is there a future for quantum chemistry

Is there a future for quantum chemistry onsupercomputers?. Computational ChemistryComputational chemistry is a branch of chemistry that usesprinciples of computer science to assist in so

Is there a future for quantum chemistry on

supercomputers?

Jürg HutterPhysical-Chemistry Institute, University of Zurich

Chemistryis the science ofatomic matter, especially itschemical reactions, but also including its properties,structure, composition, behavior, and changes as theyrelate the chemical reactions.(Wikipedia)

Chemistry is the science that invents what it studies

Theoretical Chemistry

theoretical reasoning, usually within physics and mathematics

i h ˙ψ = H ψ Quantum Mechanics

S = k log W Statistical Mechanics

Computational Chemistry

Computational chemistry is a branch of chemistry that usesprinciples of computer science to assist in solving chemicalproblems

Quantum Mechanics Molelcular Orbital Theory

Density Functional Theory

Ab Initio Calculations

Statistical Mechanics Molecular Dynamics

Monte CarloTransition State Theory

Quantum Chemistry & Transition State Theory

Construct potential energy surface for a few degrees of

freedom or find a few stationary points

Quantum Chemistry Software

• Well tested and robust programs

• User friendly and well documented

Success of Quantum Chemistry

Quantum Chemistry Software

DFT

Development of the usage of computational quantum chemistry, as measured by two different metrics The top curve gives the number of citations to software packages per year, while the lower curves provide the number of citations of particular electronic structure methods (specifically pure and hybrid density functionals) Data are from the Web

Quantum Chemistry on Supercomputers

NIC Jülich (Germany) 5 %

Oak Ridge National Lab (US) 17 %CSCS (Switzerland) 16 %

Quantum Chemistry on Supercomputers

• Most standard applications fit on mid-range computersCloud/Grid Computing

Ab Initio Molecular Dynamics

R Car and M Parrinello, Phys Rev Lett 55 2471 (1985)

Electronic Structure Calculations

+

Molecular Dynamics

Ab Initio Molecular Dynamics

System Size Number of Calculations

Standard QC 10-100 Atoms 10-100

AIMD 100-1000 Atoms 10’000 - 1’000’000

Supercomputers are needed

SIESTA, Quantum-Espresso, CPMD, CP2K

AIMD Scales on Supercomputers

History of the performance of AIMD codes on different computer

platforms (Francois Gygi, UC Davis).

AIMD: Example Application

Dye-Sensitized Solar Cells (DSSC) Grätzel Cell

1

1

Wikipedia

• Dynamics and structure of solvent (acetonitrile) atsemiconductor (TiO2) interface

F Schiffmann et al., J Phys-Cond Mat 20 064206 (2008)

• Distribution of electrolyte (I−, I−3) at the interface

F Schiffmann et al., PNAS 107 4830 (2010)

• Regeneration mechanism of dye at interface

F Schiffmann et al., PNAS 107 4830 (2010)

• Binding and IR spectra of dye (N3) on TiO2 surface

F Schiffmann et al., J Phys Chem C 114 8398 (2010)

• Electron transfer dynamics (dye → semiconductor)

F Schiffmann, Thesis UZH 2010

Computational Model

• Almost complete model of a DSSC

• 1751 atom computational cell, 864 (TiO2), 60 dye+electrolyte, 828 solvent

• 9346 electrons, 22951 basis functions

• MD simulation using PBE (DFT+U)

• CPU time on 1024 cores Cray-XT5

• SCF iteration: 13.7 seconds

• MD time step: 164 seconds

DSSC: Complex Electronic Structure

Relative position of orbital levels important for chargelocalization and for electron injection dynamics

I− Distribution at Interface

I− Distribution at Interface

• Solvent near interface cannot be described by singledielectric constant

• Non-monotonic shell structure distribution

• First layer of ACN passivates the surface (no direct contact

of TiO2 to electrolyte)

• I−concentration peaks at 10 Å from surface (all other ionsstudied have decreasing concentrations near the interface)

I−/I−2 Association (Free Energy)

I−/I−2 Association

• Barrierless association of I−/I−2 with dye molecules

• I−3 from interaction of dye/I2complex with I−

• No bimolecular reaction of I−2 in solution necessaryExperiment finds only very small concentrations of I−2 insolution

Postulated Regeneration Mechanism

• Complex systems/interfaces can be studied with AIMD

• Most complex model of a DSSC studied had 1751 atoms

• Insight in various process of DSSC has been gained

• Ion distribution at interface

• Regeneration mechanism

• Electron transfer dynamics

• Complex systems (condensed phase) with strong overlap

to material science and bio-sciences

• ab initio molecular dynamics for statistical sampling