GEMS: THE GRID EMPOWERED MOLECULAR SIMULATOR

Antonio Laganà

Department of Chemistry, University of Perugia, Italy

THE PROJECT• Step 1 - SIMBEX (Simulator of Crossed Beam

Experiments) for atom diatom trajectory studies

• Step 2 - GEMS (Grid Empowered Molecular Simulator)

• Step 3 - Grid version of GEMS

• Step 4 - Some case studies

• Step 5 - The COMPCHEM Molecular Science Virtual Organization (VO)

• Step 6 - Next

1 - SIMBEX

SIMBEX: CROSSED BEAM EXPERIMENT of Perugia

MEASURABLES- Angular and time of flight product distributions

INFORMATION OBTAINABLE- Primary reaction products- Reaction mechanisms- Structure and life time of transient- Internal energy distribution of products- Key features of the potential

THE SIMULATOR

Interaction

Observables

Dynamics

Virtual Monitors

System input

The implemented INTERACTION module

INTERACTION

DYNAMICS

Is therea suitable

PES?

Import thePES parameters

NO

YES

START

CAVEATS

PES not needed in on the fly methods.

Seldomly a PES already exists

PESs can be semiempirical

Best if from a fit of ab initio values

Often PESs are of low accuracy

The implemented DYNAMICS module

DYNAMICS

OBSERVABLES

Are trajectory

calculationsaccepta-

ble?

NO

YES

TRAJ: application

using classical mechanics

calculations

CAVEATS

Implementation with trajectories

ABCtraj for atom diatom

The implemented OBSERVABLES module

OBSERVABLESNO

YES

Is the observable

a state-to-stateone?

DISTRIBUTIONS: Virtual Monitors forscalar and vector

product distributions

Do calculated

and measuredproperties

agree?

EXTEND THECALCULATION

TO OTHERPROPERTIES

YES NOTRY USINGANOTHER SURFACE

PG

MI

PD

BO

BA

NA

RM

The prototype ChemGrid.it of grid.it

CILEA

UPV

UBCESCA

CINECA

High perfor-mance nets

GARR Fiber optics

Portals Security Communications

Resource Management MonitoringMiddleware

HP Components Problem Solving

Libraries Cost models

Program-Ming tools

Applications

Astrophysics Bioinformatics Earth observation

Geophysics Computational Chemistry

THE EGEE PRODUCTION GRID

• EGEE is a European project aimed at developing a European service grid infrastructure available to scientists.

• A prototype implementation of the Grid Molecular Simulator has been selected for the NA4 Activity of EGEE (Application Identification and Support)

                                                                                                                                                            

                                                                       

THE EGEE PRODUCTION GRID

The GRIDified atom diatom TRAJ kernel

TRAJ

return

Iterate over initial conditionsthe integration of individualtrajectories (ABCTRAJ, etc.)

Define quantities of generaluse

Collect individual trajectory results

TRAJECTORY NATURAL CONCURRENCY

SEND “ready” status messageRECEIVE seedintegrate trajectoryupdate indicatorsSEND “ready” status messageGOTO RECEIVE

Worker:

DO traj_index =1, traj_number RECEIVE status message IF worker “ready” THEN generate seed SEND seed to worker ELSE GOTO RECEIVE endIF endDO

Master:

THE VIRTUAL MONI-TORS SHOWED THE PRODUCT ANGULAR DISTRIBUTIONS FOR THE VARIOUS CHANNELS

H+ICl→Cl + HI

H+ICl→H + ICl

H+ICl→HCl+I

Using history files to rationalize mechanisms

NEAR RECROSSING IN REACTIVE PROCESSES

2 – A GENERALIZATION OF SIMBEX TO GEMS: THE GRID EMPOWERED

MOLECULAR SIMULATOR

The molecular dynamics problem

twWHtwWt

i ,,ˆ,,

Electronic Schrödinger equation:

WwWEWwH nnnelec ;;ˆ

Nuclear Schrödinger equation:

tWt

itWH nnn ,,ˆ

Separation of electronic and nuclear motions

ELECTRONIC SCHRÖDINGER EQUATION

• Programs: often standard packages

• Methods - wavefunction quantum approaches (MRCI) - density functional theory (DFT)

• Classical

transform the Schrödinger equation into a set classical mechanics equations and integrate them in time

• Quantum - Integrate the equation in time for a given (or a set of

given or an average distribution) state(s) - Integrate the (stationary) equation in space for a given

energy and all energetically open states

NUCLEAR SCHRÖDINGER EQUATION

• Semiclassical

overimpose quantum effects of the associated wave to quantum mechanics outcomes

THE QUANTUM TREATMENT

Time dependent

{W} – set of position vectors of the nuclei or choices of center of mass coordinates like the already seen Jacobi Rτ and rτ vectors

HN - nuclear Hamiltonian

METHOD – integrate the first order time dependent equation using time as continuity variable and either collocating the system wavepacket on a grid (for R and r) or by expanding it on a basis set (for Θ)

Time independent

{W} – set of position vectors of the nuclei or choices of orthogonal coordinates of which one can act as continuity variable in going from one arrangement to another

HN - nuclear Hamiltonian

METHOD – segment the continuity variable in sectors and expand locally (in each sector) the wavefunction on the remaining (orthogonal) coordinates

THE QUANTUM TREATMENT

0

)()()(

1)(

2

tdtCTQTQ

Tk fNtrans

translational partition function

Qtrans(T) R

22

32

rotational partition function

oddjj

evenjjN

j

jTQ

exp123

exp126)(2

Flux-flux correlation function

)(tC f

By exact MCTDH or approximateSC-IVR calculations

FLUX CORRELATION FUNCTION FORMULATION OF THE RATE

COEFFICIENT

THE MCTDH METHOD• Diagonalisation of the thermal flux operator

defined onto a dividing surface to build a reduced Krylov subspace (iterative diagonalisation by consecutive application of the thermal flux operator on a trial wave function). The outcome is a set of eigenvalues and eigenstates of the thermal flux operator.

• Time propagation of the thermal flux eigenstates employing MCTDH.

• Calculation of observables: k(T), N(E).

THE FLUSS PROGRAM

QDYN: the Quantum dynamics group in COMPCHEM (from COST Action D37)

• A COST Action to foster the constitution of a Molecular science community in the European Grid initiatives

• A working group (QDYN) to implement exact and approximate quantum methods

• Develop workflow and expert system tools for quantum chemical investigations

• Enhance collaborative research work in terms of service offer/request within quantum chemistry developers

• Foster the transfer of exact molecular treatments to industrial and commercial applications

MEMBERS OF QDYN

• A. LAGANA’, O. GERVASI (Perugia, Italy)

• G.G. BALINT KURTI (Bristol, UK)

• E. GARCIA (Vittoria, Spain)

• F. HUARTE (Barcelona, Spain)

• G. LENDVAY (Budapest, Hungary)

• G. NYMAN (Goteborg, Sweden)

• S. FARANTOS (Heraklion, Greece)

• M. LAUNAY (Rennes, France)

OTHER APPROACHES

• Reduced dimensionality quantum methods

• Classical, quasiclassical and molecular dynamics methods

• Semiclassical methods

3 – GRID IMPLEMENTATION

The extended INTERACTION module

INTERACTION

DYNAMICS

Is therea suitable Pes?

Are ab initiocalculationsavailable?

Are ab initiocalculations

feasible?

Import thePES routine

NO NO NO

YES YES YES

Take force fielddata and

procedures from relateddatabases

START

FITTING SUPSIM

SUPSIM: the Gridified Ab initio approach

SUPSIM

return

Iterate over the systemGeometries the call of ab

initio suites of codes (GAMESS, GAUSSIAN,

MOLPRO, etc)

Define the characteristics of the ab initio calculation, the coordinates used and the

Variable’s intervals

Collect single molecular geometry energy

The FITTING portal

FITTING

Return

Are asym-ptotic values

accurate?

Are remai-ning valuesinaccurate?

Do ab initiovalues have the

proper sym-metry?

Enforce the propersymmetry

Application using fitting programs to

generate a PESroutine

Modify asym-ptotic values

NO NONO

Modify short andlong range values

YES YESYES

The extended DYNAMICS module

DYNAMICS

OBSERVABLES

Exact quantum

calculations?

NO NO

YES YES

CLASSICALIntegration of the

Classicalequations

APPRQDYNIntegration of the approximate quantum dynamicsequations

QDYNIntegration of theexact quantum

dynamics equations

SEMICLASSICALIntegration of clas-sical equations and

of the associatedwave

YES

NO Ap-

proximate quantumcalculations?

Se-miclassical

calculations?

The QDYN PROCEDURES

QUANTUMDYNAMICS

OBSERVABLES

Single Initial

quantum state?

Multiple initial

quantum states?

NO NO

YES YES

CRP: cumulative

reaction probabilities and TransitionState theory

TI: atom diatomS matrix

elements for a single energy

TD: atom diatom S matrix elements

for several energies

MCTDH: reactive flux over the

barrier

Statespecific

(summed overfinal states)

YES

Fully averaged

Gridified time dependent approaches

TD

return

•Iterate over initial conditions•the time propagation •(RWAVEPR, CYLHYP, etc.)

Define quantities of generaluse

•Collect single initial state•S matrix element

Gridified time independent approach

TI

return

Iterate over total energy value the integration of scattering

equations

Define quantities of generaluse including the integration

bed

Iterate over the reaction coor-dinate to build the interaction

matrix

Broadcast coupling matrix

Collect coupling matrix elements

Collect state to state S matrix elements

Gridified MCTDH method

The extended MEASURABLES module

OBSERVABLESNO NO

YES YES

Is the observable

a state-to-stateone?

Is theobservable

a state specificonee?

VM for thermal and thermodynamic pro-

perties including Molecular Virtual

Reality tools

CROSS: VM for statespecific cross sections,

rate constants and maps of

product intensity

DISTRIBUTIONS: VMfor scalar and vectorproduct distributions,

and state-to-state crosssections

Do calculated

and measuredproperties

agree?END

YES

INTERACTION

NO

Beam VM for Intensity in the

Lab frame

PROGRAMS BEING IMPLEMENTED ON THE GRID FOR PERSONAL USE

Perugia, ABC (also using PGRADE), RWAVEPR, CYLHYP, DL_POLY

Bristol, DIFFREALWAVE

Vittoria, RWAVEPR, VENUS

Vienna, COLUMBUS

Budapest, ABC, VENUS, RWAVEPR

Barcelona, MCTDH

Goteborg (On the fly Q-RBA?)

Heraklion, MODTINKER

4 – SOME CASE STUDIES

The N+N2 case study

)',()(),()( 12

412

4 vNSNvNSN gg

2NN : the LEPS potential energy surface

The collinear LEPS surface

Isoenergetic contour maps 1 eV spacing

Reactive state to state probabilities

0.146 eV

0.433 eV

0.717 eV

0.997 eV

E(v)

V=0

V=1

V=2

V=3

1.270 eV

1.543eV

V=4

V=5

Threshold energies

1.359 eV

0.950 eV

0.772 eV

0.199 eV

Etr

V=0

V=1

V=2

V=4

2NN : the L3 potential energy surface

The bent L3 surface (125o transition state geometry)

Isoenergetic contour maps 1 eV spacing

2NN : the L4 potential energy surface

The bent L4 surface (125o transition state geometry)

Two higher barriers sandwiching a well

L3

L4

●

●

Rate coefficients

L3L4

LEPS

IONIC BIOLOGICAL CHANNELS

They are usually schematized as a sequence of:• Entrance gate• Bilayer pore• Selectivity filter

• Biological ionic channels play an important role in the control of ionic cellular concentrations and in synapses

ION FLOW THROUGH NANOTUBES

A nanotube model can be used to understand the ionic conductivity of cations (like Na+ or K+) through cellular membranes.

A life science application to the understanding of cellular micropores

THE CARBON NANOTUBE AS A MODEL

We considered the CNT as a model for biological ionic channels (though it has also several interesting applications in itself)

MOLECULAR DYNAMICS

• H+/D+ ions flowing through a carbon nanotube

• A quantum scattering problem solved using a 3D time-dependent technique (the problem has been already solved using classical approaches)

• Implementation of a quantum scattering formalism based on polar cylindrical coordinates to single out resonances, interferences and tunneling

A quantum approach to ion flow in nanotubes

SCATTERING IN CYLINDRICAL SYMMETRY PROBLEMS

In the nanotube problem the symmetry is about cylindricalThe most suitable coordinates are the polar cylindrical ones (r,,z) The projection of the total angular momentum on z is a good quantum number

with k being the momentum along z.

iKθnK e)Rr

(ρJθ)ψ(r,

ikzeψ(z)

BASIS SET

R is the nanotube radius K is the angular momentum component on z

ρn is the nth zero of the Bessel function JK

The radial component is a Bessel function and the angular component is an imaginary exponential

The z component of the wavefunction is given by plane waves:

THE WAVEPACKET

- The initial (t=0) wavepacket is placed at one end of the nanotube- Its shape is that of an eigenfunction of the polar component of the Hamiltonian with a given component of the total angular momentum and a given radial excitation (that of the corresponding Bessel function)- Its z component is a Gaussian times a phase factor (corresponding to the linear momentum)

ikz2σ

)z(z

ee(z)2

20

-0.003

-0.002

-0.001

0.000

0.001

L=0

-0.003

-0.002

-0.001

0.000

L=5

-0.003

-0.002

-0.001

0.000

L=10

Out

goin

g F

lux

0 500 1000 1500 2000-0.003

-0.002

-0.001

0.000

L=30

Time (atomic units)

OUTGOING FLUX PLOTS: angular momentum

H+ - Elong=0.04 h Etransv=0.01 h

An increase of the value of the angular momentum quantum number slightly delays the flux (the increase of the centrifugal potential pushes the wavepacket closer to the nanotube walls).

Docking Proteina - Molecola piccola

Recettore: Adipocita Proteina che lega i lipidi

PDB code: 1LIC

Ligando: acido Esadecan sulfonico

Barnase

Barstar

1BRS : Barnase + Barstar

Docking Proteina - Proteina

AIR POLLUTION SIMULATION

CPM10 Concentration from CHIMERE-aerosols

5 – THE COMPCHEM VIRTUAL ORGANIZATION

WHAT IS COMPCHEM

• COMPCHEM is a Virtual Organization (VO)

• VOs specialize a segment of the European Grid for specific purposes

• COMPCHEM is the VO of molecular and material sciences

• It is based, at present, on a subgrid of more than 8000 cpus (out of the 80000 of EGEE)

THE COMPCHEM APPROACH1. USER PASSIVE : Runs other’s programs ACTIVE: Implements at least one program for personal usage2. SW PROVIDER (from this level on one can earn credits) PASSIVE : Implements at least one program for other’s usage ACTIVE: Management at least one implemented program for cooperative usage 3. GRID DEPLOYER PASSIVE : Confers to the infrastructure at least a small cluster of processors ACTIVE: Contributes to deploy and manage the structure 4. STAKEHOLDER: Takes part to the development and the

management of the virtual organization• Further information at http://compchem.unipg.it

6 – FUTURE GUIDELINES

QUANTUM CHEMISTRY DATA STANDARDIZATION

• The Q5 data model and format was created for quantum chemistry (electronic structure) data by the WG 4 of D37

• Create D5 a data model for dynamics (in particular quantum dynamics)

• Extend the Q5 standard to D5

CREATE AND TEST WORKFLOWS

• Inter-job workflow

- Wrap the jobs

- Treat the jobs as objects

- Define composition rules and data links

• Intra-job workflows

- Define tools as for inter-job workflows via directives to be inserted inside the jobs

GETTING READY FOR EGI

• Broaden the molecular and material science user basis

• Introduce and gridify other suites of programs• Carry out massive calculations using the

gridified programs• Extend the usage of graphical interfaces and

virtual reality either to define initial conditions or to represent final observable properties

• Develop a credit system• Cluster COMPCHEM with other Grid VOs

ACKNOWLEDGEMENTS

• CDK group, Dept. Chemistry, Perugia (Crocchianti, Faginas, Pacifici, Skouteris, Costantini, Rampino, Manuali)

• HPC group, Dept. Math&Inf, Perugia (Gervasi, Tasso)

• Qdyn group, COST D37 (Garcia, Huarte, Lendvay, Nyman, Balint-Kurti, Farantos)

• Other groups of COST D37