# QCM-2013-2-0045 Performance of Time Dependent Density Functional Theory in the strong field photoionisation of noble gas atoms

#### MEC y MICINN

*Status: finished project*

Application Id: QCM-2013-2-0045

1.- Activity Title:

Performance of Time Dependent Density Functional Theory in the strong field photoionisation of noble gas atoms

Area: Chemistry and Materials Science and Technology

2.- Research Project Description:

a) Is this a Test Activity? No

b)Is this a Long Term Activity that will extend over several application periods? No

c) Brief description of the Project

We want to examine the performance of Time Dependent Density Functional Theory in the field of strong field photoionisation of noble gas atoms. To understand the response of these atoms exposed to a strong external field one should solve the time-dependent Schödinger Equation, but this is a complex task because the wavefunction scales exponentially with the number of electrons. Instead, we use TDDFT [Density functional theory for time-dependent systems, E. Runge and E.K.U. Gross, Phys. Rev. Lett, 52, 997, 1984; Self-Consistent Equations Including Exchange and Correlation Effects, W. Kohn and L. J. Sham, Physical Review, 140, A4: A1133, 1965] due to its computational simplicity as all observables can be extracted from the one-body electron density. However, in TDDFT, the functional dependence of the exchange correlation potential is unknown and needs to be approximated.

In particular, we want to calculate the total ionisation yield, which within TDDFT, can be obtained by following the evolution of the electronic density in time of our atoms as we apply a strong external field. However, in order to test the TDDFT performance we need to perform many calculations for several noble gas atoms (Neon, Argon, Xenon) as a function of the external laser parameters (frequency, pulse length, intensity) and of the density functional dependence of the exchange correlation potential to check the effect of its asymptotic decay (LDA, PBE, LB94, CXD-LDA) [Accurate and simple analytic representation of the electron-gas correlation energy, J.P. Perdew and Y. Wang, Phys. Rev. B, 45, 13244, 1992; Generalised Gradient Approximation Made Simple, J.P. Perdew et al., Phys. Rev. Lett., 77, 3865, 1996; Exchange-correlation potential with correct asymptotic behaviour, R. van Leeuwen and E. J. Baerends, Phys. Rev. A, 49, 2421, 1994; Prediction of the Derivative Discontinuity in Density Functional Theory from an Electrostatic

Description of the Exchange and Correlation Potential, X. Andrade and Alán Aspuru-Guzik, Phys. Rev. Lett., 107, 183002, 5, 2011].

On the other hand, these atoms are composed of several electrons, where we expect that it is a reasonable approximation to use a pseudopotential for the tightly bound core electrons, but this has to be checked from an all-electron calculation.

This research project is being supported by the Grupos consolidados UPV/EHU del Departamento de Educación, Universidades e Investigación del Gobierno Vasco (Ref. IT578-13), the European Commission projects CRONOS (Grant number 280879-2 CRONOS CP-FP7), as well as the Spanish Grant

MICINN (FIS2010-21282-C02-01)

Grant References: IT578-13 280879-2 CRONOS CP-FP7 FIS2010-21282-C02-01

d) Brief description of the Project (If this Activity takes place in the context of a Technology or Industrial Project):

e) Specific Activity proposed:

To perform our calculations we are using the Octopus code [http://www.tddft.org/programs/octopus/wiki/index.php/Main_Page] which is already installed in MareNostrum. We need to perform many calculations, as we need to test the laser parameters, the TDDFT density functionals, as well as noble gas atoms from an all-electron and pseudopotential calculation as I mentioned in section c) to have a clear idea of how TDDFT works in the strong field photoionisation regime. Our CORVO cluster currently only allows us to perform a parallelised 48 hour calculation with the "short queue" and an 168 hour calculation with the "long queue" and we only have about 1080 processors for a group of about 30 people (that is on average 35 processors per person per day), which obliges people to queue for several days many times, especially when the calculations require many nodes or when one needs to perform many calculations. This situation even becomes worse when one needs to use the "long queue", because the "short queue" calculations start first as their priority is greater.

In our case, we have to test many calculations and each one of the pseudopotential calculations (some tests have already been performed for a Neon atom, the simplest noble gas we want to test, composed of 10 electrons) require on average a minimum of 72 processors to be able to use the "short queue" as they last 48 hours (although this depends on the pulse length that we test). However, for the all-electron calculations, the "long queue" is required and using 48 processors they take 96 hours for the shortest length pulse, so then we expect that for the longest laser pulse they will take 576 hours using the same amount of processors.

On the other hand, we also have to consider that these calculations will take longer for noble gas atoms with more electrons, Argon is composed of 18 electrons and Xenon of 54 electrons.

To concretise, we have to test three noble gas atoms, three pulse frequencies (one per atom), two pulse lengths per atom, 7 intensities per functional and 8 different functionals (the 4 functionals mentioned above from a pseudopotential calculation and an all-electron one) per atom.

As the all-electron calculation is only performed to understand the physics behind the pseudopotential, we will only test these for the shorter laser pulse and the simplest noble gas atom, which in this case is Neon.

According to this application, the typical job runs would correspond to the pseudopotential calculations and the largest job runs would correspond to the all-electron calculations.

Consequently, we consider that using MareNostrum would be necessary to be able to perform our calculations, not just because of the amount of time they take, but also because of the amount of calculations we have to perform and the number of processors that we need to use.

f) Computational algorithms and codes outline:

The project involves the use of an installed computer code in MareNostrum that implements techniques of TDDFT in real-space. We want to use the Octopus code (http://www.tddft.org/programs/octopus/wiki/index.php/Main_Page) written by M. Marques et al. This is a real-time, real-space TDDFT code.

The code is parallelised in spatial domains, states and k-points. It uses the libraries FFTW3, LAPACK/BLAS, GSL, Perl, NetCDF.

This code has been tested with excellent performances on MareNostrum in previous projects carried out by our group.

3.- Software and Numerical Libraries

Software components that the project team requires for the activity

Applications + Libraries: BLAS FFTW GOTO GSL LAPACK MPICH NETCDF SCALAPACK SPARSEKIT OCTOPUS

Compilers and Development Tools: GCC IBM XL Compilers

Utilities + Parallell Debuggers and Performance Analysis Tools: PYTHON GNUPLOT LIBTOOL GRACE AUTOCONF AUTOMAKE

Other requested software:

Propietary software:

4.- Research Team Description

a) name of Team Leader: Angel Rubio Secades

Institution: Departamento de Física de Materiales, Facultad de Químicas, Universidad del Pais Vasco

e-mail: angel [dot] rubio [at] ehu [dot] es

Phone: +34-943018292

Curriculum Vitae of the Team Leader:

Angel Rubio is a Professor of Condensed Matter Physics in the Department of Materials of the Faculty of Chemistry in the Basque Country University (UPV/EHU). His research activity is internationally recognized and he has received numerous honors and awards. Among them, we would like to mention: National Prize for the best Spanish undergraduate student of Physics (1989), the Royal Spanish Physical Society Prize "Outstanding young researchers" (1992); the Sir Allan Sewell Fellowship School of Science, Griffith University, Australia (2004); the Fellow of the American Physical Society: Materials Science Division (2004); The 2005 Friedrich Wilhelm Bessel Research Award, Humboldt Foundation, Germany (2005); DuPond Prize on Science, (2006); ther “Distinguished Visiting Scientist”, Fritz Haber Institute der Max-Planck- Gesellschaft, Berlin (2009); The Outstanding Referee award, American Physical Society, (2009).

Prof. Rubio has an outstanding publication record: more than 200 scientific publications with more than 13000 citations (Hirsch index 57). His group has a deep and long-standing experience in code-development (octopus, Yambo, Abinit) and usage of high-performance computing facilities. The main research activity is in the field of theory and modeling of electronic and structural properties developing novel theoretical tools and computational codes to investigate the electronic response of solids, nanostructures to external electromagnetic fields. The group is involved in the organisation of the International Workshop and School: Time-Dependent Density-Functional Theory: Prospects and Applications, held every two years in Benasque Center for Science, Benasque, Spain. The group also collaborates with MareNostrum for developing and testing high-performance codes.

d) Names of other researchers involved in this activity:

1) Alison Crawford Uranga

Institution: Nano-Bio Spectroscopy group and ETSF Scientific Development Center, Departamento de Física de Materiales, Centro de Física de Materiales CSIC-MPC and DIPC, Universidad del País Vasco UPV/EHU, Avenida de Tolosa 72, E-20018, San Sebastián, Spain

E-mail: alisonc1986 [at] gmail [dot] com

2) Esa Räsänen

Institution: Department of Physics, Tampere University of Technology, FI-33101 Tampere, Finland

E-mail: esa [dot] rasanen [at] tut [dot] fi

3) Umberto De Giovannini

Institution: Nano-Bio Spectroscopy group and ETSF Scientific Development Center, Departamento de Física de Materiales, Centro de Física de Materiales CSIC-MPC and DIPC, Universidad del País Vasco UPV/EHU, Avenida de Tolosa 72, E-20018, San Sebastián, Spain

E-mail: umberto [dot] degiovannini [at] ehu [dot] es

4) Micael J. T. Oliveira

Institution: Center for Computational Physics, University of Coimbra, Rua Larga, 3004-516 Coimbra, Portugal

E-mail: micael [at] teor [dot] fis [dot] uc [dot] pt

5) Peter Lambropoulous

Institution: Institute of Electronic Structure and Laser

Foundation for Research and Technology

P.O. Box 1527, GR-71110, Heraklion, Greece

E-mail: labro [at] iesl [dot] forth [dot] gr

6) Stefan Kurth

Institution: Nano-Bio Spectroscopy group and ETSF Scientific Development Center, Departamento de Física de Materiales, Centro de Física de Materiales CSIC-MPC and DIPC, Universidad del País Vasco UPV/EHU, Avenida de Tolosa 72, E-20018, San Sebastián, Spain

E-mail: stefan_kurth [at] ehu [dot] es

e) The five most relevant publications, in the last five years, from the members of the research team:

Octopus: a tool for the application of time-dependent density functional theory A. Castro, H. Appel, Micael Oliveira, C.A. Rozzi, X. Andrade, F. Lorenzen, M.A.L. Marques, E.K.U. Gross, and A. Rubio,, Phys. Stat. Sol. B 243 2465-2488 (2006)

Ab-initio angle and energy resolved photoelectron spectroscopy with time-dependent density-functional theory, U. De Giovannini, D. Varsano, M. A. L. Marques, H. Appel, E. K. U. Gross, A. Rubio Physical Review A 85, 062515 (2012)

Strong-field ionization suppression by light field control, Esa Räsänen and Lars Bojer Madsen, Phys. Rev. A 86, 033426 (2012)

Simulating pump-probe photo-electron and absorption spectroscopy on the attosecond time-scale with time-dependent density-functional theory, U. De Giovannini, G. Brunetto, A. Castro, J. Walkenhorst, A. Rubio Chemphyschem 14, 1363 - 1376 (2013)

Route to direct multiphoton multiple ionisation, P. Lambropoulos et al., Phys. Rev. A, 83, 021407(R), 2011

f) Does any of the researchers involved in the activity apply to the mobility program?: No

5.- Resources

a) Estimated resources required for the Activity fot the current Application Period:

Requested machine: MareNostrum 3 (IBM System X iDataplex with Infiniband / >40000 cores)

Interprocess communication Tighltly Coupled

Typical job run:

Number of processors needed for each job 96.00

Estimated number of jobs to submit 150.00

Average job durations (hours) per job 24.00

Total memory used by the job (GBytes) 150.00

Largest job run:

Number of processors needed for each job 240.00

Estimated number of jobs to submit 28.00

Average job durations (hours) per job 48.00

Total memory used by the job (GBytes) 100.00

Total disk space (Gigabytes) Minimun 250.00 Desirable 300.00

Total scratch space (Gigabytes) Minimun 50.00 Desirable 100.00

Total tape space (Gigabytes) Minimun 50.00 Desirable 100.00

Total Requested time: (Thousands of hours) 668.16

INFORMATION: The estimated cost of the requested hours, considering only the electricity cost, is 7637.0688 euros.

The required resources have to be executed in the selected machines, the other architectures do not fit the requirements to execute the proposal.

** this option implies that if no hours in this machine/these machines are available, the acces committee will reject the full application.

6.- Abstract for publication

We want to examine the performance of Time Dependent Density Functional Theory in the field of strong field photoionisation of noble gas atoms.

We need to perform many calculations, as we need to test the laser parameters, the TDDFT density functionals, as well as noble gas atoms from an all-electron and pseudopotential calculation.

Our CORVO cluster currently only allows us to perform a parallelised 48 hours calculation with the "short queue" and an 168 hours calculation with the "long queue" and we only have about 1080 processors for a group of about 30 people.

Consequently, we consider that using MareNostrum would be necessary to be able to perform our calculations, not just because of the amount of time they take, but also because of the amount of calculations we have to perform and the number of processors that we need to use.