NUQG

From HP-SEE Wiki

Revision as of 08:32, 23 February 2012 by Antun (Talk | contribs)
Jump to: navigation, search

Contents

General Information

  • Application's name: Numerical study of ultra-cold quantum gases (NUQG)
  • Virtual Research Community: Computational Physics
  • Scientific contact: Antun Balaz (antun at ipb.ac.rs)
  • Technical contact: Antun Balaz (antun at ipb.ac.rs)
  • Developers: Scientific Computing Laboratory, Institute of Physics Belgrade, Serbia
  • Web site: http://www.scl.rs/

Short Description

The behavior of atoms at low temperatures is a research field that attracted a lot of attention after experimental realization of Bose-Einstein condensation in 1995. and several Nobel prizes for this discovery and associated effort related to cooling techniques able to produce quantum gases at temperatures as low as 50 nK. Macroscopic quantum phenomena (Bose-Einstein condensation, superfluidity, superconductivity) are now extensively studied experimentally, and numerical simulations have become an essential tool to understanding such strongly correlated systems. In the past several years optical lattices have also attracted much research interest due to tunable nature of the parameter space, allowing in-situ studies of the behavior that was previously inaccessible in the condensed matter physics. The recently introduced effective action approach by the path integral group of the Scientific Computing Laboratory of the Institute of Physics Belgrade provides an ideal framework for numerical study of Bose-Einstein condensation and response of the condensate to rotation, vortex formation and evolution in the rotating condensate, the study of the parametric resonance in the collective modes of the condensate due to the modulation of the interaction using the Feshbach resonance, and effects of disorder to global and local properties and dynamics of Bose-Einstein condensates.

NUQG application requires HPC resources due to several reasons: exact diagonalization of large systems, requiring excessive amounts of memory; tightly-coupled large-scale computations of real-time dynamics, requiring significant amounts of computing power to process large initial conformations of systems of ultra-cold quantum gases; large-scale study of the effects of disorder, which require real-time dynamics calculation for the statistically significant ensemble of disordered potentials etc.

Problems Solved

Calculation of the ground state for ultra-cold quantum gases systems, calculation of global and local properties of Bose-Einstein condensates, real-time dynamics and formation of vortices, effects of disorder on Bose-Einstein condensation, study of Bose-glass quantum phase transition and its characterization.

Scientific and Social Impact

This HPC application will be a significant numerical tool for ongoing research on the physics of ultra-cold quantum gases, being carried out at the Scientific Computing Laboratory of the Institute of Physics Belgrade and its international collaborators. The results which will be obtained will provide insights and deepen our understanding of dynamical processes in highly correlated systems of quantum gases, properties of their phases, and new quantum phase transitions due to the effects of disorder. These questions have raised a significant research interest in recent years, and are currently being investigated in leading quantum optics laboratories throughout the world. A number of questions still remain unresolved (e.g. precise characterization of the Bose-glass phase transition), and the NUQG application will contribute to their solving.

The understanding of the behavior of quantum gases is a fundamental research problem, and has various important societal impacts. It leads to the development of quantum optics and nano-science, which is essential for many aspects of the technology development, including the design of new devices, materials and meta-materials, nano-fabrication, etc. In addition to these practical societal impacts, the obtained results can be effectively used to popularize the science, especially among the young students, due to the fundamental nature of the research problems studied (especially the notion of quantum phase transition), visual appeal of the dynamical evolution and effects of disorder on Bose-Einstein condensates, and its vast possible technological applications, specifically in information and communication technologies.

Collaborations

This application is developed as a part of a long-term research at the Scientific Computing Laboratory of the Institute of Physics Belgrade. As of 2007, the development of this application also serves the collaboration between SCL and University of Duisburg-Essen, Potsdam University and Free University of Berlin. During 2011, we have also established collaboration with the experimental group from University of Sao Paulo (campus Sao Carlos).

Within the framework of the project, we have initiated collaboration with the Romanian partner IFIN-HH, who have extensive experience in variational studies of Bose-Einstein condensates. We also envisage collaboration with the Bulgarian partner IICT-BAS on the application of quasi-MC methods for generation of disordered potentials, as well as development of its GPGPU counterpart.

Beneficiaries

  • Researchers in the field of ultra-cold atoms and molecules
  • Researchers in the field of nonlinear optics
  • Researchers studying nonlinear Schroedinger equation

Number of users

18 users

Development Plan

  • Concept: Done before the project started.
  • Start of alpha stage: Done before the project started.
  • Start of beta stage: M1
  • Start of testing stage: M5
  • Start of deployment stage: M13
  • Start of production stage: M15

Resource Requirements

  • Number of cores required for a single run: From 1 to up to 1024
  • Minimum RAM/core required: 1-2 GB
  • Storage space during a single run: 10 GB
  • Long-term data storage: 4 TB
  • Total core hours required: 3 000 000

Technical Features and HP-SEE Implementation

  • Primary programming language: C/C++
  • Parallel programming paradigm: OpenMP/MPI
  • Main parallel code: NUQG is a suite of codes and contains SPEEDUP module for exact diagonalization, module for study of real-time dynamics, new module for disorder potential generation, and module for solving GP equations
  • Pre/post processing code: Included in the main code
  • Application tools and libraries: SPRNG, Intel MKL, FFTW
  • Number of cores required: From 1 to up to 1024
  • Minimum RAM/core required: 1-2 GB
  • Storage space during a single run: up to 10 GB
  • Long-term data storage: up to 4 TB

Usage Example

NUQG set of codes solves the time-(in)dependent Gross-Pitaevskii nonlinear partial differential equation in one, two, and three space dimensions in a trap using imaginary-time and real-time propagation. The Gross-Pitaevskii equation describes the properties of a dilute trapped Bose-Einstein condensate. The equation is solved using the split-step Crank-Nicolson method by discretizing space and time. The discretized equation is then propagated in imaginary or real time over small time steps.

To run any of the codes compiled with the provided Makefile, use: 

  ./<codename> -p <parameterfile>

where <codename> is a name of the compiled executable, and <parameterfile> is a parameter input file prepared by the user. Examples of parameter input files are included in the code distribution.

Example: Run imagtime2d compiled code with the parameter input file located in input/imagtime2d-input 

  ./imagtime2d -p input/imagtime2d-input

Infrastructure Usage

  • Home system: PARADOX
    • Applied for access on: 09.2010
    • Access granted on: 09.2010
    • Achieved scalability: 512 cores
  • Accessed production systems:
  1. HPCG
    • Applied for access on: 11.2010
    • Access granted on: 11.2010
    • Achieved scalability: 512 cores
  • Porting activities: The SPEEDUP module required porting from the Mathematica output for the effective action (Mathematica code is used to produce analytic expressions for higher order effective potentials). GP module required porting from the previous Fortran version of the code to C, as well as parallelization using the hybrid MPI/OpenMP programming model. DISORDER module, which will add additional features to the code, is still in development, and will be mainly based on OpenMP. All C codes are ported for compilation with GNU's gcc compiler, Intel C compiler, and IBM's xlc compiler.
  • Scalability studies: Tests with different numbers of cores.


Publications

Retrieved from "http://hpseewiki.ipb.ac.rs/index.php/NUQG"
Personal tools