Formation and evolution of layered structure in dissimilar welded joints between ferritic-martensitic steel and 316L stainless steel with fillers

1. Introduction

Dissimilar welding is widely utilized in engineering applications such as oil and gas, petro-chemistry, power-generation engineering and nuclear industry due to low costs and excellent combination of attractive properties of component materials [[1], [2], [3], [4]]. In recently developed advanced Generation VI fission reactors and future fusion reactors, dissimilar welding between reduced activation ferritic/martensitic (RAFM) steels (structural materials) and 316 L (N) stainless steel (cooling pipes and vacuum vessel) is expected by virtue of their design flexibility and technical consideration [[5], [6], [7]]. The direct welding between RAFM and 316 L stainless steel via high-energy beam (HEB) techniques including laser and electron beam welding (EBW) has been extensively studied and evaluated in the past two decades by virtue of generation of narrow fusion zone and heat-affected zone (HAZ) [2,[8], [9], [10], [11], [12], [13], [14], [15], [16]]. This dissimilar HEB welding is not merely beneficial to mitigate residual stress and weld distortions [17], but also conducive to lessen irradiation embrittlement which is more likely to occur in the weldment than that in base metals [16].

However, two problems are envisaged from the characteristics of phase transformation in HEB direct welding between RAFM and 316 L stainless steels [11]. First, due to the mixing of two steels, the composition of weld metal is between the composition of two base metals, which means that the weld metal is not adequately stable to maintain austenitic structure at room temperature, while the nickel content of resultant weld metal is high enough to retard the γ→α transformation to martensite zone. This means that the weld metal exhibits a stronger hardening effect than similar RAFM welded joints. Thomas et al. [12] have also explored higher hardness of the weld metal due to the formation of fine lath martensite in the dissimilar electron beam welding between RAFM steel and 316 L stainless steel. Second, the hardening in the heat-affected zone (HAZ) and weld metal (WM) of dissimilar direct welded joint is difficult to eliminate at the same time after post weld heat treatment (PWHT) because of different compositions. Serizawa et al. [13] have also reported that the hardness of WM was still larger than that of base metals after PWHTs. In addition, the carbon migration from the ferritic steel side to the austenitic stainless steel during the service-life is considerably easier due to the higher diffusivity of carbon in body-centered cubic (BCC) of WM, which caused a carbon-depleted region in ferritic steel close to fusion line [18,19].

To address this aspect, Serizawa et al. [14] deflected laser beam toward the 316 L stainless steel side during welding to increase the nickel content of WM to a level identical to 316 L stainless steel, and found that the hardness of WM was almost the same to that of base metals. Nevertheless, Kano et al. [15] have recently indicated that martensite continued to exist in the WM in the case of 0.2 mm beam offset, which is possibly related to the difficulty in controlling the stability of the beam.

Another feasible method to improve the metallurgical properties of dissimilar weld metal is adding appropriate austenitic metals between RAFM steels and 316 L stainless steel. Furuya et al. [20] have added proper austenitic stainless filler in dissimilar tungsten-inert gas (TIG) welding of RAFM steel to 316 L stainless steel, and found that the microstructure and hardness distribution of the bulk weld metal were relatively uniform. However, solute macrosegregation easily occurred in narrow high-energy beam weld pool because of incomplete mixing, particularly in the case of large compositional difference between the filler metal and the base metal.

In the dissimilar arc welding, macrosegregation characteristics such as base-metal-like “beaches”, “peninsulas” and “islands” were formed near the fusion boundary, which were caused by mismatch of liquidus temperature between weld metal and base materials [21]. Yang and Kou [22] have discussed the macrosegregation features of filler rich zone near the weld bottom in the study of Cu-30Ni arc welds with dissimilar filler metals, the composition and microstructure of bulk weld metal were relatively uniform. Zhang et al. [23] have also corroborated that the microstructure and mechanical properties of the unmixed zone (UMZ) near the fusion boundary were significantly different from that of the bulk weld metal in the study of dissimilar filler welding. However, because of the extremely high cooling rate in the high-energy beam welding, the partially mixed liquid of base metals that are stirred in the weld pool is easily undercooled into the miscibility gap zone to form some immiscible liquid layers [21]. The larger undercooling increased the viscosity of the liquid metal, and the layered liquids solidified as a layered structure in the weld pool before being dispersed, producing banded macrosegregation in the dissimilar welded joints. Similar results regarding banded onion-ring structure macrosegregation in the dissimilar 2205DSS/Q235 laser welding have also been reported by Du et al. [24]. Hence, different from the relatively uniform austenite microstructure of the bulk weld metal in dissimilar filler arc welded joints, the layered structure often occurs in the high-energy beam weld metal. Moreover, strong convection in the high-energy beam weld pool led to relatively smaller peninsulas and thinner beaches near the fusion boundaries.

In the present study, 310S austenitic stainless steel foil and pure nickel foil were added between China Low Activation Martensite (CLAM, one of RAFM steels recently developed in China) steel and 316 L stainless steel, respectively, to improve the metallurgical properties of WM via the electron beam welding with identical welding parameters. According to the above discussion, the banded macrosegregation should occur in both resultant weld metals. An attempt was made to obtain a solution to the following aspects: what is the nature of banded macrosegregation and what influence does it have on the microstructure of both the resultant weld metals? How does the microstructure of resultant weld metals evolve during PWHTs? Until now, there have been no related studies focusing on these issues in dissimilar high-energy beam welded joint between RAFM steel and 316 L stainless steel with filler foil metal. The present study aims to address these issues, and to improve the mechanical properties of dissimilar high-energy beam welded joints through appropriate PWHTs.

2. Experimental procedure

2.1. Welding

In this study, CLAM and 316 L austenitic stainless steel plates with a thickness of 3 mm were used as the base materials for dissimilar joining. The filler interlayers were 310S austenitic stainless steel foil (thickness: 0.4 mm) and pure nickel foil (thickness: 0.3 mm). The compositions of all the raw materials are listed in Table 1. The CLAM plates were normalized at 980 °C for 30 min followed by tempering at 760 °C for 90 min before welding. The pristine 316 L stainless steel was as-rolled plate. All the materials were cleaned by acetone and alcohol before welding.

Electron beam welding (EBW) was adopted for square butt welding with added 310S stainless steel filler foil (named as 310S-filler welded joint) and pure nickel filler foil (referred as Ni-filler welded joint), as shown in Fig. 1. The welding parameters were identical in both the cases and as listed in Table 2. Circular oscillation of electron beam was utilized in the EBW process, which could improve the gap tolerance and decrease the requirements of assembly accuracy [25].

After welding, samples for microstructural observations and mechanical properties tests were cut from the center part of two dissimilar joints, as shown in Fig. 1. Samples were subjected to two conditions of PWHTs, (I) post-weld direct tempering (PWDT, tempering at 740 °C/2 h/air-cooling) and (II) post-weld normalizing plus tempering treatment (PWNT, normalizing at 980 °C/30 min/air-cooling + tempering at 740 °C/2 h/air-cooling) to study the effect of PWHTs on the microstructure and mechanical properties of joints. Samples in various conditions (as-welded, PWDT and PWNT) for 310S-filler welded joint and Ni-filler welded joint are referred as 310S-as-welded, 310S-PWDT, 310S-PWNT, Ni-as-welded, Ni-PWDT, and Ni-PWNT, respectively.

2.2. Microstructural analysis

Specimens in the cross-section of joints were mechanically polished to mirror finish. The etching method for optical microscopy (OM) observations were identical to those described by Liu et al [11]. Samples for scanning electron microscopy (SEM) observations of 310S-WM were etched by aqua-regia (50 ml HNO₃ and 150 ml HCl) in as-welded state, and Kalling’s reagent (5 g CuCl₂, 100 ml HCl, and 100 ml ethanol) after PWHTs. The ZEISS ULTRA-55 field emission scanning electron microscope (equipped with EBSD and EDS) operated at 20 kV was used to carry out scanning electron microscopy (SEM), energy dispersive X-ray spectrometer (EDS), and electron backscattered diffraction (EBSD) studies. Sixteen locations in 310S-WM and Ni-WM shown in Fig. 2(a) and (b), respectively, were selected to ascertain the distribution of elements by EDS. The EDS mapping results are the mean composition of a square box near the microhardness indent (as marked).

The resultant weld metal in dissimilar welding between the ferritic-martensitic steel and austenite stainless steel was predicted in terms of the Schaeffler equation [26], given by:

Cr_eq=Cr+1.5Si + Mo+5V+0.5Ta+0.75 W (1)

Ni_eq=Ni+0.5 Mn + 30C+30N+0.3Cu + Co (2)

where Cr_eq and Ni_eq are chromium and nickel equivalent [27], respectively.

2.3. Mechanical tests

Three lines of hardness measurements were carried out in the upper, middle, and lower part of the cross-section of joints with 0.5 kgf load for 15 s at a regular interval of 0.25 mm, as represented by the indentations in Fig. 2. The transverse tensile tests were utilized to evaluate the reliability of the dissimilar CLAM/316 L welded joint. Longitudinal tensile tests were conducted to obtain mechanical properties of resultant weld metals. The size of tensile samples was the same as those described by Liu et al. [11]. The tensile tests were performed at ambient temperature. Transverse tensile test samples for 310S-filler welded joint and Ni-filler welded joint in various conditions are named as T-310S-as-welded, T-310S-PWDT, T-310S-PWNT, T-Ni-as-welded, T-Ni-PWDT, and T-Ni-PWNT, respectively. Longitudinal tensile samples for 310S-filler welded joint and Ni-filler welded joint in various conditions were named as L-310S-as-welded, L-310S-PWDT, L-310S-PWNT, L-Ni-as-welded, L-Ni-PWDT, and L-Ni-PWNT, respectively.

3. Results

3.1. Microstructure and mechanical properties of dissimilar joints in as-welded state

Microstructure of the cross-section of joints corresponding to the two conditions of welding, namely, 310S-filler welded joint and Ni-filler welded joint, are presented in Fig. 2(a) and (b), respectively. The optical images of these two joints indicate full penetration weldments with no indication of macro-defects such as incomplete penetration or undercut. The foil interlayers (dotted frame in Fig. 2) completely melted and mixed with part of the base metals forming welding pool during the welding process, which solidified as weld metal after cooling. Two different dissimilar welded joints consisted of four zones including base metal of CLAM (CLAM-BM), heat-affected zone in CLAM (CLAM-HAZ), weld metal (WM) and base metal of 316 L (316 L-BM), respectively. There was no HAZ on 316 L side in these welded joints, just resembling the results of Serizawa et al [13]. Samples of WM in 310S-filler welded joint and Ni-filler welded joint are referred as 310S-WM and Ni-WM, respectively.

The degree of blending of elements in the two cases of weld metals was significantly different. The compositions at sixteen positions in 310S-WM were slightly different (the degree of variation of Cr and Ni was less than 2%), as depicted in Fig. 3(a), while the variation of alloy content at different positions of Ni-WM was relatively large (Fig. 3(b)), especially the nickel content (exceeded 4%). This difference can be attributed to the large compositional difference between iron-based base metals and pure nickel filler foil. The average composition (including content of interstitial elements) of two resultant WMs was calculated (the calculation method is presented in the appendix) and the results are presented in Table 3. The Cr_eq and Ni_eq values of two different weld metals were calculated based on their average composition (Table 3) and the results are given in the table of Fig. 3(c).

Optical micrographs (OM) of 310S-WM and Ni-WM are presented in Fig. 4(a) and (e), respectively. The color of CLAM side was black by virtue of the presence of ferrite microstructure, while that of 316L-BM was white for γ-phase. As regards 310S-WM, some black body-centered cubic (BCC) layered structure was presented in austenite weld metal (white color), exhibiting onion-ring appearance (Fig. 4(a)). Whereas the microstructure of Ni-WM was homogeneous cellular grains, except for some slightly layered structure (Fig. 4(e)). Enlarged EBSD images (Fig. 4(b) and (c)) indicated that the microstructure of 310S-WM was predominantly dendritic austenite (50-100 μm), perpendicular to the fusion line, and extended to the center of weld metal. Besides, band contrast (BC) map in Fig. 4(b) and (c) clearly illustrated that the layered structure was perpendicular to the dendritic austenite. In the case of Ni-WM, the microstructure was coarse single-phase austenite columnar grains with width of 100-200 μm (Fig. 4(f) and (g)), consistent with the prediction of Schaeffler diagram (Fig. 3(c)).

The microstructure of 310S-WM matrix (Fig. 4(d)) was consistent with the predictions of empirical relations (Cr_eq˜$\widetilde{1}$8.8, Ni_eq˜$\widetilde{1}$0.9) in Schaeffler diagram (Fig. 3(c)), which is solidified dendritic austenite with a small amount of δ-ferrite (Cr and Mo enriched) distributed along the solidified dendritic boundaries, formed in [FA] solidification mode through a complicated peritectic-eutectic reaction [28]. The δ-ferrite that solidified along dendritic boundaries in 310S-WM was referred as δ_pe-ferrite to distinguish from others in this study. Fig. 4(h) shows the solidified sub-grains (cellular structure) in the single-phase austenitic Ni-WM through [A] solidification mode [29]. The boundary separating adjacent sub-grains was considered as a solidification sub-grain boundary (SSGB). The packets or groups of parallel cellular grain formed the coarse austenite columnar grain (Fig. 4(f) and (g)), and the intersection of these columnar grains was the solidification grain boundary (SGB). Kou [30] has reported that the solute and impure elements can easily concentrate at the solidified interfaces in face-centered cubic (FCC) structure, therefore, some liquid micropores with low melting temperature in austenitic WMs always occurred along the SSGB, producing some shrinkage micropores along SSGB in Ni-WM (Fig. 4(h)) after cooling.

The BC map (inset in Fig. 5(a)) depicts that the layered structure consisted of δ-ferrite (named as δ_L-ferrite) and austenite in the as-welded state of 310S-WM. Some parallel δ_L-ferrites were concentrated in the local area (as ellipse in Fig. 5(a)), producing a large amount of phase interface (δ_L/γ) in the zone. Besides, there was no obvious composition difference between δ_L-ferrite and γ-phase based on the EDS results (Fig. 5(b)). The average content of nickel element in the layered structure zone (Ni ˜$\widetilde{6}$.7%) was slightly lower (˜2%) than that in 310S-WM matrix. Fig. 5(c) shows that lath martensite existed in CLAM-HAZ in as-welded state, which consisted of coarse-grained heat affected zone (CGHAZ, 50-80 μm width) and fine-grained heat affected zone (FGHAZ). Moreover, some δ-ferrite (named as δ_CLAM-ferrite) was left in the CGHAZ near fusion boundary (Fig. 5(c)) by virtue of incomplete transformation of δ → γ at fast cooling process. In addition, the δ-ferrite has detrimental effect on the mechanical properties such as DBTT and creep properties in 9Cr-welded structure [31,32].

Vickers hardness plots across the cross-section of welded joints in as-welded state for 310S-filler welded joint and Ni-filler welded joint are presented in Fig. 6(a) and (b), respectively. The CLAM-HAZ and base metals for these two dissimilar welded joints exhibited similar hardness. The average hardness of CLAM-BM was $\widetilde{2}$28 HV, higher than 316 L-BM (150 HV). The CLAM-HAZ exhibited a high hardness with an average value of $\widetilde{3}$67 HV because of the lath martensite in the HAZ. Although the layered structure existed in as-welded 310S-WM, the hardness of 310S-WM was uniform (average value $\widetilde{1}$86 HV), and higher than 316 L-BM. The average hardness of Ni-WM was $\widetilde{1}$30 HV (Fig. 6(b)), and lower than the pristine 316 L-BM.

The tensile stress-strain behaviors of samples L-310S-as-welded and L-Ni-as-welded are presented in Fig. 6(c), and the longitudinal tensile test data are presented in the table. Similar to the lower hardness of Ni-WM, the strength of Ni-WM was lower than the pristine 316 L-BM, resulting in the fracture of transverse tensile sample of T-Ni-as-welded in Ni-WM (Fig. 6(d)). The ultimate tensile strength (UTS $\widetilde{8}$34 MPa) of 310S-WM in as-welded state was higher than the two base metals, and the fracture of sample T-310S-as-welded occurred in 316 L-BM. Moreover, the 310S-WM had a relatively higher total elongation ($\widetilde{6}$4.2%), consistent with the deformation (ΔL =1.52 mm) of 310S-WM for the transverse tensile sample T-310S-as-welded after the test (Fig. 6(d)).

3.2. Effect of PWDT on the mechanical properties of dissimilar joints

After PWDT, the hardness of CLAM-HAZ decreased rapidly (Fig. 7(a) and (b)) for tempered martensite formed in CLAM-HAZ after 740 °C/2 h, as depicted in Fig. 8(a). The δ_CLAM-ferrite continued to exist in CGHAZ. The average hardness of CLAM-HAZ was ˜234 HV, close to that of CLAM-BM. The average hardness of Ni-WM was almost unchanged after PWDT (Fig. 7(b)). The lower strength of Ni-WM (Fig. 7(c)) led to the fracture of transverse tensile sample T-Ni-PWDT in Ni-WM after the tensile test (Fig. 7(d)). Besides, the SSGB in Ni-WM partially disappeared after 740 °C/2 h, while the shrinkage micropore was not eliminated (Fig. 8(b)).

However, the hardness distribution of 310S-WM after 740 °C/2 h was remarkably inhomogeneous, some deviation in hardness occurred in the 310S-WM (Fig. 7(a)), and the average hardness increased to 221 HV. In addition, the ultimate tensile strength (UTS) of 310S-WM after PWDT decreased to $\widetilde{7}$28 MPa, but still higher than the base metals, and the fracture occurred at 316 L-BM for the transverse tensile sample T-310S-PWDT (Fig. 7(d)). It is worth noting that the ductility of 310S-WM decreased significantly (TE ˜ 47.8%) after PWDT, consistent with the deformation (ΔL =0.24 mm) of 310S-WM for the transverse tensile sample T-310S-PWDT after the test (Fig. 7(d)).

3.3. Evolution of layered structure in 310S-WM during PWHTs

The Ni-filler welded joints in the as-welded and PWDT states did not satisfy the safety design criteria due to the inferior mechanical properties of Ni-WM. The response of ductility for CLAM-HAZ and 310S-WM to PWDT was inconsistent. The balance between strength and toughness for CLAM-HAZ was obtained after PWDT, while the worse ductility occurred in 310S-WM.

The optical micrograph (OM) of 310S-WM after PWDT is presented in Fig. 9(a). Obviously clearer phase interfaces (δ_L/γ) occurred in the layered structure zones after PWDT (Fig. 9(b)), which can be attributed to the precipitates along the interfaces (Fig. 9(c)). From EDS results, it can be detected that the precipitates were enriched with Cr and Mo. Hsieh and Wu. [33] have reported that δ-ferrite can promote the formation of σ-phase (inhalation of elemental Cr and Mo) in the fusion zone of dissimilar stainless welded joints. Therefore, it can be derived that σ-phase is responsible for decreasing the ductility of 310S-WM after PWDT.

The σ-phase preferentially nucleated at phase interface and grew into δ-ferrite via the eutectoid decomposition reaction of ferrite (δ→σ+γ₂). The ferrite stabilizers (Cr, Mo) in δ-ferrite diffused into σ-phase and the nickel atoms were rejected to the ferrite-phase at the same time, forming secondary austenite (γ₂). Therefore, the intermetallic σ-phase was enriched with Cr and Mo, while the secondary austenite (γ₂) was enriched with Ni. A higher density of phase interface (δ_L/γ) in the layered structure zone provided more nucleation sites for σ-phase, producing massive continuously distributed brittle σ-phase after PWDT (Fig. 9(c)), as the evolution mechanism-Ⅰ illustrated in Fig. 10. This led to heterogeneous hardness distribution in 310S-WM. Fig. 9(d) illustrates that δ_pe-ferrite with high Cr and Mo content in 310S-WM matrix transformed into σ-phase after PWDT at the same time. However, the discontinuous σ-phase had less hardening effect in the 310S-WM matrix.

If δ_L-ferrite transformed into γ-phase (thermodynamically stable at room temperature) by normalizing treatment at higher temperature, the volume fraction of δ_L-ferrite as well as the density of phase interface (δ_L/γ) in the layered structure of 310S-WM would decrease dramatically. Thus, the normalizing treatment can effectively reduce the probability of occurrence for σ-phase, as illustrated in Fig. 10 (mechanism-Ⅱ). To confirm this idea, a post-weld normalizing (980 °C /0.5 h/air-cooling) and tempering (740 °C/2 h/air-cooling) treatment (PWNT) was carried out.

After PWNT, the microstructure of 310S-WM was mainly γ-phase (white color), and only a small amount of layered structure (black) existed in the weld metal near CLAM-BM side (Fig. 11(a)). Therefore, it can be confirmed that the majority of δ_L-ferrite transformed into γ-phase after normalizing heat treatment (980 °C/0.5 h), and the formed γ-phase was stable at room temperature. Fig. 11(b) is an enlarged SEM micrograph corresponding to the box b in Fig. 11(a). Cr and Ni content of layered structure zones was still lower than the weld metal matrix (Fig. 11(c)). In addition, the δ_pe-ferrite was not eliminated at high temperature, and σ-phase nucleated along the phase interface (δ_pe/γ) and grew into the δ_pe-ferrite in the 310S-WM matrix after subsequent tempering, as shown in Fig. 11(d).

3.4. Mechanical properties of dissimilar joints after two-step heat treatment

After PWNT, 310S-WM exhibited relatively uniform hardness distribution (Fig. 12(a)). Only the side near CLAM-BM indicated relatively high hardness, corresponding to a small amount of layered structure in 310S-WM after PWNT (Fig. 11(a)). The average hardness of 310S-WM was slightly greater than that of pristine 316 L-BM (Fig. 12(b)).

The hardness of Ni-WM was lower than the base metals after PWNT (Fig. 12(c) and (d)), consistent with the coarse columnar in Ni-WM (Fig. 13(b)). Besides, SSGB almost disappeared after PWNT, and the shrinkage micropore continued to exist along the SSGB.

The hardness distribution of CLAM-HAZ was significantly uniform after PWNT (Fig. 12(a) and (c)), and the average value was ˜220 HV, which was almost identical to CLAM-BM, as shown in Fig. 12(b) and (d). Thus, the superior balance of strength and toughness for CLAM-HAZ was obtained after PWNT. The uniform hardness distribution of CLAM-HAZ after PWNT was consistent with the homogeneous microstructural distribution (Fig. 13(a)) because of the disappearances of coarse martensitic structure and δ_CLAM-ferrite after PWNT.

The rupture of transverse tensile sample T-Ni-PWNT occurred at Ni-WM (Fig. 12(f)) due to lower strength of Ni-WM. After PWNT, the TE of 310S-WM increased to 57.7% (remarkable plastic deformation with ΔL = 1.08 mm) in comparison with the sample L-310S-PWDT. In addition, the strength of sample L-310S-PWNT was higher than that of the base metals, and fracture occurred at the base metal of 316 L stainless steel (Fig. 12(f)), which satisfied the safety requirement for the dissimilar joint.

In summary, the effects of PWHTs on the mechanical properties of CLAM-HAZ and layered structure in 310S-WM were significantly different. PWNT was conducive to both CLAM-HAZ and 310S-WM.

4. Discussion

4.1. Mechanism of formation of layered structure in dissimilar electron beam welded joints

As illustrated in Fig. 14(a), the melted three types of raw material (two types of base metals and a filler foil) mixed with each other in the stirred weld pool during the welding process. The liquidus temperature of two resultant weld metals (310S-WM and Ni-WM) and two base metals (CLAM-BM and 316 L-BM) are listed in Table-II of Fig. 14(c). It is apparent that 316 L-BM had similar liquidus temperature to the two resultant WMs (1450-1475 °C), while the liquidus temperature of CLAM-BM (1500-1525 °C) was higher than the two resultant weld metals. Therefore, the unmixed liquid CLAM-BM ahead of the solidification front (zone-I in Fig. 14(b)) was quickly frozen before mixing with the cooler surrounding liquid in the weld pool, and formed base-metal-like ‘peninsula’ near the fusion boundaries, as shown in Fig. 15(a) and Fig. 15(b).

On the other hand, the liquid CLAM-BM incompletely mixed with the liquid in the narrow weld pool due to the viscous effect of weld pool boundaries [34], as shown by the zone-II in Fig. 14(b). In view of high cooling rate in the weld pool, the partially mixed liquid of base metal easily undercooled into the miscibility gap to form some immiscible liquid layers. The larger undercooling increases the viscosity of the liquid layered base metal, which was solidified as a layered structure in the weld pool before being dispersed. This insufficient blending in the weld pool (liquid state) resulted in the banded alloy-depleted structure in two dissimilar resultant weld metals (solid state), producing compositional fluctuation in the resultant weld metals, similar to the banded elements distribution in the weld metal of F82 H/SUS316 L dissimilar laser welded joint [13]. Horizontal view of the micrograph of 310S-filler welded joint (Fig. 15(c)) demonstrated that the peninsula and layered structure were both derived from incomplete mixing of liquid layer of CLAM-BM, and only the extent of mixing for peninsula and layered structure was different.

The layered structure did not exist in Ni-WM (Fig. 15(d)), while the compositional fluctuation in Ni-WM was relatively larger than 310S-WM. The reason is illustrated using the quasi-binary phase diagram by the vertical-section of 70% iron (Fig. 14(d)) as follows:

The solidification sequence of resultant weld metal matrix as well as the alloy-depleted zone are summarized in Table 4. It is apparent that the sequence of phase transformation and final microstructure of the alloy-depleted zone in Ni-WM was identical to Ni-WM matrix, forming homogeneous austenitic cellular grain in Ni-WM. Although the composition fluctuation in 310S-WM was less than Ni-WM, the phase transformation sequence of alloy-depleted zone was altered to L→L+δ→δ→δ+A, which is different from the 310S-WM matrix (L→L+δ→L+δ+A→δ+A). The subsequent transformation of δ → γ in alloy-depleted zone of 310S-WM was suppressed due to a relatively lower phase transformation temperature (order Ⅳ in Fig. 14(d)) at fast cooling rate. Therefore, a large amount of high temeprature ferrite (δ_L-ferrite) was retained at room temperature, producing “dual-phase” layered structure (γ and δ_L-ferrite) in the 310S-WM, which consisted of phase interfaces with high-density. In addition, due to the relatively less time for elemental diffusion during δ→γ transformation, the composition of δ_L-ferrite was close to that of γ-phase in the layered structure zones.

4.2. Evolution of three types of δ-ferrite in dissimilar joints during PWHTs

The composition difference between the alloy-depleted zone and 310S-WM matrix maintained after PWNT, indicating that long-range diffusion of the substitutional elements (Cr, Ni and Mo) did not occur in the 310S-WM during PWHTs. However, short-range diffusion in the layered structure zones led to different chemical composition among different phases (δ, γ and σ), such that the electrical potential of each phase is different. Therefore, it is easy to etch the layered structure in 310S-WM after PWHTs using relatively weak acid (Kalling’s reagent). In contrast, the phase interface (δ_L/γ) in the as-welded state was necessarily etched by strong acid of aqua-regia, indicating similar composition between δ_L and γ-phase in the layered structure (Fig. 5(b)).

The evolution of three types of high temperature ferrite (δ_L-ferrite, δ_pe-ferrite and δ_CLAM-ferrite) in these dissimilar joints during the PWHTs is gathered in Table 5. The compositions of three types of δ-ferrite were different because of different formation mechanisms during the welding process. They correspond to different phase transformation temperatures, leading to different microstructural evolutions during PWHTs. There was no σ-phase precipitated in δ_CLAM-ferrite due to relatively lower chromium content (Cr $\widetilde{9}$%). For δ_L-ferrite and δ_pe-ferrite, σ-phase formed through eutectoid decomposition (δ→σ+γ₂) during PWDT. The δ_L-ferrite and δ_CLAM-ferrite transformed into γ-phase at high temperature (980 °C). The γ-phase that derived from δ_L-ferrite was thermodynamically stable at room temperature due to high nickel content (Ni $\widetilde{7}$%). While the austenite that derived from δ_CLAM-ferrite sheared into lath martensite after cooling. The δ_pe-ferrite was not eliminated after normalizing treatment (980 °C) due to higher chromium content, which enlarged the ferritic phase range and reduced the austenitic phase range. During subsequent second-step tempering, the retained δ_pe-ferrite experienced δ_pe→σ+γ₂ transformation, while the quenched martensite in CLAM-HAZ transformed to tempered martensite.

4.3. Impact of layered structure on the mechanical properties of dissimilar joints

In the as-welded state, two soft phases consisting of δ_L-ferrite and γ-phase coexisted in the layered structure, which had less detrimental influence on the ductility of 310S-WM. The relatively deeper and larger dimples in the fractograph in the as-welded state (Fig. 16(a)) implied superior plastic deformation ability of 310S-WM.

After PWDT, cracking easily occurred and propagated in the layered structure zones of 310S-WM during the deformation due to decohesion of σ-phase, forming banded crack with quasi-cleavage facet, as shown in Fig. 16(b). The quasi-cleavage facet indicated relatively little absorption energy associated with plastic deformation of sample L-310S-PWDT, which was in good agreement with its small elongation (Fig. 7(c)).

Majority of δ_L-ferrite transformed into stable γ-phase and the layered structure almost vanished in 310S-WM by PWNT. Therefore, the strength and ductility of sample L-310S-PWNT were significantly superior to sample L-310S-PWDT because of the absence of continuous brittle σ-phase. Furthermore, dense discontinuous distribution of precipitates in 310S-WM after PWHTs can act as nucleation site for voids during cold deformation, producing a large number of smaller dimples in the fractographs (Fig. 16(b) and (c)).

The strength and ductility of Ni-WM in various conditions were almost similar, and lower than 310S-WM. The parallel SSGB in the Ni-WM could not act as interfacial barrier for dislocation motion. Hence, the effective grain size of Ni-WM in various conditions was larger than 310S-WM, resulting in lower mechanical properties of Ni-WM. Furthermore, the shrinkage micropores along the SSGB also produced negative effects. These micropores cannot be eliminated after PWHTs. The high stress concentrated near the micropore during tensile deformation, led to fast growth and coalescence of crack, producing some large dimples in the fractographs, irrespective of the absence or presence of heat treatment, as depicted in Fig. 16(d)-(f). The quasi-cleavage facet and large dimples in these fractographs both indicated inferior plastic deformation behavior of Ni-WM.

Additionally, the mechanical stability of γ-phase in resultant WM was evaluated using the mechanical stability index of austenite, Md₃₀ temperature, as reported by Herrera et al. [35], and is given in wt%:

Md(γ) = 551-462(C(γ)+N(γ))-9.2Si(γ)-8.1 Mn(γ)-13.7Cr(γ)-29Ni(γ)-29Cu(γ)-18.5Mo(γ) (3)

The austenite stability index (Md₃₀) is a temperature at which 50% austenite will be transformed to martensite through cold deformation of 0.3 true strain. The composition of γ-phase in 310S-WM and Ni-WM were taken from the average composition of weld metals, respectively. It can be inferred that Ni-WM and 316 L-BM did not occur strain-induced martensite transformation at ambient temperature (20 °C), as shown in Table 6, which implied high mechanically stability of γ-phase in Ni-WM and 316 L-BM. Because the Md₃₀ temperature was 14.3 °C for γ-phase in 310S-WM, the γ-phase in 310S-WM experienced transformation-induced plasticity (TRIP) during tensile straining, resulting in a superior work hardening ability of 310S-WM. As a result, the 310S-WM had superior combination of strength and ductility than Ni-WM.

After PWNT, majority of δ_L-ferrite transformed into austenite, and the continuous coarse σ-phase in the layered structure vanished, and some dispersed fine σ-phase with γ₂ formed in 310S-WM matrix. Based on previous reports, the (σ + γ) is conducive to improve creep strength by hinder the dislocation movements [[36], [37], [38]], thus, it can be inferred that the creep strength of 310S-WM could significantly increase after PWNT.

Fig. 17(a) represents the 650 °C creep curves of two transversal welded joints after PWNT under 150 MPa. Rupture life in this condition of two joints were both several hours. Besides, the fracture position of dissimilar welded joints with 310S and nickel filler were both located in CLAM-BM (Fig. 17(c)), although Ni-WM or 316 L-BM had lower room temperature hardness in comparison to the CLAM steel side (Fig. 17(b)). This can be attributed to higher creep strength of FCC structure by virtue of slower self-diffusion coefficient [39]. In addition, the inhomogeneous martensite microstructure as well as the coarse M₂₃C₆ precipitates in CLAM-HAZ were eliminated during the normalizing heat treatment, which avoided the occurrence of VI creep failure in fine grain HAZ of joint [40]. The δ_CLAM-ferrite near the fusion boundary also be eliminated after PWNT, alleviating the premature of creep failure for joints.

5. Conclusions

In this study, dissimilar CLAM/316 L welded joints were obtained by EBW with added 310S-filler and Ni-filler foils, respectively. The formation and evolution of the layered structure in the dissimilar joints and its effect on the mechanical properties of joints were investigated. The significant results are summarized as follows:

(1) Although the compositional fluctuation of 310S-WM was less in comparison to Ni-WM, layered structure occurred in 310S-WM, rather than in Ni-WM.

(2) The abnormal phenomenon is explained in terms of the variation in phase transformation sequence in resultant WMs due to different fillers. The phase transformation sequence in Ni-WM was uniformly L→L + A→A. The 310S-WM matrix zone was solidified by [FA] mode, while the depletion zones of solute elements in 310S-WM was solidified through [F] mode and transformed by L→L+δ→δ→δ+A, leading to alternating “dual-phase” layered structure in 310S-WM.

(3) In spite of the uniform microstructure, the mechanical properties of Ni-WM were inferior to the base metals because of a relatively larger grain size, shrinkage micropores and inferior work hardening ability (no TRIP effect). In addition, the Ni-WM could not be improved by heat treatment.

(4) Although the layered structure existed in 310S-WM in as-welded state, the mechanical properties of 310S-WM were superior. However, the resistance of non-equilibrium layered structure to thermal disturbance was inferior. Severe hardening or embrittlement easily occurred in 310S-WM after the conventional direct tempering treatment because of eutectoid decomposition reaction of δ_L-ferrite (δ_L→σ+γ₂) in the layered structure.

(5) For the two-step heat treatment (PWNT), the first-step normalizing treatment made δ_L-ferrite transform into γ-phase, and was retained at room temperature, so that there was almost no σ-phase formed during subsequent second-step tempering process. Meanwhile, δ_CLAM-ferrite transformed into γ-phase at higher temperature (980℃) and sheared into lath martensite after cooling, forming homogeneous tempered martensite in CLAM-HAZ after the second-step tempering. The thermal instable phase (δ_L-ferrite) in the layered structure was eliminated via normalizing treatment, and effectively reduced the influence of thermal disturbance on the mechanical properties of 310S-WM. Therefore, a relatively uniform hardness distribution across the 310S-filler welded joint was obtained after PWNT, which was both beneficial to the mechanical properties of CLAM-HAZ and 310S-WM. The fracture of creep failure for 310S-filler welded joint and Ni-filler welded joint in PWNT condition both occurred at CLAM-BM.

Acknowledgements

This work was supported financially by the National Magnetic Confinement Fusion Program of China (Nos. 2014GB120000 and 2014GB104003) and the National Natural Science Foundation of China (No. 51571026). The author thanks Dr. Zheng weiwei in State Key Laboratory for Advanced Metals and Materials of USTB for providing creep-test.

Appendix A. Calculation of composition in the dissimilar narrow weld seam

It is well known that EDS cannot accurately measure the content of lower concentration elements and interstitial elements (C, N) in the narrow zone of dissimilar weld metals. A reliable approach that utilized the major substitutional alloying elements (Cr, Ni) to determine the dilution of the weld seam was proposed. The precondition of this method is that all elements had identical blending degree in the liquid state in the local area of weld pool. The dilution of D_CLAM, D_316L, and D_Filler was defined as volume ratio of CLAM-BM, 316 L-BM, and filler metals to the whole weld in the weld pool. The content of nickel and chromium obtained by EDS were used to calculate the dilutions as follows:

D_{CLAM *} Cr_CLAM + D_{316L *} Cr_316L+ D_{Filler *} Cr_Filler=Cr_WM (4)

D_{CLAM *} Ni_CLAM + D_{316L *} Ni_316L+ D_{Filler *} Ni_Filler=Ni_WM (5)

D_CLAM + D_316L + D_Filler = 1 (6)

Where Cr_CLAM, Ni_CLAM, Cr_316L, Ni_316L, Cr_Filler and Ni_Filler are chromium and nickel content of CLAM-BM, 316 L-BM and filler foil, respectively, which were determined by wet chemical analysis (Table 1). Cr_WM and Ni_WM were content of chromium and nickel in weld metals determined by EDS. The degree of dilution was calculated and presented in the Table 7. The average composition of two resultant weld metals was calculated based on the dilution and composition of raw materials. The results are listed in Table 3.

---