Microstructural Evolution of Reaction Layer of 1.5 GPa Boron Steel Hot-Dipped in Al-7wt%Ni-6wt%Si Alloy

: The constituents, distribution, and characteristics of the phases formed on the coating layer of boron steel hot-dipped in Al-7wt%Ni-6wt%Si were evaluated in detail. In particular, the microstructure and phase constitution of the reaction layer were characterized. Moreover, the microstructural evolution mechanism of the phase was presented with reference to the (Al-7wt%Ni-6wt%Si)-xFe from the pseudo-binary phase diagram. The solidiﬁcation layer consisted mainly of Al, Al 3 Ni, and Si phases. Reaction layers were formed in the order of Al 9 FeNi(T), Fe 4 Al 13 ( θ ), and Fe 2 Al 5 ( η ) from the solidiﬁcation layer side. In addition, the κ (Fe 3 AlC) layer was formed at the Fe 2 Al 5 ( η )/steel interface. From pseudo-binary phase diagram analysis, it was found that Fe 4 Al 13 ( θ ) can form when the Fe concentration is over 2.63 wt% in the 690 ◦ C Al-7wt%Ni-6wt%Si molten metal. When the concentration of Fe increased to 10.0–29.0 wt%, isothermal solidiﬁcation occurred in the Fe 4 Al 13 ( θ ) and Al 9 FeNi(T) phases simultaneously. Moreover, given that the T phase does not dissolve Si, it was discharged, and the Si phase was formed around the Al 9 FeNi(T) phase. The Fe 2 Al 5 ( η ) phase was formed by a diffusion reaction between Fe 4 Al 13 ( θ ) and steel, not a dissolution reaction. Moreover, Al 2 Fe 3 Si 3 ( τ 1 ) was formed at the Fe 4 Al 13 ( θ )-Fe 2 Al 5 ( η ) interface by discharging Si from Fe 4 Al 13 ( θ ) without Si solubility. Furthermore, the Fe 3 AlC( κ ) layer was formed by carbon accumulation that discharged in the Fe 2 Al 5 ( η ) region transformed from steel to Fe 2 Al 5 ( η ). The twin regions in the Fe 4 Al 13 ( θ ) and Fe 2 Al 5 ( η ) grain were due to the strains caused by the lattice transformation in the constrained state, wherein the phases are present between the Al 9 FeNi(T) layer and steel.


Introduction
In recent years, the automobile industry has been using boron steel, in which boron is added to a low alloy steel, for environmental friendliness, low fuel consumption, and low weight. When boron is added to steel, the boron is segregated in the previous austenite grain boundaries, even if only approximately 20 ppm is added, and the hardenability can be significantly improved by delaying the ferrite nucleation on cooling. Moreover, boron steel is boron-added low alloy steel, and it is a type of advanced high-strength steel (AHSS) with a tensile strength of 1.5 GPa [1]. Due to the high strength of boron steel, cold forming is difficult, and the spring-back phenomenon occurs after forming. Therefore, hot stamping is performed to maintain the ductility by heating to a temperature of A3 or higher, press forming, and rapid cooling to maintain a high strength. Hot stamping involves heating at a temperature of A3 or higher, to ensure ductility, and then quenching to maintain a high strength [2][3][4][5]. In hot stamping, surface oxidation and surface decarburization occur on heating at a high temperature; therefore, it is mainly performed with Al-10%Si coating, which has an excellent oxidation resistance at high temperatures [6][7][8]. When the Al-Si hot-dipped boron steel was laser-welded, Al and Si segregation portions were formed in the fusion zone. The portion formed a ferrite with low hardness during hot stamping, and it was reported that the tensile properties of the fusion zone decreased [9][10][11][12]. In addition, Yoon et al. reported that a hook-shaped Fe 3 (Al, Si) intermetallic compound was formed on the interface between the fusion zone and base metal during laser welding. Moreover, it was transformed to polygonal ferrite during the hot stamping processes, and the tensile properties were degraded [13].
There are several studies on the development of coating compositions for the inhibition of ferrite formation in Al-Si hot-dipped boron steel welding. Coating compositions were designed by adding Mn or Ni, which are austenite-forming elements, to Al. In general, research on the microstructures of Al-7wt%Mn [14] and the Al-7wt%Ni [15] coating layer was conducted. In the abovementioned studies, the desired Ni phase (Al 9 FeNi) and Mn phase (Al 6 Mn, Al 11 Mn 4 ) were formed; however, the Fe 2 Al 5 (η) phase layer, which is the cause of ferrite formation, was formed with a thickness of approximately 100 µm.
On the other hand, when Si is added to a molten bath when hot-dipping the carbon steel and low alloy steel, the thickness of the produced intermetallic compound Fe 2 Al 5 (η) phase during plating can be reduced. In addition, the melting point and viscosity of the molten bath can be reduced to increase the coating ability [16][17][18]. Several mechanisms were proposed to explain the decrease in the thickness of the Fe 2 Al 5 (η) phase due to the addition of Si, such as an Al diffusion delay due to the Si solubility in the Fe 2 Al 5 (η) phase vacancy [16,19], Al diffusion delay due to the Si segregated in the Fe 2 Al5(η) phase grain boundary [20], and a reduction in the Al activation coefficient due to the addition of Si [21,22]. In most studies, it was reported that Si inhibits the growth of the Fe 2 Al 5 (η) phase by reducing the diffusion and activation coefficients of Al. On the other hand, Lemmens B. et al. reported that the amount and thickness of the Al-Fe-Si ternary phase (τ 1 , τ 5 ) generated in the Fe 2 Al 5 (η) phase increased and the thickness of the Fe 2 Al 5 (η) phase decreased when the amount of Si added to the coating bath was increased [23]. Yun et al. reported that the thickness of the Fe 2 Al 5 (η) phase decreased and the amount of the Al 9 FeNi(T) phase (which they think that inhibiting ferrite formation phase when welding) increased as the amount of Si increased through hot-dip coating with (1-6) wt%Si added to Al-7wt%Ni [24]. In particular, it was reported that due to the addition of Si, 1-5 at% of Si was dissolved in the η phase to interfere with the growth of Al. Moreover, the Al 2 Fe 3 Si 3 (τ 1 ) phase was generated, and it preferentially consumed Al. However, there was no mention of the formation of a reaction layer during the coating, the characteristics, and formation mechanism of the reaction phases.
Given that the Fe2Al5(η) phase of the reaction layer that contains a minimum of 5 wt% of Si during the hot-dip coating of the Al alloy is 2 µm or lower in thickness, and the grain diameter of the Al-Fe-Si ternary phase (τ 1 , τ 5 ) is 1 µm or lower, studies that employ transmission electron microscopy (TEM) are required. However, in most studies, only X-ray diffraction (XRD), energy-dispersive X-ray spectroscopy (EDS), and electron backscatter diffraction (EBSD) analyses were conducted; thus, the accuracy of the analyses was limited. In addition, there are no reports on the formation of carbon-related phases during the coating of steels that contain a certain amount of carbon, such as boron steel (0.22 wt% of carbon).
Hence, in this study, the reaction layer formed on the boron steel coated at 690 • C for 2 min in an Al-7wt%Ni-6wt%Si molten bath was deflected by a focused ion beam (FIB), and phase identification was conducted using TEM and TEM-EDS. Moreover, Electron Backscatter Diffraction (EBSD) was used for the microstructure analysis of the macroscopic reaction layer and solidification layer. The mechanism of formation of intermetallic compounds was further investigated using a pseudo-binary phase diagram.

Materials and Methods
The base metal used in this study was boron steel (22MnB5) with a thickness of 1.4 mm, for which the chemical composition is presented in Table 1. The Al-7wt%Ni-6wt%Si alloy was prepared by appropriately mixing high-purity (99.99%) Al ingot, high-purity (99.99%) Al-15wt%Si ingot, and high-purity (99.99%) Ni ingot. The specimens to be coated were cut to a size of 40 mm × 120 mm, pickled in a solution that consisted of 40 mL of H 2 SO 4 and 200 mL of CH 3 OH for 5 min, and then rinsed using C 2 H 5 OH. The prepared hot-dip coated alloy was heated in a crucible set in a high-frequency furnace, and then maintained at a coating temperature of 690 • C for melting. The pre-treated specimens were pre-heated at 150 • C for 20 s to remove moisture from the surface, and then placed in a molten bath of Al-7wt%Ni-6wt%Si, which was maintained at 690 • C for 2 min for coating. After coating, the specimens were cooled in air.
The coated specimens were polished in the normal direction using #180-2500 SiC sandpaper, a 1 µm diamond suspension, and a 0.04 µm colloidal silica suspension. The coating layer was then etched using 49 mL of distilled water and 1 mL of HF according to ASTM B487-85. The etched coating layer was observed using an optical microscope (Olympus BX51M; Tokyo, Japan) and field-emission scanning electron microscopy (FE-SEM) system (Carl Zeiss SUPRA 45; Oberkochen, Germany). Moreover, EDS (EDAX; Mahwah, NJ, USA) and field-emission electron probe micro-analysis (FE-EPMA) (JEOL JXA-8530F; Tokyo, Japan) were performed to analyze the composition of the coating layer. For FE-EPMA, area analysis was conducted with a voltage of 10 kV, a current of 100 nA, and step sizes of 0.1-0.5 mm in the non-etched state.
For the phase analysis of the reaction layer, the TEM specimens were deflected using FIB (FEI, Scios; Hillsboro, OR, USA). The diffraction pattern and chemical composition were analyzed using FE-TEM (FEI, TALOS F200X; Hillsboro, OR, USA). In addition, EBSD (EDAX, Hikari; Mahwah, NJ, USA) was used for the phase distribution and crystal orientation analyses. In particular, phase distributions were confirmed by Chi-Scanning that using EDS and EBSD. The thickness and length of the reaction layer were measured using the Image-Plus Pro image analysis program. The equilibrium composition of each phase formed during the coating and the (Al-7wt%Ni-6wt%Si)-xFe pseudo-binary phase diagram were obtained using Thermo-Calc TM software Ver. 6.0 (Solna, Sweden), for the evaluation of the formation mechanism of each phase. Figure 1 presents the structure of the specimen coated at 690 • C for 2 min in an Al-7wt%Ni-6wt% Si molten bath, as observed using the optical microscope. The coating layer was classified into a solidification layer (SL) and reaction layer (RL). Reaction layer (RL) is boron-steel/coating layer interface that composed of gray layers that observed some black particles inside and ivory phases that attached to gray layer and grown to solidification layer.

Identification of Solidification Layer Phases
The second phases in the solidification layer were mainly formed on the grain (or dendrite) boundary. Metals 2018, 7, x 4 of 16   Figures 1 and 3 presents the results of the EBSD analysis. Given that EBSD was conducted in a non-etched state, the microstructure distribution was slightly different from that in the SEM image by re-polishing. Table 2 presents the results of the composition analysis of the phases labeled 1-11 in Figure 2 using EDS point analysis. The black needle-shaped phases 1-3 in Figure 2 represent Si. A small amount of Al was detected; however, Si did not dissolve in Al [25]. Hence, the detected Al represents the Al matrix in the EDS analysis results. In the EBSD results, the phase was Si, whereas the structures in the SEM image consisted of aggregates of fine grains, and the randomly shaped phases were parts of one grain. On the other hand, phases 4-11 exhibited a slight color difference when observed using an optical microscope and SEM. Moreover, phases 9-11 were reaction layer phases, as they were connected to the reaction layer. These phases were classified into two phase groups by EDS analysis. Phases 4-11 contained Al and Ni as the main components, but divided into two kinds: 4-6 in which Si and Fe are small in solid content and 7-11 phases in which Fe and Si contents are higher than those phases. From the EBSD results, phases 4-6 were Al3Ni (orthorhombic). Furthermore, Al3Ni was observed as a polygonal shape in the SEM image; however, from the EBSD results, it was found to be an aggregate of very fine grains. It was confirmed that phases 7-11 were Al9FeNi (T, monoclinic) phases. In Figures 2, 7 and 8, the Al9FeNi(T) phases were present in the solidification layer, which were similar to the phase indicated by the red circle in Figure 3. Moreover, these were mostly observed near the reaction layer.    Figures 1 and 3 presents the results of the EBSD analysis. Given that EBSD was conducted in a non-etched state, the microstructure distribution was slightly different from that in the SEM image by re-polishing. Table 2 presents the results of the composition analysis of the phases labeled 1-11 in Figure 2 using EDS point analysis. The black needle-shaped phases 1-3 in Figure 2 represent Si. A small amount of Al was detected; however, Si did not dissolve in Al [25]. Hence, the detected Al represents the Al matrix in the EDS analysis results. In the EBSD results, the phase was Si, whereas the structures in the SEM image consisted of aggregates of fine grains, and the randomly shaped phases were parts of one grain. On the other hand, phases 4-11 exhibited a slight color difference when observed using an optical microscope and SEM. Moreover, phases 9-11 were reaction layer phases, as they were connected to the reaction layer. These phases were classified into two phase groups by EDS analysis. Phases 4-11 contained Al and Ni as the main components, but divided into two kinds: 4-6 in which Si and Fe are small in solid content and 7-11 phases in which Fe and Si contents are higher than those phases. From the EBSD results, phases 4-6 were Al 3 Ni (orthorhombic). Furthermore, Al 3 Ni was observed as a polygonal shape in the SEM image; however, from the EBSD results, it was found to be an aggregate of very fine grains. It was confirmed that phases 7-11 were Al 9 FeNi (T, monoclinic) phases. In Figure 2, Figures 7 and 8, the Al 9 FeNi(T) phases were present in the solidification layer, which were similar to the phase indicated by the red circle in Figure 3. Moreover, these were mostly observed near the reaction layer.    Figures 1 and 3 presents the results of the EBSD analysis. Given that EBSD was conducted in a non-etched state, the microstructure distribution was slightly different from that in the SEM image by re-polishing. Table 2 presents the results of the composition analysis of the phases labeled 1-11 in Figure 2 using EDS point analysis. The black needle-shaped phases 1-3 in Figure 2 represent Si. A small amount of Al was detected; however, Si did not dissolve in Al [25]. Hence, the detected Al represents the Al matrix in the EDS analysis results. In the EBSD results, the phase was Si, whereas the structures in the SEM image consisted of aggregates of fine grains, and the randomly shaped phases were parts of one grain. On the other hand, phases 4-11 exhibited a slight color difference when observed using an optical microscope and SEM. Moreover, phases 9-11 were reaction layer phases, as they were connected to the reaction layer. These phases were classified into two phase groups by EDS analysis. Phases 4-11 contained Al and Ni as the main components, but divided into two kinds: 4-6 in which Si and Fe are small in solid content and 7-11 phases in which Fe and Si contents are higher than those phases. From the EBSD results, phases 4-6 were Al3Ni (orthorhombic). Furthermore, Al3Ni was observed as a polygonal shape in the SEM image; however, from the EBSD results, it was found to be an aggregate of very fine grains. It was confirmed that phases 7-11 were Al9FeNi (T, monoclinic) phases. In Figures 2, 7 and 8, the Al9FeNi(T) phases were present in the solidification layer, which were similar to the phase indicated by the red circle in Figure 3. Moreover, these were mostly observed near the reaction layer.    Table 2. Energy-dispersive X-ray spectroscopy (EDS) analysis results of phases denoted as 1-11 in Figure 2.

Microstructural Evolution of Solidification Layer
The formation of the Fe-containing phase in the solidification layer indicates that Fe was dissolved in the Al-7wt%Ni-6wt%Si liquid phase. Based on this, (Al-7wt%Ni-6%wtSi)-xFe pseudobinary phase diagram was obtained using Thermo-Calc TM software, to evaluate the formation of a solidification layer under phase transformation during the cooling of the (Al-7wt%Ni-6%wtSi)-xFe liquid containing Fe. Figure 4 presents the (Al-7wt%Ni-6%wtSi)-xFe pseudo-binary phase diagram obtained using Thermo-Calc TM software. In Figure 4, it can be seen that the phases formed during cooling vary depending on the Fe content. The Fe content during solidification was 0.92 wt% (0.46 at%), as measured using EDS. The phase changes occurring during cooling proceeded in the order of primary Al3Ni + Liquid→Al3Ni + Al + Liquid→Al3Ni + τ1 + Al + Liquid→Al3Ni + τ1 + Al + Si. On the other hand, given the dimensions of the supercooled specimen (thickness of 1.4 mm, width of 40 mm, and length  Table 2. Energy-dispersive X-ray spectroscopy (EDS) analysis results of phases denoted as 1-11 in Figure 2.

Microstructural Evolution of Solidification Layer
The formation of the Fe-containing phase in the solidification layer indicates that Fe was dissolved in the Al-7wt%Ni-6wt%Si liquid phase. Based on this, (Al-7wt%Ni-6%wtSi)-xFe pseudo-binary phase diagram was obtained using Thermo-Calc TM software, to evaluate the formation of a solidification layer under phase transformation during the cooling of the (Al-7wt%Ni-6%wtSi)-xFe liquid containing Fe. Figure 4 presents the (Al-7wt%Ni-6%wtSi)-xFe pseudo-binary phase diagram obtained using Thermo-Calc TM software. In Figure 4, it can be seen that the phases formed during cooling vary depending on the Fe content. The Fe content during solidification was 0.92 wt% (0.46 at%), as measured using EDS. The phase changes occurring during cooling proceeded in the order of primary Al 3 Ni + Liquid→Al 3 Ni + Al + Liquid→Al 3 Ni + τ 1 + Al + Liquid→Al 3 Ni + τ 1 + Al + Si. On the other hand, given the dimensions of the supercooled specimen (thickness of 1.4 mm, width of 40 mm, and length of 120 mm), no phase transformation due to diffusion occurred at low temperatures.
In particular, when Al 3 Ni and Al were preferentially formed at the initial stage of cooling, only Si was concentrated in the liquid; thus, Si was easily crystallized, and the τ 1 (Al 2 Fe 3 Si 3 , triclinic) phase transformation degenerated. Hence, the solidification layer was considered to comprise Al 3 Ni + Al + Si 3 phases. On the other hand, in Figure 4, when Fe was concentrated to approximately 1.3-2.0 wt%, the primary T phase could form during cooling, and even in the liquid. It was therefore considered that the solidification layer near the reaction layer contained a relatively high concentration of Fe; thus, an Al 9 FeNi(T) phase was generated.
of 120 mm), no phase transformation due to diffusion occurred at low temperatures. In particular, when Al3Ni and Al were preferentially formed at the initial stage of cooling, only Si was concentrated in the liquid; thus, Si was easily crystallized, and the τ1(Al2Fe3Si3, triclinic) phase transformation degenerated. Hence, the solidification layer was considered to comprise Al3Ni + Al + Si3 phases. On the other hand, in Figure 4, when Fe was concentrated to approximately 1.3-2.0 wt%, the primary T phase could form during cooling, and even in the liquid. It was therefore considered that the solidification layer near the reaction layer contained a relatively high concentration of Fe; thus, an Al9FeNi(T) phase was generated.  Figure 5 presents an SEM image, in which the reaction layer region indicated by B in Figure 1 is enlarged. The reaction layer can be classified into three layers (R1, R2, R3) and one phase (white phases in the R3 layer). Table 3 presents the results of the EDS point analysis of the phases labeled 1-5. There was a decrease in the quantitative reliability, given that the phases were very fine. However, the expected phase in Table 3 was qualitatively predicted from the composition.   Figure 1 is enlarged. The reaction layer can be classified into three layers (R1, R2, R3) and one phase (white phases in the R3 layer). Table 3 presents the results of the EDS point analysis of the phases labeled 1-5. There was a decrease in the quantitative reliability, given that the phases were very fine. However, the expected phase in Table 3 was qualitatively predicted from the composition. of 120 mm), no phase transformation due to diffusion occurred at low temperatures. In particular, when Al3Ni and Al were preferentially formed at the initial stage of cooling, only Si was concentrated in the liquid; thus, Si was easily crystallized, and the τ1(Al2Fe3Si3, triclinic) phase transformation degenerated. Hence, the solidification layer was considered to comprise Al3Ni + Al + Si3 phases. On the other hand, in Figure 4, when Fe was concentrated to approximately 1.3-2.0 wt%, the primary T phase could form during cooling, and even in the liquid. It was therefore considered that the solidification layer near the reaction layer contained a relatively high concentration of Fe; thus, an Al9FeNi(T) phase was generated.  Figure 5 presents an SEM image, in which the reaction layer region indicated by B in Figure 1 is enlarged. The reaction layer can be classified into three layers (R1, R2, R3) and one phase (white phases in the R3 layer). Table 3 presents the results of the EDS point analysis of the phases labeled 1-5. There was a decrease in the quantitative reliability, given that the phases were very fine. However, the expected phase in Table 3 was qualitatively predicted from the composition.   To understand the composition, distribution, and grain size of the reaction layer, the TEM sample in Figure 1b (reactive layer region) was evaluated using FIB and observed by scanning transmission electron microscopy (STEM). Figure 6a,b present the high-angle annular dark-field (HAADF) and bright-field images, respectively, and Figure 6c is an enlarged HAADF image denoted as X in Figure 6a. Moreover, Figure 7 presents the EDS mapping analysis of Al, Ni, Si and Fe. From the Figure 7, the sample was divided into several regions, with respect to the concentrations of Al, Ni, Si and Fe. Table 4 presents the phase predictions from the EDS analysis of each area.  To understand the composition, distribution, and grain size of the reaction layer, the TEM sample in Figure 1b (reactive layer region) was evaluated using FIB and observed by scanning transmission electron microscopy (STEM). Figure 6a,b present the high-angle annular dark-field (HAADF) and bright-field images, respectively, and Figure 6c is an enlarged HAADF image denoted as X in Figure 6a. Moreover, Figure 7 presents the EDS mapping analysis of Al, Ni, Si and Fe. From the Figure 7, the sample was divided into several regions, with respect to the concentrations of Al, Ni, Si and Fe. Table 4 presents the phase predictions from the EDS analysis of each area.    Table 3. EDS analysis results of phases denoted as 1-5 in Figure 5. To understand the composition, distribution, and grain size of the reaction layer, the TEM sample in Figure 1b (reactive layer region) was evaluated using FIB and observed by scanning transmission electron microscopy (STEM). Figure 6a,b present the high-angle annular dark-field (HAADF) and bright-field images, respectively, and Figure 6c is an enlarged HAADF image denoted as X in Figure 6a. Moreover, Figure 7 presents the EDS mapping analysis of Al, Ni, Si and Fe. From the Figure 7, the sample was divided into several regions, with respect to the concentrations of Al, Ni, Si and Fe. Table 4 presents the phase predictions from the EDS analysis of each area.     Figure 8 presents the diffraction pattern of the EDS point analysis region labeled 1-11 in Figure 6. Phases 1 and 2 represent Al and Si, respectively, which were solidified layer structures. From the EDS point analysis, it was confirmed that phases 3 and 4, which grew along the solidification layer, were the Al 9 FeNi (T, monoclinic) phase. It can be seen that the Al 9 FeNi(T) phases had a relatively larger grain size than other phases, and the grain size on the side of the solidification layer was larger than that on the side of the base metal.

No. at%/wt%
From the EDS point analysis, phases 5-8 consisted mainly of Al and Fe. Moreover, EDS mapping analysis revealed that the Fe concentration of phases 5 and 6, which were in close proximity to the T phases, was approximately 2 at% lower than that of phases 7 and 8, which were in close proximity to the steel side, and the Si concentration was approximately 2 at% higher. By diffraction pattern analysis, phases 5 and 6 were confirmed as a Fe 4 Al 13 (θ, monoclinic) phase, and phases 7 and 8 were confirmed as a Fe 2 Al 5 (η, orthorhombic) phase. Furthermore, labeled 9 and 10 in Figure 6a which appears as a white particle, is identified as Al 2 Fe 3 Si 3 (τ 1 , triclinic) phase. Figure 6c reveals that the τ1 phases consist of several grains with diameters of 100-200 nm, rather than one grain. Figure 9 presents a bright-field image, in which the X-Z region of Figure 6b is enlarged. Figure 9a,b are the enlarged images of the Al 9 FeNi(T) phase layer. It can be seen that the Fe 4 Al 13 (θ) phase layer formed to a thickness of less than 1 µm directly under the Al 9 FeNi(T) phase layer. The grain diameter of the Fe 4 Al 13 (θ) phase adjacent to the Al 9 FeNi(T) phase layer was 100-500 nm, and the grain diameter of the adjacent θ phase was approximately 200 nm. Most of the Fe 4 Al 13 (θ) phase was characterized by a twin formation. Moreover, between the Fe 4 Al 13 (θ) phase layer and the steel, an Fe 2 Al 5 (η) phase layer with a thickness of approximately 2.7 µm was formed. The Fe 2 Al 5 (η) phase grain had a columnar shape (aspect ratio: 1.5) and twin formations in several grains (Figure 8c).       Figure 10 presents the EBSD analysis results for the reaction layer. Figure 10a is the phase map, and Figure 10b is the inverse pole figure (IPF). The reaction layer was formed in the order of the T/θ/η layer on the solidification layer side. The very fine Al 2 Fe 3 Si 3 (τ 1 ) phase and Fe 3 AlC(κ) phases observed by TEM were not detected by EBSD, and were simply classified as Fe 2 Al 5 (η) phase layers. This suggests that EBSD analysis cannot accurately determine the size and distribution of the phase when the thickness of the reaction layer is 1 µm or less and the grain size is small.
Metals 2018, 7, x 10 of 16 Figure 10 presents the EBSD analysis results for the reaction layer. Figure 10a is the phase map, and Figure 10b is the inverse pole figure (IPF). The reaction layer was formed in the order of the T/θ/η layer on the solidification layer side. The very fine Al2Fe3Si3(τ1) phase and Fe3AlC(κ) phases observed by TEM were not detected by EBSD, and were simply classified as Fe2Al5(η) phase layers. This suggests that EBSD analysis cannot accurately determine the size and distribution of the phase when the thickness of the reaction layer is 1 μm or less and the grain size is small. From the TEM observation and EBSD analysis results, it was confirmed that the reaction layer was formed in the order of Al9FeNi(Τ)→Fe4Al13(θ)→Fe2Al5(η) from the solidification side, and the Al2Fe3Si3(τ1) phase was dispersed in the Fe2Al5(η) phase layers. The characteristics of each phase are as follows. The Al9FeNi(Τ) phase had a monoclinic crystal structure, and the grains on the solidification layer side were larger than those on the base material side. The Fe4Al13(θ) phase had a monoclinic crystal structure, and the grain diameter was approximately 100-500 nm. In addition, twin formations were observed. The Fe2Al5(η) phase had an orthorhombic crystal structure, and the grains developed in a columnar shape. Moreover, many Fe2Al5(η) grains contained twin formations. The Al2Fe3Si3(τ1) phase contained a triclinic crystal structure, and several grains with diameters of 100-200 nm were present. The Fe3AlC(κ) phase layer was formed with a thickness of approximately 10-15 nm at the Fe2Al5(η) phase layer/steel interface. Figure 11 presents a schematic diagram depicting the shape of the reaction phases and the reaction layer based on EBSD and TEM analyses of the plated layer. Al, Si, and Al3Ni were the major phases in the solidification layer, and a small Al9FeNi(Τ) phase was present near the reaction layer. The reaction layers were formed in the order of Al9FeNi(Τ), Fe4Al13(θ), and Fe2Al5(η) from the solidification side, and the average thicknesses of the respective phases were 4.7 μm, 500 nm and 2.0 μm. Moreover, Al2Fe3Si3(τ1) was present in Fe2Al5(η). Si was also formed between the Al9FeNi(Τ) phases, and the Fe3AlC(κ) layer was formed at the Fe2Al5(η) layer/steel interface with a thickness of approximately 10-15 nm. Fine twin formations were present in the Fe4Al13(θ) grains, and twin formations were present in several Fe2Al5(η) grains. From the TEM observation and EBSD analysis results, it was confirmed that the reaction layer was formed in the order of Al 9 FeNi(T)→Fe 4 Al 13 (θ)→Fe 2 Al 5 (η) from the solidification side, and the Al 2 Fe 3 Si 3 (τ 1 ) phase was dispersed in the Fe 2 Al 5 (η) phase layers. The characteristics of each phase are as follows. The Al 9 FeNi(T) phase had a monoclinic crystal structure, and the grains on the solidification layer side were larger than those on the base material side. The Fe 4 Al 13 (θ) phase had a monoclinic crystal structure, and the grain diameter was approximately 100-500 nm. In addition, twin formations were observed. The Fe 2 Al 5 (η) phase had an orthorhombic crystal structure, and the grains developed in a columnar shape. Moreover, many Fe 2 Al 5 (η) grains contained twin formations. The Al 2 Fe 3 Si 3 (τ 1 ) phase contained a triclinic crystal structure, and several grains with diameters of 100-200 nm were present. The Fe 3 AlC(κ) phase layer was formed with a thickness of approximately 10-15 nm at the Fe 2 Al 5 (η) phase layer/steel interface. Figure 11 presents a schematic diagram depicting the shape of the reaction phases and the reaction layer based on EBSD and TEM analyses of the plated layer. Al, Si, and Al 3 Ni were the major phases in the solidification layer, and a small Al 9 FeNi(T) phase was present near the reaction layer. The reaction layers were formed in the order of Al 9 FeNi(T), Fe 4 Al 13 (θ), and Fe 2 Al 5 (η) from the solidification side, and the average thicknesses of the respective phases were 4.7 µm, 500 nm and 2.0 µm. Moreover, Al 2 Fe 3 Si 3 (τ 1 ) was present in Fe 2 Al 5 (η). Si was also formed between the Al 9 FeNi(T) phases, and the Fe 3 AlC(κ) layer was formed at the Fe 2 Al 5 (η) layer/steel interface with a thickness of approximately 10-15 nm. Fine twin formations were present in the Fe 4 Al 13 (θ) grains, and twin formations were present in several Fe 2 Al 5 (η) grains. Metals 2018, 7, x 11 of 16 Figure 11. Schematic diagram of the distribution of phases in the solidification and reaction layers.
From the phase diagram, the formation of the reaction layer in the order of Al9FeNi(Τ), Fe4Al13(θ), Fe2Al5(η) in the solidification layer can be understood, in addition to the formation of Al2Fe3Si3(τ1) in Fe2Al5(η). Dipping the steel into a 690 °C Al-7wt%Ni-6wt%Si molten bath resulted in the partial dissolution of the steel. The reaction that occurred during hot-dipping at 690 °C was divided into two reactions: The reaction between the Al-7wt%Ni-6wt%Si melts and dissolved Fe, and the diffusion reaction between the reaction layer formed in the previous reaction and steel. The influence of the increase in the Fe concentration on the phase formed in the Al-7wt%Ni-6%wtSi molten metal was therefore analyzed using a (Al-7wt%Ni-6wt%Si)-xFe pseudo-binary phase diagram. Figure 12 presents the (Al-7wt%Ni-6wt%Si)-xFe pseudo-binary phase diagram obtained using Thermo-Calc™ software. Figure 12 reveals that solid Fe4Al13(θ) can be formed when Fe concentration of over 2.63 wt% is hot-dipped into the Al-7wt%Ni-6wt%Si molten metal at 690 °C. When the Fe concentration was greater than 10.0 wt%, Fe4Al13(θ) and Al9FeNi(Τ) can be formed, and when the Fe concentration was greater than 29.0 wt%, three phases of Fe4Al13(θ), Al3Ni, and Al2Fe3Si3(τ1) can be formed. In this study, Al3Ni was not formed in the reaction layer, and was only present in the solidification layer. Given that Al3Ni was not observed in the reaction layer, it was expected that the Fe concentration in the molten metal near the steel during coating was 29 wt% or lower. In addition, when the Fe concentration in the molten metal increased, θ first crystallized, and then T crystallized as the reaction progressed. This may be why the T layer was formed near the liquid phase, and the Fe4Al13(θ) layer was formed below the Al9FeNi(Τ) layer. The microstructure of the Al-(5-10)wt%Si alloy hot-dipped coating layer without Ni was compared with that of the coating layer in the study, as no previous studies were conducted on the Al-Ni-Si alloy hot-dipped coating. For the Al-(5-10)wt% Si hot-dipped steel, the reaction layer was formed in the order of Al 7 Fe 2 Si(τ 5 ), Fe 4 Al 13 (θ), Fe 2 Al 5 (η) [23,26,27] or Al 7 Fe 2 Si(τ 5 ), and Fe 2 Al 5 (η) [28] in the solidification layer. Moreover, it was reported that Al 2 Fe 3 Si 3 (τ 1 ) was formed within Fe 2 Al 5 (η) in common. Therefore, when Ni was added to the Al-Si coating, Al 9 FeNi(T) was formed instead of Al 7 Fe 2 Si(τ 5 ), Fe 4 Al 13 (θ)/Fe 2 Al 5 (η) was formed in the lower layer, and Al 2 Fe 3 Si 3 (τ 1 ) was formed in Fe 2 Al 5 (η).
From the phase diagram, the formation of the reaction layer in the order of Al 9 FeNi(T), Fe 4 Al 13 (θ), Fe 2 Al 5 (η) in the solidification layer can be understood, in addition to the formation of Al 2 Fe 3 Si 3 (τ 1 ) in Fe 2 Al 5 (η). Dipping the steel into a 690 • C Al-7wt%Ni-6wt%Si molten bath resulted in the partial dissolution of the steel. The reaction that occurred during hot-dipping at 690 • C was divided into two reactions: The reaction between the Al-7wt%Ni-6wt%Si melts and dissolved Fe, and the diffusion reaction between the reaction layer formed in the previous reaction and steel. The influence of the increase in the Fe concentration on the phase formed in the Al-7wt%Ni-6%wtSi molten metal was therefore analyzed using a (Al-7wt%Ni-6wt%Si)-xFe pseudo-binary phase diagram. Figure 12 presents the (Al-7wt%Ni-6wt%Si)-xFe pseudo-binary phase diagram obtained using Thermo-Calc™ software. Figure 12 reveals that solid Fe 4 Al 13 (θ) can be formed when Fe concentration of over 2.63 wt% is hot-dipped into the Al-7wt%Ni-6wt%Si molten metal at 690 • C. When the Fe concentration was greater than 10.0 wt%, Fe 4 Al 13 (θ) and Al 9 FeNi(T) can be formed, and when the Fe concentration was greater than 29.0 wt%, three phases of Fe 4 Al 13 (θ), Al 3 Ni, and Al 2 Fe 3 Si 3 (τ 1 ) can be formed. In this study, Al 3 Ni was not formed in the reaction layer, and was only present in the solidification layer. Given that Al 3 Ni was not observed in the reaction layer, it was expected that the Fe concentration in the molten metal near the steel during coating was 29 wt% or lower. In addition, when the Fe concentration in the molten metal increased, θ first crystallized, and then T crystallized as the reaction progressed. This may be why the T layer was formed near the liquid phase, and the Fe 4 Al 13 (θ) layer was formed below the Al 9 FeNi(T) layer.
1 Figure 12. (Al-7wt%Ni-6wt%Si)-xFe pseudo-binary phase diagram obtained using Thermo-Calc TM software. Table 5 presents the equilibrium phase composition at 690 • C in the (Al-7wt%Ni-6wt%Si)-xFe system obtained using Thermo-Calc™ software. With reference to Table 5, Si was not dissolved in Al 9 FeNi(T). During the formation and development of Al 9 FeNi(T) in the molten metal, Si phases were formed around Al 9 FeNi(T), due to the discharge and agglomeration of Si from Al 9 FeNi(T). The formation mechanism of Fe 2 Al 5 (η) was examined using the (Al-7wt%Ni-6wt%Si)-xFe pseudo-binary phase diagram. Given that the Fe concentration at which the Fe 2 Al 5 (η) was formed at 690 • C was approximately 41.0 wt% (as indicated by the pink rectangular in Figure 12), Fe 2 Al 5 (η) could not be formed by dissolving Fe in the liquid phase during coating at 690 • C. Another mechanism by which Fe 2 Al 5 (η) could be formed as the solid phase diffusion reaction between the already formed reaction phase (Fe 4 Al 13 (θ), Al 9 FeNi(T)) and steel. In particular, Fe 2 Al 5 (η) can be formed through mutual diffusion when Al, Ni, and Si are diffused into the steel in the molten metal→Al 9 FeNi(T)→Fe 4 Al 13 (θ) path, and Fe is diffused through the reaction layer steel Fe 4 Al 13 (θ)→Al 9 FeNi(T)→molten metal path. There are three phases that can be formed on the steel side by mutual diffusion in the state diagram, namely, FeAl(β), Fe 2 Al 5 (η), and Al 2 Fe 3 Si 3 (τ 1 ).
In this diffusion mechanism, the phase with the smallest concentration difference in comparison with Fe 4 Al 13 (θ) is easily formed. Therefore, Fe 2 Al 5 (η) in Table 5 is most similar to Fe 4 Al 13 (θ) and Al and Fe concentrations. In Figure 9 b, it can be seen that the grain size of Fe 4 Al 13 (θ) adjacent to Fe 2 Al 5 (η) was significantly smaller than that of Fe 4 Al 13 (θ) adjacent to the Al 9 FeNi(T) layer. This is expected that the Fe 4 Al 13 (θ) grain decreased in size and disappeared, as it was transformed into Fe 2 Al 5 (η) by the diffusion reaction.
The mechanism by which Al 2 Fe 3 Si 3 (τ 1 ) was produced within Fe 2 Al 5 (η) is as follows. Al 9 FeNi(T) and Fe 4 Al 13 (θ) formed at Fe concentrations below 30 wt% did not dissolve Si, as can be seen in Table 5. Therefore, when Al 9 FeNi(T) and Fe 4 Al 13 (θ) discharged Si into the liquid phase, the Si phase was formed in the solidification layer. However, when Si was discharged toward the steel, Al 2 Fe 3 Si 3 (τ 1 ) phases that contained Si were precipitated. In Figures 6 and 7, the significant formation of Al 2 Fe 3 Si 3 (τ 1 ) phases at the Fe 4 Al 13 (θ)/Fe 2 Al 5 (η) interface can be explained by the formation mechanism of the Al 2 Fe 3 Si 3 (τ 1 ) phases.
Another reaction phase is the Fe 3 AlC(κ) formed at the Fe 2 Al 5 (η)/steel interface. Springer recently reported that an 800 nm band that consists of FeAl(β') and Fe 3 AlC(κ) phases was formed between the steel and Fe 2 Al 5 (η) phase by inter-diffusion in the solid reaction of low-carbon steel and pure Al at 600 • C; however, there was no mention of the formation mechanism [27]. In Figure 13, carbon was not detected in the Al 9 FeNi(T) and Fe 2 Al 5 (η) layers. This confirms that the carbon solubility of Al 9 FeNi(T) and Fe 2 Al 5 (η) was very low. Given that the boron steel used in the study contained 0.22 wt% of carbon, it was suggested that the Fe3AlC(κ) layer was formed by carbon accumulation that discharged in the Fe 2 Al 5 (η) region transformed from steel to Fe 2 Al 5 (η). A study of pure Al [29] and the Al-7Ni [15] hot-dip coating on boron steels similarly reported the formation of a Fe 3 AlC(κ) layer at the grain boundaries of Fe 2 Al 5 (η) and adjacent steels. significantly smaller than that of Fe4Al13(θ) adjacent to the Al9FeNi(Τ) layer. This is expected that the Fe4Al13(θ) grain decreased in size and disappeared, as it was transformed into Fe2Al5(η) by the diffusion reaction. The mechanism by which Al2Fe3Si3(τ1) was produced within Fe2Al5(η) is as follows. Al9FeNi(Τ) and Fe4Al13(θ) formed at Fe concentrations below 30 wt% did not dissolve Si, as can be seen in Table  5. Therefore, when Al9FeNi(Τ) and Fe4Al13(θ) discharged Si into the liquid phase, the Si phase was formed in the solidification layer. However, when Si was discharged toward the steel, Al2Fe3Si3(τ1) phases that contained Si were precipitated. In Figures 6 and 7, the significant formation of Al2Fe3Si3(τ1) phases at the Fe4Al13(θ)/Fe2Al5(η) interface can be explained by the formation mechanism of the Al2Fe3Si3(τ1) phases.
Another reaction phase is the Fe3AlC(κ) formed at the Fe2Al5(η)/steel interface. Springer recently reported that an 800 nm band that consists of FeAl(β') and Fe3AlC(κ) phases was formed between the steel and Fe2Al5(η) phase by inter-diffusion in the solid reaction of low-carbon steel and pure Al at 600 °C; however, there was no mention of the formation mechanism [27]. In Figure 13, carbon was not detected in the Al9FeNi(Τ) and Fe2Al5(η) layers. This confirms that the carbon solubility of Al9FeNi(Τ) and Fe2Al5(η) was very low. Given that the boron steel used in the study contained 0.22 wt% of carbon, it was suggested that the Fe3AlC(κ) layer was formed by carbon accumulation that discharged in the Fe2Al5(η) region transformed from steel to Fe2Al5(η). A study of pure Al [29] and the Al-7Ni [15] hot-dip coating on boron steels similarly reported the formation of a Fe3AlC(κ) layer at the grain boundaries of Fe2Al5(η) and adjacent steels. During the formation of the reaction layer, It is expected that streaks occurred in the Fe4Al13(θ) and Fe2Al5(η) phases, as can be seen in Figure 9b. Streaks in each θ and some η grain are observed at regular intervals throughout the grain, not in only one specific part. And it also has a different orientation, like each grain's crystal orientation. Hence, streaks are expected to twin. Takata et al. reported that during the interfacial reaction between liquid-Al and solid-Fe, strain developed due to the volume expansion in the process of transformation from α-ferrite to the η phase, which resulted in the refinement of the η phase in contact with the liquid phase [30]. In this study, it was considered that the twin formation in Fe4Al13(θ) and Fe2Al5(η) between the Al9FeNi(Τ) and steel was caused by the strain generated by the lattice transformation (Al9FeNi(Τ)→Fe4Al13(θ), Fe4Al13(θ)→Fe2Al5(η)) that occurred in the layer interface. Moreover, T and θ have the same monoclinic structure, but their space group is different (Al9FeNi(Τ): P21/C [31], Fe4Al13(θ): C1/m1(12) [32]). The twin formation can be attributed to the expansion caused by the Al9FeNi(Τ)→Fe4Al13(θ) transformation.

Conclusion
The microstructures of the reaction layer and phases formed on boron steel hot-dipped in Al-7wt%Ni-6wt%Si were investigated using FIB, TEM, and EBSD. The formation mechanism of each phase was examined by employing a (Al-7wt%Ni-6wt%Si)-xFe pseudo-binary phase diagram obtained using Thermo-Calc TM software. The results are as follows.
1) The solidification layer consisted mainly of Al, Al3Ni, and Si phases, and a small Al9FeNi(T) phase was observed near the reaction layer. By analyzing the solidification process using a phase diagram, it was found that three phases (Al, Al3Ni and Si) were formed in the low Fe During the formation of the reaction layer, It is expected that streaks occurred in the Fe 4 Al 13 (θ) and Fe 2 Al 5 (η) phases, as can be seen in Figure 9b. Streaks in each θ and some η grain are observed at regular intervals throughout the grain, not in only one specific part. And it also has a different orientation, like each grai's crystal orientation. Hence, streaks are expected to twin. Takata et al. reported that during the interfacial reaction between liquid-Al and solid-Fe, strain developed due to the volume expansion in the process of transformation from α-ferrite to the η phase, which resulted in the refinement of the η phase in contact with the liquid phase [30]. In this study, it was considered that the twin formation in Fe 4 Al 13 (θ) and Fe 2 Al 5 (η) between the Al 9 FeNi(T) and steel was caused by the strain generated by the lattice transformation (Al 9 FeNi(T)→Fe 4 Al 13 (θ), Fe 4 Al 13 (θ)→Fe 2 Al 5 (η)) that occurred in the layer interface. Moreover, T and θ have the same monoclinic structure, but their space group is different (Al 9 FeNi(T): P21/C [31], Fe 4 Al 13 (θ): C1/m1(12) [32]). The twin formation can be attributed to the expansion caused by the Al 9 FeNi(T)→Fe 4 Al 13 (θ) transformation.

Conclusions
The microstructures of the reaction layer and phases formed on boron steel hot-dipped in Al-7wt%Ni-6wt%Si were investigated using FIB, TEM, and EBSD. The formation mechanism of