E ﬀ ect of Oxide Scale on Hot Dip Zn-Al-Mg Alloy Coating Prepared by Reduction Combined with Induction Heating

: Hot dipping Zn-Al-Mg coatings were prepared by rapid induction heating combined with gas protection. The inﬂuence of oxide scale on the structure and surface quality of a hot-dip Zn-6Al-3Mg alloy coating was studied in this paper. The results showed that the reaction of Fe-Al was suppressed by the scale on the surface of the steel plate. When the thickness of scale was 10 µ m and the steel entry temperature was 900 ◦ C, the surface quality of the coating was good. The Zn-Al-Mg coatings mainly consisted of the ternary eutectic structure of Zn / Al / MgZn 2 and Fe 4 Al 13 at the interface. When the scale thickness was 2–3 µ m with the same steel entry temperature, the surface quality of the coating was poor, and serious stripe-like protrusion defects were formed on the surface of the coating, which was mainly caused by the Fe 4 Al 13 phase separating from the substrate / coating interface into the overlay.


Introduction
Steel undergoes corrosion when exposed to different environments. Hot-dip galvanizing and its alloy coatings are widely used in steel protection [1]. With the shortage of resources and the growing problem of environmental pollution, steel protection faces new challenges. Traditional pure Zn, Zn-Al alloy coatings are unable to meet the needs of industrial production [2,3]. Al-Si coatings and Zn-Al-Mg coatings are currently widely used in production. Al-Si coating (90% Al-10% Si) is a new type of coating product commonly used in appliances and automotive industries. New automobiles are increasingly designed using advance high strength steels in crash critical parts for side impact, roof crush, and frontal impact. The Al-10% Si coating is often applied to press-hardened steels (PHS) in order to avoid significant oxidation and subsurface decarburization of the steel blank in the hot stamping process [4,5]. The Zn-Al-Mg alloy coating has become a hot spot in anti-corrosion coating because of its excellent corrosion resistance and cutting-edge protection performance. A series of Zn-Al-Mg alloy coatings with high quality, high performance has been developed, such as Superzinc, Dymazinc, SuperDyma, and ZAM [6][7][8][9] after more than 30 years of research. Foreign Zn-Al-Mg coating products have achieved large-scale industrial production, but in China, only Shandong Kerui Steel Plate Co., Ltd. (Binzhou, Shandong, China) has launched Zn-Al-Mg coated steel sheets, and several steel plants are still in the trial production stage. However, there is a big gap between domestic products in terms of quality control and international advanced levels due to the lag in basic theory and application. Even if the performance of Zn-Al-Mg coatings is excellent, it hasn't been widely promoted and applied at present because it is mainly imported and expensive. In order to promote the application of high-performance Zn-Al-Mg alloy, it has great significance to take Zn-Al-Mg alloy coating as the focus of research.
There have been a lot of reports of Zn-Al-Mg alloy coatings, focusing on the microstructure, growth mechanism, and corrosion resistance of alloy coatings [10][11][12][13][14][15]. Although the corrosion resistance of the coating is greatly improved by the addition of Mg elements [16,17], the structure of the coating is more complicated and unstable, and the process requirements are more stringent [18]. The optimized process is very important to obtain a high-quality stable coating. However, there are few systematic studies on the hot-dip Zn-Al-Mg process in the reported literature.
Solvent method is used in the traditional hot-dip galvanizing alloy process, to obtain a clean surface that is beneficial to the combination of the alloy bath and the steel matrix by pickling the steel plate [19]. However, the hot-dip Zn-Al-Mg alloy coatings by the solvent method have problems such as difficulty in the hot dipping process and environmental pollution. Therefore, many scholars pay attention to the combination of non-acid pretreatment and hot-dip galvanizing, such as shot blast, slurry blast, and a reduction of shielding gas [20] The method of reducing a steel plate with a certain content of N 2 -H 2 mixture atmosphere is more suitable for hot-dip galvanizing [21,22], and it can be extended to continuous production line applications. Researchers have conducted a series of studies on hot dip galvanizing by the reduction method, and found that the composition of oxide scale of hot-rolled steel with different ingredients was different, leading to the great difference in the reduction process [23][24][25][26][27][28]. It is generally considered that the dense oxide scale is reduced to porous spongy iron, which can improve the wettability of the matrix [29]. Some studies have paid attention to the microstructure of coatings obtained by hot-dip Zn and Zn-Al alloys after a reduction of hot-rolled steel sheets. It has been found that the thickness, degree of reduction, and microstructure of the oxide scale affect the interface reaction of the coating. For example, the oxide scale can inhibit the formation of the Fe-Zn brittle phase in hot-dip galvanizing, while in the hot-dip Zn-Al alloy, the specific surface area of the oxide scale increases after reduction, which accelerates the consumption of Al and makes it difficult to form a Fe-Al alloy layer [22,[29][30][31]. However, few studies have focused on the effect of oxide scale on the hot-dip Zn-Al-Mg alloy coatings.
In this paper, induction heating was introduced into the reduction process to prepare the Zn-6Al-3Mg alloy coatings, and effects of oxide thickness on the structure and surface quality of the coating were studied. The environmental pollution caused by the acid pickling of the pre-treatment process and the evaporation of the solvent during the hot dipping process was avoided by this technology.

Materials Preparation
Q235 steel was used as the substrate, of which the chemical composition was given in Table 1. The compositions of the Zn-6Al-3Mg alloy immersion baths were given in Table 2. The hot-rolled Q235 steel sheets were cut into 40 mm × 20 mm × 2 mm sample plates. The samples were divided into two groups. One group of the samples was polished by 150# sandpaper to remove the oxide scale on the surface of the steel plates; the other group retained the oxide skin without any surface treatment. All the samples were degreased with NaOH (10 wt. %) solution and then washed with water. The homemade hot dipping simulator was shown in Figure 1. The samples were placed in a quartz tube surrounded by the high-frequency induction heating coil. N 2 with a flow rate of 4 L/min was introduced for 5 min to drain the air in the device. The induction heater (Model: SPG-06A-II, Frequency: 220 kHz, Power: 1.5 kW) was turned on, and then the N 2 was replaced by 70% N 2 + 30% H 2 mixture with a flow rate of 2 L/min for 40 s; at this time, the sample temperature was 900 • C. The temperature of the steel plate was monitored by an infrared thermometer. The samples were dipped into the Zn-6Al-3Mg bath (460 • C) for 30 s, and the Zn-6Al-3Mg coatings were formed after solidification in the protective atmosphere of a quartz tube.

Phase Structure of Steel Plate before Hot Dipping
Iron oxide was the main component of the oxide scale formed in a hot-rolled Q235 steel plate. The structure and composition of the oxide scale, which had a large impact on the hot-dip Zn-Al-Mg coatings, were very complex. Therefore, the structure and composition of the oxide scale should be clarified first. Figure 2 shows the microstructure of the cross-section before and after the reduction of polished and unpolished steel plates.
The thickness of the oxide scale on the surface of an unpolished steel plate was about 10 μm, and the oxide scale was divided into two layers from the surface to the inside as shown in Figure 2a,b. After the reduction of the steel plate, the structure became loose, and there was no obvious stratification. However, the outer layer (region 3 in Figure 2b) had higher density than the adjacent matrix (region 4 in Figure 2b). This was obviously different from the phenomenon in the literature [22], which mentioned that the surface of iron becomes dense after reduction at 900 °C. This may be caused by using different reduction equipment. The ordinary annealing furnace was used in that experiment, while induction heating was used in this study, which was fast and takes a short amount of time. The thickness of the residual oxide scale was about 2-3 μm after polishing, and the microstructure of the cross section of the steel plate was shown in Figure 2c,d. Because the residual

Characterization and Property Evaluation
The phase structures of the Zn-6Al-3Mg alloys were examined by X-ray diffraction (XRD; D8 Advance, Bruker, Germany) at 25 • C using Cu Kα radiation (1.5406 Å). The cross-section microstructure of the sample was characterized by using Quanta 450 FEG field emission scanning electron microscopy (SEM, FEI, Hillsboro, OR, USA) and energy dispersive spectroscopy (EDS, FEI, Hillsboro, OR, USA).

Phase Structure of Steel Plate before Hot Dipping
Iron oxide was the main component of the oxide scale formed in a hot-rolled Q235 steel plate. The structure and composition of the oxide scale, which had a large impact on the hot-dip Zn-Al-Mg coatings, were very complex. Therefore, the structure and composition of the oxide scale should be clarified first. Figure 2 shows the microstructure of the cross-section before and after the reduction of polished and unpolished steel plates. oxide scale after polishing was thin and uneven, the presence of the residual oxide scale was observed by scanning electron microscopy, but the porous and loose tissue morphology of the reduced oxide scale can't be clearly observed. The elemental composition and relative content at points 1-4 marked in Figure 2 are shown in Figure 3. As can be seen from the energy spectrum, the oxygen content on the reduced surface was greatly reduced, and the oxide on the surface was almost completely reduced, while the internal oxygen content became little and the internal oxide was not reduced. The phase composition of the oxide scale on the surface of the polished and unpolished steel plates were determined by XRD, as shown in Figure 4. The surface scale of the unpolished hot-rolled Q235 steel sheet was mainly composed of Fe3O4 and α-Fe, as shown in Figure 4a. The XRD analyses indicated that the surface oxide scale has a higher oxygen content and lower internal scale content. According to the related literature [28][29][30], the surface oxide scale of a hot rolled mild steel sheet mainly consisted of surface layer Fe3O4 and an internal Fe/Fe3O4 eutectoid structure. The composition of the surface phase after the reduction was shown in Figure 4b. The scale only contains α-Fe, which indicated that it was reduced to porous sponge iron. According to the depth data of Fe and its oxide X-ray diffraction calculated by DIFFRAC EVA software in Table 3, it can be seen that under the current test conditions, the maximum test depth of the surface Fe was less than 3.5 μm, while the thickness of the entire oxide scale was about 10 μm. Combined with the high content of O element at point 4 in Figure 2, it can be inferred that there were still an unreduced oxide scale near the matrix. Primavera and Hou proved that [24,27] the oxide scale of hot-rolled mild steel was reduced in the atmosphere containing H2 in the order of Fe3O4→FeO→Fe. When the reduction temperature was higher than 750 °C, only FeO and Fe phases existed in the scale, and when the reduction temperature was higher than 900 °C, only the α-Fe phase existed in the scale. Similar explanations can also be found in the literature [21]. After reduction, the oxide scale near the matrix still had a high oxygen content, which may be due to the scale effect of induction heating. Because the temperature of internal oxide was lower than 900 °C, FeO was not reduced, as shown in Figure 4c. The residual scale on the surface of the polished steel plate was still composed of Fe3O4 and α-Fe, but the content of Fe3O4 was The thickness of the oxide scale on the surface of an unpolished steel plate was about 10 µm, and the oxide scale was divided into two layers from the surface to the inside as shown in Figure 2a,b. After the reduction of the steel plate, the structure became loose, and there was no obvious stratification. However, the outer layer (region 3 in Figure 2b) had higher density than the adjacent matrix (region 4 in Figure 2b). This was obviously different from the phenomenon in the literature [22], which mentioned that the surface of iron becomes dense after reduction at 900 • C. This may be caused by using different reduction equipment. The ordinary annealing furnace was used in that experiment, while induction heating was used in this study, which was fast and takes a short amount of time. The thickness of the residual oxide scale was about 2-3 µm after polishing, and the microstructure of the cross section of the steel plate was shown in Figure 2c,d. Because the residual oxide scale after polishing was thin and uneven, the presence of the residual oxide scale was observed by scanning electron microscopy, but the porous and loose tissue morphology of the reduced oxide scale can't be clearly observed. The elemental composition and relative content at points 1-4 marked in Figure 2 are shown in Figure 3. As can be seen from the energy spectrum, the oxygen content on the reduced surface was greatly reduced, and the oxide on the surface was almost completely reduced, while the internal oxygen content became little and the internal oxide was not reduced.
The phase composition of the oxide scale on the surface of the polished and unpolished steel plates were determined by XRD, as shown in Figure 4. The surface scale of the unpolished hot-rolled Q235 steel sheet was mainly composed of Fe 3 O 4 and α-Fe, as shown in Figure 4a. The XRD analyses indicated that the surface oxide scale has a higher oxygen content and lower internal scale content. According to the related literature [28][29][30], the surface oxide scale of a hot rolled mild steel sheet mainly consisted of surface layer Fe 3 O 4 and an internal Fe/Fe 3 O 4 eutectoid structure. The composition of the surface phase after the reduction was shown in Figure 4b. The scale only contains α-Fe, which indicated that it was reduced to porous sponge iron. According to the depth data of Fe and its oxide X-ray diffraction calculated by DIFFRAC EVA software in Table 3, it can be seen that under the current test conditions, the maximum test depth of the surface Fe was less than 3.5 µm, while the thickness of the entire oxide scale was about 10 µm. Combined with the high content of O element at point 4 in Figure 2, it can be inferred that there were still an unreduced oxide scale near the matrix. Primavera and Hou proved that [24,27] the oxide scale of hot-rolled mild steel was reduced in the atmosphere containing H 2 in the order of Fe 3 O 4 →FeO→Fe. When the reduction temperature was higher than 750 • C, only FeO and Fe phases existed in the scale, and when the reduction temperature was higher than 900 • C, only the α-Fe phase existed in the scale. Similar explanations can also be found in the literature [21]. After reduction, the oxide scale near the matrix still had a high oxygen content, which may be due to the scale effect of induction heating. Because the temperature of internal oxide was lower than 900 • C, FeO was not reduced, as shown in Figure 4c. The residual scale on the surface of the polished steel plate was still composed of Fe 3 O 4 and α-Fe, but the content of Fe 3 O 4 was relatively less than that on the surface of unpolished steel plate. Pure iron was obtained after reduction of the residual oxide scale, and the XRD pattern in Figure 4d has only an α-Fe phase. Combined with Figure 2c,d and Table 3, the maximum test depth of Fe was greater than the thickness of the residual oxide scale, indicating that the reduction was sufficient.
Metals 2020, 10, x FOR PEER REVIEW 5 of 15 relatively less than that on the surface of unpolished steel plate. Pure iron was obtained after reduction of the residual oxide scale, and the XRD pattern in Figure 4d has only an α-Fe phase. Combined with Figure 2c,d and Table 3, the maximum test depth of Fe was greater than the thickness of the residual oxide scale, indicating that the reduction was sufficient.    relatively less than that on the surface of unpolished steel plate. Pure iron was obtained after reduction of the residual oxide scale, and the XRD pattern in Figure 4d has only an α-Fe phase. Combined with Figure 2c,d and Table 3, the maximum test depth of Fe was greater than the thickness of the residual oxide scale, indicating that the reduction was sufficient.

Surface Appearances and Phase Structure of the Coating
Photographs of samples reduced for 40 s in 70% N 2 + 30% H 2 atmosphere, and dipped in Zn-6Al-3Mg alloy bath at 460 • C, were shown in Figure 5. The steel entry temperature of both sets of plates was 900 • C, and the complete coating was obtained, but the surface quality of the coating was quite different. After hot dipping of the polished steel plate (Figure 5a), many strip-like protrusion defects appeared on the surface of the coating. However, the surface of an unpolished steel plate coating was smooth with no defects (Figure 5b). In order to find the cause of this strange phenomenon, it is necessary to study the influence of oxidized scale on the coating when the entry temperature of steel was 900 • C.

Surface Appearances and Phase Structure of the Coating
Photographs of samples reduced for 40 s in 70% N2 + 30% H2 atmosphere, and dipped in Zn-6Al-3Mg alloy bath at 460 °C, were shown in Figure 5. The steel entry temperature of both sets of plates was 900 °C, and the complete coating was obtained, but the surface quality of the coating was quite different. After hot dipping of the polished steel plate (Figure 5a), many strip-like protrusion defects appeared on the surface of the coating. However, the surface of an unpolished steel plate coating was smooth with no defects (Figure 5b). In order to find the cause of this strange phenomenon, it is necessary to study the influence of oxidized scale on the coating when the entry temperature of steel was 900 °C.
The phase composition was also determined by XRD, as shown in Figure 6. It can be seen that the two kinds of coatings with different pretreatments were composed of three phases of Zn, Al, and MgZn2, that is, although the macroscopic morphology of the coating was quite different, the phase composition was the same.

Surface Characterization of the Coating
The SEM analysis was performed on the surface of the coated layer after hot dipping of the polished steel plate, that is, the area A in Figure 5, as shown in Figure 7. Many obvious striated protruding defects were formed on the surface of the polished steel sheet after hot dipping ( Figure  7a). The enlarged views of the boundary between the striated defects and the surface overlay was shown in Figure 7b (region A in Figure 7a). The local enlarged view of the junction of overlay was shown in Figure 7c (Area B in Figure 7b). The components of region 1 and region 2 in Figure 7c were shown in Figure 8. As can be seen from Figure 7, region 1 was the overlay of Zn-6Al-3Mg, while region 2 was the Fe-Al alloy layer with an atomic ratio of 1:3.25. Combined with Figure 9, region 2 was determined to be the Fe4Al13 phase. Figure 7d was a magnification of the region 2, and the distribution of the Fe4Al13 phase was dense. The phase composition was also determined by XRD, as shown in Figure 6. It can be seen that the two kinds of coatings with different pretreatments were composed of three phases of Zn, Al, and MgZn 2 , that is, although the macroscopic morphology of the coating was quite different, the phase composition was the same. SEM analysis was performed on the surface of the coating obtained from the unpolished hot dipping plate, that is, the region B in Figure 5, as shown in Figure 10. Compared with the polished steel plate, the surface of the unpolished steel plate had no strip protruding defects (Figure 10a,b). The SEM analysis revealed that the banded structure in Figure 10c was a Zn-rich phase, the darkercolored region was a Zn/Al/MgZn2 ternary eutectic structure, in which Al was interspersed between the strips of Zn and MgZn2. The unpolished steel sheet also had some small defects due to the unevenness of the scale (Figure 10d).

Surface Characterization of the Coating
The SEM analysis was performed on the surface of the coated layer after hot dipping of the polished steel plate, that is, the area A in Figure 5, as shown in Figure 7. Many obvious striated protruding defects were formed on the surface of the polished steel sheet after hot dipping (Figure 7a). The enlarged views of the boundary between the striated defects and the surface overlay was shown in Figure 7b (region A in Figure 7a). The local enlarged view of the junction of overlay was shown in Figure 7c (Area B in Figure 7b). The components of region 1 and region 2 in Figure 7c were shown in Figure 8. As can be seen from Figure 7, region 1 was the overlay of Zn-6Al-3Mg, while region 2 was the Fe-Al alloy layer with an atomic ratio of 1:3.25. Combined with Figure 9, region 2 was determined to be the Fe 4 Al 13 phase. Figure 7d was a magnification of the region 2, and the distribution of the Fe 4 Al 13 phase was dense. SEM analysis was performed on the surface of the coating obtained from the unpolished hot dipping plate, that is, the region B in Figure 5, as shown in Figure 10. Compared with the polished steel plate, the surface of the unpolished steel plate had no strip protruding defects (Figure 10a,b). The SEM analysis revealed that the banded structure in Figure 10c was a Zn-rich phase, the darkercolored region was a Zn/Al/MgZn2 ternary eutectic structure, in which Al was interspersed between the strips of Zn and MgZn2. The unpolished steel sheet also had some small defects due to the unevenness of the scale (Figure 10d).      Figure 11 showed the cross-sectional SEM microstructure of the coatings obtained after polished and unpolished steel plates, which had great difference in cross section morphology. The coating structure of the polished steel plate was thick, and its thickness was about 75 μm, which consisted of an overlay and alloy layer. The Fe4Al13 phase in the coating was separated to the surface of the overlay layer. The coating structure of the unpolished steel plate was fine and its thickness was about 48 μm, and there was no Fe-Al alloy layer appeared. The hot dipping was the process of mutual wetting between the molten Zn-Al-Mg immersion baths and the steel matrix. Local complete and incomplete wetting behaviors of grain boundaries by the molten immersion baths occurred due to the presence of oxide scale when the immersion baths reacted with the steel matrix. Such wetting behaviour caused the white Zn-rich phase to be surrounded by the black Zn/Al/MgZn2 ternary eutectic phase appeared in the coating (Figure 11b). The same structure appeared on the coating surface of the unpolished steel plate (Figure 10c,d). This similar complete and incomplete wetting of grain boundaries by the melt phase was also shown in the results of Straumal [32]. The grains of the coating of the unpolished steel plate were finer after hot dipping (Figure 11b), which had better overall performance [33]. SEM analysis was performed on the surface of the coating obtained from the unpolished hot dipping plate, that is, the region B in Figure 5, as shown in Figure 10. Compared with the polished steel plate, the surface of the unpolished steel plate had no strip protruding defects (Figure 10a,b). The SEM analysis revealed that the banded structure in Figure 10c was a Zn-rich phase, the darker-colored region was a Zn/Al/MgZn 2 ternary eutectic structure, in which Al was interspersed between the strips of Zn and MgZn 2 . The unpolished steel sheet also had some small defects due to the unevenness of the scale (Figure 10d). The coating of the unpolished steel plate retained the obvious oxide scale of about 8 μm thick near the matrix. The coating overlay was gently polished off with fine sandpaper, and the phase composition of the oxide scale was tested. It can be seen from Figure 12 that the main phase of the residual oxide scale was FeO.

Cross-Section Characterization of the Coating
After hot-dipping the Zn-Al-Mg coating on the polished steel plate, the elements distribution of the cross section was shown in Figure 13. When the steel entry induction heating temperature was 900 °C, a large number of Fe elements, as revealed by the SEM analysis of these regions, diffused from the matrix to the near surface of the coating, forming an Fe-Al alloy layer with Al elements. A small amount of Fe-Al alloy layer also existed near the substrate. The Mg element was not involved in the Fe-Al alloy reaction, but it mainly existed in the Zn/MgZn2 binary and Zn/Al/MgZn2 ternary eutectic structure.  Figure 11 showed the cross-sectional SEM microstructure of the coatings obtained after polished and unpolished steel plates, which had great difference in cross section morphology. The coating structure of the polished steel plate was thick, and its thickness was about 75 µm, which consisted of an overlay and alloy layer. The Fe 4 Al 13 phase in the coating was separated to the surface of the overlay layer. The coating structure of the unpolished steel plate was fine and its thickness was about 48 µm, Metals 2020, 10, 1519 9 of 14 and there was no Fe-Al alloy layer appeared. The hot dipping was the process of mutual wetting between the molten Zn-Al-Mg immersion baths and the steel matrix. Local complete and incomplete wetting behaviors of grain boundaries by the molten immersion baths occurred due to the presence of oxide scale when the immersion baths reacted with the steel matrix. Such wetting behaviour caused the white Zn-rich phase to be surrounded by the black Zn/Al/MgZn 2 ternary eutectic phase appeared in the coating (Figure 11b). The same structure appeared on the coating surface of the unpolished steel plate (Figure 10c,d). This similar complete and incomplete wetting of grain boundaries by the melt phase was also shown in the results of Straumal [32]. The grains of the coating of the unpolished steel plate were finer after hot dipping (Figure 11b), which had better overall performance [33]. The coating of the unpolished steel plate retained the obvious oxide scale of about 8 μm thick near the matrix. The coating overlay was gently polished off with fine sandpaper, and the phase composition of the oxide scale was tested. It can be seen from Figure 12 that the main phase of the residual oxide scale was FeO.

Cross-Section Characterization of the Coating
After hot-dipping the Zn-Al-Mg coating on the polished steel plate, the elements distribution of the cross section was shown in Figure 13. When the steel entry induction heating temperature was 900 °C, a large number of Fe elements, as revealed by the SEM analysis of these regions, diffused from the matrix to the near surface of the coating, forming an Fe-Al alloy layer with Al elements. A small amount of Fe-Al alloy layer also existed near the substrate. The Mg element was not involved in the Fe-Al alloy reaction, but it mainly existed in the Zn/MgZn2 binary and Zn/Al/MgZn2 ternary eutectic structure.    After hot-dipping the Zn-Al-Mg coating on the polished steel plate, the elements distribution of the cross section was shown in Figure 13. When the steel entry induction heating temperature was 900 • C, a large number of Fe elements, as revealed by the SEM analysis of these regions, diffused from the matrix to the near surface of the coating, forming an Fe-Al alloy layer with Al elements. A small amount of Fe-Al alloy layer also existed near the substrate. The Mg element was not involved in the Fe-Al alloy reaction, but it mainly existed in the Zn/MgZn 2 binary and Zn/Al/MgZn 2 ternary eutectic structure.  The schematic diagram of the coating formation of the polished steel plate was shown in Figure  14. Reduction efficiency was improved after the steel plate was polished, and the remaining oxide scale on the surface of the steel base was fully reduced to Fe. The hot-dip galvanizing process was the process of interdiffusion between bath atoms and steel matrix atoms, and the porous loose sponge Fe contributed to hot-dip galvanizing. It can be seen from Figure 2d that the scale remained on the steel substrate was sufficiently reduced to Fe. The diffusion of Fe atoms was promoted by the high temperature of the steel plate. The Fe atoms reacted with Al atoms in the immersion baths to form the Fe4Al13 phase, which maintained a continuous structure in the overlay due to its toughness. At high temperature, the Fe4Al13 phase separated into the surface of the overlay, resulting in serious strip-like protrusion defects on the surface of the coating. The separation of Fe4Al13 in the coating was different from the periodic layered structure formed due to diffusion and migration [34][35][36]. The formation of this structure was mainly due to the high temperature before the steel plate entered the Zn-Al-Mg baths [37], and the amount of oxide scale on the steel plate significantly decreased after polishing, resulting in more Fe atoms spreading violently from the surface of the plate into the bath. The violent diffusion forms an Fe-Al alloy layer, and finally the alloy layer was separated from the matrix, causing serious protrusion defects on the surface of the coating. The schematic diagram of the coating formation of the polished steel plate was shown in Figure 14.
Reduction efficiency was improved after the steel plate was polished, and the remaining oxide scale on the surface of the steel base was fully reduced to Fe. The hot-dip galvanizing process was the process of interdiffusion between bath atoms and steel matrix atoms, and the porous loose sponge Fe contributed to hot-dip galvanizing. It can be seen from Figure 2d that the scale remained on the steel substrate was sufficiently reduced to Fe. The diffusion of Fe atoms was promoted by the high temperature of the steel plate. The Fe atoms reacted with Al atoms in the immersion baths to form the Fe 4 Al 13 phase, which maintained a continuous structure in the overlay due to its toughness. At high temperature, the Fe 4 Al 13 phase separated into the surface of the overlay, resulting in serious strip-like protrusion defects on the surface of the coating. The separation of Fe 4 Al 13 in the coating was different from the periodic layered structure formed due to diffusion and migration [34][35][36]. The formation of this structure was mainly due to the high temperature before the steel plate entered the Zn-Al-Mg baths [37], and the amount of oxide scale on the steel plate significantly decreased after polishing, resulting in more Fe atoms spreading violently from the surface of the plate into the bath. The violent diffusion forms an Fe-Al alloy layer, and finally the alloy layer was separated from the matrix, causing serious protrusion defects on the surface of the coating. The distribution of elements in the section of the unpolished steel plate coating was shown in Figure 15. The high O content near the matrix (Figure 15e) indicated the oxide scale was not completely reduced. The Fe-Al alloy layer separated into the overlay as shown in Figure 12 was not found in the coating. Only Al enrichment occurred at the fracture of the scale, forming a small amount of Fe-Al alloy layer. The Mg elements were also uniformly distributed in the coating.
The schematic diagram of the formation of the unpolished steel plate was shown in Figure 16. The distribution of elements in the section of the unpolished steel plate coating was shown in Figure 15. The high O content near the matrix (Figure 15e) indicated the oxide scale was not completely reduced. The Fe-Al alloy layer separated into the overlay as shown in Figure 12 was not found in the coating. Only Al enrichment occurred at the fracture of the scale, forming a small amount of Fe-Al alloy layer. The Mg elements were also uniformly distributed in the coating.   The schematic diagram of the formation of the unpolished steel plate was shown in Figure 16. There was no significant separation of the Fe-Al alloy layer in the overlay of the unpolished steel plate. It can be seen from Figure 2b that after the scale of the unpolished steel plate was reduced, it had a porous structure with high porosity, and there were many micro-cracks formed in the scale, which make the unpolished steel plate have a higher specific surface area. The porous surface can provide more diffusion channels for the immersion baths, which is conducive to the aggregation of Al atoms on the interface and formation of the Fe-Al phase [22]. However, the Fe-Al phase was not obviously formed after the reduction of the unpolished steel plate in this experiment. The main reason was that a certain amount of FeO remains in the oxide scale of the unpolished steel plate, which had not been completely reduced [19]. And after reduction, the thickness of the pure iron on the surface of the steel plate was relatively thinner. The pure iron reduced on the surface can react with Al in the immersion baths to form an Fe-Al phase. After this part of iron was consumed, the unreduced FeO was in contact with the immersion baths, which can't provide enough Fe atoms for the formation of the Fe-Al phase, resulting in no obvious Fe-Al alloy layer formed on the surface of the unpolished steel plate. This phenomenon clearly indicated that the presence of oxide scale had a significant hindrance to the formation of the Fe-Al alloy layer.

Conclusions
In this study, one group of the samples was polished by 150# sandpaper to remove the oxide scale; the other group retained the oxide scale without any surface treatment. All the samples were degreased with NaOH (10 wt. %) solution and then washed with water. The samples were placed in During subsequent study, it was found that with the steel entry temperature decreased, the phenomenon of separation of the Fe-Al alloy layer into the overlay during the hot dipping process of the polished steel plate gradually weakened, and the alloy layer gradually approached the substrate. However, the oxide scale of the unpolished steel plate hindered the formation and separation of the alloy layer. The effect of oxide scale on the interface reaction of the coating was proved by choosing a higher plate temperature.

Conclusions
In this study, one group of the samples was polished by 150# sandpaper to remove the oxide scale; the other group retained the oxide scale without any surface treatment. All the samples were degreased with NaOH (10 wt. %) solution and then washed with water. The samples were placed in a quartz tube surround by the high-frequency induction heating coil. N 2 with a flow rate of 4L/min was introduced for 5 min to drain the air in the device. The induction heater (Model: SPG-06A-II, Frequency: 220 kHz, Power: 1.5 kW) was turned on, and then the N 2 was replaced by 70% N 2 + 30% H 2 mixture with a flow rate of 2 L/min for 40 s at this time. The samples were dipped into the Zn-6Al-3Mg bath (460 • C) for 30 s. In this way, a good Zn-6Al-3Mg coating can be obtained. The effect of the oxide scale on the structure and surface quality of Zn-6Al-3Mg alloy coating was investigated, and the following conclusions were obtained.

1.
The main components of the oxide scale of hot-rolled Q235 steel sheet were Fe 3 O 4 and α-Fe.
The thickness of the oxide scale on the surface of the unpolished steel plate was about 10 µm. After reduction for 40 s, the surface of the steel plate was a loose sponge, while the inside was still FeO. The thickness of the residual oxide scale on the surface of the polished steel plate was about 2-3 µm, which can be reduced to sponge iron in the N 2 -H 2 mixed atmosphere. The reduction efficiency was improved by mechanical grinding.

2.
When the temperature of the steel sheet was 900 • C, the Fe atoms in the steel matrix violently diffused into the immersion baths, resulting in the separation of the Fe 4 Al 13 alloy layer formed in the coating of the polishing steel sheet from the matrix into the overlay, even to the surface of the coating, causing serious defects on the surface of the coating.