(445f) Mechanistic Details of Formic Acid Dehydration on TiO2 and ZrO2 Catalysts

Formic acid (HCOOH) decomposes via dehydration and dehydrogenation paths at relative rates that depend on the nature of the catalysts and the reaction conditions. Such reactions serve as attractive probes of the nature of active sites on solids and of the most appropriate descriptors of reactivity and selectivity for the two decomposition modes.¹ Such routes are also of fundamental interest for reactions that are mediated by formate-type intermediates, such as water-gas shift² and methanol synthesis.³ Metal oxides, such as TiO₂ and ZrO₂, selectively dehydrate HCOOH to form CO and H₂O at mild temperatures (< 473 K). Mechanistic details of such routes and the nature of active centers, however, have remained controversial due to a lack of detailed kinetic, isotopic, and spectroscopic data at relevant reaction conditions and of theoretical assessments that confirm the nature of active centers and the elementary steps. Here, kinetic, in-situ infrared (IR), and isotopic studies are combined with density functional theory (DFT) to determine the identity and kinetic relevance of the elementary steps that mediate HCOOH dehydration routes on TiO₂ and ZrO₂ catalysts and to provide more accurate descriptors of acid-base properties of metal oxide catalysts and of binding properties that determine reactivity.

Dehydration rates (1.5 kPa HCOOH; 433 K) were similar on anatase TiO₂ (a-TiO₂) samples that were treated in He, H₂, or O₂ at 723 K before reaction, indicating that reduced centers, often proposed as the active centers for dehydration reactions on TiO₂,⁴ are either absent or catalytically inconsequential. Dehydration rates increased with HCOOH pressure (< 1 kPa), but reached a constant value as TiO₂ surfaces became saturated with HCOOH-derived species. In-situ IR spectra and DFT-derived adsorption energies suggest that such species are either molecularly adsorbed HCOOH or monodentate formate species, which become stable at high coverages (> 0.5 ML) via interactions between the adsorbed species. Bidentate formates were also detected from IR, but their decomposition rates were much slower than dehydration turnovers, indicating that such species act as spectators. The intensity of the band assigned as an asymmetric CO₂^- vibration of bidentate formate species (1554 cm^-1) was insensitive to HCOOH pressure (0.1-2.5 kPa HCOOH; 433 K), indicating that such species present at near-saturation coverages at all conditions. These trends are inconsistent with dehydration rates that increase with HCOOH pressure (< 1 kPa), suggesting that bidentate formates are formed at sites that are irrelevant to catalysis, such as rutile that may be present as a minority crystalline phase in the a-TiO₂ sample. Dehydration rates were unaffected by CO pressure (0.2- 1 kPa; 433 K), but H₂O inhibits rates (0.1-1.5 kPa; 423-453 K) by competitive adsorption with HCOOH on Ti-O pairs. DCOOH reactants led to a strong H/D kinetic isotope effect (k_H/k_D = 2.7; 433 K), but the rates were similar for HCOOD and HCOOH. DFT-derived kinetically relevant transition state (TS) involves the concurrent activation of the C-O and the C-H bond in molecularly-adsorbed HCOOH on Ti-O pairs that act as Lewis acid-base centers; this step forms CO(g) and vicinal Ti-OH and O-H species that recombine as H₂O in a subsequent elementary step to re-form Ti-O pairs. DFT-derived activation barriers and the kinetic isotopic effects calculated on Ti-O pairs in a-TiO₂(101) were in a quantitative agreement with measured values.

Dehydration turnover rates (per Ti-O pair; 1.5 kPa HCOOH; 433 K) were much lower on rutile TiO₂ (r-TiO₂; 0.94 ks^-1) than on a-TiO₂ (34 ks^-1), despite the stronger Lewis acid strength of Ti-centers in r-TiO₂ than those in a-TiO₂, estimated by DFT-derived OH^- binding energies.⁵ IR spectra measured during HCOOH dehydration on r-TiO₂ indicated the presence of the bidentate formates at near-saturation coverages. Such trends are consistent with DFT-derived adsorption energies of bidentate formates (ΔE_ads = -195 kJ mol^-1) that are much favorable than that of molecularly bound HCOOH or of monodentate formates (ΔE_ads = -124 and -56 kJ mol^-1, respectively) on r-TiO_2. Bidentate formates are much stable on r-TiO₂ than that on a-TiO₂ (ΔE_ads = -129 kJ mol^-1) because of the stronger acid strength of Ti and of the M-O distances in rutile that provide an effective H-bonding between bidenate formate and neighboring OH groups. Such strongly bound species on r-TiO₂ require a significantly higher activation barrier (ΔE^‡= 211 kJ mol^-1) than that of loosely bound species on a-TiO₂ (ΔE^‡= 157 kJ mol^-1), when the barriers are referenced to the relevant adsorbed species, resulting much lower dehydration rates on r-TiO₂ than on a-TiO₂.

Monoclinic and tetrahedral ZrO₂ catalysts (m-ZrO₂ and t-ZrO₂) exhibited much lower dehydration turnover rates per acid-base pair (0.39 and 0.33 ks^-1, respectively) than TiO₂ catalysts (34 and 0.94 ks^-1 for a-TiO₂ and r-TiO₂) at a given condition (1.5 kPa HCOOH; 433 K). These reactivity trends are consistent with much weaker acid strength of Zr than of Ti centers, as indicated by DFT-derived OH^- binding energies that are much favorable on Ti centers in a-TiO₂ (-246 kJ mol^-1) than on Zr in m-ZrO₂ (-37 kJ mol^-1).⁶ DFT-derived H⁺ addition energies are more favorable on O-atoms in m-ZrO₂ (-1616 kJ mol^-1) than in a-TiO₂ (-1175 kJ mol^-1).⁵ Such basic O-atoms on m-ZrO₂ abstract the H-atom from HCOOH via O-H activation and form unreactive bidentate formate species on the surface, leading to deactivation of the ZrO₂ catalysts over time-on-stream. This study provides mechanistic details of HCOOH dehydration steps on TiO₂ and rigorous analyses on how oxides with intermediate acid-base strength and with specific M-O distances achieve high dehydration turnover rates by stabilizing the TS relative to the relevant precursor.

The authors acknowledge the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences (DE-AC05-76RL0-1830) for financial support and the Environmental Molecular Science Laboratory (EMSL) at Pacific Northwest National Laboratory (PNNL) for computational time.

References

(1) Uemura, Y.; Taniike, T.; Tada, M.; Morikawa, Y.; Iwasawa, Y. J. Phys. Chem. C 2007, 111, 16379.

(2) Ueno, A.; Onishi, T.; Tamaru, K. Trans. Faraday Soc. 1970, 66, 756.

(3) Robbins, J. L.; Iglesia, E.; Kelkar, C. P.; DeRites, B. Catal. Letters 1991, 10, 1.

(4) Aizawa, M.; Morikawa, Y.; Namai, Y.; Morikawa, H.; Iwasawa, Y. J. Phys. Chem. B 2005, 109, 18831.

(5) Wang, S.; Iglesia, E. J. Catal. 2017, 345, 183.

(6) Wang, S.; Iglesia, E. J. Phys. Chem. C 2016, 120, 21589.