- Email: [email protected]
The topology of three-dimensional grain boundary network and its influence on stress corrosion crack propagation characteristics in austenitic stainless steel in a simulated BWR environment
Corrosion Science xxx (xxxx) xxx–xxx
Contents lists available at ScienceDirect
Corrosion Science journal homepage: www.elsevier.com/locate/corsci
The topology of three-dimensional grain boundary network and its influence on stress corrosion crack propagation characteristics in austenitic stainless steel in a simulated BWR environment ⁎
Tingguang Liua,b, Shuang Xiab, , Tetsuo Shojic, Qin Baib, Bangxin Zhoub, Yonghao Lua a b c
National Center for Materials Service Safety, University of Science and Technology Beijing, Beijing 100083, China School of Materials Science and Engineering, Shanghai University, Shanghai 200072, China Frontier Research Initiative, NICHe, Tohoku University, Sendai 980-8579, Japan
A R T I C L E I N F O
A B S T R A C T
Keywords: A. Stainless steel B. SEM C. High temperature corrosion C. Stress corrosion C. Interfaces
The intergranular cracking and the grain boundary network along the cracks in an austenitic stainless steel 316L after stress corrosion cracking (SCC) test in simulated BWR water were investigated in terms of three-dimensional characterization. It was found that the twin boundaries not only show a strong resistance to cracking, but they could also prevent their neighboring boundaries from cracking, as the cracking probability is lower for boundaries that have higher fraction of neighboring twin boundaries. The propagation of intergranular SCC can be hindered by triple junctions or quadruple junctions in the presence of twin boundaries.
1. Introduction Stress corrosion cracking (SCC) occurs in materials under the synergistic action of stress and a corrosive environment, as in the case of austenitic stainless steel in pressurized water reactor (PWR) or boiling water reactor (BWR) environment. After a long-term service at the reactor operating conditions or even new materials if heat-treating is not done properly, grain boundaries become a susceptible path to SCC initiation and propagation, and intergranular SCC (IGSCC) is a severe problem in the nuclear industry in particular [1,2]. Therefore, improvement of SCC resistance through grain boundary enhancement are getting more and more attentions in recent years [3–10]. In coincident site lattice (CSL) approach, the extent of grain-to-grain fit (∑-value) between the atoms in the two grains is characterized by the reciprocal of the ratio of the number of ‘coincidence sites’ to the total number of sites. A coincidence site is one that is shared by both grains at the two sides of the grain boundary (GB). For instance, when ∑ equals to 3, it means that there will be one atom for every three atoms that is shared by the two lattices. Therefore a boundary with low Σ would be expected to have a lower energy than one that has a high Σ [11]. GBs with different ∑-values have different lattice-structures and therefore different properties. The twin boundary (TB, i.e. ∑3 boundary) has lower boundary energy, and stronger resistance to impurities segregation/precipitation and intergranular degradation compared with random boundaries (∑ > 29) [3–10,12]. So the concept of GB-
⁎
engineering [13–18] was proposed to improve the GB-related properties by increasing the proportion of special boundaries (low-∑ CSL boundaries, boundaries with ∑ ≤ 29) in materials. Many experiments [3–10,19–22] in high temperature or corrosive solutions have shown that the GB-engineered austenitic stainless steel and Ni-based alloy 690 had remarkably higher resistance to intergranular corrosion/SCC than conventional materials, because the proportion of low-∑ CSL boundaries were increased to about 75% after GB-engineering from the original level, which was about 30 ∼ 50%. However, all these studies were based on two-dimensional (2D) microstructure characterization of the tested alloys, using optical microscope (OM), scanning electron microscope (SEM) and electron backscatter diffraction (EBSD) mapping. GB character distribution in 3D microstructure and its relationship to IGSCC are rarely studied. While 2D studies had shown that some special boundaries have stronger resistance to cracking during SCC test, the effect of these special boundaries on IGSCC was still not understood comprehensively in terms of crack propagation, because it is more important to study the topological characteristics of GB network rather than individual interfaces. The topological characteristics are correlated with not only the proportion of special boundaries but also the distribution of special boundaries in spatial GB network, and it really should be studied in 3D. It is generally found that the IGSCC propagates along random boundaries [3–5]. If the distribution of special boundaries is such a way that breaks up the long-range connectivity of IGSCC-susceptible random
Corresponding author (Shuang Xia) at: P.O. Box 269, 149 Yanchang Road, Shanghai, 200072, China. E-mails: [email protected], [email protected] E-mail addresses: [email protected], [email protected] (S. Xia).
http://dx.doi.org/10.1016/j.corsci.2017.10.003 Received 11 July 2016; Received in revised form 24 September 2017; Accepted 12 October 2017 0010-938X/ © 2017 Elsevier Ltd. All rights reserved.
Please cite this article as: Liu, T., Corrosion Science (2017), http://dx.doi.org/10.1016/j.corsci.2017.10.003
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
Fig. 1. (a) Schematically showing the slices of 3D characterization in the CT sample. (b) The thicknesses distribution of the total of 101 slices.
Fig. 2. Visualizations of the reconstructed 3D microstructures: (a) the reconstructed 3D-EBSD microstructure; (b) the reconstructed 3D-OM microstructure; (c, d) 3D visualizations of the IGSCC crack from different perspective, which were drawn using the 3D-OM data in ImageJ 3D Viewer. (See supplementary file ‘3D-crack’).
3D microstructure [23]. Therefore, further studies on GB network characteristics and their influence on IGSCC propagation by using 3D microstructure characterization are needed. In the present paper, the microstructure and crack of a 316L stainless steel after SCC test in a simulated BWR environment were investigated in three-dimensional by using 3D-EBSD and 3D-OM. Attention was particularly paid to the propagation of cracks through quadruple junctions in the presence of twin boundaries.
boundaries, it is generally expected to hinder IGSCC propagation based on the percolation theory [9,22,23]. For example, triple junctions with two twin boundaries could stop the IGSCC propagation [19,20]. However, how other types of topology structures of GB network affect crack propagation is unclear. For example, how the IGSCC propagates through a quadruple junction. Quadruple junction [24–26] is a spatial structure so that it cannot be observed in 2D maps. Quadruple junction is an assembly of six boundaries between four mutually neighboring grains, and 3 is the maximum number of twin boundaries in quadruple junction [26]. The arrangement of twin boundaries in quadruple junctions should have significant influence on IGSCC propagation. In addition, the observed connectivity of random boundaries depends on whether the observation dimension is 2D or 3D [23,27]. Simulation studies have shown that the connectivity of random boundaries could be broken when proportion of special boundaries is more than 35% in a 2D microstructure [28], but the threshold value is more than 80% for a
2. Material and experiments 2.1. Material The material used in this work is a low-carbon austenitic stainless steel 316L with a chemical composition of 17.16 wt.% Cr, 11.90 wt.% Ni, 1.32 wt.% Mn, 2.08 wt.% Mo, 0.028 wt.% C, 0.37 wt.% Si, 2
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
the previous plastic zone produced by fatigue pre-cracking in air, which would facilitate the transition from transgranular crack front (produced by fatiguing in air) to intergranular stress corrosion cracking. After the transition procedure, the loading mode was switched to a constant loading for SCC test. The initial K value was approximately 30 MPa·m0.5. The loading period of constant load test was 2500 h, and only one sample was tested in this work. 2.3. 3D characterization An intergranular crack mode was produced in the CT specimen during SCC test. In order to investigate the effects of grain boundary network on the crack propagation, a 3D material characterization method, that is serial-sectioning coupled with EBSD (electron backscatter diffraction) mapping and OM (optical microscope) mapping, was performed on the cracked specimen. First, the specimen was mechanically-polished manually under a fixed load and for fixed time to achieve a certain thickness reduction. A Buehler 40–7920 Chemomet Synthetic polishing cloth and 40–6377 MasterPrep polishing suspension (0.05 μm alumina) were used for the polishing, so that a good surface can be produced for EBSD mapping. Second, a micrometer with a precision of 1 μm was used to measure the reduction in thickness. A thickness reduction of 2.5 μm was expected for each slice. Third, microhardness indents were used to mark the region of interest for EBSD collection and OM mapping. Finally, a region of interest including the SCC crack on the polished surface was mapped by using EBSD and OM, respectively. The OM was used to take pictures of the crack path. The EBSD was used to measure the grain boundary characters. A HKL/ Channel 5 EBSD system integrated with a CamScan Apollo 300 field emission scanning electron microscope (SEM) was used here. The EBSD field of view was 1000 μm × 1000 μm with a step size of 2.5 μm. Repeated operations of the four steps were performed on the SCC tested 316L sample, and only one position, embracing the entire SCC crack, was characterized considering that the 3D characterization is time consuming. A total of 101 parallel slices were mapped for the SCC sample, as schematically shown in Fig. 1(a). The average thickness per slice was 2.65 μm, so volume of the 3D-EBSD microstructure was 1000 × 1000 × 267.65 μm3 with a voxel of 2.5 × 2.5 × 2.65 μm3. Fig. 1(b) shows the distribution of thicknesses for each slices. They are between 0 and 5 μm, and most of them are about 2 or 3 μm. The thickness resolution is relatively high in comparison with the average grain size of 52 μm. Post-processing of the measured 2D EBSD data on the 101 sections included 3D reconstruction, 3D visualization and quantification of 3D microstructural features [29–31]. Oxford-HKL Channel 5, EDAX-TSL OIM Analysis, Dream3D (v4.2.5004) [32–35], ParaView (v4.3.1) [36], HDFView and Matlab were combined to process the 3D-EBSD data in this work. The 2D-EBSD data (‘.ctf’ of Oxford-HKL or ‘.ang’ of EDAXTSL) were transformed into a native Dream3D file format, and then they were reconstructed into a 3D-EBSD data in Dream3D [32–35]. The 3D reconstruction procedure includes sections alignment, noise reduction, grain and grain boundary identification and boundary smoothing. Subsequently, the 3D data was visualized in ParaView. The regions of interest can be visualized in 3D, such as individual grain or boundary, and assemblies of certain grains or boundaries. Thirdly, the 3D microstructure should be studied quantitatively. Dream3D was used to quantify the grains, such as grain size (volume and equivalent sphere diameter), grain shape parameters, orientations, number of nearest neighbors and so on. In-house developed Matlab programs were used to quantify the grain boundaries, including the boundary area, grain surface area, numbers of neighboring boundaries and twin boundaries per boundary, and twin boundary proportions. The 3D-EBSD data could provide the crystallographic information and morphologies based on orientations, but it cannot determine the SCC crack path. The pictures of crack path on the 101 sections were taken using OM. These 2D OM maps can be reconstructed into a 3D-OM
Fig. 3. Distributions of the average grain diameters (a) and the twin boundary fractions (b) of three 2D-EBSD maps (slice 20th, slice 50th and slice 80th) and the 3D-EBSD microstructure.
0.005 wt.% S, 0.044 wt.% P and Fe in balance. The as-received material was hot rolled by 50% reduction of thickness at a starting temperature of 1200 °C, and then it was solution annealed at 1050 °C for 2.5 h and water quenching to produce a fully recrystallized microstructure with an average grain diameter of 52 μm (in 2D). The final thickness of the produced material was 80 mm approximately.
2.2. Stress corrosion cracking A compact tension (CT) specimen was machined from the as-prepared 316L stainless steel. It was a standard 1T CT (thickness B = 25 mm) with T-L orientation according to ASTM E399, where ‘L’ means longitudinal direction (rolling direction), and ‘T’ means long transverse direction. The main crack should propagate along the rolling direction during SCC test. Firstly, the CT specimen was pre-cracked by fatigue under a sinewave loading in air. The extension of the fatigue crack was about 2 mm. Subsequently, the specimen was fixed in an autoclave. The testing environment was simulated BWR primary water: deionized water, temperature at 288 °C, dissolved oxygen concentration (DO) at 20 ppm, pressure at 8 MPa. Before the SCC test started, a transition procedure, using a fatigue loading of a triangular-wave form at frequency of 0.01 Hz for 100 h and 0.001 Hz for 100 h, respectively, was performed in the testing environment in autoclave. The maximum stress intensity factor Kmax was about 30 MPa·m0.5. The load ratio (Kmin/Kmax) was 0.5. The transition procedure was employed to propagate the crack through 3
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
Fig. 4. Examples of cracked triple junctions. (a) The OM map and EBSD image of the 20th slice. The two highlighted regions ‘b’ and ‘c’ are accordance with figure b and c, respectively. (b) A cracked triple junction b262-b891-b307 and its topology structure. In the figure ‘b262′ is a grain boundary ID; ‘g93′ is a grain ID. In the topology structure, points represent grains; lines represent grain boundary, and cracked boundaries are represented by broken lines. (c) A cracked triple junction b117-b950-b1172 and its topology structure.
Fig. 5. (a) Sketch and topology structure of quadruple junction (QJ). (b) Topologically showing the modes of IGSCC propagating through quadruple junctions. As an example, QJC21 means the quadruple junction with two cracked boundaries, and the subscript number means isomerism. (c, d) The 3D schematic maps of cracking modes QJC33 and QJC41.
circle diameter) is 52 μm for the 20th, 50th and 80th 2D EBSD maps. The 3D-EBSD microstructure contains 4672 grain boundaries, including the boundaries that intersect the surfaces of the observed block. The average boundary size (equivalent circle diameter) is 33.5 μm. There are 3737 random boundaries and 935 twin boundaries. Fig. 3(b) shows the fractions of twin boundaries in the 20th, 50th and 80th 2D EBSD maps, and the 3D-EBSD volume, respectively. The measured area fraction of twin boundaries in the 3D microstructure is 41.6%, which is smaller than the length fractions in the 2D EBSD maps. In addition, the quantity fraction of twin boundaries is only 20.0%, which is much smaller than the area fraction, because the twin boundaries are larger than the random boundaries on average. The average random boundary size is 29.6 μm, and it is 49.0 μm for twin boundaries.
data by using a “3D Viewer” plugin [37] of ImageJ software [38–40]. 3. Results The main crack length after SCC testing in simulated BWR water for 2500 h is about 0.6 mm. The microstructure beside the crack was characterized in 3D by using 3D-EBSD coupled with 3D-OM. The reconstructed 3D microstructures are shown in Fig. 2. The 3D-EBSD microstructure contains 1105 grains, which includes 574 completely internal grains and 531 grains that intersect the surfaces of observed block. The average 3D grain size (equivalent sphere diameter) is 40.7 μm (including border grains), which is smaller than the measured 2D grain size, as shown in Fig. 3(a). The average grain size (equivalent 4
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
Fig. 6. An example of quadruple junction with three twin boundaries (3T-QJ) from the 3D-EBSD microstructure: (a) the four grains of the quadruple junction, having diameters of 159.1 μm (A), 78.8 μm (B), 135.1 μm (C) and 121.5 μm (D), respectively; (b) the six boundaries of the quadruple junction, in which the boundaries colored by dark red, green and pink are twin boundaries; (c) the topology structure of the quadruple junction, where points and lines were colored according to the related grains and grain boundaries, respectively. (See supplementary file ‘3D-QJ’). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
4.1. Effects of topological characteristics of triple junctions on IGSCC
The 3D-OM technology was used to map the IGSCC propagation path in 3D. Fig. 2(b) shows the 3D morphology, reconstructed using ImageJ 3D Viewer [37], from the 101 slices of OM observation. Furthermore, the 3D path of the branched crack through the depth of the observed block is shown in Fig. 2(c & d). The intergranular morphology of the 3D crack is evident. The grain boundary characters along the 3D crack path was investigated in 2D cross sectional planes at different depth. For example, Fig. 4(a) shows the OM and EBSD map of the 20th slice. It can be seen that most cracked boundaries are random boundaries. The twin boundaries scarcely cracked due to its strong resistance to IGSCC, as expected [3–5]. An example marked by a blue circle is shown in Fig. 4(a). The two random boundaries are cracked, and a short ∑9 boundary is partially cracked. The crack should propagate forward along the two twin boundaries because they are mechanically favorable oriented to the main crack propagation direction. However, this crack was stopped here. Except for the twin boundaries, other types of low-∑ CSL boundaries did not demonstrate special resistance to IGSCC, such as the arrows pointed boundaries in Fig. 4(a). The red arrow pointed boundaries is ∑23, and the black arrow pointed boundary is ∑13a. Both the two boundaries cracked during SCC.
Triple junction is an assembly of three boundaries between three mutually neighboring grains [14,23,41], for examples b262-b891-b307 and b950-b1172-b1170 in Fig. 4(b & c). The three boundaries has a common line. Triple junction can be topologically shown as a triangle, in which vertexes represent grains, and the lines between the vertexes represent grain boundaries. The broken lines in the topology structures in Fig. 4 mean cracked boundaries. A triple junction has two twin boundaries at most, and the third boundary is ∑9 in this case [24,26]. The ∑9 boundary in a triple junction with two twin boundaries should be hindered to crack because the three grains at the triple junction are bound by the two twin boundaries. From 3D results of this work, all observed ∑9 boundaries in ∑3-∑3-∑9 type of triple junctions are not cracked or just partially cracked from the side that opposites to the two twin boundaries. Therefore, the topological characteristics of triple junction have significant influence on the cracking resistance of boundaries of the triple junction. Fig. 4(c) shows a cracked triple junction, in which all the three boundaries are cracked. The 2D EBSD map shows the boundary b1172 is a low-∑ CSL boundary (∑13a), but it still cracked during SCC. Fig. 4(b) shows another cracked triple junction, in which the two boundaries b262 and b891 are cracked though b262 is a ∑17b boundary, but the random boundary b307 is uncracked though it seems to be more mechanically favorable oriented to IGSCC propagation compared with the boundary b891. Diameter of equivalent area of the boundary b307 is 98.2 μm. It has 20 neighboring boundaries which mean linearly connected boundaries with b307, and there are 2 twin boundaries. The 3D diameter of boundary b891 is 60.0 μm. It has 8 neighboring boundaries, including 2 twin boundaries. Neither of the two boundaries, b307 and b891, construct triple junctions with two twin boundaries. However, the boundary b307 is incorporated into two 2T-QJs (quadruple junction with two twin boundaries) but the boundary b891 is incorporated into only one 2T-QJ. The 2T-QJ could hinder the IGSCC propagation to some extent as discussed below. This may be a possible reason why the boundary b307 did not crack during SCC.
4. Discussion 3D characterization of the SCC tested 316L stainless steel shows that the crack propagated along GB network. Twin boundaries demonstrate a strong resistance to cracking, as reported in 2D studies [3–5]. However, a shift in focus from single boundary statistics to the topological characteristics of the entire GB network is need to reveal the mechanism of improving IGSCC resistance of materials by GB engineering process. Triple junction [14,23,41] and quadruple junction [24,25] are two typical topological structures of GB network. Effects of the topological characteristics of triple junctions and quadruple junctions in the presence of twin boundaries on the IGSCC propagation need investigation. It is well know the IGSCC would be stopped at triple junctions with two twin boundaries, but it is unclear how the IGSCC propagates through quadruple junctions. Triple junctions can be observed in 2D maps of cross-sections, but the quadruple junctions have to be studied in 3D. 5
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
Fig. 7. Topological modes of IGSCC propagating through the five types of quadruple junctions based on the quantity and arrangement of twin boundaries. The abbreviations of cracking modes at the left column are in accordance with Fig. 5(b). The top row shows the abbreviations and topology structures of the five types of quadruple junctions: 0T-QJ, 1T-QJ, 2T-QJ1, 2T-QJ2 and 3T-QJ.
account. For example, all the three modes, QJC31, QJC32 and QJC33, have three cracked boundaries, but they have different topology structure as well as different effectiveness to destroy the quadruple junction. A grain would be separated from the other grains when the case of QJC33 occurred, but the four grains would be still assembled when QJC31 or QJC32 occurred. Fig. 5(c) and (d) shows the 3D sketch maps of cracking modes QJC33 and QJC41. In Fig. 5, however, the twin boundaries were not taken into account. The twin boundaries have strong resistance to cracking, so some cracking modes of Fig. 5 should be excluded when the quadruple junction has twin boundaries. Gertsman [26] had studied the CSL theory of quadruple junctions, obtaining that the maximum number of twin boundaries in a quadruple junction is three, with two ∑9 boundaries and one ∑27 boundary in this case. For example, the quadruple junction in Fig. 6, boundaries AB, BC and CD are twin boundaries, and boundaries AC and AB are ∑9, and AD is a ∑27 boundary. The twin
4.2. Effects of topological characteristics of quadruple junctions on IGSCC Quadruple junction [24,25] is an assembly of six boundaries between four mutually neighboring grains, as schematically shown in Fig. 5(a). The four grains or the six boundaries have a common point which is generally named quadruple node. Quadruple junction is a spatial structure so that it cannot be seen in 2D maps of cross-sections. A topology method, mutually connected four points, was used to illustrate the quadruple junction, as shown in Fig. 5(a). In the topology structure, points represent grains, and lines represent boundaries. Fig. 6 shows an example of quadruple junction from the 3D-EBSD microstructure. It has three twin boundaries. The modes of IGSCC propagating through quadruple junctions were classified into ten types based on the cracked boundary number and their arrangement, as topologically shown in Fig. 5(b). The broken lines represent cracked boundaries, and isomerism structure was taken into 6
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
Fig. 8. Two examples of cracked quadruple junctions and their topology structures. The boundaries were differentiated by colors. Red lines in the topology structures represent twin boundaries, i.e. t1972, t824 and t1169. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
boundaries are cracked but the two twin boundaries. In addition, if the former five types of cracking modes occur, i.e. QJC1, QJC21, QJC22, QJC31 and QJC32, as shown in Fig. 7, the four grains of the quadruple junction will still be bound by the non-cracked boundaries, so the five cracking modes cannot destroy quadruple junctions thoroughly. However, if the latter five type of cracking modes occur, i.e. QJC33, QJC41, QJC42, QJC5 and QJC6, the four grains will be separated, and then quadruple junctions are broken thoroughly. Furthermore, if a quadruple junction has three twin boundaries (3TQJ), IGSCC would never propagate through it following the latter five types of cracking modes. The four grains of the 3T-QJ are bound by the 3 twin boundaries, so that 3T-QJ can hinder the IGSCC propagation. Even a quadruple junction has two twin boundaries (2T-QJ), only one or two of the later five cracking modes could occur. Therefore, quadruple junctions with three twin boundaries could stop the IGSCC propagation, and even quadruple junctions with two twin boundaries have considerable resistance to IGSCC propagation. Fig. 9. Statistics of the quantity of neighboring twin boundaries (TBs) and the quantity fraction of twin boundaries in total of its neighboring boundaries for randomly selected 61 grain boundaries. 44 of them are cracked boundaries, and the other 17 boundaries are uncracked but their positions are beside the IGSCC propagation path.
4.3. Neighboring twin boundaries effect on cracking of boundary The cracking susceptibility of a boundary depends on not only the character (∑-value of CSL model) of this boundary but also the characters of its neighboring boundaries. For example, a random boundary that surrounded by twin boundaries would never crack during IGSCC, because the crack-resistant twin boundaries would prevent the IGSCC from propagating to the random boundary. The boundary that has more neighboring twin boundaries should have higher resistance to IGSCC. Fig. 9 shows statistics of the quantity of neighboring twin boundaries and the ratio of neighboring twin boundary quantity to the total number of neighboring boundaries. In total, 61 boundaries that randomly selected were measured, including 44 cracked boundaries and 17 uncracked boundaries that beside the crack path. The points in Fig. 9 seem to be scattered. It seems that the cracking susceptibility of a boundary has no obvious correlation with the quantity of neighboring twin boundaries. However, the boundary with the highest quantity of neighboring twin boundaries, meanwhile, having the highest quantity fraction of neighboring twin boundaries, is uncracked. In comparison, the boundaries having the lowest fractions of neighboring twin boundaries are cracked. In addition, the average number of neighboring twin boundaries for the 44 cracked boundaries is 2.34, and it is 3.00 for the 17 uncracked boundaries. The uncracked boundaries have more neighboring twin boundaries on average. The average ratio of neighboring twin boundary quantity to the total number of neighboring boundaries for the 44 cracked GBs is 0.17, and it is 0.23 for the 17 uncracked boundaries. The uncracked boundaries have higher fraction of neighboring twin boundaries as well than the cracked boundaries.
boundaries were identified automatically by Dream3D, and the allowed deviation was defined according to Brandon Criterion [42]. Considering quantity and arrangement of twin boundaries, quadruple junctions can be classified into five types: 0T-QJ, 1T-QJ, 2T-QJ1, 2T-QJ2 and 3T-QJ, as shown in Fig. 7. Isomerism was considered here, because both 2TQJ1 and 2T-QJ2 have two twin boundaries but different topology structures. The modes of IGSCC propagating through the five types of quadruple junctions were laid out in Fig. 7. For 0T-QJ (quadruple junction without twin boundary), all the ten cracking modes can occur, but some cracking modes will not occur for other types of quadruple junctions if twin boundaries are supposed non-crackable. For example, the 3T-QJ has only four potential cracking modes. The 2T-QJ1 and 2T-QJ2 have six and eight cracking modes, respectively, though both them have two twin boundaries. Therefore, the topological characteristics of quadruple junctions have significant influence on the cracking resistance of IGSCC propagation, and not only the quantity but also the arrangement of twin boundaries in GB network should be considered. Fig. 8 shows two examples of cracked quadruple junctions along the IGSCC propagation path in the 316L sample. All the five non-twin boundaries of the quadruple junction of Fig. 8(a) are cracked except a twin boundary, so it follows QJC5. The quadruple junction of Fig. 8(b) follows the cracking mode QJC41, in which all the four non-twin 7
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
[9] S. Kobayashi, R. Kobayashi, T. Watanabe, Control of grain boundary connectivity based on fractal analysis for improvement of intergranular corrosion resistance in SUS316L austenitic stainless steel, Acta Mater. 102 (2016) 397–405. [10] A. Stratulat, J.A. Duff, T.J. Marrow, Grain boundary structure and intergranular stress corrosion crack initiation in high temperature water of a thermally sensitised austenitic stainless steel, observed in situ, Corros. Sci. 85 (2014) 428–435. [11] V. Randle, The Role of the Coincidence Site Lattice in Grain Boundary Engineering, Cambridge University Press, London, 1996. [12] H. Li, S. Xia, B. Zhou, W. Chen, C. Hu, The dependence of carbide morphology on grain boundary character in the highly twinned Alloy 690, J. Nucl. Mater. 399 (2010) 108–113. [13] T. Watanabe, Approch to grain boundary design for strong and ductile polycrystals, Res. Mech. 11 (1984) 47–84. [14] V. Randle, Twinning-related grain boundary engineering, Acta Mater. 52 (2004) 4067–4081. [15] S.A. Xia, B.X. Zhou, W.J. Chen, Effect of single-step strain and annealing on grain boundary character distribution and intergranular corrosion in Alloy 690, J. Mater. Sci. 43 (2008) 2990–3000. [16] J. Lind, S.F. Li, M. Kumar, Twin related domains in 3D microstructures of conventionally processed and grain boundary engineered materials, Acta Mater. 114 (2016) 43–53. [17] T. Liu, S. Xia, B. Wang, Q. Bai, B. Zhou, C. Su, Grain orientation statistics of grainclusters and the propensity of multiple-twinning during grain boundary engineering, Mater. Des. 112 (2016) 442–448. [18] A. Das, Grain boundary engineering: fatigue fracture, Philos. Mag. 97 (2017) 867–916. [19] S. Xia, H. Li, T.G. Liu, B.X. Zhou, Appling grain boundary engineering to Alloy 690 tube for enhancing intergranular corrosion resistance, J. Nucl. Mater. 416 (2011) 303–310. [20] C.L. Hu, S. Xi, H. Li, T.G. Liu, B.X. Zhou, W.J. Chen, N. Wang, Improving the intergranular corrosion resistance of 304 stainless steel by grain boundary network control, Corros. Sci. 53 (2011) 1880–1886. [21] F. Shi, P.C. Tian, N. Jia, Z.H. Ye, Y. Qi, C.M. Liu, X.W. Li, Improving intergranular corrosion resistance in a nickel-free and manganese-bearing high-nitrogen austenitic stainless steel through grain boundary character distribution optimization, Corros. Sci. 107 (2016) 49–59. [22] M. Michiuchi, H. Kokawa, Z.J. Wang, Y.S. Sato, K. Sakai, Twin-induced grain boundary engineering for 316 austenitic stainless steel, Acta Mater. 54 (2006) 5179–5184. [23] M. Frary, C.A. Schuh, Connectivity and percolation behaviour of grain boundary networks in three dimensions, Philos. Mag. 85 (2005) 1123–1143. [24] V. Gertsman, Coincidence site lattice theory of multicrystalline ensembles, Acta Crystallogr. A 57 (2001) 649–655. [25] S.F. Li, J.K. Mason, J. Lind, M. Kumar, Quadruple nodes and grain boundary connectivity in three dimensions, Acta Mater. 64 (2014) 220–230. [26] V.Y. Gertsman, Coincidence site lattice theory of triple junctions and quadruple points, Science and Technology of Interfaces, John Wiley & Sons Inc., 2002, pp. 387–398. [27] E.A. Holm, P.M. Duxbury, Three-dimensional materials science, Scripta Mater. 54 (2006) 1035–1040. [28] C.A. Schuh, R.W. Minich, M. Kumar, Connectivity and percolation in simulated grain-boundary networks, Philos. Mag. 83 (2003) 711–726. [29] M.A. Groeber, B.K. Haley, M.D. Uchic, D.M. Dimiduk, S. Ghosh, 3D reconstruction and characterization of polycrystalline microstructures using a FIB-SEM system, Mater. Charact. 57 (2006) 259–273. [30] A.D. Rollett, S.B. Lee, R. Campman, G.S. Rohrer, Three-dimensional characterization of microstructure by electron back-scatter diffraction, Annu. Rev. Mater. Res. 37 (2007) 627–658. [31] S. Zaefferer, S.I. Wright, D. Raabe, Three-dimensional orientation microscopy in a focused ion beam-scanning electron microscope: a new dimension of microstructure characterization, Metall. Mater. Trans. A 39a (2008) 374–389. [32] M. Groeber, M. Jackson, DREAM.3D: a digital representation environment for the analysis of microstructure in 3D, Integr. Mater. Manuf. Innov. 3 (2014) 1–17. [33] M. Groeber, S. Ghosh, M.D. Uchic, D.M. Dimiduk, A framework for automated analysis and simulation of 3D polycrystalline microstructures.: Part 1: statistical characterization, Acta Mater. 56 (2008) 1257–1273. [34] M. Groeber, S. Ghosh, M. Uchic, D. Dimiduk, A framework for automated analysis and simulation of 3d polycrystalline microstructures. Part 2: synthetic structure generation, Acta Mater. 56 (2008) 1257–1273. [35] Y. Bhandari, S. Sarkar, M. Groeber, M.D. Uchic, D.M. Dimiduk, S. Ghosh, 3D polycrystalline microstructure reconstruction from FIB generated serial sections for FE analysis, Comp. Mater. Sci. 41 (2007) 222–235. [36] U. Ayachit, The ParaView Guide: A Parallel Visualization Application, Kitware, 2015. [37] B. Schmid, J. Schindelin, A. Cardona, M. Longair, M. Heisenberg, A high-level 3D visualization API for java and ImageJ, BMC Bioinf. 11 (2010). [38] T.J. Collins, ImageJ for microscopy, Biotechniques 43 (2007) 25–30. [39] M. Abramoff, P. Magalhaes, Image processing with ImageJ, Biophoton. Int. (2004). [40] A. Ullah, G.Q. Liu, H. Wang, M. Khan, D.F. Khan, J.H. Luan, Optimal approach of three-dimensional microstructure reconstructions and visualizations, Mater. Exp. 3 (2013) 109–118. [41] G. Gottstein, L.S. Shvindlerman, Grain boundary junction engineering, Scripta Mater. 54 (2006) 1065–1070. [42] D.G. Brandon, The structure of high-angle grain boundaries, Acta Metall. 14 (1966) 1479–1484.
Furthermore, compared with the quantity of neighboring twin boundaries, the cracking susceptibility of boundary is more strongly affected by the quantity fraction. The cracking susceptibility is approximately lower for boundaries that have higher fraction of neighboring twin boundaries. 5. Conclusions A crack in a solution annealed 316L stainless steel after SCC test in simulated BWR water was characterized in three-dimensional by using 3D-EBSD coupled with 3D-OM. The effects of the topological characteristics of grain boundary network on the IGSCC propagation were investigated in 3D. It was found that the twin boundaries not only have a strong resistance to IGSCC themselves, but they could also prevent the neighboring boundaries from cracking, as the cracking probability is lower for boundaries that have higher fraction of neighboring twin boundaries. The IGSCC propagation can be hindered by triple junctions or quadruple junctions in the presence of twin boundaries. Triple junction or quadruple junction with more twin boundaries plays an important role to improve the IGSCC resistance of materials. Triple junctions with two twin boundaries can stop the IGSCC propagation. Quadruple junctions with three twin boundaries can stop the IGSCC propagation, and even quadruple junction with two twin boundaries has a considerable resistance to IGSCC propagation. Acknowledgments This work was supported by the National Natural Science Foundation of China (NSFC) (grant number 51671122), the Shanghai Science and Technology Commission (grant number 13520500500) and the fund from Jiuli Hi-Tech Metals Co., Ltd. The authors gratefully thank Gregory S. Rohrer and Xiaoting Zhong (Carnegie Mellon University, USA) for their helps on Dream3D and ParaView, and thank Asad Ullah (Karakoram International University, Pakistan) for his help on ImageJ 3D Viewer, and thank Wenyue Zheng (Natural Resources Canada. University of Science and Technology Beijing) for his help to improve the English. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.corsci.2017.10.003. References [1] S.M. Bruemmer, G.S. Was, Microstructural and microchemical mechanisms controlling intergranular stress corrosion cracking in light-water-reactor systems, J. Nucl. Mater. 216 (1994) 348–363. [2] X. Zhong, S.C. Bali, T. Shoji, Accelerated test for evaluation of intergranular stress corrosion cracking initiation characteristics of non-sensitized 316 austenitic stainless steel in simulated pressure water reactor environment, Corros. Sci. 115 (2017) 106–117. [3] B. Alexandreanu, B. Capell, G.S. Was, Combined effect of special grain boundaries and grain boundary carbides on IGSCC of Ni-16Cr-9Fe-xC alloys, Mat. Sci. Eng. A Struct. 300 (2001) 94–104. [4] V.Y. Gertsman, S.M. Bruemmer, Study of grain boundary character along intergranular stress corrosion crack paths in austenitic alloys, Acta Mater. 49 (2001) 1589–1598. [5] E.A. West, G.S. Was, IGSCC of grain boundary engineered 316L and 690 in supercritical water, J. Nucl. Mater. 392 (2009) 264–271. [6] A. Telang, A.S. Gill, D. Tammana, X. Wen, M. Kumar, S. Teysseyre, S.R. Mannava, D. Qian, V.K. Vasudevan, Surface grain boundary engineering of Alloy 600 for improved resistance to stress corrosion cracking, Mater. Sci. Eng. A 648 (2015) 280–288. [7] Z. Zhang, S. Xia, W. Cao, H. Li, B. Zhou, Q. Bai, Effects of grain boundary character on intergranular stress corrosion cracking initiation in 316 stainless steel, Acta Metall. Sin. 52 (2016) 313–319. [8] A. Telang, A.S. Gill, M. Kumar, S. Teysseyre, D. Qian, S.R. Mannava, V.K. Vasudevan, Iterative thermomechanical processing of alloy 600 for improved resistance to corrosion and stress corrosion cracking, Acta Mater. 113 (2016) 180–193.
8
Contents lists available at ScienceDirect
Corrosion Science journal homepage: www.elsevier.com/locate/corsci
The topology of three-dimensional grain boundary network and its influence on stress corrosion crack propagation characteristics in austenitic stainless steel in a simulated BWR environment ⁎
Tingguang Liua,b, Shuang Xiab, , Tetsuo Shojic, Qin Baib, Bangxin Zhoub, Yonghao Lua a b c
National Center for Materials Service Safety, University of Science and Technology Beijing, Beijing 100083, China School of Materials Science and Engineering, Shanghai University, Shanghai 200072, China Frontier Research Initiative, NICHe, Tohoku University, Sendai 980-8579, Japan
A R T I C L E I N F O
A B S T R A C T
Keywords: A. Stainless steel B. SEM C. High temperature corrosion C. Stress corrosion C. Interfaces
The intergranular cracking and the grain boundary network along the cracks in an austenitic stainless steel 316L after stress corrosion cracking (SCC) test in simulated BWR water were investigated in terms of three-dimensional characterization. It was found that the twin boundaries not only show a strong resistance to cracking, but they could also prevent their neighboring boundaries from cracking, as the cracking probability is lower for boundaries that have higher fraction of neighboring twin boundaries. The propagation of intergranular SCC can be hindered by triple junctions or quadruple junctions in the presence of twin boundaries.
1. Introduction Stress corrosion cracking (SCC) occurs in materials under the synergistic action of stress and a corrosive environment, as in the case of austenitic stainless steel in pressurized water reactor (PWR) or boiling water reactor (BWR) environment. After a long-term service at the reactor operating conditions or even new materials if heat-treating is not done properly, grain boundaries become a susceptible path to SCC initiation and propagation, and intergranular SCC (IGSCC) is a severe problem in the nuclear industry in particular [1,2]. Therefore, improvement of SCC resistance through grain boundary enhancement are getting more and more attentions in recent years [3–10]. In coincident site lattice (CSL) approach, the extent of grain-to-grain fit (∑-value) between the atoms in the two grains is characterized by the reciprocal of the ratio of the number of ‘coincidence sites’ to the total number of sites. A coincidence site is one that is shared by both grains at the two sides of the grain boundary (GB). For instance, when ∑ equals to 3, it means that there will be one atom for every three atoms that is shared by the two lattices. Therefore a boundary with low Σ would be expected to have a lower energy than one that has a high Σ [11]. GBs with different ∑-values have different lattice-structures and therefore different properties. The twin boundary (TB, i.e. ∑3 boundary) has lower boundary energy, and stronger resistance to impurities segregation/precipitation and intergranular degradation compared with random boundaries (∑ > 29) [3–10,12]. So the concept of GB-
⁎
engineering [13–18] was proposed to improve the GB-related properties by increasing the proportion of special boundaries (low-∑ CSL boundaries, boundaries with ∑ ≤ 29) in materials. Many experiments [3–10,19–22] in high temperature or corrosive solutions have shown that the GB-engineered austenitic stainless steel and Ni-based alloy 690 had remarkably higher resistance to intergranular corrosion/SCC than conventional materials, because the proportion of low-∑ CSL boundaries were increased to about 75% after GB-engineering from the original level, which was about 30 ∼ 50%. However, all these studies were based on two-dimensional (2D) microstructure characterization of the tested alloys, using optical microscope (OM), scanning electron microscope (SEM) and electron backscatter diffraction (EBSD) mapping. GB character distribution in 3D microstructure and its relationship to IGSCC are rarely studied. While 2D studies had shown that some special boundaries have stronger resistance to cracking during SCC test, the effect of these special boundaries on IGSCC was still not understood comprehensively in terms of crack propagation, because it is more important to study the topological characteristics of GB network rather than individual interfaces. The topological characteristics are correlated with not only the proportion of special boundaries but also the distribution of special boundaries in spatial GB network, and it really should be studied in 3D. It is generally found that the IGSCC propagates along random boundaries [3–5]. If the distribution of special boundaries is such a way that breaks up the long-range connectivity of IGSCC-susceptible random
Corresponding author (Shuang Xia) at: P.O. Box 269, 149 Yanchang Road, Shanghai, 200072, China. E-mails: [email protected], [email protected] E-mail addresses: [email protected], [email protected] (S. Xia).
http://dx.doi.org/10.1016/j.corsci.2017.10.003 Received 11 July 2016; Received in revised form 24 September 2017; Accepted 12 October 2017 0010-938X/ © 2017 Elsevier Ltd. All rights reserved.
Please cite this article as: Liu, T., Corrosion Science (2017), http://dx.doi.org/10.1016/j.corsci.2017.10.003
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
Fig. 1. (a) Schematically showing the slices of 3D characterization in the CT sample. (b) The thicknesses distribution of the total of 101 slices.
Fig. 2. Visualizations of the reconstructed 3D microstructures: (a) the reconstructed 3D-EBSD microstructure; (b) the reconstructed 3D-OM microstructure; (c, d) 3D visualizations of the IGSCC crack from different perspective, which were drawn using the 3D-OM data in ImageJ 3D Viewer. (See supplementary file ‘3D-crack’).
3D microstructure [23]. Therefore, further studies on GB network characteristics and their influence on IGSCC propagation by using 3D microstructure characterization are needed. In the present paper, the microstructure and crack of a 316L stainless steel after SCC test in a simulated BWR environment were investigated in three-dimensional by using 3D-EBSD and 3D-OM. Attention was particularly paid to the propagation of cracks through quadruple junctions in the presence of twin boundaries.
boundaries, it is generally expected to hinder IGSCC propagation based on the percolation theory [9,22,23]. For example, triple junctions with two twin boundaries could stop the IGSCC propagation [19,20]. However, how other types of topology structures of GB network affect crack propagation is unclear. For example, how the IGSCC propagates through a quadruple junction. Quadruple junction [24–26] is a spatial structure so that it cannot be observed in 2D maps. Quadruple junction is an assembly of six boundaries between four mutually neighboring grains, and 3 is the maximum number of twin boundaries in quadruple junction [26]. The arrangement of twin boundaries in quadruple junctions should have significant influence on IGSCC propagation. In addition, the observed connectivity of random boundaries depends on whether the observation dimension is 2D or 3D [23,27]. Simulation studies have shown that the connectivity of random boundaries could be broken when proportion of special boundaries is more than 35% in a 2D microstructure [28], but the threshold value is more than 80% for a
2. Material and experiments 2.1. Material The material used in this work is a low-carbon austenitic stainless steel 316L with a chemical composition of 17.16 wt.% Cr, 11.90 wt.% Ni, 1.32 wt.% Mn, 2.08 wt.% Mo, 0.028 wt.% C, 0.37 wt.% Si, 2
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
the previous plastic zone produced by fatigue pre-cracking in air, which would facilitate the transition from transgranular crack front (produced by fatiguing in air) to intergranular stress corrosion cracking. After the transition procedure, the loading mode was switched to a constant loading for SCC test. The initial K value was approximately 30 MPa·m0.5. The loading period of constant load test was 2500 h, and only one sample was tested in this work. 2.3. 3D characterization An intergranular crack mode was produced in the CT specimen during SCC test. In order to investigate the effects of grain boundary network on the crack propagation, a 3D material characterization method, that is serial-sectioning coupled with EBSD (electron backscatter diffraction) mapping and OM (optical microscope) mapping, was performed on the cracked specimen. First, the specimen was mechanically-polished manually under a fixed load and for fixed time to achieve a certain thickness reduction. A Buehler 40–7920 Chemomet Synthetic polishing cloth and 40–6377 MasterPrep polishing suspension (0.05 μm alumina) were used for the polishing, so that a good surface can be produced for EBSD mapping. Second, a micrometer with a precision of 1 μm was used to measure the reduction in thickness. A thickness reduction of 2.5 μm was expected for each slice. Third, microhardness indents were used to mark the region of interest for EBSD collection and OM mapping. Finally, a region of interest including the SCC crack on the polished surface was mapped by using EBSD and OM, respectively. The OM was used to take pictures of the crack path. The EBSD was used to measure the grain boundary characters. A HKL/ Channel 5 EBSD system integrated with a CamScan Apollo 300 field emission scanning electron microscope (SEM) was used here. The EBSD field of view was 1000 μm × 1000 μm with a step size of 2.5 μm. Repeated operations of the four steps were performed on the SCC tested 316L sample, and only one position, embracing the entire SCC crack, was characterized considering that the 3D characterization is time consuming. A total of 101 parallel slices were mapped for the SCC sample, as schematically shown in Fig. 1(a). The average thickness per slice was 2.65 μm, so volume of the 3D-EBSD microstructure was 1000 × 1000 × 267.65 μm3 with a voxel of 2.5 × 2.5 × 2.65 μm3. Fig. 1(b) shows the distribution of thicknesses for each slices. They are between 0 and 5 μm, and most of them are about 2 or 3 μm. The thickness resolution is relatively high in comparison with the average grain size of 52 μm. Post-processing of the measured 2D EBSD data on the 101 sections included 3D reconstruction, 3D visualization and quantification of 3D microstructural features [29–31]. Oxford-HKL Channel 5, EDAX-TSL OIM Analysis, Dream3D (v4.2.5004) [32–35], ParaView (v4.3.1) [36], HDFView and Matlab were combined to process the 3D-EBSD data in this work. The 2D-EBSD data (‘.ctf’ of Oxford-HKL or ‘.ang’ of EDAXTSL) were transformed into a native Dream3D file format, and then they were reconstructed into a 3D-EBSD data in Dream3D [32–35]. The 3D reconstruction procedure includes sections alignment, noise reduction, grain and grain boundary identification and boundary smoothing. Subsequently, the 3D data was visualized in ParaView. The regions of interest can be visualized in 3D, such as individual grain or boundary, and assemblies of certain grains or boundaries. Thirdly, the 3D microstructure should be studied quantitatively. Dream3D was used to quantify the grains, such as grain size (volume and equivalent sphere diameter), grain shape parameters, orientations, number of nearest neighbors and so on. In-house developed Matlab programs were used to quantify the grain boundaries, including the boundary area, grain surface area, numbers of neighboring boundaries and twin boundaries per boundary, and twin boundary proportions. The 3D-EBSD data could provide the crystallographic information and morphologies based on orientations, but it cannot determine the SCC crack path. The pictures of crack path on the 101 sections were taken using OM. These 2D OM maps can be reconstructed into a 3D-OM
Fig. 3. Distributions of the average grain diameters (a) and the twin boundary fractions (b) of three 2D-EBSD maps (slice 20th, slice 50th and slice 80th) and the 3D-EBSD microstructure.
0.005 wt.% S, 0.044 wt.% P and Fe in balance. The as-received material was hot rolled by 50% reduction of thickness at a starting temperature of 1200 °C, and then it was solution annealed at 1050 °C for 2.5 h and water quenching to produce a fully recrystallized microstructure with an average grain diameter of 52 μm (in 2D). The final thickness of the produced material was 80 mm approximately.
2.2. Stress corrosion cracking A compact tension (CT) specimen was machined from the as-prepared 316L stainless steel. It was a standard 1T CT (thickness B = 25 mm) with T-L orientation according to ASTM E399, where ‘L’ means longitudinal direction (rolling direction), and ‘T’ means long transverse direction. The main crack should propagate along the rolling direction during SCC test. Firstly, the CT specimen was pre-cracked by fatigue under a sinewave loading in air. The extension of the fatigue crack was about 2 mm. Subsequently, the specimen was fixed in an autoclave. The testing environment was simulated BWR primary water: deionized water, temperature at 288 °C, dissolved oxygen concentration (DO) at 20 ppm, pressure at 8 MPa. Before the SCC test started, a transition procedure, using a fatigue loading of a triangular-wave form at frequency of 0.01 Hz for 100 h and 0.001 Hz for 100 h, respectively, was performed in the testing environment in autoclave. The maximum stress intensity factor Kmax was about 30 MPa·m0.5. The load ratio (Kmin/Kmax) was 0.5. The transition procedure was employed to propagate the crack through 3
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
Fig. 4. Examples of cracked triple junctions. (a) The OM map and EBSD image of the 20th slice. The two highlighted regions ‘b’ and ‘c’ are accordance with figure b and c, respectively. (b) A cracked triple junction b262-b891-b307 and its topology structure. In the figure ‘b262′ is a grain boundary ID; ‘g93′ is a grain ID. In the topology structure, points represent grains; lines represent grain boundary, and cracked boundaries are represented by broken lines. (c) A cracked triple junction b117-b950-b1172 and its topology structure.
Fig. 5. (a) Sketch and topology structure of quadruple junction (QJ). (b) Topologically showing the modes of IGSCC propagating through quadruple junctions. As an example, QJC21 means the quadruple junction with two cracked boundaries, and the subscript number means isomerism. (c, d) The 3D schematic maps of cracking modes QJC33 and QJC41.
circle diameter) is 52 μm for the 20th, 50th and 80th 2D EBSD maps. The 3D-EBSD microstructure contains 4672 grain boundaries, including the boundaries that intersect the surfaces of the observed block. The average boundary size (equivalent circle diameter) is 33.5 μm. There are 3737 random boundaries and 935 twin boundaries. Fig. 3(b) shows the fractions of twin boundaries in the 20th, 50th and 80th 2D EBSD maps, and the 3D-EBSD volume, respectively. The measured area fraction of twin boundaries in the 3D microstructure is 41.6%, which is smaller than the length fractions in the 2D EBSD maps. In addition, the quantity fraction of twin boundaries is only 20.0%, which is much smaller than the area fraction, because the twin boundaries are larger than the random boundaries on average. The average random boundary size is 29.6 μm, and it is 49.0 μm for twin boundaries.
data by using a “3D Viewer” plugin [37] of ImageJ software [38–40]. 3. Results The main crack length after SCC testing in simulated BWR water for 2500 h is about 0.6 mm. The microstructure beside the crack was characterized in 3D by using 3D-EBSD coupled with 3D-OM. The reconstructed 3D microstructures are shown in Fig. 2. The 3D-EBSD microstructure contains 1105 grains, which includes 574 completely internal grains and 531 grains that intersect the surfaces of observed block. The average 3D grain size (equivalent sphere diameter) is 40.7 μm (including border grains), which is smaller than the measured 2D grain size, as shown in Fig. 3(a). The average grain size (equivalent 4
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
Fig. 6. An example of quadruple junction with three twin boundaries (3T-QJ) from the 3D-EBSD microstructure: (a) the four grains of the quadruple junction, having diameters of 159.1 μm (A), 78.8 μm (B), 135.1 μm (C) and 121.5 μm (D), respectively; (b) the six boundaries of the quadruple junction, in which the boundaries colored by dark red, green and pink are twin boundaries; (c) the topology structure of the quadruple junction, where points and lines were colored according to the related grains and grain boundaries, respectively. (See supplementary file ‘3D-QJ’). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
4.1. Effects of topological characteristics of triple junctions on IGSCC
The 3D-OM technology was used to map the IGSCC propagation path in 3D. Fig. 2(b) shows the 3D morphology, reconstructed using ImageJ 3D Viewer [37], from the 101 slices of OM observation. Furthermore, the 3D path of the branched crack through the depth of the observed block is shown in Fig. 2(c & d). The intergranular morphology of the 3D crack is evident. The grain boundary characters along the 3D crack path was investigated in 2D cross sectional planes at different depth. For example, Fig. 4(a) shows the OM and EBSD map of the 20th slice. It can be seen that most cracked boundaries are random boundaries. The twin boundaries scarcely cracked due to its strong resistance to IGSCC, as expected [3–5]. An example marked by a blue circle is shown in Fig. 4(a). The two random boundaries are cracked, and a short ∑9 boundary is partially cracked. The crack should propagate forward along the two twin boundaries because they are mechanically favorable oriented to the main crack propagation direction. However, this crack was stopped here. Except for the twin boundaries, other types of low-∑ CSL boundaries did not demonstrate special resistance to IGSCC, such as the arrows pointed boundaries in Fig. 4(a). The red arrow pointed boundaries is ∑23, and the black arrow pointed boundary is ∑13a. Both the two boundaries cracked during SCC.
Triple junction is an assembly of three boundaries between three mutually neighboring grains [14,23,41], for examples b262-b891-b307 and b950-b1172-b1170 in Fig. 4(b & c). The three boundaries has a common line. Triple junction can be topologically shown as a triangle, in which vertexes represent grains, and the lines between the vertexes represent grain boundaries. The broken lines in the topology structures in Fig. 4 mean cracked boundaries. A triple junction has two twin boundaries at most, and the third boundary is ∑9 in this case [24,26]. The ∑9 boundary in a triple junction with two twin boundaries should be hindered to crack because the three grains at the triple junction are bound by the two twin boundaries. From 3D results of this work, all observed ∑9 boundaries in ∑3-∑3-∑9 type of triple junctions are not cracked or just partially cracked from the side that opposites to the two twin boundaries. Therefore, the topological characteristics of triple junction have significant influence on the cracking resistance of boundaries of the triple junction. Fig. 4(c) shows a cracked triple junction, in which all the three boundaries are cracked. The 2D EBSD map shows the boundary b1172 is a low-∑ CSL boundary (∑13a), but it still cracked during SCC. Fig. 4(b) shows another cracked triple junction, in which the two boundaries b262 and b891 are cracked though b262 is a ∑17b boundary, but the random boundary b307 is uncracked though it seems to be more mechanically favorable oriented to IGSCC propagation compared with the boundary b891. Diameter of equivalent area of the boundary b307 is 98.2 μm. It has 20 neighboring boundaries which mean linearly connected boundaries with b307, and there are 2 twin boundaries. The 3D diameter of boundary b891 is 60.0 μm. It has 8 neighboring boundaries, including 2 twin boundaries. Neither of the two boundaries, b307 and b891, construct triple junctions with two twin boundaries. However, the boundary b307 is incorporated into two 2T-QJs (quadruple junction with two twin boundaries) but the boundary b891 is incorporated into only one 2T-QJ. The 2T-QJ could hinder the IGSCC propagation to some extent as discussed below. This may be a possible reason why the boundary b307 did not crack during SCC.
4. Discussion 3D characterization of the SCC tested 316L stainless steel shows that the crack propagated along GB network. Twin boundaries demonstrate a strong resistance to cracking, as reported in 2D studies [3–5]. However, a shift in focus from single boundary statistics to the topological characteristics of the entire GB network is need to reveal the mechanism of improving IGSCC resistance of materials by GB engineering process. Triple junction [14,23,41] and quadruple junction [24,25] are two typical topological structures of GB network. Effects of the topological characteristics of triple junctions and quadruple junctions in the presence of twin boundaries on the IGSCC propagation need investigation. It is well know the IGSCC would be stopped at triple junctions with two twin boundaries, but it is unclear how the IGSCC propagates through quadruple junctions. Triple junctions can be observed in 2D maps of cross-sections, but the quadruple junctions have to be studied in 3D. 5
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
Fig. 7. Topological modes of IGSCC propagating through the five types of quadruple junctions based on the quantity and arrangement of twin boundaries. The abbreviations of cracking modes at the left column are in accordance with Fig. 5(b). The top row shows the abbreviations and topology structures of the five types of quadruple junctions: 0T-QJ, 1T-QJ, 2T-QJ1, 2T-QJ2 and 3T-QJ.
account. For example, all the three modes, QJC31, QJC32 and QJC33, have three cracked boundaries, but they have different topology structure as well as different effectiveness to destroy the quadruple junction. A grain would be separated from the other grains when the case of QJC33 occurred, but the four grains would be still assembled when QJC31 or QJC32 occurred. Fig. 5(c) and (d) shows the 3D sketch maps of cracking modes QJC33 and QJC41. In Fig. 5, however, the twin boundaries were not taken into account. The twin boundaries have strong resistance to cracking, so some cracking modes of Fig. 5 should be excluded when the quadruple junction has twin boundaries. Gertsman [26] had studied the CSL theory of quadruple junctions, obtaining that the maximum number of twin boundaries in a quadruple junction is three, with two ∑9 boundaries and one ∑27 boundary in this case. For example, the quadruple junction in Fig. 6, boundaries AB, BC and CD are twin boundaries, and boundaries AC and AB are ∑9, and AD is a ∑27 boundary. The twin
4.2. Effects of topological characteristics of quadruple junctions on IGSCC Quadruple junction [24,25] is an assembly of six boundaries between four mutually neighboring grains, as schematically shown in Fig. 5(a). The four grains or the six boundaries have a common point which is generally named quadruple node. Quadruple junction is a spatial structure so that it cannot be seen in 2D maps of cross-sections. A topology method, mutually connected four points, was used to illustrate the quadruple junction, as shown in Fig. 5(a). In the topology structure, points represent grains, and lines represent boundaries. Fig. 6 shows an example of quadruple junction from the 3D-EBSD microstructure. It has three twin boundaries. The modes of IGSCC propagating through quadruple junctions were classified into ten types based on the cracked boundary number and their arrangement, as topologically shown in Fig. 5(b). The broken lines represent cracked boundaries, and isomerism structure was taken into 6
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
Fig. 8. Two examples of cracked quadruple junctions and their topology structures. The boundaries were differentiated by colors. Red lines in the topology structures represent twin boundaries, i.e. t1972, t824 and t1169. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
boundaries are cracked but the two twin boundaries. In addition, if the former five types of cracking modes occur, i.e. QJC1, QJC21, QJC22, QJC31 and QJC32, as shown in Fig. 7, the four grains of the quadruple junction will still be bound by the non-cracked boundaries, so the five cracking modes cannot destroy quadruple junctions thoroughly. However, if the latter five type of cracking modes occur, i.e. QJC33, QJC41, QJC42, QJC5 and QJC6, the four grains will be separated, and then quadruple junctions are broken thoroughly. Furthermore, if a quadruple junction has three twin boundaries (3TQJ), IGSCC would never propagate through it following the latter five types of cracking modes. The four grains of the 3T-QJ are bound by the 3 twin boundaries, so that 3T-QJ can hinder the IGSCC propagation. Even a quadruple junction has two twin boundaries (2T-QJ), only one or two of the later five cracking modes could occur. Therefore, quadruple junctions with three twin boundaries could stop the IGSCC propagation, and even quadruple junctions with two twin boundaries have considerable resistance to IGSCC propagation. Fig. 9. Statistics of the quantity of neighboring twin boundaries (TBs) and the quantity fraction of twin boundaries in total of its neighboring boundaries for randomly selected 61 grain boundaries. 44 of them are cracked boundaries, and the other 17 boundaries are uncracked but their positions are beside the IGSCC propagation path.
4.3. Neighboring twin boundaries effect on cracking of boundary The cracking susceptibility of a boundary depends on not only the character (∑-value of CSL model) of this boundary but also the characters of its neighboring boundaries. For example, a random boundary that surrounded by twin boundaries would never crack during IGSCC, because the crack-resistant twin boundaries would prevent the IGSCC from propagating to the random boundary. The boundary that has more neighboring twin boundaries should have higher resistance to IGSCC. Fig. 9 shows statistics of the quantity of neighboring twin boundaries and the ratio of neighboring twin boundary quantity to the total number of neighboring boundaries. In total, 61 boundaries that randomly selected were measured, including 44 cracked boundaries and 17 uncracked boundaries that beside the crack path. The points in Fig. 9 seem to be scattered. It seems that the cracking susceptibility of a boundary has no obvious correlation with the quantity of neighboring twin boundaries. However, the boundary with the highest quantity of neighboring twin boundaries, meanwhile, having the highest quantity fraction of neighboring twin boundaries, is uncracked. In comparison, the boundaries having the lowest fractions of neighboring twin boundaries are cracked. In addition, the average number of neighboring twin boundaries for the 44 cracked boundaries is 2.34, and it is 3.00 for the 17 uncracked boundaries. The uncracked boundaries have more neighboring twin boundaries on average. The average ratio of neighboring twin boundary quantity to the total number of neighboring boundaries for the 44 cracked GBs is 0.17, and it is 0.23 for the 17 uncracked boundaries. The uncracked boundaries have higher fraction of neighboring twin boundaries as well than the cracked boundaries.
boundaries were identified automatically by Dream3D, and the allowed deviation was defined according to Brandon Criterion [42]. Considering quantity and arrangement of twin boundaries, quadruple junctions can be classified into five types: 0T-QJ, 1T-QJ, 2T-QJ1, 2T-QJ2 and 3T-QJ, as shown in Fig. 7. Isomerism was considered here, because both 2TQJ1 and 2T-QJ2 have two twin boundaries but different topology structures. The modes of IGSCC propagating through the five types of quadruple junctions were laid out in Fig. 7. For 0T-QJ (quadruple junction without twin boundary), all the ten cracking modes can occur, but some cracking modes will not occur for other types of quadruple junctions if twin boundaries are supposed non-crackable. For example, the 3T-QJ has only four potential cracking modes. The 2T-QJ1 and 2T-QJ2 have six and eight cracking modes, respectively, though both them have two twin boundaries. Therefore, the topological characteristics of quadruple junctions have significant influence on the cracking resistance of IGSCC propagation, and not only the quantity but also the arrangement of twin boundaries in GB network should be considered. Fig. 8 shows two examples of cracked quadruple junctions along the IGSCC propagation path in the 316L sample. All the five non-twin boundaries of the quadruple junction of Fig. 8(a) are cracked except a twin boundary, so it follows QJC5. The quadruple junction of Fig. 8(b) follows the cracking mode QJC41, in which all the four non-twin 7
Corrosion Science xxx (xxxx) xxx–xxx
T. Liu et al.
[9] S. Kobayashi, R. Kobayashi, T. Watanabe, Control of grain boundary connectivity based on fractal analysis for improvement of intergranular corrosion resistance in SUS316L austenitic stainless steel, Acta Mater. 102 (2016) 397–405. [10] A. Stratulat, J.A. Duff, T.J. Marrow, Grain boundary structure and intergranular stress corrosion crack initiation in high temperature water of a thermally sensitised austenitic stainless steel, observed in situ, Corros. Sci. 85 (2014) 428–435. [11] V. Randle, The Role of the Coincidence Site Lattice in Grain Boundary Engineering, Cambridge University Press, London, 1996. [12] H. Li, S. Xia, B. Zhou, W. Chen, C. Hu, The dependence of carbide morphology on grain boundary character in the highly twinned Alloy 690, J. Nucl. Mater. 399 (2010) 108–113. [13] T. Watanabe, Approch to grain boundary design for strong and ductile polycrystals, Res. Mech. 11 (1984) 47–84. [14] V. Randle, Twinning-related grain boundary engineering, Acta Mater. 52 (2004) 4067–4081. [15] S.A. Xia, B.X. Zhou, W.J. Chen, Effect of single-step strain and annealing on grain boundary character distribution and intergranular corrosion in Alloy 690, J. Mater. Sci. 43 (2008) 2990–3000. [16] J. Lind, S.F. Li, M. Kumar, Twin related domains in 3D microstructures of conventionally processed and grain boundary engineered materials, Acta Mater. 114 (2016) 43–53. [17] T. Liu, S. Xia, B. Wang, Q. Bai, B. Zhou, C. Su, Grain orientation statistics of grainclusters and the propensity of multiple-twinning during grain boundary engineering, Mater. Des. 112 (2016) 442–448. [18] A. Das, Grain boundary engineering: fatigue fracture, Philos. Mag. 97 (2017) 867–916. [19] S. Xia, H. Li, T.G. Liu, B.X. Zhou, Appling grain boundary engineering to Alloy 690 tube for enhancing intergranular corrosion resistance, J. Nucl. Mater. 416 (2011) 303–310. [20] C.L. Hu, S. Xi, H. Li, T.G. Liu, B.X. Zhou, W.J. Chen, N. Wang, Improving the intergranular corrosion resistance of 304 stainless steel by grain boundary network control, Corros. Sci. 53 (2011) 1880–1886. [21] F. Shi, P.C. Tian, N. Jia, Z.H. Ye, Y. Qi, C.M. Liu, X.W. Li, Improving intergranular corrosion resistance in a nickel-free and manganese-bearing high-nitrogen austenitic stainless steel through grain boundary character distribution optimization, Corros. Sci. 107 (2016) 49–59. [22] M. Michiuchi, H. Kokawa, Z.J. Wang, Y.S. Sato, K. Sakai, Twin-induced grain boundary engineering for 316 austenitic stainless steel, Acta Mater. 54 (2006) 5179–5184. [23] M. Frary, C.A. Schuh, Connectivity and percolation behaviour of grain boundary networks in three dimensions, Philos. Mag. 85 (2005) 1123–1143. [24] V. Gertsman, Coincidence site lattice theory of multicrystalline ensembles, Acta Crystallogr. A 57 (2001) 649–655. [25] S.F. Li, J.K. Mason, J. Lind, M. Kumar, Quadruple nodes and grain boundary connectivity in three dimensions, Acta Mater. 64 (2014) 220–230. [26] V.Y. Gertsman, Coincidence site lattice theory of triple junctions and quadruple points, Science and Technology of Interfaces, John Wiley & Sons Inc., 2002, pp. 387–398. [27] E.A. Holm, P.M. Duxbury, Three-dimensional materials science, Scripta Mater. 54 (2006) 1035–1040. [28] C.A. Schuh, R.W. Minich, M. Kumar, Connectivity and percolation in simulated grain-boundary networks, Philos. Mag. 83 (2003) 711–726. [29] M.A. Groeber, B.K. Haley, M.D. Uchic, D.M. Dimiduk, S. Ghosh, 3D reconstruction and characterization of polycrystalline microstructures using a FIB-SEM system, Mater. Charact. 57 (2006) 259–273. [30] A.D. Rollett, S.B. Lee, R. Campman, G.S. Rohrer, Three-dimensional characterization of microstructure by electron back-scatter diffraction, Annu. Rev. Mater. Res. 37 (2007) 627–658. [31] S. Zaefferer, S.I. Wright, D. Raabe, Three-dimensional orientation microscopy in a focused ion beam-scanning electron microscope: a new dimension of microstructure characterization, Metall. Mater. Trans. A 39a (2008) 374–389. [32] M. Groeber, M. Jackson, DREAM.3D: a digital representation environment for the analysis of microstructure in 3D, Integr. Mater. Manuf. Innov. 3 (2014) 1–17. [33] M. Groeber, S. Ghosh, M.D. Uchic, D.M. Dimiduk, A framework for automated analysis and simulation of 3D polycrystalline microstructures.: Part 1: statistical characterization, Acta Mater. 56 (2008) 1257–1273. [34] M. Groeber, S. Ghosh, M. Uchic, D. Dimiduk, A framework for automated analysis and simulation of 3d polycrystalline microstructures. Part 2: synthetic structure generation, Acta Mater. 56 (2008) 1257–1273. [35] Y. Bhandari, S. Sarkar, M. Groeber, M.D. Uchic, D.M. Dimiduk, S. Ghosh, 3D polycrystalline microstructure reconstruction from FIB generated serial sections for FE analysis, Comp. Mater. Sci. 41 (2007) 222–235. [36] U. Ayachit, The ParaView Guide: A Parallel Visualization Application, Kitware, 2015. [37] B. Schmid, J. Schindelin, A. Cardona, M. Longair, M. Heisenberg, A high-level 3D visualization API for java and ImageJ, BMC Bioinf. 11 (2010). [38] T.J. Collins, ImageJ for microscopy, Biotechniques 43 (2007) 25–30. [39] M. Abramoff, P. Magalhaes, Image processing with ImageJ, Biophoton. Int. (2004). [40] A. Ullah, G.Q. Liu, H. Wang, M. Khan, D.F. Khan, J.H. Luan, Optimal approach of three-dimensional microstructure reconstructions and visualizations, Mater. Exp. 3 (2013) 109–118. [41] G. Gottstein, L.S. Shvindlerman, Grain boundary junction engineering, Scripta Mater. 54 (2006) 1065–1070. [42] D.G. Brandon, The structure of high-angle grain boundaries, Acta Metall. 14 (1966) 1479–1484.
Furthermore, compared with the quantity of neighboring twin boundaries, the cracking susceptibility of boundary is more strongly affected by the quantity fraction. The cracking susceptibility is approximately lower for boundaries that have higher fraction of neighboring twin boundaries. 5. Conclusions A crack in a solution annealed 316L stainless steel after SCC test in simulated BWR water was characterized in three-dimensional by using 3D-EBSD coupled with 3D-OM. The effects of the topological characteristics of grain boundary network on the IGSCC propagation were investigated in 3D. It was found that the twin boundaries not only have a strong resistance to IGSCC themselves, but they could also prevent the neighboring boundaries from cracking, as the cracking probability is lower for boundaries that have higher fraction of neighboring twin boundaries. The IGSCC propagation can be hindered by triple junctions or quadruple junctions in the presence of twin boundaries. Triple junction or quadruple junction with more twin boundaries plays an important role to improve the IGSCC resistance of materials. Triple junctions with two twin boundaries can stop the IGSCC propagation. Quadruple junctions with three twin boundaries can stop the IGSCC propagation, and even quadruple junction with two twin boundaries has a considerable resistance to IGSCC propagation. Acknowledgments This work was supported by the National Natural Science Foundation of China (NSFC) (grant number 51671122), the Shanghai Science and Technology Commission (grant number 13520500500) and the fund from Jiuli Hi-Tech Metals Co., Ltd. The authors gratefully thank Gregory S. Rohrer and Xiaoting Zhong (Carnegie Mellon University, USA) for their helps on Dream3D and ParaView, and thank Asad Ullah (Karakoram International University, Pakistan) for his help on ImageJ 3D Viewer, and thank Wenyue Zheng (Natural Resources Canada. University of Science and Technology Beijing) for his help to improve the English. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.corsci.2017.10.003. References [1] S.M. Bruemmer, G.S. Was, Microstructural and microchemical mechanisms controlling intergranular stress corrosion cracking in light-water-reactor systems, J. Nucl. Mater. 216 (1994) 348–363. [2] X. Zhong, S.C. Bali, T. Shoji, Accelerated test for evaluation of intergranular stress corrosion cracking initiation characteristics of non-sensitized 316 austenitic stainless steel in simulated pressure water reactor environment, Corros. Sci. 115 (2017) 106–117. [3] B. Alexandreanu, B. Capell, G.S. Was, Combined effect of special grain boundaries and grain boundary carbides on IGSCC of Ni-16Cr-9Fe-xC alloys, Mat. Sci. Eng. A Struct. 300 (2001) 94–104. [4] V.Y. Gertsman, S.M. Bruemmer, Study of grain boundary character along intergranular stress corrosion crack paths in austenitic alloys, Acta Mater. 49 (2001) 1589–1598. [5] E.A. West, G.S. Was, IGSCC of grain boundary engineered 316L and 690 in supercritical water, J. Nucl. Mater. 392 (2009) 264–271. [6] A. Telang, A.S. Gill, D. Tammana, X. Wen, M. Kumar, S. Teysseyre, S.R. Mannava, D. Qian, V.K. Vasudevan, Surface grain boundary engineering of Alloy 600 for improved resistance to stress corrosion cracking, Mater. Sci. Eng. A 648 (2015) 280–288. [7] Z. Zhang, S. Xia, W. Cao, H. Li, B. Zhou, Q. Bai, Effects of grain boundary character on intergranular stress corrosion cracking initiation in 316 stainless steel, Acta Metall. Sin. 52 (2016) 313–319. [8] A. Telang, A.S. Gill, M. Kumar, S. Teysseyre, D. Qian, S.R. Mannava, V.K. Vasudevan, Iterative thermomechanical processing of alloy 600 for improved resistance to corrosion and stress corrosion cracking, Acta Mater. 113 (2016) 180–193.
8