MA Yuanjun, WANG Jingjie, GAO Yubi, WANG Xingmao, CHEN Jianjun, DING Yutian＊

(1. State Key Laboratory of Advanced Processing and Recycling of Non-ferrous Metals, Lanzhou University of Technology, Lanzhou 730050, China; 2. School of Materials Science and Engineering, Lanzhou University of Technology, Lanzhou 730050, China)

Abstract: The feasibility of applying the grain boundary character distribution (GBCD) optimization to Inconel 625 for improving the intergranular corrosion (IGC) resistance was studied. The GBCD was obtained and characterized by electron backscatter diffraction (EBSD) analysis, and its optimization was mainly attributed to annealing twins (Σ3) and twins related to boundaries formed during thermal-mechanical processing (TMP). Through TMP of 5% cold rolling and subsequent annealing at 1150 ℃ for 5 min, the proportion of low Σ coincidence site lattice (CSL) grain boundaries of the Inconel 625 can be enhanced to about 35.8% which mainly were of Σ3n (n=1, 2, 3) type. There is an increase of 24.8% compared with the solution-treated sample, and simultaneously the large-size highly-twinned grain-cluster microstructure is formed. The grain-cluster is mainly composed of Σ3-Σ3-Σ9 or Σ3-Σ9-Σ27 triple junctions, which is mainly caused by boundary reactions during grain growth. Among them, the IGC resistance of Σ3 grain boundaries, Σ9 grain boundaries and random grain boundaries is sequentially weakened. With the increase of the low ΣCSL grain boundary fraction, the IGC resistance of Inconel 625 improves. The essential reason is the amount of Σ3 boundaries interrupting the random boundary network increases and the large grain-cluster arrests the penetration of IGC.

Key words: nickel-based superalloys; grain boundary characteristic distribution (GBCD); corrosion behavior; low Σ coincidence site lattice (low ΣCSL); EBSD

1 Introduction

Inconel 625 is a kind of solid solution strengthened nickel-base wrought superalloy with Mo and Nb as the main strengthening elements. Meanwhile, the particularly high corrosion resistance and strength offered by Inconel 625 has resulted in its application in several specialist areas including gas turbine engine, nuclear power equipment, aerospace engine and ship application[1-5]. With the rapid development of science and technology, the industry’s requirements for the service life and performance of equipment and components continue to increase. Meanwhile, studies by researchers have demonstrated that the corrosion failure of the metal materials with face center cubic (FCC) structure in different corrosion environments is closely related to the grain boundary structure[6-8]. Thus, the need to further improve the resistance to intergranular corrosion (IGC) in Inconel 625 should be considered.

It is customary to improve the corrosion resistance of alloys by composition adjustment or chemical surface treatment. The corrosion resistance of Inconel 625 mainly from the rapid formation of Cr2O3oxide film by the Cr element in the alloy to hinder the continuation of the corrosion reaction[9,10]. However, considering the cost and environmental protection, it is a better choice to improve the corrosion resistance of the alloy by adjusting the alloy structure without changing the alloy composition. The low Σ coincidence site lattice (CSL) grain boundaries, as the low energy grain boundaries (GBs), exhibit strong resistance to IGC and thus called “special grain boundaries (SBs)”. The thermo-mechanical processing (TMP) was commonly used to manipulate the SBs. By exercising control over the fraction of SBs (fSBs), the IGC resistance of alloy can be greatly enhanced[11-13]. The idea of grain boundary design and control was proposed by Watanabe[10], which was refined to be grain boundary engineering (GBE), also known as grain boundary characteristic distribution (GBCD) optimization. Recent studies have shown that the occurrence of IGC of polycrystalline alloys depends on the state of precipitates (especially carbides with Cr) near the GBs and whether precipitates are formed on the GBs[14-17]. All of the above are closely related to the effect of GBCD on the precipitates at the GBs. On the other hand, some researchers have studied how to control the microstructure and improve the IGC resistance of nickel-based alloys[18-20]by GBCD, especially Inconel 690[7], Inconel 600[21]and Inconel 617[22]. However, there are only few studies on the effect of GBCD on IGC in Inconel 625.

The effects of GBCD on IGC are particularly significant in Inconel 625 due to their being typically installed in a solution annealed (precipitate free) state[23]. Our group has conducted multiple studies on solution treatment[24-26], GBCD[27-29], and high temperature corrosion[30,31]of Inconel 625, which has an excellent research foundation. Therefore, we chose GBCD optimization to explore the effect of SBs on IGC resistance of Inconel 625 on the existing basis. And several factors related to the IGC are described such as the grain boundary connectivity, grain size, trigeminal grain boundary characteristic distribution and grain boundary chemical composition distribution. Ultimately, the present work in order to provide a solution for improving the IGC resistance of the Inconel 625.

2 Experimental

The experimental Inconel 625 used in the present study was melted by vacuum induction melting (VIM) and electro-slag remelting (ESR). Subsequent the twostage homogenization treatment at 1 140 ℃ for 10 h plus at 1 210 ℃ for 48 h, the cast ingot was hot extruded on a 10 MN horizontal extruder (the extrusion speed is 50 mm/s, the extrusion ratio is 7.4 and the extrusion temperature is 1 150 ℃). The chemical compositions of Inconel 625 used in this study are as follows (wt %): Cr: 21.77, Mo: 8.79, Nb: 3.75, Fe: 3.68, Ti: 0.4, Al: 0.21, Mn: 0.2, Co: 0.19, Si: 0.12, C: 0.042, and the balance Ni. The hot extrusion samples were solution treated at 1 150 ℃ for 1 h followed by air cooling (AC) to room temperature for the GBCD sample. The solution-treated specimen was termed here as the base material (BM). The method of single step TMP is chosen to GBCD, that is, small deformation and annealing. There are two main annealing processes, one is low temperature with long time annealing, the other is high temperature with short time annealing. Due to Inconel 625 being typically installed in a solution annealed (precipitate free) state, and the influence of SBs on IGC is mainly studied in this paper, the influence of precipitated phase should be excluded. Therefore, shortterm annealing at solid solution temperature (1 150 ℃) is selected to control the microstructure. The BM was cold-rolled by 5%, 10%, 15%, and 20% in thickness reduction, and then annealed at 1 150 ℃ for 5 min in vacuum atmosphere followed by AC. Specific parameters are shown in Table 1.

Table 1 TMP of specimens from Inconel 625

The GBCD of specimens was conducted by the electron backscattering diffraction (EBSD) in a thermal field emission gun FEI Nova Nano SEM 430 equipped with an orientation imaging microscopy (OIM) system. It operated at 15 kV and a tilt angle of 70°, with a step size of 1 μm. Before EBSD measurement, each sample was mechanically ground and then electropolished in a solution of 10vol% perchloric acid and 90vol% ethanol under 10 V for 25 s. The average grain size was estimated from the EBSD-reconstructed images. The fraction of different grain boundary types was determined according to the length fraction. In this study, GBs with Σ ≤ 29 were denoted as the ΣCSL boundaries with low energy[32,33], and the others are random boundaries with high energy. Brandon’s criterion[34](Δθ≤ 15°Σ-0.5) was adopted for the critical deviation in the grain boundary characterization. Among them, the Σ1 grain boundaries were defined as boundaries with a misorientation of 2°-15°, the random grain boundaries were defined as with a misorientation greater than 15°.

The IGC measurements of specimens with different GBCD were conducted using method A in ASTM G28. The corrosion test solution used was the 6.8 mol/L ferric sulfate plus 0.1 mol/L sulfuric acid. Before the experiment, the size, morphology and weight of the specimens were characterized. Then, they were immersed in a boiling corrosion solution for different periods (24-120 h)[35]. After each corrosion test, the specimens were taken out for cleaning and drying and subsequently reweighed to obtain weight loss and corrosion rate at different immersing times[36]. The microstructures of the tested samples were observed by JSM-6700 field emission scanning electron microscope (FESEM) equipped with Oxford INCA spectrometer, Axiovert 40 MAT optical microscopy (OM) and the corresponding EBSD measurement by SEM.

3 Results

3.1 GBCD optimization

In Fig.1, typical GBCD characterized by OIM maps is shown after TMP for the BM and the coldrolled specimens, in which the grain boundary network (GBN) consisted of grain boundaries with different types. It is found from GBN that the connectivity of the random boundaries in the BM is significantly better than that of the GBCD specimens. In OIM maps, Σ1 grain boundaries are set as the grain boundary formed when the orientation difference between two grains is less than 15° according to standard Brandon[34]. The step size of EBSD scanning in this experiment is 1μm, and the Σ1 grain boundaries were defined as boundaries with a misorientation of 2°-15°. In Fig.1, there are many fine Σ1 grain boundaries, these fine Σ1 grain boundaries indicate that there is more than 2° misorientation within a distance of 1μm, indicating that this is a deformed matrix without recrystallization. If there are no or few fine Σ1 grain boundaries, it indicates that the region has been recrystallized. Therefore, the change of Σ1 grain boundaries in Fig.1 is closely related to recrystallization. When the deformation is very small, the storage energy is not enough to drive recrystallization, the grain size is the original grain size, and the deformation matrix is less, that is, Σ1 grain boundary is less, as shown in Fig.1(b). When the amount of deformation increases to a certain extent, the distortion energy is enough to cause recrystallization. However, due to the small degree of deformation, the ratio of nucleation rate to growth rate is very small, so extremely coarse grains are obtained. Due to a certain degree of deformation and slow recrystallization, relatively more Σ1 grain boundaries are obtained, as shown in Fig.1(c). As the amount of deformation increases, the storage energy that drives nucleation and growth increases, while the nucleation rate increases rapidly, which increases the ratio of nucleation rate to growth rate, and the recrystallized grains become more and more refined, while Σ1 grain boundary becomes less and less, as shown in Figs.1(c) and 1(d). So Σ1 grain boundary in Figs.1(b)-1(d) varies with the amount of deformation.

Fig.1 OIM maps of SBs and random boundaries in the specimens cold-rolled to different deformation amounts after annealing at 1 150 ℃ for 5 min: (a) BM; (b) ε= 5%; (c) ε= 10%; (d) ε= 15%; (e) The enlarged image corresponding to the blue-framed regions in (d); (f) ε= 20%

Additionally, the large-sized grain cluster is also a typical microstructure of GBCD optimized sample which is encircled by the random boundary containing many Σ3 (red), Σ9 (blue), and Σ27 (green) grain boundaries[35]. During GBCD optimization, the grain cluster mentioned by the investigators[37-39]also occurs in the Inconel 625, as shown in Figs.1(b-f). The Σ3n(n= 1, 2, 3) grain boundaries inside of the grain cluster connect and form a large number of triple junctions, such as Σ3-Σ3-Σ9 and Σ3-Σ9-Σ27 (represent the reactions ΣA+ΣB→ΣA×B and ΣA+ΣB→ΣA/B), as shown in Fig.1(e). In other words, the clusters of large-size grains with Σ3nrelationship are formed via a large number of Σ3-Σ3-Σ9 and Σ3-Σ9-Σ27 triple junctions.

The proportion of SBs also changes significantly under different TMPs. The GBCD statistics are shown in Fig.2. From it, thefSBsis determined to be 11.0% in the BM, as well as the connectivity of the random boundary network is relatively complete, as shown in Fig.1 and Fig.2. Moreover, it can be seen that thefSBsof the GBCD optimized samples are all higher than or close to that of the BM samples. The highestfSBsof 35.8% appears in the sample annealed at 1 150 ℃ for 5 min after cold rolling by 5% reduction, which is mainly composed of a large number of annealed twins (Σ3) and a small number of multiple twin boundaries (Σ9, Σ27). With the increase of cold-rolling reduction, the trend of the fraction of Σ9 and Σ27 grain boundaries is the same as that of Σ3 grain boundaries, and both gradually decrease. In summary, it can be found that the smaller cold-rolling reduction is really beneficial to the formation of SBs during recrystallization.

Fig.2 The histogram of the fSBs and the fractions of various low ΣCSL boundaries for the BM and the specimens cold-rolled to different reductions and then annealed at 1 150 ℃ for 5 min

Studies have been reported that twin boundaries (Σ3) are grain boundaries with high symmetry and have similar grains refinement as traditional grain boundaries[40]. Therefore, during calculating the grain size, two kinds of grain sizes were calculated, twins also regarded as grains and twins no regarded as grains, as shown in Fig.3. Meanwhile, the results of reduction ratio with 5% and 20% show that the grain size distribution of the alloy under different TMP is related closely to the twin boundaries. In specimens rolled at 5%, the grains are coarser (41.52 μm) and the fractions of Σ3 boundaries is 30%. The 20% deformation of the specimen resulted in a very fine-grained structure (18.30 μm) and the fraction of Σ3 boundaries is 9.6%. Interestingly, a greater amount of cold rolling deformation produced a finer grain size and a smaller fraction of Σ3 boundaries, while smaller deformation can produce largesize grains and more Σ3 boundaries, as shown in Fig.2 and Fig.3. This phenomenon can be explained by the formation mechanism of Σ3 boundaries, which requires the migration of grain boundaries[11]. A greater number of grain boundary reactions (i e, a higher extent of Σ3 boundary regeneration) are present in coarser materials owing to the longer distance covered by the grain boundaries of growing grains[41]. As a result, after different processes of TMP, the higher the content of Σ3 grain boundary in the larger grain structure.

Fig.3 The grain size distribution of Inconel 625 under different TMPs

3.2 IGC behavior

3.2.1 Corrosion rate

To study the relationship between GBCD optimization and IGC resistance in the Inconel 625, the BM with the lowestfSBs(11.0%), the r5%-a1 150 ℃/5 min sample with the highestfSBs(35.8%), and other samples with the moderatefSBs(25.2%, 14.8%, 11.5%) are selected for IGC performance tests. According to GB/T 15260-2016 intergranular corrosion test method of nickel-base alloy, the calculation formula of the corrosion rate is as follows:

whereγcorris the corrosion rate,Kis a constant (8.76×104),ΔMis the weight loss (g),Ais the exposed area (cm2),tis the exposure time (h), andρis the density of the specimen (g/cm3). According to the Eq. (1), the varieties of corrosion rate with the change of corrosion duration after the ferric sulfate-sulfuric acid test is obtained as indicated in Fig.4. In the initial 96 h, the corrosion rate changes gently with the increase of immersing time. In the last 24 hours, as the immersing time increases, the corrosion rate increases greatly. The reason why the corrosion rate changes with the immersing time may be that the water in the solution evaporates continuously and the concentration of corrosive substances in the solution increases with the corrosion time at any time, which accelerates the corrosion of the sample. It is also found that no matter what the corrosion duration is, the corrosion rate of BM is the fastest and the corrosion rate of r5%-a1 150 ℃/5 min sample is the slowest. Additionally, thefSBsis highest in r5%-a1 150 ℃/5 min sample and least in BM. This indicates that the increase offSBscan improve the IGC resistance of Inconel 625.

Fig.4 Comparison of the corrosion rate in the Inconel 625 with different GBCD during ferric sulfate-sulfuric acid corrosion tests

3.2.2 Corrosion morphologies

The corrosion morphology of the GBCD optimized specimens via different TMP are presented in Fig.5, respectively. After immersing for 120 h, the IGC of the BM is very serious. The corrosion groove of GBs is wide, and the grain drop is significantly more than that of the GBCD optimization samples, indicating that more serious IGC occurred in the BM. Effectively, the GBCD optimization can improve the IGC resistance of the alloy. In r10% (15%, 20%)-a1 150 ℃/5 min samples, more grains dropped, and almost no grains dropped in r5%-a1 150 ℃/5 min sample. According to the width of the grain boundary corrosion groove and the number of grain drops (grains are detached from the surface microstructure of the sample due to corrosion), it was found that the improvement of IGC resistance for Inconel 625 with 5% reduction (the highestfSBs) was most obvious.

Fig.6 shows the SEM results of the cross-section of Inconel 625 after the ferric sulfate-sulfuric acid test for 120 h, which further confirms the results in Fig.5. After immersing for 120 h, the corrosion along the grain boundary can be seen more clearly from the cross-section, as shown in Figs.6(a2) and 6(c2). The GBs were seriously corroded, and the grain tended to drop off. By comparing the percolation depth of different samples, it is found that the depth of BM is about twice that of the GBCD optimization samples. GBCD optimization breaks the connectivity of random grain boundaries, so it has a significant hindering effect on corrosion along the grain boundaries, as seen in other studies[15]. In different GBCD optimization samples, the percolation depth changed little, nearly 30 μm. Among them, the corrosion of r5%-a1150℃/5 min sample is the lightest, owing to the small corrosion width and corrosion area along the grain, no grain dropping tendency, as shown in Fig.(6b2). As previously indicated in Fig.4 and Fig.5, the least grain dropping occurred in the sample with the highestfSBs, which means the IGC andfSBsare one-to-one correspondence.

Fig.6 SEM images of the cross-sections in: (a) BM, (b) r5%-a1 150 ℃/5 min, (c) r10%-a1 150 ℃/5 min, (d) r15%-a1 150 ℃/5 min and (e) r20%-a1 150 ℃/5 min specimens after 120 h ferric sulfate-sulfuric acid corrosion tests. (2) is an enlarged view of the yellow box in (1)

3.2.3 Precipitates

The element distribution of samples undergone IGC was characterized. It can be seen from the Fig.7 that there is no element segregation in the corrosion area near the eroded grain boundary, which indicates the existence of no precipitate phase at there. In the inner region (the matrix), there is a small amount of dispersed granular precipitates. According to EDS map analysis, granular precipitates are mainly the segregation of Mo and Nb elements. Based on previous studies and phase diagrams of Inconel 625[42,43], it is determined that the precipitates are mainly (Mo, Nb)C (MC carbide), which mainly from residues in the alloy casting stage. Due to Inconel 625 is a solid solution strengthened alloy, and the alloy has been solid solution treated at 1 150 ℃ before GBCD, that is, the alloy structure is an austenite structure in which most of the precipitated phase has been dissolved in the alloy matrix. After GBCD treatment, the alloy was annealed for a short time at solution temperature to adjust the alloy structure, which also resulted in re-dissolution of precipitation phase. As previously indicated in Fig.7, it can be seen that the elements in the IGC area are uniformly distributed. Therefore, the IGC rate of Inconel 625 in this paper should be mainly related to the distribution of grain boundary and the type of grain boundary structure.

Fig.7 SEM images and corresponding EDS element maps of the cross-sections in: (a) BM, (b) r5%-a1 150 ℃/5 min, (c) r10%-a1 150 ℃/5 min, (d) r15%-a1 150 ℃/5 min and (e) r20%-a1 150 ℃/5 min specimens after 120 h ferric sulfate-sulfuric acid corrosion tests

4 Discussion

4.1 GBCD formation

The GBCD optimization used in this work is a simple single-step TMP, which was much simpler than the iterative type TMPs[12,18,44]. Effectively, the GBCD optimization can improve the grain-boundaries-related properties of low stacking fault energy FCC materials (such as nickel-based alloy[45-47], austenitic stainless steel[48]and copper alloy[49]) via enhancingfSBs, especially IGC resistance[50-53]. According to Fig.2, the SBs in Inconel 625 are mainly composed of Σ3n(n=1, 2, 3) boundaries, in addition to a small amount of other low CSL boundaries, the Σ3 boundaries are dominant among these SBs. This result is similar to that of austenitic stainless steel and Inconel 600[54-56]. Besides, the misorientation between the annealing twins and the parent grains is ＜111＞/60°, which is exactly the Σ3 misorientation relationship in the CSL model. That is, the grain boundary between the twin and the parent grain is Σ3 grain boundary[57]. Therefore, the GBCD process of alloy mainly relies on increasing the ratio of annealing twins to increase the ratio of SBs related to twins, and achieve the purpose of adjusting the distribution of grain boundaries. It is found that the formation of annealing twins plays an important role in GBCD during TMP.

Generally, the lower the stacking fault energy of the material, the higher the probability of forming annealing twins. Inconel 625, as a medium and low stacking fault energy FCC structure alloy, often has annealing twins in its microstructure. Therefore, it is also very suitable for the use of GBE methods to control the microstructure. The GBCD process in Inconel 625 includes a small amount of cold working (deformation) and a high-temperature short-time annealing (recrystallization).

In the recrystallization annealing process, annealing twins will be formed due to growth accidents in grain growth and stacking faults of {111} plane nucleate and grow. The coherent Σ3 grain boundary formed in the recrystallization annealing process can product Σ3-Σ3-Σ9 triple junctions via the “Σ3 re-excitation model”[11], as shown in Fig.8. This not only increases the proportion of SBs in the alloy but is also one of the ways to produce Σ9 grain boundaries. The model believes that when the alloy is recrystallized, two recrystallized nuclei (G1, G2) containing coherent Σ3 grain boundaries grow up and contact. The grain boundaries in the contacting parts have higher mobility which is connected with the coherent Σ3 grain boundary in G1, as shown in process A in Fig.8. The highly mobile grain boundaries continue to migrate, and the coherent Σ3 grain boundary in G1 becomes longer and contacts the coherent Σ3 grain boundary in G2, resulting in a Σ9 grain boundary. This means the formation of triple junction of Σ3-Σ3-Σ9, as shown in processes B and C in Fig.8. Then, the Σ9 grain boundary will continue to migrate due to its higher mobility than the other two coherent Σ3 grain boundaries, as shown in process D in Fig.8. Until the Σ9 grain boundary contacts other coherent Σ3 grain boundaries to form a non-coherent Σ grain boundary, another new triple junction of Σ3-Σ3-Σ9 can be produced. With this evolution mechanism continuing in the process of recrystallization and grain growth, the proportion of Σ3 and Σ9 grain boundaries in the material are finally increased. It can be seen from the “grains” in Fig.8 that this structure (Σ3-Σ3-Σ9) is present in Inconel 625. This indicates that the mechanism proposed by the “Σ3 re-excitation model” applies to the Inconel 625 during the recrystallization. The study[58]showed that the increase in Σ3 and Σ9 proportions are the immediate cause for the improvement of IGC. Concurrently, it can be seen from the “grains” in Fig.8 that the IGC resistance of Σ3 and Σ9 grain boundaries is significantly better than that of random grain boundaries in Inconel 625. Therefore, the “re-excitation model” during the recrystallization is very important to Inconel 625. The “re-excitation model” occurs based on a large number of Σ3 grain boundaries which is closely related to the amount of deformation, so we need a suitable amount of deformation for GBCD optimization.

Fig.8 Schematic diagram of “Σ3 re-excitation model”

In the TMP process, more Σ3 grain boundaries can be produced with a smaller amount of deformation, as shown in Fig.2. This is because during recrystallization, the larger the deformation, the higher the nucleation density per unit volume[59,60], as a result, the smaller the grain size after recrystallization. In small grains, the expansion space of the crystal nucleus during the growth process is smaller, and the migration distance of the random grain boundary is shorter, which ultimately leads to a lower possibility of annealing twins. Therefore, the ratio of Σ3ngrain boundaries is low. On the contrary, the sample with a small amount of deformation has a small nucleation density during recrystallization. Then the growth space of the crystal nucleus is large, the migration distance of the random grain boundary is longer, so the probability of annealing twins is high. Finally, the ratio of Σ3 grain boundaries can be increased. As previously indicated in Fig.2, it can be seen that more SBs (especially Σ3 and Σ9 grain boundaries) can be produced when the deformation is controlled within a small range.

4.2 The effect of GBCD on the IGC

The GBCD optimization aims to improve the IGC resistance of some polycrystalline materials by substantially increasing thefSBs. After GBCD optimization, the Σ3 grain boundary is connected with other Σ3ngrain boundaries to form Σ3-Σ3-Σ9 or Σ3-Σ9-Σ27 grain boundaries, of which there are mainly Σ3-Σ3-Σ9 grain boundaries. These special triple junctions connect small grains to form grain clusters. Moreover, the size of the grain clusters is closely related tofSBs, higher proportion of Σ3ngrain boundaries leads to larger-sized clusters.

The larger the grain cluster, the better the IGC resistance of the alloy, as shown in Fig.9. The reasons for this result can be said from three aspects. The first aspect is that different grain boundaries have different IGC resistance. As shown in the “grain” in Fig.8 and the circle “3” in Fig.10, there is no sign of corrosion in the coherent twins inside the grain where the random grain boundary is deeply corroded, while the Σ9 grain boundary corrosion is relatively shallow. Therefore, it can be concluded that the IGC resistance of Σ3 grain boundaries, Σ9 grain boundaries, and random grain boundaries decreases in order. It is known that the larger the grain cluster, the higher the proportion of Σ3ngrain boundaries in the grain. That means the fewer random grain boundaries in the material structure, resulting in fewer grain boundaries that can be severely corroded. The IGC resistance of the alloy with a larger-sized grain cluster will be better, which is consistent with the results in Fig.9. In the second aspect, the IGC of the alloy is mainly along the random grain boundary. The larger the grain cluster, the more random grain boundaries that need to be corroded to make a grain drop, which means enhancing the IGC resistance of alloy. The third aspect is that during the IGC, the random grain boundary is the corrosion channel, as shown in Fig.6. The corrosive medium is transmitted to the inside through the corrosion channel. At the forefront of corrosion, due to the continuous chemical reaction between the corrosive medium and the alloy, it moves forward. However, as the corrosion progresses, the concentration of the corrosive medium will continue to decrease, resulting in slower corrosion. It is necessary to increase the concentration of the corrosive medium at the corrosion front via diffuse through the channel between the corrosion front and the surface of the material. If the size of the grain cluster is larger during IGC, the corresponding corrosion channel will be longer, resulting in a lower equilibrium concentration of the corrosion front. Therefore, the rate of corrosion to the inside is slower, and the rate at which the grains are dropped is also slower. The slower the structure under the corroded grains is exposed to the corrosive medium, the more protective effect on the internal structure will be. This is consistent with the results of Fig.5 and Fig.9. The surface layer is mainly corroded off by some small-sized grains, while the large-sized grains have better resistance to IGC.

Fig.9 Grain size and corrosion rate vs. fraction of SBs curves for the samples after 120 h ferric sulfate-sulfuric acid corrosion tests

Fig.10 The microstructure of r5%-a1 150 ℃/5 min specimens after 120 h ferric sulfate-sulfuric acid corrosion tests. (a) SEM images of the specimen after IGC; (b) The corresponding grain boundary distributions obtained by EBSD; (c) The corresponding SEM map of region G1 in (a); (d-f) The corresponding EBSD map of region G1 in (a)

It can be found that the deep corrosion groove appears along the uninterrupted random grain boundary, as shown in Fig.10. Notably, the IGC of random grain boundary interrupted by SBs (especially Σ3 grain boundaries) is shallower, and no obvious corrosion grooves are found at the SBs in the grains, as shown in Fig.10(c). This fully shows that the SBs which break the GBN structure can significantly inhibit the expansion of corrosion cracks and reduce the probability of grain dropping during the IGC process. Take the grain “G1” in Fig.10(a) as an example, as shown in the yellow circles “1” and “2” in Figs.10(c-f), when the corrosion extends to the Σ3 grain boundary, the growth of corrosion cracks is significantly suppressed. The Σ3n(n=1, 2, 3) grain boundaries can break the connection of grain boundaries, as seen in other studies[15]. It can also be seen in the enlarged view of Fig.1 that the GBN is broken by Σ3 grain boundaries. With the increase of Σ3 grain boundaries, it can greatly increase the block ratio of SBs to random grain boundaries. Moreover, the grain boundary (random grain boundary) of the grain is difficult to be continuously corroded and fall off from the surface of the microstructure. As a result, Inconel 625 will have a better IGC resistance. For small grains, there are fewer random grain boundaries, fewer Σ3 grain boundaries in the interior, lead to less random grain boundaries are interrupted. Therefore, random grain boundaries in small grains are easy to be continuously corroded, and thus fall off from the surface of the microstructure, accelerating the corrosion rate. At last, the comparisons of percolation depth, grain size, corrosion rate, andfSBsin the samples further demonstrate a quite beneficial effect of GBCD on the IGC improvement in the Inconel 625.

5 Conclusions

In this work, GBCD optimization has proven to be a successful process for producing Inconel 625 with highfSBs. The GBCD process includes a small amount of cold working (deformation) and a high-temperature short-time annealing (recrystallization). The reduction of cold rolling before annealing played a key role in GBCD during TMP, among which a smaller cold-rolling reduction was beneficial to the formation of SBs. It was noticed that the highestfSBs(35.8%) was observed in the the sample treated with 5% cold rolling followed by annealing at 1 150 ℃ for 5 min, which was an increase of 24.8% compared with that of BM. During TMP, the increase infSBswas mainly based on the formation of annealing twins (Σ3), and the formation of Σ9 grain boundaries was mainly through the boundary reaction which conformed to the “Σ3 re-excitation model” during grain growth. The IGC resistance of Σ3 grain boundaries and Σ9 grain boundaries were obviously better than random grain boundaries. The highfSBseffectively blocked the connectivity of the random grain boundary can inhibit the expansion of corrosion cracks near random grain boundaries under the ferric sulfate-sulfuric acid corrosion tests. In addition, the drop of a single large-sized grain cluster which mainly composed of Σ3-Σ3-Σ9 or Σ3-Σ9-Σ27 triple junctions required more random grain boundaries to be eroded, and the grains in sub-surface was being protected. Thence, IGC resisitance of the GBCD optimized sample was significantly enhanced under tests, with a corrosion rate of 0.65 g∙m-2∙h-1compared to 0.79 g∙m-2∙h-1in the BM. These findings are expected to be relevant in future studies concerning IGC resistance in Inconel 625.