Experimental material

A commercial F/M steel P92 pipe with a wall thickness of 60 mm without service was selected as the initial material in this study, and its mechanical properties were reported in Refs. [36-42]. The pipe was delivered after a final heat treatment consisting of normalizing and tempering (NT). Normalizing was conducted at 1323 K for 30 min, whereas tempering was conducted at 1038 K for 60 min. After NT, the pipe was cooled to room temperature. The chemical composition of the pipe was 0.093C, 0.14Si, 0.41Mn, 8.75Cr, 1.62W, 0.183V, 0.052Nb, 0.063N, 0.207Ni, 0.505Mo, 0.001Ti, 0.003B (wt%), and bare Fe. Previous studies [17, 43] suggested that strain-induced boundary migration (SIBM) can improve the fraction and distribution of CSLBs along the PAGB. Consequently, steel plates with dimensions of 300 mm × 60 mm × 2.5 mm were subjected to a TMP involving three cold-rolling reduction ratios, namely, 6%, 9%, and 12%, followed by austenitization at 1323 K for 40 min, tempering at 1053 K for 45 min, and air cooling. Figure 1 shows the TMP routes for all the specimens. The as-received and 6%, 9%, and 12% deformed specimens are designated as AR, 6D, 9D, and 12D, respectively. The nomenclature of each specimen is listed in Table 1. The specimens were cut normal to the rolling direction to observe the microstructure; the observation area was located at the edge of the specimen.

Fig. 1Full-screen View

(Color online) Schematic representation of thermomechanical processing of P92 steel

Microstructural investigation

Specimens for EBSD analyses with a dimension of 10 mm × 10 mm × (2.2-2.35) mm were cut from the inside TMP-treated steel plates, and the EBSD scanning area was vertical to the plate surface. The specimens were first mechanically ground with SiC paper in different grits, up to 5000 grit, and then mechanically polished with a diamond solution with different diamond grain sizes, up to 1 μm. The final polishing was carried out with 0.04 μm silica solution. The GB was quantitatively evaluated using a ZEISS Gemini SEM 500 scanning electron microscope equipped with a Nordlys Max3 EBSD probe. The acceleration voltage was 20 kV, the beam current was 10 μA, and the scanning step size was 0.09-0.2 μm. The work adopted Brandon’s criterion to categorize the CSLBs: Δθ ≤ 15°Σ^-1/2 [44], and boundaries with 2° ≤ θ < 10° were considered as low angle grain boundaries (LAGBs). After EBSD scanning, Aztec crystal 2.1 software was used to reconstruct the PAG, assuming a Kurdjumov-Sachs (K-S) orientation relationship [29, 45-48]. Aztec crystal 2.1 software was used for subsequent analyses.

The EBSD maps of the AR specimens are shown in Fig. 2. The maps were cropped to 100 μm × 50 μm for better visualization of the microstructure. A typical lath-martensite microstructure is shown in Fig. 2. The inverse pole figure (IPF) maps of AR specimen after the PAG reconstruction was shown in Fig. 2, it can be clearly observed that the martensite lath disappears and PAGBs appear during the process of PAG reconstruction. The reconstructed PAG size is ~20 μm, which is similar to the metallographic result and the result obtained by T. Sakthivel et al. [49]. Relevant statistical results will be analyzed in the following sections.

Fig. 2Full-screen View

(Color online) Inverse pole figure (IPF) map of the as-received (AR) specimen showing the character of individual gain boundaries in the martensite phase. Low angle grain boundaries (LAGBs), random high angle grain boundaries (RHAGBs), Σ3ⁿ boundaries, and other coincident site lattice boundaries (CSLBs) are indicated with grey, black, red, and green lines, respectively

Compared with the EBSD technique, metallographic analysis can observe a larger area, which can ensure the accuracy of the PAG size measurement. In the metallographic method used to reveal the microstructure, the specimens were ground, polished, and etched in a solution of 1.5 mL nitric acid (superiorly pure), 1 mL hydrofluoric acid (analytically pure), 2 mL hydrogen peroxide (30%, analytically pure), 10 mL detergent (including surfactant), and 50 mL distilled water for 2-10 s. The linear intercept method was used to measure the PAG size from the metallographic images. All images showed more than 500 PAGs and more than 30 lines were randomly drawn on each image.

Corrosion test

Electrochemical corrosion tests were performed to study the effect of the TMP on intergranular damage resistance. To perform the electrochemical corrosion test, the specimen was wire cut to a size of 10 mm × 2 mm × (2.2-2.35) mm. The electrochemical specimen was welded to a copper conductor and sealed with epoxy resin, such that the exposed working surface area was 0.22-0.235 cm². The surfaces were polished with sandpaper and wiped with anhydrous alcohol. The Tafel curve was measured in 0.03 M Na₂SO₄ and 0.03 M NaCl aqueous solution, which is consistent with the experimental environment of Pang et al. [35], at room temperature (RT) by utilizing the three-electrode system and CHI760e electrochemical workstation. The working electrode was the specimen, the reference electrode was a saturated calomel electrode (SCE), and the auxiliary electrode was a platinum sheet. The experiment began when the steady-state open circuit potential (OCP) was reached, and the Tafel curve was obtained in the scanning range of −1 to 1 V with respect to the OCP at a scanning rate of 0.005 V/s. To ensure reproducibility, all the corrosion experiments were repeated at least three times.

Prior austenite grain size

The grain structures of TMP-treated P92 steel with different reduction ratios are shown in Fig. 3. For better visualization of the structure, the images were cropped to 250 μm × 220 μm. The distribution of the PAGBs and martensitic lath boundaries is shown in Fig. 3, indicating the typical microstructure of P92 steel.

Fig. 3Full-screen View

Optical microscope images of the studied steels: (a) AR, (b) 6D, (c) 9D, (d) 12D

Considering the thickness of specimens and the low cold-rolling reduction ratios, the PAG size was measured in two areas as shown in Fig. 4(a), one just beneath the surface, identified as ‘edge’ and the other at the center of the specimen, identified as ‘middle’. Based on the metallographic analysis, the PAG sizes of the AR- and TMP-treated specimens against the reduction ratio are shown in Fig. 4(a). Compared with the AR specimen, the PAG size of all TMP-treated specimens did not change significantly, which is consistent with the results in Ref. [18]; however, the PAG sizes of the specimens with different reduction ratios were slightly different. The PAG size in the edge area decreased with an increase in the reduction ratio, whereas the PAG size in the middle area increased with increasing cold-rolling reduction ratio, up to 9%, and then decreased at a reduction ratio of 12%. The literature shows that the growth dynamics of the PAG are directly related to the degree of deformation, temperature, and time [50]. Under isothermal growth conditions, when the temperature is above the PAG growth temperature Ac1, the higher the peak temperature and the longer the holding time, the larger is the PAG size under ideal conditions. In this study, the normalizing temperature and normalizing time were the same, and the tempering temperature was lower than the PAG growth temperature. Therefore, the heat treatment process should have the same effect on the PAG size. In conclusion, the strain introduced by cold rolling was the only factor that affected the PAG size of the TMP-treated specimens. The recrystallization fraction determined by the grain orientation spread (GOS) value [51] was positively correlated with the reduction ratio in this study, as shown in Fig. 4(b), indicating that the strain introduced by cold rolling promotes recrystallization. Only grains with 0° < GOS < 1° [52, 53] were identified as recrystallized grains, and this criterion was consistent with the experimental results. To the best of our knowledge, SIBM promotes grain growth, whereas recrystallization reduces the grain size [16, 29]. The relationship between the PAG size and recrystallization fraction calculated from the EBSD data in the edge area is consistent with the above discussion. However, this seems to contradict the changing trend of the PAG size in the middle area. The following provides a detailed explanation of this contradiction. As shown in Fig. 4(a), the edge and middle areas of 12Cr F/M steel also have different change trends in PAG size when the annealing time is 45 min. Kinoshita et al. [29] noted that a short time did not appear to be long enough for grain growth in the middle area when the reduction ratio was low. In our work, the variation trend of the PAG size in the middle area with the reduction ratio was inconsistent with that of the recrystallization fraction because we collected EBSD data from the edge area.

Fig. 4Full-screen View

(Color online) (a) prior austenite grain size of the studied steel under different thermomechanical processes (TMPs), (b) Grain orientation spread distribution for all the studied steels. The data on 10Cr F/M steel and 12Cr F/M steel were extracted from Ref. [16] and Ref. [29], respectively

Based on the above analysis, we can conclude that SIBM dominates at low strains, and recrystallization dominates as the strain increases. To the best of our knowledge, GB migration is usually accompanied by the formation of twins, and recrystallization usually introduces new random boundaries. Therefore, it can be expected that a high fraction of twin boundaries, corresponding to the {111} Σ3 boundaries, will be introduced into the austenite phase for the 6D specimen, and the fraction of CSLBs along PAGBs will decrease with the increase in reduction ratio, which will be discussed in the next section.

Prior austenite grain boundary character distribution

The EBSD maps of AR specimen and the TMP-treated specimens after PAG reconstruction are shown in Fig. 5(a-d). Different types of PAGBs are marked by rendering different colored line segments. For better visualization of the structure, the maps were cropped to 100 μm × 50 μm. Compared with the AR specimen as shown in Fig. 5(a), it can be found that Σ3ⁿ boundaries still account for the majority of CSLBs along the PAGBs after the TMP. To ensure statistical accuracy, at least 500 PAGBs were sampled for statistical analysis.

Fig. 5Full-screen View

(Color online) EBSD maps of P92 steel after prior austenite grain (PAG) reconstruction for the (a) AR, (b) 6D, (c) 9D, and (d) 12D specimens; (e) coincidence site lattice boundaries (CSLBs) along the distribution of prior austenite grain boundaries (PAGBs). The low angle grain boundaries (LAGBs), random high angle grain boundaries (RHAGB), Σ3ⁿ, and other CSLBs are indicated with grey, black, red, and green lines, respectively

The fractions of CSLBs along the PAGBs of the studied steels are shown in Fig. 5(e). The TMP clearly led to higher fractions of CSLBs along PAGBs. A high fraction of Σ3 boundaries can be observed along PAGBs. Compared with the AR specimen, the fraction distribution of Σ3 boundaries along PAGBs for 6D, 9D, and 12D improves by 46.3%, 27.1%, and 8.3%, respectively. Figure 6(a) shows the fraction of Σ3ⁿ boundaries along PAGBs as a function of the reduction ratio for the AR specimen and TMP-treated specimens. The fraction of Σ3ⁿ boundaries along PAGBs was evaluated as follows: $f_{\left(\Sigma 3^n\right)}^{\left(\gamma \right)}=\frac{L_{\left(\Sigma 3^n\right)}^{\left(\gamma \right)}}{L_{\left(\text{total}\right)}^{\left(\gamma \right)}},$ (1) where $L_{\left(\Sigma 3^n\right)}^{\left(\gamma \right)}$ is the total length of Σ3ⁿ boundaries along PAGBs, and $L_{\left(\text{total}\right)}^{\left(\gamma \right)}$ is the total length of PAGBs. The variation trend of the fraction of Σ3ⁿ boundaries along PAGBs with the reduction is similar to that of PAG size. The fraction of Σ3ⁿ boundaries along PAGBs decreases with the increasing reduction ratio. Figure 6(b) shows the relationship between the fraction of CSLBs along PAGBs and the reduction ratio. The fraction of CSLBs along PAGBs also decreased with an increasing reduction ratio. This finding is consistent with the results of Shimada et al. [11] and Kokawa et al. [22], who found a peak in the fraction of CSLBs at a 5% deformation and an inverse relationship between the fraction of CSLBs and the reduction ratio when the deformation was greater than 5% in ɣ-Fe. For comparison, the data obtained from 10Cr and 12Cr F/M steels are shown in Fig. 6(a) and the data obtained from 9Cr and 10Cr F/M steels are shown in Fig. 6(b). As mentioned in Sect. 3.1, SIBM can promote twin formation, whereas recrystallization can introduce new random boundaries. However, the mechanism that dominates a given deformation is still unclear. We assume a special value of deformation, σ, for different materials and different TMPs. When the deformation is less than σ, SIBM is dominant, otherwise recrystallization is dominant instead of SIBM. Actually, σ is different for different materials and different TMPs. As can be observed from Fig. 6, σ is 10% in the work of Hirryama et al. (10Cr F/M steel) [16], and σ is 5% in the work of Kinoshita et al. (12Cr F/M steel) [29]. It can be reasonably guessed that the deformation was greater than σ in this work, which can be confirmed from the recrystallization fraction in Sect. 3.1 and the change in trend of the fraction of Σ3ⁿ boundaries along PAGBs and CSLBs along PAGBs.

Fig. 6Full-screen View

(Color online) (a) Fraction of Σ3ⁿ boundaries along PAGBs as a function of the reduction ratio, (b) relation between the fraction of CSLBs along PAGBs and the reduction ratio. The data on 9Cr F/M, 10Cr F/ M, and 12Cr F/M steels were extracted from Ref. [17], Ref. [16], and Ref. [29], respectively. The shaded area represents the error bar

Prior austenite grain boundary connectivity

Figure 7 shows the two-dimensional PAGB network of four steels, where the green and black lines denote special boundaries (Σ3-Σ29b boundaries) and random boundaries, respectively. Besides the different fraction of Σ3ⁿ boundaries along PAGBs and CSLBs along PAGBs for the four studied steels, the distribution of special boundaries and the phenomena of reciprocal connectivity significantly vary. With the application of the cold-rolling-normalizing-tempering process, the random prior austenite GBs were segmented into fragments.

Fig. 7Full-screen View

(Color online) Schematic of the prior austenite grain boundary (PAGB) connectivity of the AR and TMP-treated specimens: (a) AR, (b) 6D, (c) 9D, (d) 12D

Percolation theory is typically used to simulate GB connectivity [54]. Previous studies suggested that four types of triple junctions can occur in a material [55]: TJ0 (R-R-R), TJ1 (S-R-R), TJ2 (S-S-R), and TJ3 (S-S-S), as shown in Fig. 8(a). To assess GB connectivity, it is necessary to consider the distribution of triple junctions. To the best of our knowledge, a triple junction containing at least one random boundary is susceptible to such attacks. However, no TJ2 or TJ3 triple junctions were observed in any of the specimens, as shown in Fig. 8(b). Only the 6D specimen had more TJ1 and fewer TJ0 triple junctions than the AR specimen. The triple distributions of the other specimens were almost identical to those of the AR specimens. This means that only the 6D specimen has more broken prior austenite random boundaries than the AR specimen, whereas the other specimens show little change in prior austenite random boundary interruptions than the AR specimen.

Fig. 8Full-screen View

(Color online) (a) Four types of triple junctions (S: special boundary, R: random boundary): (Ⅰ) TJ0, (Ⅱ) TJ1, (Ⅲ) TJ2, (Ⅳ) TJ3; (b) triple junction distribution of the AR and TMP-treated specimens. The data was calculated by using Matlab software (mtex)

Carbide Particles

The SEM micrographs in Fig. 9 show the distribution of precipitates in the AR, 6D, 9D, and 12D specimens. One of the authors [56, 57] has proved that Cr-rich M₂₃C₆ is the main precipitate phase in P92 steel when tempering temperature is above 700 ℃. To the best of our knowledge, subblock boundaries and PAGBs act as preferential nucleation sites for Cr-rich M₂₃C₆ and GBE can enhance the fraction of subblock boundaries [16, 17], leading to finely dispersed precipitates in the studied steel. Furthermore, CSLBs along the PAGBs can strongly suppress the formation and coarsening of chromium carbide precipitates, as mentioned in the Introduction, because the diffusivity of CSLBs is generally lower than that of random boundaries. A quantitative analysis of the precipitate phase requires further investigation, which is beyond the scope of this study. However, based on the above analysis, it can be inferred that the 6D, 9D, and 12D specimens have more densely dispersed precipitates than the AR specimens, as shown in Fig. 9, which is consistent with the results reported by Hirayama et al. [16] and Gupta et al. [18]. This means that the GBE-treated steel can achieve better high-temperature properties than the AR specimens. Experimental measurements of high-temperature properties will also be carried out in future studies. Meanwhile, it can also be inferred that the corrosion resistance will be improved in GBE-treated materials as chromium segregation is suppressed owing to the enhanced fraction of CSLBs along the PAGBs, as discussed in Sect. 3.2.

Fig. 9Full-screen View

SEM micrographs of the morphology of precipitation particles in (a) AR, (b) 6D, (c) 9D, and (d) 12D specimens

Corrosion behavior

The increase in CSLBs along the PAGBs fraction and the disruption of prior austenite random boundaries are helpful in improving the resistance to intergranular damage [16]. From the results of the analysis in Sects. 3.1, 3.2, and 3.3, a TMP with a low reduction ratio can enhance the fraction of CSLBs along the PAGBs and disrupt the prior austenite random boundaries, thereby improving the intergranular resistance. In this study, electrochemical corrosion tests were conducted to determine whether these changes in the prior austenite GBCD and PAGB connectivity are associated with corrosion resistance. All the corrosion tests were repeated at least three times, and representative data are presented.

Potentiodynamic polarization curves were measured to determine the self-corrosion density (i_corr) and self-corrosion rate (v), as shown in Fig. 10(a), and to verify the effectiveness of the GBE. i_corr was obtained by extrapolating the potentiodynamic polarization curve, and v was calculated according to the following formula [58]: $v=\frac{M}{nF}{i}_{\text{corr}},$ (2) where M is the molecular mass of the material, n is the number of shifted electrons, and F is Faraday’s constant. The corresponding data are listed in Table 2.

Fig. 10Full-screen View

(Color online) (a) Potentiodynamic polarization curves, (b) self-corrosion current density and self-corrosion rate as a function of the fraction of CSLBs along PAGBs

The variations of i_corr, as well as corrosion rate v followed a trend contrary to the change in the fraction of CSLBs along PAGBs, as shown in Fig. 10(b). As the fraction of CSLBs along PAGBs increases, the value of i_corr is (1.02 ± 0.15) × 10^-5, (1.53 ± 0.41) × 10^-5, (3.16 ± 0.19) × 10^-5, and (3.57 ± 0.25) × 10^-5 A·cm^-2, respectively, and the value of v is 3.12 ± 0.46, 4.66 ± 1.25, 9.66 ± 0.58, 10.92 ± 0.74 mil·year^-1, respectively. The self-corrosion density and rate of the TMP-treated specimens were lower than those of the AR specimen, reflecting an improvement in the electrochemical corrosion resistance. The specimen with more CSLBs along PAGBs showed a decreasing electrochemical corrosion susceptibility, which proves that GBE is effective in improving the intergranular damage resistance of P92 steel under the presented conditions. In addition, this result is in accordance with the visual inspection of the metallographic analysis.

Typical SEM micrographs of AR and TMP-treated specimens after the potentiodynamic polarization test in 0.03 M Na₂SO₄ and 0.03 M NaCl aqueous solutions are shown in Fig. 11(a-h). The electrochemical corrosion morphologies of the AR specimens show larger and denser corrosion pits than those of the TMP-treated specimens. In addition, the high-magnification micrograph of the AR specimen indicates that corrosion cracks were generated, as shown in Fig. 11(b), which were not observed in the other specimens, as shown in Fig. 11(d, f, h). The electrochemical corrosion morphologies of the 9D and 12D specimens show fewer corrosion pits than those of the AR specimen, as shown in Fig. 11(e-h), and their surfaces are smoother than those of the AR specimen. The electrochemical corrosion morphologies of the 6D specimen exhibited the lowest number of corrosion pits and demonstrated a low degree of corrosion compared with the other specimens, as shown in Fig. 11(c, d).

Fig. 11Full-screen View

SEM micrographs of the specimens after anodic polarization. Low magnification for (a) AR, (c) 6D, (e) 9D, and (g) 12D. High magnification for (b) AR, (d) 6D, (f) 9D, and (h) 12D