Enhancing the reactivity of clean, defect-free epitaxial graphene by the substrate—Experiment and theory

Graphene (Gr) is a key nanomaterial in a variety of diverse areas including catalysis.^1,2 Although many liquid phase reactions are known that are catalyzed by Gr flakes in solution^3,4 (in particular for sulfur compounds^5–11), only a few^12–14 reactions, heterogeneously catalyzed by clean (nonfunctionalized) Gr, have been characterized at ultraclean, ultrahigh vacuum (UHV) conditions. Here, we present one of the few studies^15,16 where an actual surface reaction takes place at UHV on Gr and not just an enhanced adsorption (increased binding strength and coverage) of gas-phase species, that was reported earlier.^12–14 At low (⁓100 K) temperatures, clean, nonfunctionalized, and virtually defect-free single-layer Gr is promoting the decomposition of H₂S. This result is unexpected since adsorption of all other gas-phase species studied so far on the same system (water, alkanes, benzene, CO, and CO₂)^17–19 with the same experimental setup is just molecular (or does not take place at all at 100 K). For gaining a detailed understanding of the mechanism by which H₂S interacts with Gr/Ru (which may allow for designing other working Gr catalysts), experimental surface science data are combined with density functional theory (DFT). Generally, the surface chemistry of sulfur compounds is rich and complicated, which makes it interesting for a mechanistic basic science study. Ruthenium is a standard substrate to grow Gr since the Gr synthesis is experimentally rather simple, complicated morphologies (e.g., rotational domains) do not form, and Gr/Ru is well characterized.^20–22 Also, in a prior study,²³ Gr/Ru did show promising catalytic activity. In future projects, we will consider nonmetallic substrates. In that case, however, the graphene synthesis is more challenging.

Regarding applications, the largest industrial source of H₂S is not only generated by the hydrodesulfurization (HDS) process in petroleum refineries but also coke ovens, sewerage, paper mills, and tanneries generate it. Also, natural gas can include it. Therefore, energy efficient removal of H₂S gas is of practical interest²⁴ as well as H₂S sensors for diverse applications.^25,26 The catalyst studied here is sulfur tolerant and does decompose H₂S by forming gaseous hydrogen. A catalyst poisoning over time is, therefore, unlikely. Thus, in principle, H₂S waste generated, e.g., in HDS, could be used to support a hydrogen economy and/or H₂ could be fed back into HDS.

Considering Gr and other carbon systems is besides scientific curiosity motivated by developing (noble) metal-free catalysts.^27,28 Noble metal-free catalysts would allow for developing sustainable chemical processes since carbon materials are inexhaustible and as such sustainable, in contrast to precious metals. Highly selective catalysts can reduce unwanted side reactions, reduce waste products, and increase energy efficiency (sustainable/green chemistry). The concept of noble metal-free catalysis dates back many years but has so far mostly been explored in liquid/solid phase reactions. Gas-surface reactions are preferred for industrial processes due to easier separation and less waste formation. One could consider other 2D (two-dimensional) crystals as potential metal-free catalysts such as silicene (2D silicon) or germanene which have greater chemical reactivity than Gr due to sp²/sp³ hybridization and differences in the π bonding as compared to Gr. However, the drawback of extreme reactivity is that the reactions become likely stoichiometric and not catalytic, i.e., the 2D support just acts as a reactant and is consumed as part of the reaction. Certainly, there are many other strategies to develop catalysts such as utilizing non-noble metals, reducing catalyst loadings, etc. In a broader perspective, however, illustrating an example where clean, nonfunctionalized Gr is catalytically active, extending the use of Gr beyond the utilization of its remarkable electronic and mechanical properties.

In this study, Gr was synthesized at UHV using two different well-established procedures applied for decades in numerous projects by numerous groups. First, C₆D₆ was dissociated on ruthenium(0001) using thermal desorption spectroscopy (TDS) cycles, which allows for determining the vacancy defect density (⁠ $≪1%$ monolayer, ML) by monitoring the residual D₂ desorption while forming graphene (hereafter referred to as low temperature/low T prep., Fig. S1).^17,43 The residual D₂ desorption signal is basically zero (Fig. S1).⁴³ Thus, the vacancy defect density in Gr is practically zero, well below 1% ML. TDS is a simple surface science sample temperature ramping technique, for details see the supplementary material. According to scanning tunneling microscopy (STM)^20–22 studies, the low temperature (T) TDS preparation leads to graphene islands separated by island borders (i.e., grain boundaries or 2D line defects), but vacancy defects do not form.

In a second preparation, Gr was formed in a one-step dissociation process of C₆D₆ at high temperatures (T = 1000 K) which minimizes, according to STM (Refs. 20–22) studies, not only vacancy defects but also grain boundary defects (hereafter high T prep.). In summary, the low T prep. generates grain boundaries which are absent in the high T prep. Both synthesis procedures minimize vacancy defects. In both cases, the growth of Gr on Ru is self-terminating since benzene adsorbs only molecularly on Gr/Ru, i.e., single-layer Gr is formed.¹⁷ The UHV synthesis guarantees impurity-free Gr/Ru.²⁹ [See the supplementary material for further experimental details: Auger electron spectroscopy (AES) of Gr/Ru in Fig. S2, AES peak assignments in Tables S1–S3, synthesis summary Table S4, TDS technique, experimental setup.⁴³] Experimental uncertainties (±1.3 kJ/mol) can be estimated from the absolute accuracy of the TDS peak positions (±5 K).

Density functional theory (DFT) was exploited to perform geometry optimizations and electronic-structure calculations. Previous studies showed that DFT-GGA (generalized gradient approximation) can capture the interaction between benzene and transition metals, especially in the iron group.³⁰ We accordingly adopted the Dmol3 code,³¹ exploiting the Perdew–Burke–Ernzerhof (PBE) GGA functional. We made use of Grimme's PBE-D approach to augment PBE, introducing long-range van der Waals (vdW) effects.³² The wave functions were generated by a numerically tabulated basis set of double-ζ plus polarization (DNP) quality. DFT semicore pseudopotentials (DSSPs) were exploited to effectively describe the interaction between electrons and ions via a single effective potential. The Brillouin zone was sampled via a 2 × 2 × 1 Monkhorst–Pack grid.³³ All calculations were carried out in the spin-polarized mode. The energy barriers for molecular dissociation were computed by the nudged elastic band (NEB) approach using the same force-convergence criterion (2.0 × 10⁻³ Ha/Å) adopted for structural relaxations. The transition-state structures were confirmed to be first-order saddle points by carrying out a suitable vibrational frequency analysis.

The graphene/ruthenium(0001) (Gr/Ru) complex, also considered in the previous literature,^34,35 was simulated by means of a (6 × 6) graphene supercell, anchored on the Ru(0001) surface. The Ru substrate was described as a slab, consisting of a (5 × 5) supercell, with four-layer thickness. The Ru slab was initially built by using the optimized lattice constants [a = 2.9 Å and c = 4.3 Å]. A vertical vacuum region of 25 Å was introduced to minimize spurious interactions between periodic images. After relaxation, the cell parameters for the overall system were found to be [a = 14.3 Å (in-plane) and c = 50 Å (out-of-plane)]. Hence, a moderate lattice mismatch (about 3%) exists between the Ru substrate and the graphene sheet (with a lattice constant of 2.4 Å). In the geometry optimization of Gr/Ru, the final structures were obtained by relaxing the coordinates of all C atoms and the two upmost Ru layers. For further details, see the supplementary material.⁴³ 

Figure 1 depicts TDS data where the parent mass of H₂S was detected. Thus, a molecular adsorption/desorption pathway exists. However, importantly, AES data (Fig. 2) collected after typical TDS experiments reveal adsorbed sulfur on Gr/Ru (see the S-AES line at 153 eV). Therefore, H₂S must dissociate. The AES data were collected at room temperature and after flashing the surface above at least 350 K. Therefore, molecular H₂S is not present since it would desorb already at ⁓130 K (Fig. 1). The C- and Ru-AES peak intensities remain unaffected (Fig. 2), i.e., the C and Ru surface concentrations did not change while H₂S dissociates. Therefore, it is very unlikely that the graphene layer was altered. Thus, the process is catalytic and not stoichiometric. Indeed, regardless of the Gr preparation used, multimass TDS experiments (Fig. S3, Table S5),⁴³ indicate that H₂ is the only gas-phase reaction product (see Fig. 3).

On a closer view, the α-TDS peak (solid line in Fig. 1) consists of two contributions (dashed lines) which may indicate different adsorption sites or adsorption configurations.

It is known from prior DFT studies³⁷ that H₂S can dissociate on very specific defects in Gr which could act as seed active sites. However, H₂ formation, regardless of the Gr preparation used (Fig. 3), rules out that vacancy defects and grain boundary defects are the main active sites for H₂S dissociation on Gr/Ru. [H₂ desorption temperatures and TDS intensities (see Figs. S4 and S5)⁴³ are independent of the Gr preparation.] In other words, the formation of H₂ indicates H₂S dissociation. First, vacancy defect sites can be ruled out for both Gr preparations used and can, therefore, not dissociate H₂S. Grain boundaries are present for the low T prep. but not for the high T prep. The H₂ TDS results are identical. Thus, grain boundary defects can be ruled out. Furthermore, in multimass TDS, the intensity ratio of the H₂S parent mass and H₂ desorption signals, respectively, allow for roughly estimating a reaction probability of 0.79, taking mass spec. sensitivity factors into account. Similarly, from AES data, large sulfur, S(a), coverages of at least 0.5 ML (for 7 L H₂S exposure, Fig. 2) can be deduced (see the supplementary material for calculations).⁴³ About 2 h of Ar⁺ sputtering (1 μA sample current, 2000 eV) are needed to clean off S(a) (≤ 7L exposures), again consistent with a large S(a) coverage. Both the reaction probability and S(a) coverage are much larger than the experimentally determined vacancy defect density limit (⁠ $≪1% ML$⁠, Fig. S1)⁴³ of Gr/Ru. This indicates again that defects are not the main active sites for H₂S dissociation. Interestingly, even in the presence of S(a), a H₂ TDS signal is still detected (Fig. 3). Thus, sulfur does not poison H₂S dissociation. (Sulfur does accumulate in the course of the TDS experiments on the surface.) The DFT calculations discussed below do indeed show that adsorbed sulfur decreases the activation energies for H₂S dissociation. Thus, the process becomes autocatalytic. The desorbing H₂ originates from the Gr/Ru surface and not from subsurface sites since H₂ desorbs from Ru at much greater temperatures (200–500 K).³⁸ Certainly, various test experiments were conducted (see Fig. S7 in the supplementary material)⁴³ to rule out experimental artifacts.

The molecular desorption is dominated by a rather broad peak centered at 130 K (Fig. 1, α-peak, solid line) corresponding to a binding energy of 33.9 kJ/mol (for a first-order prefactor of 1 × 10¹³/s). However, on a closer view, two TDS peaks at 123 and 140 K may be identified (dashed lines in Fig. 1), yielding binding energies of 32.0 and 36.6 kJ/mol, in reasonable agreement with our DFT results (see below).

A H₂S condensation temperature of 110–120 K (Fig. 1, c-peak, dotted line) agrees with prior studies.³⁹ 

Assuming that H₂S dissociation occurs in sympathy with H₂ desorption at about 110–120 K (Fig. 3) and considering first-order [in H₂S(a)] dissociation kinetics, the activation energy for H₂S dissociation may be estimated as (28.5–31.2) kJ/mol (prefactor 1 × 10¹³/s), again in agreement with our DFT calculations.

Assuming that H₂ desorption by itself follows second order kinetics [with pre-exponential of 10²¹ cm²/(mol s)],⁴⁰ the H₂ desorption energy amounts to (45.5–50.0) kJ/mol. For zeroth-order H₂ desorption, which is consistent with the TDS curve shapes, we obtain an activation energy for H₂ desorption of 7.4 kJ/mol and a prefactor of 7.0 × 10¹⁰/s (Fig. S6).⁴³ The standard model for zeroth-order kinetics assumes the formation of a two-phase regime: sulfur clusters form on the surface and become covered with hydrogen. Due to a smaller number of neighboring sites, H₂ starts to desorb along the rim of the islands which are replenished by H₂ on terrace sites. Thus, the H₂ rim desorption becomes coverage independent (zeroth order).

One may note that RuO, graphene oxide, or RuS₂ are not formed on our surface. Oxygen containing species would lead to an O-AES peak at 500 eV (Ref. 14) which is not present (Fig. S2).⁴³ RuS₂ which is a known desulfurization catalyst⁴¹ can only form on vacancy defect sites. The vacancy defect density of our samples is virtually zero (Fig. S1).⁴³ The AES carbon line is not altered by H₂S adsorption (Fig. 2), i.e., H₂S adsorption does not generate defects in graphene.

Gr on Ru(0001) exhibits a pronounced corrugation due to a lattice mismatch between graphene and the Ru(0001) substrate. Therefore, the calculated adsorption energy of H₂S on Gr/Ru depends on the adsorption site. For the most stable adsorption configurations (Fig. S8),⁴³ it ranges from −39.8 kJ/mol for H₂S adsorption on “valleys” to −25.7 kJ/mol for on top sites (Fig. 4). Note that the TDS peaks (Fig. 1) are rather broad (FWHM ⁓ 70 K) consistent with different adsorption sites and/or the effect of lateral interactions of the adsorbates. Indeed, experimentally two structures may be evident in TDS (see dashed lines, Fig. 1). The adsorption energy averaged over these two sites amounts to −33 kJ/mol which is in excellent agreement with the experimental estimate for the α-peak (−33.9 kJ/mol) for a saturation coverage. Note that the adsorption energy for epitaxial Gr/Ru is much larger (in absolute value) than the one for free-standing Gr which amounts to only −16 kJ/mol, indicating the importance of the Ru substrate.

The enhancement of molecular binding is due to a combination of charge transfer, polarization, and van der Waals (vdW) forces. Orbital hybridization and charge transfer (from Ru to Gr) induces a surface dipole moment in the corrugated Gr, while vdW interactions increase due to the Ru substrate and the corrugation of Gr.

Unexpectedly, experimental evidence shows that H₂S dissociates on Gr/Ru since adsorbed sulfur is seen in AES (Fig. 2) as well as H₂ desorbs (Fig. 3). Interestingly, DFT calculations clearly show that H₂S dissociation is not favored on defect-free, unsupported (free-standing) Gr. In that case, in fact, the dissociated S(a) + H₂(g) configuration is higher in energy (by 138 kJ/mol) than the undissociated one. In addition, the activation energy would be large (466 kJ/mol), see Fig. 4.

This scenario changes dramatically in the presence of the Ru substrate: the dissociated configuration is now lower in energy than the undissociated one (by 18 kJ/mol) and the activation energy is reduced to only 38 kJ/mol (see Fig. 4). Thus, clearly the Ru substrate makes the dissociation of H₂S favorable. The DFT calculated activation energy is in good agreement with the experimental estimate (about 30 kJ/mol), confirming that the simulations describe well the dissociation process observed experimentally.

We have verified that the dissociation process of H₂S on corrugated graphene, without the Ru(0001) substrate, is also characterized by a significant (about 320 kJ/mol) activation energy. Thus, indicating that the electronic factor, related to the presence of Ru(0001), is more effective in favoring the H₂S dissociation than the geometric one.

To better elucidate the dissociation mechanism of H₂S on Gr/Ru, a detailed electronic-structure analysis has been carried out. The partial density of states (PDOS) in Fig. 5(a) shows that, upon adsorption, because of charge transfer, the sharp peak at about +6 eV (related to empty antibonding orbitals of the isolated H₂S molecule) shifts below the Fermi energy, see Fig. 5(c). The charge transfer weakens the covalent binding between H and S to facilitate the dissociation of H₂S into S(a)  + 2H(a).

Moreover, when the molecule is adsorbed on the surface, its originally antibonding orbitals hybridize more efficiently with the 4d orbitals of Ru [Figs. 5(b)–5(d)]. This effect is clearly absent if H₂S adsorbs on free-standing Gr, since in this case, some antibonding states remain above the Fermi level, implying that the S–H bonds are not easily broken [see the tail in Fig. 5(b) above E_F]. The same conclusion can also be drawn by plotting the electron density deformation characterizing the adsorption of H₂S on Gr and Gr/Ru (Fig. S10).⁴³ Evidently, the electron transfer to the antibonding orbitals of H₂S adsorbed on Gr/Ru is more pronounced than on free-standing Gr. The presence of empty, low lying 3d orbitals of the S atom leads to an effective interaction with the Ru atoms of the substrate, characterized by occupied 4d states, thus making H₂S much more reactive than other molecules like alkanes, CO₂, CO, and H₂. The crucial role of 3d orbitals of S has been observed in many different systems, including transition metal dichalcogenides like TiS₂ (see Ref. 42 and further references therein).

Figure S10 (Ref. 43) also shows that the “valley” sites of corrugated Gr can represent electron donor centers, which efficiently trap H₂S and favor its dissociation by interfacial charge transfer.⁴¹ 

Our computational results (Fig. S8)⁴³ also indicate that, after H₂S dissociation, the H₂ molecule can desorb from the Gr/Ru substrate much easier than the S atom, since the adsorption energy of S is four times that of H₂, consistent with experimental findings. Therefore, after dissociation, S atoms remain on the surface (as observed, Fig. 2) and can aggregate to sulfur clusters, in agreement with the sequence H₂S(g) → … → S(a) + H₂(g) already suggested experimentally.

Interestingly, the simulations indicate that the dissociation of another H₂S molecule is even more favored in the presence of an adsorbed sulfur atom. In fact, besides decreasing the energy barrier (to 8 kJ/mol), the configuration obtained by the second dissociation reaction is made more stable than the undissociated one by 28 kJ/mol. Thus, S atoms act as a promoter for H₂S dissociation, consistent with the experimental observation that a sulfur covered surface is still reactive (Fig. 3). It has been observed before⁴² that the by-product sulfur plays a catalytic role in its own reaction.

Experimental and theoretical evidence is presented that a sulfur compound dissociates on clean, defect-free epitaxial graphene (Gr) in ultrahigh vacuum. Density functional theory results indicate the importance of the Ru(0001) support in catalyzing a pathway with small activation energy for H₂S dissociation. Gr becomes reactive due to a complex interplay of structural and electronic effects. The dissociation of H₂S was rather unexpected regarding other probe molecules studied earlier that adsorb molecularly. A kinetics model was developed based on TDS and AES data as well as experimental kinetics parameters including binding energies and activation energies were determined. DFT calculation provided the electronics mechanism as well as binding energies and activation energies consistent with the experiment. As a potential avenue for future research, exploring nonmetallic substrates could be pivotal for advancing metal-free catalysis.

The Padova group acknowledges funding from Fondazione Cariparo, Progetti di Eccellenza 2017, “Engineering van der Waals Interactions: ….” The Fargo group acknowledges a TA position from NDSU as well as support from RBD Instruments Inc. (Bend, OR) and Princeton Scientific Corp. (Easton, PA) via various donations (equipment, tech support, repairs, and materials).

The authors have no conflicts to disclose.

T. Stach: Data curation (equal); Formal analysis (equal); Investigation (equal); Validation (equal); Visualization (equal). A. Seif: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Software (equal); Validation (equal); Visualization (equal). A. Ambrosetti: Formal analysis (equal); Investigation (equal); Software (equal); Supervision (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). P. L. Silvestrelli: Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Project administration (equal); Software (equal); Supervision (equal); Validation (equal); Visualization (equal). U. Burghaus: Conceptualization (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Supervision (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.