(189bf) Reactive Sorption of Sulfur Contaminants By Copper Oxide: A First-Principles Study

Natural gas and other gaseous fuels contain varying amounts and types of sulfur compounds, e. g. H₂S, COS and CS₂, which are corrosive to equipment and have a strong poisonous effect on catalysts. Even in trace amounts at which they are found (1-1000 ppm-mol) they have a negative effect¹. Moreover, these gases are hazardous to humans and the environment. One of the common contaminants H₂S, becomes a source of acid rain when oxidized in the atmosphere to SO₂. To this regard, processes based in adsorptive separation constitute a practical way to remove contaminants, specifically the so-called reactive separation processes, which may be defined as the coupling of chemical reaction and physical separation in a single unit operation². The use of solid phases for reactive separation through a process known as reactive sorption is advantageous versus other regenerable adsorbents because the thermodynamics are more favorable compared to physisorption, leading to higher purity streams and higher purification capacity, and also because the contaminants are permanently sequestered in a stable and environmentally benign form³ (e.g. as a metal sulfide). Although metal oxides can be used in this way to remove sulfur, little research has been conducted to understand these reactions at a fundamental level, which is the rationale under this study.

In order to improve our understanding of these reactions at the atomic level, we have performed a computational study using density functional theory (DFT) of the reaction mechanism of the reactive sorption of H₂S by CuO, which produces CuS and H₂O as products. Using models of CuO surfaces we studied all of the elementary steps that comprise this reaction. We first studied the reaction on the most stable CuO surface, i.e. {111}, for which we checked the influence of several parameters, namely the reactivity of different active centers, sulfur coverage and presence of water molecules. We then extended this study to other non-polar CuO surfaces.

In terms of methodology we have used the planewave-based formalism of DFT as implemented in the VASP software package⁴, which makes use of the PAW method to represent inner atomic cores. The PBE functional was selected for accounting exchange-correlation effects, with the dDsC correction for dispersion forces and the Hubbard U approach to account for the proper description of 3d electrons of Cu atoms. Reciprocal space was modelled using the Monkhorst-Pack scheme with a 3x3x1 grid mesh. The CuO surfaces were modelled using the supercell approach, for which 2x1 supercells were used. The slab thickness was set to 7 Å.

The first part of this study deals with the reaction mechanism on a CuO {111}. This surface exhibits four types of atoms: Cu and O atoms with 4-fold coordination (4c) and also Cu and O atoms with 5-fold coordination (5c).

Regarding the process involving 5c Cu atoms and 4c O atoms, we find that the H atoms interact with the surface O atoms and the S atom interacts with surface Cu atoms which act as Lewis acid centers. This results in the activation and cleavage of one of the S-H bonds of H₂S, which produces two adsorbed species on the surface, an SH and an OH. The OH group remains on the initial position of the original O atom whereas the SH stays mono-coordinated to a surface Cu atom. This first S-H bond activation is favorable (ΔG=-126 kJ/mol) and proceeds barrierless. After this step, the adsorbed SH species can undergo through a similar activation process, which in this case it yields a S atom and an additional OH group on the surface. Contrary to the first S-H activation in H₂S molecules, this second S-H bond cleavage is notably less favorable (ΔG=-11 kJ/mol), and it is an activated process with a barrier of 55 kJ/mol. Upon the second dissociation the two OH groups can interact, and a donation of a H atom from one to the other can result into a water molecule and surface vacancy where the O atom was. We find that this step is thermodynamically very costly (ΔG=+139 kJ/mol) and will hinder the process. However, when this water formation step is coupled with the healing of the vacancy by the S atom, the process becomes more favorable (ΔG=+7 kJ/mol), which suggest a concerted step in which water is formed and at the same time the resulting surface vacancy is healed by the S atom. Finally, the water molecule is desorbed from the surface in an endothermic process (ΔH=+32 kJ/mol) but the gain in entropy makes it favorable (ΔG=-24 kJ/mol). The overall thermodynamic balance for this process is favorable as shown by the global ΔG=-106 kJ/mol.

Interestingly, we find that the process is influenced depending on the different active sites involved, i.e. different reaction pathways using different types of Cu and O atoms. When 4c Cu atoms are involved, the equivalent elementary steps show different thermodynamics. The first S-H activation is noticeable less favorable with ΔG=-7 kJ/mol. Similarly, the second S-H activation is less favored with ΔG=+17 kJ/mol. This can attributed to their lower Lewis acidity compared to Cu atoms surrounded by five O atoms. However, the next step which involves formation of water and vacancy healing by a S atom is very favorable and compensates the previous ones with ΔG=-125 kJ/mol, compared to ΔG=+7 kJ/mol. Both this pathway and the one exposed above share the same global ΔG=-106 kJ/mol.

In addition, we find a strong influence of the type of surface O atoms involved in the reaction. When the reaction occurs using the 5c O atoms, the process becomes less favorable. The first and second S-H activations become ΔG=+4 kJ/mol and +10 kJ/mol respectively, and the formation of water gets to ΔG=+89 kJ/mol. The overall thermodynamic balance is less favorable as well, with ΔG=-52 kJ/mol compared to ΔG=-106 kJ/mol with the higher-coordinated O atoms.

We also studied the effect of sulfur coverage by modelling the adsorption several H₂S molecules up to n=4. Upon adsorption of n=3-4 molecules, which simulates a high sulfur coverage in our model, S atoms and SH adspecies have a tendency to form covalent S-S bonds and create polysulfides adsorbed on the surface, especially in the form of S₂ dimers. Interestingly, we observe that the formation of these polysulfide species turns more favorable the activation of higher-coordinated O atoms, e.g. ΔG=-76 kJ/mol versus ΔG=+10 kJ/mol for a low coverage. The occurring of these S₂ structures is important because the CuS phase experimentally found, covellite, features S₂ dimers providing a link between the early stages of the reaction and complete transformation of CuO to CuS.

Another important factor is the influence of water molecules on the surface. Reports found in the literature⁵ show that in our operational conditions (1 atm, 298K, 1000 ppm H₂O, 100 ppm O₂, μ_O= -0,41 eV, μ_H2O= -0,49 eV) a ½ monolayer of water will be covering the CuO surface. We find than when water is present in the surface the elementary steps become more favorable in most of the pathways, especially for the activation of the second S-H bond (ΔG=-11 kJ/mol vs. -43 kJ/mol in bare surfaces) which as explained before is harder.

Finally, we extended the study to other non-polar CuO surfaces ({-111}, {011} and {101}) which are more unstable than the {111} surface, i.e., they have a higher surface free energy. We find a both ΔG and activation barriers are inversely proportional to the surface free energy, i.e., the more unstable the surface (higher surface free energy) the more thermodynamically and kinetically the reaction becomes in all of its elementary steps.

In essence, we performed a computational study of the reactive sorption of H₂S by CuO taking into account several parameters. The study shows how the activation of the first S-H bond is easier than of the second, and water formation is energetically difficult but a concerted mechanism in which a S atom heals the vacancy left by water helps this process. We also conclude from the current state of this study that for a given a surface the coordination and chemical environment of active centers strongly influences the energetics of the process. In addition, we find a positive effect of water and sulfur coverage, and a proportionality between the surface free energy of CuO surfaces and their reactivity towards H₂S. This study will shed light into our understanding of this complex process at the atomic level and contributes towards developing a rational framework for designing materials for removing sulfur contaminants.

[1] Hidnay, A. J., et al., Fundamentals of Natural Gas Processing, CRC Press, 1-21 (2006).

[2] Kulprathipanja, S., Reactive Separation Processes. Hemisphere, Washington (2002).

[3] Samokhvalov, A., et al., Cat. Rev. Science and Engineering, 52, 381-401 (2010).

[4] Kresse, G., et al., Phys. Rev. B, 47, 558 (1993).

[5] Fronzi, M., et al., RSC Advaances, 7, 56721 (2017).