Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Practical and improved density functionals for computational catalysis on metal surfaces

Benjamin X. Shi, Timothy Berkelbach

Abstract

Density functional theory (DFT) has been critical towards our current atomistic understanding of catalysis on transition-metal surfaces. It has opened new paradigms in rational catalyst design, predicting fundamental properties, like the adsorption energy and reaction barriers, linked to catalytic performance. However, such applications depend sensitively on the predictive accuracy of DFT and achieving experimental-level reliability, so-called transition-metal chemical accuracy (13 kJ/mol), remains challenging for present density functional approximations (DFAs) or even methods beyond DFT. We introduce a new framework for designing DFAs tailored towards studying molecules adsorbed on transition-metal surfaces, building upon recent developments in non-self-consistent DFAs. We propose two functionals within this framework, demonstrating that transition-metal chemical accuracy can be achieved across a diverse set of 39 adsorption reactions while delivering consistent performance for 17 barrier heights. Moreover, we show that these non-self-consistent DFAs address qualitative failures that challenge current self-consistent DFAs, such as CO adsorption on Pt(111) and graphene on Ni(111). The resulting functionals are computationally practical and compatible with existing DFT codes, with scripts and workflows provided to use them. Looking ahead, this framework establishes a path toward developing more accurate and sophisticated DFAs for computational heterogeneous catalysis and beyond.

Practical and improved density functionals for computational catalysis on metal surfaces

Abstract

Density functional theory (DFT) has been critical towards our current atomistic understanding of catalysis on transition-metal surfaces. It has opened new paradigms in rational catalyst design, predicting fundamental properties, like the adsorption energy and reaction barriers, linked to catalytic performance. However, such applications depend sensitively on the predictive accuracy of DFT and achieving experimental-level reliability, so-called transition-metal chemical accuracy (13 kJ/mol), remains challenging for present density functional approximations (DFAs) or even methods beyond DFT. We introduce a new framework for designing DFAs tailored towards studying molecules adsorbed on transition-metal surfaces, building upon recent developments in non-self-consistent DFAs. We propose two functionals within this framework, demonstrating that transition-metal chemical accuracy can be achieved across a diverse set of 39 adsorption reactions while delivering consistent performance for 17 barrier heights. Moreover, we show that these non-self-consistent DFAs address qualitative failures that challenge current self-consistent DFAs, such as CO adsorption on Pt(111) and graphene on Ni(111). The resulting functionals are computationally practical and compatible with existing DFT codes, with scripts and workflows provided to use them. Looking ahead, this framework establishes a path toward developing more accurate and sophisticated DFAs for computational heterogeneous catalysis and beyond.
Paper Structure (12 sections, 6 equations, 3 figures)

This paper contains 12 sections, 6 equations, 3 figures.

Table of Contents

  1. Introduction
  2. Results
  3. Accurate adsorption energy predictions
  4. Reliable description of CO and graphene adsorption
  5. Towards right answers for the right reasons on surface barrier heights
  6. Discussion
  7. Conclusions
  8. Methods
  9. Non-self-consistent density functional approximations for metals
  10. Modeling procedure for adsorption energy and barrier heights
  11. Density functional theory calculation details
  12. Gas-phase random phase approximation calculation details

Figures (3)

  • Figure 1: Accurate and balanced performance for adsorption on transition metal surfaces. Comparison of hBEEF-vdW@BEEF-vdW and dhBEEF-vdW@BEEF-vdW against common (self-consistent) density functional approximations for the adsorptions energies of the CE39 dataset wellendorffBenchmarkDatabaseAdsorption2015. We give mean absolute deviations (per product formed) w.r.t. experimental values from Ref. wellendorffBenchmarkDatabaseAdsorption2015 for the (a) chemisorption and (b) physisorption subsets. We also compare the (c) total performance and (d) show the correlation between the physisorption and chemisorption subset MADs.
  • Figure 2: Overcoming qualitative failures of self-consistent DFAs for adsorption. Comparison of hBEEF-vdW@BEEF-vdW and dhBEEF-vdW@BEEF-vdW against common (self-consistent) density functional approximations for (a) CO adsorption on Pt(111), Rh(111) and Cu(111) and (b) graphene adsorption on Ni(111). We give the relative energy between the (on-)top and hollow (fcc) site in (a), with a negative value favoring the top site. The experimental value for the adsorption energy and distance in (b) were taken from Ref. janthonTheoreticalAssessmentGraphenemetal2013.
  • Figure 3: Improved barrier heights for the right reasons on transition metal surfaces. Comparison of hBEEF-vdW@BEEF-vdW and dhBEEF-vdW@BEEF-vdW against common (self-consistent) density functional approximations for the (a) SBH17 dataset tchakouaSBH17BenchmarkDatabase2023. We provide mean absolute deviations (MADs) against experimental values for the total SBH17 dataset as well as its CH$_4$ [visualized in (b)], H$_2$ and N$_2$ dissociation subsets. Further comparison is given for (c) gas-phase barrier heights (NBH58; the neutral subset of the BH46 and DBH22 datasets in Ref. liangGoldStandardChemicalDatabase2025), (d) gas-phase dimer non-covalent interactions (S66 dataset rezacS66WellbalancedDatabase2011) and (e) metal surface energies (SE20 dataset) from Ref. lundgaardMBEEFvdWRobustFitting2016.