Oxide Scale Formation of Stainless Steels with Different Heating Methods – Effect of Hydrogen as Fuel

The evolution from natural gas usage to new technologies, such as the use of hydrogen as fuel or electricity‐based heating, strongly influences the oxidation of the stainless steel surface in the reheating furnace. Thermogravimetric tests using different simulated combustion and induction reheating conditions are performed for austenitic AISI 301, AISI 304, and ferritic AISI 444 steel grades. Simulated furnace atmospheres in combustion methods are based on methane–air, methane–oxygen, hydrogen–oxygen, and methane–hydrogen–oxygen combinations. For induction simulations, air and nitrogen are used as furnace atmospheres. The results indicate that changes in heating conditions to H2‐fueled combustion or induction only have a minor influence on the oxidation of the ferritic grade; whereas, their effects on the austenitic grades are more pronounced. The transition from a methane–air to H2–oxyfuel combustion increases the total oxidation by 1.7 and 4 times for steel grades 304 and 301, respectively; therefore, grade 304 can be considered better suited for transition for H2–oxyfuel use. The shorter induction heating considerably decreases the amount of oxide scale for austenitic grades, but the nitrogen atmosphere produces a subscale inside the steel matrix, which can hinder the descaling process.


Introduction
The world's stainless steel production was close to 56.3 Mt in 2021, [1] and the total CO 2 emissions from stainless steel production were 2.1 t per ton steel when 85% scrap was used as raw material. [2]High temperatures are required in the various stages of the production process; and thus, emissions can be decreased by developing the heating operations toward a process using green hydrogen and green electricity as energy sources.
The industrial reheating furnace is used to elevate the temperature of slabs for the hot-rolling process.The operating condition in the reheating furnace affects the heating rate of the slabs, the CO 2 emissions, and the fuel consumption of the process. [3]Potential heating methods include fuel combustion and induction heating.Typically, the combustion method is an air-fuel combustion with mainly methane containing natural gas (NG), or alternatively propane (LPG), as fuel gases.The air-fuel combustion consumes energy to heat nitrogen gas, [4] in which the proportion is 79% in the air.Thermal efficiency of the combustion process can be increased by replacing air with pure oxygen either partially (enriched combustion) or fully (oxyfuel combustion). [5]8][9] Without nitrogen carried by air in the furnace, the mass flow rate of flue gas is also decreased, [5] while the proportion of CO 2 in flue gas is increased facilitating its recovery. [5,10,11]The high flame temperature of the conventional oxyfuel burners produces NO x emissions which can be reduced significantly using flameless oxyfuel burners. [3,12]raditional air-methane combustion produces water vapor and carbon dioxide in the furnace atmosphere, and the application of the oxyfuel method would increase the proportions of these components.In contrast, the potential changing of the fuel from methane to hydrogen would strongly increase the proportion of water vapor and exclude CO 2 from the gas atmosphere.Thus, the possibility to reduce CO 2 emissions of the combustion process is the main advantage of hydrogen fuel, and the greatest environmental value would be achieved when green hydrogen is produced by electrolysis using renewable electricity. [13,14]Because hydrogen has a lower energy density compared to methane in relation to volume, maintaining the same heat release rate requires increasing jet velocity or pipe diameter of the CH 4 -air burners, while in hydrogen-methane mixture fuels, combustion can be enabled without adjusting the burners by using a small proportion of hydrogen. [15]owever, reduction of CO 2 emissions by half from methane combustion requires an 80% proportion of hydrogen, [16] while NO x emissions increase by increasing the hydrogen proportion in the mixture fuel due to higher flame temperature. [16,17]Kikuchi et al. [18] presented that NO x emissions from hydrogen-air combustion decrease over 80% by changing conventional burners to flameless burners.In addition, oxyfuel combustion in flameless burners would be a possible technique, using hydrogen as fuel aiming for lower NO x emissions. [19]lectromagnetic induction heating is divided into longitudinal and transverse flux methods according to the magnetic flux direction. [20,21]Longitudinal flux induction heating is suitable for ferromagnetic materials under the Curie point temperature, about 770 °C; while transverse flux induction heating is also possible for nonmagnetic materials with temperatures above the Curie point. [22,23]As a heating method, induction heating is energy efficient, enables faster heating, allows the use of any atmosphere in the furnace, and enables the removal of CO 2 emissions caused by heating compared to fossil fuel combustion. [21,22,24]However, more environmentally friendly use of induction heating requires the utilization of renewable electricity in order for the CO 2 emissions of the process to remain as low as possible. [25,26]emperature, holding time, and atmosphere in the reheating furnace all affect oxide scale formation on the surface of stainless steel slabs.In the reheating furnace, fuel combustion produces an oxidizing atmosphere, when complete combustion of the fuel gas is ensured by using a sufficient excess of an oxidant gas.The amount of free oxygen in the furnace atmosphere has a strong impact on oxide scale growth.In 2 h isothermal oxidations, higher oxygen partial pressure produced a higher oxidation rate of ferritic AISI 441 at 900 °C, [27] while an oxygen content of 3% produced a higher oxidation rate and thicker oxide layer than oxygen content of 5% for ferritic AISI 430 stainless steel at 1,150 °C. [28]In addition, the lean oxyfuel method (no free O 2 ) produces significantly less oxide scale than a 3% free oxygencontaining air-fuel and oxyfuel methods of methane combustions for austenitic AISI 304 in simulating slab reheating. [29]he proportion of water vapor in the furnace atmosphere has an increasing effect on the oxide scale formation of stainless steels. [30,31]In reheating conditions, oxide scale growth in atmospheres containing water vapor has been generally studied using simulated methane combustion with air [32][33][34][35] or humid air atmospheres. [36,37]For example, simulated natural gas combustion with air produces almost an 120 mg cm À2 weight gain for ferritic AISI 430 in isothermal oxidation at 1,250 °C for 2 h, [32] while dynamic heating for 3 h causes only a 17 mg cm À2 weight gain for austenitic AISI 304 using the same target temperature. [29]Using a humid air atmosphere, higher water vapor content has been observed to increase oxide scale layer thickness of high oxidation resistance for ferritic stainless steel in isothermal oxidation at 1,100 °C for 2 h [38] and for AISI 304 at 1,100 °C for 20 min. [39]In addition, for the ferritic grade AISI 430, the oxidation rate was significantly higher using air containing 18% water vapor than dry air in isothermal oxidation at 1,090 °C for 2 h, which increased the total weight gain value from about 20 to 70 mg cm À2 . [40]n this study, simulated furnace conditions and different atmospheres of fuel gas combustion and induction heating were used to compare oxide scale formation during reheating for selected austenitic and ferritic stainless steels.The simulated furnace atmospheres are as follows: methane combustion with air; methane combustion with oxygen; methane-hydrogen combustion with oxygen; and hydrogen combustion with oxygen.Induction heating simulations were performed in air and nitrogen atmospheres.A higher heating rate of induction heating is based on shorter operation time in the furnace compared to combustion heating.The oxide scale layers are characterized in terms of structure, chemical composition, and amount, with a view to optimizing the changes of oxide scale formation in the transition toward CO 2 -free heating methods.

Experimental Section
The austenitic AISI 301, AISI 304, and ferritic AISI 444 (EN 1.4521) stainless steels were supplied by Outokumpu Stainless Oy.The chemical compositions of the steels are presented in Table 1.The samples were cut from a laboratory cast slab using the dimensions of 30 Â 20 Â 2 mm.The actual dimensions of the samples' sides were determined from the average of three measurements and the surface area of the samples was calculated from these.For heating simulations, a hanging hole of 2 mm in diameter was drilled at the top of the sample.Before heating, the samples were cleaned with acetone.
The heating curves of the samples simulated a change in the surface temperature of the industrial slab during the reheating process.The target temperatures were selected based on the temperature ranges of the individual steel grades' industrial heating processes.Therefore, the target temperatures for the austenitic grades 301/304 were 1,225 and 1,275 and 1,150 °C for the ferritic grade 444.The overall time of the simulations was 180 min for fuel combustion heating and 87 min for induction heating.After heating, the sample was removed from the furnace and allowed to cool at room temperature in ambient air.The heating curves of austenitic and ferritic grades with different heating methods are presented in Figure 1.
The laboratory furnace gas atmospheres simulated the industrial furnace heating using calculated combustion products of different fuel-oxidizer combinations.The following combinations were studied: 1) methane-air (MA); 2) methane-oxygen (MO); 3) hydrogen-oxygen (HO); 4) mixture of 50% methane and 50% hydrogen, with oxygen (MHO).Simulations of induction heating were performed in air (IA) and in nitrogen (IN).The furnace atmospheres are calculated from the fuel-oxidizer combinations with excess of oxidant gas using following reaction Equation ( 1)-( 3) The composition of furnace gas atmospheres used is listed in Table 2.The atmospheres were produced by feeding the gas mixture (air, CO 2 , H 2 O, N 2 , O 2 ) into the furnace using a gas flow rate of 2 L min À 1 .
The heating simulations were performed in the thermogravimetry (TG) furnace.The sample was placed in the vertical tube furnace to hang from the digital scale, and the test was started by adding the test gas atmosphere and increasing the furnace temperature according to the simulated temperature curve.The sample weight data was received at 5-s intervals using a computer-controlled TG scale, and the data was normalized to the samples' surface area.The thickness of the attached oxide layer and elemental depth profiles of the oxide scales formed on the sample surface were examined by glow discharge optical emission spectroscopy (GDOES).The determination limit used for the oxide scale thickness was an oxygen content of 100 ppm.For cross-section studies, the samples were cold-mounted in epoxy, sectioned, and polished.The structure and composition of the oxide scales were investigated from the cross-sections by field emission scanning electron microscopy (FESEM, Zeiss ULTRA plus) with an energy dispersive spectrometer (EDS).Backscattered electrons (BSE) were used in FESEM imaging.

Thermogravimetry
Weight gain as a function of time for simulations of different combustion methods in a reheating furnace is presented in Figure 2, and the total weight gain values of all methods and comparison to values obtained in a previous study [29] are shown in Table 3.
The results show that the induction simulations produced significantly less oxide scale compared to the combustion methods.As could be expected, induction in the nitrogen atmosphere produced notably less oxide scale compared to the air atmosphere.The total weight gain values of both austenitic grades were higher for all used atmospheres using the higher target temperature of  These values increased by over 3.5 and 2.5 times, respectively, when the methane combustion method changed from air-fuel to oxyfuel.This result highlights the fact that changing from an air-fuel burn to an oxyfuel burn can drastically increase the growth of oxide scale during slab reheating.
For the oxyfuel methods, the oxide scale growth increased by increasing the proportion of hydrogen in the fuel at both target temperatures, and the change in oxidation is particularly large between fuel gases at a higher target temperature of 1,275 °C.In addition, the onset of higher oxidation rates increased for the oxyfuel methods compared to air-fuel and decreased with an increasing target temperature of heating.The starting points were 6,700 s for MA and 6,100 s for oxyfuel methods using a target temperature of 1,225 °C, and 6,400 s for MA and 5,900 s for oxyfuels at a target temperature of 1,275 °C.
For austenitic grade AISI 304, oxyfuel combustions of methane-hydrogen and hydrogen fuels produce quite similar weight gain curves during heating at the same temperature profile.The total weight gain values of hydrogen-oxyfuel combustion methods were only 0.9 and 2.4 mg cm À2 higher than those using mixed fuel.The onset of a higher oxidation rate started at 6,300 s for MHO atmospheres using both temperatures and for an HO atmosphere using a target temperature of 1,225 °C, and at 6,100 s for HO using a target temperature of 1,275 °C.
Total weight gains of ferritic grade AISI 444 were lower compared to austenitic grades using all methods, and the combustion methods especially produced significantly less oxide scale with values of only 1.0 mg cm À2 for all studied atmospheres using a target temperature of 1,150 °C.The weight gain started to increase about after 6,600 s regardless of the combustion atmosphere.Also, there were almost no notable differences between the simulated gas atmospheres.

Cross-Section Morphology
The cross-section morphologies of oxide scales from AISI 301 after different heating simulations using target temperatures of 1,225 and 1,275 °C are presented in Furthermore, the amount of oxide scale formed was very similar for all oxyfuel methods using a target temperature of 1,225 °C.Based on FESEM cross-sections, the amount of attached oxide scale increased significantly using a higher target temperature for the heating methods of IA, MA, MHO, and HO, while it was relatively the same amount for IN and MO.
For AISI 304, the cross-section morphologies of oxide scales are presented in Figure 4.The morphologies of attached oxide scales can be described as follows: 1) oxide scale nodules for IA-1225; 2) surface and subsurface oxidation for IN methods; 3) oxide scale pockets for MHO-and HO-1225 methods; and 4) oxide scale pockets that partially combined to form internal oxide layer for MHO-and HO-1275 methods.The amount of attached oxide scale is higher for all heating methods using a target temperature of 1,275 °C, but it is significantly higher for the heating methods of MHO and HO.
For the ferritic AISI 444, a thin oxide scale layer covered the surface after all heating methods, as shown in Figure 5.  [29] 17.7 [29] 18.2 19.1 1275 4.1 0.4 20.0 [29] 22.7 [29] 33.0  Induction simulation produced a layer consisting of smaller oxide scale pieces, while after combustion simulations, oxide scale layers were more uniform.

Composition of Oxide Scale
The elemental composition of the oxide scale structure for AISI 301, using different heating methods, is presented in Figure 6.
At the target temperature of 1,225 °C, induction simulation using air atmosphere produced an oxide layer covering the entire surface, as shown in Figure 6a.EDS-maps in Figure 6b show that the oxide layer contained mainly chromium oxide with manganese enrichment, while a continuous silicon oxide layer is not formed at the oxide-metal interface.
Using IA-1275, the oxide scale nodules on the surface of the steel, in Figure 6c, have iron oxide in the outer part, Fe-Cr oxide in the inner part, and a layer-like enrichment of chromium and silicon in the oxide-metal interface, as shown in Figure 6d.The elemental composition of EDS-points is presented in Table 4.The EDS-point analysis shows that the outer part of the nodule consists of about 70 wt% iron, indicating the composition of hematite (Fe 2 O 3 ) in point 1.The inner part of the oxide nodule consists of 34 wt% iron and 30 wt% chromium in point 2, and in the darker phase inside the nodule (in point 3), the chromium content increased in Fe-Cr oxide, and the silicon content is high (8 wt%).
After induction simulations in a nitrogen atmosphere, oxide scale structures, and compositions also have more variations in the surface area.For the IN-1225 method, layer-like subsurface oxidation covers part of the steel surface, and some nodules have formed, as shown in Figure 6e.The subsurface oxidation consists of Cr-oxide with manganese enrichment in the light gray phase and silicon oxide in the dark phases; and oxide nodules have iron-rich on the oxide outer part and Cr-Fe oxide on the inner part, as shown in EDS-maps in Figure 6f.For the IN-1275 method, the subsurface oxide layer is presented in Figure 6g and the oxide scale nodule in Figure 6h.Nodules have an iron oxide upper part (point 4), higher chromium content in Cr-Fe oxide in the surface level of oxide (point 5), and internal oxidation below the oxide scale nodules containing a silicon oxide layer at the oxide-metal interface, as shown in the EDS-maps in Figure 6i.Internal oxidation contains Fe-Cr oxide with nickel enrichment, as shown in point 6.
The combustion simulations of methane-air at 1,225 °C produced oxide pockets into the steel matrix, as shown in Figure 7a.The pockets were mainly oxides containing chromium and iron, while the lighter phase contained more iron and nickel (point 7) and the darker phase more chromium (point 8).In addition, enrichment of silicon is observed on the oxide-metal interface, as seen in point 9 and the EDS-maps in Figure 7b.When using the higher target temperature for MA, the number of oxide scale pockets is higher, the shape of these pockets is rounder, and they extend deeper into the steel matrix (in Figure 3).
When using the oxyfuel methods MO, MHO, and HO at 1,225 °C, the oxide scale is structured also as pockets; however, the amount of attached oxide in pockets increased compared to MA. Figure 7c shows a detached and attached oxide scale of  MO-1225, in which the detached oxide layer contains iron oxide as seen in the EDS-maps in Figure 7d.Correspondingly, the iron oxide layer can be assumed to have broken off from the top of all pockets that formed using other heating conditions.The distribution of elements in the attached Fe-Cr oxide pockets is shown in Figure 7e, where the middle parts of the pockets have higher iron content, enrichment of nickel is observed in small particles in the outer part of the pocket, and silicon oxide-covered oxide-metal interface.
Using the MHO-1275 method, a thick layered oxide scale structure formed, as shown in Figure 7f.The upper layer consists mainly of iron oxide (point 10), while the chromium content increased in the interface of the iron oxide layer and internal oxidation in point 11.In the internal oxidation structure, the composition of oxide is Fe-Cr oxide as seen in point 12. Part of the lower oxide layer is presented in Figure 7g, showing that the darker gray phase consists of mainly Fe-Cr oxide (point 13) and the lighter gray phase of metallic iron and nickel (point 14).
Figure 8 presents the EDS results to characterize elemental composition of the oxide scale structure for AISI 304.Induction simulation in the air at 1,225 °C produced a thick chromium oxide layer and iron-oxide-containing nodules (Figure 8a).Enrichment of chromium oxides is detected at the surface level of the steel, and silicon oxide is formed at the oxide-metal interface, as shown in Figure 8b.At the target temperature of 1,275 °C, an iron oxide layer at the top of the oxide nodules has broken off so that the attached oxide scale forms pocket-like structures inside, growing into the steel matrix, as shown in Figure 4. Induction simulations at the target temperature of 1,225 °C, using a nitrogen atmosphere, produced a chromium oxide layer with enriched manganese, and silicon oxide enrichments inside the steel matrix, which are shown in Figure 8c,d    For combustion simulation methods MHO and HO, oxidescale pockets had a layer-like internal structure at the target temperature of 1,225 °C, as seen in Figure 8h.The darker gray phase consisted of mainly Fe-Cr oxide (point 19), while the lighter gray phase of metallic iron and nickel (point 20).At a higher target temperature of 1,275 °C, the structure of oxide-scale pockets was more netlike formed, and the composition of phases is similar as at a lower temperature, as shown in points 21 and 22 in Figure 8i.
The composition of oxide scales for ferritic AISI 444 at the target temperature of 1,150 °C is presented for the MA-1150 method in Figure 9a.The method produced a layer mainly consisting of chromium oxide (point 23) and some manganese enrichment at the surface of the steel, as shown in Figure 9b.Similar oxide structures formed using all other methods, in which a silicon oxide layer can be seen at the oxide-metal interface as the dark gray phase for every method as seen in point 24, and some metallic parts as light phase is detected as seen in point 25.

GDOES
Examples of GDOES depth profiles of each steel grade are presented in Figure 10.The thickness of the attached oxide layer is shown as a vertical black dashed line in each profile and the results are presented in Table 5.For AISI 301, these results only include induction heating simulations because surface unevenness caused vacuum sealing problems for the device.Respectively, the results of AISI 304 are presented without MHO-and HO-1275.
Figure 10a-d shows GDOES profiles for AISI 301.The thickness of the oxide layer formed using IA-1225 method is 13 μm.The structure of the oxide layer is typical for chromium-rich oxide layers: enrichment of manganese content on the oxidegas interface, chromium-rich oxide layer in the middle, and enrichment of silicon content on the oxide-metal interface.In addition, depleted content of chromium is observed below the oxide layer.Similarly structured oxide layer is observed for IA-1275, but the thickness of the layer is much higher than at lower temperature.For IN-methods, the peak of chromium remains shorter compared to IA-methods indicating more iron-rich oxide layer formed by limiting the amount of oxygen.However, thicknesses of those oxide layers are of the same magnitude than chromium oxide layer of IA-1225.
Selected GDOES profiles for AISI 304 are presented in Figure 10e-h.IA-1225 method produced a typical structure for chromium oxide layer, and a similar structure is also observed using nitrogen.HO-1225 profile is an example of the combustion methods at 1,225 °C, in which similar changes such as manganese enrichment and chromium depletion are not observed indicating a more Cr-Fe oxide-like structure than chromium-rich oxide layer.In addition, the shape of the profile for IA-1275 method resembles those combustion simulations at a lower temperature.However, the IN-1275 profile still has small signs of chromium depletion even if the chromium content near the surface is lower than other methods for AISI 304.
Based on the results of AISI 304, the IN-1225 method has the lowest thickness (9 μm), while changing the furnace atmosphere to a more oxidizing oxide scale thickness increased to 33 μm for the IA-1225 method, and the thickness increased even more using combustion methods MHO and HO with the same target   temperature.In addition, a higher target temperature of 1,275 °C increased thickness values for induction methods producing a similar thickness for IA-1275 methods as combustion methods using a lower target temperature.
For AISI 444, the shape of the GDOES profiles is similar for each of the methods as presented in Figure 10i-l.The thickness of the oxide scale was smaller for induction methods than combustion methods so that the thinnest layer was 8 μm for the IN method, and the thickest layer was 15 μm for MHO and HO methods as seen in Table 5.

Effect of Heating Method on Oxidation of Austenitic Grades
The selected heating method of stainless steel slabs in the reheating process can have significant effects on the oxide scale formed on the surface of the steel.In this study, the most commonly used method in the industry, methane burning with air, produced oxide scale pocket structures for the austenitic grade AISI 301.For AISI 304, a previous study by Laukka et al. [29] also found a pocket structure of scales when using methane-air.
By changing the oxidizing gas of the heating method from air to oxygen in the burning of methane, water vapor content increases by four times in the furnace atmosphere as can be seen in Table 2.[40] Potential methods of effect of water vapor have been reviewed by Saunders et al. [41] In this study, total weight gain values between MA and MO showed that oxidation increased over 3.5 times using the target temperature 1,225 °C for AISI 301 by changing the method, while it was only 9% for AISI 304.Correspondingly, the increase was 2.5 times for AISI 301 and 14% for AISI 304 at the target temperature of 1,275 °C.The significantly higher amount of oxide scale for AISI 301 between methods is a result of more progressed oxidation, because the oxide scale structure developed from few individual oxide pockets to wider pockets that cover almost the entire surface area, using both target temperatures for AISI 301.In contrast, the pocket structure did not have major differences between different methods for AISI 304. [29]Thus, based on total weight gain values and oxide scale structures, changing the heating method from MA to MO has a higher effect on oxidation for grade AISI 301 than grade AISI 304.
Modification toward nonfossil-fueled methods requires the utilization of hydrogen as the fuel gas or changing the heading method to induction.For the MHO method, adding hydrogen to the fuel increased the water vapor content by 20%; whereas, when changing the fuel completely from methane to hydrogen in oxyfuel combustion, the water vapor content increased 50% compared to the MO methods (Table 2).However, even the highest increased water vapor content had no effect on the structure of oxide scale and only had minor effects on the amount of scale formed for both austenitic grades at the target temperature of 1,225 °C.
For both steel grades, a higher amount of water vapor in the gas atmosphere, based on hydrogen combustion methods compared to methane combustion (MO), is observed to have a major impact on the formed oxide scale at the higher target temperature 1,275 °C.Relatively, the amount of oxide scale changed even more for steel grade AISI 304 than AISI 301, because the increased amount of oxide scale was 45% in MHO and 56% in HO for AISI 304 and only 11% in MHO and 48% in HO for AISI 301.The structure of oxide scales reveals a significant effect on H 2 O contents when transferring from MO to hydrogenfueled methods, when oxide pockets grow larger and wider, combining into a partial internal layer for AISI 304 and a completely internal layer beneath the iron oxide layer for AISI 301.Thus, these amounts of hydrogen fuel in oxyfuel methods cause major increases in material losses when removing the oxide scale.The air-fuel method could be an alternative for using nonfossil hydrogen fuel at the higher slab's heating temperatures from the viewpoint of oxide scale formation, because combustion will only cause approximately 30% H 2 O content to post-burn atmosphere using similar oxygen content as in this study, which is an even lower content than the MO method.Another potential method to reduce material losses from hydrogen-fueled heating could be to minimize the amount of free oxygen in the post-burn atmosphere, which on laboratory scale showed the potential to decrease the oxidation rate of AISI 304 when using a lean oxyfuel method for methane. [29]urthermore, the studied oxyfuel simulations revealed the worst-case scenario of oxidation, because the oxyfuel method allows for a shorter exposure time in the furnace than the air-fuel method due to a higher heating rate. [8]The faster heating of oxyfuel methods could affect the difference between the oxide scales of air-fuel and oxyfuel methods.Adolfi et al. [42] presented that the air-fuel method increased time of reheating by about 30% compared to the oxyfuel method in a batch furnace.However, Hu et al. [9] determined that the reduction of heating time between air-fuel and oxyfuel methods was only 6% in the slabs' reheating process of 200 min; thus, its effects on oxidation could be minor.
For induction heating, the heating time is significantly shorter compared to combustion heating.In addition to the shorter heating time, a different heating profile and atmospheres are the reasons that induction heating simulations produced considerably less oxidation and different oxide structures compared to combustion heating methods.The most commonly seen oxide structure of combustion methods was the pocket structure of oxide scale, which for induction simulations only formed for the austenitic grades on AISI 304 when using the IA-1275 method.
A notable effect from the differences in free oxygen content in the gas atmosphere on oxide scale formation was observed between the IA and IN methods.The higher oxygen content from air produced a more homogeneous oxide structure compared to the nitrogen atmosphere, which contained only minimal residual oxygen mainly based on the open-top structure of the furnace.For the IA method, the development of the oxide structure from the Cr-oxide layer to the iron-oxide-containing nodules and to the layered structure for longer exposure time [43] follows the development of oxide structures in combustion-based heating, while more heterogeneous oxide structures from the IN method differs from it.The layer-like subsurface oxide was especially observed as a characteristic oxide structure for IN methods (Figure 11), in which dark gray oxide is Cr-oxide with enrichment of manganese and almost black oxide is silicon oxide.The formation of Cr-oxide inside the steel matrix is an indication of the oxygen partial pressure being too low to form a protective and uniform chromium oxide layer.In a previous study, the use of air/CH 4 ratio, thus a low oxygen partial pressure, also prevented the formation of a protective Cr-oxide layer for AISI 304. [44]An oxide layer formed inside the steel matrix for IN methods could impair the descaling process compared to oxide formed on the surface using IA methods.Different heating profiles in induction heating concerning the surface of the slab can affect the oxide scale formation.In this study, the induction simulation profile is based on the slab heating, in which the purpose would be to keep the difference between the surface and core temperatures small. [45]The heating time can be even shorter by accelerating the heating of the slab's surface, while the temperature difference to the core increases [45,46] and thus, the difference between oxide scales from induction and combustion heating could increase even more.
The overall oxidation progression of austenitic grades in the slab's reheating simulations is presented in Figure 12, in which oxide scales are divided according to amount and structure.The structure of oxide scales is divided into five categories: Cr-oxide layer, subsurface oxidation and nodules, oxide nodules, oxide pockets, and layered oxide.Figure 12 shows that, for both short IA-methods, oxide scales of steel grades belong to different categories.The nodule formation and development into the pocket structures as the outer layer is broken off indicates earlier breakaway oxidation for AISI 304 compared to AISI 301.However, the combustion methods increase the progress of oxidation significantly more for grade AISI 301, because in the change of the heating method from MA to HO the amount of oxidation increased 4 times for grade 301 and only 1.7 times for grade 304 at the higher target temperature of 1,275 °C.MA are the only methods for AISI 301 that have a relatively similar amount of oxide scale than almost all combustion methods for AISI 304, from which all oxyfuel methods of AISI 301 differ according to the amount of oxide scale.Based on oxidation progress, steel grade AISI 304 is more suitable to fuel gas transition from methane to hydrogen, while induction heating could reduce oxide scale formation more for AISI 301.

Difference in Composition of Oxide Scales between Steel Grades
For ferritic steel grade AISI 444, the onset of breakaway oxidation is not observed as iron oxide and nodule-kind structure formation by any used heating methods; whereas, nodules and pocket structures were major structures for both austenitic grades.Correspondingly, Cheng et al. [38] noticed that the difference in the humidity of air has only minor effects on oxidation of ferritic stainless steel in reheating conditions.The lower target temperature and faster diffusion rates of chromium [47] and silicon [48] for ferritic compared to austenitic grades slows down the progression of oxidation due to the formation of a continuous thick Cr-oxide layer and Si-oxide layer beneath it, which acts effectively as a diffusion barrier.Thus, in the studied reheating conditions  and heating methods, ferritic AISI 444 is more appropriate for changing the heating method than the austenitic grades.
Austenitic steel grades have different contents of alloying elements, in which higher silicon and lower chromium contents for AISI 301 than AISI 304 are particularly important in terms of breakaway oxidation.At the beginning of breakaway oxidation, oxide nodules from heating methods IA-1275 for AISI 301 and IA-1225 for AISI 304 have a similar structure.However, in the larger oxide pocket structures, the differences between compositions are accentuated.Figure 13a,b shows an oxide pocket of MA-1275 methods for AISI 301, in which there are Ni-rich metallic particles and pores throughout the area of the internal pocket structure, no healing layer of Cr-rich oxide and a wide layer of Si-rich oxide enrichment in the oxide-steel interface.Due to increased water vapor content in the atmospheres, oxidation progressed further with oxyfuel methods already at 1225 °C target temperatures.Then, the nickel-rich particles are observed closer to the oxide-steel interface using MO-1225 method (Figure 7e), and finally, these particles disappeared making a more uniform oxide structure containing Cr, Fe, and Ni (point 26 in Table 6) using HO-1225 method, as shown in Figure 13c-e.For AISI 304, the development of oxide structure and composition happens from multilayered (Figure 13f and 8h) to network-like nickel-rich metal particles (Figure 8i), while a similar uniform Ni-containing oxide structure is not formed.
The mechanism of breakaway oxidation combined with alloy compositions would cause the difference in oxide scale structure and composition of austenitic grades.][51][52][53] The mechanisms agree that nodule formation begins by nucleation after the Cr-rich oxide layer loses its protectiveness in the center of the grains, because the protection is maintained better near the grain boundaries, which act as diffusion channels for chromium.Col et al. [49] presented that internal spinel oxide (FeCr 2 O 4 ) forms below the nodule and grows internally toward the grain boundaries, and diffusion and oxidation of chromium and iron cause pores and nickelenriched metallic particles in the pocket structure, in which nickel is not oxidized due to deficient oxygen partial pressure.
In the progression of breakaway oxidation, multilayer-like structured AISI 304 pockets are formed before a network-like structure.A similar multilayer structure of oxide and metallic phases in internal oxidation is also observed using methane combustion [29,54] and steam atmospheres. [55]Cheng et al. [30] presented a mechanism for the formation of a multilayer structure for AISI 316, which starts with the formation of Cr-rich oxide and Cr depletion below it.When critical Cr content is reached in the oxide-steel interface, internal oxidation begins by the formation of Fe-Cr oxide producing Ni-rich metallic particles in the depletion zone.Oxidation of Fe-Cr oxide continues toward the steel matrix until the depth where the content of Cr is enough for the formation of a new protective Cr-rich layer.
For AISI 301, a thick and continuous Cr-rich oxide healing layer is formed in the oxide-steel interface of pockets (point 27 in Figure 13e), which is reported to reduce the oxidation rate, [48,56] while a similar thick healing layer is not formed for the AISI 304 grade.The difference between the healing layers' formation may be influenced by higher silicon content of AISI 301 compared to AISI 304.This enables higher silicon enrichment and formation of the continuous silicon oxide layer at the Cr-oxide and steel interface, which acts as a diffusion barrier. [56]The scale of AISI 301 contains dark, silicon-rich oxide phases, in both the outer (point 28) and inner interfaces (point 29) of the healing layer, while the scale of AISI 304 has a lower Si content and a less thick and uniform region of enriched silicon in the pocket structures.The interface structure could also affect the oxidation of nickel by allowing the oxygen partial pressure to be sufficient, and thus, the nickel-rich metallic particles disappear in the pocket structure of AISI 301.Based on the results of this study, the selection of the reheating method may affect the descaling process after reheating.For ferritic grade 444, a similar layer structure of oxide scales was formed after every heating method, and thus, effects on the efficiency of descaling are expected to be minor.For austenitic grades, a thick iron oxide layer and different internal oxide structures between steel grades, temperatures, and heating methods may influence the descaling process.Mechanical descaling by bending the steel progressed in the oxide scale along the continuous chromite phases.Continuous chromite phases are formed in the grain boundaries of the steel matrix in the network-like structured internal oxidation of AISI 304.The formation of chromite phases was promoted by a longer heating and higher oxygen content in the atmosphere. [33]Hydraulic descaling by water jets has been shown to affect oxide scale by thermal shock and mechanical pressure. [57]The efficiency of hydraulic descaling is better for oxide scale structure with oxidized nickel compared to network-like structure with metallic nickel-rich phases. [53]hus, AISI 301 oxides with oxidized nickel may be more suitable for the descaling process than AISI 304.However, the shape of the internal pockets of AISI 301 could decrease the efficiency of descaling due to the rounder pockets being more protected by the steel matrix than wider pockets of AISI 304.

Conclusion
The effect of the heating method on oxide scale formation in the reheating furnace is studied for austenitic AISI 301 and AISI 304, and ferritic AISI 444 stainless steels.For ferritic grade, the change of heating method did not have significant effects on the structure or amount of formed chromium oxide scale layer.For the austenitic grades, simulated induction heating conditions considerably decreased the amount of oxide scale compared to combustion methods.IN method reduced oxidation even further, but it caused subsurface oxidation of chromium.Furthermore, the simulated combustion of either H 2 or a CH 4 -H 2 mix caused a prominently different structure and amount of oxide scale for both austenitic grades compared to CH 4 -air combustion when using the higher target temperature of 1,275 °C.
Based on the results of this study, grade AISI 301 is even more suitable for shorter induction heating, while for the transition from CH 4 to H 2 as fuel gas or for utilizing the oxyfuel method in combustion, grade AISI 304 is more appropriate.The total increase in the amount of oxidation was 4 times for AISI 301 and only 1.7 times for AISI 304 from MA to HO at 1,275 °C.Only the MA method for AISI 301 at 1,225 °C, from all combustion methods, produced less oxide scale than the corresponding methods of AISI 304.
A higher water vapor content of gas atmosphere is not as considerable for AISI 304 compared to AISI 301.Only the most demanding reheating conditions, as H 2 and CH 4 -H 2 mix fuel combustions at a higher temperature, increased oxidation significantly for AISI 304, while all oxyfuel methods, regardless of the fuel gas and temperatures, produced that kind of increase compared to the MA method for AISI 301.
The internal pocket structure below the thick iron oxide layer is the main oxide structure for both austenitic steel grades using combustion heating methods.The higher silicon content of AISI 301 compared to AISI 304 could explain the difference between their oxidation behaviors in pocket structures.Oxyfuel methods and higher target temperatures proceeded oxidation further for both austenitic grades, which created a thick healing layer of Crrich oxide with silicon oxide in the oxide-steel interface and oxidation of nickel in the pocket structure for AISI 301, and developed the internal pocket structure from multilayered to network-like structured for AISI 304.

Figure 1 .
Figure 1.Target heating curves for different heating methods: C = combustion and I = induction.

Figure 3 .
The cross-section morphologies of attached oxide scales are described as follows: 1) thin oxide scale layer for IA-1225; 2) oxide scale nodules for IA-1275; 3) layer-like subsurface oxidation and oxide scale nodules for IN-1225 and IN-1275; 4) oxide scale pockets for both target temperatures of MA and MO, MHO-1225, and HO-1225; and 5) thick layered oxide scale for MHO-1275 and HO-1275.

Figure 3 .
Figure 3. FESEM cross-section morphologies of AISI 301 oxide scales.The length of the black scale bar in each figure is 200 μm.

Figure 4 .
Figure 4. ESEM cross-section morphologies of AISI 304 oxide scales.The length of the black scale bar in each figure is 200 μm.

Figure 5 .
Figure 5. FESEM cross-section morphologies of AISI 444 oxide scales at 1,150 °C.The length of the black scale bar in each figure is 20 μm.

Figure 6 .
Figure 6.Composition of oxide scale structure of AISI 301 after induction simulations: a) IA-1225, b) EDS-maps of part (a), c) IA-1275, d) EDS-maps of part (c), e) IN-1225, f ) EDS-maps of part (e), g) subsurface oxide layer of IN-1275, h) oxide scale nodule of IN-1275, and i) EDS-maps of part (h).
. Another oxide-scale structure for IN-1225 was Fe-Cr oxide on the surface (point 15) and subsurface oxidation below it, as seen in Figure8e.The composition of subsurface oxidation was Fe-Cr oxide with silicon enrichment (point 16).For IN-1275, similarly composed subsurface oxidation had a more layer-like structure (Figure8f),

Figure 7 .
Figure 7. Oxide scale composition of AISI 301 after combustion simulations: a) MA-1225, b) EDS-maps of part (a), c) MO-1225, d) EDS-maps of detached oxide in part (c), e) EDS-maps of oxide scale pocket in part (c), f ) layered oxide structure of MHO-1275, and g) netlike-structured oxide in part (f ).
and surface oxide-scale consisted of iron oxide (point 17) and Fe-Cr oxide, as shown in point 18 in Figure8g.

Figure 8 .
Figure 8. Oxide scale composition of AISI 304: a) IA-1225, b) EDS-maps of part (a), c) oxide scale layer of IN-1225, d) EDS-maps of part (c), e) subsurface oxidation of IN-1225, f ) subsurface oxidation of IN-1275, g) surface oxide-scale of IN-1275, h) layer-like oxide pocket structure of MHO-1225, and i) netlike oxide pocket structure of MHO-1275.

Table 2 .
The compositions of furnace gas atmospheres used.

Table 3 .
Total weight gain, in mg cm À2 .

Table 5 .
Thickness of oxide scale measured by GDOES.

Table 6 .
Elemental contents of EDS points from Figure13in wt%.