Effect of Mg Contents on Catalytic Activity and Coke Formation of Mesoporous Ni/Mg-Aluminate Spinel Catalyst for Steam Methane Reforming

Ni catalysts are most suitable for a steam methane reforming (SMR) reaction considering the activity and the cost, although coke formation remains the main problem. Here, Ni-based spinel catalysts with various Mg contents were developed through the synthesis of mesoporous Mg-aluminate supports by evaporation-induced self-assembly followed by Ni loading via incipient wetness impregnation. The mesoporous Ni/Mg-aluminate spinel catalysts showed high coke resistance under accelerated reaction conditions (0.0014 gcoke/gcat·h for Ni/Mg30, 0.0050 gcoke/gcat·h for a commercial catalyst). The coke resistance of the developed catalyst showed a clear trend: the higher the Mg content, the lower the coke deposition. The Ni catalysts with the lower Mg content showed a higher surface area and smaller Ni particle size, which originated from the difference of the sintering resistance and the exsolution of Ni particles. Despite these advantageous attributes of Ni catalysts, the coke resistance was higher for the catalysts with the higher Mg content while the catalytic activity was dependent on the reaction conditions. This reveals that the enhanced basicity of the catalyst could be the major parameter for the reduction of coke deposition in the SMR reaction.


Introduction
With the increasing concerns about air pollution and global warming, the development of a clean and alternative energy source to depleting fossil fuels has drawn great attention. Hydrogen, with strengths such as cleanness and high energy content, has been considered to be one of the most promising energy carriers [1]. Although the eventual method of hydrogen production should be renewable or low-carbon, such as via electrolysis of water, it is still highly costly and the majority of hydrogen production is based on fossil fuels. Steam methane reforming (SMR) is the most common and successful process of hydrogen production from natural gas, accounting for 76% of hydrogen production worldwide [2]. From the viewpoint of energy resources, the production of hydrogen from natural gas can be a smooth transition from fossil fuel-based energy system to renewable counterpart. Various transition-metal-based catalysts have been developed as SMR catalysts, including Pt, Rh, Ru, Pd, and Ni [3][4][5][6]. Considering the activity and the cost, Ni catalysts are regarded to be the most suitable, and the commercial SMR catalyst is Ni-based as well. Although it is a highly matured process, coke formation is still a main problem for the SMR reaction, together with sintering, which easily leads to the deactivation of catalysts [7,8].
of the Mg contents on the physicochemical properties of the spinel catalysts from viewpoints of the sintering resistance and the Ni particle formation by the exsolution, which has been less studied before. Figure 1 shows TEM images of Mg-aluminates calcined at 900 • C. In the case of Mg9 and Mg19 with low Mg content (Mg/Al molar ratios of 0.10 and 0.23, respectively), well-defined and ordered mesopores formed by decomposition of the self-assembled P123, namely SDA, were clearly observed (Figure 1a,b). The insets in Figure 1a,b show line-scanned profiles displaying the width of pores and walls with regularity. The averaged pore sizes were measured to be 14.09 nm and 16.84 nm for Mg9 and Mg19, respectively. However, in the case of Mg25 and Mg30 with higher Mg content (Mg/Al molar ratios of 0.32 and 0.43, respectively), no well-defined or ordered mesopores were found (Figure 1c,d). As shown in TEM images of Supplementary Figure S1, well-defined and ordered mesopores were maintained regardless of Mg content in the case of Mg-aluminate calcined at 600 • C. This means that well-defined and ordered mesopores of Mg-aluminate with high Mg content (Mg25 and Mg30) were collapsed by sintering below 900 • C. The Mg/Al molar ratio of Mg30 is close to that of the MgAl 2 O 4 spinel phase. Koo et al. reported severe sintering of MgAl 2 O 4 as a catalyst support with the increase of calcination temperature from 800 to 1200 • C [18]. A similar trend was observed with their Ni-loaded catalysts. The BET surface area, in that study, decreased gradually from 108 to 15 m 2 /g, while the crystallite size of Ni-MgAl 2 O 4 increased from 4 to 53 nm under 800-1200 • C calcination.

Results and Discussion
Catalysts 2020, 10, x FOR PEER REVIEW 3 of 15 the Mg contents on the physicochemical properties of the spinel catalysts from viewpoints of the sintering resistance and the Ni particle formation by the exsolution, which has been less studied before. Figure 1 shows TEM images of Mg-aluminates calcined at 900 °C. In the case of Mg9 and Mg19 with low Mg content (Mg/Al molar ratios of 0.10 and 0.23, respectively), well-defined and ordered mesopores formed by decomposition of the self-assembled P123, namely SDA, were clearly observed (Figure 1a,b). The insets in Figure 1a,b show line-scanned profiles displaying the width of pores and walls with regularity. The averaged pore sizes were measured to be 14.09 nm and 16.84 nm for Mg9 and Mg19, respectively. However, in the case of Mg25 and Mg30 with higher Mg content (Mg/Al molar ratios of 0.32 and 0.43, respectively), no well-defined or ordered mesopores were found (Figure 1c,d). As shown in TEM images of Supplementary Figure S1, well-defined and ordered mesopores were maintained regardless of Mg content in the case of Mg-aluminate calcined at 600 °C. This means that well-defined and ordered mesopores of Mg-aluminate with high Mg content (Mg25 and Mg30) were collapsed by sintering below 900 °C. The Mg/Al molar ratio of Mg30 is close to that of the MgAl2O4 spinel phase. Koo et al. reported severe sintering of MgAl2O4 as a catalyst support with the increase of calcination temperature from 800 to 1200 °C [18]. A similar trend was observed with their Ni-loaded catalysts. The BET surface area, in that study, decreased gradually from 108 to 15 m 2 /g, while the crystallite size of Ni-MgAl2O4 increased from 4 to 53 nm under 800-1200 °C calcination. The change of the pore structures of the Mg-aluminates by sintering in the present study was confirmed by BET and BJH analysis ( Figure 2). There was a very clear trend of decreasing BET surface area and increasing pore size with increasing Mg content or Mg/Al ratio ( Figure 2 and Table S1 in the supplementary materials). In Figure 2a, the N2 adsorption-desorption isotherms of all of the supports show typical mesopore type IV hysteresis curves. The mesoporous supports could be classified in more detail by the Langmuir theory. The hysteresis loop was close to the H1 type, facile pore connectivity and high pore size uniformity, all of which reflect the typical pore-structure form by The change of the pore structures of the Mg-aluminates by sintering in the present study was confirmed by BET and BJH analysis ( Figure 2). There was a very clear trend of decreasing BET surface area and increasing pore size with increasing Mg content or Mg/Al ratio ( Figure 2 and Table S1 in the Supplementary Materials). In Figure 2a, the N 2 adsorption-desorption isotherms of all of the supports show typical mesopore type IV hysteresis curves. The mesoporous supports could be classified in more detail by the Langmuir theory. The hysteresis loop was close to the H1 type, facile pore connectivity and high pore size uniformity, all of which reflect the typical pore-structure form by decomposition of self-assembled P123 [19,33]. The pore size of mesoporous metal oxide prepared by the soft-templating method is dependent on the SDA. Previous studies show that, in the synthesis of alumina, triblock copolymer Pluronic F127 as the SDA generates smaller mesopores (4-7 nm) than those (upto 15 nm) formed by P123. Furthermore, surfactants such as cetrimonium bromide and sodium dodecyl sulfate as the SDA generate even smaller mesopores [35]. Considering mass transport, the catalyst support with larger mesopores would be more advantageous in the SMR reaction, which is the main reason why P123 is used in this study. The BJH plots in Figure 2b show the pore size distribution with narrow peaks between 6 and 20 nm. Smaller pores were formed for Mg-aluminate with lower Mg/Al ratio. As shown in the Table S1, the BET surface area was increased from 93 m 2 /g (Mg30) to 170 m 2 /g (Mg9). Interestingly, the pore volumes of the Mg-aluminates were similar regardless of the Mg/Al ratio (Supplementary Table S1). From these results, it can be inferred that the resistance against sintering is higher for Mg-aluminate with a lower Mg/Al ratio, and therefore, mesopores formed by decomposition of SDA are maintained better after calcination. On the other hand, for the Mg-aluminate with higher Mg/Al ratio, the nanocrystals were sintered to form enlarged intraparticle pores while the total pore volume maintained was similar. After Ni loading with calcination and after reduction, the BET surface area was decreased while the trend of BET surface area, d p value, and mean pore size according to Mg contents were similarly maintained ( Figure 2).

Results and Discussion
Catalysts 2020, 10, x FOR PEER REVIEW 4 of 15 decomposition of self-assembled P123 [19,33]. The pore size of mesoporous metal oxide prepared by the soft-templating method is dependent on the SDA. Previous studies show that, in the synthesis of alumina, triblock copolymer Pluronic F127 as the SDA generates smaller mesopores (4-7 nm) than those (upto 15 nm) formed by P123. Furthermore, surfactants such as cetrimonium bromide and sodium dodecyl sulfate as the SDA generate even smaller mesopores [35]. Considering mass transport, the catalyst support with larger mesopores would be more advantageous in the SMR reaction, which is the main reason why P123 is used in this study. The BJH plots in Figure 2b show the pore size distribution with narrow peaks between 6 and 20 nm. Smaller pores were formed for Mg-aluminate with lower Mg/Al ratio. As shown in the Table S1, the BET surface area was increased from 93 m 2 /g (Mg30) to 170 m 2 /g (Mg9). Interestingly, the pore volumes of the Mg-aluminates were similar regardless of the Mg/Al ratio (Supplementary Table S1). From these results, it can be inferred that the resistance against sintering is higher for Mg-aluminate with a lower Mg/Al ratio, and therefore, mesopores formed by decomposition of SDA are maintained better after calcination. On the other hand, for the Mg-aluminate with higher Mg/Al ratio, the nanocrystals were sintered to form enlarged intraparticle pores while the total pore volume maintained was similar. After Ni loading with calcination and after reduction, the BET surface area was decreased while the trend of BET surface area, dp value, and mean pore size according to Mg contents were similarly maintained ( Figure 2).

Figure 2.
BET and BJH analysis of (a-c) Mg-aluminate support, (d-f) after Ni loading with calcination, and (g-i) after reduction: (a,d,g) N2 adsorption-desorption isotherms, (b,e,h) BJH pore size distributions, and (c,f,i) plots of BET surface area, dp value, mean pore size versus Mg contents. The synthesized Mg-aluminate showed the spinel phase. The degree of sintering (or sintering resistance) of Mg-aluminate according to the Mg/Al ratio could be also confirmed by XRD (Figure 3a). The XRD pattern of Mg30 was close to that of MgAl 2 O 4 (JCPDS #21-1152, Supplementary Table S2). As the ratio of Mg/Al was lowered, the peaks were right shifted, and the XRD pattern of Mg9 was close to that of Mg 0.388 Al 2.404 O 4 , which is known as the defect spinel phase (JCPDS #48-0528, Supplementary  Table S2) [36]. The main cause of the high thermal stability of Mg9 and Mg19 against sintering, which leads to well-developed and ordered mesopores, as shown in Figure 1, is considered to be Mg deficiency at the octahedral sites of the spinel [36]. The peaks of Mg19 and Mg25 were between those of Mg9 and Mg30. This peak shift was observed more clearly for the peak of the (440) plane. For careful peak identifications, the peaks of the (440) plane were magnified as shown to the right of the XRD patterns in Figure 3. The 2-theta of the (440) plane decreased from 66.25 degrees for Mg9 to 65.35 degrees for Mg30.
After Ni impregnation into Mg-aluminate, NiO peaks were observed in the XRD patterns ( Figure 3b). The ICP analysis shows that all the samples have equivalent Ni contents ( Table 1). The peak intensity, however, decreased with decreasing Mg/Al ratio, and no peak was found in Mg9-Ni. The reason for the absence of an NiO peak in Mg9-Ni was the filling of Mg-deficient sites (A in AB 2 O 4 ) by Ni atoms to form the spinel MgNiAl 2 O 4 phase. On the other hand, Mg19, Mg25 and Mg30 had no or not enough defect sites of A in AB 2 O 4 to be filled by added Ni atoms, and therefore, surplus Ni atoms not included in the spinel matrix formed the NiO phase, as indicated in Figure 3b. It is notable that the diffraction peaks were left-shifted by Ni impregnation (Figure 3b,d). The width of the peak shift reflected the extent of A-site deficiency in the support samples or, equally, the amount of Ni filled into the impregnated samples. There was almost no peak shift for Mg30, and its width increased with decreasing Mg/Al ratio, which phenomenon inversely matched the intensity of the NiO peak.
After the reduction of the impregnated samples, Ni peaks were clearly observed, and the peaks of the spinel were right shifted, far more than even the pre-impregnation support. The crystallite size of Ni calculated using the Scherrer equation based on the peak at 2θ = 51.85 • , (200) plane was smaller along with lower Mg contents (Table 1). This is more evident with TEM observation (Figure 4 and Table 1). There have been several researches of Mg effect on the size of Ni in Ni-Mg-Al oxides [17,[28][29][30]. Different from previous studies, the change of Ni crystal size by Mg content could be explained based on exsolution during reduction process. The right shift of the diffraction peaks was owed to exsolution of the Ni, which resulted in renewed A-site deficiency. In the spinel (AB 2 O 4 ) or perovskite (ABO 3 ) structure, easily reducible cations at the A or B sites, respectively, were diffused out to the surface and formed metal nanoparticles under the reducing atmosphere, which effect is known as the exsolution process [37][38][39]. Here, the extent of exsolution or width of right shift was proportional to the width of left shift by Ni impregnation and the extent of A-site deficiency (Figure 3d). The peak shift of Mg30, again, was negligible (Figure 3c,d). The greater post-reduction right shift of the diffraction peak relative to the peak position of the support could be attributed to the partially exsolved Mg atoms in addition to the exsolved Ni atoms.    The H 2 -TPR analysis revealed that there were two weak reduction peaks (the satellite peaks around 500 and 600 • C) and two strong peaks at around 750 and above 800 • C ( Figure 5). The reduction peaks at around 500 and 600 • C were attributed to the reduction of bulk NiO species weakly interacting with the support. The peaks at around 750 • C could be attributed to the reduction of complex NiO x species strongly interacting with the support [32,36]. The peaks above 800 • C were attributed to the reduction of NiAl 2 O 4 with or without MgO or CaO promotion, which was not observed only for Mg30-Ni [17,19]. This was due to the fact that Mg30 was stoichiometrically full, forming the MgAl 2 O 4 spinel structure. After Ni impregnation, Ni was not embedded into the A site, and therefore, the NiAl 2 O 4 phase was not formed. On the other hand, the peak at around 750 • C, attributable to NiO species reduction and strongly interacting with the support, was not observed only for Mg9-Ni. This is consistent with the XRD data in Figure 3b showing no evident NiO for Mg9-Ni, since all of the Ni atoms had been accepted into the A site to form the NiAl 2 O 4 phase. The order of peak intensity at around 750 • C was as follows: Mg30-Ni > Mg25-Ni > Mg19-Ni. This is in line with the trends of NiO peak size and extent of peak shift in the XRD patterns (Figure 3b,d).  The H2-TPR analysis revealed that there were two weak reduction peaks (the satellite peaks around 500 and 600 °C) and two strong peaks at around 750 and above 800 °C ( Figure 5). The reduction peaks at around 500 and 600 °C were attributed to the reduction of bulk NiO species weakly interacting with the support. The peaks at around 750 °C could be attributed to the reduction of complex NiOx species strongly interacting with the support [32,36]. The peaks above 800 °C were attributed to the reduction of NiAl2O4 with or without MgO or CaO promotion, which was not observed only for Mg30-Ni [17,19]. This was due to the fact that Mg30 was stoichiometrically full, forming the MgAl2O4 spinel structure. After Ni impregnation, Ni was not embedded into the A site, and therefore, the NiAl2O4 phase was not formed. On the other hand, the peak at around 750 °C, attributable to NiO species reduction and strongly interacting with the support, was not observed only for Mg9-Ni. This is consistent with the XRD data in Figure 3b showing no evident NiO for Mg9-Ni, since all of the Ni atoms had been accepted into the A site to form the NiAl2O4 phase. The order of peak intensity at around 750 °C was as follows: Mg30-Ni > Mg25-Ni > Mg19-Ni. This is in line with the trends of NiO peak size and extent of peak shift in the XRD patterns (Figure 3b,d).   Figure 6a shows the time-on-stream CH4 conversion in the SMR reaction with increasing GHSV at 800 °C with a steam-to-carbon ratio (S/C) of 3. For comparison, a commercial catalyst was reacted under the same conditions. At the reaction temperature of 800 °C with S/C = 3, the calculated CH4 conversions at the thermodynamic equilibrium is 100%. All catalysts except Ni/Mg9 reached the thermodynamic equilibrium at the GHSV of 8000 h −1 . CH4 conversion of 100% was maintained up to the GHSV of 16,000 h −1 for the developed catalysts other than Ni/Mg9, whereas the commercial catalyst showed decreased conversion at the GHSV of 16,000 h −1 . The CH4 conversions at the GHSV of 24,000 h −1 were in the following order: Ni/Mg25 > Ni/Mg30 > Ni/Mg19 > Ni/Mg9 > commercial catalyst. At the higher GHSV, the more likely SMR will be controlled by diffusion phenomena.  Figure 6a shows the time-on-stream CH 4 conversion in the SMR reaction with increasing GHSV at 800 • C with a steam-to-carbon ratio (S/C) of 3. For comparison, a commercial catalyst was reacted under the same conditions. At the reaction temperature of 800 • C with S/C = 3, the calculated CH 4 conversions at the thermodynamic equilibrium is 100%. All catalysts except Ni/Mg9 reached the thermodynamic equilibrium at the GHSV of 8000 h −1 . CH 4 conversion of 100% was maintained up to the GHSV of 16,000 h −1 for the developed catalysts other than Ni/Mg9, whereas the commercial catalyst showed decreased conversion at the GHSV of 16,000 h −1 . The CH 4 conversions at the GHSV of 24,000 h −1 were in the following order: Ni/Mg25 > Ni/Mg30 > Ni/Mg19 > Ni/Mg9 > commercial catalyst. At the higher GHSV, the more likely SMR will be controlled by diffusion phenomena. Ni/Mg25 and Ni/Mg30, having greater pore sizes (Figure 2 and Table S1), show higher activity at 800 • C than Ni/Mg9 and Ni/Mg19 with narrower pores. The carbon balance of the reaction follows the trend of CH 4 conversion (Figure 6b). The H 2 /CO ratios of Ni/Mg19, Ni/Mg25 and Ni/Mg30 at the GHSV of 8000 h −1 were around 3.6, which were decreased at the GHSV of 16,000 h −1 but still higher than 3 (Figure 6c). The higher H 2 /CO ratio than 3, the theoretical H 2 /CO ratio in SMR, reflects that there was a further reaction of water gas shift in the presence of excess water (S/C = 3). The low CO selectivity at the low GHSV because of water gas shift is confirmed in Figure 6d. At the GHSV of 24,000 h −1 , H 2 /CO ratios of Ni/Mg19, Ni/Mg25 and Ni/Mg30 are around 3 with increased CO selectivity.  In order to understand the resistance of catalysts to coke formation, the SMR reaction was performed under accelerated deactivation conditions (S/C = 1, GHSV = 8000 h −1 , 700 °C). In contrast to the results of CH4 conversion at 800 °C with S/C = 3 (Figure 6), the conversion by Ni/Mg9 and Ni/Mg19 at 700 °C with S/C = 1 was slightly higher than that by Ni/Mg25 and Ni/Mg30 (Figure 7a,b). However, the carbon balance was highest for Ni/Mg30, which resulted in the lowest coke formation in that catalyst (Figure 7b,c). The amount of deposited coke by TGA analysis was in accordance with the carbon balance measured during the reaction (Figure 7b and Supplementary Figure S2), where the lower carbon balance reflects the more carbon consumption for the formation of coke. The H2/CO ratio was not higher than 3 under the SMR reaction with S/C = 1 where H2O was not enough for water gas shift (Figure 7d). It is notable that the trend of decreasing coke deposition with increasing Mg content was clearly observed (Figure 7b). The amount of coke formed per gram of catalysts per hour In order to understand the resistance of catalysts to coke formation, the SMR reaction was performed under accelerated deactivation conditions (S/C = 1, GHSV = 8000 h −1 , 700 • C). In contrast to the results of CH 4 conversion at 800 • C with S/C = 3 (Figure 6), the conversion by Ni/Mg9 and Ni/Mg19 at 700 • C with S/C = 1 was slightly higher than that by Ni/Mg25 and Ni/Mg30 (Figure 7a,b). However, the carbon balance was highest for Ni/Mg30, which resulted in the lowest coke formation in that catalyst (Figure 7b,c). The amount of deposited coke by TGA analysis was in accordance with the carbon balance measured during the reaction (Figure 7b and Supplementary Figure S2), where the lower carbon balance reflects the more carbon consumption for the formation of coke. The H 2 /CO ratio was not higher than 3 under the SMR reaction with S/C = 1 where H 2 O was not enough for water gas shift (Figure 7d). It is notable that the trend of decreasing coke deposition with increasing Mg content was clearly observed (Figure 7b). The amount of coke formed per gram of catalysts per hour was calculated to be 0.0050, 0.0046, 0.0040, 0.0034, and 0.0014 g coke /g cat ·h for commercial catalyst, Ni/Mg9, Ni/Mg19, Ni/Mg25, and Ni/Mg30, respectively. The effect of increasing basicity by introduction of Mg or MgO to Ni-alumina-based catalysts on coke formation in SMR and DMR reactions has been reported [20,24]. At the surface of the catalyst, increased basicity reduces coke formation by forming surface-activated O*, which can react with neighboring C* and result in formation of CO and CO 2 [19,40]. It is also widely accepted that smaller sized Ni particles with higher SMSI and higher dispersion of active sites shows higher coke resistance, and the developed catalysts with lower Mg show smaller Ni particles (Table 1 and Figure 4) [9,22]. Our results show that catalysts with higher Mg content and basicity show a more dominant effect than the effect of Ni particle size on suppressing coke formation. The CO2-TPD profiles in Figure 8 reveal the basicity of the catalysts. For all of the samples, two peaks of CO2 desorption are shown at about 120 °C (low temperature, peak I) and 400 °C (high temperature, peak II). The amount of CO2 desorbed is indicative of the number of basic sites on the catalyst [41]. The amount of CO2 desorption calculated based on the total area of the CO2-TPD (peak I + peak II) increased with increasing Mg content, as shown in Table 2. Peak I at low temperature originated from CO2 desorption from the weak basic sites, and peak II, from the medium or strong basic sites [42]. The trend was very clear: the greater the Mg content, the greater the amount of CO2 desorption from the medium/strong basic sites (Table 2). Furthermore, the ratio of peak II to peak I for Ni/Mg30 was more than double those of the others (Table 2). This confirms that the presence of The CO 2 -TPD profiles in Figure 8 reveal the basicity of the catalysts. For all of the samples, two peaks of CO 2 desorption are shown at about 120 • C (low temperature, peak I) and 400 • C (high temperature, peak II). The amount of CO 2 desorbed is indicative of the number of basic sites on the catalyst [41]. The amount of CO 2 desorption calculated based on the total area of the CO 2 -TPD (peak I + peak II) increased with increasing Mg content, as shown in Table 2. Peak I at low temperature originated from CO 2 desorption from the weak basic sites, and peak II, from the medium or strong basic sites [42]. The trend was very clear: the greater the Mg content, the greater the amount of CO 2 desorption from the medium/strong basic sites (Table 2). Furthermore, the ratio of peak II to peak I for Ni/Mg30 was more than double those of the others (Table 2). This confirms that the presence of Mg increased the basicity of the catalyst and increased the resistance to coke formation, which result is in good agreement with the amount of coke deposition as analyzed by TGA (Figure 7b and Supplementary Figure S2). Although Ni/Mg9 showed the highest surface area and smallest Ni particle size, which is typically advantageous for an Ni catalyst, Ni/Mg30 showed the highest coke resistance.

Preparation of Ni/Mg-Aluminate Catalysts
The EISA method was used to synthesize the catalyst supports [34]. The EISA method using selfassembled amphiphilic molecules as templates is a simple means of synthesizing metal oxides of mesoporous structure [19,36,37]. First, precursor was prepared by adding 6

Preparation of Ni/Mg-Aluminate Catalysts
The EISA method was used to synthesize the catalyst supports [34]. The EISA method using self-assembled amphiphilic molecules as templates is a simple means of synthesizing metal oxides of mesoporous structure [19,36,37]. First, precursor was prepared by adding 6.57 g of P123 to a mixture of 135.0 mL of ethanol and 11.0 mL of nitric acid under vigorous stirring. Next, Al(OiPr) 3 and MgCl 2 were added to the P123-containing solution, which was stirred for a further 5 h. The mixing ratio of all of the components was fixed as (Al(OiPr) 3 + MgCl 2 ):P123:EtOH:HNO 3 = 1:0.017:34.100:2.090.
The molar ratio of MgCl 2 to Al(OiPr) 3 for Mg9, Mg19, Mg25, and Mg30 was 0.10, 0.23, 0.32, and 0.43, respectively. The precursor solution was oven-dried at 60 • C for 48 h. Then, calcination was carried out at 900 • C for 5 h in air to obtain the Mg-aluminate support in the form of white powder. Ni was loaded on the Mg-aluminate support by the incipient wetness impregnation method. The weight percent of Ni metal in the catalyst was fixed to be 15 wt%. The Ni-loaded Mg-aluminate was dried at 60 • C for 4 h, followed by calcination at 900 • C for 2 h. The obtained powder was pelletized under 200 bar, then ground and sieved to separate particles of 80-100 mesh size for the catalytic reaction. The sample nomenclature is as follows: 'Mg9' for support, 'Mg9-Ni' for Ni-impregnated and calcined sample, and 'Ni/Mg9' for reduced sample (also: Mg19, 25, 30).

Catalyst Characterization
X-ray diffraction (XRD) analysis was performed with an AXS D8 diffractometer (Bruker, Billerica, MA, USA) at Cu Kα wavelength, 40 kV, and 40 mA. The crystallite sizes of metallic Ni were calculated using the Scherrer equation based on the peak at 2θ = 51.85 • , (200) plane. The main peak of metallic Ni at 2θ = 44.51 • , (111) plane, was not used for the calculation, since it is partially overlapped with the peak at 2θ = 44.83 • , (400) plane, of MgAl 2 O 4 phase. Field emission transmission electron microscopy (FE-TEM) analysis was performed using the JEM-2100F (JEOL, Tokyo, Japan) at 200 kV at the National Nanofab Center (NNFC). The nickel particle size distributions were estimated from randomly selected 100 particles in each TEM image by using the ImageJ software, version 1.8.0. Brunauer-Emmett-Teller (BET) analysis was carried out using BELSORP-mini (MicrotracBEL, Osaka, Japan). The samples were subjected to nitrogen adsorption and desorption at −196 • C after pretreatment in a vacuum at 300 • C for 3 h. The pore size distribution was analyzed according to the Barrett-Joymer-Halenda (BJH) theory. A CO 2 -temperature-programed desorption (CO 2 -TPD) analysis was carried out using the AutoChem 2920 (Micromeritics, Nocross, GA, USA) under an He atmosphere at 300 • C for 2 h and CO 2 adsorption at 40 • C for 1 h, followed by heating to 800 • C at a ramping rate of 5 • C/min. The reduction characteristics of the catalysts were analyzed by hydrogen temperature programed reduction (H 2 -TPR) using the BET-CAT (MicrotracBEL, Osaka, Japan). In the H 2 -TPR measurement, about 150 mg of samples were loaded and heated (5 • C/min) in a gas flow (30 mL/min) containing a mixture of H 2 :Ar (20:80). Prior to the H 2 -TPR experiment, the samples were pretreated under an inert atmosphere (Ar) at 200 • C for 2 h and then cooled to room temperature. Thermogravimetric analysis with differential scanning calorimetry (TGA-DSC) analysis was performed using the Labsys TGA EVO (Setaram Instrumentation, Lyon, France) at temperatures up to 1400 • C in air. Inductive coupling plasma atomic emission spectroscopy (ICP-AES) analysis was performed to determine the chemical composition of the synthesized catalysts by using an Avio500 (Perkin-Elmer, Norwalk, MA, USA) after the pretreatment of dissolving catalysts in acidic digestion.

Catalyst Activity Test of SMR
For the catalytic reaction, the synthesized catalysts were ground and sieved using mesh (150-180 µm). As a reference catalyst, the Ni-based commercial steam reforming catalyst (Tablet type) was crushed, ground, and sieved with the same mesh mentioned above. The catalytic reaction was performed in a microreactor (a quartz tube of 1/4 inch inner diameter). The temperature was controlled by a digital program controller (KP1000, CHINO, Japan), while the influx of reactants was controlled by a mass flow controller (5850E, Brooks ® , Halfield, PA, USA). The temperature difference between inside of the catalyst bed and outside of the reactor was measured, as shown in Supplementary Table S3. For the steam, distilled water heated to 180 • C was supplied by micro pump (NP-KX-200, DONGSUNG Science, Seoul, Korea). A K-type thermocouple was inserted into the reaction tube to measure the temperature of the catalyst bed. In the quartz reactor, 85 mg of catalyst was positioned between quartz wool. Reduction of the catalyst was performed under 50 sccm of 20 mol% H 2 /N 2 mixed gas at 900 and 550 • C for the developed catalysts and the commercial catalyst, respectively (by following the instructions for optimum reduction conditions). The reactions were performed under gas velocities (Gas Hourly Space Velocity, GHSV = h −1 ) of 8000, 16,000, and 24,000 h −1 at 800 • C at the steam to carbon (S/C) ratio of 3. In addition, the harsh condition reactions were performed at the S/C ratio of 1 under a space velocity GHSV of 8000 h −1 at 700 • C. He was used as the carrier gas, and the outlet gas (CH 4 , CO, and CO 2 ) was analyzed using the TCD column of the HP 6890 GC system (Agilent Technologies Inc., Santa Clara, CA, USA) every 22 min. after stabilization of the catalytic reaction for 1 h. The performance of each catalyst was determined by CH 4 conversion in consideration of a carbon balance of over 0.9.

Conclusions
The mesoporous Ni/Mg-aluminate spinel catalysts were developed by synthesizing mesoporous Mg-aluminate supports with various Mg contents using the EISA method, followed by loading Ni active sites by the impregnation method. The Mg content affects the pore structures and the Ni particle sizes of the catalysts by changing the sintering resistance and the extent of the exsolution. The higher specific surface area with smaller pore size and smaller Ni particle size, which could be advantageous for the coke resistance, were obtained with the lower Mg contents. Nevertheless, the coke formation was mainly influenced by the basicity of the catalyst, and therefore the catalysts with the higher Mg contents showed stronger resistance against coke formation.  Table S1: Pore structure analysis of mesoporous Mg-aluminate supports, after Ni loading with calcination and after reduction, Table S2: JCPDS card information of main peaks, Table S3: The temperature difference between inside of the catalyst bed and outside of the reactor during the reaction.