Evaluation of chloride stress corrosion cracking susceptibility of stainless steels

1 Introduction

Stainless steels are used in many different applications in and around ocean environments, and chloride stress corrosion cracking (Cl-SCC) can cause material failure, depending on the environmental exposure, grade, and condition of the stainless steel (SS), and the applied stress. This work presents experiments toward developing a Cl-SCC test method that is suitable for making atmospheric corrosion measurements to assess Cl-SCC susceptibility. The first part of the testing used aqueous immersion exposures of boiling MgCl₂. The corrosion susceptibility was evaluated in situ for 304 L, 316 L, 317 L, and AL-6XN, using fracture mechanics parameters as a measure of Cl-SCC susceptibility. Two parameters were used to evaluate susceptibility: the stress intensity necessary to initiate Cl-SCC (K_ISCC) and CGR at higher stress intensity values (Stage-II CGR). The second part of the testing used the in situ methodology of the aqueous immersion exposures but applied it in atmospheric conditions. In this testing, the double cantilever beam (DCB) geometry used in the immersion testing was modified to incorporate a smooth gauge section. MgCl₂ salt was placed in the smooth gauge area of the specimen, and the hygroscopic action of MgCl₂ in a controlled humidity chamber created the corrosive environment, notionally like what sea salt, which contains MgCl₂, would also do. As was the case for the immersion test, the occurrence of Cl-SCC was measured with in situ load shedding and DCB arm displacement.

2 Materials and methods

DCB stresses and deflections were predicted using FEMAP/Nei NASTRAN structural FEM software. To validate the approach used in this work, results were compared with prior studies of Speidel (1981) under nominally identical conditions. Stress intensity is not an intrinsic output of this software package, so a different approach was required to acquire the desired mode I stress intensity (K_I) versus crack depth characteristics of a given sample design. Therefore, this work utilized the virtual crack closure technique (VCCT) (Krueger (2002)) to calculate stress intensity using nodal displacements and forces in the vicinity of the crack tip that are readily obtainable with the FEMAP postprocessor. The benefit of this approach is its applicability to highly arbitrary geometries for which analytical or empirical K₁ solutions may not be available. Speidel (1981) offered a formula relating crack growth to arm displacement but did not present information regarding any limits of applicability on sample geometry for the formula including crack length and sample width (parameters a and H in Figure 8 of Speidel (1981)). The FEA/VCCT method was used to verify that the equation was in fact general enough to be applicable to the updated geometries. The method was validated against Speidel’s equation using three different specimen geometries, showing close agreement between the FEA/VCCT method and Speidel’s equation.

The test apparatus was designed to stress a DCB sample that was partially immersed in a solution at an elevated temperature. The load frame could apply up to 22.2 kN (5,000 lbf) force and was instrumented to continuously monitor the applied load and resulting beam deflection. The combination of load and beam deflection data enabled real-time estimation of crack length using DCB samples and mechanical test systems. Figure 1 shows details of the gap sensor used to measure deflections caused by cracking. The fracture test system was used to measure the Cl-SCC behavior of 304 L, 316 L, 317 L, and AL-6XN double cantilever beam (DCB) samples in an aqueous 42 % MgCl₂ solution at 130 °C. The samples were fabricated from wrought plate and consideration given to using the S-L orientation, but the plate was not of sufficient thickness for this; therefore, the orientation of the cracking kept in the T-L orientation in the initial aqueous, as well as the later atmospheric testing. All of the alloys have austenitic microstructure and corrosion resistance is expected to increase in the order of 304 L, 316 L, 317 L, and AL-6XN, primarily related to molybdenum content (increasing from 304 L to Al-6XN). Earlier work by Speidel (1981) used a CGR threshold sensitivity of 3 × 10⁻¹¹ m/s to measure K_ISCC, and this was adopted for this work. Since the displacement sensor could detect movement of 15 µm, a CGR of 3 × 10⁻¹¹ m/s corresponded to an initial dwell of 145 h at which time loads could be increased if no in situ displacement had been detected, and this was adopted in the experiments. A rising step load procedure was adopted and if cracking initiated quickly, then a lower load was used on other specimens to determine K_ISCC.

Figure 1:

DCB assembly showing gap sensor and loading pins.

Initially, the ligament length of the specimens was designed to be 65 mm, and a fatigue precrack was placed at the notch in order to initiate cracking. However, at high loads cracking was off-center and at low loads cracking initiated at the sides of the notch (Figure 2). The lack of cracking at the notch center at low loads was likely due to compressive stresses at the notch resulting from the precracking procedure as well as the initial DCB geometry. Furthermore, the low loads necessary to initiate cracking in the aggressive MgCl₂ solution meant that lower stresses at sides of the DCB groove were sufficient to nucleate cracks preferentially to the notch area. Finite element modeling (FEM) showed that by shortening the ligament from 65 mm to 20 mm, the stresses could be concentrated at the notch preferentially to the sides of the notch groove (Figure 3). Based on this, the specimen geometry was changed to a 20 mm ligament length and the precrack procedure was discontinued. Because the crack length in a DCB geometry includes the relatively long 25 mm distance from the load pins to the crack notch, a crack extension of 1 mm is a small fraction of the total crack length, and the 1 mm extension does not substantially change the calculated stress intensity. In this work, the minimum load necessary to cause arm deflection corresponding to a 1 mm crack extension was used to calculate K_ISCC, and this provides an upper-bound estimate of K_ISCC. This method was used to calculate K_ISCC in 42 % MgCl₂ at 130 °C for 304 L, 316 L, 317 L, and AL-6XN. Figure 4 shows an example of cracking, and Figure 5 shows an example of crack length data obtained using this procedure.

Figure 2:

Initial testing with long ligament produced off-center cracking (above) at high loads and crack initiation at the sides of the notch grove (below) at lower loads, necessitating a redesign of the specimen geometry.

Figure 3:

Finite element modeling (FEM) shows the calculated Mises stress of the initial specimen geometry (left) and the redesigned geometry (right) to produce cracks along the centerline that initiate at the notch. Units are in MPa√m.

Figure 4:

Cracks initiated at the notch in MgCl₂ without a precrack (below) with a shortened DCB geometry, compared to the original longer DCB geometry with a precrack that had crack closure and no cracks initiated at the notch (above).

Figure 5:

Example of crack growth measurement in 304 L DCB specimens using no precrack and with the crack initiated at the notch.

Figure 6 shows the geometry used for the test specimens for the atmospheric testing. The load frame, displacement measurement, and load measurement equipment were as in the aqueous immersion tests (Figure 1). The application of a room temperature saturated MgCl₂ solution was through the use of a fiberglass membrane, dosed with the salt solution, placed at the smooth gauge section, and held in place by a Teflon rod as shown in Figure 7. Type 304 L stainless steel DCB samples with and without salt were placed in a temperature and RH chamber. Environmental sensors from Luna Labs mapped against NIST-traceable temperature, and RH probe were used to verify temperature and RH control. Work in Prosek and Iversen (2008) evaluated the effect of FeCl₃ on Cl-SCC of stainless steels, and following this, in one case, a 1018 steel rod crevice former was placed in the smooth gauge section, recognizing that FeCl₃ would be present in the salt mixture as the 1018 steel corroded and could serve as an additional environmental stress factor.

Figure 6:

DCB-like sample geometry used for atmospheric testing. Note that the smooth gauge area has a radius of 12.5 mm and the same ligament distance (25 mm) as the earlier aqueous work. All units are in mm.

Figure 7:

Setup of smooth gauge DCB specimens with MgCl₂ salt. First, a hydrophilic fiberglass membrane was placed in the smooth gauge section (top left), then it was dosed with MgCl₂ salt, and then a Teflon rod (bottom left) was placed in the smooth gauge section.

Figure 8:

The load necessary to give the strain used in Prosek and Iversen (2008) and Prosek et al. (2009, 2014 u-bend samples was calculated and applied in the initial testing. The principal strain is shown on the left and the Mises stress is on the right.

Abaqus FEM software was used to calculate the load necessary to achieve the stress or strain condition that was targeted (e.g., Figure 8). Initially, a strain of 15 % was targeted to match the conditions of the U-bends of Cl-SCC investigations of earlier work by Prosek and Iversen (2008) and Prosek et al. (2009, 2014. However, the load that produced 15 % strain caused low-temperature creep, and substantial plastic flow (see Figure 9) consistent with Alden (1987), and the load shedding and arm displacement from creep was large enough to mask in situ displacement and load shedding measurements caused by Cl-SCC. As a result, the applied load was decreased to 90 % of YS, and at this decreased load Cl-SCC cracking could be detected in situ.

Figure 9:

When the load was set to 90 % of YS at the bottom of the smooth gauge section, there was minimal arm displacement (A, 0.08 mm on the left), and when the load was set to achieve 15 % principal strain, there was substantial creep reflected by the arm displacement that was not related to SCC (B, 13 mm additional displacement on the right). The DCB in (A, left) corresponds to that shown in Figure 14A, and the DCB in 14B (right) corresponds to that shown in Figure 14C.

3 Results and discussion

3.1 Cracking behavior in boiling 42 % MgCl₂ at 130 °C

A substantial part of this work consisted in the setting up of experimental protocols and procedures. The work in aqueous MgCl₂ on 304 L was compared to earlier work of Speidel (1981). The upper limit for K_ISCC for 304 L was measured to be approximately 2.5 MPa√m, whereas Speidel (1981) measured K_ISCC to be approximately 8 MPa√m, and this comparison is shown in Figure 10. No precrack was used in the modified DCB geometry where cracks initiated at the notch, and the aggressive nature of MgCl₂ solution acted as a source of crack initiation. The details of the precrack procedure for Speidel (1981) are not clear, but based on the difficulty of initiating a preferential crack with precracked samples in the present work, the higher K_ISCC measured by Speidel (1981) may have been created by residual compressive stresses, as was the case for the initial samples of this work. As explained earlier, in this work K_ISCC was estimated for 1 mm crack extension calculated from DCB arm deflection, with the crack length including the 25 mm distance from the load point to the notch.

Figure 10:

With the redesigned DCB geometry, the stress intensity necessary to induce Cl-SCC (K_ISCC) in type 304 SS was measured to be lower than that measured by (Speidel 1981).

The cracking behavior for 316 L, 317 L, and AL-6XN was also measured, and the CGR as a function of K₁ is in Table 1 and Figure 11. Types 304 L, 316 L, and 317 L all were measured to have low K_ISCC of approximately 2.5 MPa√m in boiling concentrated MgCl₂ (two 304 L samples cracked immediately when K was raised from 2.5 to 5 MPa√m). AL-6XN was measured to have a much higher K_ISCC than the other alloys at approximately 20 MPa√m. The stage-II CGR as measured by arm deflection is also summarized in Figure 12, which shows CGR to be highest to lowest in the order of 304 L, 316 L, 317 L, and AL-6XN, with AL-6XN having a much lower CGR than the other alloys. This may be summarized as follows:

CGR304L > CGR316L≳CGR317L≫CGRAL−6XN

Table 1:

Summary of crack initiation in aqueous 130 °C MgCl₂.

304 L		316 L		317 L		AL-6XN
K (MPa√m)	Initiation time (h)	K (MPa√m)	Initiation time (h)	K (MPa√m)	Initiation time (h)	K (MPa√m)	Initiation time (h)
2.5	1	2.5	19	2.5	8	25	240b
5	146a	2.5	31	2.5	8	25	240b
5	146a	2.5	46	2.5	12	20	49
						20	241c
						20	241c
						15	143

1. ^aHeld for 145 h at 2.5 MPa√m, then K stepped to 5 MPa√m, upon which cracking occurred. ^bHeld for 239 h at 20 MPa√m, then stepped to 25 MPa√m, upon which cracking occurred. ^cHeld for 240 h at 15 MPa√m, then stepped to 20 MPa√m, upon which cracking occurred.

Figure 11:

Crack growth rate as a function of stress intensity for all the materials in this test.

Figure 12:

K_ISCC for 304 L, 316 L, 316 L, and AL-6XN (above, A) and stage-II crack growth rate for 304 L, 316 L, 317 L, and AL-6XN (below, B).

Visual observations of cracks were not as precise but were confirmatory of the in situ CGR measurements. It was hard to distinguish K_ISCC behavior of 304 L, 316 L, and 317 L because of the very aggressive property of the boiling concentrated MgCl₂, where only the very resistant AL-6XN alloy required a substantially higher K_ISCC to induce cracking, and consistent with this AL-6XN had a much slower stage-II CGR than the other alloys. The limitation of these measurements is that measured values for K_ISCC are only upper-bound estimates for the aggressive MgCl₂ aqueous environment and were made based on the DCB crack length including the distance from the load pins.

3.2 Atmospheric corrosion measurement

The atmospheric corrosion evaluation combined elements of work by Denhard (1960), Prosek and Iversen (2008) and Prosek et al. (2009, 2014. Denhard (1960) measured the effect of stress on Cl-SCC in boiling MgCl₂, while Prosek and Iversen (2008) and Prosek et al. (2009, 2014 studied the effect of atmospheric corrosion of chloride salts in U-bends having nominally one strain state (15 %). This work used a DCB configuration allowing varied load conditions with correspondingly different stress states in the DCB smooth gauge and dosing with MgCl₂ salt in controlled temperature humidity. This enables the Cl-SCC investigation of the effect of both the stress state and atmospheric chloride condition.

Atmospheric corrosion testing was performed on Type 304 L SS using MgCl₂ salts at 50 °C and 35 % RH. The aim was to develop a test that could detect the onset of Cl-SCC as a function of applied load. Abaqus FEM was performed as described earlier to determine the loads to use. Loads that gave high plastic strain (load to get 15 % strain) lead to low-temperature creep, as discussed earlier, masked the in situ detection of Cl-SCC. However, lower loads that delivered a peak stress of 90 % YS caused Cl-SCC cracking that was able to be detected in situ. Figure 13 shows that this stress in the presence of MgCl₂ caused cracking as indicated by in situ load shedding measurements compared to the control specimen that had no salt in the smooth gauge section. The specimen with a 1018 steel crevice also had load shedding, although the load shedding was delayed. Post-test observations showed Cl-SCC cracking (Figure 14). The specimen with a 1018 steel crevice former exhibited cracking predominantly at the edges of the crevice former. The reason for this is not immediately obvious but may be related to galvanic corrosion of 1018. Figure 13 shows that the sample with the 1018 crevice former had initially less load shedding than the sample loaded with salt without the 1,018 steel crevice former, and then at the end of the test (∼650 h, ⪅4 weeks), it had a steeper rate of change in the load than any of the other samples. Repeated testing will be necessary to confirm these results. Figure 14C also shows Cl-SCC cracking on a specimen with higher loads that produced maximum of 15 % strain in the smooth gauge section. The higher loads produced creep that prevented in situ load shedding and compliance measurements from distinguishing Cl-SCC from low temperature creep. Cl-SCC cracks were most often associated with pits, and there was increased crack nucleation as indicated by a higher density of cracks than at the lower load (90 % of YS), although the surface cracks were shorter than at the lower load condition.

Figure 13:

Load shedding was detected related to Cl-SCC in 304 SS with an applied stress of 90 % of YS.

Figure 14:

Cracks on surface of sample exposed to MgCl₂ salts at 35 % RH with sample at 0.9Ys without a 1,018 steel crevice former (A) and with a 1,018 steel crevice former (B). Cl-SCC with a load corresponding to max 15 % plastic strain (C). Cracking occurred in this sample, but low temperature creep prevented in situ detection of cracking (no clear difference from a corresponding high-load control). Note: the feature in the center is debris – the surface was intentionally not subjected to further cleaning.

4 Summary

This work investigated two experimental methods of evaluating the susceptibility of stainless steels to Cl-SCC. First, stainless steel alloys using DCB specimens in aqueous boiling MgCl₂ were evaluated by measuring fracture mechanics parameters K_ISCC and stage-II CGR. Second, a method for applying specific loads in more prototypic atmospheric test conditions was investigated. The findings of aqueous immersion MgCl₂ testing may be summarized as follows:

1. The geometry of DCB specimens needs careful consideration in order to get reasonable crack growth direction and crack initiation at the notch.

2. It is possible to use the notch in a DCB geometry without a precrack to get reasonable upper-bound measurements of K_ISCC as well as reasonable measurements of stage-II CGR.

3. The DCB geometry makes it possible to estimate an upper-bound to K_ISCC with a 1 mm crack extension because of the comparatively longer distance from the load point to the crack notch.

4. The aggressive nature of boiling MgCl₂ makes it difficult to make clear distinctions in K_ISCC between of 304 L, 316 L, and 317 L, measured to be between 2.5 and 5 MPa√m, although the K_ISCC of AL-6XN was measured to be substantially higher (15–25 MPa√m).

5. The measured stage-II CGRs were consistent with expected Cl-SCC susceptibility of the alloys and may be summarized as 304 L > 316 L ≳ 317 L >> AL-6XN.

The atmospheric testing may be summarized as follows:

1. A test was developed that could detect the in situ initiation of Cl-SCC.

2. Appropriate controls and loads were required to distinguish low temperature creep from Cl-SCC.

3. The effect of stress state on Cl-SCC can be evaluated using FEM to set loads and reasonable experimental techniques to measure loads and displacements.

4. This test protocol should have the capability to be extended to other materials and environmental conditions.