Hydrogen Diffusion and Trapping in Laser Additively Manufactured Ultra-High Strength AerMet100 Steel Open Access

Ultra-high strength steels (UHSS), with yield strengths of at least 1,350 MPa, are widely used for critical load-bearing components in the aerospace and other industries, including landing gears in aircrafts and rotor shafts of helicopters.¹  AerMet100^† (UNS K92580⁽¹⁾) steel (23Co14Ni11Cr3Mo) is a high Co-Ni secondary hardening high-alloyed UHSS. When slightly overaged at 482°C, tempered microstructures of UNS K92580 steel featuring fine dispersive rod-like M₂C carbides amongst highly dislocated Fe-Ni martensite plates and (∼3 nm to 8 nm) thin film-like retained/reverted austenite along martensite plate interfaces, resulting in a good combination of high yield strength (YS ≥ 1,750 MPa) and high fracture toughness (K_IC ≥ 115 MPa√m). In comparison with conventional low-alloyed UHSS (such as AISI 4340 [UNS G43400] and 300M [UNS K44220]), UNS K92580 steel has a much better hydrogen-assisted stress corrosion cracking resistance, as well as a higher fracture toughness.^2-4  Therefore, it has been utilized in highly stressed fasteners, arresting hooks, and the landing gear components of aerospace vehicles.^1,3,5 

Hydrogen embrittlement (HE) of UHSS is usually caused by a small amount of diffusible hydrogen when high strength levels are present,⁶  and has been the subject of keen interest for decades.^1,7-8  Under high stress conditions, the delayed brittle fracture of UHSS can be observed along with a decrease in both strength and ductility after the absorption of substantial atomic hydrogen in humid or marine environments.^6,9-10  In addition, the accumulation of atomic hydrogen in the fracture process zone beyond the crack tip can prompt hydrogen-related crack propagation of the steel at stresses far below the critical stress required when considering a hydrogen-free condition.¹¹  In general, the HE behavior of UHSS is simultaneously influenced by hydrogen trapping states, diffusible hydrogen concentration and its distribution, the effective hydrogen diffusion coefficient (D_eff), and the supply of hydrogen from the crack tip to a critical depth to reach the fracture process zone to trigger fracture.^4-6,8,12-16 

The hydrogen trapping states of UHSS are mainly related to microstructural features (such as precipitates, matrix phase interfaces, and prior-austenite grain boundaries) and crystalline defects (such as vacancies and dislocations), which control hydrogen uptake, diffusion and repartitioning, and the subsequent fracture modes of the steel.^{4-5,8-10,12,14-15}  Extensive reversible and irreversible H trapping involving at least three unique trap states (each categorized by a different detrapping activation energy, E_d) was previously identified in the tempered microstructure of wrought UNS K92580 steel.^5,14  The reversible hydrogen trap state with an E_d of 21.4 kJ/mol to 21.6 kJ/mol was attributed to the precipitation of M₂C carbides, while the irreversible hydrogen trap sites (with E_d values of 71.3 kJ/mol to 72.2 kJ/mol and 99.1 kJ/mol to 99.9 kJ/mol) were likely associated with martensite plate interfaces, prior-austenite grain boundaries, mixed dislocation cores, and undissolved carbides.¹⁴  Due to the substantial hydrogen trap sites in the microstructure, the D_eff of tempered UNS K92580 steel was relatively low, with a room temperature value of approximately 1 × 10⁻⁹ cm²/s to 8.96 × 10⁻⁹ cm²/s.^3,5,17-18  With an increase in diffusible hydrogen concentration, the value of D_eff increased, and the threshold stress intensity (K_th) of H-charged UNS K92580 steel decreased from a K_IC value of ∼130 MPa√m to ∼12 MPa√m.¹²  Furthermore, the fracture mode of the steel gradually changed from microvoid coalescence to brittle transgranular cracking with increasing diffusible hydrogen.⁵  Although a high density of reversible hydrogen trap sites has the beneficial effect of slowing down the stage-II crack propagation rate (da/dt_II) of hydrogen environment-assisted cracking (HEAC) of UNS K92580 steel through a decrease in D_eff, hydrogen trap sites can also act as a reservoir of mobile H for the internal hydrogen embrittlement of the steel, supporting H repartitioning to nearby martensite lath interfaces and cleavage planes in the region of high hydrostatic tensile stress in front of the crack tip. This enabled brittle transgranular cracking of the steel.^4,12 

Laser additive manufacturing (LAM) is an additive manufacturing technology that can be used for near-shape forming of large, complex, and dense UNS K92580 steel parts by melting and solidification of inert-gas-carried prealloyed powders in a layer-by-layer deposition mode.^19-21  Advantages of LAM include no required mold, reduced material waste and postdeposition machining, shorter production cycle, lower production cost, and unprecedented design flexibility.^22-23  Generally, as-deposited microstructures of LAM UNS K92580 steel are mainly composed of large-size columnar grains with interior unidirectional cellular-dendrite solidification structures, complex room-temperature bainite microstructures, and high contents of retained austenite, resulting in low strength and anisotropic ductility of the steel.^21,24  However, after proper post strengthening and toughening heat treatments, LAM UNS K92580 steel can achieve improved tensile mechanical properties, fracture toughness, and fatigue crack propagation resistance due to microstructure optimization (including grain refinement, reduction of retained austenite, dissolution of M₃C carbides, and dispersive precipitation of coherent M₂C carbides).^20-21,24-28  It has been noted that microstructure optimization caused by heat treatment can change the type, amount, and distribution of hydrogen trap sites, and thus can influence the hydrogen diffusion behavior and embrittlement susceptibility of the steel.^15,29-30  Although a recent review describes the effect of the additive manufacturing (AM) process on corrosion behavior,³¹  relatively little information is available on HE or HEAC of AM materials. In order to study the HE behavior and the corresponding delayed cracking mechanism of LAM UNS K92580 steel, it is necessary to first understand fundamental hydrogen-metal interactions, including microstructure-related hydrogen trap types and trapping parameters (detrapping activation energy, E_d, and trap density, N_TR) of the steel, as well as hydrogen diffusion behaviors.

In this paper, microstructural characteristics, thermal desorption spectroscopy and electrochemical hydrogen permeation behavior of as-deposited specimens (AD), and three types of heat-treated specimens (bainite microstructure [BM], tempered bainite and martensite microstructure [TBMM], and tempered martensite [TM]) of LAM UNS K92580 steel are examined. The detrapping activation energy E_d of different hydrogen trapping sites and the effective hydrogen diffusion coefficient D_eff of different microstructures are analyzed. Finally, hydrogen trapping states, D_eff and the diffusible hydrogen concentration (C_H,diff) of UNS K92580 steel in different metallurgical conditions are compared and discussed.

An AD UNS K92580 steel thick plate with the dimensions of 240 mm (length) × 45 mm (width) × 370 mm (height) was first built (Figure 1). Chemical compositions of the pre-alloyed powders (wt%) with diameters from 75 μm to 250 μm were as follows: 13.50 Co, 11.26 Ni, 3.00 Cr, 1.25 Mo, 0.22 C, 0.022 Nb, 0.022 Si, <0.005 Mn, 0.011 Al, 0.0007 S, <0.005 P, <0.005 Ti, 0.0041 O, 0.0009 N, with the balance of Fe. Detailed information on manufacturing process of LAM UNS K92580 steel can be found in literature.²¹  Furthermore, three specimen blanks with dimensions of 40 mm × 45 mm × 80 mm were cut from the middle-upper part to middle part of the as-deposited plate and subsequently heat-treated (Table 1). After heat treatments, three kinds of fine-grained microstructures were designed as follows: (1) BM; (2) TBMM; and (3) TM. The tensile mechanical properties of these microstructures are listed in Table 2. Small cubic bulks with dimensions 15 mm × 15 mm × 15 mm were then prepared for each microstructure. The XOZ planes of these specimens were ground, mechanically polished, etched using a mixture of 4% Nital and saturated picric acid water solution, and investigated using a Leika-DM 4000^† optical microscope (OM) and a JEOL JSM-6010LA^† scanning electron microscope (SEM). To study the boundary characteristics of each microstructure, some ground specimens were additionally electro-polished in a mixed solution of 4% perchloric acid and 96% ethanol, and subsequently examined by a JSM-7001F^† SEM with a Pegasus XM2^† electron backscatter diffraction (EBSD) with a resolution of 0.15 μm. The average sizes of prior-austenite grains and bainite/martensite packets were measured by a linear intercept method.³²  The retained/reverted austenite contents of specimens were identified using x-ray diffraction (XRD) with Cu K_α, scanning rates 0.5°/min, voltage 40 kV, and current of 200 mA, and further quantitatively analyzed with the structure refinement using the Rietveld method with GSAS-II^† software.³³  The Vickers-hardness of each specimen was measured using a FM800^† micro-hardness tester under a load of 500 gf with a load-dwell time of 15 s. Alloy carbides were investigated in thin foil specimens and carbon film replicas using a Tecnai G2 F30 S-TWIN^† field-emission high-resolution transmission electron microscope (TEM) operated at 300 kV and equipped with an EDAX^† energy-dispersive x-ray spectroscopy (EDS) system. Preparation methods for TEM specimens are described in the literature.²⁴ 

Electrochemical hydrogen permeation of LAM UNS K92580 steel specimens was tested in a double-cell Devanathan–Stachurski setup³⁴  associated with a CorrTest CS 2350^† dual-unit electrochemical workstation (Figure 2). Assuming that hydrogen permeation of LAM UNS K92580 steel with specimen thickness of at least 0.25 mm is diffusion controlled,^3,35  specimens with dimensions of 35 mm × 30 mm × 0.5 mm for hydrogen permeation tests were first cut parallel to the XOZ plane (Figure 1), ground using 1200 grit water abrasive paper, degreased in acetone, rinsed in ethanol, dried, and stored in a desiccator. Prior to each test, the specimen with thickness of 0.25 mm was clamped between the charging cell and the oxidation cell to expose a circular region of diameter 16 mm to the 0.1 M NaOH aqueous solution (pH∼12.7) on each side. In order to increase the efficiency of hydrogen absorption, 3 g/L of NH₄SCN was added into the solution of the charging cell as a hydrogen recombination poison (pH ∼12.5).³⁶  High-purity nitrogen gas (>99.999%) was continuously purged into both cells 2 h before and during the experimental test for deaeration. The bare specimen was shared by both electrochemical cells, with one surface serving as the anode in the oxidation cell and the parallel opposite surface as the cathode in the charging cell (Figure 2). Initially, in the oxidation cell, the thin specimen was potentiostatically maintained at +200 mV_Hg/HgO reference electrode to achieve a stable background oxidation current (<0.05 μA/cm²). Then, hydrogen was generated on the entrance (left) side of specimen by galvanostatic polarization at a cathodic current density of −10 mA/cm². Meanwhile, the anodic current at exit (right) side of thin specimen was recorded as a function of time from the onset of hydrogen generation at the entrance side. This current corresponds to the oxidation of H diffusing through the foil. When absorbed hydrogen atoms diffuse through the membrane, the increased anodic current in the oxidation cell gives a direct measure of rising hydrogen diffusion flux above background. After charging H for 48 h, the value of the cathodic current density was decreased to ∼0.5 mA/cm² for the subsequent 48 h. In this period, the anodic current began to decay. The permeation cycle of rising period and decay period for each specimen was repeated two times.

To analyze the hydrogen permeation behavior, the first step is to fit experimental results with a theoretical model by determining the appropriate model (initial and boundary conditions) and mathematical equations for calculation of effective diffusion rate D_eff.³⁷  Based on different initial and boundary conditions, it is established that the anodic current in the permeation rising stage follows Equations (1) and (2) for a fixed D_eff assumption at constant concentration (CC) model (C[0, t] = C₀, t > 0) or for a constant flux (CF) model (⁠⁠., t > 0) separately.³⁸ 

Prior to the thermal desorption spectroscopy (TDS) test, hydrogen capacity test, and diffusible hydrogen concentration test, the AD specimen and three types of heat-treated specimens with dimensions of 7 mm × 6.4 mm × 0.5 mm were first prepared using the same method as the hydrogen permeation specimens (with electrical contact enabled with an alligator clip), painted with water-resistant lacquer, and dried in a desiccator. Then, the specimens with exposed dimensions of 5 mm × 6.4 mm × 0.5 mm were immersed in a base solution of 0.1 M NaOH + 3 g/L NH₄SCN (pH∼12.5) and precharged with H using a galvanostatic method at a cathodic current density of −10 mA/cm² for 72 h. After precharging H, the water-resistant lacquer on the specimens was immediately stripped away, and the bare specimens were slightly ground with 1200 grit water abrasive paper, flushed using deionized water, ultrasonically bathed in acetone, and dried within 2 min.

The total H concentration (C_H,tot) of the precharging specimens was also determined using a LECO TCH600^† gas analyzer device. Prior to testing, the interval time for preparing specimens was less than 15 min, and the weight of each precharging H specimen was recorded using an analytic balance with an accuracy of 0. 1 mg. Then, the precharging H specimen was completely melted fusion in the heating chamber of the gas analyzer device with a power of 4,800 W. The released hydrogen was carried by the flow of argon gas through a CuO tube chamber and was oxidized into water, which was then detected by an infrared method. The amount of released hydrogen from the precharging H specimen was shown in the form of integral area, and finally quantified by comparison of the 1 g standard steel specimen with a known reference hydrogen concentration of 6 ppm. In order to evaluate hydrogen concentration in irreversible hydrogen traps of precharging H specimens, additional groups of precharging H specimens were first baked at 190°C for 12 h, then similarly examined by total H concentration tests. For each C_H,diff (or C_H,tot) test, at least two specimens were examined.

In order to determine the types, amounts, and detrapping activation energies (E_d) of hydrogen traps in LAM UNS K92580 steel, thermal desorption spectroscopy (TDS) tests of LAM UNS K92580 steel specimens after precharging H for 72 h were conducted in the temperature range of 50°C to 550°C in an integrated quadrupole mass spectrometer in a vacuum system and proportional-integral-derivative (PID)-controlled heating muffle furnace system. The quadrupole mass spectrometer in vacuum chamber was controlled using MOTF^† software for acquiring hydrogen partial pressure (P_H2) data. The background-corrected results of P_H2 released from precharging H specimens were obtained by subtracting the P_H2 results of the system without specimens from the P_H2 results of the system with the precharging H specimens, and then converting those P_H2 results into hydrogen desorption rate (dC/dt) using the same method utilized by Smith and Scully.⁴¹ 

The AD LAM UNS K92580 steel consisted of large-size (approximately 150 μm to 600 μm) prior-austenite columnar grains with interior epitaxial unidirectional growth cellular-dendrite solidification structure along the deposition direction (Figures 3[a] and [c]). There were several microstructural features present such as the grain boundary allotriomorphic ferrite (GBA), grain interior irregular proeutectoid ferrite, plate-like upper bainite, needle-like lower bainite, and a high content of (∼8.6 wt%) retained austenite (Figure 3[c] and Ran, et al.²¹ ). The retained austenite phases were observed as large-size blocky morphologies in inter-dendrite regions and small-size blocky morphologies in regions between bainite packets/blocks, and as film-like morphologies in regions between bainite plates (in Figures 4[a] and [c]). Furthermore, alloy carbides in the as-deposited specimens were fine, short rod-like Nb-rich MC carbides and needle-like M₃C carbides (Figure 5). The Vickers-hardness of AD microstructure was 518±6 HV.

After heat treatments, the large-size columnar grains transformed into fine (∼18 μm) equiaxed grains with elimination of the cellular-dendrite solidification structure and decrease of packet size (Figures 3[b] and [d] and Table 3). Grain-interior microstructures of the BM specimens (Figure 3[d]) include proeutectoid ferrite, plate-like upper bainite, needle-like lower bainite, and a high content (∼7.9 wt%) of retained austenite. The retained austenite was observed as small-size blocky morphologies in regions between bainite packets/blocks (Figure 4[b]) and as film-like morphologies in regions between bainite plates (Figure 4[c] and Ran, et al.²⁹ ). Besides spherical MC carbide and needle-like M₃C carbides, a few Mo-rich M₆C carbides and Cr-rich M₂₃C₆ carbides were observed in BM microstructure. It was noted that the amount of M₃C carbide in the BM microstructure was much less than that in the AD microstructure (Figures 5[c] and 6[c]). The Vickers-hardness of the BM microstructure was 537±8 HV. In comparison with the AD specimen and BM specimen, the TBMM specimen and TM specimen were dominated by tempered martensite plates (Figures 3[e] and [f]) and slightly inter-plate film-like retained/reverted austenite (Figure 4[d]). However, no apparent tempered bainite could be distinguished in the TBMM specimen (Figure 3[e]). In both the TBMM and the TM specimen, four types of alloy carbides (including fine spherical MC carbides, fine rod-like M₂C carbides, large-size spherical Mo-rich M₆C carbides, and Cr-rich M₂₃C₆ carbides) were observed (Figure 6). Although a few of the MC carbides with lower Nb and higher Cr were larger (∼32 nm to 51 nm), most of the MC carbides were quite fine (∼10 nm to 20 nm) and enriched with Nb and Mo.²⁴  Fine rod-like M₂C carbides (6 nm to 10 nm long, 1 nm to 2 nm in diameter, and spaced at least 6 nm to 8 nm apart), with a stoichiometry composition of (Cr_0.63Mo_0.27Fe_0.10)₂C, were present in a dispersed distribution and high coherency between these rods and the martensite matrix (Figures 6[d] and [e] and Ran, et al.²⁴ ). Furthermore, its composition is different from that reported in wrought UNS K92580 steel as (Cr_0.75Fe_0.13Mo_0.12)₂C. Furthermore, a semicoherent interface was at the ends of the rod-like M₂C carbides, where some mismatch between the matrix lattice and the carbide lattice with a few misfit dislocations could be observed (Ran, et al.²⁴  and Figure 6[e]). The amount and size of Mo-rich M₆C and Cr-rich M₂₃C₆ carbides in the TBMM specimen and the TM specimen were similar to those in the BM specimen.

EBSD results (Table 3 and Figure 7) suggest that the prior-austenite grain boundaries and the grain-interior matrix phase interfaces of the AD microstructure as well as different heat-treated microstructures are mainly high-angle boundaries, and that the misorientation angle is mainly distributed in the range of 50° to 65°. In comparison to the TM microstructure, the increased fraction of high-angle boundaries (∼60°) in the TBMM microstructure is related to an increased misorientation angle brought about due to the presence of tempered bainite. When a larger amount of blocky retained austenite is present in LAM UNS K92580 steel, the high-angle boundary percentage (especially the 45° misorientation angle) of the α/γ interface apparently increased. Besides the high-angle boundary, there is a high percentage of sub-grain boundaries due to a high density of dislocation and twinning substructures. In addition, compared to the lower Vickers hardness values and yield strengths of the AD specimen and the BM specimen, those of the TBMM specimen and the TM specimen are subsequently increased due to the precipitation of M₂C carbides and decreased retained/reverted austenite content (Tables 2 and 3).

Figure 8 mainly shows TDS results from lattice interstitial sites and reversible hydrogen traps with low E_d.¹⁴  With an increase in temperature, three apparent H desorption peaks (i.e., 1st Peak, 2nd Peak, and 3rd Peak) in the TDS result for each specimen could be gradually observed (Figure 8). Furthermore, most of the released atomic hydrogen was from hydrogen traps corresponding to the low-temperature 1st Peak. Compared to the AD specimen and the BM specimen, the TBMM and the TM specimens have much higher dC/dt peak values and higher peak width temperatures (∼50°C), indicating a higher reversible hydrogen concentration (C_H,re) in the steel (Figure 9). Moreover, the dC/dt peak values for the above four types of specimens are higher than that for as-quenched specimen (i.e., martensite specimen without tempering treatment). Therefore, an important effect of fine precipitates in matrix can be on the C_H,re of LAM UNS K92580 steel. In addition, in the temperature range of 300°C to 550°C, the dC/dt values of the heat-treated specimens are much higher than that of the AD specimen (Figure 8), indicating proper heat treatment can improve irreversible hydrogen trapping capacity of LAM UNS K92580 steel. TDS results for both the precharging H AD specimen as well as three types of heat-treated LAM UNS K92580 steels that were precharged and tested at a heating rate of 5°C/min. At the beginning of the heating period (T<100°C), a stage of hydrogen desorption (measured by hydrogen desorption rate dC/dt) can be observed (Figure 8[a]), which corresponds to the release of atomic hydrogen

Figure 10 and Table 4 further illustrate plots and the determined detrapping activation energy (E_d) of hydrogen traps in LAM UNS K92580 steel. E_d values of hydrogen traps in LAM UNS K92580 steel in different metallurgical conditions are either below 20 kJ/mol or above 80 kJ/mol. In pure iron and low-alloyed steel, dislocations, low-angle boundaries, and ferrite/cementite interfaces have been suggested as the main reversible trapping sites of hydrogen.^42,44  Comparison of the AD microstructure and the BM microstructure of LAM UNS K92580 steel shows that similar M₃C carbides are in bainitic ferrite. Therefore, the hydrogen trapping peaks with an E_d of 17.1 kJ/mol for the AD specimen and an E_d of 16.1 kJ/mol for the BM specimen are mainly related to the M₃C carbides in a bainitic matrix. For the TM specimen, the E_d value of dominantly reversible hydrogen trap is 18.0 kJ/mol, which is mainly related to M₂C carbides in a martensite matrix. Dislocations (or low-angle boundaries) are also likely have low trap binding energies. However, the trap number of dislocation sites (or low-angle boundary sites) is less than carbide sites and also of lower binding energy so they are difficult to detect in TDS results (Figure 8). In contrast, hydrogen traps with higher E_d values corresponding to the 2nd Peak and the 3rd Peak in LAM UNS K92580 steel are assigned to MC carbides, large-size spherical M₆C carbides and M₂₃C₆ carbides, prior-austenite grain boundary, as well as highly misoriented packet boundaries.^14,45-47  It is shown that the E_d values of hydrogen traps corresponding to the 2nd Peak and the 3rd Peak in BM specimen are the highest followed by those in AD specimen.

Figure 11 shows electrochemical hydrogen permeation behavior of LAM UNS K92580 steel. Breakthrough time (t_b), confirmed by linear extrapolation of initial rise transient to the baseline, is the time corresponding to the onset of rise in exit hydrogen oxidation current density. In general, a longer t_b indicates a higher density and intensity of hydrogen trap sites within the diffusion path. Comparison of a short t_b for the BM specimen with that for the AD specimen (Table 5) indicates that grain refinement does not have a significant influence on the t_b. The values of t_b for the TM specimen and the TBMM specimen are an order of magnitude higher than those for the AD specimen and the BM specimen. The t_b of the TM specimen is the greatest of all heat treatment types. These results could indicate that the hydrogen trapping states present in tempered martensite specimens are more influential on t_b than in the case of bainite microstructures.

Figure 12 shows a part of normalized experimental permeation curves and the corresponding theoretical curves (dash line) plotted by first-order approximation of Equations (1) through (3). Results indicate that the rise stages of the AD and BM specimens follow the initial and boundary conditions determined by the CF model (Figure 12[a]); in contrast, the rise stages of the TBMM specimen and the TM specimen follow the initial and boundary conditions determined by CC model (Figure 12[b]). After rearranging the first-order approximation results of Equations (4) through (6), the values of D_eff can be further calculated from the slope of the vs. t plots for the rise transient period and the slope of the vs. t plots corresponding to the initial transient (< 3 × 10⁵ s) of decay period.

The results of D_eff for LAM UNS K92580 steel are shown in Table 5. The values of D_eff for the TM specimen and the TBMM specimen are apparently lower than those for the AD specimen and the BM specimen. Moreover, the values of D_eff for TBMM specimen are very close to those for TM specimen because of a high amount of tempered martensite plates as well as the increased density of carbides after the 482°C treatment.

Table 6 reports the values of the C_H,tot and C_H,diff for different precharging H specimens of LAM UNS K92580 steel. In all of the precharging H specimens, the TM specimen has the highest value of the C_H,tot and C_H,diff, and the TBMM specimen has slightly smaller values. In contrast, the C_H,tot and C_H,diff of the AD specimen and the BM specimen are much smaller. These results indicate that specimens with tempered martensite microstructures have stronger hydrogen trapping capacities and higher diffusible hydrogen concentrations.

For LAM UNS K92580 steel, hydrogen trapping states are closely influenced by microstructural characteristics as in wrought materials. There are at least three kinds of hydrogen trap sites with different E_d values in the AD microstructure and three kinds of heat-treated microstructures of LAM UNS K92580 steel. The main reversible hydrogen trap site in the AD microstructure of the steel is the M₃C carbide/ferrite interfaces. It should be noted that M₃C carbides precipitating in large-size LAM UNS K92580 steel are mainly attributed to the slow cooling rates of the solidification work piece from an accumulated temperature by continuously using a laser heat input with high power (>6,000 W). Moreover, the site density of reversible hydrogen traps in the AD microstructure of LAM UNS K92580 steel is low (∼10¹⁹ sites/cm³) due to the low density of M₃C carbides forming in a short cooling period. After heat treatments, the reversible hydrogen trapping effects of LAM UNS K92580 steel can be greatly increased by precipitation of highly coherent dispersed M₂C carbides. Considering a similar medium intensity of the reversible hydrogen traps (Table 4), the significant difference in the reversible hydrogen trapping capacity of LAM UNS K92580 steel in different metallurgical conditions is mainly attributed to the difference in the amount of fine dispersive alloy carbides. In addition, according to the TDS results (Figure 8), hydrogen concentration in the reversible hydrogen traps should be more than 95% of the C_H,tot in LAM UNS K92580 steel. However, for precharged H in LAM UNS K92580 steel specimens with the tempered martensite microstructure, it is noted that C_H,diff is only about 70% of the C_H,tot. This phenomenon is also present in hydrogen precharged wrought UNS K92580 steel.⁵¹  These results suggest that one more kind of reversible hydrogen trap with a higher E_d value is present in the steel where some atomic hydrogen cannot be easily diffused at room temperature. For the TDS result of precharging H wrought AF1410 steel (a similar high Co-Ni secondary hardening UHSS), a similar situation exists. That is, besides the presence of some reversible hydrogen traps with an E_d of 20.2 kJ/mol corresponding to the dominant peak, there are other hydrogen trap states. Moreover, another reversible hydrogen trap with an E_d of 24.6 kJ/mol corresponding to a small peak at a higher temperature was observed as well.⁵²  After refitting TDS data, three hidden peaks with E_d values of 15.3 kJ/mol, 20.4 kJ/mol, and 26.3 kJ/mol corresponding to the 1st apparent peak can be obtained for the hydrogen precharged TM specimen of LAM UNS K92580 steel. It is clear that the tempering treatment prompts formation of both reverted austenite and M₂C carbides in LAM UNS K92580 steel,²⁶  associated with the additional presence of the stronger reversible hydrogen trap with an E_d of 26.3 kJ/mol. However, the increased presence of austenite boundaries is not associated with the aforementioned detected stronger reversible hydrogen trap, otherwise such a stronger reversible hydrogen trap would have been detected in the AD microstructure of the steel (Figure 8). In addition, it is well known that⁴⁶  the atomic hydrogen can be mainly trapped by strong attractive force from a tensile triaxial stress field (at solute atoms, coherent or semicoherent interfaces, front of crack tip, etc.), lattice defects (high-angle boundary, incoherent carbide/matrix interface, nonmetallic inclusion/matrix interface and pores) or their mixed type (dislocation) in structural materials.

Considering the large difference in E_d values between reversible (<20 kJ/mol) and irreversible hydrogen traps (>80 kJ/mol) in LAM UNS K92580 steel, hydrogen traps related to the M₂C carbides are attractive sites for hydrogen due to the tensile stress field around coherent interfaces or semicoherent interfaces associated with misfit dislocations (Figure 6[e]). Therefore, the highly coherent relationship between M₂C carbides and the martensite matrix is suggested to provide the formation of mostly reversible hydrogen traps. The semicoherent interfaces at the ends of short rod-like M₂C carbides associated with misfit dislocations may also prompt the formation of the stronger reversible hydrogen traps. Wei and Tsuzaki⁵³  agreed that the semicoherent interface of carbides associating with a high activation energy could trap hydrogen at the core of the misfit dislocation with cathodic charging. The corresponding work of characterizing hydrogen atoms accumulating in these interfaces is very difficult and could be further performed by a hydrogen isotope method in future investigations.^54-55 

It was shown that a microstructure consisting of a uniform distribution of fine and strong hydrogen traps could maximize an alloy’s resistance to hydrogen permeation/embrittlement as long as they do not create a susceptible crack path.⁵⁶  However, the requirement to meet this condition is to have very strong traps in a closed system. Another requirement is that the trap is strong enough to resist repartitioning of H to crack-tip enhanced stress fields.⁵⁷  That is also an issue in closed systems.² Generally, the hydrogen trapping propensity for an MC carbide is related to its size, stoichiometry, morphology, and the prevailing interface relation. After heat treatment, short rod-like Nb-rich MC carbides in LAM UNS K92580 steel change into spherical Nb-rich MC carbides with a bimodal size distribution (7 nm to 18 nm or 32 nm to 51 nm). Moreover, semicoherent or incoherent large-size MC carbides (such as TiC, VC, and NbC) can prompt the formation of irreversible hydrogen traps in low-alloy steels.^47,58-59  However, there was also a view that incoherent MC carbides could not trap hydrogen through cathodic charging.^53,60  Even though MC carbides have a strong hydrogen trapping capability, the hydrogen trapping benefit from MC carbides in LAM UNS K92580 steel is limited due to a low concentration (∼0.022 wt%) of Nb. In addition to MC and M₂C carbides, heat treatment also prompts the formation of a few M₂₃C₆ carbides and M₆C carbides in the steel, associated with presence of hydrogen traps with E_d values of 74.2 kJ/mol for the BM specimen and 83.0 kJ/mol for the TM specimen (Tables 3, 4, and Ran, et al.²⁴ ). In quenched and tempered Fe-C-Cr or Fe-C-Mo steels, the E_d value of large-size spherical M₂₃C₆ carbides and M₆C carbides was reported to be at least 51 kJ/mol and 84 kJ/mol, respectively.⁴⁷  After examining irreversible hydrogen traps in UHSS using atom probe tomography (APT) and high-resolution TEM, Cheng, et al.,⁶¹  found that M₆C carbides (compared to NbC carbides) are not the dominant irreversible hydrogen trapping sites for H. Therefore, it can be argued that the irreversible hydrogen trapping effect of M₆C and M₂₃C₆ carbides in LAM UNS K92580 steel is quite limited (Figure 8 and Table 4). In contrast, grain refinement of LAM UNS K92580 steel can prompt the increase of irreversible hydrogen traps by increasing grain boundary and matrix packet/block/plate boundaries (Table 3 and Figure 8). Different trap activation energy in a range can be corresponding to the grain boundary of different misorientation.⁶⁰  Furthermore, the intensity of these irreversible hydrogen traps can be further increased by increasing the misorientation angles of matrix packet/block/plate boundaries (Tables 3 and 4, and Figure 7). Therefore, the E_d values of hydrogen trap corresponding to the 2nd and 3rd Peaks of the BM specimen with higher misorientation angle are higher than those of the AD specimen and the TM specimen (Table 4).

In order to achieve good mechanical properties (strength, ductility, fracture toughness, and fatigue crack propagation resistance), fine-grain tempered martensite microstructures is required for LAM UNS K92580 steel.^24-25,29  In comparison to wrought UNS K92580 steel, LAM UNS K92580 steel has a similar size and morphology of prior-austenite grain and grain-interior martensite packet/block/plate, but both MC and M₂C carbides have slight differences. It should be noted that fine rod-like M₂C carbides in LAM UNS K92580 steel have lengths of 6 nm to 10 nm and diameters of 1 nm to 2 nm with average stoichiometric composition of (Cr_0.63Mo_0.27Fe_0.10)₂C,²⁴  but in the wrought condition of UNS K92580 steel, M₂C carbides have a length of 6 nm to 10 nm and diameter of 3 nm to 5 nm with average stoichiometric composition of (Cr_0.75Mo_0.12Fe_0.13)₂C.²⁶  Changes in alloying elements in carbides such as C-rich and Fe-poor coherent M₂C carbides decrease the elastic strain energy and can prompt the formation of semi- or incoherent interfaces for carbides.⁶²  The difference in diameter and average stoichiometric composition of M₂C carbides indicate that the elastic stress field and lattice distortion around M₂C carbides are not necessarily the same for steels fabricated by different methods. Therefore, the E_d values of reversible hydrogen traps corresponding to M₂C carbides are slightly different (Table 4 and Li, et al.¹⁴ ). However, the reversible hydrogen trapping capability of UNS K92580 steel fabricated by different methods (wrought or LAM) is comparable, as both of these fabrication methods impart a similar high volume density of M₂C carbides (Table 7). Moreover, the similar volume density and coherency interface relation of M₂C carbides is indirectly indicated by the similar yield strengths (∼1,765 MPa to 1,810 MPa) for both tempered LAM UNS K92580 steel and wrought counterparts.^12,26,29  Second, in contrast with the Ti-rich MC particles of size 5 nm to 12 nm in wrought UNS K92580 steel,²⁶  the MC carbides in LAM UNS K92580 steel are Nb-rich with a bi-modal size distribution. Therefore, the E_d values of corresponding irreversible hydrogen traps for LAM UNS K92580 steel are slightly different from those for wrought UNS K92580 steel (Table 7), but neither have dominant influences on irreversible hydrogen trapping effect of the steel due to the low concentrations of Nb or Ti.

The D_eff of LAM UNS K92580 steel is not only related to the lattice interstitial hydrogen diffusion coefficient (D_L) but also influenced by reversible hydrogen trapping states. Oriani⁷  proposed that atomic hydrogen in steel is in dynamic equilibrium between reversible hydrogen traps and surrounded lattice interstitial sites. Therefore, atomic hydrogen diffusion through LAM UNS K92580 steel can be slowed with an increase in density and intensity of reversible hydrogen trap. It has been confirmed that two kinds of alloy carbides (M₂C and M₃C) with medium trapping intensity are the dominant reversible hydrogen trap sites in LAM UNS K92580 steel. With the precipitation of M₃C carbides, the D_eff of AD LAM UNS K92580 steel is 7 × 10⁻⁹ cm²/s to 8 × 10⁻⁹ cm²/s (Figure 5[c] and Table 7), which is only about 9% to 10% of D_L.¹⁴  After heat treatment, with only fine and dispersed coherent M₂C carbides instead of M₃C carbides in matrix phase, the D_eff of LAM UNS K92580 steel can be further decrease toward ∼2.8 × 10⁻⁹ cm²/s due to an apparently higher reversible hydrogen trap density (Figures 6[d] and [e] and Table 7). Besides the influence of reversible hydrogen traps in ferrite or martensite phases, austenite phases with low H diffusivity (∼10⁻¹² cm²/s⁶³ ) can play an important role in the D_eff of alloy steel. Furthermore, many studies^3,64  revealed that film-like reverted austenite was beneficial in decreasing the D_eff of martensite microstructure in Ni-rich high-strength steel. For LAM UNS K92580 steel, film-like retained/reverted austenite in the as-deposited and heat-treated microstructures should also effectively impede hydrogen diffusion. However, not all of the austenite in LAM UNS K92580 steel can effectively decrease the D_eff. For example, the high amount of retained austenite with blocky morphology in inter-cellular regions of the AD specimen and regions between bainite packets of the BM specimen does not appear to apparently influence the D_eff values (Figures 4[a] and [b], Tables 3 and 5). It should be noted that the difference in D_eff of the LAM UNS K92580 steel after heat treatment is mainly related to the density of dominant reversible hydrogen traps (Table 7). Therefore, although the austenite content in the tempered martensite microstructure of LAM UNS K92580 steel undergoes an apparent decrease, the D_eff of the steel can decrease due to an increase in the density of reversible hydrogen traps. In addition, in the similar precharging H environment, C_H,diff in different specimens of LAM UNS K92580 steel mainly depends on the hydrogen atoms in reversible hydrogen traps. Furthermore, with the increase in reversible hydrogen trap density, C_H,diff values of the TBMM specimen and the TM specimen with tempered martensite microstructure are much higher than those of the AD specimen and the BM specimen (Table 6).

In comparison, the values of D_eff in wrought UNS K92580 steel with tempered martensite microstructure (tempered at 482°C for 5 h) reported in different literature sources^{3,17-18,51,65}  was over a large range due to the use of different testing methods (Table 7). Sundaram and Marble¹⁷  and Figueroa and Robinson³  used electrochemical methods to monitor the potential of precharging H specimen as a function of time and found that the room temperature D_eff of wrought UNS K92580 steel was 2.91 × 10⁻⁹ cm²/s and 1 × 10⁻⁹ cm²/s, respectively. However, room temperature D_eff extrapolated from the higher temperature (120°C to 200°C) D_eff values from TDS isothermal test results was slightly higher (8.96 × 10⁻⁹ cm²/s),¹⁸  which might be due to lack of consideration that some reversible hydrogen traps with lower E_b can additionally impede H diffusion in lower temperature range (<120°C). Recently, Hu, et al.,⁶⁵  examined hydrogen permeation behavior of wrought UNS K92580 steel in a 0.5 mol/L H₂SO₄ + 0.25 g/L As₂O₃ atmosphere environment using a modified double-cell Devanathan-Stachurski setup and found the D_eff of tempered microstructure at 45°C was 3.35 × 10⁻⁸ cm²/s by utilization of the time-lag method. Without correction of time lag through the consideration of relaxation time during a decay transient, the value of D_eff achieved by the time-lag method is an overestimate.³⁴  Therefore, the room temperature D_eff of wrought UNS K92580 steel can be argued to be no more than 8.96 × 10⁻⁹ cm²/s. Besides data scatter caused by using different testing methods, the difference in D_eff between tempered LAM UNS K92580 steel such as the TM condition (Table 6) and the wrought condition at a similar strength and ductility level (Table 2) can be mainly attributed to changes in M₂C carbide characteristics (such as diameter and composition) slightly modifying the reversible hydrogen trapping states of the steel (literature^14,24,26  and Table 4).

• At least three types of hydrogen trap sites are identified in LAM AerMet100 (UNS K92580) steel, and most of the internal hydrogen atoms are contained in reversible hydrogen traps. The main reversible hydrogen traps in the as-deposited microstructure are M₃C carbides/ferrite interfaces with an E_d value of 17.3±0.2 kJ/mol. After heat treatments, dominantly reversible hydrogen traps with an E_d value of 19.3±0.5 kJ/mol are identified in tempered martensite microstructure due to the formation of highly coherent M₂C carbides. In comparison, given the reported E_d value of ∼21.4 kJ/mol for main reversible hydrogen traps in wrought UNS K92580 steel, the lower E_d value in LAM UNS K92580 steel is closely related to the composition change of M₂C carbides. Furthermore, the main irreversible hydrogen traps with E_d value of over 80 kJ/mol in the steel are high-angle boundaries of grains and grain-interior martensite, which are improved by grain refinement.

• In all of the precharging H specimens, the diffusible and total hydrogen concentration of the TM specimen and the TBMM specimen is about three times as high as that of the AD specimen and the BM specimen. Furthermore, the TM specimen with tempered martensite microstructure has the highest diffusible and total hydrogen concentration due to the high density of dominantly reversible hydrogen traps.

• The effective hydrogen diffusion coefficient (D_eff) of LAM UNS K92580 steel is in the order of 10⁻⁹ cm²/s and is strongly influenced by reversible hydrogen trap density (N_TR). After heat treatment, the N_TR of LAM UNS K92580 steel increased from 2.99 × 10¹⁹ sites/cm³ to 2.85 × 10²⁰ sites/cm³, associated with a decrease of D_eff from 7.7 × 10⁻⁹ cm²/s to 2.8 × 10⁻⁹ cm²/s. In contrast, the effect of simultaneous decreased blocky retained austenite amount on the D_eff of the steel is much less.

This work was supported by the National Key Research and Development Program of China (Grant No. 2018YFB1106000) and Youth Program of the National Natural Science Foundation of China (Grant No.51901010). In addition, senior engineer L.H. Han from the school of Metallurgical and Ecological Engineering (University of Science & Technology Beijing) is acknowledged for help with the LECO gas analysis equipment.