Schuit Institute of Catalysis

1
Schuit Institute of Catalysis
Applications of computational chemistry to
electrocatalysis
Marc Koper Berkeley August 30 - September 1, 2001
2
Methods of computational chemistry

• Quantum chemistry
• adsorbate-substrate bonding
• ab initio molecular dynamics
• Molecular dynamics
• electric double layer
• electron transfer reactions
• Monte Carlo simulations
• Simulation of adsorbate isotherms, ordering
• Electrocatalytic reactions

3
Typical time and system scales

• Density functional theory (DFT) calculations and
• ab initio molecular dynamics
• 10-100 atoms, 1-5 ps
• Molecular dynamics
• 1000-10,000 atoms, several ns
• Monte Carlo simulations
• 1,000,000 particles, seconds - minutes

4
Slab geometry
z
?3x?3 unit cell 4 metal layers 6 equivalent
layers of vacuum
Pt2Ru
PtRu2
x,y
5
CO on different PtRu surfaces

• 1/3 ML CO adsorption
• Pt(111)
• Ru(0001)
• RuML/Pt(111)
• PtML/Ru(0001)
• Pt-Ru alloys with
• 12 and 21 mixing ratio
• Pt-Ru alloys with
• Pt or Ru monolayer

6
CO on homogeneous Pt-Ru surface
7
CO on PtML/PtRu

• Mixing Pt with Ru weakens CO bond to the Pt
  surface.
• There is no relation between the binding energy
  and C-O vibration!
• Calculations suggest correlation with Pt-C
  vibration.

8
CO on RuML/PtRu

• Mixing Ru with Pt gives a slightly stronger CO
  bond to the Ru sites.
• There is no relation between the binding energy
  and C-O vibration!
• Calculations suggest correlation with Ru-C
  vibration.

9
Interpretation of alloying effect

• Hammer-Norskov d-band shift model.
• Mixing Pt with Ru leads to a lowering of the
  center
• of the d band on Pt due to transfer of d
  electrons to
• Ru. This leads to a weaker back donation and a
• weaker CO chemisorption bond.
• B.Hammer, J.K.Norskov, Adv.Catalysis (2001)
• Bond energy conservation model
• Ru-Ru and Pt-Ru bonds are stronger than
• Pt-Pt bond.
• Hence, mixing with Ru leads to a weaker Pt-CO
• bond if bond energy conservation principle
  applies.

10
Be careful with IR spectroscopy!

• Lin et al. (J.Phys.Chem., 2000) ?
  Our results
• ?Ru(0001) 2005 cm-1
  ?Ru(0001) 1979 cm-1
• ?RuML/Pt(111) 2012 cm-1
  ?RuML/Pt(111) 2007 cm-1
• ?Pt(111) 2060 cm-1
  ?Pt(111) 2050 cm-1
• Conventional interpretation higher C-O frequency
  implies weaker back donation and hence weaker
  bond on Ru/Pt(111) than on Ru(0001).
• However the CO bond to Ru/Pt(111) is the
  strongest bond
• of all Pt-Ru surfaces considered.

11
CO and OH on Pt-Ru
Two groups - high binding energy
(coordination to Ru site) - weak binding
energy (coordination to Pt site) for both
adsorbates
12
Pt-Ru conclusions

• Mixing with Ru weakens CO bond to Pt
• Mixing with Pt strengthens CO bond to Ru
• Sites that bind CO strongly also bind OH
  strongly
• PtML/Ru is the surface with the weakest CO bond
• Both Pt and Ru sites must available at the
  surface
• to have sites that bind CO weakly and OH
  strongly
• Pt-Ru is not a prototype bifunctional catalyst

M.T.M.Koper, T.E.Shubina, R.A.van Santen,
submitted
13
CO and OH on PtMo
PtMo is somewhat similar to PtRu, but Mo is not
such a strong CO binder
T.E.Shubina, M.T.M.Koper, in preparation
14
CO and OH on Pt3Sn(111)
Pt
Sn
CO -1.37 OH -2.32
CO -1.67 OH -2.28
CO -0.42 OH -2.01
CO 0 OH -2.24
CO -0.02 OH -2.24
CO -1.86 OH -1.74
CO does not bind to Sn, but OH binds equally
strong to both Pt and Sn
15
Field-dependent CO chemisorption
16
Outer-Sphere Electron Transfer
Oxn e- Red(n-1)
Sequence of events 1. The reactant moves close
to the electrode surface, but does not
adsorb (outer Helmholtz plane, say) 2. The
solvent assumes a suitable intermediate
non-equilibrium configuration (the transition
state) 3. The electron is exchanged
radiationless 4. The system (solvent) relaxes to
its new equilibrium configuration
Oxn
e-
Red(n-1)
17
The Marcus Potential Energy Surface
transition state
free energy
1. Minima at q n-1 (Red) and q n (Ox
e-) because these are the equilibrium
solvent configurations. 2. Deviations from
equilibrium are assumed to be harmonic
VRed and VOxe are parabolic in q. 3. VOxe
can be shifted up and down by changing the
electrode potential.
VRed
VOxe
n-1
n
generalized solvent coordinate q
18
The solvent reorganization energy l
l is the difference in energy between a
non-equilibrium Ox species with a Red solvation
shell and an Ox species with its proper
equilibrium solvation shell, taking into account
only the slow modes of solvation (i.e. the
electronic polarization is always equilibrated)
free energy
Red
Ox
l
q
19
Testing the Marcus theory
Generate potential energy surfaces by Molecular
dynamics simulations
20
Nonlinear solvent reorganization

• Free energy surface for neutral species not
  parabolic
• Reorganization energy is charge- (or DE)
  dependent

C.Hartnig, M.T.M.Koper, J.Chem.Phys., in press
21
Comparing Marcus and MD

• Strongest changes in electrostriction from q0
  to -1
• Dielectric saturation from qgt1

22
Charge-dependent electrostriction
Cl-O radial distribution function
Cl-H radial distr. function
23
The Anderson-Newns Hamiltonian describes the
exchange of an electron between an
isolated orbital (the adsorbate) and a continuum
of levels (metal). Electronic part Helec
ea na Sk ek nk SkVak cack Vka
ckca Solvent part Hsolv lq2 2l(z - na)q
e
metal
adsorbate
ek
ea
24
The electronic interaction parameter D D
2pSkVak2d(e-ek)
electronic energy e
solution
metal
eF
ea
D
density of states
D describes the broadening of the adsorbate
energy level due to electron exchange.
25
Concerted bond breaking and electron transfer
e.g. methylchloride reduction CH3Cl e-
CH3 Cl- The methylchloride does not
adsorb onto the metal electrode
R
X
e-
R
X-
J.M.Saveant, J.Am.Chem.Soc. 109 (1987) 6788
26
A Hamiltonian for adiabatic bond breaking ET H
Helec Hsolv Hbond-breaking Hbond-breaking is
modeled by a kind of switching function Hbond-br
eaking 1-na VR-X na VRX- where na is
the number operator of the antibonding LUMO
orbital of the R-X molecule
r, distance between R and X
M.T.M.Koper, G.A.Voth, Chem.Phys.Lett. 282 (1998)
100
27
Potential energy surface for bond breaking
Small D, weak electronic interaction
28
Activation energy of bond breaking
(lDe-h)2
DGact
4(lDe)
transfer coefficient a - -
1
h
2
2(lDe)
amount of charge transferred to the antibonding
orbital
29
MD simulation of bond breaking ET methylchloride
reduction at a Pt(111) electrode
A.Calhoun, M.T.M.Koper, G.A.Voth, J.Phys.Chem.B
103 (1999) 3442
30
Adsorption of molecules backdonation the metal
donates electronic charge to the antibonding
orbital leading to a weakening of the
intramolecular bond.
de-
Large D, strong electronic interaction
31
From DFT to macroscopic behavior Monte Carlo
simulations

• Study of properties of a mechanism from the
• elementary steps and interactions
• role of island formation and surface diffusion
• role of substrate structure and composition
• role of lateral interactions and ordered
  overlayers
• To go beyond the mean-field approximation in
• which reaction rates are expressed in average
• coverages.

32
Dynamic Monte Carlo simulations
A Dynamic Monte Carlo simulation samples
the real-time behavior of a system, based on a
kinetic Master equation
CARLOS code written by Johan Lukkien
http//wwwpa.win.tue.nl/johanl/carlos
33
CO oxidation on PtRu surfaces

• Presumed bifunctional mechanism
• H2O Ru ? OHads,Ru H e-
• COads,Pt OHads,Ru ? CO2 Ru Pt H
  e-
• No lateral interactions
• Model surface square lattice

M.T.M.Koper, J.J.Lukkien, A.P.J.Jansen, R.A.van
Santen, J.Phys.Chem.B 103 (1999) 5522
34
Composition influence of Ru content
H.A.Gasteiger et al., J.Phys.Chem. 98 (1994) 617
35
Influence of CO diffusion on activity
CO diffusion essential for catalytic activity
36
Composition influence of Ru islands
High D Reaction occurs at the perimeter of Ru
islands Low D Reaction occurs at Ru
and Pt (two peaks)
37
CO oxidation on non-homogen. PtRu
CO on Pt CO on Ru OH on Pt OH on Ru
38
Summary Monte Carlo

• Monte Carlo simulations are necessary to study
• influences of adsorbate diffusion and surface
• inhomogeneities.
• Mean-field approaches using average coverages
• only work if adsorbates mix well and do not
• exhibit islanding or ordering
• CO diffusion is essential for the bifunctional
• mechanism to predict catalytic enhancement

39
Acknowledgments

• Electron transfer/molecular dynamics
• Christoph Hartnig (TU/e),
• Wolfgang Schmickler (Ulm)
• August Calhoun, Gregory Voth (Utah)
• DFT slab studies on alloys
• Tatyana Shubina (TU/e)
• Frank de Bruijn, Dimitri Papageorgopoulos
  (ECN)
• Monte Carlo studies
• Johan Lukkien, Peter Hilbers (Computing
  Science, TU/e)
• Tonek Jansen, Rutger van Santen (TU/e)
• Royal Netherlands Academy of Arts and
  Sciences
• Netherlands Energy Research Foundation
• Netherlands Organization for
  Scientific Research