1. Introduction
In the welding process, the formation and presence of welding slag are crucial because it not only protects the weld but also improves the welding quality [
1]. The welding slag covers the weld during the welding process, forming a layer of silicate melt that effectively prevents the intrusion of oxygen, nitrogen, and other harmful gases, thus protecting the weld metal from oxidation or nitridation [
2]. Additionally, the presence of slag can absorb impurities and inclusions generated during welding, reducing defects in the weld and enhancing its mechanical properties and corrosion resistance [
3,
4]. Currently, the evaluation of the welding slag quality mainly considers its chemical composition, fluidity, viscosity, cooling rate, and compatibility with the base material [
5,
6,
7]. Therefore, optimizing the performance of welding slag at different service stages is essential to ensure the welding quality. To optimize the various properties of welding slag mentioned earlier, the composition and structure of the slag are crucial aspects. Consequently, research on the elemental composition and doping modification of flux-cored welding wires has become a key focus in this area [
8,
9].
Studies have shown that the addition of TiO
2 and Al
2O
3 can significantly improve the fluidity and slag removal performance of welding slag [
10]. The introduction of TiO
2 can significantly reduce the viscosity, melting temperature, and surface tension of the slag, thereby enhancing its fluidity [
11]. Conversely, an increase in Al
2O
3 will initially slightly increase the viscosity, melting temperature, and surface tension of the slag, but then significantly decrease them [
12]. Additionally, the composite phase of BaO and TiO
2 occupies a significant proportion in the slag, playing an important role in slag formation and removal [
9]. By controlling the ratios of TiO
2/CaO and SiO
2/CaO, better slag removal can be achieved [
13].
In summary, optimizing the service characteristics of welding slag essentially involves the regulation and optimization of the silicate melt structure under high-temperature conditions during welding. Therefore, returning to the essence of materials research, the introduction of doping elements will become the core variable affecting the various properties of welding slag, including the viscosity of the silicate melt, the composition of the short-range ordered structure in the melt, and the integrity of the glass network structure. Existing research has proven that the most basic glass network structure in silicate melts is formed by [SiO
4] tetrahedra connected by shared corners [
14]. Since the angles between any two connected [SiO
4] tetrahedra are random, the structure of a silicate melt is characterized by short-range order and long-range disorder, typical of amorphous structures [
15]. Analyzing the phenomenon in which the introduction of Al
2O
3 can increase the viscosity of welding slag [
16], it has been found that Al
2O
3 does not exist in the form of Al
3+ ions under high-temperature conditions during welding but is distributed in the slag matrix in the form of [AlO
4] tetrahedra [
17]. The widely existing [SiO
4] tetrahedra in the slag can structurally connect with the [AlO
4] tetrahedra through bridging oxygen bonds to form a complete silicate glass network structure. Therefore, the increase in the welding slag viscosity due to Al
2O
3 can be attributed to the repair process of the glass network structure by [AlO
4] tetrahedra. Conversely, due to their larger ionic radii and coordination numbers, CaO and BaO mainly exist in the form of [CaO
6] or [BaO
6] octahedra in the glass matrix. These octahedral coordinated metal cations cannot be embedded into the glass network formed by [SiO
4] and [AlO
4] tetrahedra; thus, the doping of CaO or BaO into welding slag will lead to the destruction of the slag glass network, reducing the viscosity and ensuring good slag removal.
Furthermore, rare earth elements, known as “industrial vitamins”, exhibit similar effects on metallurgical slag in the steel industry [
18]. For example, the introduction of La
2O
3 not only significantly reduces the viscosity of metallurgical slag but also promotes the modification of MnS [
19], causing the originally “sharp-edged” MnS to undergo spheroidization, reducing the stress on the surrounding structure, and optimizing the mechanical properties of the steel. However, the high costs of raw materials for rare earth elements greatly limit their large-scale application in the steel industry, despite their ability to optimize various properties of metallurgical slag [
20].
For flux-cored welding wires, the selection, dosage, and added value of rare earth elements do not significantly affect the production cost. Therefore, developing rare earth-doped flux-cored welding wires is of great significance for the development of high-performance welding processes. Understanding the impacts and mechanisms of rare earth oxide doping is crucial in effectively optimizing the properties of welding slag and improving the weld quality. This study aims to use modern characterization tools to investigate the effects of the La2O3 content on the distribution of short-range ordered structures, degree of polymerization (DOP), phase composition, and thermal stability of welding slag.
2. Materials and Methods
As this study involves tracking the concentrations of rare earth oxides in the melt, analytically pure Na
2O, K
2O, B
2O
3, CaO, MgO, Al
2O
3, SiO
2, and La
2O
3 were used as the primary raw materials for the preparation of the welding flux core in a moisture-free atmosphere (
Table 1). After grinding the above ingredients using a 200-mesh sieve, 200 g of the gradient was weighed and milled for 24 h. The powder obtained after the ball milling procedure was divided into two parts. The first part was placed in a platinum crucible and heated to 1200 °C using an electric heating furnace. The melt was cast into a stainless-steel plate preheated at 600 °C to simulate the protection of the flux core from room temperature to the pyrolysis temperature and the subsequent fast cooling.
A differential scanning calorimeter (DSC NETZSCH 449, NETZSCH, Selb, Germany) was used to detect the thermal effect of the welding slag prepared from the oxide mixture. Based on the nucleation (Tg) and crystallization temperature (Tp) obtained from the DSC experiment, the powder was placed in the electric furnace and heated to the corresponding temperature. It was then directly poured into DI water for quenching to obtain the phase transition products of the raw material at different temperatures.
The samples obtained after quenching were fully ground. XRD (Panalytical Powder, Malvern Panalytical, Malvern, United Kingdom) was used to characterize the existing crystal phases of the welding flux core at different temperatures. Molecular vibrational spectroscopy (Raman, HR800, Horiba Jobin Yvon, Palaiseau, France) and scanning electron microscopy (ZEISS Suppra55, Carl Zeiss AG, Oberkochen, Germany) were used to obtain structural information, such as the short-range ordered structure distribution, DOP, microstructure morphology, and elemental distribution of the welding slag. EBSD technology (NordlysNano EBSD) was used to identify the crystalline phases and their distribution in the welding slag at the submicron to micron scale.
3. Results
3.1. Effects of La2O3 Doping on Thermal Effect of Welding Flux Core
As shown in
Figure 1, La
2O
3 doping (and its concentration) modified the thermal effect of the welding slag. Significant effects were observed on the softening (Tg) and crystallization temperature (Tp). As the content of La
2O
3 increased, the Tg and Tp of the welding slag increased significantly: the Tg increased from 645 °C (undoped samples) to 667 °C (with 5 wt.% of La
2O
3) and the Tp increased from 900 °C (undoped samples) to 946 °C (with 5 wt.% of La
2O
3) as the content of La
2O
3 increased. As demonstrated, the effects of the DOP on the Tg and Tp of the welding slag were significant.
3.2. Effects of La2O3 Doping on Polymer Network Structure of Welding Slag
Figure 2 shows the Raman profiles of the welding slag with La
2O
3 addition. In this work, the PEAKFIT software (wire 5.1) for Raman spectra was used for data processing, and the Gaussian formula was used for deconvolution processing. The definitions and relative concentrations of the corresponding network ligands are listed in
Table 2.
Doping with La
2O
3 reflects the complexity of a welding slag network modification in many ways. As reported, the total number of bands changes significantly after La
2O
3 doping [
23]. Specifically, new absorption peaks appear, while some disappear. For instance, a new absorption peak (denoted as ~352) is observed at 350 cm
−1 after doping. Additionally, the original peak at ~496 in the undoped sample is split into two sub-peaks (~496 and ~488) when La
2O
3 is added. According to
Table 2, these absorption peaks can be attributed to the vibration mode of La-O bonds in the slag network and the vibration mode of oxo-bridged oxygen bonds in fully connected tetrahedra.
Notably, the signal for the La-O bonds is found in La2O3 crystals, while the T-O-T oxo-bridges are found in feldspar crystals. Hence, the appearance of a peak at ~352 means that La-O bonds appear in the original slag network. Likewise, with the increase in the La2O3 content, the change of ~488 reveals the existence of La-O entities in the slag network in the form of hexagonal [LaO6], due to the high coordination number of [LaO6]. In contrast, this will destroy the Si-O tetrahedron in the original network due to the competition for O. Apparently, [LaO6] participates in modifying the slag network. Therefore, the appearance of more oxo-bridges in La2O3 increases the absorption peak area of ~496.
When the content of La2O3 in the welding slag reaches 1 wt.%, the peak at ~571 disappears entirely after the emergence of a new absorption peak. Meanwhile, the absorption peak intensity at ~688 and ~743 is also weakened significantly in this case. After La2O3 was introduced, the rearrangement degree of the network was modified considerably. As the content of La2O3 was further increased to 3 wt.% ~ 5 wt.%, the peak at ~688 reappeared. Meanwhile, this process was also accompanied by the enhancement of the bands and the complete disappearance of ~743. Because ~571 is related to a Q4 structure of complete polymerization, the values of ~608, ~688, and ~743 are attributed to the tetracoordinated SiO4 and AlO4, the bending vibration of oxo-bridges, and the symmetric stretching vibration of the Si-O tetrahedron. Therefore, the changes in the absorption peaks are ultimately caused by the increased concentration of oxo-bridges via La2O3 doping.
Notably, ~743 represents the vibration of the Si-O tetrahedron, so the weakening of this absorption band after modification with La2O3 is mainly caused by the depolymerization of La ions in the slag network. Additionally, when the content of La2O3 reached 5 wt.%, the newly emerged ~571 (300 cm−1 further) proved that the basic structure of La2O3 crystals appeared in the slag matrix, demonstrating that the solubility in the slag of La2O3 was 3~5 wt.%.
To clarify the effects of La doping on the DOP of the slag network, the DOP was calculated based on the average number of oxo-bridges in each Si-O tetrahedron; see Equation (1).
where
n refers to the relative content of the structural units of the network;
Qn refers to structural units with different numbers of oxo-bridges, while
n = 0~4 in a silicate slag.
The effects of La
2O
3 on the DOP of the welding slag network are shown in
Figure 3. It can be seen that, for the welding slag without La
2O
3, its DOP value was 2.2, suggesting that each SiO
4 tetrahedron in the welding slag contained an average of 2.2 oxo-bridges. After this, only 1 wt.% La
2O
3 reduced the original DOP (2.2) to 1.7. Upon increasing the La
2O
3 content further, the DOP increased again to 1.8 and 1.9, demonstrating that high content of La
2O
3 could promote the formation of oxo-bridges in the slag network, improving the DOP. Nevertheless, the samples with 3 wt.% and 5 wt.% La
2O
3 showed DOP values of 1.81 and 1.8, reinforcing the hypothesis regarding the solubility of La
2O
3 in the slag network. In this case, the excess La
2O
3 did not react with the Si-O network.
Deconvolution resulted in five signals (Q0~Q4). Q0 was identified as a typical isolated structure independent of the slag network. Even with the increase in the La2O3 content, the increase rate of the peak intensity Q0 decreased, demonstrating that the La2O3 modification of the primary slag network was rather significant.
The variation law of the Qn peak area after La2O3 addition was more complex, reflecting the deep rearrangement of the slag network structure to a certain extent. Additionally, the new absorption peaks proved the emergence of novel network structural units in the slag network (e.g., the intensity of Q3 decreases drastically after just 1 wt.% of La2O3). Similar changes have been observed in the intensity distributions of other Qn signals.
3.3. Phase Identification of Welding Slag with La2O3 Addition
X-ray diffraction analysis was carried out on the samples involved in this experiment to further verify the crystal phase composition of the welding slag after La
2O
3 was added (
Figure 4). Except in the pristine samples, no diffraction peaks were observed. Undoubtedly, the introduction of La
2O
3 seriously inhibits the nucleation and crystal growth of the welding slag.
As mentioned before, the heat input during high-temperature welding can not only cause the melting of the spacecraft components in the welding flux core but also provide energy for the formation of crystals in the slag melt. However, the welding process does not consider the insulation requirements for the preparation of traditional materials. To a certain extent, the quantities of precipitated crystals are often challenging to analyze by XRD.
To better elucidate the crystal precipitation mechanism of LaTi
2O
3-doped welding slag, the heat treatment time of the components was extended to 40 h, and the XRD patterns were registered again (
Figure 5). The crystallinity of the welding slag was significantly improved following the long-term heat treatment. For the samples containing 0~1 wt.% La
2O
3, the sharp diffraction peaks proved that all of the crystalline phases in the samples were diopside crystals with high crystallinity. When the content of La
2O
3 increased to 3 wt.%, new diffraction peaks were observed. After calibration, it was found that the newly emerged phase was feldspar: sodium-containing (albite, NaAlSi
3O
8) and calcium-containing (anorthite, CaAl
2Si
2O
8) feldspar.
No calcium feldspar or sodium feldspar is known in the pure phase in nature. Because the Na+ and Ca2+ in the feldspar structure could form continuous solid solutions, they could be identified as a single feldspar phase. The only difference was the degree of substitution of Na+ and Ca2+. Additionally, the spectra of the samples with 3 wt.% La2O3 showed a diffraction peak, indicating that La2O3 had solubility of ~3wt.% in the CMAS system adopted in this study.
The previously precipitated feldspar suggested that the modification with La2O3 could further modify the connectivity of the slag network and promote its evolution from the chain-type to a network-type silt-like structure. As for samples with 5 wt.% La2O3 (although La2O3 had reached saturation), the precipitation of CaLa4(SiO4)3O, visible in its XRD pattern, further proves that La doping can effectively promote the precipitation of crystalline nanoparticles. Additionally, diopside with different content of La2O3 reveals that La2O3 influences the crystal cell parameters, showing a shift in the reflections to a higher angle.
3.4. Influence on Microstructure Evolution of Welding Slag with La2O3 Addition
Due to the absence of crystal formation in samples containing La
2O
3 after conventional treatment, a 40 h heat treatment was applied to further research the crystallization products following La
2O
3 addition.
Figure 6 shows the microstructure of the welding slag after 40 h of heat treatment. To better reflect the distribution of the phases containing La
2O
3, the BSE mode was adopted when registering the SEM micrographs. If the samples contained 0~1 wt.% La
2O
3, there were two distinct crystal phases in the welding slag (
Figure 6A,B). According to the XRD results (
Figure 5), the area with low contrast was diopside, while the bright crystals were La
2O
3. Due to the relatively small proportion of this phase in the matrix, the diffraction information of La
2O
3 was not significant. Compared with the bare samples, the size of the diopside crystals was significantly smaller when 1 wt.% La
2O
3 was added. This phenomenon could be related to the lower crystal growth rate due to the inhibition of La
2O
3 diffusion. When the content of La
2O
3 was increased to 3 wt.% and 5 wt.%, the diopside crystals were aggregated. This demonstrates that the nucleation rate of the diopside crystals was further inhibited.
Meanwhile, the appearance of La
2O
3 and feldspar crystals in the slag matrix was consistent with the XRD results. Notably, the microstructure of the welding slag was characterized via BSE, and the contrast was directly related to the average atomic number of the containing elements. Therefore, with the increase in the La
2O
3 content, the gradual growth of the residual slag phase occurred, as shown in
Figure 6. Additionally, the samples with the slag phases containing 3 wt.% and 5 wt.% La
2O
3 were relatively close in contrast, demonstrating that the solubility of La
2O
3 in the slag matrix was approximately 3 wt.%.
3.5. Characterization of Welding Slag Samples with 5 wt.% La2O3 by EBSD
To further study the specific crystal phases and their distribution in the welding slag samples with 5 wt.% La
2O
3 (after crystallization), EBSD technology was used (
Figure 7). The slag matrix showed a black appearance in the BC diagram due to the absence of a basic long-range ordered structure in the slag phase. As shown in
Figure 7, diopside (red), CaF
2 (blue), and feldspar (yellow) were the main phases in this area.
4. Discussion
The diffusion process often controls the phase transition process in silicate slags. However, to enhance the slag removal performance, the welding slag needs to have a specific thermal mismatch with the welded metal, and the wettability of the slag–gold interface should be as low as possible. This requires structural modifications that can comprehensively control the viscosity, thermal stability, mechanical performance, and crystallization ability of the silicate solution. Conventionally, it is believed that rare earth–metal oxides reduce the slag viscosity by disturbing the silicate slag network. By reducing the viscosity, the diffusion process is facilitated, thus promoting the nucleation of the welding slag and crystal growth. In this case, the fluidity of the welding slag will be increased, including the welding quality.
Nevertheless, introducing La
2O
3 leads to the severely inhibited crystallization of the welding slag, which is inconsistent with previous studies. To further investigate the reason behind this phenomenon, Raman spectroscopy was applied (
Figure 2). It was found that the added La
2O
3 could react with the primary slag network. Even 1 wt.% of La
2O
3 will cause significant structural rearrangements in the slag by interfering with the connectivity of the welding slag network. The newly emerged absorption peaks at ~352 and ~488 indicate that La
2O
3 exists in the slag matrix in the form of [LaO
6] octahedra. In addition, there are many oxygen-containing bonds (except oxo-bridges) in the slag matrix, which leads to the depolymerization of the slag network structure.
The introduction of La
2O
3 into slag leads to an increase in the oxo-bridges in the slag network. This process promotes the transition of the slag structure from chain-like to network-like. It promotes the precipitation of silicate crystals, such as calcium and sodium feldspars. According to the Raman data (
Figure 2), the peak area at ~437, belonging to the depolymerized structure, is significantly increased in samples with 1 wt.% La
2O
3. This is due to the high coordination number of La ions. To meet the requirement for their coordination number when producing [LaO
6], La ions must compete for O with the Si-O network.
Therefore, the original slag network will be depolymerized when the La ions take away the O in the Si-O network. As a result, the peak area at ~437 increases significantly after La doping. Additionally, the peak area at ~571 in samples with 1 wt.% La2O3 is also considerably reduced. Since ~571 represents the Q4 structure in the state of complete polymerization, its decrease (with the increase in the content of La2O3) further proves that the depolymerization effect of La ions is rather significant. If the La2O3 content is further increased to 3 ~ 5 wt.%, the area at ~571 will increase.
Undoubtedly, an excess of La leads to an increase in oxo-bridges. However, the reason for the abnormal changes in the ~571 peak area after La2O3 addition remains to be further investigated. As mentioned above, when La ions enter the slag network, [LaO6] will be formed. The Na+ inside the slag will be gathered around the [LaO6] to ensure electroneutrality. Since alkaline metal ions such as Na+ often cause the breakage of oxo-bridges, the concentration of Na+ near [LaO6] will inevitably lead to a decrease in Na+ in other positions. There is no doubt that the preferred distribution of Na+ is the reason that the content of La2O3 and the number of oxo-bridges increase. In this case, the connectivity and viscosity of the slag matrix will increase significantly, and the diffusion resistance will increase accordingly. Therefore, it can be concluded that the excess La2O3 outside the network will prevent the welding slag from achieving the expected nucleation and crystal growth rate in the heat treatment process.
Additionally, when introducing La2O3, the original structure of the welding slag changes; thus, ~496 splits into ~496 and ~488. The welding slag structure gradually changes to the T-O-T connection mode observed in feldspars. Moreover, 1 wt.% La2O3 doping resulted in many units similar to the feldspar-like structure. This process also provides the base material for the sodium or calcium feldspars precipitated in the subsequent “crystallization” process.
To further study the effects of La
2O
3 on the precipitation of diopside crystals, the samples with 5 wt.% La
2O
3 were characterized by EBSD and “crystallized” for 40 h before the investigation.
Figure 8A–D show the SEM micrographs and EDS analysis of the samples. The Kikuchi pattern of the selected crystal shows significant monoclinic characteristics. The crystal was identified as diopside. This characterization strategy is similar to the SAED used with TEM. However, the above method lacks specific elemental data and can only confirm the crystal structure of the phase.
According to the EDS results (
Figure 8B), a significant amount of La was found, in addition to Ca, Mg, Al, Si, and O, commonly found in diopside crystals. According to the related theories in structural chemistry, the chain structure formed by the Si-O tetrahedron connected end to end by oxo-bridges is identified as Q
2, which is fixed by [CaO
6] as a single-chain structure. Since Ca
2+ and La
3+ ions have a very close ionic radius and can form [MO
6] octahedra, most researchers tend to believe that La mainly replaces Ca in diopside crystals.
Nevertheless, La can exist in a single slag matrix in [LaO
6] forms. From the perspective of structural chemistry, [SiO
4] and [CaO
6] in diopsides can be replaced by [LaO
6]. Given the valence difference between Ca and La ions (La-O partially replaces [CaO
6] and [SiO
4] since the ionic radius of Ca
2+ is 0.1 nm and is smaller than that of La
3+), Na
+ will accumulate near La-O bonds to ensure electroneutrality. The lack of Na
+ from the network damages the oxo-bridges and further increases the DOP. The structural parameters of the diopside crystal after solid solution by La are lower than those of the pure phase. This explains why the dominant diffraction peaks of diopside crystals shift towards higher 2θ angles as the La
2O
3 content increases. According to the XRD results (
Figure 5), 1 wt.% La
2O
3 has a relatively weak effect on the pattern of diopsides. After increasing the La
2O
3 content to 3 wt.%, a significant amount of feldspar was identified, demonstrating that the complete connection between Si-O and Al-O in the form of tetrahedra is the main prerequisite for the formation of crystals.
It has been demonstrated that the coordination number of La ions in silica slag is higher than that of other group elements. La ions will break the existing Si-O bonds to satisfy the high coordination number of La ions. Additionally, due to the differences in the valence states of La, Si, and Al, Na tends to be preferentially located near La-O bonds to neutralize the remaining negative charges. However, Na+ can destroy the Si-O or Al-O network by generating defects similar to Si-O-Na in the slag network. Hence, once the distribution of Na+ is preferential, the slag structure ratio at a high DOP is bound to increase. The difficulty of nucleation and crystal growth in the slag network with a high DOP will increase due to its low diffusion coefficient. However, with the continued increase in the content of La2O3, the original amount of Na+ failed to balance the negative charges. At this time, alkaline earth metal ligands with similar effects will be distributed near La-O ligands instead of Na+. Additionally, there will be regions with high concentrations of Ca, Na, Al, and Si species in the slag network, meeting the requirements for feldspar crystal formation. Therefore, feldspar crystals may exist in samples modified with a large amount of La2O3.
In addition to the phases mentioned above, a significant amount of (CaLa4(SiO4)3O) was also found in the XRD patterns of samples containing 5 wt.% La2O3. This phenomenon could be related to the lower La content in the previous samples. EBSD was used to conduct an in-depth analysis of (CaLa4(SiO4)3O). The influence of La2O3 on the final crystalline phases formed from the CMAS welding slag is mainly realized by modifying the structure of the slag network and the preferential distribution of Na+ and Ca2+. However, the network structure of the welding slag can yield a large number of crystalline nanoparticles after modification with 1~3 wt.% La2O3. The lack of La can lead to the precipitation of (CaLa4(SiO4)3O) in the samples, which occurred when 5 wt.% La2O3 was applied, so only (CaLa4(SiO4)3O) crystals could appear in samples with 5 wt.% La2O3.
In summary, the introduction of La2O3 can lead to an increase in the silicone melt viscosity and a decrease in the ability to achieve crystal precipitation. Even if sufficient amounts of feldspar crystals can be precipitated via short-term heat treatment, the substantial thermal mismatch between the slag matrix/feldspar and weld metals will further promote the slag removal process at a later stage. Additionally, the increase in melt viscosity mentioned above is due to the combined effect of the increase in the content of Q4 ligands in the silica network and the enrichment of alkali metal ions around the rare earth elements. Therefore, the brittleness of the slag matrix containing rare earth elements is significantly greater than that of the welding slag under the same circumstances. Apparently, in this case, the complex and brittle welding slag matrix is prone to cracks. In this way, the slag removal performance is ensured after the final welding.