Combining Theory and Experiment for Multitechnique Characterization of Activated CO 2 on Transition Metal Carbide (001) Surfaces

Early transition metal carbides (TMC; TM = Ti, Zr, Hf, V, Nb, Ta, Mo) with face-centered cubic crystallographic structure have emerged as promising materials for CO 2 capture and activation. Density functional theory (DFT) calculations using the Perdew-Burke-Ernzerhof exchange-correlation functional evidence charge transfer from the TMC surface to CO 2 on the two possible adsorption sites –namely MMC and TopC–, and the electronic structure and binding strength differences are discussed. Further, the suitability of multiple experimental techniques with respect to (1) adsorbed CO 2 recognition and (2) MMC/TopC adsorption distinction is assessed from extensive DFT simulations. Results show that ultraviolet photoemission spectroscopies (UPS), work function changes, core level X-ray photoemission spectroscopy (XPS), and changes in linear optical properties could well allow for adsorbed CO 2 detection. Only infrared (IR) spectra and scanning tunnelling microscopy (STM) seem to additionally allow for MMC/TopC adsorption site distinction. These findings are confirmed with experimental XPS measurements, demonstrating CO 2 binding on single crystal (001) surfaces of TiC, ZrC, and VC. The experiments also help resolving ambiguities for VC, where CO 2 activation was unexpected due to low adsorption energy, but could be related to kinetic trapping involving a desorption barrier. With a wealth of data reported and direct experimental evidence provided, this study


INTRODUCTION
The surface chemistry of CO 2 is an active field of research, 1-3 largely motivated by an urgent need for efficient CO 2 capturing materials. Availability of such materials would facilitate the implementation of technologies for CO 2 capture and storage 4,5 (CCS) or its conversion to value added chemicals, i.e. CO 2 capture and usage (CCU). Such routes are ultimately promising for climate change mitigation 6 and worldwide ocean acidification. Still, a relatively high adsorption energy is needed for CO 2 to be captured by a materials surface, only encountered on some privileged materials 1-3 due to the high stability of CO 2 molecules.
Moreover, moderate to strong CO 2 adsorption often involves a significant activation, a process requiring charge transfer from the substrate, 1 leading then to weakened C-O bonds, and resulting in a concomitant molecular bending.
As a versatile class of economic materials, transition metal carbides (TMCs) 7,8 have recently proven interesting for CO 2 activation: 9 Results from density functional theory (DFT) based calculations showed the high potential of molybdenum carbides to strongly activate CO 2 on different stable surfaces. [10][11][12][13] In fact, experiments on different molybdenum carbide phases 10,13-15 indicate CO 2 dissociation at room temperature, as well as catalytic activity for CO 2 hydrogenation at elevated temperatures, thus indirectly implying the formation of an activated CO 2 moiety. Especially the high CO selectivity and conversion in CO 2 hydrogenation over hexagonal α-Mo 2 C powder catalysts is remarkable in the light of practical applications. 13 Aside from molybdenum carbide, other carbides received less attention, but DFT based calculations on TiC 16 and WC 17 also showed their potential for CO 2 activation. In fact, catalytic tests on TiC, WC, ZrC, NbC, and TaC show CO 2 hydrogenation activity on the respective carbide powders, although with lower conversion compared to Mo 2 C. 15 With this promising catalytic activity in mind, a recent comparative DFT study on the most stable (001) surfaces of rocksalt crystal structure and non-magnetic character TiC, ZrC, HfC (group 4), NbC, TaC (group 5), and δ-MoC (group 6) reported strong CO 2 adsorption and activation. 18 In more detail, strength of CO 2 activation was found to vary depending on the transition metal, but always possible on two different adsorption sites exposed on the (001) surface, either in a three-fold hollow position neighboring two metal and one carbon surface metal atom (MMC) or on top of a surface C atom (TopC).
The above mentioned studies demonstrate the potential of TMCs for CO 2 activation, but, apart from evidence coming from DFT calculations, and the implications from catalytic experiments mentioned above, basic surface science experiments on the interaction of CO 2 with well defined single crystal surfaces are a missing piece in this puzzle, although available and well summarized for other materials. 1,2,19,20 Such studies could provide a detailed microscopic picture of CO 2 adsorption by characterizing adsorbed moieties, commonly achieved by combining multiple complementary experimental techniques, 19 where we refer to some illustrative textbook studies here. [21][22][23][24] Still, when trying to interpret such experimental data, a direct comparison to DFT based predictions is of great aid, see for instance combined studies on CO 2 adsorption on Ni, 25 ZnO, 26

COMPUTATIONAL AND EXPERIMENTAL DETAILS
Briefly, to assess the properties of CO 2 adsorbed on the different TMCs, 18 periodic DFT calculations were carried out, using slab models and optimized surface-adsorbate geometries previously obtained. 18 All calculations were carried out using the Vienna Ab Initio Simulation Package -VASP code. 29 The Perdew-Burke-Ernzerhof (PBE) 30  chamber that had capabilities for XPS and low-energy electron diffraction (LEED). The metal carbide substrates were prepared and cleaned as described in references. 35,36 These generated surfaces had a very good LEED pattern and a carbon/metal ratio in the range of 0.96-0.98.
The metal carbide surfaces were exposed to CO 2 at a temperature of 200 K to avoid physisorption of the molecule. Upon heating to 350 K, the CO 2 desorbed from the carbide substrates without any trace of decomposition (i.e. the corresponding C 1s and O 1s spectra were very similar to those measured before adsorption of the CO 2 molecule).

XPS Experiments
We investigated the adsorption of sub-monolayer amounts of CO 2 on TiC, ZrC, and VC (001) surfaces using XPS. The adsorption of the CO 2 molecules was carried out at 200 K, and by 350 K the adsorbate desorbed without any signs of decomposition, see Figure 1 The amount of CO 2 chemisorbed on TiC(001) and ZrC(001), after a dose of 100 Langmuir (L), was close to 0.5 monolayer (ML). In the case of VC(001), the CO 2 coverage was only 0.15-0.2 ML and could be mainly a consequence of interactions with defects or imperfections instead of interaction with flat terraces (see below). For the clean carbides the C 1s peak appeared in the range of 281 to 282 eV in good agreement with XPS data previously reported [35][36][37] The C 1s features for adsorbed CO 2 were centered from 285 to 286 eV. This implies a downward shift of ~4 eV with respect to the C 1s peak of the carbides. The O 1s for adsorbed CO 2 was detected at binding energies between 532 and 533 eV. These XPS results provide direct evidence for CO 2 activation on single crystal (001) surfaces of TiC, ZrC, and VC. We will further compare these results to our theoretical modelling of core-level XPS, see section 3.7.

CO 2 adsorption strength
The experimental results presented above give direct evidence for CO 2  CO 2 are also depicted in Figure 2, here restricted to the case of TiC (001), while being similar for all the carbides, see discussion below.
The case of VC has to be discussed apart. Here, CO 2 activation on MMC and TopC sites seems unfavorable leading to adsorption energies of +0.11 and +0.25 eV, respectively; It still remains noticeable how CO 2 activation on VC is only mildly favorable, whereas TiC and NbC carbides -direct neighbor carbides in the periodic table-activate CO 2 in a highly favored way. The effect can however be understood by decomposing the overall adsorption energy into contributions arising from cost due to geometry changes on (1) the surfaces active sites, and (2) in the CO 2 molecule going together with (3) an attachment energy arising from interaction of both systems at the final geometry. Results given in Table 2 show that the three contemplated TMCs indeed interact strongly with CO 2 , leading to highly favored attachment energies (3), while induced geometry changes (1) and (2) counteract this to a differing extent. Especially for VC, attachment energy (3) is almost cancelled by geometry changes (1) and (2). Still, the interaction between surface and CO 2 is distinct and seems similar to other TMCs, making the occurrence of the metastable MMC adsorption state and the given desorption barrier understandable.
Turning back to adsorbate geometry, see Figure 2, MMC and TopC adsorption slightly differ in their relative orientation against the surface with CO 2 being inclined in an MMC hollow adsorption, while TopC CO 2 is oriented perpendicular over a long bridged adsorption mode. Still, the bonding principles seem rather similar, concerning C↔C and metal↔O interactions between CO 2 and surface sites, to what has been described before. 10,18 We suspect that such similarity could make a clear experimental distinction between these structures difficult, as rather similar electronic properties are likely involved. To ascertain this assumption, we in the following briefly analyze several electronic structure descriptors and next turn to a discussion of experimental techniques that could allow for identification of activated CO 2 and for a distinction between MMC and TopC adsorption.

Electronic structure analysis
To confirm that a charge transfer from the surface to CO 2 leads to activation and a concomitant bending of the adsorbed molecule, a Bader charge-analysis 38 has been carried out. Results in Table 1 show that adsorbed CO 2 is highly activated, receiving a considerable amount of charge ΔQ of around -0.7 and -1.1 e over all cases. The involved charge redistribution can be analyzed in more detail from the CDD plots.
We here focus on the illustrative case of TiC (001)  pair density between surface carbon and the CO 2 carbon atom evidences the newly formed covalent bond between adsorbate and surface. Note here, that projected densities of states are provided below, adding further information to the discussion of electronic structure.

Ultraviolet spectroscopy
Ultraviolet Photoemission Spectroscopies (UPS) have been conveniently used to deduce information about molecular adsorbates on surfaces. 40  assuming final state effects on energy levels to be negligible. 44 We therefore here provide the local density of states of the systems of interest, discussing first the representative example of CO 2 adsorption on TiC (001), see Figure 5. Note that we focus here on CO 2 induced changes, a thorough discussion of the bare surface DOS is provided in Ref. 39. Red traces in Figure 5 specify the contribution of CO 2 related states. Labels were assigned by visual inspection of the partial charge density in the respective energy range.
For physisorption on the bare surface, the DOS can merely be understood as a combination of bare surface states and states of gas-phase CO 2 in D ∞h symmetry, as was expected. CO 2 activation on MMC or TopC changes this situation distinctly: The wealth of CO 2 related states are then better understandable by looking at the states of a bent CO 2 in gasphase of C 2v symmetry, reported in Figure 5 as well. A significant energetic shift relative to the respective gas-phase orbitals is however seen for the 6a 1 /1b 1 labeled orbitals. Note that this labeling was used as unambiguous assignment of 6a 1 and 1b 1 is difficult, due to significant hybridization of these orbitals with surface states. This hybridization serves well as an explanation for the strong and covalent bonding of CO 2 to the surface discussed above.
Further, the antibonding 6a 1 orbital is now populated in the adsorbed CO 2 , ultimately the cause for weakened and elongated C-O bonds, 1,19 illustrating nicely the concept of adsorption induced activation.
The DOS for the other studied cases than CO 2 are given in Figure S2 in the Supporting Information. For all cases appearing peaks in the DOS are distinctly different for CO 2 physisorption and activation. UPS should then give good indications for CO 2 activation on the studied systems; it should be noted, however, that peak energy differences between MMC and TopC adsorption are quite small for all cases; a distinction of adsorption sites therefore would be hampered by experimental resolution.

Work function changes
In experimental and theoretical works considering CO 2 activation on Ni 25,43 and Co, 46,47 a ~1.0 eV increase in the work function was induced when exposing the bare substrate to CO 2 . Such an increase is often related with charge transfer from the substrate to CO 2 and by analogy is expectable from CO 2 adsorption on TMC (001) surfaces as well. A mere focus on charge transfer however oversimplifies the picture, given that a strong dependence on the substrate is well known. 48 Therefore, to clarify the situation for TMC (001) surfaces, the difference between naked and CO 2 covered surface (Δϕ = ϕ !/!! ! -ϕ S ) has been calculated and is reported in

Vibrational frequencies
Vibrational frequencies of CO 2 on surfaces can be distinguished from experimental techniques such as high resolution electron energy loss spectroscopy (HREELS) 25,26 or by methods based on IR spectroscopy, see e.g. Refs. 13, 28. Especially, these references show how a combination of DFT predicted vibrational frequencies an experiment offers a clear cut interpretation of these spectra. Therefore, to help in a future peak assignment for CO 2 adsorption on the TiC (001) surface, Figure 6 provides its simulated IR spectra.
In more detail, Figure 6 shows For an experimental detection, the symmetric stretch however seems better suited given its higher predicted relative intensity. For CO 2 adsorption on all other carbides similar observations can be made, while the differences between MMC and TopC can in some cases be even more pronounced, see e.g. the case of -9 -adsorption on TaC. We refer to Figure S1 of the Supporting Information for an overview of the remaining cases.

Simulated Core-level XPS
To interpret XPS spectra of each TMC + CO 2  Before turning to CO 2 adsorption however, a qualitative evaluation of the method performance is interesting for the bare TMC (001) surfaces, possible here by comparing to high-quality experimental data. Surface CLBES for C 1s are provided for all carbides in Table   S1 of the Supporting Information, calculated from the binding energy difference of a first and Addressing now the CLBES expected for CO 2 adsorption, we first discuss the case of the C 1s of CO 2 in the different cases i-iii). These CLBES are given and discussed with reference to a surface carbon atom, a viable choice in an UHV experiment as shown in Figure   1. Note that, in this way, the CLBEs reported are in fact core level binding energy shifts.
Results for all TMCs are listed in Table 3  The experiments presented in Figure 1 imply a shift of ~4 eV with respect to the C 1s peak of  Table 4 and we again turn to CO 2 adsorption on TiC

Simulation of STM images
STM has proven powerful to unambiguously identify CO 2 adsorption geometries on well-defined surfaces, especially when compared to DFT predicted adsorption minima and the derived STM images. Experimental studies on Au nanoclusters, 53 as well as combined studies on Ni (110) 54 and TiO 2 (110) 27 provide illustrative examples. Here, we provide simulated STM images of the bare TMC surfaces, and the ones derived from CO 2 adsorption on MMC and TopC adsorption geometries. For the exemplifying case of TiC (001), images are given in Figure 7 and will be discussed in the following; for all other cases we refer to Figure S3 of the Supporting Information.
To produce well-resolved simulated STM images for each system, different pairs of bias voltages and tunneling currents were systematically tested, arriving at the optimum values given below each image in Figure 7. For most cases, to image bare surfaces a simulated tip bias voltage of ± 0.1 V proved valuable at a constant current of ~1.0 nA. In fact rather similar conditions were used in STM experiments on NbC (001) 55 and VC 0.8 (111), 56 noting that a reversal of bias remained without a noticeable influence on image appearance, as we can affirm from simulated images in most cases. As expected, the TiC (001) bare surface then exhibits the expected ordered pattern with C (Ti) atoms appearing bright (dark).
Simulated images (for ±1.0 V and 1.0 nA) for CO 2 adsorption in both cases clearly show this adsorbed molecule as a bright species, surrounded by dark areas, likely a consequence of charge transfer to the adsorbate. A note of caution however, is that in this formalism, a STM tip is modeled as an infinitely small point source leading to high resolution, possibly diminishing under experimental conditions. Nevertheless, according to these predictions, MMC and TopC adsorption modes of CO 2 would be distinguishable by appearance, given they display one or two symmetry planes, respectively. Determining the lattice directions in an uncovered surface area could simplify the assignment further, given the differing relative orientations between MMC and TopC adsorption.

Linear optical properties
Adsorption of molecules on a materials surface can introduce changes in the linear optical response. 57,58 In more detail, changes in the materials dielectric functions are then induced by adsorption, leading to respective changes e.g. directly measurable by absorption-, reflectance-or in electron energy-loss spectroscopy. To predict changes expected from CO 2 adsorption on the TMC (001) surfaces, we here evaluate the respective linear dielectric response of each model neglecting local field effects. 59  and local field effects could improve the description in this energy range. It is also notable that experimental reflectance spectra show overall higher reflectance than reproduced by calculations on surface slabs, while bulk model predictions often agree with experimental values. Including a larger number of layers in the slab model could therefore remedy this behavior. 59 Note however, that the obtainable spectra heavily depend on sample preparation.
Still, in the following, we focus on the relative changes induced upon CO 2 adsorption, thus keeping the model.
Knowing that the present theoretical model performs reasonably well, qualitatively predicting the linear optical properties of the three bulk carbides, we now assess changes induced upon CO 2 adsorption, focusing here first on TiC (001). In Figure  An overview for CO 2 adsorption on other TMCs is provided by Figures S4-S9 of the Supporting Information.

SUMMARY AND CONCLUSIONS
In summary, a combined theoretical and experimental study of CO 2 adsorption and detection on the stable (001)  With a wealth of simulations carried out on an important example, we expect the present study to provide also a concise exploration into multitechnique simulation of surface adsorption by means of state-of-the-art DFT. A proof of this concept is provided by experimental results for transition metal carbides TiC, ZrC, and VC, lending direct first significant evidence of the previously computationally predicted CO 2 activation on the (001) surfaces.

ASSOCIATED CONTENT
Supporting Information: The Supporting Information is available free of charge on the ACS Publications website at DOI: XXXXXXXXXXX. Further computational details, information on the calculation of charge density differences (CDD), work function changes, the simulation of i) IR spectra, ii) core level binding energies, iii) STM images, iv) linear optical properties.
Computed work functions for bare and CO 2 adsorbed surfaces. Complete set of simulated IR spectra, LDOS, STM images, and linear optical properties.