Grain boundary character distribution and sensitisation behaviour of grain boundary engineered stable austenitic stainless steel (AISI 316L)

Introduction

Intergranular corrosion of austenitic stainless steels has always been a conventional problem during welding and high temperature use. One of the major reasons for intergranular corrosion is sensitisation, ¹ i.e. chromium depletion due to chromium carbide precipitation at grain boundaries. Evidences have shown that the extent and morphology of carbide precipitation is dependent on grain boundary (GB) structure and energy.^2–4 Low energy grain boundaries such as coincidence site lattice boundaries (CSLBs) or special boundaries (SBs) are considered to be more immune from sensitisation.

Grain boundary engineering (GBE), has emerged as a better alternative to improve intergranular properties^4–9 of low stacking fault energy materials. Grain boundary engineering has some advantages over traditional methods of suppressing interface related degradation and thus has received increasing attention recently. Research has mainly concentrated on the GBE processes to introduce a high fraction of low Σ coincidence site lattice boundaries (CSLBs) and the evolution mechanisms involved.^4,10–16

In addition to the evolution of CSLBs, other metallurgical changes, such as grain size, have also occurred. While most of the works reported have correlated CSLBs with sensitisation, changes in grain size during GBE process have not been extensively investigated. A few models describe sensitisation only based on grain size^17,18 or CSLBs, ⁴ but limited investigations simultaneously consider their effects on sensitisation.^19–21 Moreover, grain size distribution has also been shown to affect polycrystalline metal response. ²² Thus, it is necessary to take account of such changes when assessing sensitisation resistance of a GBE processed material.

An attempt has been made in this paper to investigate how a GBE processing affects the evolution of GB character distribution (GBCD) and the sensitisation behaviour in type 316L austenitic stainless steel. A wide range of prestrain (5–80) followed by annealing was used to complete the process. Furthermore, other metallurgical changes, i.e. the grain size, have also been correlated with the degree of sensitisation (DOS).

Experimental

The material used was type 316L austenitic stainless steel with a chemical composition of 17·01Cr–12·03Ni–0·40Ni–0·40Si–1·40Mn–0·025C–0·028P–0·003S–2·05Mo (wt-). The as received material was solution treated at 1373 K for 30 min to form a base material (BM) sample with CSLBs fraction f_CSL of 14·4 and high angle boundary (HAB) network fully connected (Fig. 1). The BM sample, 12×12×100 mm in size, was thermomechanically processed by one-step prestrain plus annealing. The prestrain was given by unidirectional cold rolling at room temperature, resulting in 5–80 thickness reduction. The annealing was completed in vacuum sealed quartz capsules at 1240 K for 72 h followed by water quenching. Specimens for electron backscatter diffraction (EBSD) studies were prepared by standard metallographic techniques and were subsequently electropolished in 5 perchloric acid in ethyl alcohol. The frequency of CSLBs and the GBCD were analysed using an HKL Technology Channel 5 EBSD detection system interfaced to a ZEISS ULTRA 55 field emission gun scanning electron microscope. In order to ensure the statistical significance, three areas of at least 1000×1000 μm in each condition were assessed. The fractions of the different GB types were determined on the basis of the length proportion. In the present study, grain boundaries with 3≤Σ≤29 were classified as low Σ CSLBs or SBs. The Brandon criterion ²³ was adopted for the critical deviation in the GB characterisation. The sensitisation propensity was evaluated by a double loop electrochemical potentiokinetic reactivation (DL-EPR) method ²⁴ in 0·5M H₂SO₄+0·01M KSCN solution after sensitisation treatment at 948 K for 20 h. Microstructures were examined by scanning electron microscopy (SEM) after electroetching in 10 oxalic acid solution (as per practice A, A262 ASTM). ²⁵ Moreover, the grain size was measured by the linear intercept method taking twins into account. The sensitisation treatments did not alter the grain size or GBCD.

OPEN IN VIEWER

1

Electron backscatter diffraction reconstructed grain boundaries for BM; both special and general high angle boundaries (HABs) are shown in a and only HABs are shown in b to assess random boundary connectivity

Results and discussion

The fractions of total CSLBs f_CSL, Σ3 (f_Σ3) and (Σ9+Σ27) (f_Σ9+Σ27) in BM and the specimens, which were strained by 5–80 and then annealed at 1240 K for 72 h, are given in Fig. 2.

OPEN IN VIEWER

2

Effect of roll reduction ratio on length proportion of total CSLBs, Σ3 and (Σ9+Σ27) boundaries in AISI 316L stainless steel thermomechanically processed by cold rolling and subsequent annealing at 1240 K for 72 h; error bars describe standard deviation of three measurements

As can be clearly seen, application of GBE processing leads to noticeably increased proportions of total CSLBs in AISI type 316L austenitic stainless steel, and the fractions are relevant to the cold rolling reduction. The maximum f_CSL is 58 in 5 GBE material (GBEM), which is above four times as much as the value of the BM. Relatively moderate but noticeable increases are found in strain annealed specimens of medium to high levels of prestrains (20–80). Compared with a medium to high level of strain, a low strain is more prone to inspire CSLBs during subsequent annealing. This phenomenon has been previously discussed in detail elsewhere. ¹⁵ Briefly, it may be attributed to the different stored strain energies resulted from different prestrains. Shimada et al. ⁴ reported that the f_CSL was enhanced to 87 in a 304 stainless steel at the strain of 5 followed by annealing at 1240 K for 72 h, but we could not reproduce the result with a 316L in our study. This may be attributed to the following two reasons since twins play vital roles during the GBCD evolution process. First, the higher stacking fault energy of 316L decreased twinning propensity. ²⁶ Second, the pre-existing twin proportion of initial solid solution state in our study is less than one-fourth of that in their 304. ²⁶ It is also observed from Fig. 2 that, under all rolling conditions, CSLB of type Σ3 is very dominant with other Σ3ⁿ (1<n≤3) frequency nearly negligible. The f_Σ3 has the same variation trend with f_CSL as the strain enhanced from 5 to 80. Earlier studies indicated that Σ3s could be generated by virtue of twin events, ²⁷ which indicates that increase in twin events is expected to introduce more Σ3s. Therefore, the higher f_Σ3 in 5GBEM is indicative of more twin events. Significant increase in summation fraction of higher order twin boundaries (Σ9 and Σ27) is only found in 5GBEM as indicated by arrow in Fig. 2. These issues will be further discussed below.

Research on GBE has shown that the connectivity interruption of random high angle GB (HAB) network based on triple junction distribution is considered to be equally important for improved property. ²⁸ As EBSD reconstruction of grain boundaries can straightforwardly give the information of connectivity interruption of boundary networks, the corresponding maps for GBEMs are presented in Fig. 3. As can be clearly seen, many clusters of SBs, ²⁹ with the size of ∼250 μm, emerge (regions A and B in Fig. 3a) and the HAB networks are effectively disrupted (Fig. 3a’) in the 5GBEM. However, the connectivity of HAB networks for the other three GBEMs is less interrupted than that of 5GBEM, as shown in Fig. 3b’–d’. These differences may be interpreted with the model of migration and interactions of incoherent twin boundaries proposed by Wang and Guo. ³⁰ Briefly, unlike the coherent twins either grain spanning or terminated within a single grain in Fig. 3b–d, the Σ3 boundaries in Fig. 3a are mostly identified as incoherent twin boundaries and generally located in the triple junctions. Incoherent twin boundaries are highly mobile and will therefore promote further encounters with other Σ3s; hence, higher order Σ3ⁿ (Σ9 and Σ27) boundaries can be developed. Their presence, usually as short segments between Σ3s, replaces segments of random GBs. As a result, the triple junctions consisting of Σ3–Σ3–Σ9 or Σ3–Σ9–Σ27 (see region A in Fig. 3a) are introduced, ³¹ and further the connectivity of HAB network in 5GBEM is interrupted significantly. This corresponds to the more twin events and noticeably increased f_Σ9+Σ27 in 5GBEM mentioned above. However, in the case of intermediate to large strain (20–80), HAB may be easily produced during recrystallisation. The extensive migration of newly developed HAB promotes the formation of CTBs with the mechanisms proposed by Mahajan et al. ³² The immobile coherent twins do not contribute to interrupting the connectivity of HAB network directly.

OPEN IN VIEWER

3

Electron backscatter diffraction reconstructed grain boundaries for a, a’ 5GBEM, b, b’ 20GBEM, c, c’ 60GBEM and d, d’ 80GBEM: both special and general high angle boundaries (HABs) are shown in a–d, only HABs are shown in a’–d’

It is worth mentioning here that abnormal grain growth (AGG) happens in the 20GBEM (as shown in region C in Fig. 3b). Abnormal grain growth in 316L stainless steel during further annealing treatment after being compressed to 30 has also been reported. ³³ The AGG in our study may be attributable to the inhomogeneous distribution of local strain coupled with intrinsic misorientation dependent boundary mobility. ³⁴ Evidence has been reported that AGG is beneficial to break down the HAB connectivity. ³⁴ Abnormal grain growth may play a very important role in the evolution of optimised GBCD because a small number of HABs migrate widely for a long time and consequently increases the interaction opportunity with Σ3s and other SBs so as to produce more low energy segments. ⁴ These segments are not necessarily CSLBs but have lower GB energy and thus better property than the initial random boundary. The schematic of optimisation process is presented in Fig. 4. Well distributed low energy segments (indicated by arrows in Fig. 4b) in the GB network create a discontinuous chain of chromium depletion and thus reduce the extent of sensitisation. ⁴ This is a possible reason contributing somewhat to the reduction in DOS in 20GBEM, which will be discussed below.

OPEN IN VIEWER

4

a before abnormal grain growth and b after abnormal grain growth

All specimens were subjected to sensitisation treatments at 948 K for 20 h. The variation of DOS (as assessed by the Ir/Ip ratio resulting from the DL-EPR test) is displayed in Fig. 5. As can be expected, the DOS of the GBEM is greatly reduced compared with that of the BM. Such a general improvement in sensitisation resistance is linked to the optimisation of GBCD. After 20 h of sensitisation, Ir/Ip ratios for the four GBEMs range from 0·011 to 0·116 (Fig. 5b). The lowest DOS is seen in 5GBEM with the highest f_CSL and an effective discontinuity of the HAB network. The 5 and 20GBEM have ratios of 0·011 and 0·049 respectively, which would be classified as only slightly sensitised according to British Standard BS ISO 12732:2006. ³⁵ The other two GBEMs (prestrains of 60 and 80) display ratios of 0·116 and 0·069 respectively, which are above the threshold ratio of 0·05 stipulated by British Standard BS ISO 12732:2006 and would be categorised as sensitised. ³⁵ Unsurprisingly, the BM (a non-GBE microstructure) displays the highest DOS with Ir/Ip ratio of 0·684. Images (SEM) for the BM and 5GBEM after electroetched are shown in Fig. 6. Images (SEM) reveal that the sensitised BM is severely etched with deep and continuous chain of corrosion pits, i.e. chromium depletion, along the boundaries. High angle boundaries are more preferentially sensitised because of selective chromium carbide precipitation due to higher GB energy. ³⁶ Sensitised boundaries provide connected networks for corrosion (Fig. 6a). However, the 5GBEM is just slightly attacked shallowly in the surface with the chain of corrosion pits always interrupted by the substitution of SB segments, which are much more immune from sensitisation (Fig. 6b). This is consistent with the f_CSL and the extent of connectivity of HAB network mentioned above. In spite of the fact that GB plane also shows some effects on the properties of materials,^37,38 CSLBs fraction and GB connectivity are confirmed to be the most efficient and effective ones. Even so, it appears to us that resistance to sensitisation modified by GBE process is not dependent on GB character alone. For instance, as shown in Figs. 2 and 3, the fraction of the CSLBs is the highest for 60GBEM among the three GBEMs (prestrains of 20–80) with similar extent of HAB connectivity; the 60GBEM, however, shows the lowest resistance to sensitisation.

OPEN IN VIEWER

5

a current changes versus activation and reactivation potential obtained on basis of DL-EPR test for BM and thermomechanically processed AISI 316L stainless steels after sensitisation treatment at 948 K for 20 h and b variation of DOS and average grain size with roll reduction ratio; error bars describe standard deviation of three measurements

OPEN IN VIEWER

6

Images (SEM) for a, a’ BM and b, b’ 5GBEM specimens; rectangle areas in a and b are shown at higher magnifications in a’ and b’ respectively

In addition to the evolution of GBCD, grain size, which can also be an important factor affecting the sensitisation process,^17,18 has changed. The average grain size measured by the linear intercept method including twin boundaries is displayed in Fig. 5b. Thermomechanical GBE processing resulted in moderate grain growth as a consequence of strain induced GB migration. The grain size increases from 27·42 μm to a maximum of 52·66 μm in the 5GBE specimen (Fig. 5b). For the GBEMs, the average grain size decreases as rolling reduction enhanced from 5 to 80. A higher rate of nucleation (recrystallisation) during the annealing stage and thus a more fine grained structure may be assumed to occur when a larger scale of deformation is applied. The average grain size of the 60GBEM is the only exception to this trend with the reason unclear. Figure 5b displays the relationship between grain size and DOS for all specimens. No relationship between the DOS and grain size seems to be found. However, closer observation reveals a general decrease in DOS with decreasing average grain size in GBEMs (prestrain of 20–80). The significantly improved resistance to sensitisation for 20GBEM may be partly attributed to the abnormal grain growth and therefore interrupted HAB network mentioned above. Cihal ¹⁷ has concluded that the required effective Cr concentration to suppress sensitisation increases for larger grain sizes for a stainless steel with a given C concentration. In fine grained structures, more chromium carbides will nucleate at the GBs, which restricts the growth and size of each carbide nucleus. Thus the extent of Cr depletion across the GB, i.e. sensitisation, will be reduced.^19,20 Yet, the similar grain size with BM but greatly reduced DOS in 80GBEM just confirms the contribution of GBE structure to suppressing sensitisation again. Thus, the effect of grain size on sensitisation in 5GBEM is masked by the optimum GBCD.

Although a proper model combining the effects of both the GBCD and the grain size remains to be formulated, a message of the present study remains: ‘the influence of changing grain size on the sensitisation behaviour should be borne in mind when assessing sensitisation propensity of GBE processed specimens’. Briefly, minimising the average grain size during GBE optimisation process could have a high potential to improve the sensitisation resistance of a 316L stainless steel.

Conclusions

Grain boundary engineered microstructures were obtained by application of thermomechanical processing. Through the analyses of GB character distribution and sensitisation behaviour, the main conclusions could be summarised.