Quantification of SCC mechanisms in austenitic alloys under PWR primary water conditions

1 Introduction

Stress corrosion cracking (SCC) is a progressive failure mode that requires a specific environment (cooling water), stress (applied or residual) and a susceptible material (e.g. stainless steels or nickel base alloys), affecting many components in light water nuclear reactors. As their operating life increases, SCC is more likely to occur, leading to unexpected economic losses from service outages and costs for replacement components. For this reason, a great international research effort is in place in order to understand the underlying mechanisms for stress corrosion cracking in steels and Ni-base alloys. To reliably predict the susceptibility, initiation times and stress corrosion crack growth rates for new systems, it is essential that the operating mechanisms are understood and quantified. From the different appearances of crack surfaces, the different time scales – from hours to decades – and different electrochemical conditions, it is clear that no one single mechanism can be responsible for SCC in every system. Researchers have tried to quantify and predict stress corrosion susceptibly and crack growth rates (CGRs) analytically and empirically (Magnin et al. 1996; Satoh et al. 1998; Scott 1999). Original equations have been altered and adapted for new materials and conditions (Hall 2008, 2009). Some of the suggested equations, mechanisms and models appear to describe the characteristics of SCC in certain systems accurately. However, all proposed mechanisms/models cannot explain on their own all the observed phenomena. This should not be a surprise, since the majority was proposed using data from indirect measurements, which did not take into consideration the true scale of the microstructural and chemical processes involved. Only in the last two decades, advances in characterization techniques and, more importantly, a change in the approach by the corrosion and nuclear communities have enabled a critical review of the current situation.

A large number of advanced experiments covering an extensive and impressive parameter space have been published in the literature in the last decades. Some of these data have been reviewed and summarized by, e.g. Scott (2004), Arioka et al. (2008), Staehle (2016) and Shoji et al. (2011). A review of high-resolution characterization of SCC (Lozano-Perez et al. 2014) was presented in 2014 by the author. As the reader will notice, characterization and testing techniques have continuously evolved, contributing to more reliable and accurate data over the years.

Traditionally, autoclave testing over long periods of time was used to measure initiation times and crack growth rates in high-temperature aqueous corrosion. Fracture surfaces were analysed in the scanning electron microscope (SEM), but transmission electron microscopy (TEM) was rarely used to understand the alloy microstructure, due to the complexity of the crack tip sample preparation. In more recent years, with the availability of modern focused ion beam (FIB), techniques for TEM (Huang et al. 2002; Langford et al. 2001) and atom-probe site-specific sample preparation (Meisnar et al. 2015a), high-resolution techniques with high chemical sensitivity such as analytical transmission electron microscopy (ATEM) and atom-probe tomography (APT) have started providing unique datasets that are helping understand key degradation processes (Kruska et al. 2019, 2012; Langelier et al. 2016). These techniques can provide information on the crucial chemical and mechanical processes occurring at the crack tip. FIB 3D (imaging/EBSD/EDX) and electron or X-ray tomography are now finally producing high-resolution 3D data that help visualizing the real nature of defects and cracks (Lozano-Perez et al. 2009a; Persaud et al. 2018a). Electron back-scatter diffraction (EBSD) and digital image correlation (DIC) have also been employed to understand stress fields around cracks, map strain and understand its role in crack propagation (Di Gioacchino and Clegg 2014).

In this work, we will use new and published data by our group to rationalize our current understanding of the mechanisms controlling SCC in austenitic alloys under the conditions that are relevant to PWR primary water systems. For this purpose, we will base our approach on Coriou’s and Arioka’s work, studying a range of austenitic alloys that exhibit higher SCC resistance for a range of temperatures and compositions, aiming to provide a mechanistic explanation. Over 50 years ago, Coriou et al. (1966) identified a ‘safer’ range of compositions with Ni wt% between 20 and 60 where alloys would be less susceptible to SCC. Initially, this study did not receive much attention and was highly criticized, but it was later demonstrated to be pointing in the right direction. Recently, it has been the subject of a comprehensive review by Staehle (2016) and the validity of the original results extensively investigated by Arioka under much more controlled testing conditions. Arioka performed a series of constant load tests on 1/2CT specimens in simulated PWR primary water, generating a matrix of 44 samples: 11 alloys of different compositions were tested at 4 different temperatures (290, 320, 340 and 360 °C) and SCC CGRs measured from fracture surfaces. All results were reported in (Arioka et al. 2014), although they were not fully mechanistically explained. A crucial piece of information that transpired from these results is that the ‘safer’ range of alloy compositions at which SCC CGR decreases occurs at temperatures above 340 °C (Figure 1B) but it disappears below 320 °C (Figure 1A). Furthermore, the beneficial effect of Cr is obvious at temperatures between 320 °C and 340 °C but appears to be negligible for higher and lower temperature. His work led to the conclusion that crack growth rates (CGRs) are indeed strongly affected by these parameters (alloy composition and/or temperature), but a more mechanistic understanding was lacking.

Figure 1:

IGSCC growth versus Ni wt% at 320 °C (A) and 360 °C (B) in PWR water. Crosses refer to samples where the SCC crack was not complete across the specimen and indicate an upper bound for CGR. The beneficial effect of Cr addition (red points) is only obvious below 340 °C. Data from (Arioka et al. 2014).

Since 2012, we have systematically studies selected samples from the matrix of austenitic alloys tested by Arioka as described in (Arioka et al. 2014). If we think of this matrix as having columns for test temperature and rows for alloy composition, the row containing the 10 wt% Ni alloy (316SS) was the first to be analysed. A temperature dependence (including a peak which moved with cold-work level), already reported by Arioka et al. (2007, 2008), was confirmed. Careful microstructural and chemical analysis revealed that, at least, two mechanisms operated: a thermally activated diffusion-related and a localized deformation-related (Meisnar et al. 2016), but with their relative contributions affected by temperature. In order to further test these hypothesis, the approach was successfully used to explain the lack of peak temperature at the other end of the matrix, for A600 (Shen et al. 2019e), where no peak temperature was observed. The data collected from this sample was used to identify a series of precursors occurring before crack propagation (Shen et al. 2019b).

When these austenitic alloys are deformed, there is an excess of dislocations generated in order to accommodate plasticity; however, there are differences on how this happens. The dominant mode of deformation in the high-Ni alloys is dislocation glide, while the low-Ni ones (Fe-rich) undergo considerable deformation twinning. This result can be rationalized by conventional theories of stacking fault energies (SFEs). SS316 has the lowest SFE, which literature estimates to be around 15 mJm⁻² (Byun 2003; Lu et al. 2016). SFE theory predicts that martensitic transformations and deformation twinning are the dominant deformation mechanisms (Molnár et al. 2019), but only twinning is observed in this sample. This agrees with published data (Das 2016). The SFE of both A800 and A600 is more ambiguous and is not as well studied as that of SS316. SFE of A800 is reported to be between 29 and 150 mJm⁻², while the SFE of A600 is reported as high as 420 mJm⁻² (Totsuka et al. 2005). The SFE of A800 is an intermediate of SS316 and A600, but it deforms by dislocation glide instead of deformation twinning. As a result of no twins, there is more strain localization along {111} planes. As it will be shown in the next section, our GND observations support the trends in SFE of these materials but, more importantly, they allow for an accurate characterization and quantification of the deformation at the key regions. Increased dislocation densities do not only affect the mechanical properties, such as yield stress, but also changes the local diffusion kinetics. Therefore, it is important to characterize the dislocation distribution in the most relevant regions for SCC, such as the crack tips.

It has become apparent that more than one mechanism is needed to explain SCC in PWRs. Intergranular oxidation ahead of the crack tip for both Ni-base alloys and stainless steels was demonstrated (Lozano-Perez et al. 2009b; Meisnar et al. 2015a) and acknowledged as a required precursor for crack propagation. However, the morphology of the crack tip and crack flanks suggests that local plasticity has also occurred, in addition to other phenomena such as diffusion-induced grain boundary migration (DIGBM). Thus, a mechanistic explanation for all observed behaviours is the main topic of this manuscript. We also aim to provide a comprehensive explanation to the following questions raised from Figure 1:

1. Why do ‘medium-Ni’ alloys experience a reduced CGR only at ‘high’ T?

2. Why ‘low-Ni’ and ‘high-Ni’ alloys, being so different, exhibit an equally high CGR?

3. Why an increased Cr content is particularly beneficial at ‘low’ T?

2 Materials and methods

Selected samples from the matrix of 44 specimens (11 compositions and 4 test temperatures) tested as 1/2CT specimens under constant load by INSS as described by Arioka in (2014) and additional alloys provided by EDF have been selected for a comprehensive high-resolution characterization. The K values of testing were approximately 30 MPa m^1/2. PWR primary water, which contained boric acid (H₃BO₃, 500 ppm as B), lithium hydroxide (LiOH, 2 ppm as Li), and dissolved hydrogen (30 cc STP H₂/kg H₂ O) in the range of temperature between 290 °C and 360 °C. In addition to commercial nuclear grade SS316, Alloy 800, Alloy 690 and Alloy 600, model alloys A16Ni and A60Ni were added to the study to better understand the effect of Ni and Fe (note that all alloys studied, except A800 and A690, have ∼16 wt% Cr). 60Ni Alloy was solution-treated at 1,030 °C for 1 h in air, while the other alloys had it at 1,075 °C. All alloys were subsequently water-cooled. Grain boundary carbide coverage was nearly 0 for all alloys except A690 (50 %), A800 (80 %) and A600 (20 %). Alloy compositions and average grain size are shown in Table 1. All alloys were cold rolled to a reduction in thickness of 20 % prior to autoclave testing under simulated PWR primary water (hydrogenated water: 500 ppm B + 2 ppm Li + 30 cc/kg dissolved H₂). The SCC tests were conducted in an autoclave at a constant load of 30 MPa m^1/2 by the institute of nuclear safety systems (INSS) using pre-cracked 1/2T compact tension (CT) specimens, which were extracted in the T–L orientation (crack growth direction parallel to the rolling direction and crack plane perpendicular to the plane of the plate).

A MTS XP nanoindenter with a spherical tip was used for the nanoindentation and hardness measurements at room temperature of SS316, A800 and A600, all polished to a mirror finish. The results presented are from grain with a <001> surface normal. The depth of each nanoindent was 300 nm, needed to contain all generated deformation (dislocations) inside the TEM lift outs and, thus, enable high-resolution characterization. In addition, these deformation dimensions are comparable with those observed at crack tips. Target grain orientations were determined by EBSD, and coordinates for indentation were generated prior to testing. Indents are discarded if they are near microstructural features such as grain boundaries or carbides.

A X-treme high temperature nanoindenter was used for the tests at 320 and 360 °C on mirror-polished samples of SS316, A800, A690 and A600. They were mounted on the same heating stage using cement and in vacuum (<0.05 atm). Both load and displacement-controlled tests were carried out. The target load and displacement was 50 mN and 1 μm, respectively. Indents were completed in 2 × 8 arrays, with a spacing of 100 μm. This large spacing was used to mitigate anisotropic effects by sampling multiple crystal orientations. The samples were tested at room temperature before being heated to 280 °C and 360 °C. No thermal cycling occurred. The samples were mounted using omega Bond 600 cement and heated at a rate of 1.6 °C m⁻¹ until the target temperature was achieved. A carbon–boron nitride (cBN) Berkovich tip was used as diamond has a tendency to react with ferrous and carbide forming materials. Once at temperature, the indentation started when the tip and sample were at thermal equilibrium ±1 °C, as measured by thermo-couples in the tip and stage. The data collection method used is described by Oliver and Pharr (Oliver and Pharr 1992) with no continuous stiffness measurement (CSM). This means only a single value of hardness and modulus is achieved for each indent.

High-resolution electron back-scattered diffraction (HR-EBSD) was completed on a Zeiss Merlin and recorded on a CCD camera (Electron Back-scattered Patterns (EBSD): 800 × 600 pixels). The SEM conditions were typically 20 keV accelerating voltage and a 5 nA beam current. All EBSD maps were collected using a 125 nm step size.

Transmission Kikuchi diffraction (TKD) was performed using a Zeiss Crossbeam 540 FEG-SEM equipped with an Oxford Instrument (OI) EBSD Nordlys Max 3 detector and a Zeiss Merlin FEG-SEM equipped with a Bruker Optimus TKD head, always on TEM samples. The step size used was the same for all results in the same plots, typically 10 nm. After the TKD testing, the data were subsequently post-processed. Image quality, inverse pole figure (z-axis) (IPFZ), Kernel average misorientation (KAM) maps and geometrically necessary dislocation (GND) maps were calculated automatically.

The KAM maps, where each pixel represented the misorientation (MO) value in this pixel with regard to the average MO value of all pixels in the same grain, can be used to visually evaluate the local dislocation density in the samples based on a colour temperature spectrum (warmer areas such as red, orange and yellow indicated higher dislocation density than colder areas such as blue and green). It should be reminded that MO values shown in the MO line-profiles are relative (to the first pixel) and only the local changes (e.g., MO gradients) are important to understand local dislocation densities.

GNDs calculations were performed using a correlative technique (Hielscher et al. 2019; Meisnar et al. 2015b) when adequate data were available. In this study, the relatively high-level of cold-work (20 %) in the samples and the fact that some were obtained from regions additionally deformed, such as below indents or around crack tips, meant that the individual TKD patterns were not clear enough in some regions for an efficient cross-correlation. For this reason, GND densities in the most heavily deformed regions were calculated from the local misorientation and curvature data, instead of cross-correlation of diffraction patterns, which is also a less computationally demanding method. As a consequence, the misorientation-based GND densities have an absolute resolution of 1,012 dislocations/m². While this is approximately ten-times greater than cross-correlation (Dunne et al. 2012), it is considered adequate for this application.

Some of the data presented in this study, relies GND data obtained from local KAM gradients, due to limitations in computational power at the time the data were acquired. If TKD KAM maps alone are used, one can measure the extent of the localized deformation zone (LDZ) around the indent, local lattice rotation and local deformation intensity, both within the LDZ. Nye (1953) explains how deformation intensity can be associated with dislocation density. To demonstrate that KAM deformation intensity measurements can be efficiently related to GND densities, TKD maps from several austenitic samples used in this study were used to calculate GND density and KAM maps. KAM line profiles originating at crack tips and highly deformed areas were extracted using the method presented in (Shen et al. 2019e), in order to measure the local deformation intensities. That way, local GND density and deformation intensity (DI) from the same region can be compared. The results show acceptable linearity (R = 0.98) and the following relationship: GND density (m⁻²) = 9 × 10¹⁴ DI (°/μm) ± 4 × 10¹⁴, which is also correlated with the measured hardness (Figure 2).

Figure 2:

GND density and hardness versus Ni content (wt%).

In general, TKD is a powerful tool to analyse dislocations (Yu et al. 2019); however, there are limitations to this methodology. They will be briefly reviewed next, together with the approach used to minimize them. As noted by Liu et al. (2019), the rotations and shift in the Kikuchi pattern is a function of both sample thickness and step size. This error was minimized by using identical TKD conditions, 10 nm step size, in the data collection process. The process of creating a sample can result in data loss, due to relaxation of the elastic strain, while this is non-negligible, it is measured by Karamched et al. to be approximately 20 % (Karamched et al. 2021). Due to consistent sample preparation, this is assumed to be constant across the samples. The correlative techniques only capture GNDs, and statistically stored dislocations (SSDs) should not be ignored, as ρ_total = ρ_GND + ρ_SSD. Implications of this potential issue can be seen in the comparability of the pseudo bright-field images and the GND maps. Unlike TEM, the stage is less versatile and the sample cannot be tilted. This limitation affects comparative studies, as no corrections can be made in the chamber to ensure similar orientations. Finally, accurate GND calculations rely on a suitable ‘strain-free’ reference pattern. In the results reported, a strain-free region should be treated with caution, particularly if the materials have undergone prior cold-work and already contain significant deformation. The best compromise is to select a central point at the bottom of the lamella. The thinned sample’s edge is thought to provide the lowest possible levels of deformation, as the dislocations can escape the lamella and was chosen for this purpose. This point is marked by a cross on Figures 3–6. Fore-scattered diode images were collected using 20 kV & 1 nA.

Figure 3:

(A, top) Fore-scattered cross section images (both pseudo bright- and dark-field) of a spherical indent in SS316 along a near <101> zone axis. (A, bottom) Higher magnification images of regions of interest, selected from the fore-scattered images. (B) A calculated GND density map.

Figure 4:

(A, top) Fore-scattered cross section images (both pseudo bright- and dark-field) of a spherical indent in A800 along a near <101> zone axis. (A, bottom) Higher magnification images of regions of interest, selected from the fore-scattered images. (B) A calculated GND density map.

Figure 5:

(A, top) Fore-scattered cross section images (both pseudo bright- and dark-field) of a spherical indent in A600 along a near <101> zone axis. (A, bottom) Higher magnification images of regions of interest, selected from the fore-scattered images. (B) A calculated GND density map.

Figure 6:

Young’s modulus (A) and hardness (B) results at room temperature (RT), 280 and 360 °C. Relative error for all results is ∼5 %.

TEM/TKD samples were prepared as plan view FIB lift-outs containing crack tips. More details about the sample preparation with FIB can be found in (Lozano-Perez 2008). Final thickness was always below 50 nm. TEM analysis was performed with a JEOL 2100 (LaB₆, operating voltage 200 kV) for TEM imaging and selected area diffraction. High-resolution analysis, including scanning TEM (STEM) and electron energy loss spectroscopy (EELS), was conducted with a JEOL ARM200F (cold-field emission gun with a typical energy resolution of 0.6 eV) operating at 200 kV and equipped with a Quantum Gatan image filter (GIF) spectrometer. EELS spectra (low- and core-loss) were acquired correcting the spatial drift and recalibrated in energy using the zero-loss peak as a reference. Principal components analysis (for denoising) was performed using Hyperspy 1.3 (open source) software. The convergence and collection half-angles were 31 and 41 mrad, respectively, and a dispersion of 0.25 eV/channel was used. The relatively low thickness of all areas analysed (<50 nm) allowed for a reliable quantification without removing plural scattering. This was checked for regions of known composition (e.g. matrix). The relative errors in EELS quantification were dominated by uncertainties in theoretical cross sections and can be assumed to be ∼10 % relative for most measurements.

3 Characterization of SCC: results

3.1 Mechanical properties

Arioka et al. (2014) showed that yield strength of the alloys used in this study range between 500 and 650 MPa at room temperature and experience a decrease of around 10 % at 320 °C. Alloys with medium Ni content have the lowest yield strengths, around 20 % lower than those with high (Alloy 600) or low Ni (SS316). These measurements are macroscopic and, although they reflect an influence of temperature and composition, they do not capture how the deformation/strain is generated and where.

Figure 2 summarizes the mean calculated GND density in the matrix in the tested materials (in as-received condition) after HR-EBSD. The error bars represent the inter-quartile range of the measured GND information, this was chosen to show any skew in the recorded densities. As can be seen, these results correlate well with the hardness measurements from the nanoindentation tests and capture the trend revealed by the macroscopic yield strength measurements.

In a previous study, GND density measurements revealed how grain boundaries in A600 become more susceptible to SCC initiation as dislocations gradually pile-up after deformation (Saravanan et al. 2020). The alloys tested in this study (see Figure 1) exhibit a distinct change in CGR with Ni-content or temperature. To determine if dislocation behaviour under deformation plays an important role, selected samples were studied via nanoindentation at different temperatures, followed by FIB sample extraction under the indent for subsequent TEM and TKD characterization.

3.1.1 Room temperature deformation

Room temperature nanoindentation was performed on 3 alloys with different SFE and Ni/Fe ratios in an attempt to capture the effect on dislocation behaviour in the plastic zone below the indent. Cross-sectional focused ion beam (FIB) lift-outs containing the spherical indents on the three alloys were taken from similar crystallographic orientations with a <001> surface normal. The lift-outs were made close to a <110> plane normal to facilitate comparison. The high magnification results have been collected using transmission Kikuchi diffraction (TKD).

Deformation, in both the previous indentation experiments and SCC crack growth, requires the generation and movement of dislocations throughout the matrix. Figures 4–6 shows cross-sectional images of the region below the spherical indents for the three alloys, together with GND density maps that illustrate their differences. Figure 4 shows the plastic zone underneath a spherical indent (centre–top) in SS316. During indentation, dislocation generation has occurred. Dislocations are visible as black or white features in the pseudo bright- and dark-field images, respectively. The numerous dislocations have resulted in multiple visible slip traces. The normal of the slip traces have been determined to fall along the <111> directions from the pole figures. Some regions of interest (ROIs) have been highlighted in all figures to show dislocations with better clarity and illustrate how they can be counted ‘manually’ (as opposed to the GND calculation algorithm). Region of interest (ROI) 1 is chosen for the ρ_GND calculations, directly under the indent and representing the theoretical maximum measured dislocation density. ROI 3 is chosen in a region far away from the indent, as ‘undisturbed’ as possible, for the ρ_background measurements. Dislocations were also manually counted on the image in ROI 3. Both methods show a similar trend in dislocation densities and show reasonable correlation, as shown in Table 2.

Table 2:

Summary of the background, counted and under-the-indent GND dislocation densities from the three samples.

Material	ρ background (m−2)	ρ counted (m−2)	ρ GND (m−2)
SS316L	2.5 ± 1.1 × 1013	7.2 ± 0.4 × 1013	6 ± 4 × 1015
A800	1.3 ± 0.6 × 1013	2.1 ± 0.3 × 1014	5 ± 3 × 1015
A600	2.3 ± 0.8 × 1014	3.0 ± 0.3 × 1014	6 ± 3 × 1016

A800 shows the greatest normalized increase in dislocation density between the surrounding matrix and the region under the indent. Dislocation generation is the primary deformation mechanism seen in A800 and, as such, the number of dislocations to accommodate the deformation has increased the most.

The GND calculations and bright-field images show similarities with results from Cackett et al. (2019), who mapped GND densities in cross sections of spherical indents in copper. An observed trend across both GND and counting methods is that as the distance from the indent increases, the density of dislocations decreases.

SS316 and A800 exhibited clear deformation by slip after nanoindentation, in line with previous experiments reported in the literature (Das 2016; Lu et al. 2016; Murr 1969). Along the slips, dislocation densities of up to 4 × 10¹⁵ m⁻² were measured. In these alloys, there was a clear difference in dislocation density (both by GND calculations and directly counted on the image) between the slips and the region in between (∼3× more). In contrast, A600 exhibited a very similar dislocation density across the LDZ. In addition, as shown in Table 2, A800 appears to be the alloy that accommodated more dislocations after deformation (16× more), followed by SS316 (3× more) and A600 (1.3×). The results on A600 are interesting, since it was the alloy with higher dislocation density to start with, but with fewer generated after deformation. The novel approach presented has allowed the direct observation of dislocation behaviour after deformation, together with an accurate quantification via GND calculations.

3.1.2 Operating temperature deformation in vacuum

Mechanical properties tests are typically performed at room temperature. However, it is not clear whether this is representative of PWR operating T. Arioka’s results revealed a substantial difference in CGR between temperatures above and below 340 °C for alloys with a mid-Ni content. As an example, a 20 °C increase from 320 to 340 °C caused the CGR of SS316, A600 and A690 to increase by approximately one order of magnitude, while it dropped by two orders of magnitude for A800. In order to better understand the T dependence of mechanical properties and any related mechanism on SCC CGR, selected alloys were tested at 280 and 320 °C, representing the ‘low-T’ behaviour and at 360 °C, representing the ‘high-T’ behaviour. Tests consisted of nanoindendation at 280 and 360 °C and TKD characterization of SCC crack tips from samples autoclaved at 320 and 360 °C. Nanoindentation results at room temperature (RT) are provided for reference purposes. Both hardness and Young’s modulus results are provided as an average of both depth- and load-controlled tests. An error analysis of all tests on the same alloy and T reveals a representative relative error of ∼5 % for the hardness and Young’s modulus measurements. All results can be seen in Figure 6.

The calculated hardness and modulus values at RT are within the reported values found in the literature (Nomura et al. 2002). A600 is the hardest alloy at all T, while SS316 exhibits the lowest values. At PWR temperatures, the difference in hardness is minimal for A800 and greatest for SS3160 followed by A690. The observed softening is caused by a reduction in the lattice resistance to dislocation motion, which is typical of FCC materials (Frost and Ashby 1982), where the SFE tends to increase with T and dislocation glide is favoured (Molnár et al. 2019). The pre-existing dislocation network can contribute to the increased hardness, requiring more force to move dislocations, as shown in Figure 2. The Young’s moduli values exhibit a similar trend with T, with almost no change in A800 and the greatest for SS3160 followed by A690 (Badrish et al. 2019; Torres et al. 2016). This is somewhat expected, as movement of dislocations is favoured as the temperature increases, decreasing the occurrence of dislocation pile-ups and reducing the Young’s modulus as explained in (Benito et al. 2005).

3.1.3 Operating temperature deformation in autoclave

The deformation behaviour around crack tips has been studied traditionally using microhardness tests, qualitative TEM studies or SEM EBSD (Chimi et al. 2019; Ehmstén et al. 2009; Gussev et al. 2018; Lin et al. 2022; Nagashima et al. 2007). Increased dislocation densities (not quantified) and the occurrence of slip deformation was reported. However, the data lacked the spatial resolution required to appreciate the deformation localization characteristic of SCC around crack tips. As part of this study, deformation localization was studied in all samples that exhibited SCC at 320 and 360 °C. Shen and Meisnar (Meisnar et al. 2015b; Shen 2018; Shen et al. 2019b, 2019e, 2019f, 2022) demonstrated that the local deformation around a SCC crack tip can be characterized and quantified for A600 and SS316 when using TKD. The method, as described in Section 2, involves the acquisition of high-resolution TKD data, where the local lattice rotations gradients (referred to as deformation intensities) can be measured from MO profiles. Results from selected alloys in Figure 1 were obtained and are presented in this paper. An example result is shown in Figure 7 for crack tips extracted from the model alloys A60Ni and A16Ni tested at 360 °C. Deformation intensities were measured from line profiles and a summary of all results is presented in Table 3. These results can be converted into GND densities as explained in Section 2.

Figure 7:

TKD maps and misorientation (MO) profiles extracted around the dominant crack tips of (A) A60Ni and (B) A16-Ni model alloys, tested at 360 °C, using a step size of 10 nm. (A) top: TKD indexation quality map (left), IPFZ map (middle), and average MO map (right); bottom: MO profiles; local deformation zone (LDZ) size ∼295 nm/lattice rotation 1.7 ± 1.0° (MO profile 1), LDZ size ∼270 nm/lattice rotation 1.3 ± 1.1° (MO profile 2); (B) top: TKD indexation quality map (left), IPFZ map (middle), and average MO map (right); bottom: MO profiles; local deformation zone (LDZ) size ∼165 nm/lattice rotation 1.3 ± 0.9° (MO profile 1), LDZ size ∼230 nm/lattice rotation 1.2 ± 0.8° (MO profile 2).

Table 3:

Summary of results for the IOZ length, DIGBM dimensions, deformation intensity around the crack tip and CGR at 320 and 360 °C.

	Alloy	IOZ (nm)	DIGBM length (nm)	DIGBM amplitude (nm)	Deformation intensity (°/μm)	CGR (mm/s)
320 °C	SS316	80 ± 20	75 ± 22	10 ± 12	2.1 ± 0.3	6.2 × 10−7
	A16Ni	140 ± 30	62 ± 28	10 ± 10	1.5 ± 0.4	1.1 × 10−7
	A60Ni	180 ± 40	95 ± 21	72 ± 17	0.7 ± 0.4	6.0 × 10−7
	A600	30 ± 20	121 ± 53	31 ± 11	1.2 ± 0.6	5.6 × 10−7
360 °C	SS316	70 ± 20	112 ± 18	27 ± 11	1.4 ± 0.5	1.6 × 10−7
	A16Ni	170 ± 30	311 ± 15	112 ± 22	0.6 ± 0.5	1.3 × 10−8
	A60Ni	480 ± 50	412 ± 41	249 ± 34	0.6 ± 0.4	8.4 × 10−8
	A600	440 ± 60	556 ± 82	43 ± 12	1.2 ± 0.5	4.8 × 10−6

3.2 Corrosion-related behaviour

The oxidation behaviour at the crack tip of the samples that exhibited SCC at 320 and 360 °C (only SS316, A16Ni, A60Ni and A600) was characterized in detail. Some of the results for A600 and SS316 can be found in (Meisnar et al. 2016; Shen et al. 2018, 2019a, 2019b, 2019e) and for A60Ni in (Shen et al. 2021). As a reminder, the alloys with Ni compositions between 16 and 60 wt% didn’t experience SCC at high T and were not included in this study (see Arioka et al. 2014). For all crack tips, the following phenomena were studied and characterized using STEM HAADF and EELS:

1. Phase and composition of the intergranular oxides: An intergranular oxidation zone (IOZ) was observed ahead of the crack tip. This observation has also been reported by several other groups on various austenitic steels (Bruemmer and Thomas 2010; Kuang et al. 2019; Persaud et al. 2013, 2018b).

2. Dimensions of the diffusion-induced grain boundary migration (DIGBM) region: A Cr–Fe depleted zone is always observed ahead of the crack tip (which can also be referred to as Ni-enriched) and corresponds to the region where DIGBM has occurred. DIGBM occurrence is now frequently confirmed by high-resolution studies, whenever high-Ni regions are observed ahead of the oxidation from (Langelier et al. 2017; Persaud et al. 2018a; Shen et al. 2019c). This region can be characterized by its length along the grain boundary and width or ‘amplitude’ perpendicular to it.

3. Porosity in the oxide: Oxide nanoporosity was found in all intergranular oxides ahead of the crack tip, as shown for A600 in (Shen et al. 2018, 2019e). Thus, the intergranular oxide ahead of the crack tip cannot be considered a fully protective oxide. This was observed by other groups in alloys exposed to similar environments (Kuang et al. 2022; Thomas and Bruemmer 2000).

A minimum of 2 crack tips that appeared to be active at the time of stopping the autoclave experiment were extracted as TEM samples and characterized by STEM and TKD. Figures 8 and 9 show representative ones. Table 3 summarizes the measurements from all samples with their respective errors. Details of the methodology used for the measurements can be found in (Shen et al. 2019e).

Figure 8:

HAADF image and EELS elemental maps from selected SCC crack tips in alloys tested at 320 °C. (A) SS316, (B) A16Ni, (C) A60Ni and (D) A600.

Figure 9:

HAADF image and EELS elemental maps from selected SCC crack tips in alloys tested at 360 °C. (A) SS316, (B) A16Ni, (C) A60Ni and (D) A600.

4 Discussion

One of the first challenges that any rationalization of SCC-related phenomena faces is identifying external and internal phenomena affecting SCC. We proposed to define ‘global factors’ as those pre-set parameters or conditions affecting the alloy during exposure to the environment. Normally, they will not change globally during exposure, although they could change locally. An example of these are alloy composition, initial alloy microstructure (e.g. grain size, grain boundary type, phases present, carbide coverage, cold work … ), temperature or water chemistry. Once the alloy is exposed to the environment, local changes begin to occur that will ultimately contribute to facilitate or prevent SCC. We will refer to those as ‘precursors’. Identifying the precursors, as can be seen in the results section, requires a high-resolution multi-technique approach. Changes often only involve regions of several nm in size and are, therefore, challenging to characterize. The key regions to focus are the surface for initiation and the crack tip for propagation. However, it is rare that the same high-resolution methodology is systematically applied to an extensive series of samples. In this report, this was only possible over the course of a decade, involving several researchers but in turn generating comparable data spanning over different variables.

In this work, we have focused on the crack tip region to better understand the mechanisms controlling SCC propagation mechanisms. The concept of precursors is promising to understand the local changes that lead to SCC susceptibility, as demonstrated by Shen et al. (2018), Persaud et al. (2018b) or Kruska et al. (2019). Our work has identified the following precursors for SCC in PWR primary water side:

4.1 Localized deformation

Localized deformation is observed around grain boundaries due to dislocation pile-up caused by the prior cold-working and enhanced by the constant-load test in the autoclave. It can also occur around deformation bands, if present, as shown in (Lozano-Perez et al. 2009c). If we plot the deformation intensity results and compare them to the observed CGR at each temperature (see Figure 10), some interesting trends are observed. For high-Ni alloys, temperature has little effect on the local dislocation density around the crack tip. As the Ni content decreases, the effect of T is more pronounced, with dislocation densities decreasing by a bigger factor (up to 30 %) from 320 to 360 °C in the case of SS316. This information can be presented as the crack tip region becoming softer, thus making crack propagation harder. The other important observation is that alloys with medium Ni-content have the lowest dislocation densities, particularly at higher T. These results correlate well with the CGR results presented in Figure 1 and suggest that localized deformation around the crack tip is one of the controlling crack growth mechanisms. They also explain the observed peak temperature behaviour by which some austenitic alloys exhibit lower SCC CGRs after a certain T (Shen et al. 2020).

Figure 10:

Effect of Ni content on deformation intensity versus CGR in tests at (A) 320 °C and (B) 360 °C.

4.2 Intergranular oxidation

An intergranular oxidation zone (IOZ) was always found ahead of crack tips in all samples and temperatures in this study. It is enhanced by local high dislocation densities, which facilitate diffusion, and solute depletion, which in turn restricts the formation of a protective oxide. The oxides in this region formed following the same sequence observed in surface oxidation (Kruska et al. 2013; Lozano-Perez et al. 2012; Meisnar et al. 2015a, 2016; Scenini et al. 2019; Shen et al. 2019b). The most favoured oxide thermodynamically is Cr₂O₃, but this phase is only observed briefly and in the form of small nanoparticles (see Meisnar et al. 2015a; Shen et al. 2019b), due to the lack of available Cr locally. A more permanent spinel phase similar to chromite (FeCr₂O₄) is observed instead. However, in all of the alloys studied with a Cr content of ∼16 wt%, there were not enough trivalent cations to form a stoichiometric chromite. This is proposed as an explanation for the high density of vacancies observed and the further formation of porosity (Terachi et al. 2011). The high local density of vacancies can also explain the amount of substitutional diffusion observed, responsible for the local composition observed ahead of the IOZ, referred to as diffusion induced grain boundary migration (DIGBM). It should be noted that IOZ measurements were only performed on samples where the grain boundary was close to perpendicular respect to the TEM lamella plane, to avoid projection artefacts.

As can be seen in Figure 11, for the alloys tested, the length of the IOZ generally increases with T, although marginally for low-Ni alloys and greatly for high-Ni ones. Interestingly, at 360 °C, there is a clear anti-correlation with CGR, suggesting that this is not a general SCC controlling mechanism.

Figure 11:

Effect of Ni content on length of intergranular oxidation zone (IOZ) IOZ versus CGR in tests at (A) 320 °C and (B) 360 °C.

4.3 Occurrence of diffusion-induced grain boundary migration (DIGBM)

This phenomenon typically occurs around grain boundaries but also in regions of high local deformation. In PWR primary water conditions, the formation of the IOZ ahead of the crack tip promotes the diffusion of Fe and Cr towards the free surface of the crack tip to react with O. This is only possible in the presence of vacancies, which ultimately promote the local rearrangement of Ni atoms, occupying the vacancies left by Fe and Cr and resulting in DIGBM. A high local vacancy density has been recently confirmed experimentally (Yang et al. 2023) in this region. Ni is the most noble of the 3 solute atoms and requires a higher electrochemical potential to be oxidized. No Ni-rich oxide was found intergranularly. As can be seen in Figure 12, there is no clear correlation between the length of the DIGBM along the grain boundary and SCC CGR. However, when we look at its width/amplitude, there is a clear trend in the samples tested at 360 °C (see Figure 13). The anti-correlation between the DIGBM width and the CGR can be explained when looking at the path the crack follows in the A16Ni and A60Ni (see Figure 9). The data presented in (Shen et al. 2019c, 2021) confirm that the presence of a DIGBM with a high amplitude (perpendicular to the grain boundary) forces the crack to deviate from its original path along the grain boundary into directions that are no longer as affected by the applied stress (typically, perpendicular to the grain boundaries the crack chooses to propagate). This observation should be taken into consideration when proposing SCC controlling mechanisms. Note that in our previous publications, the length of the DIGBM is referred to as ‘Fe–Cr depletion’ zone.

Figure 12:

Effect of Ni content on length of the DIGBM versus CGR in tests at (A) 320 °C and (B) 360 °C.

Figure 13:

Effect of Ni content on amplitude of the DIGBM versus CGR in tests at (A) 320 °C and (B) 360 °C.

5 Summary and conclusions

The data presented in the previous sections can be rationalized by explaining SCC in selected austenitic alloys under PWR primary water conditions as being mainly controlled by two types of mechanisms. One is diffusion-related and the other one mechanical properties-related. This idea was already explored by Meisnar and Shen (Meisnar et al. 2016; Shen et al. 2019e). In this study, we add our recent results and expand the number of precursors involved in SCC propagation in order to better explain the observed SCC CGR presented in Figure 1.

In the diffusion-related mechanism, grain boundaries (and their surrounding region, high in dislocations due to pile-ups) act as fast-diffusion paths. Their effectiveness is enhanced by applied stress, dislocation density and temperature. The precursors involved are intergranular oxidation and the formation of DIGBM. We have quantified this mechanism by analytical STEM and atom-probe tomography.

In the mechanical properties-related mechanism, dislocations around the crack tip control the local mechanical properties. Factors such as cold-work level, slip transfer between neighbouring grains or SFE will control their initial density and how they evolve under an applied stress. If dislocations can move ‘easily’ and leave the area around the crack tip, the region is locally soft and stresses don’t build up as quickly. We have quantified this mechanism by nanoindentation, micromechanical testing and TKD.

Intergranular oxidation is required for SCC to propagate. No crack tip was found without an IOZ in any of the samples tested in this research. In previous studies, applied stresses ranging between 300 and 800 MPa are sufficient to fracture oxidized grain boundaries in austenitic alloys (Couvant et al. 2019; Dohr et al. 2017; Dugdale et al. 2013; Fujii et al. 2011). Thus, the local mechanical properties around the crack tip are responsible for stress concentration. A ‘threshold’ stress needs to be reached for oxidized grain boundary to fracture and the crack to propagate.

Regarding the questions we posed in Section 1:

1. Why do ‘medium-Ni’ alloys experience a reduced CGR only at ‘high’ T?

Although there is no SCC data from the ‘medium-Ni’ samples in our study (they did not fracture at the ‘high’ T), one can speculate that this is not a direct consequence of their oxidation resistance, but more of their different mechanical properties. Extrapolating from the A16Ni and A60Ni results, they are likely to experience the highest local ‘softening’ around the crack tip when the temperature is increased to 360 °C, which mitigates stress concentration. In addition, they are also likely to exhibit the biggest DIGBM amplitudes, deviating the crack path into lower effective stress directions.

2. Why ‘low-Ni’ and ‘high-Ni’ alloys, being so different, exhibit an equally high CGR?

SS316 is harder around the crack tip that A600 at both T. In turn, A600 intergranular oxidation (IOZ length) extends much further than in SS316 and it is more affected by T. 316SS becomes gradually softer with T (peak T effect as reported by [Andresen 1993; Arioka et al. 2014; Shen et al. 2019e]), while its IOZ remains practically unaffected, reducing the CGR. The role of one mechanism compensates for the other, resulting in both alloys experiencing similar CGRs.

3. Why an increased Cr content is particularly beneficial at ‘low’ T?

The results from recent studies confirm that alloys with higher Cr exhibit better quality oxides (continuous and with less porosity), as shown in (Shen et al. 2019e) for surface oxides. In this study, two alloys with higher Cr were studied, A690 and A800. Since neither of them exhibited SCC at 360 °C (see Figure 1), the crack tip region could not be characterized. However, one would expect that the oxides at the IOZ would be more protective, as already shown in (Shen et al. 2019e). In addition, both A690 and A800 are the ‘softest’ in the crack tip region (see Table 3), which would help decreasing stress localization. Finally, these alloys have a higher intergranular carbide coverage than the rest which, as shown in (Langelier et al. 2017; Shen et al. 2019d), can deviate cracks from their ‘linear’ path (along the grain boundary), into others around the carbide requiring higher local stresses to fracture the intergranular oxide.

We can now propose that the ideal austenitic alloy for the PWR primary water circuit should be one that:

1. Has a low dislocation density around the crack tip when under stress.

2. Exhibit low intergranular oxidation under stress.

3. Exhibit DIGBM with a high ‘amplitude’.

4. Has a higher Cr content than 16 wt%, in order to experience lower SCC susceptibility in the full operating T range.

With these in mind, it is no surprise that A690 has had an excellent in-service record, since it fulfils all conditions (e.g. see the NUREG-1841 report). The other commercial alloy that stands out is A800, which is in the right Ni-content window and has a higher Cr content. Interestingly, some recent results have shown than A800NG, with extra Cr, exhibits even better SCC resistance (Arioka et al. 2018) in PWR primary conditions.

In this study, we have shown that a comprehensive high-resolution characterization of the SCC phenomena has allowed the quantification of some of the key mechanisms contributing to SCC. They are clearly affected by alloy composition, microstructure and operating temperature. All findings point towards a better mechanistic explanation of SCC, which could be potentially relevant to other environmentally assisted degradation phenomena, and will contribute to the development of better alloys and predictive models.