of after long-term use.
The test material was a weld joint made using 9Cr-1Mo-V-Nb steel pipe. Chemical compositions of the base metal and filler metal are presented in Table 1. The weld joint was fabricated using TIG welding, which was performed with one pass per layer. The as-welded joint macrostructure is shown in Fig. 1, in which welding defects are not visible. The vertical direction in Fig. 1 corresponds to the thickness direction of the pipe. The upper and lower parts respectively correspond to the outer and inner surface sides of the pipe. Points vertically observed in Fig. 1 are indentations of micro-Vickers hardness testing described later. For this study, weld metal except for regions near the first and final layer was observed to investigate microstructures in weld metal where welding was performed stably under certain conditions. The weld joint was annealed at 1013 K/7.7 h as post-weld-heat-treatment (PWHT). Subsequently, 1013 K/100 h aging heat treatment was applied for a part of the weld joint. Assuming the value of constant C for tempering parameter as 20,27,28) 1013 K/100 h corresponds to 873 K/340000 h.
| C | Si | Mn | P | S | Cu | Ni | Cr | Mo | V | Sol-Al | Nb | N | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Base metal | 0.10 | 0.29 | 0.48 | 0.010 | 0.001 | 0.03 | 0.15 | 8.85 | 0.98 | 0.21 | 0.007 | 0.08 | 0.058 |
| Filler metal | 0.08 | 0.16 | 0.99 | 0.007 | 0.004 | 0.02 | 0.66 | 9.14 | 0.89 | 0.18 | 0.05 | 0.02 |
As-welded and aged samples were cut and embedded in thermosetting resin. The sample preparation for microstructural observation was conducted as described below. Wet polishing was done using emery paper (~#2000) and polishing cloth with 0.3 μm diamond slurry, followed by colloidal silica. Then, etching by Vilella reagent (mixture of picric acid, hydrochloric acid and alcohol) was conducted. Microstructural observation was conducted using optical microscopy (OM), scanning electron microscopy (SEM) and electron backscattered diffraction (EBSD) method. The step size of EBSD measurement was 0.2 μm. OIM Analysis (ver. 7.3.0) manufactured by TSL was used for EBSD data analysis. To evaluate interparticle spacing of precipitates, observation with secondary electron (SE) imaging was conducted. Two SE images (10000× magnification) were continuously observed so that the observation area was about 200 μm2, and the combined image was defined as one observation field. The average value evaluated by observation of four fields was used as the representative value of interparticle spacing. Micro-Vickers hardness testing with 0.2 kgf (1.96 N) load was conducted to obtain hardness profiles of the weld metal.
Figure 2 shows a hardness profile of the as-welded metal. Hardness was measured from top to bottom in Fig. 1: from the outer to inner surface side of the pipe. Figure 2 indicates that the hardness was changed locally in the weld metal, and that the local maximum and local minimum appeared at certain intervals. The maximum and minimum hardness were approximately 420 HV and 300 HV. Figure 3(a) shows an OM image of the as-welded metal. As in the case with Fig. 1, the vertical direction in Fig. 3(a) corresponds to the thickness direction of the pipe. Magnified images of the frames presented in Fig. 3(a) are depicted in Figs. 3(b) and 3(c). The fields of view in Figs. 3(b) and 3(c) were regions with low and high hardness. Martensite structures were observed in Figs. 3(b) and 3(c). Prior austenite grains observed in Fig. 3(b) were coarse. In contrast, those in Fig. 3(c) were fine. Consequently, regions with coarse or fine microstructure were observed at certain intervals, as confirmed in Fig. 3(a). In addition, the interval between regions with coarse and fine microstructure approximately coincided with that between local minimum and local maximum of hardness. As described later, it was inferred that regions with coarse microstructure correspond to so-called Sub-critically HAZ (SCHAZ)29) and that regions with fine microstructure were Fine Grain HAZ (FGHAZ).29) Hereinafter, the former is defined as the coarse grain region and the latter as the fine grain region.
Figures 4(a) and 4(b) show SE images of the coarse and the fine grain regions. In the coarse grain region, precipitates were located on grain boundaries with high density (Fig. 4(a)). In the fine grain region, however, precipitates were observed not only on grain boundaries but also inside grains as indicated by an arrow (Fig. 4(b)). The volume fraction of precipitates observed in the fine grain region was qualitatively lower than that in the coarse grain region. Although phase identification was not conducted for this study, most precipitates observed in Fig. 4 were inferred as M23C6, judging from the volume fraction, particle size and precipitation site.24)
Based on the results presented above, a schematic illustration of microstructure of the as-welded metal is exhibited in Fig. 5. The test material in this study was welded with one pass per layer. Because each layer is heated by welding of the next or subsequent layers, microstructure and precipitate distributions are locally different in the weld metal. Consequently, regions among N layers subjected to temperatures below Ac1 by welding of N+1 or subsequent layers would be the coarse grain region. The microstructure in this region is equivalent to SCHAZ29) because the hardness was low and precipitates were observed densely on grain boundaries (Fig. 4(a)). In contrast, regions among N layers that are short-time heated to temperature above Ac1 by welding of N+1 or subsequent layers would be the fine grain region. In this region, hardness was high, with sparser precipitates than in the former region. Furthermore, the microstructural observation presented in Fig. 4(b) shows that some precipitates were in line inside grains. It is possible to consider that these microstructures are formed as described hereinafter. During short-time heating to temperature above Ac1 by welding, martensite is transformed reversely to austenite. Insufficient grain growth of austenite because of short-time heating occurred. Subsequently, austenite is transformed again into a martensite structure during cooling. Precipitates would be dissolved in matrix when the material is heated for briefly to temperature above Ac1 during the welding thermal cycle. Because of short-time holding at high temperatures, however, un-dissolved precipitates near original grain boundary would remain,15) and solute atoms derived from dissolved precipitates would be insufficiently diffused.26) The solute atoms are consumed by re-precipitation or growth of un-dissolved precipitates. As a result, the microstructure becomes fine martensite structure where some precipitates are located inside the present grains and in line along the original grain boundaries before reverse transformation. Based on the considerations presented above, the microstructure in the fine grain region is equivalent to FGHAZ.29) Strictly speaking, Coarse Grain HAZ (CGHAZ)29) and Inter-critically HAZ (ICHAZ)29) are expected to exist, but distinguishing these regions clearly in the test material was difficult. In the following sections, therefore, we particularly examine microstructural changes after aging at 1013 K in the coarse and the fine grain regions which are respectively equivalent to SCHAZ and FGHAZ.
Figures 6(a) and 6(c) respectively show OM images in the coarse and the fine grain regions after PWHT. Martensite structures were observed in both regions. Average hardness measured three times in each region is shown in Fig. 6. Hardness in the coarse and the fine grain regions were 229 HV and 234 HV, respectively. No significant difference was found between the two. It was implied that the local hardness change in the as-welded metal disappeared and that the hardness in the weld metal was approximately uniform after PWHT. Figures 6(b) and 6(d) show OM images in the coarse and the fine grain regions after 100 h aging. Microstructures observed in the former region were not significantly changed after 100 h aging. In the latter region, however, microstructure with different grain size compared to the surroundings has been formed, as indicated by an arrow. Hardness in the coarse and the fine grain regions after aging were 201 HV and 195 HV, respectively. Quantities of hardness reductions after aging in these regions were almost identical.
Then EBSD analysis was performed to investigate microstructural changes in the fine grain region after aging. Figures 7(a) and 7(b) show IPF and KAM maps obtained by EBSD, respectively. It was observed that coarse grains with low KAM were surrounded by fine grains with high KAM. Actually, KAM has been reported as a parameter that indicates the orientation difference between adjacent pixels and which is correlated with dislocation density.30) Because lath boundaries, low-angle grain boundaries, densely exist within martensite structure and the dislocation density of martensite is high, the high KAM regions would represent martensite structure.7,26) As shown in Fig. 7(b), on the other hand, coarse grains with low KAM suggest that ferrite with low dislocation density was formed and grown during 100 h aging.
Figures 8(a)–8(d) show SE images in the coarse and the fine grain regions after PWHT and 100 h aging. Precipitates were distributed almost uniformly in both regions after PWHT (Figs. 8(a) and 8(c)). Coarsened precipitates were observed after 100 h aging (Figs. 8(b) and 8(d)). Particularly, precipitates observed in the fine grain region were qualitatively coarser and interparticle spacing was larger than in the coarse grain region. As presented in Fig. 8(d), coarse precipitates of approximately 1 μm were observed inside ferrite grains and martensite structure. Precipitates with similar size were also observed in other fields of the fine grain region after 100 h aging. Because MX in addition to M23C6 is precipitated in 9Cr-1Mo-V-Nb steel, the possibility exists that coarsened MX were observed in Figs. 8(b) and 8(d). However, MX is generally smaller than M23C6.24,32) Its volume fraction and coarsening rate are both about one-tenth of those of M23C6.32) Although Laves phase and Z phase might be precipitated during aging, it is expected from previous research that these phases are not precipitated during aging at 1013 K/100 h.33,34,35) Therefore, most precipitates observed in Figs. 8(a)–8(d) were likely to be M23C6.
Figures 9(a)–9(d) show histograms indicating precipitate diameter in the coarse and the fine grain regions after PWHT and 100 h aging. Histograms were obtained by evaluating diameters of all precipitate particles (over 400 in total) observed in each observation field. In Fig. 9, all histograms showed asymmetrical distributions around each mode. As shown in Figs. 9(a) and 9(c), no significant difference was found between the two histograms in the coarse and fine grain regions after PWHT. However, as shown in Figs. 9(b) and 9(d), the proportions of coarse precipitate particles in the fine grain region after 100 h aging were higher than those in the coarse grain region.
Next, we discuss the reason why ferrite was formed in the fine grain region after 100 h aging. Tsuchiyama et al. reported that ferrite was observed in ultra-low-carbon steel with martensite structure after aging.36) They explained the conditions for ferrite formation based on the relation between dislocation density and interparticle spacing, assuming that such microstructural changes would be recrystallization due to the bulging mechanism.36) In this study, microstructural changes of ferrite formation could not be observed directly. Therefore, it remains unclear whether the formation mechanism was the bulging mechanism or nucleation and growth of the recrystallized grain. However, the dislocation density inside ferrite grains was low, as shown in Figs. 7(a) and 7(b), suggesting that high dislocation density contained in martensite would be the driving force for microstructural change in either mechanism.12) However, from Figs. 8(d) and 9(d), the proportion of coarse precipitate particles was high in the region where ferrite was formed. In light of the results presented above, it can be inferred that ferrite grains were formed and grew in the regions where the pinning energy by precipitates was locally reduced.12)
Based on the report by Tsuchiyama et al.,36) we focused on dislocation density ρ and interparticle spacing λ in martensite structure. Assuming recrystallization attributable to the bulging mechanism,36) we investigate whether such microstructural change can be explained by the relation between ρ and λ.
Dislocation density ρ of martensite structure can be estimated by its hardness HV.36)
| (1) |
where HV max and ρ max represent the hardness and dislocation density of as-quenched martensite, respectively. The relations between hardness and dislocation density in 9Cr-1Mo-V-Nb steel are reported from previous studies.31,37,38) Pešička et al.37) reported the hardness and dislocation density of as-quenched martensite in steel containing 0.075% C. Because C content of the steel which they used resembles that of the weld metal in this study, we use values reported by Pešička et al. (400 HV, 4.2 × 1014 m−2).37) HVmin and ρ min represent the hardness and dislocation density of the fully annealed steel, respectively. We use values reported from the previous study (120 HV, 1012 m−2).36) Although some concern exists that the hardness in the fine grain region after 100 h aging (195 HV) might be affected by ferrite grains, this value is sufficiently high compared to that of ferrite observed in 9Cr-1Mo-V-Nb steel weld metal after long-term use (140 HV).23) Therefore, hardness measured in this study would reflect the martensite structure hardness. It seems reasonable to estimate the dislocation density using Eq. (1). Because of asymmetrical distributions around the modes (Figs. 9(a)–9(d)), interparticle spacing λ is estimated using the following equation.27,39,40)
| (2) |
where d represents the average diameter of precipitates, d2bar and d3bar respectively denote the average values of the square and cube of average diameter, and f stands for the volume fraction of precipitates.
Figure 10 shows the relation between dislocation density ρ and interparticle spacing λ in the coarse and the fine grain regions. The higher the dislocation density or the larger interparticle spacing becomes, the more easily ferrite grains are formed. In Fig. 10, although the relation between dislocation density and interparticle spacing was similar in the coarse grain and the fine grain regions after PWHT, it was different between the two after 100 h aging. This difference was attributable to the variation in interparticle spacing and no significant difference was found in the dislocation density. The error bars in Fig. 10 represent the standard deviations of the interparticle spacing or dislocation density between observation fields. Particularly, the large standard deviation of interparticle spacing in the fine grain region after 100 h aging suggests that the microstructure with larger interparticle spacing was locally observed. Therefore, assuming that the ferrite formation occurred through recrystallization due to the bulging mechanism,36) it can be explained qualitatively that ferrite formation in the fine grain region resulted from local reduction of the pinning energy. Clarifying the formation process of ferrite grains in 9Cr-1Mo-V-Nb steel weld metal is an issue left to future research.
Finally, we discuss the reason why interparticle spacing in the fine grain region became sparse after 100 h aging. One factor is an effect of grain boundaries. As shown in Fig. 3(c), fine martensite structure was observed in the fine grain region. This observation implies that grain boundaries, a kind of fast diffusion path, were present at high density. In other words, it seems possible to infer that atomic diffusion would be encouraged in the fine grain region during aging, resulting in the promotion of precipitate coarsening.41) In addition, effects of the initial distribution of precipitates are considered.42) As portrayed in Fig. 4(b), in the fine grain region of the as-weld metal, some precipitates were in line inside grains because of the welding thermal cycle. If such precipitates already exist in the as-welded state, then the supersaturation of the constituent elements of the precipitates is decreased. Consequently, the volume fraction or number density of fine particles reprecipitated during PWHT was decreased. Furthermore, as shown in Fig. 4(b), lined up precipitates were slightly coarser than the surrounding particles. Because these precipitates become coarser during aging,42) the proportion of coarse precipitate particles is expected to be increased in the fine grain region. It is possible that these effects were superimposed on the coarsening of precipitates in the fine grain region. Consequently, it is considered that ferrite formation resulted from microstructural changes, for which interparticle spacing was locally sparse during aging.
To elucidate microstructural changes in 9Cr-1Mo-V-Nb steel weld metal caused by long-term use, microstructures in the weld metal before and after aging at 1013 K were investigated. The following conclusions were obtained.
(1) In the weld metal, martensite structures with different prior austenite grain sizes were observed at certain intervals. In the as-welded metal, most of the precipitates were located on grain boundaries in the region where prior austenite grains were coarse (the coarse grain region). In the region where prior austenite grains were fine (the fine grain region), precipitates were observed not only on grain boundaries but also inside grains. However, the volume fraction of precipitates observed in the fine grain region was qualitatively lower than those observed in the coarse grain region.
(2) Results suggest that the coarse grain region corresponded to the region where the maximum heating temperature due to welding thermal cycle was below the transformation point. The fine grain region was the region heated above the transformation point.
(3) No remarkable microstructural change was observed in the weld metal after PWHT (1013 K/7.7 h). After high-temperature and short-time aging (1013 K/100 h), ferrite grains with low dislocation density were observed in the fine grain region.
(4) After aging at 1013 K/100 h, precipitates in the fine grain region were coarser and more sparsely distributed than those in the coarse grain region. Nevertheless, no significant difference was found in the dislocation density of martensite between the coarse and the fine grain region. Assuming that ferrite was formed through recrystallization thorough the bulging mechanism, it can be explained qualitatively that ferrite formation resulted from local reduction of pinning energy in the fine grain region.
