Parallelized Screening of Characterized and DFT-Modeled Bimetallic Colloidal Cocatalysts for Photocatalytic Hydrogen Evolution
Eric M. Lopato, Emily A. Eikey, Zoe C. Simon, Seoin Back, Kevin Tran, Jacqueline Lewis, Jakub F. Kowalewski, Sadegh Yazdi, John R. Kitchin, Zachary W. Ulissi, Jill E. Millstone and Stefan Bernhard, “Parallelized Screening of Characterized and DFT-Modeled Bimetallic Colloidal Cocatalysts for Photocatalytic Hydrogen Evolution”, ACS Catalysis 10(7), 4244-4252 (2020). https://doi.org/10.1021/acscatal.9b05404
This summary was written by a language model from the paper and has not been checked line by line. Rely on the published paper for details.
The extracted text is the accepted manuscript version (titled with “DFT-modelled” and “Co-catalysts”); figures are not readable in it, so only the text, captions and stated numbers are summarized here.
Question
Bimetallic water reduction catalysts had been studied extensively by computation and in electrocatalysis, but the authors found little work on bimetallic colloids as cocatalysts for visible-light, solution-phase photocatalytic water reduction. They asked whether a fast parallel screening method could identify synergistic metal pairs among mostly untested combinations, and whether the Sabatier principle, as expressed by electrocatalytic hydrogen evolution volcano plots, also governs photocatalytic systems with nanoparticles formed in situ (Lopato et al. 2020).
Approach
The authors built a 96-well photoreactor: an aluminum plate holding 1 mL shell vials, illuminated from below by two water-cooled 100 W blue LEDs (440 to 450 nm), and sealed with a commercial H2-sensitive chemochromic tape under FEP film, silicone and a clamped plexiglass cover. Photographs taken from above track the darkening of the tape over each vial. Calibration by injecting known H2 amounts gave a linear relation between normalized intensity change and H2 content. A plate of identical reactions gave a standard error of 9.8% in maximum H2 and 12.3% in maximum rate. Reactions ran in air for 750 minutes with 0.25 mM of the iridium photosensitizer [Ir(Fmppy)2dtbbpy]PF6, triethanolamine as sacrificial donor, and a 10:1 DMSO to aqueous TEOA mixture (440 µL). Metal chloride salt concentrations were varied from 0 to 0.65 mM across rows and columns for eight pairs: Pd/Sn, Pd/Mo, Pd/Ru, Pd/Pb, Pd/Ni, Ni/Sn, Mo/Sn and Pt/Sn. A Mathematica image pipeline extracted kinetic traces, and the maximum rate came from a smoothed numerical derivative.
Synergy was quantified as a “combined activity”: the bimetallic maximum rate minus the sum of the single-metal rates at the same concentrations, with each single metal described by a fourth-order polynomial fit. Positive values mean synergy and negative values antagonism. The most active Pd/Sn particles were characterized by TEM, HAADF-STEM with EDS mapping and linescans, and XPS. On the computational side, Pd-Sn intermetallics from the Materials Project were modeled with VASP and RPBE, using all unique surfaces with Miller indices up to 2 and all unique H sites; a machine learning model predicted H binding where DFT was impractical. Binding energies were converted to free energies with a constant 0.24 eV correction.
Key findings
More than 700 bimetallic water reduction catalysts were tested. Pd alone peaked between 0.1 and 0.2 mM, where it outpaced even Pt, and lost activity at higher concentration, while Sn alone was nearly inactive. Clear synergy appeared for Pd/Sn, Pd/Mo, Pd/Ru and Pt/Sn, with Pd/Sn and Pt/Sn giving the highest combined activities. The strongest synergy for Pd/Sn occurred at only 7.7% Pd, versus 33% Pt for Pt/Sn, which made Pd/Sn the focus. Adding Sn kept the rate high above 0.4 mM Pd, where Pd alone declines; the absolute maximum rates lay along a roughly 1:1 ridge, peaking at 0.2 mM Pd with 0.2 mM Sn (and 0.3/0.3 mM), and rate versus Pd mole fraction followed a quadratic peaking at 0.5.
The 0.2/0.2 mM sample formed pseudospherical particles of 4.7 ± 0.8 nm in situ, and EDS maps showed Pd and Sn within the same particles. XPS showed Sn 3d peaks shifted to lower and Pd 3d peaks to higher binding energy in mixed samples relative to the single metals (Sn alone gave mainly SnOx), consistent with Pd-Sn mixing.
DFT showed H binding strengthening with Pd fraction. The strongest binding on PdSn was 0.6 eV stronger than on Sn2Pd, which the authors link to new Pd3 fcc hollow sites, and higher Pd fractions bound H too strongly. Most Sn surfaces bound H weakly (about 0.5 eV), with some α-Sn sites near 0.1 eV. A linear fit of the strongest binding free energy against Pd fraction (R2 = 0.90) crosses 0 eV near a Pd fraction of 0.5. Plotting measured rates against this composition-weighted free energy gave a volcano-like parabola (R2 = 0.98) peaking at -0.036 eV, close to the 0 eV electrocatalytic optimum. The authors note that other explanations are possible.
Significance
The paper introduced the parallel photoreactor and colorimetric H2 detection used in later work from these groups, and showed that bimetallic colloids can beat pure Pd or Pt while using much less precious metal. Linking experimental photocatalytic rates to DFT hydrogen adsorption free energies suggests that electrocatalytic HER descriptors can guide screening of in situ formed photocatalytic cocatalysts, pointing toward computationally guided, abbreviated searches.