**1. Introduction**

Brazing or welding Al to steel has become a hot subject in recent years because Al–steel composite metal materials have been found to have wide applications in industry. In order to develop brazing technology, the improvement of flux is a key problem that must be solved. CsF-AlF3 eutectic flux has been widely used in brazing aluminum alloys [1] and exhibits an excellent performance in brazing dissimilar metals matched with Zn-Al filler [2,3], such as brazing Al to Al [4] and Al to Cu [5]. However, CsF-AlF3 flux is not so good when brazing aluminum to steel, and the activity of removing the oxides on Al and steel simultaneously needs to be improved.

It was reported that matching CsF-RbF-AlF3 flux and the 6061 aluminum alloy to 304 stainless steel can be executed [6]. According to Ref. [7], adding 6 mol.% of ZnCl2 enhanced the activity of CsF-AlF3 when removing the alumina during brazing. It is well known that the Cl- ion is a harmful ion to aluminum and steel, so developing a novel flux without a Cl- ion as an activator represents meaningful work.

In this paper, by adding a micro-nano Ga2O3 particle into an CsF-RbF-AlF3 flux as an activator, the spreadability and wettability of Zn-Al filler metal on the aluminum, along with steel, were investigated. XDR analysis and the reaction mechanism were also studied. The results would be helpful for brazing AA5052 to Q235 steel, without the risk of Cl- ion corrosion.

## **2. Materials and Methods**

The experimental base metals used for spreading tests were commercially supplied AA5052 and Q235 steel (Changshu Huayin Filler Metals Co., Ltd., Suzhou, China) with dimensions of 50 mm × 50 mm × 3 mm. The filler metal was a 2 mm diameter commercial Zn-2Al alloy wire (Zhejiang Xinrui Welding Materials Co., Ltd., Shengzhou, China). The flux used in this research was prepared using nano Ga2O3 powder, RbF (AR purity, Shanghai Jiuyi Chemical Reagent Company, Shanghai, China), and commercial CsF-AlF3 flux (Zhejiang Xinrui Welding Materials Co., Ltd., Shengzhou, China). A total of 0.5 wt% of RbF and nano Ga2O3 powder in the range of 0.0001–0.1 wt.% were added to commercial CsF-AlF3 flux. In order to obtain homogeneous CsF-RbF-AlF3−xGa2O3 flux, accurately weighted Ga2O3 powder, RbF, and CsF-AlF3 were scattered in deionized water, together with ultrasonic concussion, for an hour, in which the weight deviations of Ga2O3 and RbF were controlled to ±1 mg. For experimenting conveniently, all specimens were pretreated: they were degreased with acetone, oxides were removed with NaOH solution for AA5052 and HCl for Q235 steel, and the specimens were then washed with deionized water and dried naturally.

The spreading test was carried out according to China's National Standard GB/T11364-2008. A total of 0.12 g filler metal was placed on base metal, covered with the prepared CsF-RbF-AlF3-Ga2O3 flux, and put into the VULCAN3-130 mu ffle furnace (Nanjing University of Aeronautics and Astronautics, Nanjing, China) holding at 600 ◦C for 60 s. To obtain accurate results, each flux was tested five times in the same conditions. The flux residue formed during the test was carefully sampled. Then, the experimental specimens were completely cleaned with hot deionized water and dried naturally. The Ga2O3 particle was tested by a Hitachi S-4800 SEM (Ningbo Institute of Materials Technology&Engineering, Ningbo, China), and by using the energy spectrum device (energy dispersion X-ray spectrometry, EDX) attached to the equipment, microstructure observation and joint composition analysis were carried out on the solder and joint samples. The spreading areas of filler metals were measured and calculated by Image Pro-plus software, and the flux residue was analyzed with a Bruker D8 XRD analyzer (Ningbo Institute of Materials Technology&Engineering, Ningbo, China).

#### **3. Results and Discussion**

#### *3.1. Morphology and Size of the Ga2O3 Micro-Nano Particle*

In order to improve the activity of CsF-RbF-AlF3 flux, the particle size of added Ga2O3 has to be small enough. A range from 100 nano-meters to several microns is perfect. The size of the Ga2O3 particle was tested by SEM, and the results are shown in Figure 1, in which the size of the particle is below 5 μm. The EDX elemental mapping of Ga2O3 particles is shown in Figure 2. It was previously reported [8] that Ga2O3 particles presenting as short rods have more importance in the integrality of the crystal structure than those presenting as long cylinders.

**Figure 1.** Morphology of Ga2O3 particles: (**a**) 10,000 times; (**b**) 20,000 times.


**Figure 2.** EDX elemental mapping of Ga2O3 particles.

#### *3.2. Application Properties of the Zn-2Al Filler Metal*

For determining the effect of the Ga2O3 particle on the spreadability and wettability of Zn-2Al filler metal, tests were carried out over 5052 aluminum and Q235 low-carbon steel with the new flux, whose spreading areas were measured and averaged. Figure 3 shows the relationship between the Ga2O3 concentration and spreading areas, and partial spreading tested graphs are presented in Figure 4.

**Figure 3.** Effect of the Ga2O3 concentration on spreading areas of Zn-2Al filler metal on base metal: (**a**) 5052 aluminum; (**b**) Q235 low-carbon steel.

Figure 3 indicates that Ga2O3 addition to the CsF-RbF-AlF3 flux clearly improved the spreadability and wettability of Zn-2Al filler metal on the base metal. It was found that the maximum spreading area of Zn-2Al on 5052 aluminum occurred when the concentration of Ga2O3 was 0.0009 wt.%. For Q235 steel, the maximum value was 0.003 wt.%. Under this circumstance, the spreading area of Zn-2Al filler metal was approximately 350 mm<sup>2</sup> over 5052 aluminum, which was an increase of 75% compared to the applications of CsF-RbF-AlF3 flux. Additionally, it was 160 mm<sup>2</sup> over Q235 steel, representing an improvement of 122%. However, the spreading area mildly decreased with Ga2O3 addition continually rising. Based on the above results, the conclusion was that the CsF-RbF-AlF3-Ga2O3 flux could be applied to brazing Q235 steel to AA5052 matched with Zn-2Al filler metal, and the additional concentration of Ga2O3 particle range was 0.0009–0.003 wt.%. In Figure 4, partial spreading images are shown. It was obvious that the spreading area of Zn-2Al filler metal was better with the Ga2O3 particle added in CsF-RbF-AlF3 flux.

**Figure 4.** Spreading tests of Zn-2Al filler metal on base metals with CsF-RbF-AlF3-xGa2O3 flux: (**a–e**) 5052 aluminum; (**f–i**) Q235 low-carbon steel.

#### *3.3. Interfacial Induction E*ff*ect of the Ga2O3 Particle*

The CsF-RbF-AlF3-Ga2O3 flux improved the spreadability and wettability of Zn-2Al filler metal over 5052 aluminum and Q235 steel obviously, but the reason for this was not clear and still needed to be investigated. Therefore, we attempted to employ the theory of interfacial tension [9] to identify some explanations. According to Young's equation, the liquid balanced over a solid surface can be determined by the relationship between the interfacial tension of the solid and liquid. In this investigation, a balance of these interfacial tensions was shown between the base metal and molten flux (σs-flux), molten filler metal and molten flux (σl-flux), and base metal and molten filler metal (σs-l).

Dezellus [10] mentioned that there were two kinds of ways in which molten metal wets a solid—non-reactive wetting and reactive wetting—and the reaction of a liquid and solid decreases their interfacial tension (σs-l) to improve spreading. Because the molten flux reacted with the surface metallic oxides over the base metal to remove it, the explanation for the interfacial reactions discussed in this research was that of reactive wetting. While balancing, the relationship between interfacial tensions is in accordance with Equation (1):

$$
\sigma\_{s-flux} = \sigma\_{l-flux} \cdot \cos \theta + \sigma\_{s-l} \tag{1}
$$

According to the investigation of Qian [11], alumina reacted with melted XF-AlF3 (X = Na, K, Rb, and Cs) flux and formed AlF3 to be dissolved, and the interfacial tension of solid–liquid (σs-l) changed from Zn-Al alloy-alumina to Zn-Al alloy-aluminum. Meanwhile, owing to the excellent intersolubility of Zn and Al [12], the interfacial tension of the AA5052 and Zn-2Al alloy (σs-l) decreased rapidly. Therefore, the spreading area of the Zn-2Al alloy over AA5052 increased obviously.

However, the surface oxide was the maximum obstacle to the brazing process for Q235 steel. No research has reported that the CsF-AlF3 series flux reacted with iron oxidation immediately. Therefore, the reason why the Zn-Al alloy spread over Q235 should be due to the reaction between the Zn-Al alloy and oxidation. The thermodynamically calculated results shown in Table 1 indicated that iron oxides reacted with Al to form simple substance Fe and dissolved [13]. The phase diagram of Fe-Zn showed that intermetallics, such as FeZn and FeZn4, formed below 600 ◦C. Precisely because of the reaction of Zn and Fe [14], the Zn-2Al filler metal spread over Q235 steel, and their interfacial tension (σ's-l) decreased along with it.


**Table 1.** Reaction Gibbs energy of iron oxidation and Al (T = 873.15 K).

Because of the interfacial reaction, molten Zn-Al filler metal wetted the AA5052 alloy and Q235 steel and spread on them, which broke the balance of Young's equation by reducing the σs-l shown in Equations (1) and (2). When the gravity was larger than the molten flux, as shown in Table 2, the Ga2O3 particle settled from molten flux freely to the molten filler metal surface. According to the result of the thermodynamic calculation shown in Table 1, it was found that the Ga2O3 particle released Ga atoms by reacting with the Zn-Al alloy, where Ga was surface active, and could enrich the surface of filler metal to decrease its interfacial tension between metals (σ's−<sup>l</sup> < <sup>σ</sup>s−l) [15]:

$$
\sigma\_{s-flux} \ll \sigma\_{l-flux} \cdot \cos \theta + \sigma\_{s-l}' \tag{2}
$$

**Table 2.** Density of the compound.


Considering the activity of molten Zn-2Al filler metal, Ga could be released and tested to demonstrate its release from Ga2O3. Due to having a similar property to Al [16], Ga permeated in the molten filler metal spontaneously. The microstructure at the interface of the joint is shown in Figure 5. There is no obvious defect in the interface, the filler metal is fully spread out on the surface of the Q235 steel, and the interface is well-combined. The reaction layers at the interface between the braze joint and base metal di ffered in thickness: the light gray layer near the braze joint and the dark gray layer near the base metal. The thickness of the interfacial IMC layer highlighted in the pictures was calculated to be 2.82 μm. The thickness of the light gray layer was 1.56 μm, and that of the dark gray layer was 1.26 μm. The EDX results in Figure 6 showed that the weak peak of Ga appeared in the interface of base metal and filler metal. With the increase of Ga2O3 addition, Ga2O3 particles displayed more activity by presenting as short rods. Owing to the decrease in σs-l, as well as the interfacial reactions, the original balance was broken and the spreading area increased obviously. A description of the variation of the interfacial tensions is shown as Figure 7. According to the Fe-Al binary phase diagram, the dark gray layer should be the θ-phase Fe-Al intermetallic compound. The θ-phase is in equilibrium with the α-Al phase in the Fe-Al phase diagram and has a monoclinic crystal structure. The atomic ratio of Al to Fe ranges from 3 to 3.5 in the θ-phase, and the crystal structure is usually represented by Fe4Al13, FeAl3.2, and FeAl3.

**Figure 5.** Microstructure at the interface of the joint: (**a**) 5052 aluminum; (**b**) Q235 low-carbon steel.

**Figure 6.** The interface of AA5052 and Zn-2Al filler metal: (**a**) SEM scanning image; (**b**) EDX results of the section.

**Figure 7.** Diagrammatic sketch of Zn-Al spreading on base metal.

#### *3.4. Analysis of the Flux Residue*

An XRD analysis experiment of the flux residue collected from the surface of AA5052 and Q235 steel was carried out. It revealed that the residue products over AA5052 were silicate, oxidation, fluoride, MgMnSi2O6, SiO2, Al2SiO5, MgSiO3, Al2O3, AlF3, MnF2, ZnGa2O4, and Cs11O3; however, they were MnAlF5, SiO2, Al2O3, MnF2, AlF3, ZnO, Al2SiO5, AlPO4, and FePO4 over Q235 steel.

According to a previous report [17], the compounds found over the AA5052 surface were mainly Mg2Si, MgO, MgAl2O4, and amorphous Al2O3 and a little Cu, Mn, Cr, and Fe. When heating to over 520 ◦C, Mg, Mn, and Si were enriched. The Cs11O3, silicate, and fluoride were from the reaction between the oxidation over the base metal and CsF-RbF-AlF3-Ga2O3 flux. However, ZnGa2O4 was identified from the residue, which certainly came from the reaction of trace amounts of Ga2O3 in the

flux. It was previously reported [18] that additional Ga in the filler metal was enriched on its surface, which meant that Ga2O3 might have been enriched. It could be speculated that the reaction equation of ZnGa2O4 can be written as follows:

$$\text{ZnO} + \text{Ga}\_2\text{O}\_3 \rightarrow \text{ZnGa}\_2\text{O}\_4\\\text{AlG} = -653.15 \text{ K}\\ \text{J/mol (T = 600 } ^\circ \text{C)}\tag{3}$$

It was reported that the mechanism through which CsF-AlF3 flux removed alumina was a process of dissolution and reaction, and the active ingredients were F-, HF, SiF6<sup>2</sup><sup>−</sup>, and Zn2<sup>+</sup> ions. In this study, a massive Zn element was present in the filler metal, with a small Si element on the base metal surface, which provided the forming conditions for Zn2<sup>+</sup> and SiF6<sup>2</sup>− ions during the spreading test. According to the basic principles of the chemical reaction [19], the major flux reaction was listed in Equation (4) [20]. Because of the imperfections of flux preparation technology, a trace amount of H2O, NH4F, and NH4AlF4 appeared in the CsF-RbF-AlF3-Ga2O3 flux, which released active HF during the experiments and accelerated the removal of alumina:

$$\text{CsAlF}\_4 \overset{\Delta}{\leftrightarrow} \text{CsF} + \text{AlF}\_3 \tag{4}$$

$$NH\_4AlF\_4 \overset{\Delta}{\leftrightarrow} NH\_4F + AlF\_3 \tag{5}$$

$$NH\_4F \xrightarrow{\Delta} NH\_3 + HF \tag{6}$$

$$2AlF\_3 + 3H\_2O \overset{\Delta}{\leftrightarrow} Al\_2O\_3 + 6HF \uparrow \tag{7}$$

However, the XRD result of the residues over Q235 steel was not the same as that for AA5052, which included MnAlF5, SiO2, AlF3, MnF2, Al2SiO5, Al2O3, ZnO, AlPO4, and FePO4. Research has shown that Fe2O3, Fe3O4, and FeO orderly cover Q235 steel [21]. Therefore, iron oxide removal was the first step for brazing conveniently. It could be found that F- and HF reacted with Mn compounds and AlF3 to form MnF2 and MnAlF5. Fe2SiO5 and Al2SiO5 were obtained from the reaction of SiO2 and iron oxidation and alumina. Nevertheless, a trifle phosphate, FePO4, and AlPO4 appeared in the XRD result. As it was shown, there was no phosphorus in the flux, whose component in Q235 was below 0.04 wt.% [22]. Therefore, FePO4 and AlPO4 were formed by the reactions between metallic oxides and phosphorus in Q235 steel as

$$2M\_2O\_3 + 4P + 5O\_2 \to 4MPO\_4 (M = Fe, Al). \tag{8}$$

Because the content of phosphorus in Q235 was below the XRD detectability, it meant that P was enriched over the Q235 surface, which is a normal phenomenon that occurs upon heating a P-contained alloy in which phosphorus is lost through burning and phosphorus oxide is formed. Although the existence of P removed a little metallic oxide, it did not remove all oxide films. Therefore, the main reaction mechanism of removing metallic oxides was the reaction between Al and iron oxidations shown in Table 1.

It was concluded that the main reactions between the CsF-RbF-AlF3-Ga2O3 flux and surface compounds on the base metal could be described as Equations (3)–(15). Due to the formation of active substances—SiF6<sup>2</sup><sup>−</sup>, HF, and F−—the molten flux cleaned up the surface oxidation over AA5052, and active Al in the filler metal and enriched P removed the iron oxidation over Q235 steel. Both of these mechanisms improved the spreadability and wettability of molten Zn-2Al filler metal on the base metal.

However, the reactions between silicate and metallic oxide shown in Equation (9) decreased the flux activity by consuming the CsF in CsF-AlF3 and inhibited its flowability. Owing to the formation of Ga, the spreadability and wettability of Zn-2Al filler metal were enhanced, despite the activity loss:

$$28\text{CsF} + 3\text{SiO}\_2 + 10\text{HF} + \text{MnO} \rightarrow 2\text{Cs}\_{11}\text{O}\_3 + 2\text{Cs}\_2\text{SiF}\_6 + 5\text{MnF}\_2 + 5\text{H}\_2\text{O} \uparrow\tag{9}$$

$$\rm Mg\_2Si + 2O\_2 \to 2MgO + SiO\_2 \tag{10}$$

$$MgAl\_2O\_4 + SiO\_2 \to MgSiO\_3 + Al\_2O\_3 \tag{11}$$

$$Al\_2O\_3 + SiO\_2 \to Al\_2SiO\_5 \tag{12}$$

$$\rm MnO + MgO + 2SiO\_2 \to MgMnSi\_2O\_6 \tag{13}$$

$$\text{Cs}\_2\text{SiF}\_6 \xrightarrow{\Delta} \text{2CsF} + \text{SiF}\_4 \uparrow \tag{14}$$

$$2MnO + 4HF + AlF\_3 \to MnF\_2 + MnAlF\_5 + 2H\_2O \uparrow \tag{15}$$

According to the proposed mechanisms, it was indicated that molten flux reacted with oxidation over base metals to promote the spread of filler metal simultaneously. However, the rapidly e ffectiveness loss of molten flux decreased the e fficiency of alumina removal and further strangled the spreading speed of filler metal, which was mainly due to the consumption of active substances—SiF6 2−, HF, and F−—and the appearance of silicates.

Compared to the lower e fficiency of CsF-RbF-AlF3 flux on iron oxidation, the Ga2O3 addition obviously improved the activity of the flux because of its enrichment and production of Ga, which helped to decrease the interfacial tension between base metal and molten filler metal. Meanwhile, Ga2O3 partially reacted with ZnO to dissolve it, in order to promote the further spreading of molten Zn-2Al filler on the base metal.
