Submerged arc welding (SAW) is widely applied for the welding of thick steel grades due to its inherently high deposition rate.[1] During SAW, no shielding gas is required as the weld pool is protected by a layer of molten slag and granular flux from atmospheric contamination.[2]

Flux is the primary source of O for weld metal (WM), and flux basicity is an indirect indication of the flux O potential.[3,4,5,6] Generally, higher flux basicity means lower flux O potential and better mechanical properties, especially low-temperature toughness.[7,8] Among several flux categories, basic-fluoride fluxes are widely applied to achieve excellent mechanical properties as the addition of CaF2 improves basicity and minimizes flux O potential.[5,9,10,11,12,13]

In recent years, to further improve mechanical properties or counter toughness deterioration of the weld under high heat input welding, TiO2, a typical acicular ferrite (AF) promoting component, is introduced into basic-fluoride fluxes.[11,14,15,16] As such, Ti and O are transferred to the WM via chemical reactions in the arc and weld pool zones.[11,13,17] Subsequently, Ti-containing inclusions are dispersed, AF formation is facilitated, and the enhancement of mechanical properties of the weldment is anticipated.[18]

When TiO2-bearing basic-fluoride fluxes are applied, it is essential to understand the roles of TiO2 in the control of WM compositions to ultimately achieve desired mechanical properties.[11,14,19] Several authors investigated the effects of TiO2 content in varying basic-fluoride fluxes over a wide range of WM compositions. For instance, Kohno et al.[14] developed CaF2-SiO2-BaO-Al2O3-TiO2 fluxes, which could transfer Ti and O to the WMs via the reduction of TiO2 in the fluxes, concluded that the Ti content in the WM increased with lower flux potential. Roy et al.[20] investigated the influence of TiO2 content on WM compositions employing CaF2-SiO2-MgO-Al2O3-MnO-TiO2 fluxes, and found that the Ti level of the WM was promoted with TiO2 addition from 0 to 12.5 wt pct. Zhang et al.,[11] on the other hand, performed SAW using CaF2-SiO2-Al2O3-MgO-TiO2 fluxes with varying TiO2 addition from 1 to 16 wt pct, and revealed that WM Ti content reached a maximum when 6 wt pct TiO2-containing flux was applied.

However, the studies reviewed above only considered WM’s change in final compositions, in which the elemental contribution from the flux is not quantified and evaluated. In SAW engineering, a namely Δ value is widely applied to express the element transfer between flux and WM: a positive Δ value indicates flux contribution to elemental gain for the WM, whereas a negative Δ value means corresponding elemental loss for the WM.[5,7,9,10,12,16,21,22,23,24] Such quantitative measure is essential to understand the transfer mechanisms between flux and WM, and to provide guidelines matching appropriate welding consumables.[9,12,25] Influences of SiO2, MnO, and CaO on the element transfer behaviors have been extensively investigated;[9,10,12,21,22,26] nonetheless, the role of TiO2 on this subject remains at large.

This study focuses on the influence of TiO2 on WM composition control using TiO2-bearing basic-fluoride fluxes. It is well known that oxides tend to provide O2 via decomposition in the welding arc, improve weld pool O level, and influence the transfer of alloying elements.[2,13] Therefore, only CaF2, which intrinsically exerts no O potential for the weld, is selected as extra component to exclude such effects.[2,13,25,26,27] Additionally, CaF2 is an effective modifier for the slag physical properties, especially melting point and viscosity.[9,28] In this regard, binary CaF2-TiO2 fluxes are designed, and elemental transfer between flux and WM is quantified. Furthermore, probable reactions that govern the transfer of O, Ti, Si, Mn, and C are investigated and evaluated from thermodynamic perspectives.

It is accepted that a flux must have a melting range lower than the WM so that the molten flux solidifies after the molten WM does.[7] Also, it is reported that flux should have adequate viscosity (higher than 0.2 Pa·s at 1673 K) to cover and protect the weld pool from atmospheric gases.[7,29] Based on these criteria, phase diagram and viscosity data of CaF2-TiO2 melts are referenced,[30,31,32] and five CaF2-TiO2 binary fluxes were designed with formulas summarized in Table I.

Table I Formulas of Fluxes Employed (Weight Percent)
Full size table

Fluxes were prepared by the fused method.[7] Reagent grade CaF2 and TiO2 powders were selected as raw materials, and the fluxes were melted at 1823 K in graphite crucibles following the procedures stated in our previous studies.[9,12] The fluxes were baked at 973 K for 2 hours to remove moisture and exclude C contamination before welding.[21,22] Evaporation of the fluoride species from the melt fluxes was expected,[28] and the losses of CaF2 and TiO2 determined by X-Ray Fluorescence (XRF, model S4 Explorer) were within 5 pct of their initial compositions, similar to our previous studies.[9,12]

EH36, a typical shipbuilding steel, was selected as the base metal (BM). Lincoln Electric Power Wave AC/DC 1000 SD (Lincoln Electric, China) was employed as the welding equipment. Bead-on-plate double-wires (DC-850 A/32 V for electrode forward, AC-625 A/36 V for electrode backward, 60 kJ/cm) SAW was performed on 24 mm thickness BM, and the welding speed was maintained at 500 mm/min. The chemical composition analysis methods for BM, electrode, and WM are stated elsewhere.[9,12,26] The compositions of BM and electrode are shown in Table II.

Table II Chemical Compositions of BM and Electrode (Weight Percent)
Full size table

The welds were cross-sectioned, polished, and etched by 4 wt pct nital solution to determine WM geometries for the calculation of WM nominal composition (composition considering only simple physical mixing of BM and electrode).[9,12,21,22] As is mentioned previously, the Δ value is used to quantify the element transfer between flux and WM. The calculation methods of nominal composition and Δ value have been stated in previous studies.[12,21,22] Analytical compositions (with subscript ‘A’), nominal compositions (with subscript ‘N’), and Δ values of WMs are given in Table III.

Table III Chemical Compositions of WMs (Weight Percent)
Full size table

First, the transfer of O is investigated as it fundamentally dictates the transfer of alloying elements.[9,17,33] ΔO values, which reflect the O potentials of the fluxes, are plotted in Figure 1 as a function of TiO2 content. As shown in Figure 1, ΔO value increases from 90 to 190 ppm when the TiO2 level is improved from 20 to 40 wt pct.

Fig. 1
figure 1

ΔO values as a function of TiO2 content in fluxes

Full size image

Chai et al.[13,27] and Mitra et al.[17,34] reported that TiO2 was susceptible to decomposition, promoted the release of O2, and improved the O level of the droplets and weld pool in the presence of the welding arc, although the decomposition mechanisms were not fully understood. Mitra et al.[17] and Lau et al.[3] concluded that the O uptake from the flux was governed by the level of \( p_{{{\text{O}}_{ 2} }} \) in the plasma.[22] Thus, it is deduced that \( p_{{{\text{O}}_{ 2} }} \) is promoted with increasing TiO2 addition owing to higher extent of oxide decomposition, similar to the case when binary CaF2-SiO2 and CaF2-MnO fluxes were applied.[9,10,13] It should be noted that Chai et al.[13] also investigated the O potential of binary CaF2-TiO2 flux with one formula; however, the dilution effects of BM and electrode were not taken into account, and the compositional contribution from the flux (ΔO) was not quantified.

Compositions of Ti as a function of TiO2 variation in the flux are plotted in Figure 2. Ti is transferred to the WM with TiO2 addition in the flux according to positive Δ values observed. Zhang et al.[11] and Erokhin[35] reported that Ti was transferred from slag to weld pool via Reactions [1] and [2] at the slag-metal interface in SAW.

Fig. 2
figure 2

ΔTi values as a function of TiO2 content in fluxes

Full size image
$$ ( {\text{TiO}}_{ 2} ) {\text{ = [Ti] + 2[O]}} $$
(1)
$$ 2 [ {\text{Fe] + (TiO}}_{ 2} ){\text{ = [Ti] + 2(FeO)}} $$
(2)

From an observation in Figure 2, when the TiO2 content in the flux is lower than 30 wt pct, ΔTi increases from 0.009 to 0.031 wt pct with TiO2 addition. However, when the TiO2 content exceeds 30 wt pct, transfer of Ti is suppressed. As is discussed above, the O content in the weld pool increases with higher TiO2 content. Also, Indacochea et al.[22] concluded that the activity of FeO at the slag-metal interface was proportional to the weld pool O level. Therefore, the increase of O and FeO activity levels suppresses the transfer efficiency of Ti by driving both Reactions [1] and [2] to the left sides. Such reduction of Ti transfer efficiency with higher flux O potential was also observed in the work of Kohno et al.[14] and Zhang et al.[11] when TiO2-bearing basic-fluoride fluxes were employed.

Additionally, Devletian et al.[36] and Sikorski[37] reported the possibility of TiF4 gas evolution in welding processes. Thus, the chemical interaction between CaF2 and TiO2 (Reaction [3]), which may promote the reduction of Ti via the formation of TiF4 gas, should be considered.[28]

$$ 2 ( {\text{CaF}}_{ 2} ) {\text{ + (TiO}}_{ 2} ){\text{ = (CaO) + TiF}}_{ 4} ({\text{g}}) $$
(3)

It is noted that precise information on the gas compositions in the plasma is still not available, as is reviewed by Sengupta et al.[1] To quantitatively analyze the influence of Reaction [3] on the loss of Ti, the precise partial pressure of TiF4 in the plasma is anticipated.[25]

To further confirm the element transfer mechanisms above, compositional changes between slags and their corresponding fluxes, which are defined as δ values, are quantified and illustrated in Figure 3.[9,12] The observed negative δTiO2 values indicate that Ti can be transferred to the WMs by the reduction of TiO2 in the fluxes.[14] By comparing Figures 2 and 3, the magnitude of δTiO2 generally increases with higher level of TiO2 addition, although the ΔTi value undergoes a spike at 30 wt pct TiO2. The difference of the changing trends between the magnitude of δTiO2 and ΔTi implies that Ti is lost to the gas phase, most probably via Reaction [3].[28,36,37]

Fig. 3
figure 3

δ values of TiO2 and FeO as a function of TiO2 content in fluxes

Full size image

Furthermore, FeO was analyzed in the slags. It is reported that Fe is oxidized at the slag-metal interface via Reaction [4].[17]

$$ {\text{Fe(l) + [O] = (FeO)}} $$
(4)

Reaction [4] generally proceeds only forward, as is concluded by Mitra et al.[17] As shown in Figure 3, δFeO value is directly proportional to the concentration of TiO2 due to improved O level in the weld pool,[17] and Reaction [4] is shifted to the right side, which is consistent with our previous conclusions.[9,12]

Si and Mn are basic elements and deoxidizers for submerged arc welded metal.[27] Si and Mn are lost to slag via Reactions [5] and [6] as no SiO2 and MnO are contained in the fluxes initially,[9] which is reflected by negative ΔSi and ΔMn values illustrated in Figure 4.

Fig. 4
figure 4

Values of ΔSi, ΔMn, and ΔC as a function of TiO2 content in fluxes

Full size image
$$ [ {\text{Si] + 2[O] = (SiO}}_{ 2} ) $$
(5)
$$ [ {\text{Mn] + [O] = (MnO)}} $$
(6)

With the increase of TiO2 content from 20 to 40 wt pct, the loss of Si increases from 0.031 to 0.046 wt pct, and the loss of Mn increases from 0.181 to 0.215 wt pct, due to higher O level in the weld pool, which drives both Reactions [5] and [6] to the right sides.

Decarburization tends to occur due to O2 generated in the plasma and O uptake in the weld pool from the flux, and C is lost at the plasma-metal interface via Reactions [7] and [8].[22,25] The contribution of flux to the extent of decarbonization (ΔC) is quantified and shown in Figure 4.

$$ 2 [ {\text{C] + O}}_{ 2} ( {\text{g) = 2CO(g)}} $$
(7)
$$ [ {\text{C] + [O] = CO(g)}} $$
(8)

Furthermore, the possibility of bubble nucleation under slag is investigated as it determines whether decarburization is feasible at the bubble-metal interface.[22,25,38] The bubble nucleation possibility under silicate-network based slags has been investigated; however, similar studies involving titanate-network based fluxes remain scarce.[25] In this study, no significant bubble is observed under detached-slag surfaces, possibly due to bubble nucleation being hindered by the large surface tensions of the slags.[22,25] Therefore, it is expected that Reactions [7] and [8] are only possible at the plasma–metal interface for the present flux system.

In summary, by designing and employing fused CaF2-TiO2 binary fluxes, transfer behaviors have been quantified, and possible mechanisms have been evaluated from thermodynamic perspectives. The following conclusions can be drawn:

  1. 1.

    TiO2 tends to decompose, release O2, and ultimately improve the O level of the WM.

  2. 2.

    Ti is transferred to the weld pool via slag-metal reactions. When the TiO2 addition level in the flux is higher than 30 wt pct, the transfer of Ti to the WM is suppressed due to improved flux O potential and chemical interaction between CaF2 and TiO2.

  3. 3.

    Si and Mn in the weld pool are oxidized at the slag-metal interface. C is lost via decarburization reactions, which likely occur at the plasma-metal interface.