1 Introduction

P92, a typical heat-resistant steel renowned for exceptional service temperature and high creep strength, has found extensive utilization as a preferred candidate for thermal power piping systems and nuclear power equipment.[1] Submerged arc welding (SAW) with deep penetration and satisfactory weld quality has been practiced for manufacturing heat-resistant steels.[2] However, high-temperature and high cooling rate of the weld pool render it inevitable that certain incomplete phase transformation products, such as δ-ferrite, could be retained in the weld metal (WM) after solidification. Wang et al.[3] investigated phase transformation products of P91 steel and found that heterogeneous microstructures caused by retained δ-ferrite showed diverse crystallographic characteristics, especially on Kernel average misorientation (KAM) value. It is pointed out that retained δ-ferrite always demonstrates low KAM value compared with other products, potentially leading to deteriorated hardness, thereby invariably imposing adverse effects towards creep strength.[4,5]

Generally, differing from the reheat zone in multi-pass welding, retained δ-ferrite content in the WM is largely dictated by the influence of major alloying element contents, such as Cr, C, Mn, etc.[6,7] The fraction of retained δ-ferrite could be enhanced by the increase of the Cr content.[8] Alkan et al.[9] changed Cr content in the 12Cr-Ni steel from 11.52 to 13.64 wt pct and the retained δ-ferrite fraction was found to experience appreciable enhancement from 5.5 to 33 pct. Meanwhile, Sam et al.[10] and Arivazhagan et al.[11] mentioned that the increase of C and Mn contents could reduce the amount of retained δ-ferrite in the WM microstructure. Wang et al.[12] demonstrated that, for the 9Cr2WVTa steel, the accrual of C content from 0.08 to 0.16 wt pct was able to lower retained δ-ferrite from 6.7 to 3.5 pct. In addition, for Fe−21Cr–3Ni–1.7Mo–2.4C, varying Mn content from 0.4 to 2.4 wt pct incurred δ-ferrite fraction to decrease from 4 to 2.8 pct.[13]

In SAW, a feasible approach achieving desirable WM composition and microstructure is to adjust welding fluxes.[14,15] Under the effect of ultra-high temperatures, vehement chemical reactions could occur between flux (slag), molten pool, and arc, influencing the transfer of alloying elements.[16,17,18,19,20] It is particularly worth mentioning that for P92 heat-resistant steel, Cr could serve as a solid solution strengthener and carbide former, and is regarded as the most important alloying element for heat-resistant steels. Low Cr content typically lead to coarsening of carbides and reduced mechanical properties. Generally, Cr in the WM is prone to loss through either evaporation or slag formation under high-temperature metallurgical reactions of SAW. Available studies have revealed that, by welding chromoly 4135 steel through SiO2–TiO2–CaO–MgO–Al2O3–MnO–CaF2 flux, Cr content decreased while O content increased in the WM.[21] However, the addition of a Cr2O3 active agent-based SiO2–B2O3–N2O–Al2O3 flux in the welding process on 304 stainless steel enhanced Cr content in the WM.[22] It is evident that the addition of Cr2O3 in flux could effectively compensate for the lost Cr in the WM. However, illustrated effect of the Cr2O3 addition on the variations of other alloying elements, such as C and Mn, has not been fully documented. It is worth noting that current commercial fluxes are basically multicomponent, and it is difficult to single out the role of Cr2O3 oxide. Notably, existing research predominantly explores the regulation of composition in low-alloy steel through flux.[16] However, there is a lack of studies on the impact of specific oxides in the flux on the WM for high-alloy steel.

In the present study, a simple while versatile CaF2–Cr2O3 binary agglomerated flux system geared toward SAW has been designed for P92 heat-resistant steel. The variation of microstructure in the WM, especially retained δ-ferrite, has been investigated. Meanwhile, possible chemical reactions during welding have been postulated to elucidate element transfer behaviors between the fluxes and the WM.

2 Experimental Procedure and Methods

Five CaF2–Cr2O3 agglomerated fluxes were prepared, and the compositions of the fluxes are given in Table I. To match a suitable welding temperature and viscosity,[23,24] Cr2O3 content was designed for 5, 10, 15, 20, and 25 wt pct, respectively, and CaF2 was set to compensate for the rest. The fluxes were made from analytical reagent grade chemicals of CaF2 (> 98.5 wt pct) and Cr2O3 (> 99.0 wt pct). After powders were mechanically mixed for 1 hour to ensure homogeneity, a sodium silicate binder whose weight is 20 pct relative to the weight of the mixed reagent was added for granulation into 10 to 50 mesh. The fluxes were then baked at 700 °C in air for 2 hours to improve particle strength and remove residual moisture of the fluxes. Bead-on-plate double-wires single-pass SAW (MZS-1250, Shangdong Aotai Electric, China) trials were performed on 25.4 mm thickness P92 heat-resistant steel. Thermanit MTS 616 was used as the welding wire. The welding speed was set as 500 mm/min, and a constant heat input of 60 kJ/cm (DC-850 A/32 V for electrode forward, AC-625 A/36 V for electrode backward) was employed.

Table I Chemical Compositions of Employed Fluxes (Weight Pct)
Full size table

Chemical compositions of base metal (BM), welding wire, and WM were determined. C and O were carried out by a LECO analyzer (ONH836, LECO Co.), and other elements were detected by an inductively coupled plasma optical emission spectrometry (ICP-OES, Optima 8300DV, Perkin Elmer). Compositions of BM, welding wire, and WM are summarized in Table II. For convenience, corresponding WMs formed by the five prepared fluxes are designated as ‘WM1’, ‘WM2’, ‘WM3’, ‘WM4’, and ‘WM5’, representing the content of Cr2O3 in corresponding agglomerated flux for 5, 10, 15, 20, and 25 wt pct, respectively.

Table II Chemical Compositions of Base Metal, Welding Wire and Weld Metal (Weight Pct)
Full size table

After welding, samples from the central region of the WM were cut, mechanically grounded, polished, and chemically etched by Vilella’s solution (1 g picric acid + 5 mL hydrochloric acid + 100 mL ethanol). Microstructures of the WMs were observed using a scanning electron microscope (SEM, MIRA3, TESCAN, Czech, acceleration voltage 20.0 kV). Crystallographic features were revealed by electron backscatter diffraction (EBSD, Nordlys Nano, Oxford Instruments, UK, operating voltage of 20.0 kV) system and HKL Channel5 software. The average δ-ferrite fraction of the WM area was determined through metallographic analysis. Specifically, after grinding, polishing, and etching each layer, over 30 fields measuring 0.5 × 0.5 mm2 each were selected from SEM micrographs. The measurement was performed on at least three layers. Phase diagrams were calculated by FactSage 8.1 software using Fstel databases.[25]

3 Results and Discussion

Figure 1 illustrates the detailed room temperature microstructures in the WMs subjected to varied CaF2–Cr2O3 fluxes. Generally, all WM microstructures demonstrate predominately martensite but with a small yet ineligible amount of δ-ferrite, which is distinguished by polygonal morphology and distributed randomly, as shown by the yellow dotted lines in Figures 1(a) through (e). In addition, the fraction of δ-ferrite increases with increasing Cr2O3 content in the fluxes. The quantitative relationship between the relative fraction of δ-ferrite and Cr2O3 content is shown in Figure 1(f). The result reveals that as the Cr2O3 content in the fluxes increases, there is a rise in the fraction of δ-ferrite in the WM, escalating from 1.87 ± 0.87 to 6.86 ± 1.49 pct. The variation represents an approximately 3.7-fold enhancement. The result is similar to the findings by Cai et al.[26] Generally speaking, the phenomenon could be enabled by the synergistic effects of enhanced ferrite stabilizing elements (Cr, W, Mo, et al.) and reduced austenite stabilizers (C, Mn, et al.).[27,28]

Fig. 1
figure 1

Typical microstructures of: (a) WM1, (b) WM2, (c) WM3, (d) WM4, and (e) WM5; (f) relative fraction of δ-ferrite in the weld metal as a function of Cr2O3 (Color figure online)

Full size image

KAM maps of WMs, serving as the most intuitive indicators clearly distinguishing δ-ferrite and martensite, are shown in Figures 2(a) through (e). KAM maps are plotted based on the first neighbor kernel parameter with a maximum misorientation angle of 5 deg. KAM maps indicate that martensite laths show high KAM values while δ-ferrites possess low KAM values (almost 0 deg).[29,30] From WM1 to WM5, it can be seen that the blue area, about 0 deg KAM, gradually reduces. Statistic results of KAM distributions in the selective area are shown in Figure 2(f), and the average KAM values of WM1 to WM5 are 1.46, 1.42, 1.36, 1.35, and 1.30, respectively. It is well-known that higher misorientation angles indicate increased dislocation density and higher degree of strain localization.[31] With the increase of δ-ferrite, the average KAM value displays a depreciated density of dislocations, imparting lower resistance that must be overcome during plastic deformation.[32]

Fig. 2
figure 2

Kernel average misorientation map of WMs: (a) WM1, (b) WM2, (c) WM3, (d) WM4, (e) WM5, and (f) KAM distribution (Color figure online)

Full size image

δ-Ferrite retention is closely related to alloying elements in WMs, which can be estimated by empirical formulas.[3,10] Generally, when the ferrite factor (FF), that is, the difference between the chromium equivalent (Creq) and the nickel equivalent (Nieq), is less than 10, no ferrite could exist in the microstructure. Creq and Nieq are given by the Newhouse formula,[33] as displayed in Eqs. [1] and [2]:

$$ {\text{Cr}}_{{{\text{eq}}}} = {\text{Cr}} + {\text{6Si}} + {\text{4Mo}} + {\text{11V}} + {\text{5Nb}} + {1}.{\text{5W}} + {\text{12Al}} + {\text{8Ti}} $$
(1)
$$ {\text{Ni}}_{{{\text{eq}}}} = {\text{2Mn}} + {\text{4Ni}} + {4}0{\text{C}} + {3}0{\text{N}} + {\text{2Co}} + {\text{Cu}} $$
(2)

where components are in weight percentages. Table III shows the calculated results from the chemical composition of the WMs according to analyzed chemical compositions (Table II). It can be seen that all FF values are more than 10, implying retained δ-ferrite could be present for all WMs.

Table III Ferrite Factor of WMs Predicted by the Empirical Formula
Full size table

Figure 3 shows WM phase diagrams calculated by the FactSage 8.1 software. Calculated results show that the δ-ferrite phase region expands while the γ-austenite phase region shrinks with the addition of Cr2O3 in the fluxes, as displayed in the black (WM1) and purple (WM5) stars in Figure 3. It is further shown that the transformation temperature of δ → γ is lowered, while that of γ → martensite is increased, meaning the dwell time at δ + γ and γ phase fields is shortened during the cooling process.[34] Previous studies have demonstrated that retained δ-ferrite is enabled by the incomplete δ → γ phase transformation.[35] The shorter the time spent in the δ + γ and γ phase fields, the higher the content of the δ-ferrite retained in the microstructure.[10] The aforementioned phenomenon may be caused by the variation of ferrite or austenite stabilizing elements.[36] Both the increase of ferrite stabilizer (Cr) and the decrease of austenite stabilizing element (Mn or C) could reduce the diffusion coefficient of C atoms, which may possibly delay the austenite transformation (as shown by the points A to C in Figure 3).[37,38] In addition, the reduction of C content increases supercooling of martensite transformation, thus the austenite to martensite transformation temperature is enhanced, as displayed the points B to D in Figure 3.[39]

Fig. 3
figure 3

WM equilibrium phase diagrams (Color figure online)

Full size image

It is well-known variation of alloying elements in the WM could be incurred by metallurgical reactions.[40] In SAW, Cr2O3 in the fluxes could decompose and release O2 under the influence of high-temperature plasma of the welding arc in the arc cavity (Eq. [3]). Oxides in the flux provide the total source of O for SAW WM due to flux preventive effect on oxidization by the atmosphere.[41] O2 in the arc cavity could supply O from the fluxes to the WM through gas-metal reactions occurring at the gas-slag interface (Eq. [4]),[41] enhancing O content in the WM. Further, a previous study has shown the negative Gibbs free energy (ΔG) of Cr2O3 decomposition reaction (Eq. [5]), indicating that the main source of Cr in the WM is the decomposition of Cr2O3 rather than the transition of suboxide (CrO). Cr2O3 reacting at the slag-metal interface and transferring Cr and O from the flux into the WM.[42] As the addition of Cr2O3 in flux increases from 5 to 25 wt pct, the activity of Cr2O3 increases from 0.447 to 0.557, while ΔG values vary from − 160.47 to − 62.17 kJ/mol (calculated by FactSage 8.1), indicating a reduced driving force of Cr2O3 decomposition. Thus, contents of Cr and O in the WM are enhanced, which vary from 0.043 to 0.105 wt pct and 9.121 to 9.868 wt pct, respectively, as shown in Table II.

$$ \left( {{\text{Cr}}_{{2}} {\text{O}}_{{3}} } \right) = {\text{CrO}}\left( {\text{g}} \right) + {\text{O}}_{{2}} \left( {\text{g}} \right) $$
(3)
$$ {1 \mathord{\left/ {\vphantom {1 2}} \right. \kern-0pt} 2}{\text{O}}_{{2}} \left( {\text{g}} \right) = \left[ {\text{O}} \right] $$
(4)
$$ \left( {{\text{Cr}}_{{2}} {\text{O}}_{{3}} } \right) = {2}\left[ {{\text{Cr}}} \right] + {3}\left[ {\text{O}} \right] $$
(5)

Furthermore, O in the weld pool could react with certain alloying elements at the slag-metal or gas-metal interface to form oxides, resulting in the loss of alloying elements in the WM.[16,41] Mn can combine with O in the weld pool and transfer to the slag in the form of MnO through slag-metal reactions, as shown in Eq. [6]. With the addition of Cr2O3 content, O content in the WM enhanced, which may significantly increase the driving force for Eq. [6]. Therefore, with the Cr2O3 varying from 5 to 25 wt pct, Mn content decreases from 0.394 to 0.254 wt pct in the WM. Similarly, Si loses from the WM to the slag in the form of Si-oxide through slag-metal reactions.[43] C loses from the WM and transitions into the arc cavity in the form of CO, as shown in Eq. [7].[43] In addition, it needs to be mentioned that the reduction in Ni concentration within the WM is attributed to two primary factors: (1) a decline in Ni content with the escalating dilution of the WM,[44] and (2) the transfer of Ni from the WM to the slag.[45]

$$ \left[ {{\text{Mn}}} \right] + \left[ {\text{O}} \right] = \left( {{\text{MnO}}} \right) $$
(6)
$$ \left[ {\text{C}} \right] + \left[ {\text{O}} \right] = {\text{CO}}\left( {\text{g}} \right) $$
(7)

4 Conclusions

In summary, the present study illustrates a variation of retained δ-ferrite in the WMs of P92 heat-resistant steel as a function of Cr2O3 in CaF2–Cr2O3 flux, and the phenomena are elucidated via salient chemical reactions. The main conclusions are as follows:

  1. 1.

    The microstructure of WM is composed of martensite and retained δ-ferrite. As the content of Cr2O3 in the fluxes improves from 5 to 25 wt pct, the retained δ-ferrite content increases from 1.87 ± 0.87 to 6.86 ± 1.49 pct, and corresponding average KAM values reduce from 1.46 to 1.30.

  2. 2.

    When the addition of Cr2O3 in the fluxes increases, a series of metallurgical reactions in the weld pool occurs, enhancing Cr content while decreasing the contents of Mn, C, and Ni in the WMs, leading to appreciable increase of retained δ-ferrite fraction.