Effect of Microstructure on Hydrogen Permeation in EA4T and 30CrNiMoV12 Railway Axle Steels

: A comparative study was conducted to reveal the effect of microstructure on hydrogen permeation in the EA4T and 30CrNiMoV12 railway axle steels. Unlike the EA4T with its sorbite structure, 30CrNiMoV12 steel shows a typical tempered martensitic structure, in which a large number of ﬁne, short, rod-like, and spherical carbides are uniformly dispersed at boundaries and inside laths. More importantly, this structure possesses plentifully strong hydrogen traps, such as nanosized Cr 7 C 3 , Mo 2 C, VC, and V 4 C 3 , thus resulting in a high density of trapping sites ( N = 1.17 × 10 22 cm − 3 ). The hydrogen permeation experiments further demonstrated that, compared to EA4T, the 30CrNiMoV12 steel not only delivered minimally effective hydrogen diffusivity but also had a high hydrogen concentration. The activation energy for hydrogen diffusion of the 30CrNiMoV12 steel was greatly increased from 23.27 ± 1.94 of EA4T to


Introduction
Railway axles are among the critical components of high-speed trains, which work under cyclic rotating and bending conditions [1,2]. An axle with high strength, toughness, and fatigue life is therefore always desired for the safe use of high-speed trains [3][4][5]. As we know, hydrogen is a ubiquitous element that can enter steels from different processes, such as smelting, acid pickling, heat working and corrosion [6][7][8][9]. Unfortunately, hydrogen in an axle would cause hydrogen-assisted stress, corrosion cracking and hydrogen embrittlement (HE) [10,11], which gives rise to a huge risk for the safe operation of high-speed railway systems. Thus, further exploration of hydrogen mobility in steels is imperative for advancing the railway axles.
Achieving H-traps in steel usually decreases hydrogen diffusivity and increases critical hydrogen concentration, which is considered to be an effective method for enhancing resistance to HE [12,13]. Vacancies, dislocations, interfaces, microvoids, inclusions, and carbides are considered to be possible trapping sites, which can be classified as reversible and irreversible according to their binding energy for hydrogen atoms [14][15][16]. Out of these, a material with finely distributed irreversible traps is less susceptible to HE because these traps reduce hydrogen accumulation at potential crack initiation sites [14,17]. Previous studies have further indicated that carbides are often identified as strong irreversible traps to decrease hydrogen diffusivity and increase critical hydrogen concentration in steel [18][19][20][21], and thus the carbides of Ti, Mo, Nb, and V such as TiC, MoC, NbC, and VC are considered to be "good traps" [18,19,[22][23][24][25]. Similar to other Cr-and Mo-rich carbides [21,26], Mo-rich M 2 C carbides were also identified as beneficial traps [23,27,28]. In particular, the fine and coherent carbide precipitates in tempered martensitic steels would significantly improve the hydrogen permeation behaviors. For example, the small and coherent TiC and V 4 C 3 with the size of about 10 to 30 nm presented the best trapping efficiency [29,30], whereas the coarser Mo 2 C and Cr 23 C 6 only trapped a Metals 2019, 9, 164 2 of 10 limited amount of hydrogen [31,32]. These results strongly indicate that the ability to reasonably control the category, morphology, size, and distribution characteristics of carbides is critical for improving the hydrogen permeation behaviors of the tempered steels.
With the need for faster train running speed, some high-strength Cr-Mo and Cr-Ni-Mo alloy steels such as 25CrMo4 and 34CrNiMo6 were designed for high-speed axles [1,33]. Most recently, we developed a new quenched and tempered (Q&T) Cr-Ni-Mo-V (30CrNiMoV12) high-speed axle. The tensile strength of the 30CrNiMoV12 steel was remarkably increased to 1000 MPa from 650 MPa of the standardized European EA4T, although the toughness was no less than that of EA4T. This increase in strength is suggested to be related to the HE susceptibility, which is also related to the fatigue life compared to a controlled environment [34,35]. Unfortunately, few works on the fatigue life by Wöhler curve (S-N curve) measurements have been done to support this so far, to the best of our knowledge. Therefore, the purpose of the present work is to investigate the hydrogen permeation behaviors of the Q&T 30CrNiMoV12 axle by combining an electrochemical method with transmission microscopy. Moreover, the EA4T steel was also studied in parallel to reveal the effect of the microstructure (H-traps) on hydrogen permeation.

Experimental Details
Standard EA4T and 30CrNiMoV12 railway axle steels were produced by Maanshan Iron and Steel Co., Ltd. of China (Maanshan, China), and their chemical compositions are listed in Table 1. The vacuum-melted ingot was forged into a round bar (Φ200 mm × 2300 mm), approximate to a full-size solid axle. Then, the bars were normalized twice at 900 • C for 4 h with the aim of producing a fine and homogeneous microstructure. Finally, the bars were austenitized at 880 • C for 2 h, followed by oil cooling to produce a quenched martensite microstructure, and finally tempered at 600 • C for 3 h. Samples for both hydrogen permeation experiments and microstructure examinations were cut from the external surface of the Q&T bars with a wire cutting technique. Hydrogen permeation experiments were carried out using an electrochemical method developed by Devanathan and Stachurski [36]. For permeation experiments, the disc-shaped samples were electrolytically polished in a solution of perchloric acid, washed several times in distillated water, and degreased with acetone. Both sides of the samples were electroplated with Ni. Each sample was used as a membrane separating an electrochemical cell into two halves. During hydrogen charging, the hydrogen entry side (cathode) was galvanostatically polarized at a constant density of 10 mA/cm −2 . During discharging, the cathodic current was terminated. During the permeation experiment, the hydrogen exit side (anodic) was held at a constant potential of 0.20 V vs. saturated calomel electrode (SCE), and the resulting anodic current was recorded using a Solartron SI 1287 electrochemical interface. Both cathodic and anodic solutions were 0.2 N NaOH, which were deaerated with gas argon to reduce the background current. The effective hydrogen diffusivity (D eff ) can be calculated from each hydrogen permeation curve using the time lag method [36,37]: where L is the sample thickness and t L is the lag time. The lag time can be obtained by spotting the time at which the permeation rate is 0.63 times the steady-state value [36,37]. The microstructure observation specimens after electro-polishing etching were examined by scanning electron microscopy (SEM) using a (FEI, Hillsboro, OR, USA) at an accelerating voltage of 20 kV. To detect the carbides, the specimens were mechanically ground to 30 µm and chemically thinned in a dual submerged jet polisher, using a solution of 450 mL CH 3 COOH and 50 mL perchloric acid electrolyte. The thin foils were further examined in a JEM-2010 high resolution transmission electron microscope (HRTEM, JEOL, Tokyo, Japan) operated at 200 kV. Selected area electron diffraction (SAD, JEOL, Tokyo, Japan) patterns combined with energy-dispersive X-ray spectrometry (EDX, JEOL, Tokyo, Japan) analysis were used to identify the carbides.

Microstructural Characterization
SEM micrographs of the Q&T EA4T and 30CrNiMoV12 samples are shown in Figure 1. It is clear that the tempered sorbite was achieved in the EA4T steel (see Figure 1a), consisting of recrystallized ferrite (equiaxed α-Fe grain), stripped carbides, and spherical carbides (see Figure 1b). More interestingly, the 30CrNiMoV12 steel has a typical tempered martensitic structure (see Figure 1c), where these grains are divided into packets that are groups of parallel laths with the same habit plane. The mean width of lath is about 0.5 µm, as measured by using the linear intercept method on an SEM micrograph. Moreover, a large number of fine, short, rod-like, and spherical carbides are uniformly distributed at lath boundaries as well as within laths, which is consistent with the reported results for the tempered Cr-Ni-Mo steels in a range of 300 to 650 • C [18,21,38]. During tempering, the matrix was depleted in solute carbon, i.e., the ratio c/a of the body-centered tetragonal (bct) martensitic structure decreased and the martensite transformed gradually to a bct or body-centered cubic (bcc) ferrite, which is largely determined by the chemical compositions of steel. A comparison of the microstructures shown in Figure 1 clearly indicates that the additions of Ni, Mo, and V greatly enhanced the tempering resistance of the 30CrNiMoV12 steel. On the basis of the aforementioned results, it can be reasonably inferred that the different hydrogen permeation behaviors should be achieved for the 30CrNiMoV12 sample due to its altered microstructure.  Figure 2a,b shows the hydrogen permeation curves for the EA4T and 30CrNiMoV12 samples at various temperatures. It can be seen that hydrogen diffusion is accelerated with the increase of temperature. At high hydrogen charge temperature, thus the elevated steady-state permeation current (I∞) is observed in Figure 2, which corresponds to the steady-state hydrogen permeation flux  Figure 2a,b shows the hydrogen permeation curves for the EA4T and 30CrNiMoV12 samples at various temperatures. It can be seen that hydrogen diffusion is accelerated with the increase of temperature. At high hydrogen charge temperature, thus the elevated steady-state permeation current (I ∞ ) is observed in Figure 2, which corresponds to the steady-state hydrogen permeation flux (J ∞ ) through the cathodic side of the specimen. Compared with EA4T, the hydrogen diffusion in the 30CrNiMoV12 sample is remarkably slow at the same measurement temperature. According to Fick's second law, the more diffusion time (t), the greater cathodic side flux. Therefore, the permeation current of the 30CrNiMoV12 sample is greatly higher than that of EA4T at the same temperature. The lag times (t L ) and the effective hydrogen diffusivities (D eff ) at various temperatures are listed in Table 2. For EA4T and 30CrMoNiV12, the calculated values of D eff at 298 K using Equation (1) are 1.81 × 10 −6 and 1.98 × 10 −7 cm 2 s −1 , respectively. Significantly, the effective hydrogen diffusivity of the 30CrNiMoV12 sample is approximately nine times lower than that of the EA4T sample. This low D eff could be related to the distinct microstructure characteristic as shown in Figure 1.    The permeation results also suggest that the 30CrNiMoV12 sample has a larger hydrogen concentration and higher activation energy (Q) for hydrogen diffusion. Furthermore, the concentration of the subsurface (C 0 ) corresponds to the concentration of hydrogen at the cathode side, which can be estimated by the following expression [39]:

Hydrogen Permeation Behavior
where i ∞ is the steady-state permeation current density and F is the Faraday constant. According to this expression, the values of C 0 in the EA4T and 30CrMoNiV12 samples were confirmed to be 2.62 × 10 −6 and 9.05 × 10 −5 mol cm −3 at 298 K, respectively. The value of C 0 in 30CrMoNiV12 is close to the reported value in the high strength 3Ni-Cr-Mo-V steels [21]. This means that the significant decrease in mobility of hydrogen, i.e., the high activation energy, can be achieved in this steel. To calculate the value of Q, the Arrhenius plots were drawn in Figure 2c. Accordingly, the values of Q for the EA4T and 30CrMoNiV12 samples were obtained to be 23.27 ± 1.94 and 47.82 ± 2.14 kJ mol −1 , respectively. This result strongly suggests that a vast quantity of sites act as efficient H-traps in the 30CrMoNiV12 sample. Hydrogen Enhanced Localized Plasticity (HELP) is considered to be the dominant HE mechanism taking place in non-hydride forming steels [40,41]. HELP proposes an increase in dislocation mobility in the presence of H, which results in highly localized plastic deformation and faster failure [40]. Significantly, some nanosized alloy carbides are often mentioned to act as irreversible traps to relieve the interaction of dislocation with hydrogen [41,42]. Thus, the carbide characteristics are further investigated in the following section.

Density and Nature of H-Traps
There is no doubt that the hydrogen permeation behaviors are highly dependent on the density and nature of traps. A recently developed mathematical model allows the density of trapping sites (N) to be determined through the direct application of the values of D eff and C 0 as Equation (3) [43].
where N is the density of trapping sites per unit volume, D L is the lattice diffusivity of hydrogen in α-Fe = 1.28 × 10 −4 cm 2 s −1 and N A is the Avogadro constant [39,44]. According to the model, the calculated value of N of the 30CrNiMoV12 sample is 1.17 × 10 22 cm −3 at 298 K, which is approximately 320 times higher than 3.67 × 10 19 cm −3 of the EA4T sample. Figure 3a shows a typical low-magnification HRTEM image of carbides in the EA4T sample. Most of the precipitates are the stripped carbides with a size of 300 to 500 nm, which is consistent with the morphology of SEM. Figure 3b presents the SAD pattern of the stripped carbide marked by a red arrow in Figure 3a, which reveals a crystal structure of Fe 3 C (cementite). The Cr-and Mo-carbides were not found by TEM observation, which implies that the atoms of Cr/Mo are dissolved in matrix and/or cementite. Figure 3c shows that the spherical cementites with a diameter of about 120 nm (marked by a blue arrow) and some of the very small particles (marked by a red circle) are uniformly distributed in the lath matrix of the 30CrNiMoV12 sample. By comparing microstructures, the high density of trapping sites in the 30CrNiMoV12 sample can be explained as follows: (1) Aside from grain boundaries, a large number of lath and packet boundaries can provide more interfaces to serve as trapping sites. (2) The microstructure feature of the 30CrNiMoV12 sample clearly indicates that tempering does not completely restore the quenched microstructure. Thus, the large quantity of trapping sites can be related to the residual intra-lath dislocations. (3) The carbides are very small, the total carbide/matrix interfacial area is large and, therefore, the trapping ability reaches its maximum. Usually the reversible traps with low bonding energies include dislocations, boundaries, and substitutional elements, etc. However, the binding energies in the range of 10 to 50 kJ mol −1 H were observed in cementite, corresponding to the weak-to-strong interaction of hydrogen with the traps [45]. On the other hand, irreversible traps with high binding energies are those trapping hydrogen permanently at temperatures close to ambient temperature [46], which can not only reduce hydrogen diffusivity, but also markedly increase hydrogen concentration in steel [13].  Figure 4 displays the high-magnification HRTEM images of the small particles marked by the red circle in Figure 3c. As shown in Figure 4, HRTEM/SAD analyses found that the particles of Cr7C3 (70-80 nm) and Mo2C (30-40 nm) precipitated at lath boundaries and in lath, respectively. A binding energy of 55 kJ mol −1 H was reported for M7C3 by Symons et al. [47] For Cr7C3 particles at lath boundary, H atoms are not trapped solely at the interface, but also in the carbon vacancies within Cr7C3. Thus, the interactions of Cr7C3 and lath boundaries can provide deeper trapping sites. Recently, Liu et al. [21] found that the Mo2C particles were coherent with the matrix from nucleation for the tempered 3Ni-Cr-Mo-V steels, and that the trapped hydrogen was not easily diffusible at ambient temperature. Mendibide and Sourmail [26] concluded that the coarse M2C and M6C carbides are formed in steel when the Mo-content is higher than the critical value of 0.75 wt%, which is supposed to be harmful with respect to HE. Similarly, the beneficial Mo2C traps (30-40 nm) were achieved as expected in the 30CrNiMoV12 steel with the Mo-addition of 0.58 wt%. If the particles remain smaller, then their interfacial area is higher and the trapping effect is stronger [20]. For the 30CrNiMoV12 sample, significantly, further smaller carbide particles marked  Figure 4 displays the high-magnification HRTEM images of the small particles marked by the red circle in Figure 3c. As shown in Figure 4, HRTEM/SAD analyses found that the particles of Cr 7 C 3 (70-80 nm) and Mo 2 C (30-40 nm) precipitated at lath boundaries and in lath, respectively. A binding energy of 55 kJ mol −1 H was reported for M 7 C 3 by Symons et al. [47] For Cr 7 C 3 particles at lath boundary, H atoms are not trapped solely at the interface, but also in the carbon vacancies within Cr 7 C 3 . Thus, the interactions of Cr 7 C 3 and lath boundaries can provide deeper trapping sites. Recently, Liu et al. [21] found that the Mo 2 C particles were coherent with the matrix from nucleation for the tempered 3Ni-Cr-Mo-V steels, and that the trapped hydrogen was not easily diffusible at ambient temperature. Mendibide and Sourmail [26] concluded that the coarse M 2 C and M 6 C carbides are formed in steel when the Mo-content is higher than the critical value of 0.75 wt%, which is supposed to be harmful with respect to HE. Similarly, the beneficial Mo 2 C traps (30-40 nm) were achieved as expected in the 30CrNiMoV12 steel with the Mo-addition of 0.58 wt%.
If the particles remain smaller, then their interfacial area is higher and the trapping effect is stronger [20]. For the 30CrNiMoV12 sample, significantly, further smaller carbide particles marked by the red arrows were found by the HRTEM measurement as shown in Figure 5a. Figure 5b,c exhibits the HRTEM images of the particles marked as 1 and 2 in Figure 5a  . Thus, it is believed that these "good" carbides were indeed precipitated in the tempered 30CrNiMo12 steel. V-carbides have been proven to be strong irreversible traps by many researchers [25,30,48]. Recently, Van den Eeckout et al. [48] discussed the effect of W 2 C, Cr 23 C 6 , TiC, and V 4 C 3 on hydrogen diffusivity. Owing to the strongest trapping ability of V 4 C 3 , the Fe-C-V alloy exhibited the lowest diffusivity. Moreover, the higher binding energy (82.3 kJ mol −1 H) was observed in nano-VC [49]. Obviously, the achieved nanosized alloy carbides as irreversible traps, such as Cr 7 C 3 , Mo 2 C, VC, and V 4 C 3 , play a crucial role in improving the hydrogen permeation behaviors of the 30CrNiMo12 steel. In this work, our results fully prove that the composition and microstructure optimizations can provide effective routes to reducing HE susceptibility for high-strength steels. Figure 4 displays the high-magnification HRTEM images of the small particles marked by the red circle in Figure 3c. As shown in Figure 4, HRTEM/SAD analyses found that the particles of Cr7C3 (70-80 nm) and Mo2C (30-40 nm) precipitated at lath boundaries and in lath, respectively. A binding energy of 55 kJ mol −1 H was reported for M7C3 by Symons et al. [47] For Cr7C3 particles at lath boundary, H atoms are not trapped solely at the interface, but also in the carbon vacancies within Cr7C3. Thus, the interactions of Cr7C3 and lath boundaries can provide deeper trapping sites. Recently, Liu et al. [21] found that the Mo2C particles were coherent with the matrix from nucleation for the tempered 3Ni-Cr-Mo-V steels, and that the trapped hydrogen was not easily diffusible at ambient temperature. Mendibide and Sourmail [26] concluded that the coarse M2C and M6C carbides are formed in steel when the Mo-content is higher than the critical value of 0.75 wt%, which is supposed to be harmful with respect to HE. Similarly, the beneficial Mo2C traps (30-40 nm) were achieved as expected in the 30CrNiMoV12 steel with the Mo-addition of 0.58 wt%. If the particles remain smaller, then their interfacial area is higher and the trapping effect is stronger [20]. For the 30CrNiMoV12 sample, significantly, further smaller carbide particles marked by the red arrows were found by the HRTEM measurement as shown in Figure 5a. Figure 5b,c exhibits the HRTEM images of the particles marked as 1 and 2 in Figure 5a, respectively, showing their irregular spherical shape with the size of about 20-30 nm. Correspondingly, Figure 5d,e presents the inverse fast Fourier transform images of the selected area in particles 1 and 2, respectively. It was found that the values of d200 and d107 were 0.2084 and 0.2130 nm, respectively, which are close to the reported values d200 = 0.2083 nm for VC (JCPDS card no. 65-7885) and d107 = 0.2132 nm for V4C3 (JCPDS card no. 24-1390). Thus, it is believed that these "good" carbides were indeed precipitated in the tempered 30CrNiMo12 steel. V-carbides have been proven to be strong irreversible traps by many researchers [25,30,48]. Recently, Van den Eeckout et al. [48] discussed the effect of W2C, Cr23C6, TiC, and V4C3 on hydrogen diffusivity. Owing to the strongest trapping ability of V4C3, the Fe-C-V alloy exhibited the lowest diffusivity. Moreover, the higher binding energy (82.3 kJ mol −1 H) was observed in nano-VC [49]. Obviously, the achieved nanosized alloy carbides as irreversible traps, such as Cr7C3, Mo2C, VC, and V4C3, play a crucial role in improving the hydrogen permeation behaviors of the 30CrNiMo12 steel. In this work, our results fully prove that the composition and microstructure optimizations can provide effective routes to reducing HE susceptibility for high-strength steels.

Conclusions
The microstructures and hydrogen permeation behaviors of the railway axle steels EA4T and 30CrNiMoV12 were comparatively studied in the present work. The principal conclusions were summarized as follows: (1) Unlike the EA4T steel with its tempered sorbite structure, the 30CrNiMoV12 steel has a

Conclusions
The microstructures and hydrogen permeation behaviors of the railway axle steels EA4T and 30CrNiMoV12 were comparatively studied in the present work. The principal conclusions were summarized as follows: (1) Unlike the EA4T steel with its tempered sorbite structure, the 30CrNiMoV12 steel has a typical tempered martensitic structure due to its high resistance to tempering. Moreover, a large number of fine, short, rod-like, and spherical carbides are uniformly distributed at lath boundaries and inside.
(2) The measured D eff of the 30CrNiMoV12 sample was approximately 1.98 × 10 −7 cm 2 s −1 , nine times lower than that of the EA4T sample. Moreover, the C 0 and Q values of the 30CrNiMoV12 were also greatly increased from 2.62 × 10 −6 to 9.05 × 10 −5 mol cm −3 at 298 K and from 23.27 ± 1.94 to 47.82 ± 2.14 kJ mol −1 , respectively.
(3) The hydrogen permeation behaviors of the railway steels strongly depend on the density and nature of H-traps. The high density of trapping sites (N = 1.17 × 10 22 cm −3 ) and a large number of irreversible traps, such as nanosized Cr 7 C 3 , Mo 2 C, VC, and V 4 C 3, are believed to be responsible for the low hydrogen diffusivity achieved in the 30CrNiMoV12 steel.