Phase Composition of Al-Si Coating from the Initial State to the Hot-Stamped Condition

The chemical and phase composition of the coating and the coating/substrate interface of an Al-Si-coated 22MnB5 hot stamped steel was investigated by means of SEM-EDS, XRD, micro-XRD and electron diffraction. Moreover, the surface profile was analyzed by XPS and roughness measurements. The XPS measurements showed that the thickness of the Si and Al oxide layers increased from 14 to 76 nm after die-quenching, and that the surface roughness increased as well as a result of volume changes caused by phase transformations. In addition to the FeAl(Si) and Fe2Al5 phases and the interdiffusion layer forming complex structures in the coating, electron diffraction confirmed the presence of an Fe2Al5 phase, and also revealed very thin layers of Fe3(Al,Si)C, Fe2(Al,Si)5 and Al-bearing rod-shaped particles in the immediate vicinity of the steel interface. Moreover, the scattered nonuniform layer of the Fe2Al8Si phase was identified in the outermost layer of the coating. Despite numerous studies devoted to researching the phase composition of the Al-Si coating applied to hot stamped steel, electron diffraction revealed very thin layers and particles on the coating/substrate interface and outermost layer, which have not been analyzed in detail.


Introduction
With incessantly increasing pressure on the automotive industry, manufacturers are being forced to develop new technologies and materials to reduce car body weight. This can be achieved by hot stamping technology capable of producing press-hardened steel (PHS), reaching an ultimate tensile strength of up to 2000 MPa. Hot stamping significantly suppresses the spring-back effect compared to conventional cold press forming, and ensures the geometric accuracy of pressed parts. PHS is used mostly to produce safety components such as A-pillars, B-pillars, bumpers, roof rails, rocker rails and tunnels. Due to its extremely high strength, PHS enables sheet thickness reduction and helps to fulfill the constantly stricter standards of CO 2 emissions. Moreover, PHS enhances the toughness of the vehicle frame and passenger safety. Therefore, its proportional content in car bodies attracts increasing attention [1][2][3].
Hot stamping technology offers two different variants of processing: (i) direct hot stamping, and (ii) indirect hot stamping. In both variants, the input material is low-alloyed steel with a ferritic/pearlitic microstructure and a strength of approximately 600 MPa [2].
to the steel substrate [2,6,19]. Windmann et al. [6,17] identified the layers in the coating at the substrate/coating interface as follows: the interdiffusion layer was formed by the solid solution α-Fe(Al,Si) adjacent to the FeAl continuous layer, and the matrix of the coating consisting of Fe 2 Al 5 and the FeAl phase forming a continuous layer or separate islands was embedded in the Fe 2 Al 5 matrix. Gui et al. [20] and Liang et al. [21], using very similar processing conditions, also observed an α-Fe(Al,Si) interdiffusion layer and the FeAl phase was dispersed, although in the Fe 2 Al matrix of the coating. Furthermore, the continuous layer or island-like phase, denoted as FeAl in previous works, was described as Fe 2 SiAl 2 (τ 1 ) and Fe 2 Si 2 Al 5 (τ 2 ) in Refs. [5,22,23], with an interdiffusion layer formed of Fe 3 Al [5,23]. The most recent work from Cho et al. [24] stated that the three-phase interdiffusion layer consisted of α-Fe(Al), Fe 3 Al, FeAl, the matrix of Fe 2 Al 5 , and a continuous layer or islandlike phase distributed in the matrix as a mixture of FeAl and an unidentified Fe x Al x Si z ternary phase.
Despite these numerous studies of the coating, the findings of the resulting phase composition, coating/substrate interface, and void distribution in the coating are inconsistent. This discrepancy might be caused by the presence of non-stoichiometric phases in the Al-Fe system with different solubilities of Si and a rather high intricacy of Al-Fe and Al-Fe-Si systems. Therefore, SEM-EDS and XRD analyses might struggle to identify the phase composition and the presence of fine layers and particles. Moreover, a detailed TEM analysis of the coating/substrate interface and phases across the coating is still missing, to our best knowledge. As such, this work is focused on the TEM-SAED analysis of the coating/substrate interface, and the description of phases through the coating, as well as an analysis of the top surface layer by means of XPS and roughness measurements to increase the general knowledge of Al-Si coatings.

Materials and Methods
Commercially available manganese-boron steel (22MnB5) with a ferritic/pearlitic microstructure and Al-Si hot-dipped coating was used in the present study. The steel sheet had a thickness of 1.5 mm and was double-side-coated (150 g/m 2 ). The Al-Si coating was commercially produced by continuous immersion in a bath with an approximate chemical composition of 90 wt.% Al and 10 wt.% Si. The as-received steel sheets with dimensions of 200 × 300 mm were first cut into smaller specimens (80 × 40 mm) for further investigation. The chemical composition of the as-received steel (Table 1) was measured with an optical emission spectrometry (OES) system (Bruker G8 Galileo, Billerica, Massachusetts, MA, USA). The specimens were heated up to a temperature of 920 • C in an electrical resistance furnace (Laboratorní pece Martínek, Kladno, Czech Republic) with an 8 min dwell time, close to the industry conditions. After that, the specimens were transported (2 s) and die-quenched to achieve martensitic transformation without any imposed deformation. The steel die was sufficiently robust to achieve the martensitic transformation and was air-cooled. The samples used for microstructural observation were prepared by a standard metallography procedure, which included grinding the samples with SiC abrasive papers (P320-P2000), polishing with polycrystalline diamond suspensions (9-1 µm), and etching the samples in 5% nital [25]. The microstructure was observed using an optical microscope (Olympus PME-3, Olympus, Tokyo, Japan) and a scanning electron microscope (Tescan Vega 3 LMU, 20 kV, SE + BSE detectors, TESCAN, Brno, Czech Republic) equipped with an energy dispersive spectroscopy detector (Oxford Instruments INCA 350, 20 mm 2 , Oxford Instruments, Abingdon, England). For further phase analysis, X-ray diffraction data were collected at room temperature with an X'Pert3 Powder θ-θ powder diffractometer (PANanalytical, Almelo, Netherlands) with a parafocusing Bragg-Brentano geometry using CuK α radiation (λ = 0.15418 nm, U = 40 kV, I = 30 mA). The micro-X-ray diffraction data were obtained by a D8 Discover microdiffractometer (Bruker, Billerica, Massachusetts, MA, USA) with a 0.1 mm collimator at room temperature in a parallel geometry using the wavelength of CoK α radiation (λ = 0.27903 nm, U = 35 kV, I = 40 mA). Two Debye-Scherrer frames were obtained in the θ-θ geometry for 2θ = 26 • and 56 • with a VÅNTEC-500 2D detector (Bruker, Billerica, Massachusetts, MA, USA) and a measurement time of 30 min per frame. The thicknesses of the coatings and interfacial layers in the hot-dipped condition and after austenitization and die-quenching were measured by image analysis using ImageJ software 1.8.0_172. The measured thickness is an average value from at least 60 measurements. The coating surface of the as-received 22MnB5 in the hot-dipped condition and after die-quenching was studied using an X-ray photoelectron spectroscopy (XPS) system (ESCAprobe P, Omicron Nanotechnology Ltd., Taunusstein, Germany) equipped with an Al K α (λ = 1486.7 eV) X-ray source. The spectra were measured with an energy step size of 0.05 eV and normalized to the binding energy of the C1s peak (285.0 eV). The data used for the chemical state evaluation were obtained from the NIST X-ray Photoelectron Spectroscopy Database. The roughness of the coated samples was measured with a distance of 17.5 mm, a cutoff of 2.5 mm, and a 2 µm sensor tip, and a Keyence VHX-7000 Series digital microscope (Keyence, Osaka, Japan) was used for the roughness measurements and surface topology profile.
TEM observations were carried out by using a Philips Inc. CM-20 ® microscope (Amsterdam, The Netherlands) operated at 200 kV with a double tilt specimen holder equipped with a liquid nitrogen cooling stage. TEM discs were prepared by cross-sectioning the two sections together that were embedded in an alcohol-based resin. The two sections were embedded in a 2 mm inner diameter pure copper tube with a thickness of 1 mm. This wafer material was cut to a thickness of 200 µm, grinded, and polished down to 100-90 µm. Then, this material was dimpled on both sides to obtain a final central thickness ranging from 20 to 25 µm. The final preparation method used to achieve electron transparency was performed by a Gatan Inc. precision ion polishing system (PIPS) (Pleasanton, CA, USA) with double Ar + flux set at an initial incident angle of 8 • , which was operated for a few minutes, and then the angle was changed to 6 • and finally 4 • . PIPS was performed for 2 h, and the sample was kept cool by flowing liquid nitrogen beneath the sample stage. Phase identification was performed by TEM by indexing the selected area electron diffraction (SAEDP) patterns. To properly obtain SAEDPs from the small particles and small areas, a converged beam (CB) was used.

Chemical and Phase Composition of the As-Received 22MnB5 in the Hot-Dipped Condition
The hot-dipped Al-Si coating was first investigated as the initial state for austenitization and die-quenching by means of SEM-EDS, XRD and micro-XRD. The as-received coating had a thickness of 27.3 ± 3.7 µm. The top surface of the coating (Figure 1a) was covered by a very thin oxide layer enriched in Al and Si, as a result of these materials' high affinity for oxygen. The matrix of the coating was composed of an Al-Si mixture with a nearly eutectic composition. The silicon particles cannot be clearly distinguished from the α-Al solid solution using a back-scattered electron detector because of the close atomic numbers of Al and Si. However, the regions enriched with Si were clearly visible in the EDS elemental distribution maps (Figure 2), and the presence of Si within the coating was also confirmed by XRD analysis, as was the presence of Al (Figure 3, Si JCPDS card no. 04-016-4861, Al JCPDS card no. 01-072-3440). The needle and platelet-like particles, as well as the continuous layer present at the interface, with an approximate thickness of 3.6 ± 0.6 µm, were formed by the Al-Si-Fe ternary phase.          According to the chemical composition determined by SEM-EDS (Table 2), this ternary phase can be identified as Fe 2 Al 8 Si (also denoted as τ 5 ) with a hexagonal structure and the P63/mmc space group [18]. The presence of the τ 5 phase was also confirmed by XRD and micro-XRD analysis ( Figure 3, JCPDS card no 00-020-0030). The irregular shape of the τ 5 /Al-Si coating interface can be explained by the solidification in the Al-Si melt [6,26]. These findings are in good agreement with the literature [4,6,18,20,26]. The purpose of adding Si to the hot-dipping bath is to form this ternary τ 5 inhibition layer, which impedes the further diffusion of Fe into the coating, and Al in the opposite direction, thus forming a thick layer of intermetallics, especially the Fe 2 Al 5 brittle intermetallic phase. Furthermore, the addition of Si increases the resistance to both corrosion and elevated temperatures, which is strongly desirable during hot stamping (at temperatures in the 900-930 • C range). Despite the τ 5 inhibition layer, a thin sublayer reaching thicknesses of 0.7 ± 0.2 µm was formed directly at the steel interface. According to the chemical composition (point 5, Figure 1b), this sublayer was identified as η-Fe 2 Al 5 with an orthorhombic structure and the Cmcm space group, which has also been mentioned in Refs. [4][5][6]18,20,27,28]. The content of Si (Table 2) in this phase corresponded to that observed by Windmann et al. [6], who reported an even slightly higher content of Si in the η-Fe 2 Al 5 phase, reaching up to 6.1 at.%. In addition, a very thin light layer containing elements with higher atomic numbers, such as Si, was observed within the η-Fe 2 Al 5 phase (marked with a dashed red frame in Figure 1b). However, this phase was below the resolution limits of the EDS detector, and its volume fraction was low, thus it was not detected by XRD either. This phase was also mentioned in the works of other authors [6,20,27] who identified it as Fe 3 Al 2 Si 3 (further denoted as τ 1 ) with a triclinic structure. Its formation was promoted by the addition of Si into the hot dipping bath, and resulted from the low solid solubility of Si in η-Fe 2 Al 5 . Contrary to this, the authors in Refs. [4,5] identified, in addition to the Fe 2 Al 5 and Fe 2 Al 8 Si phases, a FeAl 3 phase. The identification was based on the chemical composition determined by SEM-EDS. The presence of FeAl 3 was also described in Ref. [26], where it formed isolated particles or a continuous layer depending on the dipping time within Fe 2 Al 5 . Note, however, that FeAl 3 was found after hot-dipping in a bath containing a significantly lower Si content (up to 2%) for an immersion time that was considerably longer than the typical operating time. Shin et al. [18] studied the microstructural evolution of hot-dipped Al-Si coating on 22MnB5 boron steel, and stated that the intermetallic sublayer at the steel substrate interface consisted of Fe 2 Al 5 and Fe 3 Al 2 Si 3 when the Si content exceeded 5 wt.%. The authors also provided a TEM analysis of the phase evolution during hot-dip aluminizing. The analyses confirmed the presence of the major τ 5 phase (hexagonal structure) with grains elongated in the direction of the heat flow and η-Fe 2 Al 5 (orthorhombic structure), which was separated by a continuous layer of Fe 3 Al 2 Si 3 (τ 1 ) with a triclinic structure. Moreover, the authors observed the presence of a carbon-enriched zone at the steel intermetallic layer interface, which was also discussed in Ref. [26], and identified this layer as the Fe 3 AlC carbide zone.

Chemical and Phase Evolution after Austenitization and Die-Quenching
The as-received coating with a thickness of 27.3 ± 3.7 µm and the composition of a eutectic mixture of α-Al + Si, Fe 2 Al 8 Si (τ 5 ), Fe 2 Al 5 and Fe 3 Al 2 Si 3 (τ 1 ) phases underwent significant transformation during austenitization and die-quenching. The coating thickness increased to 36.8 ± 3.1 µm, and the coating formed from a multilayered system as a result of the strong diffusion of Fe into the coating, and Al and Si into the steel substrate ( Figure 4). The cracks reaching almost to the steel substrate and voids were further present in the coating, which will be discussed later.

Chemical and Phase Evolution after Austenitization and Die-Quenching
The as-received coating with a thickness of 27.3 ± 3.7 µm and the composition of a eutectic mixture of α-Al + Si, Fe2Al8Si (τ5), Fe2Al5 and Fe3Al2Si3 (τ1) phases underwent significant transformation during austenitization and die-quenching. The coating thickness increased to 36.8 ± 3.1 µm, and the coating formed from a multilayered system as a result of the strong diffusion of Fe into the coating, and Al and Si into the steel substrate ( Figure  4). The cracks reaching almost to the steel substrate and voids were further present in the coating, which will be discussed later.  The influence of austenitization and die-quenching on the chemical composition and microstructure of the coating was first investigated by SEM-EDS analysis. The chemical composition of each of the structural components found in the layer (marked in Figure  4a) is shown in Table 3. The top-most layer of the coating was enriched with oxygen, as is clearly visible from the SEM-EDS map in Figure 5. The matrix of the coating was attributed to the chemical composition of the Fe2Al5 phase (point 3, Table 3). The content of Si in the Fe2Al5 phase was homogeneously distributed in the range of 1.9-2.7 at.%, indicating the relatively poor solubility of Si in this phase with an orthorhombic structure. Other studies [4,20,29], however, identified the matrix of the coating as FeAl2, and the presence of both FeAl2 and Fe2Al5 phases was even observed. The predominant formation of the Fe2Al5 phase can be ascribed to a kinetic factor associated with a highly open structural arrangement of the atoms of this phase. The c-axis of its orthorhombic structure is described to contain only Al atoms, of which about 30% are absent. This large number of vacancies enables a greater diffusion rate of reactant species than in FeAl2, and causes directional growth along this c-axis [26,30]. Authors in [26] described the Fe2Al5 phase as a dominant reaction product at the Fe/Al melt interface following the parabolic growth of this phase in a temperature range of 715-944 °C. The inhibiting effect of Si on the growth of Fe2Al5 is well known from hot-dip aluminizing. Except for the formation of Al-Si-Fe ternary phases retarding the diffusion of Al and Fe, and thus Fe2Al5 growth, it was suggested that Si occupies a large number of vacancies on the c-axis and blocks the fast diffusion path of reactant species. Nevertheless, in the present work, the measured Si content was relatively low, reaching from 1.9 to 2.7 at.% [26,31]. The influence of austenitization and die-quenching on the chemical composition and microstructure of the coating was first investigated by SEM-EDS analysis. The chemical composition of each of the structural components found in the layer (marked in Figure 4a) is shown in Table 3. The top-most layer of the coating was enriched with oxygen, as is clearly visible from the SEM-EDS map in Figure 5. The matrix of the coating was attributed to the chemical composition of the Fe 2 Al 5 phase (point 3, Table 3). The content of Si in the Fe 2 Al 5 phase was homogeneously distributed in the range of 1.9-2.7 at.%, indicating the relatively poor solubility of Si in this phase with an orthorhombic structure. Other studies [4,20,29], however, identified the matrix of the coating as FeAl 2 , and the presence of both FeAl 2 and Fe 2 Al 5 phases was even observed. The predominant formation of the Fe 2 Al 5 phase can be ascribed to a kinetic factor associated with a highly open structural arrangement of the atoms of this phase. The c-axis of its orthorhombic structure is described to contain only Al atoms, of which about 30% are absent. This large number of vacancies enables a greater diffusion rate of reactant species than in FeAl 2 , and causes directional growth along this c-axis [26,30]. Authors in [26] described the Fe 2 Al 5 phase as a dominant reaction product at the Fe/Al melt interface following the parabolic growth of this phase in a temperature range of 715-944 • C. The inhibiting effect of Si on the growth of Fe 2 Al 5 is well known from hot-dip aluminizing. Except for the formation of Al-Si-Fe ternary phases retarding the diffusion of Al and Fe, and thus Fe 2 Al 5 growth, it was suggested that Si occupies a large number of vacancies on the c-axis and blocks the fast diffusion path of reactant species. Nevertheless, in the present work, the measured Si content was relatively low, reaching from 1.9 to 2.7 at.% [26,31].   The island-like phase (labeled with points 1, 2, and 4 in Figure 4a) contained from 52.1 to 41.4 at.% Al, and 41.5 to 46.6 at.% Fe. According to the chemical composition, this phase was identified as the FeAl(Si) phase. The decreasing content of Al from the top of the coating to the substrate interface and the increasing content of Fe in the opposite manner can be explained by the diffusion path of Al to the steel substrate, and Fe to the coating during austenitization. The measured content of Si in the FeAl(Si) phase was in the range of 5.9-11.1 at.%, which was significantly higher than in Fe2Al5. The close positions of Al and Si with similar atomic radii allows for the easy substitution of these elements in cubic BCC (B2) crystalline structures [28,32]. Similar contents of Si in the FeAl(Si) phase were also observed in Refs. [6,21]. The higher solid solubility of Si and the Fe-enrichment of the coating by the diffusion of Al towards the steel substrate, and Fe into the coating, with elongating dwell time were stated as the reasons for the FeAl(Si) phase's formation in Refs. [6,17]. Cho et al. [24] noted that the island-like phase, in addition to the FeAl phase, was formed by an adjacent FexAlySiz ternary phase that was not identified due to its fine nature and probably low symmetry.
A relatively thick layer, which was also described as an interdiffusion layer with an average thickness of 11.5 ± 1.0 µm, was formed at the substrate/coating interface with two distinguishable sublayers. The first sublayer on the top (labeled with point 5 in Figure 4a) had a chemical composition similar to the compositions at points 1, 2 and 4, which corresponded to the presence of the FeAl(Si) phase. Its occurrence at the Fe2Al5 interdiffusion layer interface can be explained by the binary Al-Fe phase diagram ( Figure 6). The diffusion of Fe towards the coating and of Al to the steel substrate led to the Fe-enrichment of the interface, resulting in the transformation of Fe2Al5 to the FeAl(Si) iron-richer phase. The sublayer closer to the steel interface (labeled with points 6 and 7 in Figure 4a) is described in the majority of the literature [6,17,[20][21][22]29] as an α-Fe(Al,Si) solid solution, which is reasonable, since Al and Si are strong ferrite stabilizers that diffuse to the steel substrate during austenitization. However, some studies [5,23] identified this layer as Fe3Al, which has also been reported by others [4,32], wherein the authors described the phase transformation of high-temperature α-Fe(Al, Si) to the low-temperature stable Fe3Al (D03) ordered phase. In the present study, the concentration gradient of Al and Si in this sublayer (α-Fe(Al, Si)/Fe3Al) was noticed, which is shown in the SEM-EDS map in The island-like phase (labeled with points 1, 2, and 4 in Figure 4a) contained from 52.1 to 41.4 at.% Al, and 41.5 to 46.6 at.% Fe. According to the chemical composition, this phase was identified as the FeAl(Si) phase. The decreasing content of Al from the top of the coating to the substrate interface and the increasing content of Fe in the opposite manner can be explained by the diffusion path of Al to the steel substrate, and Fe to the coating during austenitization. The measured content of Si in the FeAl(Si) phase was in the range of 5.9-11.1 at.%, which was significantly higher than in Fe 2 Al 5 . The close positions of Al and Si with similar atomic radii allows for the easy substitution of these elements in cubic BCC (B2) crystalline structures [28,32]. Similar contents of Si in the FeAl(Si) phase were also observed in Refs. [6,21]. The higher solid solubility of Si and the Fe-enrichment of the coating by the diffusion of Al towards the steel substrate, and Fe into the coating, with elongating dwell time were stated as the reasons for the FeAl(Si) phase's formation in Refs. [6,17]. Cho et al. [24] noted that the island-like phase, in addition to the FeAl phase, was formed by an adjacent Fe x Al y Si z ternary phase that was not identified due to its fine nature and probably low symmetry.
A relatively thick layer, which was also described as an interdiffusion layer with an average thickness of 11.5 ± 1.0 µm, was formed at the substrate/coating interface with two distinguishable sublayers. The first sublayer on the top (labeled with point 5 in Figure 4a) had a chemical composition similar to the compositions at points 1, 2 and 4, which corresponded to the presence of the FeAl(Si) phase. Its occurrence at the Fe 2 Al 5 interdiffusion layer interface can be explained by the binary Al-Fe phase diagram ( Figure 6). The diffusion of Fe towards the coating and of Al to the steel substrate led to the Fe-enrichment of the interface, resulting in the transformation of Fe 2 Al 5 to the FeAl(Si) iron-richer phase. The sublayer closer to the steel interface (labeled with points 6 and 7 in Figure 4a) is described in the majority of the literature [6,17,[20][21][22]29] as an α-Fe(Al,Si) solid solution, which is reasonable, since Al and Si are strong ferrite stabilizers that diffuse to the steel substrate during austenitization. However, some studies [5,23] identified this layer as Fe 3 Al, which has also been reported by others [4,32], wherein the authors described the phase transformation of high-temperature α-Fe(Al, Si) to the low-temperature stable Fe 3 Al (D0 3 ) ordered phase. In the present study, the concentration gradient of Al and Si in this sublayer (α-Fe(Al, Si)/Fe 3 Al) was noticed, which is shown in the SEM-EDS map in Figure 5, and the chemical composition is shown in Table 3. Furthermore, the Si-enriched area was observed within the top sublayer of the interdiffusion layer (point 5) extending further towards the steel, which is in accordance with Ref. [6]. Nevertheless, we cannot clearly confirm whether both phases (α-Fe and Fe 3 Al) are present based solely on the chemical composition. The micro-XRD analysis revealed the presence of Fe 2 Al 5 ; however, other phases were not detected by micro-XRD or XRD analysis due to the large thickness of the coating and due to the relatively low volume fraction of the phases.  Table 3. Furthermore, the Si-enriched area was observed within the top sublayer of the interdiffusion layer (point 5) extending further towards the steel, which is in accordance with Ref. [6]. Nevertheless, we cannot clearly confirm whether both phases (α-Fe and Fe3Al) are present based solely on the chemical composition. The micro-XRD analysis revealed the presence of Fe2Al5; however, other phases were not detected by micro-XRD or XRD analysis due to the large thickness of the coating and due to the relatively low volume fraction of the phases. As previously mentioned, the voids present in the coating were located in the interdiffusion layer and beneath the top surface of the coating. The mechanism of void formation in the interdiffusion layer is correlated with the Kirkendall effect [4,6,20,29]. During the first few minutes of austenitization, the Fe diffusion is more pronounced, which is supported by the Al-Si coating melting at its eutectic temperature (575 °C) [6,20,26,34]. Subsequently, Fe enrichment leads to the solidification of the partially melted material, allowing Al-Fe-Si intermetallics to form with higher melting temperatures, especially the Fe2Al5 phase, because of the extremely rapid growth along the c-axis compared to other phases [5,17]. Therefore, the further diffusion of Fe, Al and Si atoms occurs in the solid state, where the diffusion of Al atoms becomes more active towards the steel substrate. The higher diffusion rate of Al can be attributed to the high mobility of Al atoms along the c-axis of the orthorhombic Fe2Al5 phase with high vacancy concentration, which is responsible for the void formation [26,35]. This corroborates Windmann et al.'s results [6], as they observed the formation of voids in a fully solidified coating composed of Al-Fe-Si intermetallic phases as well as their increasing volume fraction, which accompanied the increasing austenitization temperature. The authors also observed that the diffusion coefficients of aluminum in the Al-Fe-rich intermetallic phases formed in the coating were larger than those of Fe. The faster self-diffusion of Al than Fe in three different Fe-Al alloys (25.5 at.%, 33.0 at.%, 48.0 at.% Al) was also described in Ref. [36]. Therefore, it might be concluded that the Kirkendall voids are caused by the diffusion of Al into the steel substrate, being more pronounced than Fe diffusion in the opposite direction, considering diffusion in the solid state. The voids beneath the top surface of the coating were probably As previously mentioned, the voids present in the coating were located in the interdiffusion layer and beneath the top surface of the coating. The mechanism of void formation in the interdiffusion layer is correlated with the Kirkendall effect [4,6,20,29]. During the first few minutes of austenitization, the Fe diffusion is more pronounced, which is supported by the Al-Si coating melting at its eutectic temperature (575 • C) [6,20,26,34]. Subsequently, Fe enrichment leads to the solidification of the partially melted material, allowing Al-Fe-Si intermetallics to form with higher melting temperatures, especially the Fe 2 Al 5 phase, because of the extremely rapid growth along the c-axis compared to other phases [5,17]. Therefore, the further diffusion of Fe, Al and Si atoms occurs in the solid state, where the diffusion of Al atoms becomes more active towards the steel substrate. The higher diffusion rate of Al can be attributed to the high mobility of Al atoms along the c-axis of the orthorhombic Fe 2 Al 5 phase with high vacancy concentration, which is responsible for the void formation [26,35]. This corroborates Windmann et al.'s results [6], as they observed the formation of voids in a fully solidified coating composed of Al-Fe-Si intermetallic phases as well as their increasing volume fraction, which accompanied the increasing austenitization temperature. The authors also observed that the diffusion coefficients of aluminum in the Al-Fe-rich intermetallic phases formed in the coating were larger than those of Fe. The faster self-diffusion of Al than Fe in three different Fe-Al alloys (25.5 at.%, 33.0 at.%, 48.0 at.% Al) was also described in Ref. [36]. Therefore, it might be concluded that the Kirkendall voids are caused by the diffusion of Al into the steel substrate, being more pronounced than Fe diffusion in the opposite direction, considering diffusion in the solid state. The voids beneath the top surface of the coating were probably caused by the diffusion of Al to form Al 2 O 3 , and the reaction with the furnace atmosphere [4].

TEM Characterization and Coating Phase Identification
All the different Al-Si coating layers were identified and then characterized by TEM. In this regard, at least four chemically different zones were already identified by scanning electron microscopy ( Figure 4). As such, TEM layer characterization was carried out starting from the hardened martensite structure of the 22MnB5 steel, and then proceeding through the Al-Si coating, ending at the outermost coating zones. Figure 7 reports a representative bright-field (BF) image showing the typical martensitic structure of the coated steel. This was mainly composed of martensite laths with 70 ± 20 nm thicknesses (Figure 7 inset). caused by the diffusion of Al to form Al2O3, and the reaction with the furnace atmosphere [4].

TEM Characterization and Coating Phase Identification
All the different Al-Si coating layers were identified and then characterized by TEM. In this regard, at least four chemically different zones were already identified by scanning electron microscopy ( Figure 4). As such, TEM layer characterization was carried out starting from the hardened martensite structure of the 22MnB5 steel, and then proceeding through the Al-Si coating, ending at the outermost coating zones. Figure 7 reports a representative bright-field (BF) image showing the typical martensitic structure of the coated steel. This was mainly composed of martensite laths with 70 ± 20 nm thicknesses (Figure  7 inset). The martensite directly in contact with the coating revealed the presence of some Albearing rod-shaped particles whose stoichiometry was fully compatible with Fe3Al. This was determined by CB-SAEDP, and the related representative image and diffraction pattern are reported in Figure 8. The very fine nature of the Fe3Al rod-shaped particles can be clarified with a closer look at the Fe-Al phase diagram ( Figure 6). The Fe3Al phase is stable in a concentration range from 22.5 at.% to 36.5 at.% Al and transformed to FeAl or α-Fe at temperatures above 545 °C [24]. Considering that the typical temperature is 800-850 °C at the beginning of hot stamping and the die-quenching in the tool is maintained usually for 8-12 s [37], there is not enough time for transformation and phase growth, resulting in the very fine nature of the rod-shaped Fe3Al particles. The transformation of α-Fe to Fe3Al was described in Ref. [32] after austenitization at 1050 °C, deformation and cooling. Furthermore, the gradual transition from the bcc crystal structure of α-Fe to D03 of Fe3Al and to B2 of FeAl at the substrate/coating interface was reported by Cho et al. [24]. However, the authors did not observe any evidence of interfaces between these three different crystal structures. The martensite directly in contact with the coating revealed the presence of some Al-bearing rod-shaped particles whose stoichiometry was fully compatible with Fe 3 Al. This was determined by CB-SAEDP, and the related representative image and diffraction pattern are reported in Figure 8. The very fine nature of the Fe 3 Al rod-shaped particles can be clarified with a closer look at the Fe-Al phase diagram ( Figure 6). The Fe 3 Al phase is stable in a concentration range from 22.5 at.% to 36.5 at.% Al and transformed to FeAl or α-Fe at temperatures above 545 • C [24]. Considering that the typical temperature is 800-850 • C at the beginning of hot stamping and the die-quenching in the tool is maintained usually for 8-12 s [37], there is not enough time for transformation and phase growth, resulting in the very fine nature of the rod-shaped Fe 3 Al particles. The transformation of α-Fe to Fe 3 Al was described in Ref. [32] after austenitization at 1050 • C, deformation and cooling. Furthermore, the gradual transition from the bcc crystal structure of α-Fe to D0 3 of Fe 3 Al and to B2 of FeAl at the substrate/coating interface was reported by Cho et al. [24]. However, the authors did not observe any evidence of interfaces between these three different crystal structures.
The coating layer immediately in contact with the steel was mainly composed of a thin layer of Fe 3 (Al,Si)C, which featured a fine nanometric granular morphology with a mean thickness of 180 ± 90 nm. This layer was not always continuously distributed, and its thickness was quite scattered, ranging from a minimum of almost null to a maximum of up to 340 nm. A representative image of this thin layer is reported in Figure 9 along with the related SAEDP (inset). This carbide phase was identified as a product of solidstate interdiffusion between low carbon steel and pure Al at 600 • C in Ref. [30], and was further found at the interface of the 22MnB5 steel and Al-Si hot-dipped coating in Ref. [18]. This layer is probably formed by the inability of carbon to diffuse through the Fe-Al intermetallic phases, and carbon becomes concentrated at the steel interface where it forms ternary carbides [26]. These carbides seem to remain unchanged after austenitization and die-quenching. Further away from the thin Fe 3 (Al,Si)C layer, another thin layer was detected. This was more compact and continuous than the former layer. This second layer was found to be composed of Fe 2 (Al,Si) 5 with a thickness of 350 ± 30 nm. Further away from the 22MnB5 steel, the Al-Si coating was formed by a thicker layer of Fe 2 Al 5 . Within this layer, some nanometric particles of Fe 2 (Al,Si) 5 were also detected. These particles were fairly scattered and cuboid-shaped. Both the coating layer and nanoparticles are shown in the TEM image of Figure 10. In particular, the inset SAEDP of Figure 10b clearly shows that there is a crystallographic correlation between the small cuboid-like Fe 2 (Al,Si) 5 particles and the surrounding layer of Fe 2 Al 5 ; this was determined to be (-211)Fe 2 (Al,Si) 5  The coating layer immediately in contact with the steel was mainly composed of a thin layer of Fe3(Al,Si)C, which featured a fine nanometric granular morphology with a mean thickness of 180 ± 90 nm. This layer was not always continuously distributed, and its thickness was quite scattered, ranging from a minimum of almost null to a maximum of up to 340 nm. A representative image of this thin layer is reported in Figure 9 along with the related SAEDP (inset). This carbide phase was identified as a product of solidstate interdiffusion between low carbon steel and pure Al at 600 °C in Ref. [30], and was further found at the interface of the 22MnB5 steel and Al-Si hot-dipped coating in Ref. [18]. This layer is probably formed by the inability of carbon to diffuse through the Fe-Al intermetallic phases, and carbon becomes concentrated at the steel interface where it forms ternary carbides [26]. These carbides seem to remain unchanged after austenitization and die-quenching. Further away from the thin Fe3(Al,Si)C layer, another thin layer was detected. This was more compact and continuous than the former layer. This second layer was found to be composed of Fe2(Al,Si)5 with a thickness of 350 ± 30 nm. Further away from the 22MnB5 steel, the Al-Si coating was formed by a thicker layer of Fe2Al5. Within this layer, some nanometric particles of Fe2(Al,Si)5 were also detected. These particles were fairly scattered and cuboid-shaped. Both the coating layer and nanoparticles are shown in the TEM image of Figure 10. In particular, the inset SAEDP of Figure 10b clearly shows that there is a crystallographic correlation between the small cuboid-like Fe2(Al,Si)5 particles and the surrounding layer of Fe2Al5; this was determined to be (-211)Fe2(Al,Si)5//(-101)Fe2Al5.  Further away, the coating changed its composition again, and this layer had the same chemical composition and crystallographic structure as the innermost Fe 2 (Al,Si) 5 layer, which was followed by a thin layer of Fe 2 Al 5 . This latter layer was followed by the outermost coating layer with a chemical composition of Fe 2 Al 8 Si. This was indeed a fairly scattered layer with a nonuniform thickness. Figure 11 shows a representative image of the interface between Fe 2 Al 5 and Fe 2 Al 8 Si. SAEDPs were recorded and are also reported in Figure 11 insets. The Fe 2 Al 8 Si ternary phase is a major reaction product formed at the interface of the steel and Al-Si hot-dipped coating, as well as partly dispersed in the Al-Si coating matrix. The presence of this ternary phase in the outermost coating layer can be explained by the gradual enrichment of the coating with Fe and vice versa by the diffusion of Al and Si into the steel substrate. As a result of this Fe enrichment, the ternary phase was shifted towards the coating surface, where it formed a discontinuous layer due to its partial conversion to the Fe richer phases. Windmann et al. [17] reported the occurrence of Fe 2 Al 8 Si after 1 min of austenitization at 920 • C, and its transformation to an iron-richer phase of type Fe 2 Al 5 with increasing dwell time. However, the Fe 2 Al 8 Si ternary phase was clearly identified in the coating, even after an 8 min dwell time at 920 • C in the present study. Further away, the coating changed its composition again, and this layer had the same chemical composition and crystallographic structure as the innermost Fe2(Al,Si)5 layer, which was followed by a thin layer of Fe2Al5. This latter layer was followed by the outermost coating layer with a chemical composition of Fe2Al8Si. This was indeed a fairly scattered layer with a nonuniform thickness. Figure 11 shows a representative image of the interface between Fe2Al5 and Fe2Al8Si. SAEDPs were recorded and are also reported in Figure 11 insets. The Fe2Al8Si ternary phase is a major reaction product formed at the interface of the steel and Al-Si hot-dipped coating, as well as partly dispersed in the Al-Si coating matrix. The presence of this ternary phase in the outermost coating layer can be explained by the gradual enrichment of the coating with Fe and vice versa by the diffusion of Al and Si into the steel substrate. As a result of this Fe enrichment, the ternary phase was shifted towards the coating surface, where it formed a discontinuous layer due to its partial conversion to the Fe richer phases. Windmann et al. [17] reported the occurrence of Fe2Al8Si after 1 min of austenitization at 920 °C, and its transformation to an iron-richer phase of type Fe2Al5 with increasing dwell time. However, the Fe2Al8Si ternary phase was clearly identified in the coating, even after an 8 min dwell time at 920 °C in the present study.

XPS Analysis of the Surface Condition
Since SEM-EDS analysis can be used to analyze the deeper parts of materials, the exact nature of the present oxidic products was further examined by obtaining a depth profile of the chemical composition measured by X-ray photoelectron spectroscopy (Figure 12). The concentration dependence is plotted over time, since the etching speed is not constant over time. From the initial oxide layer of the as-received sample with a thickness of 14 nm (Figure 12a), the austenitization led to the strong enrichment of the surface with oxygen, creating an oxide layer with a total thickness of 76 nm (Figure 12b). The oxides detected on the surface of the as-received sample consisted of Al2O3, SiO2, and Al2O3 after austenitization and die-quenching ( Figure 13). In industrial conditions, the whole austenitization process and transferral to the press do not occur under a protective atmosphere, as was the case in the experiment in the present work. As such, the coating meets the air at high temperatures in the order of minutes. The high affinity of aluminum to oxygen, as well as the elevated temperature, cause the diffusion of Al to the coating surface and sig- Figure 11. TEM micrograph of the interface between the Fe 2 Al 5 layer and the outermost coating layer. Lenticular-shaped zones were identified by SAEDPs (insets) as Fe 2 Al 8 Si.

Changes in the Surface Condition after Die-quenching 3.3.1. XPS Analysis of the Surface Condition
Since SEM-EDS analysis can be used to analyze the deeper parts of materials, the exact nature of the present oxidic products was further examined by obtaining a depth profile of the chemical composition measured by X-ray photoelectron spectroscopy ( Figure 12). The concentration dependence is plotted over time, since the etching speed is not constant over time. From the initial oxide layer of the as-received sample with a thickness of 14 nm (Figure 12a), the austenitization led to the strong enrichment of the surface with oxygen, creating an oxide layer with a total thickness of 76 nm (Figure 12b). The oxides detected on the surface of the as-received sample consisted of Al 2 O 3 , SiO 2 , and Al 2 O 3 after austeniti-zation and die-quenching ( Figure 13). In industrial conditions, the whole austenitization process and transferral to the press do not occur under a protective atmosphere, as was the case in the experiment in the present work. As such, the coating meets the air at high temperatures in the order of minutes. The high affinity of aluminum to oxygen, as well as the elevated temperature, cause the diffusion of Al to the coating surface and significant oxygen enrichment due to the formation of a protective Al 2 O 3 layer, which further protects the coating itself and the steel substrate from scaling and decarburization [4,37]. However, the thick oxide layer influences the contact resistance during resistance spot welding in the automotive industry. Moreover, the presence of oxides at the surface can lead to inhomogeneous current flow and violent heat generation [11,38]. In contrast to the smooth and uniform features characteristic of the hot-dipped condition (Figure 1), the occurrence of voids and microcracks accompanies phase transformations in the coating and steel during austenitization and die-quenching (Figure 4). The microcracks nucleated at the top of the coating as a spot of increased stress concentration In contrast to the smooth and uniform features characteristic of the hot-dipped condition (Figure 1), the occurrence of voids and microcracks accompanies phase transformations in the coating and steel during austenitization and die-quenching (Figure 4). The microcracks nucleated at the top of the coating as a spot of increased stress concentration

Roughness Measurement of the Surface
In contrast to the smooth and uniform features characteristic of the hot-dipped condition (Figure 1), the occurrence of voids and microcracks accompanies phase transformations in the coating and steel during austenitization and die-quenching ( Figure 4). The microcracks nucleated at the top of the coating as a spot of increased stress concentration and then propagated through the coating [29]. Note that these microcracks were intercepted within the interdiffusion layer and did not reach the steel substrate. The interdiffusion layer was rich in iron (75.9 to 90.8 at.%) and tended to show a ductile behavior [20,29,32].
A reasonable explanation for the microcrack's formation is the thermal stress induced during austenitization and rapid cooling. The phase development during this heat cycle is connected both with the volume changes and with differences in the thermal expansion coefficients of the newly formed phases, resulting in thermal stresses. Gui et al. [20] stated that the densities of the main intermetallic phase Fe 2 Al 8 Si (τ 5 ) at the steel/coating interface under the hot-dipped condition were 3.62 g/cm 3 and 4.11 g/cm 3 for Fe 2 Al 5 and 5.37 g/cm 3 for the FeAl phases. Both the abovementioned phases (Fe 2 Al 5 and FeAl) were also present in the die-quenched condition. The density of Fe is generally known to be 7.7 g/cm 3 . Ruan et al. [39] studied the thermal expansion coefficient of the Al-Fe system (10-75 at.% Al) by a dilatometric method. They observed an increase in the thermal expansion coefficient at the same temperature and sensitivity to temperature with increasing the Fe content from 1.62 × 10 −5 K −1 (Al75Fe25) to 2.23 × 10 −5 K −1 (Al30Fe70), and then decreasing to 1.41 × 10 −5 K −1 (Al10Fe90). Moreover, Fe 2 Al 5 is known to be fragile with a low fracture toughness (1.2 MPa·m 1/2 ), which promotes microcrack growth and propagation [20,29]. When the maximum strength of the coating is exceeded by the action of thermal stresses and volume changes, the microcracks nucleate and easily grow in the fragile coating. These microcracks then may become wide and can intersect the voids in the coating, leading to coating delamination during die-quenching. More importantly, areas with exposed substrate behave as potential weak spots for corrosion [29].
In addition to crack occurrence in the coating, the volume changes also resulted in a change in surface topography. The surface profile and roughness were measured by a Keyence digital microscope and roughness tester using three linescans. The topography profile of the as-received sample with a hot-dipped coating is shown in Figure 14. The roughness parameters R a and R pc were 1.88 µm and 53 peaks/cm, respectively, and both values confirmed that the surface was relatively smooth. Austenitization and die-quenching caused the roughness parameters to increase to R a = 2.50 µm and R pc = 142 peaks/cm ( Figure 15). The hot-dipped Al-Si coating melts during austenitization and resolidifies again due to Fe enrichment, leading to the formation of Al-Fe-Si intermediate phases. These phases were inhomogeneously distributed and, as was described above, have different volumes and thermal expansion coefficients. It can be assumed that some of the Al-Fe-Si phases were exposed to the surface, or their formation allowed the surface roughness to increase. In addition, the diffusion-related transport of atoms might also contribute to the formation of shallow valleys, as shown in Figure 15. This presumption also corresponds to the dimensions of these phases, which have already been shown in Figure 4.
These phases were inhomogeneously distributed and, as was described above, have different volumes and thermal expansion coefficients. It can be assumed that some of the Al-Fe-Si phases were exposed to the surface, or their formation allowed the surface roughness to increase. In addition, the diffusion-related transport of atoms might also contribute to the formation of shallow valleys, as shown in Figure 15. This presumption also corresponds to the dimensions of these phases, which have already been shown in Figure 4.

Conclusions
It was found that the initial hot-dipped Al-Si coating on commercially available 22MnB5 steel was composed of Al, Si matrix, FeAl8Si2 (τ5) phase and two thin sublayers of η-Fe2Al5 and Fe3Al2Si3 (τ1) at the coating/substrate interface. Heat treatment at 920 °C for 8 min followed by die-quenching resulted in strong Fe diffusion to the coating, changing the phase composition and causing an increase in the thickness. In addition to the phases FeAl(Si) and Fe2Al5, and the interdiffusion layer forming the multilayered structure of the coating, electron diffraction revealed layers and phases of a very fine nature. The thin Fe3(Al,Si)C coating layer (180 ± 90 nm), detected by other authors after hot-dipping, was found at the site of contact with the steel. However, the present layer was retained even after heat treatment, which has not been reported in the works of others. Albearing rod-shaped particles in the immediate vicinity of the steel, determined as Fe3Al, were the result of rapid cooling during die-quenching from the austenitization temperature. The presence of an Fe2Al5 phase further away from the steel substrate (matrix of the coating) was confirmed by electron diffraction, and cuboid-shaped nanometric particles of Fe2(Al,Si)5 were detected within this phase. The residues of Fe2Al8Si phase in the outermost coating layer can be attributed to the shift of this phase due to the significant enrichment of the coating with Fe, and to its partial conversion to Fe-richer phases.
Furthermore, the surface of the Al-Si coating was covered by the oxide layer initially formed by Al2O3 and SiO2, the thickness of which increased during austenitization and die-quenching from 14 to 76 nm. After this heat treatment, the layer was composed mostly

Conclusions
It was found that the initial hot-dipped Al-Si coating on commercially available 22MnB5 steel was composed of Al, Si matrix, FeAl 8 Si 2 (τ 5 ) phase and two thin sublayers of η-Fe 2 Al 5 and Fe 3 Al 2 Si 3 (τ 1 ) at the coating/substrate interface. Heat treatment at 920 • C for 8 min followed by die-quenching resulted in strong Fe diffusion to the coating, changing the phase composition and causing an increase in the thickness. In addition to the phases FeAl(Si) and Fe 2 Al 5 , and the interdiffusion layer forming the multilayered structure of the coating, electron diffraction revealed layers and phases of a very fine nature. The thin Fe 3 (Al,Si)C coating layer (180 ± 90 nm), detected by other authors after hot-dipping, was found at the site of contact with the steel. However, the present layer was retained even after heat treatment, which has not been reported in the works of others. Al-bearing rodshaped particles in the immediate vicinity of the steel, determined as Fe 3 Al, were the result of rapid cooling during die-quenching from the austenitization temperature. The presence of an Fe 2 Al 5 phase further away from the steel substrate (matrix of the coating) was confirmed by electron diffraction, and cuboid-shaped nanometric particles of Fe 2 (Al,Si) 5 were detected within this phase. The residues of Fe 2 Al 8 Si phase in the outermost coating layer can be attributed to the shift of this phase due to the significant enrichment of the coating with Fe, and to its partial conversion to Fe-richer phases.
Furthermore, the surface of the Al-Si coating was covered by the oxide layer initially formed by Al 2 O 3 and SiO 2 , the thickness of which increased during austenitization and die-quenching from 14 to 76 nm. After this heat treatment, the layer was composed mostly of an Al 2 O 3 .
The above-described phase development during the heat treatment, connected both with the volume changes and with differences in the thermal expansion coefficients of the newly formed phases, resulted in a surface roughness increase from 1.88 to 2.50 µm (in terms of R a value), and to the microcracks formation.