Effect of Sulfide and Chloride Ions on Pitting Corrosion of Type 316 Austenitic Stainless Steel in Groundwater Conditions Using Response Surface Methodology

Abstract

This study investigates the corrosion resistance of Type 316 stainless steel as a candidate material for radioactive waste disposal canisters. The viability of stainless steel is examined under groundwater conditions with variations in pH, bisulfide ions (HS⁻), and chloride ions (Cl⁻) concentrations. Utilizing response surface methodology, correlations between corrosion factors and two crucial response variables, passive film breakdown potential and protection potential, are established. Cyclic potentiodynamic polarization tests and advanced analytical techniques provide detailed insights into the material’s behavior. This research goes beyond, deriving an equation through response surface methodology that elucidates the relationship between the factors and breakdown potential. HS⁻ weakens the passive film and reduces the pitting corrosion resistance of the stainless steel. However, this study highlights the inhibitory effect of HS⁻ on pitting corrosion when Cl⁻ concentrations are below 0.001 M and at equivalent concentrations of HS⁻. Under these conditions, immediate re-passivation occurs from the destroyed passive film to metal sulfides such as FeS₂, MoS₂, and MoS₃. As a result, no hysteresis loop occurs in the cyclic polarization curve in these conditions. This research contributes to the understanding of Type 316 stainless-steel corrosion behavior, offering implications for the disposal of radioactive waste in geological repositories.

1. Introduction

Nuclear power stands out as a crucial energy resource in the battle against climate change due to its minimal carbon emissions. However, the disposal of radioactive spent nuclear energy, a byproduct of nuclear power generation, presents a formidable challenge. This waste, with half-lives extending from 10,000 to 1,000,000 years, demands secure storage deep underground with geologically stable conditions for prolonged periods [1,2]. Although disposal canisters for spent nuclear fuel are shielded by backfill materials and isolated from the soil environment, groundwater infiltration may occur after a long period of time [3]. Therefore, the canister materials need to be corrosion-resistant to withstand long-term exposure to the groundwater environment.

Numerous studies have identified that copper, Ni alloys, and Ti alloys are suitable materials for disposal canisters [4,5,6]. Notably, research in Sweden and Finland delves into copper canisters, while research in the United States has focused on Ni alloys and Ti alloys [7]. Although stainless steel is also a material with excellent corrosion resistance and high strength, it was evaluated as unsuitable for a spent nuclear fuel canister. This oversight is attributed to its vulnerability to localized corrosion in soil environments. Previous research has shown that sulfur, which originates from the oxidation of pyrite in the host rock and sulfate-reducing bacteria in soil environments, could enhance localized corrosion of austenitic stainless steels in the presence of chloride ions (Cl⁻) [7,8,9,10]. Furthermore, Lee et al. reported that hydrogen sulfide ions, so-called bisulfide ions (HS⁻) were found to generate defective and less resistive passive film on Type 316L stainless steel [11,12].

However, A. Elbiache and P. Marcus reported that molybdenum (Mo) could improve the resistance to sulfur of Fe-Cr-Ni-Mo alloys in sulfuric acid conditions [13]. Additionally, Akiko et al. demonstrated that Mo could prevent localized corrosion of Ni-Cr-Mo-Fe alloys by producing a molybdenum sulfide passive film with cation selectivity in the NaCl and H₂S environments [14]. Recent research by Xiao et al. indicated that sulfur ions (S²⁻) enhanced the re-passivation ability of 2205 duplex stainless steel and prevented pitting corrosion in the NaCl solution with the high pH [15].

According to the SKB-TR-10-67 report, when the canister was closed and the bentonite used as backfill was saturated (up to 100 years, oxygen existed), the values of pH, HS⁻ and Cl⁻ in the groundwater were expected to be 6.8–8, 0–1.3 mM, and 0.05–0.25 M, respectively [16]. Therefore, unlike previous studies, this research aims to explore the corrosion behavior of Type 316 stainless steel and the effects of pH, HS⁻, and Cl⁻ under weak alkaline conditions with very low concentrations in groundwater.

To achieve this, response surface methodology (RSM) is used in this study. It is a highly efficient statistical analysis method that is useful for finding the relationship between several factors and one or more responses with a reasonable number of experiment designs [17]. Additionally, RSM can be used to generate a meaningful quadratic model to help predict the localized corrosion of stainless steel. In this study, RSM was employed to discern the relationship between the three corrosion factors (pH, HS⁻, and Cl⁻) and two response variables: the passive film breakdown potential (E_break) and the protection potential (E_prot) of stainless-steel Type 316 (UNS S31600) [18,19,20]. Cyclic Potentiodynamic Polarization tests (CPP) were utilized to obtain the two response variables, and the experiment with the three factors was designed using the Box-Behnken design (BBD) to model E_break and E_prot. Additionally, X-ray photoelectron spectroscopy (XPS) and cyclic voltammetry (CV) were conducted to analyze the passivation film concerning the corrosion factors.

2. Materials and Methods

2.1. Specimen Preparation

In this study, commercial-grade stainless steel Type 316, rolled to a thickness of 500 μm, was utilized. The chemical composition of the specimens is presented in

Table 1. Chemical composition of Type 316 stainless steel (wt.%).

Specimens		Composition (wt.%)
	C	Cr	Ni	Mo	Mn	P	S	Si	N	Fe
Type 316	0.02	16.5	10.2	2.2	1.4	0.02	0.001	0.5	0.04	Bal.

. Before all electrochemical testing, the specimens underwent polishing with Si-C paper ranging from grit 220 to 600 [21]. This step ensured uniformity among the specimens, establishing a standardized surface for testing. The exposed surface area of each specimen was precisely adjusted to 1 cm² using silicon paste.

2.2. Design of Experiments for Response Surface Methodology

RSM, a modern statistical and mathematical technique, was employed to model the response variables influenced by multiple factors. This study aimed to investigate how three factors, namely pH (A), log[HS⁻] (B), and log[Cl⁻] (C), influence the response variables E_break and E_prot under groundwater conditions. The BBD was chosen for the experiment matrix design due to its efficiency when dealing with three factors. Minitab (Minitab 19, Minitab, LLC., State College, PA, USA) and Statistita (Statistita v10, Stasista, Hamburg, Germany) software were used for the design of experiments and statistical analysis [22].

To ensure a comprehensive investigation, this study focused on three factors: pH and concentrations of HS⁻ and Cl⁻.

Table 2. The investigated value range of factors in this study.

Investigated Factor	A, pH	B, log[HS− (mM)]	C, log[Cl− (M)]
Investigated range	8 to 10	−2 to 0	−3 to −1

outlines the ranges of pH and concentrations of HS⁻ and Cl⁻ examined in this study [23]. To simplify the adjustment and accuracy of the solutions, concentrations of HS⁻ and Cl⁻ were converted to a logarithmic scale. For statistical calculations, variables were coded using the minimum, medium, and maximum points of each variable separately, with values of −1, 0, and 1, respectively. The number of experiments was determined using the formula N = k² + k + c_p, where k represents the number of factors, 3, and c_p represents the number of experiments running at the center point, 3 [24]. The schematic of the design of experiments is illustrated in Figure 1 and

Table 3. Experiments are designed based on BBD.

Standard Run	Coded Parameter			Real Parameter
	A	B	C	pH	HS−	Cl−
					(mM)	(M)
1	$-$1	$-$1	0	8	0.01	0.01
2	$-$1	0	$-$1	8	0.1	0.001
3	$-$1	0	1	8	0.1	0.1
4	$-$1	1	0	8	1	0.01
5	0	$-$1	$-$1	9	0.01	0.001
6	0	$-$1	1	9	0.01	0.1
7	0	0	0	9	0.1	0.01
8	0	1	$-$1	9	1	0.001
9	0	1	1	9	1	0.1
10	1	$-$1	0	10	0.01	0.01
11	1	0	$-$1	10	0.1	0.001
12	1	0	1	10	0.1	0.1
13	1	1	0	10	1	0.01
14	0	0	0	9	0.1	0.01
15	0	0	0	9	0.1	0.01

, including 3 repeated center points and 12 different points [17]. Therefore, the total number of experiments to be conducted (N) was 15. Then, through the BBD results, a quadratic regression model can be created as shown in Equation (1).

$$\mathrm{Y}={\mathsf{\beta }}_0+\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathsf{\beta }}_{\mathrm{i}}{\mathrm{X}}_{\mathrm{i}}+\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathsf{\beta }}_{\mathrm{ii}}{\mathrm{X}}_{\mathrm{i}}^2+\sum _{\begin{array}{c}\mathrm{i},\mathrm{j}=1\\ \mathrm{i}\ne \mathrm{j}\end{array}}^{\mathrm{n}}{\mathsf{\beta }}_{\mathrm{ij}}{\mathrm{X}}_{\mathrm{i}}{\mathrm{X}}_{\mathrm{j}}$$

(1)

where Y is the response, β₀ is a constant coefficient, β_i, β_ii, and β_ij are linear, quadratic, and interaction coefficients, respectively, and X_i and X_j are the coded values of the variables [24]. Responses were set to E_break and E_prot, and the model quality was confirmed by analysis of variance (ANOVA).

2.3. Electrochemical Tests

Electrochemical tests, including CPP and CV, were conducted using a three-electrode method in a 1000 mL Pyrex glass cell. The specimens were connected to the working electrode, while a saturated calomel electrode (SCE) served as the reference electrode, and two glassy carbons were employed as the counter electrode. The experimental medium was distilled water, with HS⁻ concentration controlled using Na₂S and Cl⁻ concentration controlled with NaCl. pH was adjusted using a 1 M sodium hydroxide solution and a saturated boric acid solution. All electrochemical tests were conducted using a VSP-300 instrument (Bio-Logic SAS, Seyssinet-Pariset, France) at 23 °C. The experiments included an open-circuit potential (OCP) for 24 h, and CPP scans started from 0 V vs. OCP with a rate of 0.166 mV/s. The back scan was initiated when the current density reached 0.1 mA/cm² at the same rate. Surface images of all specimens were collected for comparative analysis using an optical microscope (OM), and for the 0.001 M Cl⁻ conditions, additional pitting information was collected using a confocal laser scanning microscope (CLSM). CV tests were also employed with OCP for 24 h, and the scan was repeated from −1.7 V_SCE to 0.7 V_SCE with a rate of 50 mV/s for 30 cycles. The solutions used for CV were pH 8, 1 mM HS⁻, and 0.1 M Cl⁻ and pH 8, 0.01 mM HS⁻, and 0.1 M Cl⁻.

2.4. XPS Analysis

XPS analysis was performed to analyze the chemical composition of the passive film after immersion for 24 h in different concentrations of HS⁻. The solutions used for XPS analysis were also the same with CV tests (pH 8, 1 mM HS⁻, and 0.1 M Cl⁻ and pH 8, 0.01 mM HS⁻, and 0.1 M Cl⁻). XPS analysis was carried out using an ESCA system (AXIS SUPRA, Kratos, Manchester, UK) with a monochromatic Al Kα radiation source (photoelectron energy of 1486.6 eV). High-resolution spectra were recorded in increments of 0.1 eV with a pass energy of 20 eV. The binding energies of the elements were calibrated to the carbon 1 s orbital at 284.8 eV.

3. Results and Discussion

3.1. Cyclic Polarization Measurements

To accurately determine the type of sulfur present in the solution, an understanding of the dissociation of H₂S in aqueous solutions is essential. The reactions and equilibrium equations in the solution are as follows [11,15]:

$${\mathrm{H}}_2\mathrm{S}{=\mathrm{HS}}^{-}+{\mathrm{H}}^{+},{\mathrm{K}}_{a1}$$

(2)

$${\mathrm{HS}}^{-}={\mathrm{S}}^{2-}+{\mathrm{H}}^{+},{\mathrm{K}}_{a2}$$

(3)

$${\mathrm{C}}_{{\mathrm{Na}}_2\mathrm{S}}={\mathrm{C}}_{{\mathrm{H}}_2\mathrm{S}}\times (1+\raisebox{1ex}{$K_{a1}$}\!\left/ \!\raisebox{-1ex}{${\mathrm{C}}_{{\mathrm{H}}^{+}}$}\right.+K_{a1}\times \raisebox{1ex}{$K_{a2}$}\!\left/ \!\raisebox{-1ex}{${{\mathrm{C}}_{{\mathrm{H}}^{+}}}^2$}\right.)$$

(4)

Sun et al. suggested approximate values of K_a1 and K_a2 for Equations (2) and (3) at room temperature to be 10⁻⁷ mol/L and 10⁻¹²–10⁻¹⁹ mol/L, respectively [25]. Considering the pH range of 8–10 in this study, the concentration of H₂S can be negligible because it is approximately 11 to 10³ times lower than the concentration of Na₂S according to Equation (4), and the dissociation of HS⁻ is also negligible because of very low K_a2. Thus, it was assumed that the concentration of HS⁻ used in this study was equivalent to the concentration of dissolved Na₂S.

Figure 2a–c depict the cyclic polarization curves with the designed experiment BBD according to the concentration of Cl⁻ and the parameters of the results of CPP are presented in

Table 4. The results of the CPP are based on BBD.

Run	pH	HS− (mM)	Cl− (M)	EOCP (mVSCE)	ipass (μA/cm2)	Ebreak (mVSCE)	Eprot (mVSCE)
1	8	0.01	0.01	−64	1.14	999	227
2	8	0.1	0.001	−126	12.96	1238	483
3	8	0.1	0.1	−121	0.99	652	42
4	8	1	0.01	−313	10.19	942	339
5	9	0.01	0.001	−79	5.09	1209	389
6	9	0.01	0.1	29	1.32	649	33
7	9	0.1	0.01	−86	1.79	982	312
8	9	1	0.001	−398	39.6	1263	1263
9	9	1	0.1	−313	10.23	593	177
10	10	0.01	0.01	−144	1.33	984	271
11	10	0.1	0.001	−168	24.22	1310	409
12	10	0.1	0.1	−83	1.14	725	−11
13	10	1	0.01	−373	11.03	758	323
14	9	0.1	0.01	−108	2.14	1022	213
15	9	0.1	0.01	−176	2.53	1052	257

. The surfaces of the specimens after CPP were measured using OM and CLSM, as shown in Figures S1 and S2. An important point in surface observation after CPP is that no pitting was found under No. 8 experiments (pH 9, HS⁻ 1 mM, Cl⁻ 0.001 M).

E_break values were selected at the point where the current density sharply increased within the passive region (<1 V_SCE), while E_break reaching the transpassive region (>1 V_SCE) was considered the point where the anodic curve’s slope changed. The current density at the E_break was set to the passive current density (i_pass). Additionally, during the back scan, the point where the polarization curve intersects was defined as E_prot. Regression models for E_break and E_prot were created and analyzed through ANOVA based on the results in Table 4.

3.2. Quadratic Regression Model and Analysis of Variance

The response variable, E_break, was analyzed using Minitab 19 software to obtain a quadratic regression model as shown below:

$$\begin{gathered} {\mathrm{E}}_{\mathrm{break}}\left({\mathrm{mV}}_{\mathrm{SCE}}\right)=-1200+358·\mathrm{pH}+139·\log \left[{\mathrm{HS}}^{-}\right]-389·\log \left[{\mathrm{Cl}}^{-}\right]-22.6·{\mathrm{pH}}^2-75.3·(\log {\left[{\mathrm{HS}}^{-}\right])}^2 \\ -14.8·(\log {\left[{\mathrm{Cl}}^{-}\right])}^2-42.2·\mathrm{pH}·\log \left[{\mathrm{HS}}^{-}\right]+0.2·\mathrm{pH}·\log \left[{\mathrm{Cl}}^{-}\right]-27.5·\log [{\mathrm{HS}}^{-}]·\log [{\mathrm{Cl}}^{-}] \end{gathered}$$

(5)

The statistical significance of Equation (5) should be assessed through the F-test and p-value [26]. The ANOVA results and coefficient of determination (R²) are presented in

Table 5. Results of the ANOVA of Equation (5) for predicting E_break.

ANOVA Fcritical (DFregression, DFresidual, α) (9, 5, 0.05) = 4.77
Source	Degree of Freedom (DF)	Adj. Sum of Square	Adj. Mean Square	F-Value	p-Value
Regression	9	763,514	84,835	15.39	0.004
Residual	5	27,564	5513
Lack of fit	3	25,097	8366
Pure error	2	2467	1233
Total	14	791,078
Coded Coefficient
Tcritical (DFresidual, α) (5, 0.05) = 2.57
Term	Coefficient	Standard Error Coefficient	T-value	p-value	Remark
Constant	1018.7	42.9	23.76	0.000	<0.05
A	−6.7	26.3	−0.26	0.807
B	−35.6	26.3	−1.36	0.233
C	−300.1	26.3	−11.43	0.000	<0.05
AA	−22.6	38.6	−0.58	0.584
BB	−75.3	38.6	−1.95	0.109
CC	−14.8	38.6	−0.38	0.717
AB	−42.2	37.1	−1.14	0.307
AC	0.2	37.1	0.01	0.995
BC	−27.5	37.1	−0.74	0.492
Summary Model
R2 = 96.52%		R2 (adj.) = 90.24%		R2 (pred.) = 48.54%

. ANOVA was employed to examine the relationship between the response variables and predictor variables. Generally, the significance level (α) was set to 0.05, and if the p-value of the regression model is less than α, it suggests that the model is statistically significant. In this model, the obtained F-value from the F-test was 15.39, exceeding the traditional threshold of F_critical (F_{(9, 5, 0.05)} = 4.77), and the p-value was below 0.05. Therefore, the model can reject the null hypothesis, which means that all coefficients are zero (${\mathsf{\beta }}_1={\mathsf{\beta }}_2={\mathsf{\beta }}_3={\mathsf{\beta }}_{11}={\mathsf{\beta }}_{22}={\mathsf{\beta }}_{33}={\mathsf{\beta }}_{12}={\mathsf{\beta }}_{13}={\mathsf{\beta }}_{23}=0)$ and the model is statistically significant. However, only variables with T-values outside the range of T_critical (T_{(5, 0.05)} = 2.75) are meaningful, and the results in Table 5 reveal that only the variable log[Cl⁻] (C) is significant in this model. Additionally, the R² (pred.) value is only 48.54%, indicating that the model has an overfitting problem.

To address non-significant variables, a systematic revision of the model was conducted by eliminating each non-significant variable. The detailed R² of revised Equations (5.1) and (5.6) summaries are provided in Table S1 of the supplementary section. The optimized Equation (6) was selected for the best model of E_break of the specimen.

$${\mathrm{E}}_{\mathrm{break}}\left({\mathrm{mV}}_{\mathrm{SCE}}\right)=288.8-180.9·\log \left[{\mathrm{HS}}^{-}\right]-300.1·\log \left[{\mathrm{Cl}}^{-}\right]-72.7·(\log {\left[{\mathrm{HS}}^{-}\right])}^2$$

(6)

With information from the optimized Equation (6) in

Table 6. Results of the ANOVA of the revised Equation (6) for predicting E_break.

ANOVA Fcritical (DFregression, DFresidual, α) (3, 11, 0.05) = 3.59
Source	Degree of Freedom (DF)	Adj. Sum of Square	Adj. Mean Square	F-Value	p-Value
Regression	3	750,464	250,155	67.75	0.000
Residual	11	40,614	19,710
Lack of fit	9	38,147	4239
Pure error	2	2467	1233
Total	14	791,078
Coded Coefficient
Tcritical (DFresidual, α) (11, 0.05) = 2.20
Term	Coefficient	Standard Error Coefficient	T-value	p-value	Remark
Constant	997.3	24.6	40.57	0.000	<0.05
B	−35.6	21.5	−1.66	0.125
C	−300.1	21.5	−13.97	0.000	<0.05
BB	−72.7	33.7	−2.31	0.041	<0.05
Summary Model
R2 = 94.87%		R2 (adj.) = 93.47%		R2 (pred.) = 90.19%

, it is evident that the model has been significantly simplified and improved. The F-value is notably higher than F_critical (F_{(3, 11, 0.05)} = 3.59). Importantly, only one variable (B) is non-significant in this refined model, according to the T-test. Notably, R² (pred.) exceeds 90%, while both R² and R² (adj.) remain higher than 90%, indicating a substantial improvement over the original model, and the model no longer exhibits signs of overfitting as observed previously.

Figure 3a–c are response surfaces and contour lines from Equation (6). According to Equation (6) and Figure 3, E_break is entirely independent of pH and depends solely on Cl⁻ and HS⁻. This result affirms that the neutral-to-slightly basic pH of groundwater environments does not affect E_break, aligning with previous studies [27,28]. Additionally, this model bears a striking resemblance to Strehblow and Titze’s Equation (7), illustrating the relationship between E_break and aggressive ion concentrations [29].

$${\mathrm{E}}_{\mathrm{break}}\left({\mathrm{mV}}_{\mathrm{SCE}}\right)=\mathrm{A}-\mathrm{B}·\log {\mathrm{C}}_{\mathrm{agg}}$$

(7)

In Equation (7), A and B are constants, where B signifies the aggressiveness of the specific ions. In our study, Equation (6) exhibits similarities but introduces an additional quadratic term for HS⁻. This implies that the relationship with log[HS⁻] is not purely a negative linear effect, as observed with log[Cl⁻] under the given conditions. The E_break attains its maximum value when the concentration of HS⁻ is 0.058 mM (10^−1.24 mM), as depicted in Figure 3c. However, given that the concentration of HS⁻ in groundwater is significantly lower than the concentration of Cl⁻, its influence appears unclear due to the dominating effect of Cl⁻. Particularly when the concentration of Cl⁻ is 0.1 M, confirming the effect of HS⁻ on localized corrosion of the specimen becomes challenging, as illustrated in Figure 2a.

The impact of HS⁻ is evident in E_prot, which increased with a concentration of 1 mM HS⁻ at all Cl⁻ concentrations, as shown in Figure 4a–c and Table 4. To statistically analyze the change in E_prot by each factor, a regression model of E_prot was created. The model for E_prot is presented in Equation (8) and

Table 7. Results of the ANOVA of the revised Equation (8) for predicting E_prot.

ANOVA Fcritical (DFregression, DFresidual, α) (9, 5, 0.05) = 4.77
Source	Degree of Freedom (DF)	Adj. Sum of Square	Adj. Mean Square	F-Value	p-Value
Regression	9	1,103,426	122,603	4.34	0.060
Residual	5	141,283	28,257
Lack of fit	3	136,363	45,454
Pure error	2	4921	2460
Total	14	1,244,710
Coded Coefficient
Tcritical (DFresidual, α) (5, 0.05) = 2.57
Term	Coefficient	Standard Error Coefficient	T-value	p-value	Remark
Constant	260.7	97.1	2.69	0.044	<0.05
A	−12.4	59.4	−0.21	0.843
B	147.8	59.4	2.49	0.055
C	−287.9	59.4	−4.84	0.005	<0.05
AA	−102.7	87.5	−1.17	0.293
BB	132	87.5	1.51	0.192
CC	72.8	87.5	0.83	0.443
AB	−15	84	−0.18	0.865
AC	5.3	84	0.06	0.953
BC	−182.5	84	−2.17	0.082
Summary model
R2 = 88.65%		R2 (adj.) = 68.22%		R2 (pred.) = 00.00%

.

$$\begin{gathered} {\mathrm{E}}_{\mathrm{prot}}\left({\mathrm{mV}}_{\mathrm{SCE}}\right)=-8277+1832·\mathrm{pH}+182·\log \left[{\mathrm{HS}}^{-}\right]-226·\log \left[{\mathrm{Cl}}^{-}\right]-102.7·{\mathrm{pH}}^2+132.0·(\log {\left[{\mathrm{HS}}^{-}\right])}^2 \\ +72.8·(\log {\left[{\mathrm{Cl}}^{-}\right])}^2-15.0·\mathrm{pH}·\log \left[{\mathrm{HS}}^{-}\right]+5.2·\mathrm{pH}·\log \left[{\mathrm{Cl}}^{-}\right]-182.5·\log [{\mathrm{HS}}^{-}]·\log [{\mathrm{Cl}}^{-}] \end{gathered}$$

(8)

However, as listed in Table 7, the F-value in Equation (8) was lower than F_critical, while R² (pred.) was also 0, indicating that this model was not statistically significant and was under an overfitting problem. Despite attempts to optimize the model depicted in Table S2, this issue persisted. The overfitting can be attributed to the significant deviation observed in experiment No. 8. In this experiment, with the same Cl⁻ concentration of 0.001 M (equivalent to 1 mM) and HS⁻ concentration, the polarization curve of the stainless steel did not even exhibit a hysteresis loop, suggesting that the specimen resisted pitting corrosion under these conditions. These results emphasize that HS⁻ is not exclusively an aggressive ion; it can act as an inhibitor, contributing to an increase in the resistance to pitting corrosion of the specimen, particularly when considering the ratio of HS⁻ concentration in the Cl⁻ condition.

3.3. Surface Analysis and Corrosion Mechanism

XPS analysis was performed to provide evidence of surface changes and clarify the effect on the passive film depending on the concentration of HS⁻. Type 316 stainless steel is an Fe-Cr-Ni-Mo alloy, and Fe, Cr, and Mo can mainly affect the properties of the passivation film. The specimens used in the XPS analysis were immersed in 0.01 mM and 1 mM HS⁻ solutions for 24 h, with a fixed pH 8 and Cl⁻ concentration of 0.1 M, respectively. Figure 5a–f show the results of the XPS spectra of Cr 2p, Fe 2p, Mo 3d, O 1p, S 2s, 2p, and Ni 2p, respectively.

Table 8. Binding energies of the primary compounds of the passive film.

Element	Peak	Species (Binding Energy/eV)	Reference
Cr	2p3/2	Crmet (573.6 ± 0.1); Cr2O3 (576.0 ± 0.1); Cr(OH)3 (577.3 ± 0.1)	[36]
Mo	3d5/2	Momet (227.3 ± 0.1); Mo4+ (228.5 ± 0.1); Mo6+ (231.8 ± 0.1)	[37,38,39]
	3d3/2	Momet (230.5 ± 0.1); Mo4+ (232.9 ± 0.1); Mo6+ (234.8 ± 0.1)
Fe	2p3/2	Femet (706.5 ± 0.1); FeS2 (707.3 ± 0.1); Fe3O4 (709.5± 0.1); Fe2O3 (710.7 ± 0.1); FeOOH (711.8 ± 0.1); Fe2(SO4)3 (713.0 ± 0.1)	[15,40,41]
O	1s	O2− (529.9 ± 0.1); OH- (531.2 ± 0.1); H2O (532.6 ± 0.1)	[42,43]
S	2s	S2− (225.8 ± 0.1)	[44,45,46,47]
	2p3/2	S2− (161.2 ± 0.1); S22− (162.8 ± 0.1); SO42− (168.3 ± 0.1)
	2p1/2	S2− (162.6 ± 0.1); S22− (164.8 ± 0.1); SO42− (169.6 ± 0.1)
Ni	2p3/2	Nimet (852.5 ± 0.1); NiO (853.7 ± 0.1)	[48]

shows the binding energies of the primary compounds of the passive film obtained from XPS spectra deconvolution. The XPS spectra of Type 316 stainless steel showed no significant difference in its chemical compositions, even when the HS⁻ concentration changed. However, as the concentration of HS⁻ increased, the area fraction ratios of Cr(OH)₃/Cr₂O₃ and OH⁻/O²⁻ increased, as shown in

Table 9. Area fraction ratio of deconvolted XPS spectra of passive film.

Concentration of HS− (mM)	Area Fraction Ratio
	Cr(OH)3/Cr2O3	OH−/O2−
0.01	0.70	0.91
1	0.79	1.08

. Compared to Cr₂O₃, the content of Cr(OH)₃ in the passive film has less protective resistance and more defects [30,31]. Therefore, when the concentration of HS⁻ increases, the protection of the passivation film decreases, so that i_pass increases, as shown in Figure 2 and Table 4. On the other hand, when HS⁻ is present in the solution, Ni is not significantly affected, but Fe and Mo sulfides, such as FeS₂, MoS₂, and MoS₃, can be generated and act as a passive film, as shown in Figure 5b,c,e [32,33,34]. These results can also be confirmed through the Pourbaix diagram. The Pourbaix diagram of Fe-H₂O-S and Mo-H₂O-S systems presented by Davoodi et al. showed that FeS₂ and MoS₂ are stable phases in OCP measured in this study [35].

Additionally, CV tests were performed for passive film analysis according to concentrations of HS⁻, as depicted in Figure 6a,b. When CV was performed depending on the concentration of HS⁻, a notable difference occurred in the oxidation peaks. Also, as the CV cycles were repeated, it was confirmed that the intensity of the peaks became more clear due to the metal ions eluted from the specimen. The peak I between −0.9 and −1.0 V_SCE indicates that Fe is oxidized to a Fe²⁺ ion, and the reduction peak of the reaction is displayed by peak I′. The pick II between −0.5 and −0.6 V_SCE indicates that the Fe²⁺ ion is oxidized to the Fe³⁺ ion, and the reduction peak of the reaction is represented by peak II′ [49]. The peak IV indicates that the Cr³⁺ ion is oxidized to the Cr⁶⁺ ion, and peak V indicates the cathodic reduction of dissolved oxygen [50]. The response of peak I depending on the concentration of HS⁻ is explained by the following equations [51].

$$\mathrm{Fe}+\mathrm{H}{\mathrm{S}}^{-}\text{→}\mathrm{Fe}(\mathrm{HS}{)}_{\mathrm{ads}}+{\mathrm{e}}^{-}$$

(9)

$$\mathrm{Fe}(\mathrm{HS}{)}_{\mathrm{ads}}+{\mathrm{OH}}^{-}\text{→}\mathrm{FeS}+{\mathrm{H}}_2\mathrm{O}+{\mathrm{e}}^{-}$$

(10)

Then, Haleem et al. suggested that peak III (~0.189 V_SCE) also generated FeS₂ with the following equation [52].

$$\mathrm{Fe}+2\mathrm{H}{\mathrm{S}}^{-}+2\mathrm{O}{\mathrm{H}}^{-}\text{→}\mathrm{Fe}{\mathrm{S}}_2+{2\mathrm{H}}_2\mathrm{O}+4{\mathrm{e}}^{-}$$

(11)

On the other hand, other studies suggested that the peaks III and III′ are the reactions of MoS₄²⁻ ions with the following equations [53,54].

$$\mathrm{Mo}{\mathrm{S}}_4^{2-}\text{→}\mathrm{Mo}{\mathrm{S}}_3+\frac{1}{8}{\mathrm{S}}_8+2{\mathrm{e}}^{-}$$

(12)

$$\mathrm{Mo}{\mathrm{S}}_3+\frac{1}{8}{\mathrm{S}}_8+2{\mathrm{e}}^{-}\text{→}\mathrm{Mo}{\mathrm{S}}_4^{2-}$$

(13)

$$\mathrm{Mo}{\mathrm{S}}_4^{2-}+2{\mathrm{H}}_2\mathrm{O}+2{\mathrm{e}}^{-}\text{→}\mathrm{Mo}{\mathrm{S}}_2+2\mathrm{H}{\mathrm{S}}^{-}+2\mathrm{O}{\mathrm{H}}^{-}$$

(14)

In peak III, MoS₃ is generated by Equation (12). On the other hand, the peak III′ overlaps the reduction reactions of Equations (13) and (14). First, the reduction reaction of MoS₃ occurs according to Equation (13), and then MoS₂, a new reductive deposition material, is generated according to Equation (14) below about −0.55 V_SCE.

Through XPS and CV analyses, it is clear that a new passive film is formed on the surface due to HS⁻. The presence of HS⁻ leads to an increased ratio of OH⁻/O²⁻, causing the dense and stable Cr₂O₃ on the passive film to decrease and become more susceptible to destruction. However, conversely, the broken passive film on the surface is re-passivated into new materials such as FeS₂, MoS₂, and MoS₃ by the abundant HS⁻. Particularly when the concentrations of Cl⁻ and HS⁻ were equal in the groundwater condition, it was confirmed that the destroyed passivation film immediately underwent re-passivation, providing resistance to the corrosion of Type 316 stainless steel. Additionally, as depicted in Figure 7, no hysteresis loop was observed in the specimens after conducting CPP under conditions of pH 8 and 10, 0.001 M Cl⁻, and 1 mM HS⁻. This indicates that the resistance to pitting corrosion of Type 316 stainless steel remains unaffected by pH under these specific mildly alkaline conditions.

4. Conclusions

This study aimed to assess the suitability of Type 316 stainless steel for spent nuclear fuel disposal canisters, employing statistical analysis to scrutinize the impact of pH, HS⁻, and Cl⁻ on localized corrosion in groundwater. Key findings include:

• The concentrations of Cl⁻ and HS⁻ were significant factors affecting E_break on a logarithmic scale. Especially HS⁻ reduces the ratio of Cr₂O₃, which plays a crucial role in the passive film of stainless steel.
• On the other hand, HS⁻ reacts with Fe and Mo, the main components of the passive film on Type 316 stainless steel, forming new metal sulfides on the passive film and contributing to the re-passivation behavior of Type 316 stainless steel.
• Under specific conditions (Cl⁻ concentration 0.001 M, equal HS⁻ concentration), the alloy re-passivates the broken film immediately with metal sulfides (FeS₂, MoS₂, MoS₃), displaying resistance to pitting corrosion.
• Therefore, regardless of pH, Type 316 stainless steel exhibits no hysteresis loop in the cyclic polarization curve under these conditions, suggesting negligible localized corrosion concerns.

In conclusion, this study identifies critical Cl⁻ and HS⁻ concentrations, affirming Type 316 stainless steel’s pitting resistance in weak alkaline groundwater with oxygen and holding valuable implications for its use in the disposal of radioactive waste in geological repositories.