Abstract
This work addresses a longstanding theoretical discrepancy using Density Functional Theory (DFT) with experimental observations of CO binding trends on electrocatalytically relevant metals for the CO2 reduction reaction (CO2RR). By introducing thermal fluctuations using appropriate statistical mechanical NVT and NPT ensembles, we show that DFT with universal dispersion interactions yields qualitatively better metal surface strain trends and CO binding energetics, consistently predicts the correct site preference for all metals due to thermally induced surface distortions that preferentially exposes the undercoordinated atop site for Cu(111) and Pt(111), and for the weak binding Ag(111) and Au(111) surfaces at finite temperatures shows CO-metal interactions that are a mixture of chemisorbed and physisorbed species. This study better places theory as an equal partner to experimental heterogeneous catalysis by demonstrating the need to fully account for finite temperature fluctuations to make contact with surface science experiments.
Introduction
Density Functional Theory (DFT) has long served as the cornerstone of computational catalysis. However, it faces a persistent challenge known as the "CO puzzle": standard 0 K DFT calculations often predict that carbon monoxide (CO) prefers multi-coordinated sites (e.g., hollow or bridge sites) on transition metal surfaces like Cu(111) and Pt(111). In contrast, experimental surface science consistently identifies the atop site as the preferred adsorption position at low coverage. This discrepancy is particularly problematic for modeling the $CO_2$ reduction reaction ($CO_2RR$), where CO is a central intermediate.
Computational Methodology
To resolve this discrepancy, this study employs Ab Initio Molecular Dynamics (AIMD) to account for finite temperature effects at 300 K. Two functional approaches were compared: the RPBE functional, a standard GGA known for its performance in chemisorption, and the B97M-rV meta-GGA functional, which incorporates universal dispersion interactions. Simulations utilized NVT and NPT ensembles on $6 \times 6$ supercells with six metal layers to accurately capture bulk-like surface dynamics and thermally induced distortions.
Results and Discussion
Thermal Fluctuations and Surface Relaxation
At 300 K, the metal surface is no longer a static lattice. Thermal motion induces significant interlayer relaxation and surface distortions. As shown in Figure 6, the interlayer spacing deviates from the ideal bulk values, with the top layers exhibiting characteristic expansion or contraction depending on the metal and the functional used. These fluctuations are essential for capturing the true physical state of the catalyst under operating conditions.
Resolving the CO Adsorption Site Preference
The study finds that these thermally induced distortions are the primary driver for correcting the CO site preference. By allowing the surface atoms to fluctuate, the undercoordinated atop sites become energetically accessible and ultimately preferred for Cu(111) and Pt(111) when using the B97M-rV functional. This aligns theoretical predictions with experimental observations, effectively solving the CO puzzle by moving beyond the limitations of absolute-zero modeling.
Dispersion and the Mixture State on Ag and Au
For weakly binding metals like Ag(111) and Au(111), dispersion interactions are critical. While standard RPBE predicts negligible binding at 300 K, the B97M-rV functional correctly captures the weak CO-metal interaction. Interestingly, the simulations reveal a dynamic "mixture" state where CO exists in both chemisorbed and physisorbed configurations. This suggests that experimental measurements likely report a thermal average of these distinct states, a nuance missed by static calculations.
Conclusion
This work demonstrates that accounting for finite temperature fluctuations and dispersion interactions is essential for bridging the gap between theory and experiment in electrocatalysis. By moving beyond static 0 K models, DFT can more accurately predict the behavior of complex catalytic interfaces, establishing it as a more reliable partner for experimental discovery in the quest for efficient $CO_2$ reduction technologies.
Summary updated