First-Principles Modeling of Catalysts: Novel...
Transcript of First-Principles Modeling of Catalysts: Novel...
CH3ReO3 / H2O2
15 °C, CH3CN(aq)
First-Principles Modeling of Catalysts: Novel Algorithms and Reaction Mechanisms Bryan R. Goldsmith
Fritz Haber Institute of the Max Planck Society, Theory Department
Research highlights
Methods B. Data analytics tools to find patterns and descriptors
Pollution abatment Catalyst stability and dynamics under reaction conditions
Methods A. First-principles modeling and molecular simulation techniques
Theme 1. Amorphous catalytic solids: beyond ordered materials
Theme 2. Metal oxides and MOFs as catalysts and catalyst supports
Amorphous catalysts are widely used in industry and are often superior to crystalline catalysts Catalysis is prerequisite for more than 20% of all production in the industrial world
Theme 3. Homogeneous catalysis for specialty chemical production
Subgroup Discovery find physically interpretable local models
of a target property in materials data
[1] Descriptive features, 𝑎1, … ,𝑎𝑚 ∈ 𝐴
e.g., energy, bonding topology, number of atoms
[2] Target features, 𝑦1, … ,𝑦𝑛 ∈ 𝑌
e.g., HOMO-LUMO gap
[3] Basic selectors, 𝑐1, … , 𝑐𝑘 ∈ 𝐶 → {false, true}
e.g., Is the band gap greater than 1 eV?
[4] Find selector 𝜎 = 𝑐1 ⋅ ∧ ⋯∧ 𝑐𝑙 ⋅
that maximizes quality 𝑞 = ext 𝜎𝑃
𝛼 𝑢(𝑌𝜎)1−𝛼
ext 𝜎𝑃
is the coverage of points where 𝜎 is true
𝑢 𝑌𝜎 is the utility function (optimization criteria)
Density functional theory Minima and saddle search algorithms
Slab models Cluster models
Supported nanoparticle deactivation and disintegration
Structure-property analysis of nanoclusters
Amorphous oxide catalysts Kinetic characterization of homogeneous organometallic catalysts
Homogeneous catalysis by organometallic complexes is critical in the specialty chemical industry Data analytics tools applied to big-data of materials offers opportunities to accelerate materials discovery
Formation of nanoparticles from molecular precursors
Gas-phase gold clusters
[Cu25H22(PPh3)12]Cl
Distribution of silanol groups as a function of temperature (PH2O = 10−6
bar) predicted by the model (points) and compared with experimental values (lines). Silanols are defined as: vicinal silanols, which exhibit a hydrogen-bonded OH group, including geminal and non-geminal; isolated single silanols, ≡SiOH; free geminal silanediols, =Si(OH)2.[2]
Ab initio thermodynamics and rate theories
Structure exploration e.g., replica-exchange molecular dynamics Free energy estimates e.g., multistate bennett acceptance ratio
[1]. B. Qiao et al. Nat. Chem. 634 (2011) [2]. G. Vayssilov et al. Angew. Chem. Int. Ed. 42 (2003) [3]. Y-G. Wang et al. Nat. Commun. 6 (2015) [4]. M. Eddaoudi et al., Science 295 (2002) [5]. J. C. Matsubu et al., J. Am. Chem. Soc. 137 (2015) [6]. L. M. Ghiringhelli et al., Phys. Rev. Lett. 114 (2015)
[1]. G-J Cheng et al., J. Am. Chem. Soc. 136 (2014) [2]. L. Ackermann and J. Li, Nat. Chem. 7 (2015) [3]. X. Zhang et al., Acc. Chem. Res. 49 (2016) [4]. H. Hu and W. Yang, Annu. Rev. Phys. Chem. 59 (2008)
STM images during high pressure ethylene hydrogenation on Rh(111) for: (a) 20 mtorr H2 and 20 mtorr C2H4 and (b) 20 mtorr H2, 20 mtorr C2H4 and 5 mtorr CO. The formation of an ordered adsorbate structure caused by CO coadsorption with ethylidyne inhibits catalytic activity. The images were recorded at 298 K.[1] Metal-Organic Frameworks as Catalysts and Supports
C-H Bond Activation of Arenes by Organometallic Complexes
Amorphous Oxides as Catalyst Supports: Development of Improved Models
Target: HOMO-LUMO energy gap
Apply subgroup discovery to examine 24 400 neutral gas-phase gold cluster configurations (of sizes 5-14 atoms)
Elucidate how the heterogeneity of amorphous oxide surfaces influences the properties of supported catalysts
Some catalysts which are initially crystalline become amorphous under reaction conditions
WOx supported on SiO2 for the formation of propene from ethene and butene
Single Atom and Nanocluster Catalysis
The research objectives are to increase understanding of catalysis by metal ions and metal clusters dispersed on amorphous supports like silica and silica-alumina, as well as to develop a systematic framework for modeling amorphous catalysts
The objectives of this research are to understand single atoms, nanoclusters, and nanoparticles dispersed on metal oxides and metal-organic frameworks for catalysis, and to engender a framework for their use in sustainable chemical applications
Target: intra-cluster van der Waals energy
The main aims of this research are to investigate organometallic-catalyzed C-H bond activation for specialty chemical production, and to help develop a framework for the accurate modeling of thermodynamics and kinetics for solution-phase reactions
[1]. G. A. Somorjai and M. Yang, Top. Catal. 24 (2003) [2]. C. Ewing et al., Langmuir 30 (2014) [3]. B. Peters and S. L. Scott, J. Chem. Phys. 142 (2015)
The nature of the active site(s) of WOx/SiO2 is still a matter of debate WO3 crystallites can be ruled out as the catalytically active species
What (meta)stable structures do Rh nanoclusters adopt under CO2 + H2 reaction conditions? What are the size-dependent reaction pathways? Why does cluster size influence the selectivity of catalytic methanation relative to rWGSR?
Extract descriptors for molecule adsorption strength in doped isoreticular metal-organic frameworks (IRMOFs)
Illustrations of various classes of heterogeneous catalysts and catalyst supports that will be considered in my research
Single atoms Nanoclusters Nanoparticles Metal-organic frameworks
Distribution of silica sites
Supported nanoclusters can disintegrate to form smaller clusters and single atoms that coexist simultaneously
Mechanistic Hypothesis Testing and Benchmarking Theory with Experiment
Zeolite cluster RuO2(110) slab model
Plausible mechanisms for aromatic C−H activation (a) Electrophilic Aromatic Substitution (b) Heck Type (c) Oxidative Addition (d) Concerted Metalation Deprotonation
Reverse water gas shift reaction
Catalytic methanation
Molecules considered: Alkenes, alkynes, H2, CO, CO2, H2S, NH3, H2O
Dopants considered: Mg2+, Ti2+, V2+, Mn2+, Fe2+, Co2+, Cu2+, Zn2+
[5]
T. Sperger et al., Chem. Rev. 115 (2015)
Meta-selective C-H activation of arenes[1]
Find template and linker design variables for high activity and selectivity by applying subgroup discovery to data generated by DFT
In silico template and linker design for meta-selective C-H activation of arenes
Apply LASSO+𝑙0 to find a low dimensional and physically meaningful descriptor[6]
Descriptive features: electronegativity, radii of s, p, and d orbitals of the dopants and adsorbates, chemical hardness, ionization potential, electron affinity… and physically meaningful linear & nonlinear combinations.
MOFs considered: IRMOF-n (n = 1-7)
Subgroups →
2015 – current, Humboldt Postdoctoral Fellow, FHI (Matthias Scheffler) | 2010 – 2015, PhD Chemical Engineering, UCSB (Baron Peters)
A detailed understanding of catalysts and materials requires an accurate description of their electronic and geometrical properties under realistic conditions
B. R. Goldsmith et al., J. Am. Chem. Soc. 137 (2015) T. Hwang,* B. R. Goldsmith* et al., Inorg. Chem. 52 (2013)
B. R. Goldsmith et al., J. Chem. Phys. 138 (2013) B. R. Goldsmith et al., In Reaction Rate Constant Computations, RSC (2013) B. R. Goldsmith et al., (Invited Perspective, ACS Catal.) (2017)
B. R. Goldsmith et al., J. Phys. Chem. C 118 (2014)
T-A. Nguyen, Z. Jones, B. R. Goldsmith et al., J. Am. Chem. Soc. 137 (2015); T-A. Nguyen, B. R. Goldsmith et al., Chem. Eur. J. 21 (2015); B. R. Goldsmith et al., to be submitted
Classify 82 octet binary semiconductors as either rocksalt (RS) or zincblende (ZB)
Rocksalt (RS) Zincblende (ZB) Find relations between geometrical and electronic properties
Subgroup discovery methodology
�̂�𝐿𝐴𝐿𝐿𝐿(λ) = argmin𝛽
12
𝑦 − 𝑫𝛽 22 + λ 𝛽 1
𝑙1-norm: Sum of absolute
value of coefficients
Root mean squared error Regularization
parameter
y = target property = ∆Eads D = feature matrix β = coefficients
For vitreous SiO2
[1]. J. P. Perdew and K. Schmidt, AIP Conf. Proc., 577 (2001) [2]. M. K. Sabbe et al., Catal. Sci. Tech. 2 (2012) [3]. B. R. Goldsmith et al., to be submitted [4]. K. Reuter and M. Scheffler, Phys. Rev. B 65 (2002) [5]. K. Reuter et al., Phys. Rev. Lett. 93 (2004) [6]. B. Peters, J. Phys. Chem. B 119 (2015)
[1-2]
QM/MM and Implicit Solvent
CH3Cl + Cl−
gas phase
explicit water solvent with QM/MM[4]
Nudged elastic band; Cerjan-miller; Dimer method; BFGS
Rh(NO)2 / TiO2(110)
Model the system at realistic chemical potentials
Reaction rates
vs
The primary goal of this research is to develop and apply data analytics tools to find materials-science insights Here we focus on advancing a local pattern discovery algorithm called subgroup discovery
B. R. Goldsmith,* C. Gardner* et al., in prep. Special thanks: Susannah L. Scott, Trevor Hayton, Wei-Xue Li, Luca M. Ghiringhelli, and Runhai Ouyang
[2]
𝐸{𝐑𝐼} 𝑛 = 𝑇𝑠 𝑛 + �𝑑3𝑟 𝑣 𝐑𝐼nuc 𝐫 𝑛 𝐫
+ 12�𝑑3𝑟𝑑3�́�
𝑛 𝐫 𝑛 �́�𝐫 − �́�
+ 𝐸xc 𝑛
200 K
Ρ = 68%
26%
1%
6%
[3]
‘Real system’ Model system
[3] Vicinal Geminal Isolated Bridging
[1] [2] [3] [4] Zn4(O)O12C6 clusters benzene links
[4]
[5]
[6] Big-data analytics toolkit and tutorials, https://www.nomad-coe.eu/
Ab initio molecular dynamics
Atomically Dispersed Catalysts on Amorphous Supports
J. G. Howell et al., ACS Catal. 6 (2016)