Discussions and Additional Comments

The IFP presentations, while meant to be a sample overview of the IFP’s research activities in SBES, demonstrate that the institute is deeply grounded in the fundamental mathematical, computational, science, and engineering concepts that form the foundation of SBES projects in the service of technological advances. The coupling of this sound foundation with the institute’s significant and ongoing contributions to the solution of “real world” problems and challenges of energy, environment, and transportation form an excellent showcase for the present capabilities of SBES and its even greater future potential. This site offered many excellent examples for illustrating the main themes of the WTEC panel’s report chapters.

This site visit and discussion gave the WTEC visiting team an opportunity to observe the interplay between government policy (e.g., green policy, Kyoto accords, etc.) and a government-owned institute’s applied research with a long-time horizon. The IFP research is more applied than that typically found in universities, and its time horizon starting from fundamental science is of longer duration than that typically found in industry development projects. Use of SBES to develop thermophysical properties tables of complex fuel mixtures (biodiesels) at high pressures is a good example of precompetitive research and a platform that benefits the entire automotive industry and all bio-fuel industry participants. The strategy of developing an extensive intellectual property (IP) portfolio for the refining of heavy crude with the view of a refinery feedstock mix shifting in that direction (e.g., increasing share of transportation fuels originating from tar sands) is another illustration of the long time horizon of the IFP.

The IFP provides interesting examples of an emerging model for funding public goods R&D. Thanks to its growing IP portfolio and permitted ownership (typically minority stakes) of companies that are spun-out of its research discoveries, in the future, the institute expects to generate 40% (as recalled from oral statements) of its budget from such sources. This mechanism should be of wide, global interest as a win-win incentive structure to accelerate the pace of research discoveries in government laboratories.

The institute is interested in recruiting high-quality graduate students and postdocs from all over the world. Opportunities are also available, among others, for visiting faculty on sabbaticals.

Site: Institute of Fluid Mechanics of Toulouse (IMFT)

Tel: 33 5 6128 5839; Fax: 33 5 6128 5899

Email: braza@imft.fr

Guillaume Barbut, PhD Student

Rémi Bourguet, PhD Student

Background

The Institute of Fluid Mechanics of Toulouse (Institut de Mécanique des Fluides de Toulouse, IMFT) is a public research Institute belonging administratively to the CNRS, the Institut National Polytechnique of Toulouse (INPT) and the University Paul Sabatier (UPS). It was founded in 1918 with focus on flow experiments in small-scale models but today its activities are equally divided in computational and experimental fluid mechanics. It employs about 200 permanent staff while more than 120 faculty and 100+ PhD students and postdocs are involved in the research activities. There are six research groups focusing on turbulence, aeronautics (including controls, optimization, sensitivity analysis, shape optimization, reduced order modeling), combustion, two-phase flows, environmental flows, and hydrology. IMFT has one of the most advanced fluids experimentation facilities in France, including a historic experimental subsonic wind tunnel designed by G. Eiffel, established in 1938 and still in use, that General de Gaulle visited in 1959. IMFT is the biggest fluid mechanics laboratory in France (there are other ones: in Lyon – LMFA (smaller 80 faculty); Marseille – IRPHE (phenomena beyond equilibrium); LABM, (Laboratoire d’Aérodynamique et de Biomécanique du Mouvement), Ecole Centrale Nantes (biomechanics & aerodynamics) – LEA (Laboratoire d’Etudes Aérodynamiques – Poitiers), IMFL (Institut de Mécanique des Fluides de Lille) and LADYX, Ecole Polytechnique Paris). There are also other public research Institutes doing CFD as the INRIA (Institut National de Recherche en Informatique et Automatique).

The goal of IMFT is to advance research in the mechanics of fluids with close experimental validation for applications in the areas of energy, processes, the environment and health. The group of Dr. Braza, Director of Research CNRS, has been working on direct (DNS) and large-eddy simulations (LES) of turbulent prototype flows for a long time, and more recently on the detached eddy simulation (DES) of industrial-complexity flow applications. Specifically, DES was the central theme of a European initiative with the name DESider, (Detached Eddy Simulation for Industrial Aerodynamics), (http://dd.mace.manchester.ac.uk/desider) – a 10 M Euro program motivated by the increasing demand of the European aerospace industries to improve their CFD-aided tools for turbulent aerodynamic systems with massive flow separation. DES is a hybrid of LES and statistical turbulence modeling and it is designed as a multiscale framework for modeling and capturing the inner and outer scales of turbulence. DESIder involved many academic and industrial groups with EADS (European Aeronautics & Defense Systems) acting as the coordinator of the program and IMFT a main participant providing turbulence simulations but also unique experimental data for validation of DES codes. Other companies participating are Dassault, Volvo, Peugeot/Citroen, Airbus, etc. All together, 18 organizations were involved (8 from industry, 5 from research institutions, and 5 from Universities) from 8 different European countries. DESider was supported by the 6^th Framework program (2002-2006) on “Aeronautics and Space.”

Furthermore, IMFT participates in the UFAST (Unsteady Effects in Shock-Wave-induced separation) project of the European Sixth Framework Programme, which is coordinated by the Polish Academy of Science (Gdansk) Institute of Fluid-Flow Machinery, and which includes about 18 partners. The objectives are to perform closely coupled experiments and numerical investigations concerning unsteady shock wave boundary layer interaction (SWBLI) to allow for feeding back numerical results to the experiments, and vice versa, for the sake of physics and modelling of compressibility effects in turbulent aerodynamic flows. Using RANS/URANS and hybrid RANS-LES methods, UFAST aims at assessing new methods for turbulence modelling, in particular for unsteady, shock dominated flow. UFAST investigates the range of applicability between RANS/URANS and LES for transonic and supersonic flows.

A follow up-to DESider is the project EMORPH, a multiphysics project (called FP7-AAT-2008-RTD-1/CP-FP) that couples smart materials (electro-active morphing) with CFD and with structural mechanics. Its objective is the development of integrated design tools, controller design and flight control (use distributed actuators that use vibration energy). IMFT is the lead of the project in collaboration with Airbus Toulouse and the LAPLACE (Laboratoire Plasma et Conversion d’Energie) laboratory that has expertise on new electro-active and energy-efficient materials. IMFT has very close collaborations with Airbus Toulouse and very often Airbus engineers co-supervise PhD students associated with IMFT. However, Airbus researchers have their own turbulence code (ELSA, developed by ONERA) and do not use directly the in-house codes of IMFT.

In terms of validation and uncertainty quantification, in addition to grid tests for numerical accuracy, turbulence model accuracy is evaluated by comparisons against unique 3D particle image velocimetry (PIV) data gathered from experiments at IMFT. Such extensive experimental databases are required in validating a new tensorial turbulence, eddy-viscosity model that researchers at IMFT have developed. In addition, researchers conduct tests for upstream noisy inflow conditions and have recently published joint work with Airbus on sensitivity of lift and drag forces due to such random disturbances.

Computing Infrastructure

IMFT is the biggest user of supercomputing cycles in France. It uses primarily the three national supercomputing facilities (Paris IDRIS funded by CNRS, 207 Tflops; Montpelier CINES with mainly IBM machines since they have a factory nearby, 50 Tflops; and Toulouse CALMIP in Université Paul Sabatier). Overall, the computing resources are adequate, and allocation of resources is done through grants. Other available supercomputing resources include CEA in a suburb of Paris (Atomic Energy) and Project Grid 5000—connecting clusters all over France.

Of the 100+ PhD students affiliated with IMFT, most are working on simulation work and are supported by IMFT and national fellowships (Ministry of Education and Research). The Ministry of Defense (DGA-Délégation Générale pour l’Armement, http://www.recherche.dga.defense.gouv.fr) provides fellowships for postdocs on selected topics such as aeroelasticity, biomechanics, and fluid mechanics (about 20 nationally in FM). There is a recent emphasis by the French government to encourage doctoral and postdoctoral training abroad through the Lavoisier program (http://www.egide.asso.fr). In addition, CNRS funds permanent staff and senior researchers to visit abroad for six months or so.

IMFT is a leading fluid mechanics laboratory in Europe with a focus on both fundamental research on simulation and also industrial applications. It is closely interacting with Airbus Toulouse and other companies (Peugeot, Renault) and has been involved in interesting European projects involving many universities, companies, and other research institutions. Its unique experimental facilities serve the simulation work very well, as they provide very useful pointwise measurements required for model validation, especially for unsteady turbulent flows with massive separation.

Site: IRIT (Institut de Recherche en Informatique de Toulouse), and
ENSEEIHT (Ecole Nationale Supérieure d´Electrotechnique, d´Electronique, d´Informatique, d´Hydraulique et des Télécommunications)

(At ENSEEIHT) 2, rue Charles Camichel B.P. 7122

31071 Toulouse Cedex 7, France

http://www.irit.fr/sommaire.php3?lang=en

http://www.enseeiht.fr/en/index.html

http://gridtlse.org
Date Visited: February 25, 2008
WTEC Attendees: A. Deshmukh (report author), G. Karniadakis, G. Lewison
Hosts: Professor Patrick Amestoy, IRIT, Parallel Algorithms and Optimization (APO) Team

Tel: (33) 05 61 58 83 85; Fax: (33) 05 61 58 83 06

Email: Patrick.Amestoy@enseeiht.fr

Ronan Guivarch, Assistant Professor; head, Computer Science and Mathematics Laboratory, Institut National Polytechnique de Toulouse

Tel: (33) 05 61 58 84 08; Email: Ronan.Guivarch@enseeiht.fr

Daniel Ruiz, ENSEEIHT Department of Computer Science and Applied Mathematics

Victoria Moya, Visiting Student from Zaragoza, Spain

Professor Michel Daydé (not present), ENSEEIHT Department of Telecommunications and Networks; Head, Grid-TLSE project
Tél: (33) 05 61 58 82 70; Email: Michel.Dayde@enseeiht.fr

BACKGROUND

ENSEEIHT (Ecole Nationale Supérieure d´Electrotechnique, d´Electronique, d´Informatique, d´Hydraulique et des Télécommunications)

In 1907, the city of Toulouse assisted its seven-century-old university in the creation of an institute dedicated to teaching and research, the Institute of Electrotechnology and Applied Mathematics of the University of Toulouse. Its principle goals were to train the engineers and managers necessary for the hydraulic and electric renovation of the southwest region of France and to contribute to the technological development of these scientific disciplines. The increasing number of students and technical progress led to the creation of the Electrical Engineering and Hydraulics Departments in 1955 followed, in 1956, by the Electronics Department, and in 1959, for the first time in a French engineering school, a Department of Applied Mathematics. In 1967, it became the Department of Computer Science. Today presided over by Professor Alain Ayache, ENSEEIHT is structured into 5 departments: Electrical Engineering and Automation; Electronics and Signal Processing; Computer Science and Applied Mathematics; Hydraulics and Fluid Mechanics; and Telecommunications and Networks. There are 5 closely linked.research labs.

IRIT was founded in 1990. Since then, it has played a prominent role in Toulouse computer science research. It brings together more than 400, members among which are 300 researchers, faculty members, and PhD students affiliated with CNRS (Centre National de la Recherche Scientifique), INPT (Institut National Polytechnique de Toulouse), UPS (Université Paul Sabatier) and UT1 (Université Toulouse1 Sciences Sociales). Research at IRIT covers most of the fields where computer and information science is in progress, be it in its core, ranging from computer architecture to software engineering and computer networks, or in its most contemporary developments like artificial intelligence and cognitive systems, multimedia man-machine interaction, and image interpretation and synthesis. IRIT promotes interdisciplinary research so as to cross fertilize information science and technology with other disciplines, such as linguistics, psychology, ergonomics, and neurosciences, which in turn can benefit from new information-driven concepts and tools. The researchers at IRIT are involved in major national and European research projects.

The goal of the Grid-TLSE (Test for Large Systems of Equations) project is to design an expert site that provides a user-friendly test environment for expert and nonexpert users of sparse linear algebra software. Sparse linear algebra software provides sophisticated algorithms for pre/post processing of the matrices. The selection of the most efficient solver depends on several problem and computational parameters, such as ordering, amount of memory, computer architecture, libraries available, etc. Hence, there is a need for a decision support system to aid users in solving this multiparametric problem to select the most appropriate solver for their problem and computational environment.

This is a multiyear project that started in January 2003. It was originally funded under the ACI GRID Program by the French Ministry of Research. Currently it receives funding from ANR (Agence Nationale de la Recherche) under the Solstice and Lego projects, and from CNRS under the REDIMPS program. The following French research laboratories are involved in this project: ENSEEIHT, CERFACS–Toulouse, IRIT–Toulouse, LaBRI–INRIA, and LIP ENS–Lyon/INRIA. This project also involves the following industrial partners: CEA-CESTA, CNES, EADS, EDF, and IFP.

The Grid-TLSE project aims to develop a Web portal that will provide easy access to tools allowing comparative analysis of solvers available for sparse matrix computations. Users can submit their own problems or choose from a library of matrices, which include domain matrix collections, such as the Rutherford-Boeing and University of Florida sparse matrix collections.

The Grid-TLSE architecture consists of four levels, as shown in Figure 1. The user describes the problem to be solved using a Web portal, WebSolve. The encoded problem description is used by the second level, Weaver, to generate corresponding candidate solver/problem instance list. This list is converted to XML format by the Gridcom level. This specification can be sent to direct solvers on grid resources (Grid 5000 in France) using the DIET middleware, which is a simpler version of Globus developed in GRAAL at Ecole Nationale Supérieure de Lyon. Direct solvers are launched onto a grid of remote servers using CORBA based DIET middleware. Currently, MUMPS (INRIA Bordeaux), UMFPACK (Florida), and SuperLU (Berkeley) solvers are being considered for implementation. This architecture collects statistics on the performance of these solvers for different parameters and operating constraints. Details of the Grid-TLSE project are available at http://gridtlse.org.

CONCLUSIONS

At the time of the WTEC visit, the Grid-TLSE project was still in the design phase. Project team members expect it to go into production mode in a year’s time (early 2009). They anticipate that this utility will lead to significant savings in researcher time in selecting the best solver for problems at hand—from weeks to a day or less. The researchers also plan to extend the current framework for domains beyond linear algebra.

Site: Paris Simulation Network
Participating Universities

École Nationale Supérieure de Chimie de Paris (ENSCP; Hosts)

11, rue Pierre et Marie Curie

75231 Paris, Cedex 05, France

http://www.enscp.fr/

Ecole Normale Supérieure (ENS)

45, rue d’Ulm

F-75230 Paris Cedex 05, France

http://www.ens.fr

Université Pierre et Marie Curie (UPMC)

4 Place Jussieu

75005 Paris, France

http://www.upmc.fr

l’Université Paris-Sud 11 (UPS)

Bât. 300

91405 Orsay Cedex, France

http://www.u-psud.fr

Université d'Evry-Val-d'Essonne (UEVE)

Boulevard François Mitterrand

91025 Evry Cedex, France

http://www.univ-evry.fr/
Date Visited: February 28, 2008
WTEC Attendees: P. Cummings (report author), K. Chong, M. Head-Gordon, S. Kim
Hosts: Professor Alain Fuchs, Director, ENSCP

Email: directeur@enscp.fr

Carlo Adamo, Laboratoire d’Électrochimie et Chimie Analytique, ENSCP

Email: carlo-adamo@ enscp.fr

Anne Boutin, Le Laboratoire de Chimie Physique, UPS

Email : anne.boutin@lcp.u-psud.fr

Damien Laage, Département de Chimie, ENS

Email: damien.laage@eas.fr

Rodolphe Vuilleumier, Laboratoire de Physique Théorique de la Matière Condensée, UPMC. Email: Rodolphe.vuilleumier@lptmc.jussieu.fr

The Paris Simulation Network is a loose association of researchers within Paris engaged in electronic, atomistic, and coarse-grained simulations of chemical, material, and biological systems. It involves researchers Alain Fuchs and Carlo Adamo from the Ecole Nationale Supérieure de Chimie de Paris, (ENSCP); Daniel Borgis, J.T. (Casey) Hynes, Damien Laage, and Rodolphe Vuilleumier from the Ecole Normale Supérieure (ENS); Bertrand Guillot and Pierre Turq from the Université Pierre et Marie Curie (Paris 6, UPMC); Anne Boutin and Bernard Rousseau from the Université de Paris-Sud (Orsay, UPS); and Marie-Pierre Gaigeot and Riccardo Spezia from the Université d’Evry-Val-d’Essonne (UEVE). Several of these eminent researchers hosted the WTEC team (see list of hosts, above).

The participating institutions are quite varied. ENSCP has just 300 undergraduates majoring in the chemical sciences and chemical engineering and 100 doctoral and post-doctoral candidates, with 59 teaching and research staff members. ENS is likewise small, enrolling students in both the humanities and sciences after two years of study at another institution. It is essentially a graduate-only school—students do not receive baccalaureate degrees from the ENS, only graduate degrees. UPMC and UPS are both large institutions, with approximately 30,000 students each. Pierre & Marie Curie University (UPMC) is one of the largest universities teaching science and medicine in France, and indeed in Europe, with 4000 researchers and teaching academics/researchers, 180 laboratories, and 8000 of its 30,000 students in graduate studies. UPS has 1800 teaching staff, 1300 engineers, technicians, administrative staf,f and maintenance staff, as well as 1200 research scientists and 900 technical and administrative staff of the national research organizations (CNRS, INSERM, INRA, CEA). UEVE is a new university, established in 1991, and is mid-sized, with just over 10,000 students.

R&D ACTIVITIES

The members of the Paris Simulation Network (PSN) cover a wide range of research areas, from first-principles molecular dynamics (Gaigeot, Laage, Spezia, and Vuilleumier) and quantum chemistry (Adamo), to atomistic simulation (Boutin, Fuchs, including Gibbs ensemble Monte Carlo simulation for fluid phase equilbria), to coarse-grained techniques such as dissipative particle dynamics (Rousseau), forcefield development (Guillot, particularly for water), chemical reaction and solution dynamics dynamics (Bougis and Hynes), and theory of electrolyte solutions (Turq). There are quite a number of researchers focused on using simulation methods to understand spectroscopy (IR, NMR, UV-Vis, EPR, RX absorption and diffraction) for both gas and condensed phases (including biological systems). The WTEC team’s meeting with the PSN scientists did not focus on the research activities of the members per se, but rather on the questions and issues raised in the questionnaire sent earlier to the participants by WTEC. For specifics of the research activities of PSN members, please refer to the websites given in the references section (some of which are a little dated, but nevertheless give an accurate flavor of the research of each individual). The main research targets of PSN members can be summarized as follows:

• 
  Multiscale simulations
  
• 
  Realistic and direct calculation/prediction of experimental observables
  
• 
  Development/implementation of theoretical models
  
• 
  Quantum-classical simulations
  

• 
  Forcefield matching
  
• 
  Mixed quantum-classical simulation, with application to
  

• 
  Chemical reactions (liquid phase and/or biological systems)
  
• 
  Radiolytic processes
  
• 
  Catalysis (homogeneous and inhomogeneous)
  

• 
  Micro-mesoscopic simulations, with applications to
  

• 
  Polymer systems
  
• 
  Colloids
  

The PSN focuses that are related to energy are in materials design for carbon dioxide capture and storage, radioactive waste treatment at the atomistic level, enzymatic catalysis, and photovoltaic cells. The PSN members pointed to opportunities raised by the new European community regulation on chemicals and their safe use (EC 1907/2006), known by the acronym REACH (Registration, Evaluation, Authorization and restriction of CHemical substances, http://ec.europa.eu/environment/chemicals/reach/reach_intro.htm). The REACH regulation requires specific data on every chemical produced and/or transported into and out of the EU. It has apparently been established that data derived from quantum mechanical and atomistic simulations are permissible in the REACH database, thus establishing the opportunity for modelers to replace costly experiments with accurate computationally derived properties.

The software codes used within the PSN are 50% developed in-house. These codes include Gibbs (a general Monte Carlo code); Newton and MDVRY (classical molecular dynamics codes); a mesoscopic simulation code (for dissipative particle dynamics and Brownian dynamics); QCMD (a mixed quantum/classical molecular dynamics code); a classical DFT code; and many post-processing codes (e.g., to analyze Car-Parrinello molecular dynamics, CPMD, output files). About 30% of the codes used are of the open-source/noncommercial variety: DL_POLY, CPMD, PWSF, CP2K, GROMACS, GAMESS, and NW-CHEM. Some PSN researchers are involved in the development of extensions of CPMD and PWSCF. The remaining 20% of the codes used are commercial: GAUSSIAN, ADF, CRYSTAL, AMBER and CHARMM. Again, one PSN member (Adamo) is active in development for GAUSSIAN. For the in-house codes, at each release the code is validated on a standard set of test problems. Beyond this, the PSN researchers do not perform validation beyond what many researchers do—that is, compare with experiment whenever possible, test against known results on test problems for a given algorithm, etc.

In the area of big data, PSN researchers will be participating in a new CECAM initiative to develop a database of ab initio molecular dynamics trajectories. For PSN researchers, the breakdown of computing resources used is as follows: local clusters account for 80%, European computer centers (CINECA, CEA, Edinburgh, IFP, …) constitute 10% (often made available through collaborations), and the CNRS national center (IDRIS, http://www.idris.fr/) accounts for the remaining 10%. Codes are typically used in parallel, particularly on the external resources. In the future, they would like to see order-N algorithms for first-principles calculations, improvements in processor-to-processor communication and memory access speeds.

In education and training, the WTEC team’s hosts viewed the role of universities to be teaching theory rather than computer science. They view training in simulation methods as being achieved primarily through participation in the training programs of CECAM and the UK CCPs (Collaborative Computational Projects, http://www.ccp.ac.uk/). The PhD students at PSN institutions are essentially all French; post-doctoral researchers are primarily from Europe (Italy, UK, Germany, Spain, and others). Of the graduates of the PSN groups, 20% end up as teachers at undergraduate (nonresearch) universities, 60% end up in academia, and 20% in industry.

Funding in the areas relevant to the PSN comes from the federal government (France’s Ministry of Research, “Agence Nationale de la Recherche” [ANR], CNRS, etc), Europe (the European Science Foundation, http://www.esf.org, and other EU sources), CECAM, and industry. Generally, the trend in funding for research in SBES is positive.