High Temperature Corrosion Behaviour of Aluminide-Coated Cast Iron for an Exhaust Manifold Application

To reduce the pollution emission from vehicles, an improvement on the combustion process is expected, leading to increased exhaust gas temperature. As a result, the development of new materials for an exhaust manifold used at higher temperatures is required. A cost-effective cast iron exhaust manifold treated by aluminising pack cementation was developed in the present work to combat the high temperature corrosion. Its kinetics under cyclic oxidation in N2–12%O2–10%H2O at 850 ◦C was parabolic with the rate constant (kp) of 5.66 × 10−12 g2 cm–4 s–1, about two orders of magnitude lower than that of the bare cast iron, which indicated the protectiveness of the applied coating. These results relate to the protective alumina formation for the aluminised cast iron and the formation of the less protective iron oxides for the bare cast iron after oxidation, as evidenced by the XRD and Raman spectroscopy results. The addition of 10% water vapour to N2–12%O2 thickened the aluminide layer from 344 μm for the sample oxidised in dry atmosphere to 409 μm for the sample oxidised humidified one. It accelerated the oxidation rate of the aluminised cast iron as the kp value increased by 8.5 times, and also increased the hardness of the aluminised surface, as it was 364 HV for the sample exposed to dry atmosphere and 420 HV for the sample exposed to humidified one. The latter result implied the possibility of the hydrogen dissolution into the metal surface. The roles of hydroxyl ion and dissolved hydrogen on the oxidation and evolution of the aluminide layer after exposure to water vapour were proposed.


Introduction
Towards a low carbon society, a combustion process in the automobile engine has been improved and, as a result, the exhaust gas temperature tends to be raised [1][2][3]. Traditionally the actual air-to-fuel ratio per the ideal air-to-fuel ratio, the λ value, is about 0.9, which is the optimised value for the combustion efficiency and material durability [4,5]. In this case, the excess fuel can cool the engine as well as the exhaust gas [4]. In order to combust the fuel more efficiently, the λ value has to be increased, and consequently the cooling effect due to the excess fuel is eliminated, resulting in the increased exhaust gas temperature, which can be up to about 1050 • C [4]. Thus, there has been a quest for the

Materials and Methods
The exhaust manifold material made of cast iron was used in this study. Figure 1 presents its microstructure exhibiting the distribution of vermicular and nodular graphite in a matrix, showing that the cast iron was the compacted-graphite type typically alloyed with 3.1-4.0 wt.% C and 1.7-3.0 wt.% Si, 0.015-0.035 wt.% Mg and 0.06-0.13 wt.% Ti [35]. The sample was cut into the dimensions of 8 mm × 10 mm × 5 mm. It was then polished on SiC-papers up to 600 grit, cleaned by acetone in an ultrasonic machine and dried in air. Afterwards, it was aluminised at 1050 • C for 5 h in the pack consisting of 25 wt.% Al (particle size:~70 µm), 70 wt.% Al 2 O 3 (~40 µm) and 5 wt.% NH 4 Cl in argon (purity: 99.995%). The horizontal tube furnace was used with the tube inner and outer diameters of 4.8 and 5 cm and with the tube length of 58 cm. After aluminising, the sample was taken for the cyclic oxidation test at 850 °C for 14 cycles. In each cycle, the sample was heated up from room temperature to 850 °C and held at this temperature for 8 h in the atmosphere, containing 12% of oxygen without water vapour (N2-12O2) and with 10% of water vapour (N2-12O2-10H2O), and then cooled down in a furnace to room temperature under laboratory air. The bare cast iron was also tested in N2-12O2-10H2O to compare its oxidation behaviour with the cast iron treated by aluminising. To prepare the atmosphere containing water vapour, the N2-12O2 gas mixture was flowed into the flask containing water with the temperature of 44.5 °C. From the thermodynamic calculation using Barin's data [36], water vapour at this temperature was 10% by volume. In this manner, the N2-12O2-10H2O gas mixture was produced and further fed to the furnace.
For the physico-chemical characterisation, an X-ray diffractometer (Bruker AXS D8 DISCOVER, Billerica, MA, USA) using Kα (1.5406 Å) Cu radiation and Laser Raman spectroscopy (Renishaw RM 1000, Glouchestershire, UK) were performed for phase identification. A scanning electron microscope (SEM, JSM-6610LV, Tokyo, Japan) equipped with an energy dispersive X-ray analyser was used for imaging and elemental analyses. Figure 2 shows mass gains of the bare and aluminide-coated cast irons as a function of the oxidation cycle. It can be seen that aluminising significantly reduced the oxidation rate of the cast iron in N2-12O2-10H2O. For example, after oxidation for 14 cycles, mass gains of the bare and aluminide-coated cast irons were 12.91 and 1.64 mg cm -2 , respectively, showing 87% reduction in mass gain after aluminising treatment. It can also be seen that increasing water vapour in the atmosphere enhanced the oxidation. For example, after 14 testing cycles, mass gains of the aluminidecoated cast irons oxidised in N2-12O2 and N2-12O2-10H2O were 0.49 and 1.64 mg cm -2 , respectively. After aluminising, the sample was taken for the cyclic oxidation test at 850 • C for 14 cycles. In each cycle, the sample was heated up from room temperature to 850 • C and held at this temperature for 8 h in the atmosphere, containing 12% of oxygen without water vapour (N 2 -12O 2 ) and with 10% of water vapour (N 2 -12O 2 -10H 2 O), and then cooled down in a furnace to room temperature under laboratory air. The bare cast iron was also tested in N 2 -12O 2 -10H 2 O to compare its oxidation behaviour with the cast iron treated by aluminising. To prepare the atmosphere containing water vapour, the N 2 -12O 2 gas mixture was flowed into the flask containing water with the temperature of 44.5 • C. From the thermodynamic calculation using Barin's data [36], water vapour at this temperature was 10% by volume. In this manner, the N 2 -12O 2 -10H 2 O gas mixture was produced and further fed to the furnace.

Mass Gain of the Bare and Aluminide-Coated Cast Irons after Oxidation
For the physico-chemical characterisation, an X-ray diffractometer (Bruker AXS D8 DISCOVER, Billerica, MA, USA) using Kα (1.5406 Å) Cu radiation and Laser Raman spectroscopy (Renishaw RM 1000, Glouchestershire, UK) were performed for phase identification. A scanning electron microscope (SEM, JSM-6610LV, Tokyo, Japan) equipped with an energy dispersive X-ray analyser was used for imaging and elemental analyses. Figure 2 shows mass gains of the bare and aluminide-coated cast irons as a function of the oxidation cycle. It can be seen that aluminising significantly reduced the oxidation rate of the cast iron in N 2 -12O 2 -10H 2 O. For example, after oxidation for 14 cycles, mass gains of the bare and aluminide-coated cast irons were 12.91 and 1.64 mg cm -2 , respectively, showing 87% reduction in mass gain after aluminising treatment. It can also be seen that increasing water vapour in the atmosphere enhanced the oxidation. For example, after 14 testing cycles, mass gains of the aluminide-coated cast irons oxidised in N 2 -12O 2 and N 2 -12O 2 -10H 2 O were 0.49 and 1.64 mg cm -2 , respectively.     . depicts the X-ray diffraction (XRD) pattern of the bare cast iron after cyclic oxidation in N2-12O2-10H2O for 14 cycles. Only hematite peaks were detected. Since the analytical depth of the XRD analysis was about 1-3 μm [37], this result indicated that hematite existed as the outermost layer of the scale. To further characterise the inner oxide scale, the sample was cross-sectioned and Raman spectroscopy was applied to identify the oxide phases at the inner and the outer parts of the scale as shown in Figure 5. It can be seen that at the outer part of the scale, the hematite peaks were dominantly observed while a slight signal of magnetite was detected. On the contrary, at the inner part of the scale, the strong peaks of magnetite were observed. Hematite was also found but with lower intensities than the ones detected at the outer part of the scale. The peak of fayalite was also observed. The results from XRD and Raman patterns indicated that hematite existed as the outer    . depicts the X-ray diffraction (XRD) pattern of the bare cast iron after cyclic oxidation in N2-12O2-10H2O for 14 cycles. Only hematite peaks were detected. Since the analytical depth of the XRD analysis was about 1-3 μm [37], this result indicated that hematite existed as the outermost layer of the scale. To further characterise the inner oxide scale, the sample was cross-sectioned and Raman spectroscopy was applied to identify the oxide phases at the inner and the outer parts of the scale as shown in Figure 5. It can be seen that at the outer part of the scale, the hematite peaks were dominantly observed while a slight signal of magnetite was detected. On the contrary, at the inner part of the scale, the strong peaks of magnetite were observed. Hematite was also found but with lower intensities than the ones detected at the outer part of the scale. The peak of fayalite was also observed. The results from XRD and Raman patterns indicated that hematite existed as the outer  . depicts the X-ray diffraction (XRD) pattern of the bare cast iron after cyclic oxidation in N 2 -12O 2 -10H 2 O for 14 cycles. Only hematite peaks were detected. Since the analytical depth of the XRD analysis was about 1-3 µm [37], this result indicated that hematite existed as the outermost layer of the scale. To further characterise the inner oxide scale, the sample was cross-sectioned and Raman spectroscopy was applied to identify the oxide phases at the inner and the outer parts of the scale as shown in Figure 5. It can be seen that at the outer part of the scale, the hematite peaks were dominantly observed while a slight signal of magnetite was detected. On the contrary, at the inner part of the scale, the strong peaks of magnetite were observed. Hematite was also found but with lower intensities than the ones detected at the outer part of the scale. The peak of fayalite was also observed. The results from XRD and Raman patterns indicated that hematite existed as the outer layer, whereas magnetite and fayalite were in the inner part of the oxide scale. Comparing to literature, the recent work on the oxidation of the spheroidal cast iron containing 4.4 wt.% Si and 0.5 wt.% Mo in humid air at 700 • C has reported the formation of hematite as the outer external scale and the formation of magnetite and fayalite as the inner scale by using the XRD technique [11]. However, it is noted that wustite was not observed in the scale in the present work.

Characterisation of the Bare Cast Iron after Oxidation
Coatings 2020, 10, x FOR PEER REVIEW 5 of 15 layer, whereas magnetite and fayalite were in the inner part of the oxide scale. Comparing to literature, the recent work on the oxidation of the spheroidal cast iron containing 4.4 wt.% Si and 0.5 wt.% Mo in humid air at 700 °C has reported the formation of hematite as the outer external scale and the formation of magnetite and fayalite as the inner scale by using the XRD technique [11]. However, it is noted that wustite was not observed in the scale in the present work.

Characterisation of the Aluminide-Coated Cast Iron before Oxidation
Figure 6a depicts the SEM cross-section of the cast iron after aluminising. The aluminide layer consisted of two distinct parts: the outer and inner sub-layers. The outer sub-layer which was adjacent to the atmosphere mainly consisted of the darker matrix dispersed by white phases. The inner sublayer which was adjacent to the metal substrate consisted of the matrix precipitated by elongated darker phases. By measuring from two SEM images, the thicknesses of the inner and outer sub-layers were respectively 35 ± 3 μm and 219 ± 67 μm, giving the total aluminide thickness of about 254 μm. Coatings 2020, 10, x FOR PEER REVIEW 5 of 15 layer, whereas magnetite and fayalite were in the inner part of the oxide scale. Comparing to literature, the recent work on the oxidation of the spheroidal cast iron containing 4.4 wt.% Si and 0.5 wt.% Mo in humid air at 700 °C has reported the formation of hematite as the outer external scale and the formation of magnetite and fayalite as the inner scale by using the XRD technique [11]. However, it is noted that wustite was not observed in the scale in the present work.

Characterisation of the Aluminide-Coated Cast Iron before Oxidation
Figure 6a depicts the SEM cross-section of the cast iron after aluminising. The aluminide layer consisted of two distinct parts: the outer and inner sub-layers. The outer sub-layer which was adjacent to the atmosphere mainly consisted of the darker matrix dispersed by white phases. The inner sublayer which was adjacent to the metal substrate consisted of the matrix precipitated by elongated darker phases. By measuring from two SEM images, the thicknesses of the inner and outer sub-layers were respectively 35 ± 3 μm and 219 ± 67 μm, giving the total aluminide thickness of about 254 μm.

Characterisation of the Aluminide-Coated Cast Iron before Oxidation
Figure 6a depicts the SEM cross-section of the cast iron after aluminising. The aluminide layer consisted of two distinct parts: the outer and inner sub-layers. The outer sub-layer which was adjacent to the atmosphere mainly consisted of the darker matrix dispersed by white phases. The inner sub-layer which was adjacent to the metal substrate consisted of the matrix precipitated by elongated darker phases. By measuring from two SEM images, the thicknesses of the inner and outer sub-layers were respectively 35 ± 3 µm and 219 ± 67 µm, giving the total aluminide thickness of about 254 µm. Coatings 2020, 10, x FOR PEER REVIEW 6 of 15 To characterise the outer sub-layer, the XRD technique was applied on the surface of the sample giving the pattern (a) shown in Figure 7. It can be seen that the major peaks observed were Fe2Al5. To characterise the outer sub-layer, the XRD technique was applied on the surface of the sample giving the pattern (a) shown in Figure 7. It can be seen that the major peaks observed were Fe 2 Al 5 . The most intense peak of FeAl 2 Si was also detected. To further characterise the phases located deeper from the surface than the XRD analytical depth, the energy dispersive X-ray analysis (EDX) was applied to the cross-sectioned sample. By measuring on three different areas, the atomic ratio of aluminium to iron in the matrix of the outer scale was in the range of 1.99 to 2.25 at.%. The dispersed white phases contained a significant amount of silicon by 3.3 at.%. For the inner sub-layer, the atomic ratio of aluminium to iron in the matrix was 1.05:1, showing that this phase should be FeAl.
Coatings 2020, 10, x FOR PEER REVIEW 7 of 15 The most intense peak of FeAl2Si was also detected. To further characterise the phases located deeper from the surface than the XRD analytical depth, the energy dispersive X-ray analysis (EDX) was applied to the cross-sectioned sample. By measuring on three different areas, the atomic ratio of aluminium to iron in the matrix of the outer scale was in the range of 1.99 to 2.25 at.%. The dispersed white phases contained a significant amount of silicon by 3.3 at.%. For the inner sub-layer, the atomic ratio of aluminium to iron in the matrix was 1.05:1, showing that this phase should be FeAl.   Figure 6b shows the SEM cross-section of the aluminide-coated cast iron after oxidation in N2-12O2 for 14 cycles. It can be seen that, after oxidation, most of the aluminide layer remained unattacked and still consisted of two distinct parts, like the one observed before oxidation. However, the inner part was considerably thicker to be 153 ± 12 μm thick while the outer layer was slightly thinner to be 191 ± 11 μm thick. These two parts gave the total thickness of the aluminide layer of about 344 μm, which was thicker than the one before oxidation (254 μm). Furthermore, it should be noted that, even though a large crack was contained in the outer sub-layer, the aluminide layer was still protective compared with the aluminide layer oxidised in N2-12O2-10H2O ( Figure 2).

Characterisation of the Aluminide-Coated Cast Iron After Oxidation
For the outer sub-layer, pattern (b) in Figure 7 shows the XRD result of the aluminide-coated cast iron after oxidation in N2-12O2 for 14 cycles. Comparing to the XRD peaks of the aluminide layer before oxidation, the Fe2Al5 peaks were still dominantly observed, whereas the peaks of Al2O3 were additionally detected. The existence of the latter peaks confirmed the formation of the protective alumina layer after oxidation. By using the alumina density of 3.95 g cm -2 [15], the alumina thickness calculated from the mass gain after 14 oxidation cycles was 2.64 μm. The thickness of alumina could be lower than this value because of the increased surface area due to the rough surface [15]. Because the alumina layer could be relatively thin and the aluminide layer tended to be fractured along the metal surface, as shown in Figure 6b, the alumina layer was then not obviously observed in the SEM image, even though its existence was confirmed by the XRD. It can be further seen from Figure 6b that the Fe3(Al,Si)5, once observed as the white phase dispersed in the matrix of the outer sub-layer of the aluminide-coated sample before oxidation, disappeared. The outer aluminide sub-layer; thus, looks homogeneous with the atomic ratio of aluminium to iron in the range of 2.36 to 1.98 at the locations 14 to 116 μm away from the surface. For the inner sub-layer, comparing to the aluminidecoated sample before oxidation, this region still consisted of the matrix precipitated by elongated  Figure 6b shows the SEM cross-section of the aluminide-coated cast iron after oxidation in N 2 -12O 2 for 14 cycles. It can be seen that, after oxidation, most of the aluminide layer remained unattacked and still consisted of two distinct parts, like the one observed before oxidation. However, the inner part was considerably thicker to be 153 ± 12 µm thick while the outer layer was slightly thinner to be 191 ± 11 µm thick. These two parts gave the total thickness of the aluminide layer of about 344 µm, which was thicker than the one before oxidation (254 µm). Furthermore, it should be noted that, even though a large crack was contained in the outer sub-layer, the aluminide layer was still protective compared with the aluminide layer oxidised in N 2 -12O 2 -10H 2 O (Figure 2).

Characterisation of the Aluminide-Coated Cast Iron after Oxidation
For the outer sub-layer, pattern (b) in Figure 7 shows the XRD result of the aluminide-coated cast iron after oxidation in N 2 -12O 2 for 14 cycles. Comparing to the XRD peaks of the aluminide layer before oxidation, the Fe 2 Al 5 peaks were still dominantly observed, whereas the peaks of Al 2 O 3 were additionally detected. The existence of the latter peaks confirmed the formation of the protective alumina layer after oxidation. By using the alumina density of 3.95 g cm -2 [15], the alumina thickness calculated from the mass gain after 14 oxidation cycles was 2.64 µm. The thickness of alumina could be lower than this value because of the increased surface area due to the rough surface [15]. Because the alumina layer could be relatively thin and the aluminide layer tended to be fractured along the metal surface, as shown in Figure 6b, the alumina layer was then not obviously observed in the SEM image, even though its existence was confirmed by the XRD. It can be further seen from Figure 6b  before oxidation, this region still consisted of the matrix precipitated by elongated phases. The EDX analysis showed that the atomic ratio of aluminium to iron of the matrix was 1.12:1, respectively, indicating that this phase should still be FeAl. Figure 6c shows the SEM cross-section of the aluminide-coated cast iron after oxidation in N 2 -12O 2 -10H 2 O for 14 cycles. Comparing to the aluminide-coated sample oxidised in N 2 -12O 2 , the aluminide layer exposed to this atmosphere was considerably thicker, at 409 µm, with less obvious contrast between two sub-layers. The XRD result of aluminide-coated sample oxidised in this condition for 14 cycles is shown in pattern (c) of Figure 7. The strong peaks of Fe 2 Al 5 and the existence of Al 2 O 3 peaks were observed. The EDX spot analysis showed that the atomic ratio of aluminium to iron of the aluminide layer was reduced from 2.36 to 0.81 at the locations 136 to 362 µm away from the metal surface, respectively.

Oxidation Kinetics
Mass gain from Figure 2 was squared and plotted as a function of time by considering that each testing cycle gave 8 h of isothermal oxidation. The linear relationship between the square of mass gain and time can be seen in Figure 8. It can be expressed in the following form [37,38]: where (∆m/A) is the mass gain, which is the change of mass with respect to the mass before oxidation per unit surface area of the sample; k p is the parabolic rate constant; and t is the oxidation period of time. The parabolic relation according to Equation (1) indicates that the rate limiting step of the oxidation was diffusion-controlled and the oxidation rate reduces when the oxidation period is longer [37,38]. From this figure, the parabolic rate constant of the bare cast iron oxidised in N 2 -12O 2 -10H 2 O was calculated at 4.34 × 10 -10 g 2 cm -4 s -1 . It was higher than the parabolic rate constants of the aluminide-coated samples, which were 6.69 × 10 -13 and 5.66 × 10 -12 g 2 cm -4 s -1 for the oxidation in N 2 -12O 2 and N 2 -12O 2 -10H 2 O, respectively. Comparing with literature, when the studied cast iron was treated by aluminising, its parabolic rate constants reduced and approached those of the alumina-forming alloys, which were in the range of 1 × 10 -13 to 6.2 × 10 -11 g 2 cm -4 s -1 [39][40][41][42][43][44]. In the case of iron, oxidation rate also tends to be higher when water vapor is present in the atmosphere [45].
Coatings 2020, 10, x FOR PEER REVIEW 8 of 15 phases. The EDX analysis showed that the atomic ratio of aluminium to iron of the matrix was 1.12:1, respectively, indicating that this phase should still be FeAl. Figure 6c shows the SEM cross-section of the aluminide-coated cast iron after oxidation in N2-12O2-10H2O for 14 cycles. Comparing to the aluminide-coated sample oxidised in N2-12O2, the aluminide layer exposed to this atmosphere was considerably thicker, at 409 μm, with less obvious contrast between two sub-layers. The XRD result of aluminide-coated sample oxidised in this condition for 14 cycles is shown in pattern (c) of Figure 7. The strong peaks of Fe2Al5 and the existence of Al2O3 peaks were observed. The EDX spot analysis showed that the atomic ratio of aluminium to iron of the aluminide layer was reduced from 2.36 to 0.81 at the locations 136 to 362 μm away from the metal surface, respectively.

Oxidation Kinetics
Mass gain from Figure 2 was squared and plotted as a function of time by considering that each testing cycle gave 8 h of isothermal oxidation. The linear relationship between the square of mass gain and time can be seen in Figure 8. It can be expressed in the following form [37,38]: where (m/A) is the mass gain, which is the change of mass with respect to the mass before oxidation per unit surface area of the sample; kp is the parabolic rate constant; and t is the oxidation period of time. The parabolic relation according to Equation (1) indicates that the rate limiting step of the oxidation was diffusion-controlled and the oxidation rate reduces when the oxidation period is longer [37,38]. From this figure, the parabolic rate constant of the bare cast iron oxidised in N2-12O2-10H2O was calculated at 4.34 × 10 -10 g 2 cm -4 s -1 . It was higher than the parabolic rate constants of the aluminide-coated samples, which were 6.69 × 10 -13 and 5.66 × 10 -12 g 2 cm -4 s -1 for the oxidation in N2-12O2 and N2-12O2-10H2O, respectively. Comparing with literature, when the studied cast iron was treated by aluminising, its parabolic rate constants reduced and approached those of the aluminaforming alloys, which were in the range of 1 × 10 -13 to 6.2 × 10 -11 g 2 cm -4 s -1 [39][40][41][42][43][44]. In the case of iron, oxidation rate also tends to be higher when water vapor is present in the atmosphere [45].

Formation of Thermal Oxide Scales on the Bare Cast Iron
At equilibrium, the standard Gibbs free energy for the oxide formation reaction is proportional to the absolute temperature T and the logarithm of the equilibrium oxygen partial pressure, and may also be written in the form of A-BT where A and B are, respectively, the standard enthalpy and entropy for the oxide formation reaction [46,47]. From these relations, the logarithm of the equilibrium oxygen partial pressure can be expressed as a linear function of the inverse temperature, as comprehensively derived in the literature [46,47]. Figure 9 presents this relation for the formation of oxides relevant to the present study (i.e., hematite, magnetite, wustite, fayalite, silica and alumina). The diagram was constructed using standard thermodynamic data compiled by Kubaschewski and Evans [48].  Figure 9. Equilibrium diagram of the formation of oxides relevant to the present work constructed using thermodynamic data from Kubaschewski and Evans [48].
For the bare cast iron oxidised in N2-12O2-10H2O, XRD and Raman patterns showed that hematite, magnetite and fayalite were formed. To analyse the thermodynamic stability of these oxides, the partial pressure of oxygen in the atmosphere has to be primarily determined. Oxygen in

Formation of Thermal Oxide Scales on the Bare Cast Iron
At equilibrium, the standard Gibbs free energy for the oxide formation reaction is proportional to the absolute temperature T and the logarithm of the equilibrium oxygen partial pressure, and may also be written in the form of A-BT where A and B are, respectively, the standard enthalpy and entropy for the oxide formation reaction [46,47]. From these relations, the logarithm of the equilibrium oxygen partial pressure can be expressed as a linear function of the inverse temperature, as comprehensively derived in the literature [46,47]. Figure 9 presents this relation for the formation of oxides relevant to the present study (i.e., hematite, magnetite, wustite, fayalite, silica and alumina). The diagram was constructed using standard thermodynamic data compiled by Kubaschewski and Evans [48].

Formation of Thermal Oxide Scales on the Bare Cast Iron
At equilibrium, the standard Gibbs free energy for the oxide formation reaction is proportional to the absolute temperature T and the logarithm of the equilibrium oxygen partial pressure, and may also be written in the form of A-BT where A and B are, respectively, the standard enthalpy and entropy for the oxide formation reaction [46,47]. From these relations, the logarithm of the equilibrium oxygen partial pressure can be expressed as a linear function of the inverse temperature, as comprehensively derived in the literature [46,47]. Figure 9 presents this relation for the formation of oxides relevant to the present study (i.e., hematite, magnetite, wustite, fayalite, silica and alumina). The diagram was constructed using standard thermodynamic data compiled by Kubaschewski and Evans [48].  Figure 9. Equilibrium diagram of the formation of oxides relevant to the present work constructed using thermodynamic data from Kubaschewski and Evans [48].
For the bare cast iron oxidised in N2-12O2-10H2O, XRD and Raman patterns showed that hematite, magnetite and fayalite were formed. To analyse the thermodynamic stability of these oxides, the partial pressure of oxygen in the atmosphere has to be primarily determined. Oxygen in Figure 9. Equilibrium diagram of the formation of oxides relevant to the present work constructed using thermodynamic data from Kubaschewski and Evans [48].
For the bare cast iron oxidised in N 2 -12O 2 -10H 2 O, XRD and Raman patterns showed that hematite, magnetite and fayalite were formed. To analyse the thermodynamic stability of these oxides, the partial pressure of oxygen in the atmosphere has to be primarily determined. Oxygen in the atmosphere could be from the oxygen directly fed to the reactor (0.12 bar) or the oxygen dissociated from water vapour. For one mole of N 2 -12O 2 -10H 2 O, it was calculated using standard thermodynamic data [48] that the amount of oxygen dissociated from water vapour at equilibrium of the reaction H 2 O = H 2 + (1/2)O 2 was very small (i.e., 1.05 × 10 -9 mol at 850 • C). Thus the partial pressure of oxygen in this atmosphere was about the partial pressure of oxygen directly fed to the reactor (0.12 bar). This oxygen partial pressure was clearly higher than the ones at equilibrium of the hematite, magnetite and fayalite formation as shown in Figure 9. The formations of these oxides were then thermodynamically possible.
However, wustite was absent from the scale even though thermodynamics indicates its feasibility to be formed. It was reported by Ahtoy [49] that, for the silicon-containing steel oxidised in water vapour, silicon oxide and wustite were formed in the early period of oxidation. When time elapsed, these two oxides reacted to each other and formed fayalite [49]. From the calculation using standard data [48] at 850 • C, the reaction SiO 2 + 2FeO = Fe 2 SiO 4 is thermodynamically possible at the standard state since its Gibbs free energy is negative, that is −25,714 J. In addition, Nishimoto et al. [50] applied the EBSD technique to analyse the scale formed on the silicon-containing steel oxidised in air, and found that increasing Si in the steel tended to reduce wustite while fayalite formed. The diffusivity of iron through fayalite was reported to be lower than that through wustite [49][50][51]. Thus, when fayalite formed, the outward diffusion of iron to form iron oxides was reduced. This led to the substantial formation of the iron oxides which had higher ratios of oxygen to iron (i.e., hematite or magnetite), as observed in the previous work [50] as well as the present one.

Formation of the Aluminide Layer after Aluminising
To analyse the stability of Fe-Al intermetallic compounds, it is necessary to determine the standard Gibbs free energies of their formation at high temperatures. In this analysis, the Ellingham approximation was adopted. It was assumed that the standard enthalpy and entropy of reaction did not depend on temperature [46,47]. Maitra and Gupta [52] reported that Gibbs free energies of formation of FeAl 3 , Fe 2 Al 5 , FeAl 2 and FeAl at 900 • C were −93.91, −168.58, −77.36 and −54.7 kJ.mol -1 , respectively. Kubaschewski and Evans [48] reported the values of the enthalpies of formation of FeAl 3 , FeAl 2 and FeAl at 25 • C, as well as the latent heat of fusion of aluminium at its melting point. From these data [48], the standard enthalpies of formation of those three compounds from solid iron and liquid aluminium at 900 • C can be calculated. For Fe 2 Al 5 , literature reported the values of the enthalpy of formation of solid Fe 2 Al 5 from liquid iron and liquid aluminium [53] and the latent heat of fusion of iron at its melting point [48]. From these data [48,53], the standard enthalpy of formation of Fe 2 Al 5 from solid iron and liquid aluminium at 900 • C could also be derived. When the standard Gibbs free energies and enthalpies of formations of those compounds (FeAl 3 , Fe 2 Al 5 , FeAl 2 and FeAl) at 900 • C were known, their standard entropies of formation at that temperature could be obtained. By the Ellingham approximation, the standard enthalpies and entropies of formation at 900 • C were considered to be same as the ones at other temperatures. Thus, the standard Gibbs free energies to form Fe-Al compounds from solid iron and one mole of liquid aluminium were calculated at different temperatures, as shown in Figure 10. At the standard state, all the Fe-Al compounds (FeAl 3 , Fe 2 Al 5 , FeAl 2 and FeAl) are feasible to form at the aluminising temperature of 1050 • C because their Gibbs free energies are negative. However, some of them might not be observed if the rates of their formation were too slow.
From the characterisation results of the outer aluminide sub-layer, Fe 2 Al 5 was detected at the surface by XRD. It was reported for the Fe-Al system that FeAl 2 had the atomic ratio of aluminium to iron in the range of 1.91 to 2.02, whereas Fe 2 Al 5 had the ratio in the range of 2.33 to 2.74 [52]. As previously reported in Section 3.3, the atomic ratio of aluminium to iron of the matrix of the outer sub-layer of the aluminide-coated sample was in the range of 1.99 to 2.25 at.%. Thus the chemical formula of the matrix should lie between FeAl 2 and Fe 2 Al 5 . The existence of these two phases in the same layer was also reported in the aluminide coating of the hot-dipped low carbon steel [54]. From the characterisation results of the outer aluminide sub-layer, Fe2Al5 was detected at the surface by XRD. It was reported for the Fe-Al system that FeAl2 had the atomic ratio of aluminium to iron in the range of 1.91 to 2.02, whereas Fe2Al5 had the ratio in the range of 2.33 to 2.74 [52]. As previously reported in Section 3.3, the atomic ratio of aluminium to iron of the matrix of the outer sub-layer of the aluminide-coated sample was in the range of 1.99 to 2.25 at.%. Thus the chemical formula of the matrix should lie between FeAl2 and Fe2Al5. The existence of these two phases in the same layer was also reported in the aluminide coating of the hot-dipped low carbon steel [54].

Evolution of the Aluminide Layer during Oxidation
After oxidation in N2-12O2, alumina was formed at the surface as observed by XRD. This is because the oxygen partial pressure in the atmosphere (0.12 bar) was drastically higher than the one at equilibrium of the alumina formation, as shown in Figure 9. It was reported that alumina was grown at high temperatures mainly by the oxygen inward diffusion [15,[55][56][57], while the outward aluminium diffusion was also possible [57]. The total aluminide layer was thicker than that before oxidation, indicating the diffusion of aluminium into the base cast iron. The dilution of aluminium after oxidation contributed to the increased ratio of iron to aluminium in the aluminide layer and the thicker inner sub-layer. It was also reported that the solubility of silicon in FeAl of the quenched specimen was in the range of 10.38 to 16.24 at.% at the temperature range of 800 to 1020 °C [52]. These values were substantially higher than the solubility of silicon in Fe2Al5, which was in the range of 2.00 to 4.4 at.% [52]. Because the FeAl sub-layer which had high silicon solubility was thicker after oxidising, it was then possible that silicon in the outer sub-layer diffused to the inner FeAl sub-layer and dissolved in it. This then resulted in the disappearance of the silicon-rich Fe3(Al,Si)5 phase once observed after aluminising.
When the aluminide-coated cast iron was oxidised in N2-12O2-10H2O, alumina was also formed but the oxidation rate in this atmosphere was higher than the one in N2-12O2. This should be because water vapour helped form the oxygen-containing species that could diffuse through the oxide faster than the oxygen vacancy, which was the predominant defect for the oxidation in N2-12O2. Galerie et al. [25,26] discussed that the ionic radius of the hydroxyl ion (100-120 pm) dissociated from water vapour was less than that of the oxygen ion (140 pm), and the charge of the former species was also less than that of the latter one. As a result, the diffusion of oxygen through the oxide in the hydroxyl form was more rapid than the diffusion of oxygen through the oxide via the oxygen vacancy [25,26].
Water vapour also helped increase the thickness of the aluminide layer. This should be from the existence of the species sprung from water vapour that dissolved in the aluminide layer. Water vapour can dissociate giving hydrogen and oxygen according to the reaction 2H2O = 2H2 + O2. As

Evolution of the Aluminide Layer during Oxidation
After oxidation in N 2 -12O 2 , alumina was formed at the surface as observed by XRD. This is because the oxygen partial pressure in the atmosphere (0.12 bar) was drastically higher than the one at equilibrium of the alumina formation, as shown in Figure 9. It was reported that alumina was grown at high temperatures mainly by the oxygen inward diffusion [15,[55][56][57], while the outward aluminium diffusion was also possible [57]. The total aluminide layer was thicker than that before oxidation, indicating the diffusion of aluminium into the base cast iron. The dilution of aluminium after oxidation contributed to the increased ratio of iron to aluminium in the aluminide layer and the thicker inner sub-layer. It was also reported that the solubility of silicon in FeAl of the quenched specimen was in the range of 10.38 to 16.24 at.% at the temperature range of 800 to 1020 • C [52]. These values were substantially higher than the solubility of silicon in Fe 2 Al 5 , which was in the range of 2.00 to 4.4 at.% [52]. Because the FeAl sub-layer which had high silicon solubility was thicker after oxidising, it was then possible that silicon in the outer sub-layer diffused to the inner FeAl sub-layer and dissolved in it. This then resulted in the disappearance of the silicon-rich Fe 3 (Al,Si) 5 phase once observed after aluminising.
When the aluminide-coated cast iron was oxidised in N2-12O 2 -10H 2 O, alumina was also formed but the oxidation rate in this atmosphere was higher than the one in N 2 -12O 2 . This should be because water vapour helped form the oxygen-containing species that could diffuse through the oxide faster than the oxygen vacancy, which was the predominant defect for the oxidation in N 2 -12O 2 . Galerie et al. [25,26] discussed that the ionic radius of the hydroxyl ion (100-120 pm) dissociated from water vapour was less than that of the oxygen ion (140 pm), and the charge of the former species was also less than that of the latter one. As a result, the diffusion of oxygen through the oxide in the hydroxyl form was more rapid than the diffusion of oxygen through the oxide via the oxygen vacancy [25,26].
Water vapour also helped increase the thickness of the aluminide layer. This should be from the existence of the species sprung from water vapour that dissolved in the aluminide layer. Water vapour can dissociate giving hydrogen and oxygen according to the reaction 2H 2 O = 2H 2 + O 2 . As previously discussed in Section 4.2, the amount of oxygen dissociated from water vapour in N 2 -12O 2 -10H 2 O was very small compared with the oxygen directly fed to the reactor, and could then be neglected. Thus, the species that occurred from the water vapour and affected the oxidation and properties of the aluminide layer should be hydrogen. It was reported that steel could be embrittled by the dissolved hydrogen with the increased hardness in the zone adjacent to the edge of the fracture [58]. The hardness value of the B2 FeAl was also found to be increased when hydrogen was present in the alloy [59]. Thus, the existence of hydrogen in the aluminide layer could be checked by the increase in the hardness of the aluminide layer. From the hardness measurement on three different areas of the outer sub-layer, it was found that the average hardness values were 364.4 HV (S.D. = 37.9 HV) and 420.2 HV (S.D. = 13.4 HV) for the samples exposed to N 2 -12O 2 and N 2 -12O 2 -10H 2 O, respectively. This confirmed the higher hardness of, and therefore the higher amount of the hydrogen dissolved in, the aluminide layer exposed to the humidified atmosphere. It was reported that the segregation of solute, which was the dissolved hydrogen in the present study, to vacancies lowered the formation energy of those defects [60] and; therefore, increased their concentration [61]. For the Fe-Al system, the formation energy of iron vacancy was reported to be substantially lower than that of aluminium vacancy [61,62]. Thus, the dissolved hydrogen mainly affected the former defect [61] by reducing its formation energy and increasing its concentration. With the increased concentration of iron vacancies, aluminium diffused faster through these sites resulting in the higher diffusivity of aluminium and; therefore, the thicker aluminide layer as observed in the present work.

Conclusions
As a consequence of reducing the exhaust emission from the automotive engine, the exhaust gas temperature tends to be raised, and thereafter the durability of the materials of the exhaust components has to be improved. The present work applied the aluminising pack cementation to the cast iron exhaust manifold with the aim at increasing the high temperature corrosion resistance of this part. The following conclusions can be drawn.

1.
The cyclic oxidation kinetics of the bare cast iron in N 2 -12O 2 -10H 2 O and the aluminide-coated cast iron in N 2 -12O 2 -10H 2 O and N 2 -12O 2 at 850 • C were parabolic, indicating that their oxidation rates reduce when the period of oxidation is longer. Furthermore, aluminising can reduce the oxidation rate of the studied cast iron in both dry and humidified atmospheres.

2.
The presence of water vapour in the atmosphere increased the oxidation rate of the aluminide-coated cast iron and helped thicken the aluminide layer. The higher hardness value of the coated sample oxidised in humidified atmosphere was also observed. This indicated the possible dissolution of hydrogen in the aluminide layer, which related to the enhanced oxidation rate and the thickening of the aluminide layer.