Effects of TiO 2 activating flux concentration on A-TIG welding characteristics of inconel 738LC nickel-based superalloy

1 Introduction

Nickel-based superalloys are regarded as one of the most important classes of materials for advanced industrial applications, such as aerospace and gas turbines, owing to their excellent mechanical properties at elevated temperatures, as well as their superior creep resistance, oxidation resistance, and hot-corrosion resistance (Darolia, 2019; Long et al., 2018; Rodenkirchen et al., 2022). These alloys can generally be classified into two categories: precipitation-hardened and solid-solution-strengthened superalloys. The strength of precipitation-hardened superalloys mainly originates from primary γ′ precipitates, with an approximate chemical composition of Ni₃(Al,Ti) (Chu et al., 2023; Adapa et al., 2025). Among these materials, Inconel 738LC is a widely used nickel-based superalloy and is commonly employed in the manufacture of rotating blades and stationary vanes for aero-engines and industrial gas turbines (Holländer et al., 2016; Mazur et al., 2005; Chaharlang et al., 2019). Because this superalloy is expensive, component replacement after service-induced damage results in high costs; therefore, repair and remanufacturing are of considerable practical importance (Sidhu et al., 2009; Chiang and Chen, 2009). Although welding is an effective method for the remanufacturing and repair of superalloy components, these alloys generally exhibit poor weldability and are prone to defects, particularly solidification cracking and cracking in the heat-affected zone (HAZ) (Danis et al., 2010; Montazeri and Ghaini, 2012; Zhong et al., 2005). Therefore, improving the weldability of nickel-based superalloys remains a key challenge that urgently needs to be addressed in related industries.

Gas tungsten arc welding (GTAW), also known as tungsten inert gas (TIG) welding, is a commonly used welding process for superalloys and can produce high-quality welds. However, its typical characteristics are shallow penetration and a relatively wide weld bead. Such weld geometry is undesirable for thick-section welding because multiple passes are required to fill the joint volume, leading to prolonged processing time and increased heat input, thereby increasing the likelihood of welding cracks (Sonar et al., 2021; Singh and Khanna, 2021; Rakoczy et al., 2015). To overcome this limitation, activated flux tungsten inert gas welding (A-TIG) has been developed. In this process, a flux is applied to the workpiece surface, and the preplaced active flux modifies the physical characteristics of the arc and the fluid-flow behavior of the molten pool. Under the same welding current, this process can substantially increase weld penetration and enable single-pass welding of thick sections. As a result, the number of weld passes and processing time can be effectively reduced. More importantly, the cumulative heat input associated with multi-pass welding can be fundamentally decreased, thereby narrowing the HAZ and suppressing the initiation of HAZ liquation cracking and weld solidification cracking (Lin and Wu, 2012; Lin et al., 2014; Ramk et al., 2015; Kumar et al., 2019).

Considerable efforts have been devoted to improving the weldability of nickel-based superalloys from different perspectives. Yue et al. (2023) conducted welding experiments on 10 mm-thick plates and showed that the formation of deep-penetration welds was associated with changes in Marangoni convection within the molten pool. They also discussed the effects of activating fluxes on HAZ width and mechanical properties. Chen et al. (2018) investigated HAZ liquation cracking in Inconel 738 alloy using GTAW combined with SEM, EPMA, and fracture analysis, and confirmed that cracking was mainly induced by low-melting-point complex phases in interdendritic regions and grain-boundary liquid films. They further suggested that controlling heat input and optimizing pre-weld heat treatment are key measures for reducing cracking susceptibility. Sidhu et al. (2007) compared the cracking behavior of Inconel 738 under autogenous GTAW, different pre-weld heat treatments, and post-weld heat treatment (PWHT), and found that UMT pretreatment could reduce crack propagation during welding and PWHT, whereas the amount of cracking in the SHT condition was significantly higher than that in the UMT condition. Durocher and Richards (2007) deposited Inconel 625, 718, and 722 filler materials on cast Inconel 738 substrates using a low-heat-input microwelding process, explored a repair welding route for reducing HAZ damage, and evaluated the effects of deposition rate, coating quality, and substrate material on the repair performance. Soffel et al. (2022) repaired Inconel 738LC using direct metal deposition (DMD) combined with laser preheating, and demonstrated that increasing the substrate temperature significantly reduced hot-cracking density; in some specimens, no cracks were detected. Their work provides a useful reference for thermal-stress control during the repair of nickel-based superalloys. Singh et al. (2024) compared conventional TIG, Double-TIG, and SiO₂-assisted A-TIG welding of Hastelloy C-276 and reported that A-TIG produced greater weld penetration. Balos et al. (2026) performed SiO₂-assisted GTAW on Inconel 718 nickel-based alloy and achieved full penetration at a relatively low welding current, confirming that SiO₂ can significantly increase the penetration depth in TIG welding of nickel-based alloys. More recently, Tong et al. (2025) investigated autogenous GTAW of LPBF-fabricated ZGH4142 nickel-based superalloy and demonstrated that heat input significantly affected weld-pool geometry, cooling conditions, microstructural evolution, and the associated strengthening mechanisms. In a subsequent study using electron beam welding, Tong et al. (2026) further showed that variations in heat input markedly altered the weld-pool profile, cooling rate, dendritic morphology, carbide evolution, and mechanical response of LPBF-fabricated ZGH4142 alloy. Nevertheless, existing studies have mainly focused on the optimization of heat-treatment schedules and the control of low-heat-input repair processes for Inconel 738LC nickel-based superalloy, while research on the application of activated flux-assisted TIG welding in the repair of this alloy remains relatively limited.

In this study, Inconel 738LC nickel-based superalloy was selected as the research material, and TiO₂ powder was used as the activating flux. A-TIG welding experiments were carried out using different coating concentrations. Combined with optical microscopy, scanning electron microscopy, and real-time arc imaging, the effects of flux concentration on weld geometry, molten pool morphology, and HAZ dimensions were systematically analyzed. Furthermore, the possible mechanisms underlying the concentration-dependent changes in weld penetration and molten-pool behavior were discussed based on arc imaging, compositional characterization, and high-temperature surface-tension measurements. This study is expected to provide a theoretical basis and process reference for the efficient and low-defect repair welding of Inconel 738LC nickel-based superalloy components.

2 Materials and methods

2.1 Sample preparation and experimental parameters

Analytical-grade TiO₂ powder was used as the activating flux in this study and was purchased from Zhongke Keyou Co., Ltd. Anhydrous acetone was used as the dispersing solvent. The weighed TiO₂ powder was added to the acetone solution and magnetically stirred for 10 min to ensure complete dispersion, thereby obtaining a homogeneous and stable flux suspension. Four different concentrations, namely, 0.33, 0.50, 1.00, and 2.00 g/mL, were prepared. The as-prepared TiO₂ suspensions were uniformly applied using a soft brush to a band-shaped region centered along the welding path. The coating width was fixed at 10 mm for all specimens. To quantitatively characterize the actual amount of TiO₂ remaining on the substrate after solvent evaporation, the dry TiO₂ loading was determined gravimetrically. The specimens were weighed before coating and again after complete evaporation of acetone using an analytical balance. The mass difference was divided by the known coated surface area to obtain the dry TiO₂ loading in mg/cm². After coating, the specimens were dried in a ventilated environment at 25 °C ± 2 °C until two consecutive mass measurements taken 30 min apart differed by less than 1 mg. Welding was performed only after this criterion was met. After welding and cooling to room temperature, loose TiO₂-containing residue was removed by gentle brushing followed by acetone wiping, and the specimens were dried before subsequent characterization.

Three independently coated specimens were measured for each suspension concentration, and the results are reported as mean ± standard deviation. The measured dry TiO₂ loadings corresponding to suspension concentrations of 0.33, 0.50, 1.00, and 2.00 g/mL were 3.2 ± 0.3, 4.8 ± 0.4, 9.5 ± 0.6, and 18.6 ± 1.2 mg/cm², respectively. These experimentally measured values were used to evaluate the relationship between the amount of activating flux deposited on the specimen surface and the resulting welding behavior.

After coating, the specimens were placed in a ventilated environment and naturally dried by taking advantage of the high volatility of acetone. After complete evaporation of the solvent, a dense and uniform solid TiO₂ film formed on the weld surface. The treated plates were fixed on the welding workbench using a rigid fixture to prevent displacement or deformation during welding. A-TIG welding was performed under direct current electrode negative polarity. High-purity argon was used throughout the welding process for both front-side and back-side shielding to avoid the influence of air oxidation on the arc behavior and weld microstructure. The schematic procedures are shown in Figures 1A–D. Figure 1B illustrates the arc_column constriction mechanism, and Figure 1D presents two typical forms of Marangoni convection. The chemical composition difference between the alloy used in this study and standard Inconel 738LC is listed in Table 1, and the welding parameters adopted in this study are given in Table 2.

TABLE 1. Comparison of chemical compositions between the alloy used in this study and standard inconel 738LC alloy.

TABLE 2. Welding parameters adopted in this study.

To ensure experimental reproducibility and avoid pseudoreplication, three independent welding experiments were performed for each TiO₂ concentration, including the flux-free condition and concentrations of 0.33, 0.50, 1.00, and 2.00 g/mL. For each independent replicate, the TiO₂ suspension was separately applied to the specimen, followed by independent specimen mounting and welding under identical process parameters.

2.2 Arc image acquisition and analysis

To further investigate the effect of TiO₂ activating flux on arc behavior during A-TIG welding, a high-speed camera system (EXSY-2000-P1) with a frame rate of 2000 fps was used to record the arc morphology in real time. The camera was positioned at the side of the welded specimen and kept perpendicular to the welding direction to obtain stable and clear arc-profile images. To avoid interference from intense arc light, protective glass and optical filters were placed in front of the camera.

For arc characterization, arc images were acquired from three independently produced welds for each condition. Multiple frames within the stable welding period were analyzed for each weld, and the measurements obtained from these frames were first averaged to obtain a single weld-level value. The three independent weld-level values were subsequently used to calculate the group mean and standard deviation. During all welding experiments, the nominal travel speed was set to 1.10 mm/s, while the welding current, arc length, and shielding-gas flow rate were maintained at the prescribed values. Owing to small variations in the actual motion of the welding carriage during individual experiments, the measured effective travel speed ranged from 1.07 to 1.20 mm/s. Therefore, the actual travel speed of each weld, rather than the nominal set value, was used to calculate the corresponding heat input reported in Table 3. Image-processing software was subsequently used to quantitatively calculate the arc column diameter and arc-profile area, thereby quantitatively characterizing the concentration-dependent changes in the visible arc morphology.

TABLE 3. Weld formation characterization at different TiO₂ concentrations.

2.3 Preparation of weld cross sections and microstructural characterization

After welding, transverse cross-sectional specimens were cut perpendicular to the welding direction to analyze the weld formation characteristics. The sectioned specimens were sequentially mounted, ground with abrasive papers, mechanically polished, and chemically etched to clearly reveal the boundaries of the weld zone, fusion zone, and heat-affected zone. Glycerol aqua regia was used as the etchant, with a composition of 15 mL HCl, 10 mL acetic acid, 10 mL glycerol, and five drops of HNO₃.

The cross-sectional morphology of the welds was observed using microscopy, including a HUVITZ HRM-300 optical microscope and a TESCAN VEGA three scanning electron microscope. Geometrical parameters, including weld penetration depth, weld width, and molten pool cross-sectional area, were measured. Transverse cross-sectional specimens were separately obtained from three independently produced welds for each condition. Multiple measurements within each cross section were treated as technical measurements and were averaged to obtain one representative value for each independent weld. Weld penetration depth, weld width, molten pool area, heat-affected-zone width, and microstructural area fractions are therefore reported as mean ± standard deviation based on three independent welds. Based on the measured penetration depth, D, and weld width, W, the depth-to-width ratio, D/W, was further calculated to evaluate the effect of the activating flux on weld formation ability.

Primary dendrite arm spacing (PDAS) and secondary dendrite arm spacing (SDAS) were quantitatively measured from calibrated metallographic/SEM images of the weld metal using ImageJ software. Three independently produced welds were analyzed for each TiO₂ condition. From each independent weld, three transverse cross sections were examined, and five representative fields were selected from each cross section.

In addition, the dendritic morphology, grain size, precipitate distribution, and defect characteristics, such as cracks, in the weld zone were further analyzed. An identical mapped area of 13 mm², covering the fusion zone and adjacent HAZ, was analyzed for each specimen. By comparing the microstructural differences between the flux-free welded specimen and the specimens welded with different TiO₂ concentrations, the influence of the activating flux on the microstructural evolution of Inconel 738LC alloy during welding was investigated.

2.4 Microhardness and mechanical property testing

To evaluate the variation in mechanical properties across different regions of the welded joints, hardness testing was performed on the weld cross sections using a microhardness tester. A standard diamond pyramidal Vickers indenter was used with an applied load of 300 gf (HV0.3, equivalent to approximately 2.94 N) and a dwell time of 10 s. Indentations were made along a line extending from the base metal on one side, through the heat-affected zone and weld center, to the base metal on the opposite side. The center-to-center spacing between successive indentations was maintained at 0.5 mm to minimize interaction between adjacent plastic deformation zones.

Meanwhile, transverse tensile specimens were extracted perpendicular to the welding direction, with the weld centerline located at the center of the gauge section. Three tensile specimens, each obtained from an independently produced weld, were tested for each condition (n = 3). The specimens had a gauge length of 25 mm, gauge width of 6 mm, and thickness of 3 mm, and were prepared in accordance with ASTM E8/E8M. Room-temperature tensile tests were performed at a crosshead speed of 0.15 mm/min. Strain was measured using a contact extensometer with a gauge length of 25 mm, while elongation after fracture was determined from the change in gauge length after rejoining the fractured specimen.

To verify the influence of TiO₂ activating flux on the surface tension of the molten pool in Inconel 738LC nickel-based superalloy, the high-temperature sessile drop method was used to measure the variation in surface tension of the alloy melt with temperature under different TiO₂ contents. Before testing, the furnace chamber was evacuated, and high-purity argon was introduced for repeated purging cycles to reduce the partial pressures of residual oxygen and water vapor in the furnace, thereby preventing oxidation of the alloy specimen from interfering with the surface tension measurements. Before testing, the chamber was evacuated to approximately 5 × 10⁻³ Pa and purged three times with 99.99% Ar. Measurements were performed under a slightly positive Ar atmosphere at 1400 °C–1600 °C with an interval of 50 °C, which was above the reported liquidus temperature of approximately 1350 °C for IN738LC. Three independent specimens were tested at each condition. After holding for 5 min at each target temperature, the stabilized droplet profile was recorded and fitted using the Young–Laplace equation to determine the surface tension. The temperature coefficient, dγ/dT, was obtained by linear regression of the measured surface-tension–temperature data.

After atmosphere replacement, the specimen was heated to above the liquidus temperature of Inconel 738LC alloy under a slightly positive pressure of high-purity argon. The specimen was held isothermally until it was completely melted and formed a stable droplet. Several isothermal testing points above the liquidus temperature were selected at fixed temperature intervals. At each temperature, after the droplet morphology became stable, droplet-profile images were captured using a high-temperature imaging system. Image-analysis software was then used to extract profile parameters, including droplet height, width, and surface curvature. Based on the axisymmetric droplet assumption and the Young–Laplace equation, the profile curve was numerically fitted to determine the surface tension, γ, of the alloy melt at different temperatures.

2.5 Investigation of the mechanism of TiO₂ flux

To further evaluate the relationships among process parameters, weld geometry, microstructural characteristics, and mechanical properties, Pearson correlation analysis was performed among TiO₂ concentration, penetration depth, depth-to-width ratio (D/W), HAZ width, equiaxed dendrite area fraction, ultimate tensile strength (UTS), and elongation after fracture. Correlation coefficients (r) range from −1 to +1, where positive and negative values indicate positive and inverse associations, respectively. The correlation analysis was performed using individual replicate values from all independently produced welds (n = 15, i.e., three replicates per each of the five TiO₂ conditions). For each weld, the corresponding values of penetration depth, D/W ratio, HAZ width, equiaxed dendrite fraction, UTS, and elongation were paired with its TiO₂ concentration. Because TiO₂ concentration is an ordered categorical factor rather than a truly continuous variable, the Pearson coefficients are reported as descriptive measures of linear association.

Flux-free specimens and TiO₂-flux-assisted specimens were selected for analysis. EDS point analysis and elemental mapping were performed on the molten pool surface, the near-surface region of the molten pool, and the heat-affected zone using scanning electron microscopy. The accelerating voltage was set to 15 kV, the working distance was maintained at 10 mm, and spectra were acquired for 60 s at each analysis position. The detector dead time was maintained within the recommended operating range, and quantitative analysis was performed using the standardless ZAF correction procedure provided by the EDS analysis software, the effective X-ray generation volume extended over the near-surface micrometer-scale region. The atomic fractions of elements in the near-surface region of the molten pool were quantitatively compared between the two groups of specimens to analyze the mechanism by which TiO₂ flux affects the welding process.

3 Results

3.1 Comparison of arc morphology

As shown in Figure 2A, the visible arc morphology changed systematically with increasing TiO₂ concentration. At 0.33 and 0.50 g/mL, the arc profile remained relatively broad and divergent, with a relatively wide arc root and luminous region. At 1.00 g/mL, the visible arc envelope and arc root became noticeably narrower, and the luminous region became more concentrated. When the concentration was further increased to 2.00 g/mL, the apparent arc diameter and luminous area increased slightly.

Combined with the quantitative results in Figure 2B, the equivalent maximum arc diameter, total luminous area, and arc-root diameter first decreased and then slightly increased with increasing TiO₂ concentration. The minimum values of all three parameters were obtained at 1.00 g/mL. These results demonstrate a concentration-dependent change in visible arc morphology, with the most concentrated luminous region being observed at 1.00 g/mL. However, these optical measurements alone do not independently establish overall plasma constriction.

3.2 Microstructural analysis of the weld and heat-affected zone

As shown in Figure 3; Table 3, the morphologies of the weld and heat-affected zone varied markedly with TiO₂ flux concentration. In the absence of TiO₂ flux, the weld exhibited a typical shallow and wide morphology, with a relatively small penetration depth and a large weld width. This indicates that, under conventional GTAW conditions, the arc heat was mainly concentrated on the surface of the workpiece, resulting in a limited ability of the molten pool to develop in the depth direction.

The penetration depth was significantly affected by TiO₂ concentration (one-way ANOVA, p < 0.05). The flux-free specimen exhibited a mean penetration depth of 1.02 ± 0.03 mm, whereas the maximum mean penetration depth of 1.72 ± 0.03 mm was obtained at a TiO₂ concentration of 1.00 g/mL, corresponding to an increase of approximately 68.6%. Tukey’s post hoc analysis confirmed that the penetration depth at 1.00 g/mL was significantly higher than that of the flux-free condition (p < 0.05). However, no statistically significant difference was observed between the 0.50 and 1.00 g/mL conditions (p > 0.05), although the latter exhibited the highest mean penetration depth (see Table 4 for full pairwise comparison results). When the TiO₂ concentration was further increased to 2.00 g/mL, the penetration depth decreased to 1.42 ± 0.04 mm and was significantly lower than that obtained at 1.00 g/mL (p < 0.05). This result demonstrates that, within the investigated concentration range, TiO₂ flux significantly improved both weld penetration and molten pool area.

TABLE 4. Tukey’s HSD Post-Hoc Multiple Comparisons of Weld Penetration Depth among Different TiO₂. Values were obtained from three independently produced welds for each TiO₂condition. Pairwise comparisons of penetration depth were performed using Tukey’s honestly significant difference (HSD) post hoc test following one-way ANOVA.

Mean difference represents the mean penetration depth of the first group minus that of the second group. Adjusted pvalues account for multiple pairwise comparisons. Differences were considered statistically significant at p< 0.05

These results suggest that excessive TiO₂ flux does not further improve weld formation quality. Instead, an overly thick flux layer may affect arc stability and heat-transfer efficiency, hinder internal molten pool flow, and consequently lead to reduced penetration depth and unstable weld formation. Combined with the welding speed and heat-input data, the measured travel speed and calculated heat input exhibited limited run-to-run fluctuations around the nominal welding condition. Therefore, the variations in weld penetration and molten pool morphology were not simply caused by an increase in heat input, but were mainly associated with the modification of arc behavior and molten pool flow mode induced by TiO₂ flux.

It is worth noting that the nominal welding travel speed was maintained constant for all TiO₂ concentration conditions in order to isolate the effect of activating-flux concentration. The values reported in Table 3 represent the experimentally determined average travel speeds rather than different preset welding speeds. The actual travel speed was calculated from the measured welding distance and corresponding travel time for each independent welding experiment. Therefore, small run-to-run deviations from the nominal travel speed were observed. The experimentally measured travel speed and calculated heat input are reported as mean ± standard deviation. Differences among the TiO₂ concentration groups were evaluated using one-way ANOVA. No statistically significant difference in the measured travel speed was detected among the five welding conditions (one-way ANOVA, p = 0.128), and the corresponding calculated heat input also showed no statistically significant difference (p = 0.111).

As shown in Figure 4, observation of the heat-affected zone in the Inconel 738LC alloy showed that the HAZ width was 341.66 ± 11.29 μm at a TiO₂ concentration of 1 g/mL, whereas that of the specimen without TiO₂ flux was 433.47 ± 12.19 μm, corresponding to a reduction of 21.2%. This indicates that TiO₂ flux significantly reduced the HAZ width. The reduction in HAZ width is consistent with the more concentrated visible arc morphology and enhanced downward heat transport associated with the modified molten-pool flow. Because the present optical measurements do not directly quantify plasma temperature or current-density distribution, the narrower HAZ is interpreted as a consequence of altered heat redistribution rather than as direct evidence of overall arc constriction.

Inconel 738LC alloy is susceptible to liquation cracking. This susceptibility is mainly associated with the formation of liquid films in the interdendritic regions of the HAZ. When these liquid films interact with the shrinkage stress generated by the rapid precipitation of the γ′ phase during cooling, cracks may form. As shown in Figure 5A, although the specimen welded with 1 g/mL TiO₂ exhibited the best performance in terms of penetration depth and microstructural refinement, liquation cracking was not completely eliminated. A continuous crack was observed, originating from the HAZ at the bottom of the molten pool and extending along the weld centerline toward the middle of the molten pool. The measured crack length was 857 μm. High-magnification observations showed that the crack propagated along the columnar dendrites near the weld edge and eventually extended toward the weld pool center.

The crack originated from the interdendritic region, where signs of liquation were observed at the crack initiation site. A γ′-depleted region was present near the crack, indicating that local liquation had occurred in this area, causing γ′ dissolution and the subsequent formation of a γ′-depleted zone after resolidification. SEM observation and EDS analysis revealed the presence of fine Nb- and Ti-rich particles near the crack propagation path. Nb is an element with a strong segregation tendency and a partition coefficient lower than unity; therefore, it is readily rejected into the interdendritic liquid phase during solidification.

Combined with the high-magnification SEM morphology in Figure 5B, the initiation and propagation mechanism of liquation cracking in welded Inconel 738LC can be further clarified. A distinct γ′-depleted region was observed at the crack initiation site, directly confirming that local high-temperature liquation occurred in the interdendritic region of the HAZ. During the welding thermal cycle, the peak temperature exceeded the dissolution temperature of the γ′ phase, causing the intragranular γ′ strengthening phase to dissolve completely into the liquid phase. During cooling, the weld metal resolidified, and solute elements were rapidly rejected into the interdendritic regions. The γ′ phase could not reprecipitate in time, eventually forming a γ′-depleted zone. The plasticity of this region was significantly deteriorated, making it a preferential weak path for crack initiation.

As shown in Figure 5C, EDS analysis showed pronounced enrichment of Nb and Ti in the particles adjacent to the crack path. These features are therefore referred to here as Nb- and Ti-rich particles or interdendritic constituents. Although their composition is consistent with enrichment of carbide-forming elements, EDS alone is insufficient for definitive identification as carbides. These features are therefore conservatively described as Nb- and Ti-rich interdendritic constituents. Because Nb has a partition coefficient lower than unity, it readily undergoes microsegregation during solidification and continuously enriches in the interdendritic liquid phase. It then combines with C and Ti to form carbides, which are continuously distributed along the grain boundaries. The continuous carbide films weaken grain-boundary cohesion and further reduce the ability of the grain boundaries to accommodate plastic deformation. Meanwhile, the solidification shrinkage tensile stress caused by rapid γ′ precipitation continuously acts on the interfaces between the interdendritic liquid films and carbides. The coupled effect of these factors promotes stable crack propagation along the columnar dendrite boundaries, ultimately forming a through-type liquation crack with a total length of 857 μm.

To quantitatively evaluate the effect of TiO₂ concentration on liquation cracking, the number of cracks, crack number density, total crack length per unit analyzed area, and maximum crack length were determined from three independently produced welds for each condition. As shown in Figure 6; Table 5, the flux-free condition exhibited the highest overall cracking severity, with 11.3 ± 1.5 cracks, a crack number density of 0.86 ± 0.12 cracks/mm², and a crack length density of 0.94 ± 0.10 mm/mm². With increasing TiO₂ concentration, all three parameters progressively decreased and reached their minimum values at 1.00 g/mL, where the number of cracks, crack number density, and crack length density were 5.3 ± 0.6, 0.40 ± 0.05 cracks/mm², and 0.46 ± 0.05 mm/mm², respectively. The quantitative crack analysis demonstrates that TiO₂ addition reduced the severity of liquation cracking within an appropriate concentration range.

TABLE 5. Quantitative characterization of liquation cracking in IN738LC welds at different TiO₂ concentrations.

It should be emphasized that the term optimal is not used here to imply a formal mathematical multi-objective optimization. Rather, the different TiO₂ concentrations were comparatively evaluated by considering weld penetration, D/W ratio, HAZ width, liquation-cracking susceptibility, tensile strength, elongation, and arc behavior. Among the investigated concentrations, 1.00 g/mL exhibited the most favorable overall combination of these responses, including the highest penetration depth, reduced HAZ width and cracking severity, and the highest UTS and elongation. Nevertheless, liquation cracking was not completely eliminated.

The solidification microstructure plays an important role in determining the final quality and performance of the weld; therefore, it is necessary to investigate its characteristics. The microstructures in the weld pool regions of the optimal flux-assisted specimen and the flux-free specimen were compared. As shown in Figure 7A, significant differences were observed in the morphology and distribution of the solidification structures. In flux-free GTAW, long and extended columnar dendrites occupied most of the weld pool region, whereas the equiaxed dendrite region was relatively small. In contrast, when TiO₂ flux was applied, the columnar dendrite region was clearly suppressed, while the equiaxed dendrite region was significantly expanded.

For each specimen, the total area of the weld pool and the area covered by columnar dendrites were accurately measured. The area fractions of the columnar dendrite and equiaxed dendrite regions relative to the entire weld pool area were then calculated, and the results are listed in Table 6. The data indicate that an appropriate amount of TiO₂ flux can suppress the growth of columnar dendrites and promote the expansion of the equiaxed dendrite region. In addition to the area fractions of columnar and equiaxed dendritic regions, the degree of dendrite refinement was quantitatively evaluated using primary dendrite arm spacing (PDAS) and secondary dendrite arm spacing (SDAS). Both parameters showed a clear dependence on TiO₂ concentration.

TABLE 6. Quantitative characteristics of dendritic microstructures under different TiO₂ concentrations.

The microhardness values in the two regions were also measured, and the results are shown in Figure 7B. Overall, the hardness in the weld center region was higher. As the measurement position extended from the weld center toward the fusion line, the hardness gradually decreased and reached a minimum in the heat-affected zone. A slight increase in hardness was then observed near the base metal side.

3.3 Analysis of weld mechanical properties

Figure 8 shows the EDS spot-scan spectra of the weld pool surface under different TiO₂ activating flux concentrations in region A, together with the tensile mechanical properties of the welded joints in region B. As indicated by the stress–strain curves of the five groups of specimens and the dual-axis bar chart of ultimate tensile strength (UTS) and elongation after fracture, no obvious yielding behavior was observed in any specimen. Instead, all specimens exhibited the typical continuous work-hardening behavior of nickel-based superalloys. The variations in strength and ductility were highly consistent: both first increased with increasing TiO₂ concentration, reached their peak at 1.00 g/mL, and then decreased slightly. The optimal specimen welded with 1.00 g/mL TiO₂ exhibited a UTS of up to 810 MPa and an elongation after fracture of 10.8%, showing the best overall mechanical properties among all specimens.

The quantitative EDS results obtained from the post-weld near-surface region showed that the atomic fractions of the main alloying elements, including C, Al, Nb, Mo, Cr, Co, Ta, and W, remained relatively stable among the investigated conditions. In contrast, the detected O and Ti signals first increased and then decreased with increasing TiO₂ coating concentration, with the highest measured values occurring at 1.00 g/mL.

It should be emphasized that the EDS oxygen signal represents oxygen-containing species within the post-solidification electron-beam interaction volume of the analyzed near-surface region. The detected oxygen may originate from surface oxides, subsurface oxygen-containing phases, post-weld oxidation, and other oxygen-bearing constituents within this volume. Therefore, these EDS measurements cannot be used to quantify, or as a proxy for, the concentration of dissolved surface-active oxygen at the molten-pool free surface during welding. The EDS results are used here only to characterize the chemical composition of the analyzed post-weld near-surface region.

The tensile properties and fracture locations of the IN738LC welded joints as a function of TiO₂ concentration are summarized in Table 7. Figure 8C presents the effects of temperature and active oxygen content on the surface tension of the Inconel 738LC high-temperature melt. When the active oxygen content was lower than 0.07%, the temperature coefficient of surface tension, dγ/dT, remained negative. With a further increase in active oxygen content, dγ/dT increased rapidly at first and then decreased slightly.

TABLE 7. Tensile properties and fracture characteristics of IN738LC welded joints produced at different TiO₂ concentrations.

To quantitatively examine the relationships among welding parameters, weld geometry, microstructure, and mechanical properties, Pearson correlation coefficients were calculated for TiO₂ concentration, penetration depth, D/W ratio, HAZ width, equiaxed dendrite fraction, UTS, and elongation. The resulting correlation matrix is shown in Figure 9.

Overall, the correlation analysis supports a coordinated process, structure, property relationship in which increased weld penetration and D/W ratio are associated with a narrower HAZ and a greater equiaxed dendrite fraction, while the latter is positively associated with improved tensile strength and elongation. Given that TiO₂ concentration was treated as an ordered factor with five levels and the analysis is exploratory in nature, these correlations should be interpreted as descriptive associations rather than evidence of direct causality or statistically confirmed relationships.

4 Discussion

In a conventional flux-free molten pool, outward Marangoni convection occurs. The temperature is highest at the center of the molten pool, and the surface tension of the liquid metal decreases with increasing temperature. As a result, the surface molten metal spreads from the pool center toward the fusion lines on both sides, leading to severe lateral heat dissipation. This ultimately produces a wide and shallow weld, accompanied by a large heat-affected zone and severe thermal damage to the base metal. The EDS results showed that the total oxygen signal in the near-surface region increased and then decreased with TiO₂ concentration, peaking at 1.00 g/mL; independently, sessile-drop measurements confirmed that increasing dissolved active oxygen reverses dγ/dT in this alloy. The concordance of these two independent measurements indicates that, at the optimal concentration, sufficient oxygen was transferred to reverse the surface-tension coefficient and induce inward Marangoni convection. Note that the EDS value of 4.81 at% reflects total oxygen in the interaction volume and cannot be equated with the equilibrium active-oxygen concentration at the melt surface. As schematically illustrated in Figure 10, the high-temperature region at the center of the molten pool exhibited higher surface tension. Consequently, the surface molten metal was pulled inward by the lower-temperature metal near the pool edge, while the high-temperature molten metal sank downward along the central axis of the molten pool. This flow pattern transferred arc heat vertically into the depth of the base metal, thereby significantly increasing weld penetration. The cooled metal at the bottom then flowed upward along the fusion line, completing a closed circulation loop. This circulation is expected to enhance molten-pool mixing and may facilitate the transport of entrapped gas bubbles toward the free surface. However, porosity was not quantitatively evaluated in the present study; therefore, no direct conclusion regarding porosity reduction is drawn.

This interpretation is consistent with Yang et al. (2003), who reported that TiO₂ increased penetration mainly through modification of the surface-tension gradient, with no appreciable change in arc voltage. Xu et al. (2007) similarly showed that reversed Marangoni convection played a dominant role in determining weld shape in Ti-oxide-assisted A-TIG welding of Nimonic 263. In the present study, the reduced apparent arc diameter and luminous area at 1.00 g/mL indicate a concentration-dependent change in arc morphology. However, these optical observations alone are insufficient to quantitatively confirm plasma constriction. Therefore, Marangoni convection is considered the primary mechanism, while changes in arc-energy distribution may provide an additional contribution.

The flux-free weld exhibited larger PDAS and SDAS values and a predominantly columnar solidification structure. With the addition of 1.00 g/mL TiO₂, both dendrite spacings decreased and the equiaxed dendrite fraction increased, indicating a refined solidification structure. This microstructural refinement, together with the reduced HAZ width and altered molten-pool flow, was associated with the higher UTS and elongation observed at 1.00 g/mL. Nevertheless, liquation cracking was not completely eliminated, indicating that microstructural refinement alone was insufficient to fully suppress cracking and that interdendritic liquation and solidification-related tensile stresses remained operative.

5 Conclusion

• TiO₂ flux exerted a significant regulatory effect on weld geometry, the 1.00 g/mL condition exhibited the most favorable combination of the measured responses.
• Liquation cracks in welded Inconel 738LC alloy originated from the interdendritic regions. Their formation was induced by both low-melting-point liquid films formed by Nb/Ti segregation during interdendritic liquation and solidification shrinkage tensile stress.
• TiO₂ flux transferred active oxygen to the surface layer of the molten pool through high-temperature decomposition. The atomic fractions of O and Ti in the molten pool surface layer first increased and then decreased with increasing flux coating concentration, reaching their maximum values at 1.00 g/mL.
• TiO₂-derived active oxygen altered the surface-tension gradient, promoting inward Marangoni convection and downward heat transfer, which is considered the primary mechanism for enhanced penetration. Changes in arc morphology may also contribute to heat redistribution, although overall plasma constriction was not independently confirmed.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Rajendran Chinnasamy, Sri Krishna College of Engineering and Technology, India

Reviewed by: Sanjeev Kumar Gupta, GLA University, India

Ahmed Mohammed Abbas, Northern Technical University (NTU), Iraq

Yaxing Tong, Harbin Institute of Technology, China