<?xml-model href='http://www.tei-c.org/release/xml/tei/custom/schema/relaxng/tei_all.rng' schematypens='http://relaxng.org/ns/structure/1.0'?><TEI xmlns="http://www.tei-c.org/ns/1.0">
	<teiHeader>
		<fileDesc>
			<titleStmt><title level='a'>Achieving Product Control in Furfural Hydrogenation Using Intermetallic Catalysts</title></titleStmt>
			<publicationStmt>
				<publisher>American Chemical Society</publisher>
				<date>11/15/2024</date>
			</publicationStmt>
			<sourceDesc>
				<bibl> 
					<idno type="par_id">10640499</idno>
					<idno type="doi">10.1021/acscatal.4c04091</idno>
					<title level='j'>ACS Catalysis</title>
<idno>2155-5435</idno>
<biblScope unit="volume">14</biblScope>
<biblScope unit="issue">22</biblScope>					

					<author>Charles J Ward</author><author>Minda Chen</author><author>Andrew Lamkins</author><author>Claudio Ordonez</author><author>Rong Sun</author><author>Puranjan Chatterjee</author><author>Minghui Niu</author><author>Ruoyu Cui</author><author>Da-Jiang Liu</author><author>Wenyu Huang</author>
				</bibl>
			</sourceDesc>
		</fileDesc>
		<profileDesc>
			<abstract><ab><![CDATA[Not Available]]></ab></abstract>
		</profileDesc>
	</teiHeader>
	<text><body xmlns="http://www.tei-c.org/ns/1.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xlink="http://www.w3.org/1999/xlink">
<div xmlns="http://www.tei-c.org/ns/1.0"><p>ABSTRACT: Intermetallic nanoparticles (iNPs) have garnered much attention as effective catalysts, but the impact of tuning surface properties to induce steric effects is relatively unexplored. Here, we report on the strategy of governing steric hindrance in bimetallic catalysts as a method to alter product selectivity in furfural hydrogenation using Rh-based iNPs by varying the size of the secondary metal atoms. RhGa, RhIn, and RhBi nanoparticles were synthesized within confined mesoporous silica wells (MSWs) and assessed for the vapor-phase hydrogenation of furfural. RhGa and RhIn iNPs enable product control with an enhanced selectivity to furan and furfuryl alcohol (&gt; 90 %) compared to the monometallic Rh@MSW. Adding Bi to Rh inhibits the transformation of furfural almost entirely. In-situ diffuse reflectance infrared Fourier transform spectroscopy studies and density functional theory-based machine-learning accelerated molecular dynamics simulations reveal that the secondary metal's identity strongly impacts the preferred furfural adsorption mode on the active sites, leading to the observed catalysis control. The mesoporous silica shell of the RhM@MSW catalyst provides protection against NP aggregation under reaction and regeneration conditions, as supported by good stability during recycling studies. </p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="1.">Introduction</head><p>Intermetallic compounds (IMCs) are a type of ordered alloy with defined stoichiometry and crystal structure. Intermetallic nanoparticles (iNPs) have been recognized as effective catalysts with excellent activity, selectivity, and durability in various reactions. <ref type="bibr">1,</ref><ref type="bibr">2</ref> The addition of a secondary metal enables control over the catalyst's electronic and atomic level structures, leading to electronic, geometric, steric, and ordering effects that greatly influence their catalytic performance. <ref type="bibr">3</ref> The impact of modifying the electronic and geometric properties of active metals by forming intermetallic compounds has been widely explored. <ref type="bibr">[4]</ref><ref type="bibr">[5]</ref><ref type="bibr">[6]</ref><ref type="bibr">[7]</ref> In contrast, utilizing the welldefined iNP surface to create sterically hindered active sites remains under-explored.</p><p>The limited examples of designing an iNP surface to restrict access to the active sites indicate that it is a viable strategy for controlling selectivity. Komatsu et al. described a "molecular recognition" capability in the hydrogenation of nitroarenes using RhM/SiO2, where they observed a preferential nitro group adsorption on the more sterically hindered Rh. <ref type="bibr">8</ref> Komatsu further explored this behavior in the hydrogenation of dienes to monoenes. It was shown that for intermetallic compositions with larger secondary metals, such as Sn or Bi, adsorption of the inner C=C bond was inhibited, leading to the selective hydrogenation of the terminal C=C bond and preventing overhydrogenation to alkanes. <ref type="bibr">9</ref> This surface-tuning strategy could be especially effective in more complex reactions that include multiple reaction pathways and possible products such as the hydrogenation of furfural.</p><p>With the growing economic and environmental demand to replace nonrenewable feedstocks in the chemical industry, the conversion of biomass to generate valuable fine chemicals is receiving increased attention. <ref type="bibr">[10]</ref><ref type="bibr">[11]</ref><ref type="bibr">[12]</ref> The conversion of furfural, a biomass-derived platform chemical, has received much interest in this regard as it is used to produce numerous downstream chemicals, such as furan, furfuryl alcohol (FOL), tetrahydrofuran (THF), and 2-methylfuran (2-MF) which are then used in various products including resins, adhesives, solvents, and biofuels. <ref type="bibr">13- 15</ref> Furfural is primarily obtained from the hydrolysis and dehydration of agricultural waste-based sugars making it a sustainable platform compound without relying on food or fuel sources.</p><p>Transition metal catalysts, including Cu, Ni, Co, Pt, Pd, and Rh, have been explored for the conversion of furfural. <ref type="bibr">[16]</ref><ref type="bibr">[17]</ref><ref type="bibr">[18]</ref><ref type="bibr">[19]</ref><ref type="bibr">[20]</ref><ref type="bibr">[21]</ref><ref type="bibr">[22]</ref><ref type="bibr">[23]</ref><ref type="bibr">[24]</ref><ref type="bibr">[25]</ref><ref type="bibr">[26]</ref> Many of the reported non-noble metal catalysts suffer from coking, require harsh conditions to achieve high conversions, and deactivate rapidly at high temperatures due to sintering. <ref type="bibr">[27]</ref><ref type="bibr">[28]</ref><ref type="bibr">[29]</ref> Noble metal-based catalysts, with enhanced hydrogenation abilities, have superior activity compared to non-noble metals but typically suffer from poor selectivity due to the reaction complexity and number of possible pathways, as displayed in Scheme 1. Previous studies have explored manipulating this selectivity and found it to be sensitive to many factors, including catalyst composition, catalyst support, active site size and morphology, and reaction conditions. Somorjai et al. demonstrated changing selectivity towards furfuryl alcohol and furan dependent on the size and shape of Pt NPs. <ref type="bibr">30</ref> The effect of Pt NP size was further explored by Kyriakou and coworkers, who also investigated the impact of solvent choice on reaction activity and selectivity towards FOL. <ref type="bibr">31</ref> Chen et al. reported tuning atomic ratios in Cu-Fe catalysts and reaction temperatures to produce FOL or 2-MF selectively. <ref type="bibr">32</ref> While some tunability was achieved, overall selectivity remained low (&lt;50% 2-MF). MIL-101 was used as an acidic support to increase the selectivity towards cyclopentanone using Ru-based catalysts in the transformation of furfural by Li and co-workers. <ref type="bibr">33</ref> Even with the aforementioned examples, designing catalytic systems capable of effectively targeting multiple major products using a broadly applicable strategy remains challenging.</p><p>Achieving product control with a high degree of selectivity is highly desirable. In our previous work, PtSn, PtZn, and PtPb intermetallic catalysts exhibited high selectivity (&gt;97% for PtSn) to furfuryl alcohol (FOL) and significantly improved activity compared to monometallic Pt. <ref type="bibr">34</ref> This enhanced activity and selectivity towards FOL was attributed to furfural having a different adsorption configuration on the PtSn and Pt catalyst as suggested by DFT calculations. Similar results were observed using a Ni-Sn IMC system by Wei and co-workers, where increasing the secondary metal content switched the selectivity from tetrahydrofurfuryl alcohol to furfuryl alcohol with the Sn isolating the Ni sites and limiting the furan ring interaction with the active sites. <ref type="bibr">35</ref> This apparent correlation between adsorption geometry and reaction selectivity makes furfural hydrogenation an ideal reaction to display the effectiveness of IMC-induced steric effects.</p><p>This study aims to further explore and apply the characteristics of iNP catalysts, with a focus on steric effects, for the hydrogenation of furfural with the goal of achieving tunable product control while establishing a blueprint for future catalyst design. We detail an effective method of directing the reaction pathway of furfural hydrogenation with high selectivity by changing the inert metal component of RhM (M =Ga, In, Bi) iNPs and introducing surface steric effects. Vapor-phase hydrogenation of furfural resulted in highly selective production of furan over RhGa@MSWs, FOL for RhIn@MSWs, and a sharply diminished activity for RhBi@MSW. In-situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) experiments and DFT-based machine learning (ML) accelerated molecular dynamics (MD) simulations were used to link the observed product control to the furfural adsorption configuration. Our results display the ability to alter reaction pathways by controlling the identity of the secondary metal in iNP catalysts. The nanoparticles are encapsulated in well-defined mesoporous silica wells (MSWs) to prevent sintering. <ref type="bibr">[36]</ref><ref type="bibr">[37]</ref><ref type="bibr">[38]</ref> Catalyst regeneration and recycling tests demonstrated the stability of the RhM@MSW due to the unique synthesis strategy.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.">Results and Discussion</head></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.1">Synthesis and Characterization</head><p>Rh-based iNPs are encapsulated and confined between a solid silica core and a mesoporous silica shell with aligned channels, mimicking mesoporous wells (MSWs). The MSWs act as isolated nano-reactors and inhibit particle sintering during high-temperature treatments and catalytic reactions. The confined Rh nanoparticles remain accessible to metal precursors and small organic molecules via the channels in MSW. <ref type="bibr">18</ref> This allows the addition of a secondary metal into the Rh@MSW to form the corresponding intermetallic compound nanoparticles (iNPs) and the furfural to reach the catalyst surface and interact with the active sites. An illustration of the synthesis process is summarized in Scheme 1, with transmission electron microscopy (TEM) images of the Rh NPs and the monodisperse spherical silica cores provided in Figures <ref type="figure">S1</ref> and <ref type="figure">S2</ref>.</p><p>Scheme 1. Schematic synthesis procedure for RhM@MSWs showing the Rh decoration of the functionalized silica spheres followed by the mesoporous silica shell growth and the introduction and reduction of the secondary metals to form the intermetallic nanoparticles confined within the mesoporous silica wells.</p><p>The secondary metals (Ga, In, and Bi) were deliberately chosen to study the steric effect on the catalytic property of Rh. Based on prior reports, Bi (1.8 &#197;) was selected to be the largest of the secondary metals.  TEM studies were conducted to characterize the catalyst structure and the iNPs further. A few things can be learned from these TEM characterizations. First, the high-angle annular darkfield scanning transmission electron microscopy (HAADF-STEM) images confirm the morphology of the monometallic Rh@MSWs (Figure <ref type="figure">S3</ref>) and the RhM@MSWs iNPs (Figure <ref type="figure">2a</ref>, 2c, and 2e). The mesoporous silica layer can be resolved and clearly distinguished from the solid silica core. Rh and RhM nanoparticles are well-dispersed between the mesoporous silica layer and the silica core, with an average particle of 3.7 &#177; 0.6 nm for Rh, 8.2 &#177; 1.3 nm for RhGa, 7.0 &#177; 1.3 nm for RhIn, and 9.3 &#177; 1.2 nm for RhBi. The average particle sizes are calculated by counting &gt;200 particles (Figure <ref type="figure">S4</ref>-S7) and match those values obtained from PXRD peak width (Figure <ref type="figure">1</ref>). In addition, the energy dispersive x-ray spectroscopy (EDS) maps (Figures <ref type="figure">2a</ref>, <ref type="figure">2c</ref>, and <ref type="figure">2e</ref>) show the atomic distribution of the RhM@MSWs iNPs, confirming that the Rh and the secondary metal are distributed homogeneously with no monometallic particles or phase segregation observed, indicating all Rh NPs are converted to RhM iNPs. Lastly, atomic resolution HAADF-STEM images and EDS mapping images confirm the crystal structure of the iNPs, as shown in Figure <ref type="figure">2b</ref> for RhIn, Figure <ref type="figure">2d</ref> for RhGa, and Figure <ref type="figure">2f</ref>   N2 physisorption measurements were performed for MSW, Rh@MSW, and RhM@MSW (Figure <ref type="figure">S12</ref> and Table <ref type="table">S1</ref>). All the materials show a Type IV isotherm, indicative of mesopores being present, as expected. The Brunauer-Emmett-Teller (BET) surface area of the starting, undecorated MSW is 439 m 2 /g. The monometallic Rh@MSW has a surface area of 430 m 2 /g, and the RhM@MSW BET surface areas range from 310-409 m 2 /g. The trend in BET surface area is inversely related to the trend in nanoparticle size, indicating that the larger nanoparticles could be partially filling or blocking the mesopores. The Rh loading ranges from ~1.3-1.5 wt% according to the inductively coupled plasma mass spectrometry (ICP-MS) results in Table <ref type="table">S2</ref>. The target stoichiometry for Rh: M2 was 1:1 for all cases except for RhGa@MSW, which required excess Ga to form pure-phase RhGa iNPs.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.2">Catalysis: Vapor-Phase Furfural Hydrogenation</head><p>The synthesized intermetallic catalysts were assessed for the vapor-phase hydrogenation of furfural using a plug flow reactor. As shown in Scheme 2, the furfural hydrogenation can take various pathways leading to many possible products. The complex reaction pathways can easily lead to a mixture of products, making achieving high selectivity challenging. Three possible reaction pathways for the hydrogenation of furfural are displayed in Scheme 2. Pathway I consists of the hydrogenation of the C=C bonds of the furan ring followed by the hydrogenation of the C=O carbonyl group to give tetrahydrofurfural and tetrahydrofurfuryl alcohol, respectively. Pathway II forms furfuryl alcohol (FOL) as an intermediate from the selective hydrogenation of the carbonyl group. Hydrogenolysis of C-O in FOL results in 2-methylfuran (2-MF) production. Pathway III starts with a decarbonylation step to form furan. Further hydrogenation and hydrogenolysis of the furan lead to butane as the primary product.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head>Scheme 2. Reaction pathways of furfural hydrogenation and decarbonylation.</head><p>A summary of product distribution is shown in Figure <ref type="figure">3</ref>, with the conversion held at ~20% for all reactions except RhBi, which could only achieve ~2% conversion. The reported selectivity values are averaged from 3 trials with the standard deviation displayed in Figure <ref type="figure">S13</ref>. The turnover frequency (TOF) was calculated based on the amount of converted FAL and Rh dispersion obtained from CO chemisorption measurements of each catalyst (Figure <ref type="figure">3</ref>). The calculated carbon balance (Table <ref type="table">S3</ref>) reveals that approximately 100% of the input carbon is accounted for in each catalytic reaction. Monometallic Rh@MSWs yields various primary and secondary reaction products at 160 &#176;C, with a majority (77%) being furan. FOL (1.9%) and 2-MF (6.2 %) are observed, along with a host of other products, including tetrahydrofuran (THF), butane, and tetrahydrofurfuryl alcohol (THFOL). The number and variety of products highlight the need for an improved catalyst design to enhance selectivity.</p><p>Adding a secondary metal to form iNPs resulted in different major products, as confirmed by GC-MS, being formed at the steady state for each metal combination. The intermetallic catalysts showed enhanced selectivity over the monometallic Rh@MSW. Adding Ga to form RhGa@MSW promoted furan selectivity to 89% by limiting further hydrogenation and hydrogenolysis to THF and butane and eliminating THFOL production. A decreased TOF is also noted, dropping from 81.5 h -1 for Rh@MSW to 24.4 h -1 for RhGa@MSW at 160 &#176;C. For reference, Pd, a more typical decarbonylation catalyst, has achieved a TOF of 265.8 s -1 for the conversion of furfural to furan at 230 &#176;C, 1 atm with an H2:furfural feed ratio of 25. <ref type="bibr">24</ref> RhIn@MSW-catalyzed furfural hydrogenation produced FOL with a selectivity of 97% and a TOF of 322.0 h -1 . The enhanced activity of RhIn@MSW compared to Rh@MSW could be attributed to the changed adsorption configuration and strength of furfural and furfuryl alcohol on the intermetallic surface, similar to the observation on intermetallic PtSn. <ref type="bibr">34</ref> Alloying Rh and Bi, on the other hand, resulted in a significantly lowered activity but with an enhanced selectivity to FOL (93%). The TOF for the RhBi could not be accurately calculated due to the weak CO adsorption, leading to inaccurate chemisorption results and Rh dispersion value. The catalyst-dependent selectivity was studied at high conversion and across different reaction temperatures (Figure <ref type="figure">S14</ref>). All the catalysts maintained the expected selectivity at higher conversions and only RhBi saw a decrease in selectivity as more furan was produced with increasing temperature. Additional experiments using the major product as the substrate for each catalyst (Table <ref type="table">S4</ref>) showed that Rh@MSW is prone to overhydrogenation, explaining the lower selectivity for the monometallic catalyst. In contrast, the lack of significant conversion when using the RhM@MSWs further validates that the observed selectivity is intrinsic to the catalyst and not a function of specific reaction conditions. Metal-free MSWs and the monometallic secondary metals show no activity. Rh and RhM catalysts were also synthesized without a mesoporous silica shell (RhM/NH2-SiO2) and tested for the hydrogenation of furfural under the same reaction conditions used for the RhM@MSWs. The catalyst without a shell resulted in increased furfural conversion, exemplified by the RhIn/NH2-SiO2 TOF of 624.9 h -1 compared to the RhIn@MSW TOF of 322.0 h -1 . This increased activity indicates that the mesoporous shell blocks furfural from accessing some of the active sites. The product distribution shown in Figure <ref type="figure">S15</ref> is nearly identical to the results reported for the RhM@MSWs (Figure <ref type="figure">3</ref>) in all cases. The only exception is the Rh/NH2-SiO2, which produced more THF and butanes without the shell, but the selectivity to furan is similar. We attribute the changing chemoselectivity primarily to intermetallic-induced steric effects governed by the radius of the secondary metal. Electronic effects were disregarded as a significant factor because the trend in secondary metal electronegativity, and the resulting electron density of Rh, do not correlate well to the trend in catalytic behavior. These results demonstrate that the identity of the iNPs can be tuned to induce steric effects and control reaction selectivity.  <ref type="table">S3</ref>). The catalysts were dispersed in 500 mg of quartz sand and packed into a U-shaped quartz reactor. Furfural = 0.3 &#181;L/min, H2 = 11.4 mL/min, He = 8.6 mL/min, 160 &#176;C. *RhBi is only able to achieve approximately 2% conversion. The conversion-based TOF of each catalyst is designated by the star ( ) symbol.</p><p>MSWs are designed to prevent particles from sintering at high temperatures and increase the catalyst stability and recyclability. For furfural hydrogenation, it is especially critical for the catalyst to be stable at high temperatures, as carbon deposition from furfural can lead to the catalyst deactivation. <ref type="bibr">23,</ref><ref type="bibr">[39]</ref><ref type="bibr">[40]</ref><ref type="bibr">[41]</ref> A high-temperature oxidation and reduction treatment is necessary to regenerate the catalytic surface after this carbon deposition. <ref type="bibr">33,</ref><ref type="bibr">42</ref> Therefore, to demonstrate the effectiveness of the mesoporous silica shell as a protective measure, the RhIn@MSW recyclability was explored and compared to a control catalyst without the porous silica shell, RhIn/NH2-SiO2.</p><p>The catalysts were exposed to reaction conditions with the conversion tuned to approximately 50% by adjusting the amount of catalyst used. The catalysts were intentionally deactivated after 2 h by doubling the FAL flow (He= 17.2 mL/min) and adjusting the H2 flow rate to 2.8 mL/min to maintain 20 mL/min total flow. The corresponding conversion and selectivity to furfuryl alcohol are displayed in Figure <ref type="figure">4</ref>. Three deactivation-regeneration cycles were completed, with the RhIn@MSWs maintaining approximately 50% conversion and 95% FOL selectivity across all runs. Conversely, the RhIn/NH2-SiO2 experienced on-stream deactivation and was unable to recover the initial activity after the regeneration steps, dropping from 56 % conversion in the first cycle to 44 % in the third cycle. Additionally, the selectivity towards FOL decreased with each cycle, indicating the deactivation-regeneration cycle impacted the structural properties of the RhIn NPs. PXRD patterns of the RhIn@MSWs (Figure <ref type="figure">S16</ref>) catalyst look nearly identical before and after the regeneration cycles with a calculated particle size of 6.4 and 7.2 nm, respectively. Similarly, the collected TEM images (Figure <ref type="figure">S17</ref>) show very little change in the catalyst, with an average measured particle size of 7.0 &#177; 1.3 nm for the as-synthesized catalyst and 7.2 &#177; 1.4 nm after the recycling test (Figure <ref type="figure">S18</ref>). The same characterization was done for the RhIn/NH2-SiO2 with the PXRD (Figure <ref type="figure">S19</ref>) showing an increase in particle size from 11.4 to 13.0 nm. The average measured particle size from TEM (Figures <ref type="figure">S20-22</ref>) increases from 12.8 &#177; 4.1 to 15.4 &#177; 6.9 nm. It should be noted that the initial RhIn NPs without the protective shell are larger and have a broader size distribution than the fresh RhIn@MSW NPs, pointing to significant aggregation during the initial synthesis. The catalytic stability and material characterization demonstrate the effectiveness of the mesoporous silica wells at preventing NPs from aggregating during furfural hydrogenation and the corresponding catalyst regeneration steps.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.3">DRIFTS Study</head><p>Diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) experiments were performed to study the surface of the catalysts using CO as a probe molecule (Figure <ref type="figure">5</ref>). For monometallic Rh@MSW, a strong peak at 2068 cm -1 is observed, which is assigned to the atop adsorption of CO on Rh. The presence of bridged and geminal CO on Rh is detected as well. The bridged CO appears as a broad peak at 1917 cm -1 , with the geminal CO symmetric and asymmetric stretching at 2095 and 2024 cm -1 . Rh-based catalysts are rarely used in furfural hydrogenation but have been reported for the production of THFOL, FOL, and 1,5-pentanediol. <ref type="bibr">26,</ref><ref type="bibr">[43]</ref><ref type="bibr">[44]</ref><ref type="bibr">[45]</ref><ref type="bibr">[46]</ref> The formation of furan has only been reported using Rh(I) catalysts, whereas Rh(0) was inactive. <ref type="bibr">47</ref> The geminal CO peaks indicate the presence of positively charged Rh species that could be the cause of the Rh selectivity to furan. <ref type="bibr">[48]</ref><ref type="bibr">[49]</ref><ref type="bibr">[50]</ref><ref type="bibr">[51]</ref> RhGa@MSW and RhIn@MSW show only the atop CO peak at 2036 and 2024 cm -1 , respectively. Performing the same CO DRIFTS experiment on RhBi@MSW results in a very weak linear CO peak at 2083 cm -1 . The lack of pronounced peaks indicates that CO interacts very weakly with RhBi iNPs, with only a small amount being adsorbed. The higher electronegativity of Rh compared to Ga and In results in more electron density on the alloyed Rh than the monometallic Rh and an enhanced electron back donation to the antibonding orbital of CO, lowering the frequency of the C&#8801;O vibration. <ref type="bibr">52</ref> Bi, on the other hand, is more electronegative than Rh, causing a more electropositive Rh upon alloying and a shift in the C&#8801;O vibration to a higher wavenumber. The CO DRIFTS results highlight the difference between the monometallic and intermetallic surfaces, with CO showing multiple adsorption modes on pristine rhodium, only one type of adsorption on RhIn and RhGa, and a lack of adsorption in the RhBi case. The lack of the bridging CO peak (1917 cm -1 ) indicates that the insertion of a secondary metal removed the typical bridge site on the Rh surface. The interactions of furfural with Rh, RhBi, RhGa, and RhIn under reaction conditions were investigated using FAL-DRIFTS (Figure <ref type="figure">6</ref>) to provide insight into the relationship between the observed product selectivity and adsorption configuration. We first measured bare MSWs without Rh as a reference and observed no furfural adsorption on the SiO2 support at 160 &#176;C. Therefore, any peaks seen in the DRIFTS spectra of Rh@MSW and RhM@MSW are due to the adsorption of furfural or a reaction intermediate on the metal surface. For monometallic Rh@MSW, the characteristic carbonyl (1686 cm -1 ) and furan ring (1581 and 1474 cm -1 ) peaks are not observed. <ref type="bibr">53</ref> Instead, two peaks at 2005 and 1851 cm -1 attributed to linear and bridging CO formed from the decarbonylation of furfural are observed. According to the metal-surface selection rule, the absence of carbonyl and furan ring peaks indicates the furfural adsorbs parallel to the Rh surface, as shown in Figure <ref type="figure">8a</ref>. <ref type="bibr">54</ref> For RhGa@MSW, a strong peak at 1628 cm -1 ascribed to the linear adsorption of the carbonyl group is observed, indicating that the aldehyde is bonded in the &#951;1(C) configuration with the C=O bond perpendicular to the surface. Weak furan ring peaks at 1566 and 1457 cm -1 suggest the furfural adsorbs in a tilted geometry that is nearly parallel to the RhGa surface (Figure <ref type="figure">8a</ref>). Similar to the CO DRIFTS, these peaks are red-shifted due to the electron transfer from Ga to Rh. The high C=O/C=C peak intensity ratio (13.8) signifies a strong interaction between the furan ring and the metal surface, meaning the furfural is only slightly tilted away from the surface. <ref type="bibr">55</ref> The RhIn results are very much the opposite. The carbonyl peak at 1630 cm -1 is almost negligible, while the furan ring peaks are evident. This DRIFTS result indicates that the carbonyl is nearly parallel to the RhIn surface in the &#951;2(C,O) configuration, with the double bonds of the ring tilted away from the surface (Figure <ref type="figure">8a</ref> ). The C=O/C=C peak ratio (0.2) is much lower than the RhGa results, indicating that the ring is further tilted away from the surface and interacts less with the catalyst. Some linearly adsorbed CO is also observed for the RhGa (2004 cm -1 ) and RhIn (1988 cm -1 ). The production of CO on RhGa aligns well with the observed selectivity to the decarbonylation pathway but is somewhat unexpected for RhIn. Although some decarbonylation (2.1%) is observed on the RhIn, we propose the presence of an obvious CO peak during FAL-DRIFTS is primarily due to the strong adsorption of CO on RhIn resulting in CO accumulation.</p><p>Additionally, the FAL-DRIFTS was done without the presence of hydrogen, meaning there is no hydrogenation pathway to compete with the decarbonylation pathway. The furfural DRIFTS of the largely inactive RhBi is similar to the RhIn case but with significantly weakened furan ring peaks and no observable free CO. This is due to the lack of furfural adsorption on RhBi (Figure <ref type="figure">8a</ref>). The DRIFTS results correlate well with the observed catalytic results. The multiple CO adsorption modes and the parallel furfural adsorption on monometallic Rh explain the lower selectivity, as numerous products and pathways are possible based on the many plausible interactions with the active sites. For RhGa, very weak furan ring peaks indicate that the C=C bonds are slightly tilted further away from the active surface than in the Rh case. This adsorption geometry leads to fewer possible reaction pathways within the furan ring and, thus, higher furan selectivity. It has also been reported that a higher furan selectivity is achievable through the &#951;1(C) configuration rather than the &#951;2(C,O) configuration observed on Rh. <ref type="bibr">60</ref> The presence of only linear CO on RhIn shows effective active site isolation, preventing multiple reaction pathways from taking place at once. The furfural interacts with the RhIn in a tilted &#951;2(C,O) configuration, leading to the selective hydrogenation of the carbonyl rather than the double bonds of the furan ring. When testing RhBi, the furfural DRIFTS spectra show very weak peaks compared to the other iNPs. This indicates furfural is not efficiently adsorbed on the RhBi catalyst, correlating well to the observed lack of activity.</p><p>We attribute the observed behavior majorly to the surface steric effects of the iNPs. As the secondary metals' atomic radius increases, access to the Rh active sites becomes more restricted. This increased steric hindrance prevents the bulkier furan ring from lying flat on the surface, leading to the tilted configuration and a change in the reaction pathway, as is the case for RhIn.</p><p>Ga, with an atomic radius closer to Rh, does not hinder access to the active site as much as In and thus follows the same reaction pathway as monometallic Rh. The increased selectivity towards furan is most likely due to the differences in how the carbonyl interacts with the Rh and the RhGa NPs. Alloying Rh with Bi introduces an even more significant atomic radius differential than the RhIn case, almost completely blocking the Rh sites, preventing furfural adsorption and conversion. Electronic effects were dismissed as the cause for product control as the trend in secondary metal electronegativity does not correlate well with the selectivity data.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="2.4">DFT-based ML Accelerated Molecular Dynamics Studies</head><p>To reveal the surface structure of the iNPs, we performed molecular dynamics (MD) simulations using machine learning (ML) potentials derived from density-functional theory (DFT) calculations. To avoid bias towards a particular surface orientation, we simulated a randomly aggregated NP by sequentially adding Rh and Ga/In/Bi atoms with a fixed stoichiometry of 1:1, subject to a minimum distance requirement, and then equilibrated at 300 K for 10 ps after each addition. The resulting NP was then annealed from 300 K to 1600 K at various heating rates and then rapidly quenched back to 300 K. Figure <ref type="figure">7</ref> shows the real-space configurations of the Rh, RhGa, RhIn, and RhBi NPs obtained from the described procedures. For the Rh NP, regions of (111) and (100) surfaces are formed. For the RhGa NP, (100) and (110) facets are evident. For RhIn, the surface is dominated by (100) facets. Conversely, the RhBi NP did not fully crystallize, even at the slowest simulated heating rate (20 K/ns). MD simulations of NPs formed from a Rh:M ratio of 2:1 were also performed to study the impact of atomic ratios on the particles' surfaces. The simulated results displayed in Figure <ref type="figure">S23</ref> show that Ga, In, and Bi all prefer to be near the surface, as compared to Rh which occupies the interior of the particle. This tendency is the strongest for Bi and weakest for Ga. The propensity for Bi to be on the surface is such that the nanoparticle undergoes phase separation, forming a region with a pure Rh interior covered by a single layer of Bi and a Bi-rich intermetallic region. We also performed machine learning accelerated molecular dynamics simulations to further study the interaction of furfural with each catalyst. First, we used the ML potentials for Rh, RhGa, RhIn, and RhBi to synthesize a slab of the respective metal or bimetallic system in a (1.9 &#215; 1.9 &#215; 3.8) nm 3 prismatic simulation cell with periodic boundary conditions at 1600 K. The slabs were then cooled down to 300 K at 1 K/ps. Furfural molecules were then added to the simulation cells containing the slabs sequentially at 300 K, subject to the condition that no atom was inserted at a distance less than 2.8 &#197; away from the existing atoms, including both the metal and preexisting furfural molecules. After each furfural molecule was inserted, the system was equilibrated with a short 1 ps ML MD run. Adsorption of furfural molecules on the Rh and RhM nanoparticles was also simulated using the ML potentials. Using a cubic simulation box of 9.6 nm in edge length with periodic boundary conditions, 256 furfural molecules are inserted into the simulation box sequentially within 16 ps at 300 K. The sample is then heated to 600 K at a heating rate of 1 K/ps.</p><p>Figure <ref type="figure">S24</ref> shows snapshots of the system at 600 K. For clarity, only furfural adsorbed on the NP are shown.</p><p>From ML simulations, we found that the furfural molecules adsorb on Rh, RhGa, and RhIn surfaces but not on the RhBi surface. Furfural molecules lie mostly flat on the Rh surface but are slightly tilted on RhGa and nearly completely vertical on RhIn, as shown in Figure <ref type="figure">8a</ref>. The differences between the adsorption of the furfural molecule on the RhGa and RhIn are more subtle. Figure 8b shows the angle between the normal vectors of the furan ring and the metal surface as a function of time for a single furfural molecule on Rh, RhGa, and RhIn. An angle near Rh RhGa RhIn 175 0 25 50 75 100 125 (ps) 0 2 4 8 6 10 Rh Rh n RhGa RhIn RhBi a) b) Rh RhGa RhIn 175 0 25 50 75 100 125 150 Time (ps) 0 2 4 8 6 10 Angle (degrees) Rh Rh n RhGa RhIn RhBi a) b)</p><p>0&#176; or 180&#176; means the molecule is flat, and an angle near 90&#176; means the molecule is vertical. The simulations show that for the Rh surface, the furfural molecule lies flat during the entire 10 ps simulation at 300 K. On the RhGa surface, the molecule can rotate quite easily and be flat or vertical. On the RhIn surface, the furfural molecule remains mostly vertical throughout the simulation time frame. In general, the furfural molecules favored the edges between facets for Rh, RhGa, and RhIn. On RhBi, the furfural molecules only physisorb to the surface and maintain a high degree of mobility (Figure <ref type="figure">S24</ref>). Since there are no dominating facets for furfural adsorption, the effects of surface orientations can only be taken into account by a statistical analysis. However, the difference in furfural adsorption behavior due to the chemical composition of the nanoparticles is much more apparent than the simulations suggest. Our analysis shown in Figure <ref type="figure">8</ref>, using a slab geometry with somewhat amorphous configurations, is representative. The simulated adsorption geometries and behavior agree well with the FAL DRIFTS results and provide further evidence for the proposed steric control of product selectivity.</p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="3.">Conclusion</head><p>In summary, by synthesizing and applying Rh-based intermetallic catalysts, we show that the identity of the secondary metal used to form iNPs can direct the selectivity of furfural hydrogenation. The secondary metal introduces surface steric effects that tune the furfural adsorption mode and enable product control. RhBi@MSW predominately resulted in blocked active sites and dramatically decreased interaction with furfural. Furan and furfuryl alcohol were selectively produced using RhGa@MSW and RhIn@MSW, respectively. The RhM@MSWs resulted in upwards of 85 % selectivity to the major product formed in each case. CO Synthesis and characterization methods, molecular dynamics simulations details, additional TEM, EDS, BET, and ICP characterization of the RhM@MSW catalysts, and XRD and TEM comparing the as-synthesized RhIn@MSW to the recycled RhIn@MSW. </p></div>
<div xmlns="http://www.tei-c.org/ns/1.0"><head n="4.5">ACKNOWLEDGEMENTS</head></div></body>
		</text>
</TEI>
