Influence of Thermal Treatment on SCC and HE Susceptibility of Supermartensitic Stainless Steel 16Cr5NiMo.

A 16Cr5NiMo supermartensitic stainless steel was subjected to different tempering treatments and analyzed by means of permeation tests and slow strain rate tests to investigate the effect of different amounts of retained austenite on its hydrogen embrittlement susceptibility. The 16Cr5NiMo steel class is characterized by a very low carbon content. It is the new variant of 13Cr4Ni. These steels are used in many applications, for example, compressors for sour environments, offshore piping, naval propellers, aircraft components and subsea applications. The typical microstructure is a soft-tempered martensite very close to a body-centered cubic, with a retained austenite fraction and limited δ ferrite phase. Supermartensitic stainless steels have high mechanical properties, together with good weldability and corrosion resistance. The amount of retained austenite is useful to increase low temperature toughness and stress corrosion cracking resistance. Experimental techniques allowed us to evaluate diffusion coefficients and the mechanical behaviour of metals in stress corrosion cracking (SCC) conditions.


Introduction
Supermartensitic stainless steels have been developed to offer increased corrosion resistance, especially in sour environments (containing hydrogen sulphide H 2 S) [1][2][3]. They also present remarkable weldability and fracture toughness even at low temperatures, in comparison with traditional martensitic stainless steels. Thanks to all these characteristics, they are suitable for various applications such as compressor impellers, offshore piping, naval propellers and aircraft components.
Corrosion resistance and ductility can be improved by reaching a homogeneous tempered martensitic structure with a certain amount of retained austenite. The martensitic transformation end temperature is typically close to room temperature [4][5][6], which is why a sub-zero cooling is often performed to avoid excessive retained austenite formation.
The high nickel content reduces the austenitizing temperature Ac 1 , therefore a low tempering temperature must be chosen to avoid the excessive formation of reversion austenite γ rev . Reversion austenite can become unstable during cooling and transform into fresh martensite (instead of retained austenite), implying a hardness increase, which is usually not acceptable [7]. On the other hand, the retained austenite fraction is stabilized by carbon and manganese enrichment. Samples were obtained by hot-forged rotors for centrifugal compressors. Considering that chromium carbides Cr 23 C 6 formed during stress relieving after forging are responsible for stainless steel's sensitization to corrosion, a solubilization treatment was performed to allow their dissolution. A typical tempering process is carried out at 630 • C, but the temperature can be varied due to mechanical requirements. In the present work, three different tempering temperatures were applied after quenching in order to obtain different amounts of retained austenite in the final microstructure [14].
The investigated material was the same as that of De Sanctis et al., thus austenite fractions were deduced according to [16] with reference to the alloy A results. The steels' identifying letters and thermal treatment conditions are summarized in Table 2. The differences in terms of austenite fractions were expected to modify the steels' susceptibility to stress corrosion cracking and hydrogen embrittlement. Thanks to the face-centered cubic (FCC) lattice, with elevated hydrogen solubility, the austenite phase works as a hydrogen trap, avoiding further diffusion towards cracks or other critical points [17,18].
After polishing, Beraha CdS etching was used for light microscope images. The presence of carbides, highlighted in purple-blue colours by etching, was evident for all treatment conditions. They were principally precipitated in correspondence with grain boundaries and are indicated by the red arrows in the figures (Figures 1a, 2a and 3a), while SEM images (Figures 1b, 2b and 3b) reported precipitated details with a higher magnification. Metallographic analysis showed a grain size within the range of 50-150 µm.

Permeation Tests
Test samples consisted of metal sheets 50 × 50 mm, 0.5 mm thick. They were subjected to hydrogen permeation after a prior mechanical surface cleaning by means of abrasive papers and washed with an ultrasonic cleaner with acetone. A test solution with H2SO4 1 N + As2O3 10 mg/L in distilled water was used and a current density of 1 mA/cm 2 was applied for all three materials. An Ag/AgCl reference electrode was used for voltage measurements. A nitrogen purging flux was applied inside the electrochemical cell to avoid any sample surface oxidation during the test execution.

Permeation Tests
Test samples consisted of metal sheets 50 × 50 mm, 0.5 mm thick. They were subjected to hydrogen permeation after a prior mechanical surface cleaning by means of abrasive papers and washed with an ultrasonic cleaner with acetone. A test solution with H2SO4 1 N + As2O3 10 mg/L in distilled water was used and a current density of 1 mA/cm 2 was applied for all three materials. An Ag/AgCl reference electrode was used for voltage measurements. A nitrogen purging flux was applied inside the electrochemical cell to avoid any sample surface oxidation during the test execution.

Permeation Tests
Test samples consisted of metal sheets 50 × 50 mm, 0.5 mm thick. They were subjected to hydrogen permeation after a prior mechanical surface cleaning by means of abrasive papers and washed with an ultrasonic cleaner with acetone. A test solution with H2SO4 1 N + As2O3 10 mg/L in distilled water was used and a current density of 1 mA/cm 2 was applied for all three materials. An Ag/AgCl reference electrode was used for voltage measurements. A nitrogen purging flux was applied inside the electrochemical cell to avoid any sample surface oxidation during the test execution.

Permeation Tests
Test samples consisted of metal sheets 50 × 50 mm, 0.5 mm thick. They were subjected to hydrogen permeation after a prior mechanical surface cleaning by means of abrasive papers and washed with an ultrasonic cleaner with acetone. A test solution with H 2 SO 4 1 N + As 2 O 3 10 mg/L in distilled water was used and a current density of 1 mA/cm 2 was applied for all three materials. An Ag/AgCl Materials 2020, 13, 1643 5 of 14 reference electrode was used for voltage measurements. A nitrogen purging flux was applied inside the electrochemical cell to avoid any sample surface oxidation during the test execution.
The diffusion coefficient was calculated by the time lag method according to the ISO 17081:2014 standard [19]; see Equation (1). Once calculated, the permeation flux integral curve, t L , was determined as the abscissae intercept of the steady state tangent line, while l was the sample thickness.

Slow Strain Rate Tests
Mechanical testing was carried out on as-received samples to determine the reference behaviour of each treatment condition after hydrogen charging by means of MTS CERT machine 647 All-temperature Hydraulic Wedge Grips (equipped with dedicated autoclave). The samples' geometry was in agreement with NACE TM0177 and dedicated threaded gripping ends were machined ( Figure 4). Tests were performed with reference to NACE TM0198 and ASTM G129 [20,21], and charging methods are reported in Section 2.3.1.
Materials 2020, 13, x FOR PEER REVIEW 5 of 14 The diffusion coefficient was calculated by the time lag method according to the ISO 17081:2014 standard [19]; see Equation (1). Once calculated, the permeation flux integral curve, tL, was determined as the abscissae intercept of the steady state tangent line, while l was the sample thickness. (1)

Slow Strain Rate Tests
Mechanical testing was carried out on as-received samples to determine the reference behaviour of each treatment condition after hydrogen charging by means of MTS CERT machine 647 Alltemperature Hydraulic Wedge Grips (equipped with dedicated autoclave). The samples' geometry was in agreement with NACE TM0177 and dedicated threaded gripping ends were machined ( Figure  4). Tests were performed with reference to NACE TM0198 and ASTM G129 [20,21], and charging methods are reported in Section 2.3.1. The test results were finally analyzed in terms of embrittlement index, calculated with reference to the samples' elongation at break, Equation (2). A reduction of 30% was considered as the threshold value in agreement with reference standard ASTM STP 962 [22]. (2)

Hydrogen Charging
Some samples were subjected to mechanical testing after electrochemical hydrogen charging in a cathodic protection condition (scheme reported in Figure 5), simulating subsea applications. Test specimens were electrically connected to a magnesium sacrificial anode in a saline solution (3.5% NaCl). The electrochemical potential difference between Mg and stainless steel gives rise to a current, because of the galvanic corrosion setup. The voltage between the anode and the cathode of the galvanic coupling was measured by means of a reference electrode and corresponded to 1.96 V. The test temperature varied from room temperature to 60 °C by means of a thermostatic bath.
Other samples were subjected to slow strain rate in an autoclave in order to simulate hydrogen absorption during operation, as in the external hydrogen embrittlement phenomenon. Specimens were immersed in NACE A solution during mechanical testing. Standard conditions were applied (pH = 3.5, hydrochloric acid [Cl − ] = 100 ppm, buffer CH3COONa, ptot = 1 bar, pH2S = 10% ptot, room temperature) [23]. The test results were finally analyzed in terms of embrittlement index, calculated with reference to the samples' elongation at break, Equation (2). A reduction of 30% was considered as the threshold value in agreement with reference standard ASTM STP 962 [22].

Hydrogen Charging
Some samples were subjected to mechanical testing after electrochemical hydrogen charging in a cathodic protection condition (scheme reported in Figure 5), simulating subsea applications. Test specimens were electrically connected to a magnesium sacrificial anode in a saline solution (3.5% NaCl). The electrochemical potential difference between Mg and stainless steel gives rise to a current, because of the galvanic corrosion setup. The voltage between the anode and the cathode of the galvanic coupling was measured by means of a reference electrode and corresponded to 1.96 V. The test temperature varied from room temperature to 60 • C by means of a thermostatic bath.
Other samples were subjected to slow strain rate in an autoclave in order to simulate hydrogen absorption during operation, as in the external hydrogen embrittlement phenomenon. Specimens were immersed in NACE A solution during mechanical testing. Standard conditions were applied (p H = 3.5, hydrochloric acid [Cl − ] = 100 ppm, buffer CH 3 COONa, p tot = 1 bar, p H2S = 10% p tot , room temperature) [23].

Instantaneous Strain Hardening Coefficient
By means of a rheological model with an instantaneous strain hardening coefficient, the plastic strain range was analytically described. The main aim of this investigation was to evaluate the possibility of observing the transformation of plastic strain-induced austenite to fresh martensite. Considering the study of Dimatteo [24], some materials present an inflection point in the strain hardening trend, thus an experimental data regression with a fourth grade polynomial was made to eventually appreciate this feature (Equation (3)).
The best fitting parameters and instantaneous strain hardening coefficient (Equation (4)) were evaluated for all materials.

Permeation Tests
Effective diffusion coefficients were calculated according to the time lag method (see paragraph 2.2) and the results are reported in Table 3. For completeness, permeation flux curves are plotted in Figure 6.

Instantaneous Strain Hardening Coefficient
By means of a rheological model with an instantaneous strain hardening coefficient, the plastic strain range was analytically described. The main aim of this investigation was to evaluate the possibility of observing the transformation of plastic strain-induced austenite to fresh martensite. Considering the study of Dimatteo [24], some materials present an inflection point in the strain hardening trend, thus an experimental data regression with a fourth grade polynomial was made to eventually appreciate this feature (Equation (3)).
The best fitting parameters and instantaneous strain hardening coefficient (Equation (4)) were evaluated for all materials.

Permeation Tests
Effective diffusion coefficients were calculated according to the time lag method (see Section 2.2) and the results are reported in Table 3. For completeness, permeation flux curves are plotted in Figure 6.

Slow Strain Rate Tests
The slow strain rate test results are reported in Table 4, two specimens were tested in as-received conditions with different strain rates.

Slow Strain Rate Tests
The slow strain rate test results are reported in Table 4, two specimens were tested in as-received conditions with different strain rates. In Figure 7, embrittlement indexes determined on the samples' elongation at break, as described in Equation (2), were correlated to the hydrogen charging method to compare the behaviour of groups I, X and O.

Slow Strain Rate Tests
The slow strain rate test results are reported in Table 4, two specimens were tested in as-received conditions with different strain rates.
In Figure 7, embrittlement indexes determined on the samples' elongation at break, as described in Equation (2), were correlated to the hydrogen charging method to compare the behaviour of groups I, X and O.

SEM Fractographic Analysis
Considering the test conditions described in dedicated standard references, with the sample immersed in NACE A solution, fracture surfaces were observed by a scanning electron microscope to evaluate the main fracture mode. Images related to all three material groups are reported in  to evaluate the main fracture mode. Images related to all three material groups are reported in Figures  8-10.
Samples of group I showed intergranular brittle fractures with some secondary cracks and some transgranular areas. For groups X and O, larger transgranular quasi-cleavage areas were found. The fracture mode therefore appears to be correlated with the amount of retained austenite and quasicleavage was dominant for higher austenite amounts [2,25].  to evaluate the main fracture mode. Images related to all three material groups are reported in Figures  8-10. Samples of group I showed intergranular brittle fractures with some secondary cracks and some transgranular areas. For groups X and O, larger transgranular quasi-cleavage areas were found. The fracture mode therefore appears to be correlated with the amount of retained austenite and quasicleavage was dominant for higher austenite amounts [2,25].

Instantaneous Strain Hardening Coefficient
Polynomial regression best-fitting parameters are reported in Table 5.  Samples of group I showed intergranular brittle fractures with some secondary cracks and some transgranular areas. For groups X and O, larger transgranular quasi-cleavage areas were found. The fracture mode therefore appears to be correlated with the amount of retained austenite and quasi-cleavage was dominant for higher austenite amounts [2,25].

Instantaneous Strain Hardening Coefficient
Polynomial regression best-fitting parameters are reported in Table 5.

Discussion
The experimental work was done in order to differentiate and classify the microstructure obtained with various tempering temperatures in terms of stress corrosion cracking and hydrogen embrittlement susceptibility.
The final microstructure of supermartensitic stainless steel 16Cr5NiMo was biphasic (martensite-austenite). Previous studies by means of TEM (Transmission Electron Microscope) showed the austenite fraction as elongated interlath films in a martensite matrix (lattice constant close to 1 as a body-centered cubic because of the low carbon content) [16].
Permeation tests were carried out at room temperature using a patented new technology and method [26]. The equipment was based on a solid-state gas sensor and was able to perform fast permeation of sheet metal samples, evaluating the effective hydrogen diffusion coefficient [27,28].
The permeation test time, and consequently the diffusivity of metal, is a function of temperature and microstructure. For example, it is lower in austenitic stainless steels (FCC) compared to ferritic stainless steels (BCC): it is D FCC ≈ 10 −16 m 2 /s and D BCC ≈ 10 −12 m 2 /s, respectively [29,30]. Thus, an austenite lattice in a martensitic matrix can be considered a hydrogen trap.
The microstructures of the steels were different in terms of retained austenite amounts, and a variation in hydrogen diffusivity was found [31]. Consequently, confirming the other literature studies, the hydrogen diffusion coefficient has a certain proportionality with the austenitic phase content. The martensite tempering temperature and eventual carbides precipitation may have also had an effect. A good exponential correlation was found including the data of the supermartensitic stainless steel under investigation and generic ferritic and austenitic stainless steel values ( Figure 11). The hydrogen diffusivity and plastic behavior of the steels under investigation were clearly influenced by the presence of a certain amount of retained austenite. The higher the fraction of the austenitic phase, the lower the effective hydrogen diffusivity of the steel. Group O, with a second tempering temperature of 670 °C and containing the maximum quantity of retained austenite (according to [16]), showed the lowest diffusivity [30].
Considering that in the real operative life of typical components realized in 16Cr5NiMo, the The hydrogen diffusivity and plastic behavior of the steels under investigation were clearly influenced by the presence of a certain amount of retained austenite. The higher the fraction of the austenitic phase, the lower the effective hydrogen diffusivity of the steel. Group O, with a second tempering temperature of 670 • C and containing the maximum quantity of retained austenite (according to [16]), showed the lowest diffusivity [30].
Considering that in the real operative life of typical components realized in 16Cr5NiMo, the yield strength of metal is usually never exceeded and no plastic strain-induced transformation into martensitic microstructure is expected, retained austenite acts as a trap for hydrogen. In fact, the hydrogen accumulation in the austenitic phase prevents it from moving towards the crack tip, with a consequent increase in hydrogen embrittlement resistance [32,33]. A different evolution is possible for the parts which work in a plastic strain field.
It was also evident that steel's susceptibility to hydrogen embrittlement was more influenced by the specific hydrogen charging method than the amount of retained austenite, especially because a big number of precipitates were found at the grain boundaries. Chromium carbides, M 23 C 6 type are typical, as reported in [16], where Alloy A corresponds to the same material as in the present work.
Finally, considering the instantaneous strain hardening coefficient, a correlation between austenite presence and plastic behavior is evident, but its role should be investigated more deeply.
The true stress-true strain curve was calculated from tensile test machine data. A plastic deformation range, up to necking initiation, was selected and reported in a bi-logarithmic diagram. A polynomial regression curve of group X is reported in Figure 12a as an example. Figure 12b presents instead the comparison between instantaneous strain hardening coefficients calculated by a derivative of the polynomial (orange) and a finite difference method on experimental data (blue) of the same steel. The expected trend is very close to the calculated curve. The hydrogen diffusivity and plastic behavior of the steels under investigation were clearly influenced by the presence of a certain amount of retained austenite. The higher the fraction of the austenitic phase, the lower the effective hydrogen diffusivity of the steel. Group O, with a second tempering temperature of 670 °C and containing the maximum quantity of retained austenite (according to [16]), showed the lowest diffusivity [30].
Considering that in the real operative life of typical components realized in 16Cr5NiMo, the yield strength of metal is usually never exceeded and no plastic strain-induced transformation into martensitic microstructure is expected, retained austenite acts as a trap for hydrogen. In fact, the hydrogen accumulation in the austenitic phase prevents it from moving towards the crack tip, with a consequent increase in hydrogen embrittlement resistance [32,33]. A different evolution is possible for the parts which work in a plastic strain field.
It was also evident that steel's susceptibility to hydrogen embrittlement was more influenced by the specific hydrogen charging method than the amount of retained austenite, especially because a big number of precipitates were found at the grain boundaries. Chromium carbides, M23C6 type are typical, as reported in [16], where Alloy A corresponds to the same material as in the present work.
Finally, considering the instantaneous strain hardening coefficient, a correlation between austenite presence and plastic behavior is evident, but its role should be investigated more deeply.
The true stress-true strain curve was calculated from tensile test machine data. A plastic deformation range, up to necking initiation, was selected and reported in a bi-logarithmic diagram. A polynomial regression curve of group X is reported in Figure 12a as an example. Figure 12b presents instead the comparison between instantaneous strain hardening coefficients calculated by a derivative of the polynomial (orange) and a finite difference method on experimental data (blue) of the same steel. The expected trend is very close to the calculated curve.
(a) (b) Figure 12. Example of (a) true stress-true strain plastic range up to necking initiation in a bi-logarithmic diagram and regression with fourth degree polynomial, (b) instantaneous strain hardening coefficient comparison between experimental data and polynomial regression on group X.
In Figure 13, the comparison between the as-received material and hydrogen charged samples was reported for each material group.
The first results showed that a major retained austenite content (as for materials X and O), makes a negligible hydrogen absorption influence on instantaneous strain hardening. The instantaneous strain hardening curve is probably mainly correlated to and dependent on austenite transformation into fresh martensite induced by plastic straining.
To summarize, the principal causes were both large grain size and precipitates at grain boundaries, and thus were mainly correlated to high temperature processes rather than tempering treatment variations. In fact, in all specimens, hydrogen mostly diffused through grain boundaries, showing intergranular brittle surfaces with numerous secondary cracks. All three examined variants of 16Cr5NiMo steel, independently from tempering treatment, resulted in susceptibility to stress corrosion cracking both in the case of SSR testing after hydrogen pre-charging (internal hydrogen embrittlement) and SSR testing of immersion in NACE solution (external hydrogen charging). Few differences were observed between as-received and charged materials; the major effect was the embrittlement behavior of the BCC matrix. The retained austenite effect was limited; this is correlated to the low values of the retained austenite percentages in the microstructure; between the three steels, there were small differences.
Materials 2020, 13, x FOR PEER REVIEW 11 of 14 Figure 12. Example of (a) true stress-true strain plastic range up to necking initiation in a bilogarithmic diagram and regression with fourth degree polynomial, (b) instantaneous strain hardening coefficient comparison between experimental data and polynomial regression on group X.
In Figure 13, the comparison between the as-received material and hydrogen charged samples was reported for each material group. The first results showed that a major retained austenite content (as for materials X and O), makes a negligible hydrogen absorption influence on instantaneous strain hardening. The instantaneous strain hardening curve is probably mainly correlated to and dependent on austenite transformation into fresh martensite induced by plastic straining.
To summarize, the principal causes were both large grain size and precipitates at grain boundaries, and thus were mainly correlated to high temperature processes rather than tempering treatment variations. In fact, in all specimens, hydrogen mostly diffused through grain boundaries, showing intergranular brittle surfaces with numerous secondary cracks. All three examined variants of 16Cr5NiMo steel, independently from tempering treatment, resulted in susceptibility to stress corrosion cracking both in the case of SSR testing after hydrogen pre-charging (internal hydrogen embrittlement) and SSR testing of immersion in NACE solution (external hydrogen charging). Few differences were observed between as-received and charged materials; the major effect was the embrittlement behavior of the BCC matrix. The retained austenite effect was limited; this is correlated to the low values of the retained austenite percentages in the microstructure; between the three steels, there were small differences.

Conclusions
The results obtained in the present work prove that different tempering processes cause the formation of different amounts of retained austenite in 16Cr5NiMo steels, which give rise to complex phenomena:

Conclusions
The results obtained in the present work prove that different tempering processes cause the formation of different amounts of retained austenite in 16Cr5NiMo steels, which give rise to complex phenomena: -Retained austenite has a role in the effective hydrogen diffusion coefficient, together with other significant parameters such as the martensite tempering process and secondary phases precipitations (carbides); -Retained austenite transformation due to plastic strain flow, which can be appreciated by means of mechanical tests; - The embrittlement effect, at least for the investigated conditions, appears to be not strictly correlated with the retained austenite amounts, but mainly to large dimensions of prior austenitic grain size.
The investigation techniques used in this study allowed for very long permeation times, and thus they are useful for studying the behaviour of supermartensitic stainless steels. However, more microstructural and hydrogen embrittlement susceptibility investigations are needed, and it would be desirable to test materials with a finer grain size, where the damage effect could be correlated to the presence of retained austenite amounts.