香京julia种子在线播放

The calculations presented in this study were carried out using the Vienna Ab initio Simulation Package (VASP) (Kresse and Furthmüller, 1996; Kresse and Hafner, 1993), where the latest edition of this software (6.1.2 standard) was used. The core electrons were treated with VASP’s projector augmented waves (PAW), 2017 edition (Lejaeghere et al., 2016), to expedite the calculation of the Kohn–Sham equations with an energy cutoff of 500 eV for all calculations. For the surface calculations the energy tolerance was 1 × 10⁻⁴ eV, while the force tolerance was set to 0.03 eV/Å with Gaussian smearing and a sigma value of 0.1 eV. The k-point mesh was (2 × 1 × 1) for the surface calculations and (7 × 7 × 10) for the bulk structure. The bulk structure, being hexagonal close-packed, had the lattice constants of a = b = 4.479 Å, and c = 3.113 Å (see Supplementary Figure S1 for details) using the Strongly Constrained and Appropriately Normed (SCAN) functional. Electronic exchange and correlation was treated at the generalized gradient approximation level using the Perdew–Burke–Ernzerhof (PBE) functional (Perdew et al., 1996). This level of theory was compared to the more sophisticated meta-GGA functional SCAN (Sun et al., 2015) and the van der Waals–DF functional, optB86b-vdW (Becke, 1986) for the BDO molecule in the gas phase. The SCAN functional has been demonstrated to work effectively with spin-polarized metal oxides (Sun et al., 2015) and therefore was used on all surface calculations presented. While RuO₂ has been determined to be antiferromagnetic (Berlijn et al., 2017) as the magnetic ordering of the system delays calculations and is unlikely to significantly affect the adsorption energy calculations presented here. As such, all calculations performed in this study were not spin-polarized.

The rutile structure for RuO₂ was chosen as it is the most stable under standard conditions (Haines et al., 1996), where the bulk structure was cut for the (110) facet where an O_bridge termination was determined to be the most stable surface termination. To emulate a semi-infinite surface, four tri-layers (the tri-layers being a function of the rutile stacking) were used with a 13 Å vacuum layer, with dipole layer corrections being applied in the z ^ -direction (Reuter and Scheffler, 2002). The bottom two layers were fixed with the top two relaxed to allow for the adsorption of BDO. The adsorption energy was calculated as follows: E a d s = E B D O / R u O 2 ( 110 ) − E R u O 2 ( 110 ) −   E B D O ( g ) , (1) where E B D O / R u O 2 ( 110 ) , E RuO 2 ( 110 ) , and E BDO ( g ) are the total energies of 2,3-butanediol adsorbed on RuO 2 ( 110 ) , the clean RuO 2 ( 110 ) surface, and 2,3-butanediol in the gas phase, respectively. The distortion energy was calculated as follows: E dist = E distorted geometry molecule −   E free gas molecule , (2)

E free gas molecule is the total energy of the molecule converged in vacuum space, and E distorted geometry molecule is calculated from a single-point calculation of the adsorbed molecule on the surface after the surface had been deleted. These two energy calculations have been included to give information about the configuration of the structures and as a comparison point in the adsorption analysis. Additionally, an electronic analysis was carried out, resulting in partial charge density distributions and differential charge density calculations. The differential charge density calculation is based on the following equation: Δ ρ ( r → ) = ρ BDO / RuO 2 ( 110 ) ( r → ) − ρ RuO 2 ( 110 ) ( r → ) − ρ BDO ( g ) ( r → ) (3) where ρ BDO / RuO 2 ( 110 ) ( r ⇀ ) , ρ RuO 2 ( 110 ) ( r ⇀ ) , and ρ BDO ( g ) ( r ⇀ ) are the ground-state charge distribution for adsorbed system, the clean surface, and the gas-phase molecule that is fixed in its distorted adsorption geometry, respectively. Additionally, a Bader charge analysis (Henkelman et al., 2006) was performed to quantify the atomic interaction during adsorption.

A density of states analysis was carried out such that the Fermi level was set to zero. For the gas-phase BDO the Fermi level was taken to be halfway between the HOMO and LUMO states as was done by (Mittendorfer and Hafner, 2001). In addition to the density of states a d-band analysis was performed where the center (Eq. 4) and width (Eq. (5) were identified by using the following equation (Kitchin et al., 2004) (Hensley et al., 2016): ε d = ∫ − ∞ ε Fermi E ρ ( E ) d E ∫ − ∞ ε Fermi ρ ( E ) d E , (4) w d = ( ∫ − ∞ ε Fermi E 2 ρ ( E ) d E ∫ − ∞ ε Fermi ρ ( E ) d E ) 1 / 2 , (5) where E is the given energy from the DOS analysis, ρ ( E ) is the density of the electronic states for E, and ε Fermi is the Fermi level.

To determine the most favorable adsorption site with an ensemble of predetermined binding sites, an ensemble-based approach was performed so that “ensemble-based” refers to the process that was conducted without the aid of a global optimization algorithm. As such, it is based on chemical intuition. Within this framework, seven unique possible adsorption sites were determined and tested for two different ways: with the 2,3 C-C bond parallel to the a -axis (horizontal) and with the 2,3 C-C bond parallel to the b -axis (vertical). The conformations of these seven sites are given in Supplementary Figure S2.

We employed the Northwest Potential Energy Surface Search Engine (NWPEsSe) software (Zhang et al., 2020) coupled with the xTB program (Bannwarth et al., 2021) and the VASP program to identify the energetically favorable adsorption configurations of BDO on RuO₂(110). The xTB program provides a semiempirical extended tight-binding package to accurately predict molecular structures and properties without the need to switch to a computationally more expensive model such as the DFT-based method in VASP. In this package, the GFN2-xTB (Bannwarth et al., 2019) method can quickly perform calculations of structures and interactions in molecular structures, and the GFN-FF (Spicher and Grimme, 2020) is a generic force field for even faster evaluation of structures and dynamics for large molecules. Similar to our quantum calculations using VASP, the 110 surface of the rutile structure of RuO₂ was used. The surface of our RuO₂ for the non-periodic boundary conditions (PBC) model (used in GFN-FF and GFN2-xTB) is around 17 × 17 Å², and the depth is approximately 12 Å corresponding to four layers of Ru atoms. To accelerate the geometric optimization of the BDO/RuO₂(110) adsorption system, all atoms in the RuO₂ model were fixed using the input files from the ensemble-based adsorption file and their positions were not changed during the optimization. The geometric search of a favorable BDO/RuO₂(110) adsorption configuration was conducted in three steps: 1) BDO was added to the center region of the RuO₂(110) surface and a geometry optimization was carried out using GFN-FF (Spicher and Grimme, 2020) with NWPEsSe. A total of 20,000 structures were generated. The last 1,200 optimized structures with the lowest adsorption energies were kept. 2) The obtained structures in the previous step were further optimized by GFN2-xTB (Bannwarth et al., 2019) to achieve more accurate geometries and energies. The last 10 optimized structures with the lowest adsorption energies were kept. 3) The obtained 10 structures were further optimized using the VASP code, which were used to obtain the corresponding ground-state adsorption energies.

The first step in analyzing the dehydration of BDO is to determine the reaction pathways in the absence of a catalyst based on what is known in the literature. Dehydrating BDO can have a variety of different products such as 3-hydroxy-2-butanone and 2,3-butanedione, but there are three distinct products whose pathways lead to butene (Zheng et al., 2015). These products are 1,3-butadiene, 2-butanone (methyl ethyl ketone), and 2-methylpropanal and are denoted by their respective pathways in Figure 1 and the accompanying molecular diagrams in Supplementary Table S1. Beginning with Pathway 1, BDO undergoes a double dehydration where both hydroxyl groups are removed in one step leaving behind carbon-carbon double bonds in 1,3-butadiene. As this structure is more stable than the final product of butene it is likely for the reaction to get stalled here. For Pathway 2, the step from BDO to 2-butanone is the most dependent on the functional choice with the two extremes being endothermic for PBE and exothermic for SCAN; the van der Waals optB86b-vdW functional is only endothermic by 0.06 eV. The energetic hurdle for this pathway is the transition from 2-butanone to 2-butanol, which is accomplished through hydrogenation. Finally, the transition to butene is favorable with the total process being endothermic to varying degrees depending on the functional choice. In the last pathway, Pathway 3, the dehydration process is very similar to Pathway 2, which accounts for the similarity in their energy diagrams. The difference arises from the placement of the double-bonded oxygen. For 2-butanone, the double-bonded oxygen forms a ketone, whereas for 2-methylpropanal the double-bonded oxygen shifts to an end carbon. In both cases the dehydration of one water molecule occurs. Similarly, as in Pathway 2, 2-methylpropanal is then hydrogenated before the final dehydration to butene—the hydrogenation step requiring the greatest amount of energy input.

Comparing the three pathways, the SCAN functional produced the lowest energies out of the three levels of theory, two exceptions being the intermediary structures in Pathways 2 and 3—2-butanol and 2-methyl-1-propanol. The energetic difference is especially clear when looking at the relative reaction energies for the dehydration process which was 0.395, 0.058, and 0.012 eV for the PBE, VDW, and SCAN functionals, respectively. The SCAN functional improves upon GGA as it has a better capability to distinguish between covalent and metallic bonds, additionally describing them as semi-local (Sun et al., 2015). This distinction should result in less self-interaction error during the DFT calculations.

To understand the bonding more accurately between BDO and RuO₂(110), the electronic configuration of the gas phase BDO was analyzed through partial density of states (PDOS). We also analyzed the partial charge density distribution of its four highest occupied molecular orbitals (E1–E4) as well as its lowest unoccupied molecular orbital (E5). This information will later be used in the study to identify changes in the electronic configuration upon adsorption to the surface. The PDOS shown in Figure 2 was performed for optimized configurations of BDO in the gas phase. The p-states of each of the oxygens in the hydroxyl group and their bonded carbons were examined. s-states were left out of this plot as they will largely not participate in the bonding to the surface. The partial charge energy intervals that were used for these plots are given in Supplementary Table S2. The Fermi level was calculated to be −3.07 eV away from the HOMO level.

Examining the p-states shows that the HOMO, identified as E4, is largely occupied by the two oxygens in the hydroxyl groups. The HOMO having the highest magnitude of states with the hydroxyl oxygens suggests that these oxygens are going to be the main participants in the chemisorption of BDO. This is further evidenced with E3 as well as E4 whose partial charge density distribution shows defined p-orbitals surrounding the oxygen species. Moving lower in energy to the E1 and E2 peaks, the charge density is more clearly including the hydrogen species in the hydroxyl group as well as the hydrogen species bonded to each respective carbon. E2 additionally has a very defined σ -cloud around C2 and C3. The LUMO peak at E5 has a significantly smaller peak with the charge density distributing more on the hydrogen species within the hydroxyl group as well as the corresponding bonded carbons.

To determine the most favorable adsorption site of 2,3-butanediol, the orientation of the hydroxyl groups relative to the surface—the main interest in the eventual dehydration reaction—needs to be analyzed. Therefore, for 2,3-butanediol, two alternatives were considered: one with hydroxyl groups facing toward the surface (Figure 3A) and the other with hydroxyl groups facing away from the surface (Figure 3B). As anticipated, the hydroxyl groups facing toward the surface was the more favorable configuration, with an adsorption energy of −2.34 eV, classifying it as chemisorbed to the surface. This is evident in Figure 3A, where the charge differential image clearly shows that charge is exchanged from the hydroxyl groups between the oxygen’s p orbital and the Ru 5-coordinated atom’s d orbital. The oxygen acts as an electron acceptor that is donated from Ru. Additionally, stabilization from the subsurface Ru species is also evident as charge is donated to the surface. In the case when the hydroxyl group faces away from the surface, virtually no charge exchange occurs (Figure 3B). This is consistent with the weak adsorption energy of −0.30 eV, which is indicative of physisorption. Therefore, it can be concluded that the difference between chemisorption and physisorption to the surface depends on the hydroxyl group orientation. This orientation of the hydroxyl groups in the diol is also supported in the literature when one attempts to carry out a dehydration reaction (Acharya et al., 2013).

This differential charge analysis was based on placement at Site 1 in Supplementary Figure S2 using the ensemble-based approach. Additionally, the configuration chosen was supported by an NWPEsSe-based analysis. The most favorable structure, determined by the NWPEsSe global optimizer, shows that the hydroxyl group orientation faces the surface (Figure 3C), where, again, charge transfer is observed between the oxygens in the hydroxyl group and the Ru 5-coordinated atom. This results in a more favorable adsorption energy of −2.61 eV.

Analyzing the three different reaction pathways toward butene formation, as shown in Figure 1, makes it clear that Pathway 1 is unlikely as it has the highest potential to get stalled in the intermediary product of 1,3-butadiene. This pathway has been explored in the literature as 1,3-butadiene is an important additive in the production of rubber (Sun et al., 2020); it was found to be hard to produce from BDO as 2-butanone was more favorable (Malcolm Winfield, 1945). This is consistent with the reaction energy pathway comparison as seen in Figure 1. Pathways 2 and 3 are more likely as that thermodynamic sink is not present in the same capacity. This has also been found experimentally that both 2-methylpropanal and 2-butanone are the main intermediates in the dehydration of BDO to butene (Zheng et al., 2015).

The HOMO level being primarily occupied by the oxygen p-states indicate that they will mainly determine the reactivity with the surface, as shown in Figure 2. This is later supported by Figure 3, which shows that having the oxygen functional groups face the surface are far more favorable, resulting in chemisorption. This is further supported by the results of the global optimization study. As evidenced by Figure 4B, the least favorable structures obtained through a thorough testing of 1,200 structures were the ones where the BDO functional groups are directed toward the vacuum layer. As the BDO surface-facing orientation is required to promote its eventual dehydration, we expect that this adsorption configuration would be more favorable.

The adsorption analysis of BDO on RuO₂(110) began by understanding its surface orientation. As is evident by the differential charge analysis (Figure 3) and the global optimization search (Figure 4), having the oxygen functional groups face the surface leads to the most favorable adsorption site. This is an expected result, as the PDOS reveals that the HOMO level was mostly occupied by the oxygen p-states. Additionally, for an optimal chemisorption configuration, the hydroxyl groups need to be parallel with the Ru_5c atoms as evidenced by Table 1. While other orientations of the BDO molecule with the surface could have been tested manually, the results from our NWPEsSe-based analysis confirm that the horizontal configuration is the most favorable when BDO adsorbs on a pristine RuO₂(110) surface.

Further confirmation of the differential charge analysis at the surface can be obtained through a Bader charge analysis that is given in Table 4, which was completed for both the NWPEsSe-based and ensemble-based determined global minima. In both cases, the Bader charge on the Ru_5c atoms decreases with respect to the clean surface indicating a depletion of electrons. On the other hand, the Bader charge analysis of the O₁ species, one of the hydroxyl oxygens, increases upon adsorption of BDO as compared to the clean surface indicating an accumulation of electrons. The hydroxyl oxygens Bader net atomic charge doesn’t necessarily indicate a net gain of electrons and that is most likely attributed to how the charge is partitioned in the calculation. Analyzing Figures 3A–C it’s clear that there is both a charge gain and a charge loss occurring around each oxygen species in BDO’s hydroxyl functional groups. Therefore, the Bader analysis is most likely including both in its calculation. Interestingly, in the case of the ensemble-based structure the Bader charge of the O₂ species within the adsorbed BDO molecule is unchanged as compared to its gas phase value, while there is slight loss for the O₂ species in the NWPEsSe-based structure as compared to the gas phase species. However, these changes are minor as compared to those occurring for the Ru_5c species. As such, we conclude that the Ru_5c species donates electrons to the hydroxyl oxygen to chemisorb BDO to the metal oxide surface.

The electronic configuration of the two different systems is similar. The PDOS of the main species that are involved in the adsorption of BDO (Figure 5) and the resulting partial charge density distributions (Figure 6) further confirm that both the ensemble-based and the NWPEsSe-based structure chemisorb to the surface. The overlap between the hydroxyl oxygen’s p-states and Ru_5c d-states suggest the exchange of electrons. This is evidenced by the distinct partial charge density distributions, specifically between the higher energy peaks shown in Figure 6 (F3, G3, F4, and G4). In these analyses, a p-orbital is clearly seen on the hydroxyl oxygens, and, in the case of F4 and G4, there is significant charge distributed on the Ru cations. The final peak in the p-states, G3 and F3, show the most similarity to the E4 HOMO peak found for the BDO molecule in the gas phase (Figure 2). The corresponding peaks are shifted 0.6 and 0.4 eV lower than the gas phase values for F3 and G3, respectively. The similarity of the partial charge densities is more clearly seen when comparing the F3 and E4 partial charge distributions as the ensemble-based configuration looks more like the converged gas phase structure. In both cases, F3 and G3, the p-orbital cloud is distinct in the hydroxyl oxygens, and the σ cloud is distinct between the C2-C3 bond and the accompanying methyl group attachments to C2 and C3. As these are higher energy peaks, they are less stable and have slight overlap with the Ru cations they are bonded to (Figure 6). The chemisorption is also further supported in Table 3 where the d-band center and width decreases and becomes broader as the Ru_5c cations exchange electrons with BDO. The d-band center value of the NWPEsSe-based structure being slightly more negative indicates a stronger chemical bond as compared to the structure identified by our ensemble-based method.

Between the two analyses, ensemble-based and NWPEsSe-based, the ground-state structure had the same placement on the lattice indicated as Site 1 in Supplementary Figure S2 of the SI. However, the global minimum of the NWPEsSe-based structure has a lower adsorption energy (−2.61 eV) as compared to the ensemble-based local counterpart (−2.34 eV). The lower energy from the NWPEsSe-based structure is likely attributed to the orientation of the methyl groups. With the NWPEsSe-based structure the methyl groups point toward the vacuum layer, making them perpendicular to the surface, as identified by the large dihedral bond angle with the oxygen functional group. This orientation, while having a comparable distortion energy to the ensemble-based structure from the gas-phase BDO, results in shorter bond lengths. Therefore, we hypothesize that in the ensemble-based structure, the flattening of the methyl groups to become parallel with the surface causes a strain on the hydroxyl bond to the surface and the bond lengthens to accommodate it, resulting in a less favorable adsorption energy. This conjecture is supported by the results from the calculations on the 10 lowest structures from the GFN2-xTB program. Their configuration details and adsorption energies are supplied in Supplementary Table S4. Structures 1–3,5, and 6 all have similar dihedral angles, bond lengths, and adsorption energies as compared to the ensemble-based structure. Therefore, we conclude that the ensemble-based structure adsorption site analysis was able to determine one of the lowest energy adsorption configurations but was limited in identifying the global ground-state structure.

We compare the 10 lowest structures from the GFN2-xTB run and their corresponding DFT-based optimizations in Figure 7. The relative adsorption energies are shown as the GFN2-xTB program was not specifically parameterized for metal oxide adsorbate systems and therefore provides unrealistic absolute adsorption energies. As such it highlights the need to further optimize these structures using DFT-based calculations to get more accurate adsorption energies. There is good agreement between the two relative energies (although there are two defined outliers seen in orange and purple, Structures 7 and 9, respectively). Structure 7 has a lower energy as it is closest configuration-wise to the ground-state structure seen in pink, Structure 10. Conversely, Structure 9 has the highest relative energy, making it less favorable. As shown in Figure 7, Structure 9 did not in fact form a strong bond with the surface and instead is flattened out by interacting with the neighboring O_n’s through hydrogen bonding. The lowest adsorption energy is Structure 10, making it the most favorable conformation and the ground-state structure identified, while Structure 9 is the least favorable. When examining Supplementary Table S4 it is also clear from the bond lengths, dihedral angles, and adsorption energies that Structure 7 is the structure that has the closest conformation to Structure 10. The agreement between the two systems confirms the accuracy of the calculations in determining the ground-state structure.

To take part in the Resource Identification Initiative, use the corresponding catalog number and RRID in your current article. For more information about the project and for steps on how to search for an RRID, click here.