VIEWS: 7 PAGES: 20 POSTED ON: 11/14/2011
Invited Contributions I01 - Computational design of nanostructures Lett. 89, 196803, (2002). [2] J-Y. Raty, G. and nanostructured materials Galli, A.Van Buuren and L. J. Terminello, Phys. Galli G. 1 Rev. Lett.90, 037401 ( 2003); J-Y Raty and 1 Lawrence Livermore National Laboratory, G.Galli, Nature Materials 2, 792 (2003). [3] A. Livermore USA Puzder, A. J. Williamson, J. C. Grossman and G. Galli, J. Chem. Phys. 117, 6721 (2002); Phys. Computational tools are playing an increasingly Rev. Lett. 88, 097401 (2002) and JACS 125, important role in understanding and control- 2786 (2003). [4] E. Draeger, J. Grossman, A. ling matter at the microscopic scale, by pre- Williamson and G. Galli Phys. Rev. Lett. 90, dicting with quantitative, thermodynamic accu- 167402 (2003). [5] F.Reboredo, E.Schwegler racy the property of materials based on their and G.Galli, JACS 125, 15243 (2003). atomic and molecular constituents. In the next I02 - Self-standing and metal-supported decade, we expect the coming of age of ﬁrst molecular nanostructures principles theories of matter and related com- Di Felice R. 1 putational techniques –as well as the growth of 1 INFM-S3 Modena, Italy computer power– to allow for simulations of a wide variety of alternative materials with de- The drive towards novel electronic devices at sired properties, and thus for the engineering of the nanoscale has attracted the interests of re- optimized materials from ﬁrst principles. searchers worldwide to study molecules and In this talk, computer simulations using First molecular materials. A huge investigation ef- Principles Molecular Dynamics (FPMD) and fort is currently devoted to understanding the Quantum Monte Carlo (QMC) techniques are structural and electronic properties of several employed to solve speciﬁc nanoscience prob- candidates for technological exploitation in me- lems, in particular to investigate the physical chanical and electrical devices. The application properties of group IV semiconductor nanos- of ﬁrst-principle DFT calculations to complex tructures and their possible use as chemical and molecules and molecular overlayers is playing biological labels. an ever growing prominent role in unraveling While robust experimental results have been the intimate features of the charge carriers in established for II-VI nanocrystals in the last free molecules and at their interface with inor- decade, group IV elemental nanostructures are ganic substrates. much less well characterized. The interplay be- In this presentation I will give a brief tween quantum conﬁnement effects and surface overview about the simulation of selected properties has not been fully understood, and the biomolecular nanowires [A. Calzolari, R. Di Fe- effects of preparation conditions on the physi- lice, E. Molinari, A. Garbesi, Electron chan- cal properties of Group IV nanoparticles remain nels in biomolecular nanowires, J. Phys. Chem. an open issue. Our simulations are aimed at B 108, 2509 (2004); R. Di Felice, H. Zhang, understanding the physical and chemical prop- A. Calzolari, Ferromagnetic alignment of metal erties of C, Si, SiC and Ge nanoparticles with ions in a DNA-mimic double helix, preprint diameters up to 2-3 nm. In particular, we will (2004)] and nanoglobules [S. Corni, F. De present investigations of optical gaps [1,2] and Rienzo, R. Di Felice, E. Molinari, Redox ac- surface properties[2,3], and simulations of the tivity at the Azurin active site by means of ab- effect of different preparation conditions [4] on initio electronic structure calculations, preprint the structure of Si nanoparticles. Recent simu- (2004)], to illustrate the kind of information that lation results on the early stages of quantum dot can be extracted about their potential electrical functionalization will also be discussed [5]. behavior. Then I will focus on the interface [1] A. J. Williamson, J.C. Grossman, R.Q. between thiols (cysteine and methanethiol) and Hood, A. Puzder and G. Galli, Phys. Rev. the gold (111) surface. The ammino acid cys- 9 Invited Contributions teine is used both to make self-assembled or- at least at moderate temperatures. The “icy” gi- ganic monolayers, and to link complex macro- ants Uranus and Neptune are instead thought to molecules (DNA, proteins, etc.) to the inor- be mostly composed of the so-called planetary ganic world. I will discuss the structure and “ices”: water, ammonia, and methane. High- the enegetics of partial cysteine monolayers [R. pressure and high-temperature simulations in- Di Felice and A. Selloni, Adsorption Modes dicate however that these molecules do not of Cysteine on Au(111): Thiolate, Amino- sustain extreme conditions, and become fully Thiolate, Disulﬁde, J. Chem. Phys. 120, 4906 ionised (water and ammonia) or disproportion- (2004); R. Di Felice, A. Selloni, E. Molinari, A ate (methane), with important implications for DFT study of cysteine adsorption on Au(111), planetary models. Terrestrial planets are ﬁnally J. Phys. Chem B 107, 1151 (2003)], showing composed of an iron core surrounded by a man- the relative role of different functional groups tle of silicates and oxides. Simulations of iron (SH, NH2, SS) in metal-molecule bonding. I can provide constraints on the temperature at will then present a thorough analysis of the the Earth’s center and explain why seismic shear electronic structure, that allows one to identify waves are anomalously slow in the solid core. the metal-molecule hybridization, the degree of Unfortunately, a number of crucial high-P high- chemisorption versus physisorption, the modiﬁ- T properties of mantle minerals such as viscos- cation of the molecular orbitals upon deposition ity, thermal conductivity, thermoelasticity, and onto the inorganic support, and the HOMO-EF rheology, are presently very difﬁcult to address alignment that determines the injection barriers with ab-initio methods, due to their intrinsi- at electrodes. I will ﬁnally show an application cally large time and size scales. The problem is of DFT simulations to compute the energy bar- presently coped with by constructing effective rier for methanethiol pulling from a Au surface classical potential tailored to the ab-initio dy- [A. Calzolari and R. Di Felice, unpublished], namics. Very encouraging results obtained for which is relevant for the mechanical properties liquid silica and magnesium oxide will be pre- and in AFM measurements. sented. I03 - Exploring planetary interiors with ab- I04 - Ab-initio simulation of NMR and EPR initio molecular dynamics spectra in solids Scandolo S. 1 Mauri F. 1 Pickard C. 2 Yates J. 3 1 The Abdus Salam International Cen- 1 Laboratoire de Minralogie-Cristallographie de tre for Theoretical Physics (ICTP) and Paris, France , 2 Cavendish Laboratory, Cam- INFM/Democritos National Simulation Center, bridge, Uk , 3 Cavendish Laboratory, Cam- Trieste, Italy bridge, Uk Because the conditions of pressure and tempera- The orbital magnetic response is at the origin of ture of relevance for planetary interiors, includ- important physical properties that are exploited ing the Earth, extend beyond those currently to determine the microscopic structure of ma- achievable in laboratory experiments, ab-initio terials using the electron paramagnetic reso- atomistic simulations are instrumental to pre- nance (EPR) or the nuclear magnetic resonance dict how planetary matter behaves at extreme (NMR). Indeed, the knowledge of the orbital conditions. The giant planets Jupiter and Sat- currents linearly-induced by a uniform external urn, for example, are mostly composed of hy- magnetic ﬁeld, allows us to compute two impor- drogen, which is known to loose its molecu- tant quantities measured in magnetic resonance lar character and become metallic at sufﬁciently spectroscopies: the EPR g-tensor of defects in high pressures. Ab-initio simulations indicate solids and the NMR chemical shifts. that the transition of ﬂuid hydrogen from molec- I will present a formalism for the ab initio ular to non-molecular is likely to be ﬁrst-order, calculation of NMR chemical shits, and elec- 10 Invited Contributions tric ﬁeld gradients, and of the EPR g-tensor, by molecular evolution, deﬁne the pathways of in molecules and solids based on density func- folding and ensure the avoidance of misfolded tional theory. conformations. In another application, results The reconstruction of the all-electron prop- from protein engineering experiments are used erties with the Projector-Augmented-Waves to determine the transition state for folding for (PAW) and the Gauge-Including-PAW (GIPAW) several proteins. An analysis of the resulting approaches, the use of the scalar-relativistic cor- structures suggests that, in the nucleation con- rections and of ultra-soft pseudopotentials allow densation mechanism for folding, the transition us to describe most of the atoms of the periodic state is reached when the native-like interac- table, and to deal with systems containing hun- tions of few key residues are formed. Taken to- dreds of atoms in the unit cell. gether, these results illustrate how a combina- The accuracy of the predicted NMR and EPR tion of computational and experimental data al- spectra establishes our method as an useful tool lows the investigation of the general principles to interpret and analyze the NMR and EPR spec- of macromolecular self-organization. tra of real materials. I06 - Models of Solvation for Complex So- I will illustrate our approach with applications lutes to bio-molecular and inorganic systems. Marchi M. 1 Borgis D. 2 ´ I05 - Determination of partially folded states 1 Commissariat a l’Energie Atomique, DSV- of proteins at atomic resolution DBJC-SBFM, Centre d’Etudes de Saclay, F- Vendruscolo M 1 e 91191 Gif sur Yvette C´ dex, FRANCE , 1 Department of Chemistry, University of Cam- e e 2 D´ partement de Physique et Mod´ lisation, bridge, UK e Laboratoire de Mod´ lisation des Syst` mes e e e Mol´ culaires Complexes, Universit´ d’Evry, A combination of computational and experi- 91025 Evry C´ dex, FRANCE e mental methods is making it possible to describe protein folding pathways at atomic resolution. We have introduced novel methods to describe Much of the recent progress has been driven by the solvation of complex solutes in computer advances in experimental techniques that pro- simulation. Two classes of techniques have been vide residue-speciﬁc information about the suc- developed which represent the solvent either cessive intermediate states that are populated with a dielectric continuum approach or with during the folding process. Detailed molecular a particle–based model. Both methods enforce dynamics simulations have been used to com- the requirement that they yield either an exact or plement the experimental results and to suggest approximate representation of the macroscopic new measurements. In an alternative approach laws of electrostatics. that I present here, experimental data are used In a ﬁrst approach1 , the solvent is described directly to build an artiﬁcial energy function that by dielectric continuum and the interaction be- guides computer simulations to sample the re- tween the solvent and the molecular degrees of gions of the conformational space of a protein freedom is described by means of a polariza- most compatible with the experimental results. tion density free energy functional which is min- In one recent application, NMR data obtained imum at electrostatic equilibrium. After a pseu- at increasing concentrations of denaturant were dospectral expansion of the polarization and a used for the determination of the free energy discretization of the functional, we construct the landscape of the molten globule state at pH 2 equations of motion for the system based on a of alpha-lactalbumin. The resulting landscape Car–Parrinello technique. In the limit of the adi- is characterized by deep valleys that are ro- abatic evolution of the polarization ﬁeld vari- bust against changes in the external conditions. ables, our method provides the solution of the These deep valleys in the landscape, created dielectric continuum problem “on the ﬂy”, while 11 Invited Contributions the molecular coordinates are propagated. water phase. We compute the effect of the chain In the particle–based method2 , electrostatics length of the alcohol on the phase behaviour of of continuous media is formulated in terms of the membrane. At low concentrations of alco- a polarization density free energy functional, hol the membrane has domains of the interdigi- which is projected on randomly distributed dis- tated phase that are in coexistence with the nor- crete Lennard-Jones pseudoparticles. The re- mal membrane phase. We use our model to clar- sulting model is that of a polarizable ﬂuid, in ify some of the experimental questions related to which the induced dipoles describe both orien- the structure of the interdigitated phase and put tational and electronic polarization. This model forward a simple model that explains the alco- has been implemented in molecular dynamics hol chain length dependence of the stability of simulations and its solvation properties have this interdigitated phase. been compared to continuous electrostatics for I08 - Phase transitions, micellization and simple solutes such as spherical ions or ion shear alignment in polymer and surfactant pairs. systems The computational efﬁciency of both meth- ods have also been tested against more com- Panagiotopoulos A. 1 plex solutes such as proteins. We have shown 1 Princeton University how our implicit solvent molecular dynamics methods can be successfully applied to hydrated This presentation summarizes recent work in the biomolecules, with low cost compared to free author’s group on modeling phase transitions in energy simulations with explicit solvent. model polymer and surfactant systems. A key characteristic of these ﬂuids is the close inter- 1 M. Marchi, D. Borgis, N. Levy, and P. Bal- play between microstructure and macroscopic lone, J. Chem. Phys. 114, 4377 (2001). properties and the existence of interactions on multiple length and time scales. Two com- 2 T. HaDuong, S. Phan, M. Marchi, and D. plementary approaches are used to render the Borgis, J. Chem. Phys. 117, 541 (2002). computational problem tractable, namely dras- tic simpliﬁcation of the model studied to retain I07 - Mesoscopic simulations of phase transi- only essential physical characteristics and de- tions in biological membranes velopment of appropriate sampling methodolo- Smit B. 1 gies to avoid getting trapped in local free en- 1 University of Amsterdam, The Netherlands ergy minima. We have developed a methodol- ogy based on grand canonical Monte Carlo com- Various factors can induce phase transitions bined with histogram reweighting to distinguish in biological membranes. To understand the between phase separation on one hand and mi- molecular mechanisms underlying these transi- cellization to ﬁnite-size aggregates on the other. tions, we have developed a mesoscopic mem- The effects of chain ﬂexibility on polymer and brane model. Molecular simulations on this surfactant phase and aggregation behavior have model nicely reproduce the experimental phase been examined in detail. Ongoing investiga- diagrams. A particular example the will be tions of surfactant systems under shear will also discussed is the effect of alcohol molecules or be described, in connection to the development other small amphiphilic molecules. In the cell of long-range order observed experimentally in membrane these molecules can induce signiﬁ- PS-PEP thin ﬁlms. cant changes in the structure of the membrane. We ﬁnd that alcohol can induce an interdigitated I09 - Fluid and kinetic simulations of inertial structure in which the normal bilayer structure conﬁnement fusion plasmas changes into monolayer in which the alcohol Atzeni S. 1 Schiavi A. 2 Temporal M. 3 Cali- molecules screen the hydrophobic tails from the fano F. 4 Cattani F. 5 Cornolti F. 6 Lisseikina 12 Invited Contributions T. 7 Macchi A. 8 Pegoraro F. 9 matter interaction, plasma and radiation trans- 1 Dipartimento di Energetica, Universit` di a port processes, realistic materials equation-of- Roma “La Sapienza” and INFM, Italy , 2 Dipar- state, fusion reactions and transport of fusion timento di Energetica, Universit` di Roma “La a products. As sample applications we show sim- Sapienza” and INFM, Italy , 3 ETSII, Universi- ulations of the complete evolution of a laser- dad de Castilla-La-Mancha, Ciudad Real, Spain driven target, and of the linear and nonlinear , 4 Dipartimento di Fisica, Universit` di Pisa and a evolution of multi-mode RTI during the forma- INFM, Italy , 5 Dipartimento di Fisica, Univer- tion of the central hot spot. We also present sim- sit` di Pisa and INFM, Italy , 6 Dipartimento a ulations of beam-induced ignition of precom- di Fisica, Universit` di Pisa and INFM, Italy a pressed fuels, which have helped to determine , 7 Dipartimento di Fisica, Universit` di Pisa a the beam requirements for ignition of both DT and INFM, Italy , 8 INFM and Dipartimento di and T-lean fuels. Fisica, Universit` di Pisa, Italy , 9 Dipartimento a We have studied key aspects of the fast ignitor di Fisica, Universit` di Pisa and INFM, Italy a scheme using relativistic electromagnetic codes. Particle-in-Cell simulations show that the inter- We outline the main features of codes we devel- action of an ultraintense laser beam with a solid oped to study inertial conﬁnement fusion (ICF) surface stimulates surface instabilities and re- physics problems, and present recent results on sults in the formation of fast electron jets with selected key issues. characteristic energy and angular distributions. Essential ingredients of ICF are compression Relativistic 3D collisionless ﬂuid simulations of the deuterium-tritium (DT) fuel to at least one have been used to study electron beam prop- thousand times solid density and heating of a hot agation in the dense plasma. The simulations spot which triggers a fusion burn wave. Com- provide valuable results concerning the devel- pression is obtained by laser- or beam-driven opment of the Weibel instability caused by the implosion of a spherical hollow shell, contain- interaction of the beam current with the neutral- ing a frozen DT layer. The hot spot can either izing return current provided by the cold plasma. be produced at the center of the imploded shell by hydrodynamic cumulation (central hot spot I10 - Simulating the Ether: Local algorithms ignition) or generated in the compressed fuel by for Coulombs Law an ultraintense laser or particle beam (fast igni- Maggs A.C. 1 tor scheme). 1 ESPCI, Paris, France. Key issues for ICF are limitation of plasma instabilities (hindering coupling of the laser We have introduced a local algorithm for the energy to the plasma) and hydrodynamic Monte-Carlo simulation of charged systems in- Rayleigh-Taylor instabilities, RTI (hindering teracting via the long ranged Coulomb potential. shell acceleration and compression, and, in cen- This local algorithm is very efﬁcient (with O(N) tral ignition, hot spot formation). Fast ignition scaling) when compared with traditional meth- relaxes hydro-stability requirements, but relies ods based on Ewald summation. It Is based on on not fully understood relativistic laser plasma the constrained dynamics of an auxiliary vector processes. ﬁeld coupled to the physical charge density. ICF target physics involves multiple time- In this talk we will discuss recent progress and space-scales, and requires the use of both ki- in applying cluster methods to further improve netic and ﬂuid models. We study full target per- the efﬁciency of the updates. We demonstrate formance and sensitivity to hydrodynamic in- a considerable improvement in the efﬁciency of stabilities using the 2D hydro-radiative-nuclear the algorithm compared with our original local DUED code. It employs a multi-temperature, Metropolis formulation, especially when work- single-ﬂuid model, including collisional laser- ing with low densities of charged particles. 13 Invited Contributions Generalizations to molecular dynamics are framework has also a solid microscopic founda- also possible. tion that has allowed us, in particular, to derive from microscopic principles the hydrodynamic I11 - Routes to gel formation equations of a ﬂuid mixture that can phase sepa- Sciortino F. 1 rate. The essential assumption is that of van der 1 Dipartimento di Fisica, Universita’ La Waals (separation of the molecular potential in Sapienza Roma Italy a repulsive and attractive parts, the latter being treated in mean ﬁeld). The continuum equations I will discuss some recent numerical studies have surface tension terms in a form common to of colloidal models which undergo dinamical other diffuse interface models. Yet, some ambi- structural arrest. Both glass and gel states will guities concerning the deﬁnition of internal en- be discussed. In particular I will focus on mod- ergy are clariﬁed in the derivation. els where the gel line can be approached from The continuum equations are the basis for a the equilibrium liquid side. Finally I will dis- construction, again within the GENERIC frame- cuss the possibility that gel formation can be in- work, of a discrete model of ﬂuid particles. terpreted in the same theoretical frameworks de- Smoothed particle hydrodynamics, which is a veloped for glass formation. Lagrangian discretization of the Navier-Stokes References: equation and other continuum equations, is used Saika-Voivod, E. Zaccarelli, F. Sciortino, S.V. in order to express the different terms of the Buldyrev, P. Tartaglia Effect of bond lifetime discrete equations not speciﬁed by GENERIC on the dynamics of a short-range attractive col- in a way that ensures that the discrete model loidal system cond-mat/0403320 is a faithful representation of the continuum Francesco Sciortino, Stefano Mossa, equations. We present numerical simulations in Emanuela Zaccarelli, Piero Tartaglia A simple cases (binary mixtures without viscous new route to the formation of colloidal gels: processes) to show the potential of the model. Short range attraction and long range repulsion Non-equilibrium non-isothermal phase separat- cond-mat/0312161 ing processes can be studied with full inclusion I12 - Thermodynamically consistent of cross effects of the Soret type. Smoothed Particle Hydrodynamic model for phase separating ﬂuids I13 - Adaptive sampling techniques to deter- ˜ Espanol P. 1 Thieulot C.A.P. 2 mine transition pathways, free energy, and 1 U.N.E.D Madrid (Spain) , 2 Dept. of Chemi- rates in complex systems cal Engineering, Groningen (The Netherlands) Vanden-Eijnden E. 1 1 Courant Insitute, USA Phase separating ﬂuids are challenging due to the delicate interplay between the complex equi- Various techniques have been developed to sam- librium thermodynamic behavior and the hy- ple efﬁciently the free energy associated with drodynamic behavior of the ﬂuid. In order to some reaction coordinates. Most of these ap- approach the construction of models for these proaches presuppose that the reaction coordi- systems, we have found that the recently intro- nate appropriate to the description of some rare duced framework known as GENERIC (Gen- transition event is known beforehand. This as- eral Equation for Non-Equilibrium Reversible- sumption, however, may lead to a free energy Irreversible Coupling) offers an invaluable tool. landscape that has no dynamical signiﬁcance in This framework captures in an elegant and eco- the sense that it cannot be used to determine nomic way the contents of the First and Sec- the transition region for the rare event, nor its ond Laws and has great potential for the gen- rate. In this talk, I will present a new tech- eration of new models for complex ﬂuids. The nique, termed the string method, which allows 14 Invited Contributions Figure 1: [I12] Four needles at undercritical temperature are introduced in a supercritical mixture. The mixture starts phase separating as a consequence. to search adaptively for the right reaction co- Nose’-Hoover chain, isokinetic scheme allows ordinate associated with some rare event, and molecular dynamics time steps on the order of thereby determine the correct free energy for the 100 femtoseconds to be employed without loss event and the associated transition rate. of accuracy, thus, ameliorating, considerably, the multiple time scale problem. Applications I14 - New methods for sampling rough energy of both methods to model and realistic problems landscapes are presented. Martyna G.J. 1 1 IBM Research, Division of Physical Sciences, I15 - Computing the mechanism, reaction co- Yorktown Heights, USA ordinate, transition states and rate constants of complex processes with path sampling. Atomistic computer simulation studies of com- Bolhuis P.G. 1 plex phenomena allow new insights to be ob- 1 University of Amsterdam, The Netherlands tained that would be difﬁcult to achieve via ex- perimental studies or analytical theory. Indeed, In this presentation I will give a brief overview simulations which give access to the positions of the transition path sampling (TPS) method. and velocities of atoms at all times during im- Designed for studying activated processes in portant events can be used to probe structure, complex environment, the TPS technique har- reactivity and mechanism at an extraordinary vests a collection of transition paths that connect level of detail. However, many processes are the reactant with the product states. This en- impeded by high barriers which are difﬁcult to semble of true dynamical paths allows detailed cross except at very long times inhibiting simu- understanding of the kinetics and mechanism lation from impacting many classes of problems of the reaction. In addition, rate constants can from protein folding to the formation of com- be computed. The main advantage is that the plex interfaces in semiconductor manufacture. method does in principle not need prior knowl- Similarly, signiﬁcant separations in time scales edge of the reaction coordinate. Processes as make designing new methods difﬁcult. For in- diverse as cluster isomerization, auto dissocia- stance, in protein folding, large scale structural tion of water, ion pair dissociation, the folding rearrangements take place on the time scale of of a polypeptide and reactions in aqueous solu- microseconds while carbon-carbon bond vibra- tion have been studied with TPS (see Ref.[1] for tions occur on the femtosecond scale. In this an overview). talk, two new approaches will be described. The rate constant calculation is rather time The ﬁrst, the reference system potential warping consuming and improvements in efﬁciency were scheme permits barriers to be overcome with- achieved by the introduction of the Transition out changing the equilibrium distribution speed- Interface Sampling approach. This approach is ing conﬁgurational sampling by many orders of based on the sampling of paths that cross hy- magnitude. The second, the multiple time step, persurfaces in phase spaces deﬁned by an or- 15 Invited Contributions der parameter that does not have to be equal to chanical Properties of Nanostructures the reaction coordinate. TIS allows for a vari- Louie S.G. 1 able path length, thus greatly enhancing efﬁ- 1 Department of Physics, University of Califor- ciency. Recently, we improved the efﬁciency nia at Berkeley, and Materials Sciences Divi- even more for the calculation of rate constants sion, Lawrence Berkeley National Laboratory, for the case diffusive processes by sampling Berkeley, CA 94720 USA only partial paths. Having introduced the meth- ods I will show their applicability by giving a Calculation of the various properties of materi- few examples. als often requires very different theoretical and There are also disadvantages to TPS, TIS and computational approaches because of the com- related techniques. For instance, qualitatively plex interactions and diverse behaviors in con- different reactive pathways are hard to sample. densed matter. In particular, the restricted ge- Another problem is the existence of unknown ometry and symmetry of nanostructures often local traps (metastable states) in trajectory space give rise to interesting quantum conﬁnement, which can hamper the sampling. I will discuss enhanced many-electron interaction, and other the problems occurring in the sampling of paths effects related to reduced dimensionality. These and their possible solutions. effects can lead to novel physical properties and [1] P.G. Bolhuis, D. Chandler, C.Dellago, and phenomena, which also are potentially useful in P.Geissler, Annu. Rev. Phys. Chem., 53, 291 applications. In this talk, I will discuss some (2002). of our recent studies on the electron transport, I16 - Time correlation functions for mixed optical, and mechanical properties of nanotubes quantum classical dynamics and molecular junctions. For examples, molec- Coker DF 1 Bonella S 2 Causo S 3 Ciccotti G ular electronic devices, i.e., electrical transport 4 through a single molecule can produce highly nonlinear I-V characteristics. The optical spec- 1 Boston University, U.S.A , 2 Boston Univer- tra of small diameter carbon nanotubes exhibit sity, U.S.A , 3 University of Rome, La Sapienza, dramatic excitonic effects. Also friction on the Italy , 4 University of Rome, La Sapienza, Italy nanoscale may be very different from that on A new method is developed for computing ther- the human scale. The talk will be on the the- mal equilibrium time correlation functions of ory and computation of these phenomena. We quantum subsystem operators. The trajectory will present results on the nonlinear transport based approach is obtained from a rigorous sta- behavior of molecular junctions calculated us- tionary phase approximation to the full path in- ing a newly developed ab initio scattering-state tegral expression for the correlation function. method, the optical response of nanotubes em- The method is implemented using adiabatic ap- ploying a ﬁrst-principles many-particle Greens proximations to the thermal and temporal propa- function approach, and the behaviors of me- gators and we explore how nonadiabatic dynam- chanical energy dissipation in double-walled ical effects can be incorporated. The approach is carbon nanotube oscillators from molecular dy- tested on a suite of model problems where exact namics simulations. The physical origin of the results can be obtained and, as a proof of princi- calculated behaviors will be examined. ple check, is employed to test the Green-KuboI18 - Combining ab-initio and classical ap- relation for excess electron transport in metal proaches for the simulation of complex ma- molten salt solutions. The method will also be terials: the case of semiconductor-dielectric employed to study vibrational relaxation in con- interfaces. densed phase systems. Curioni A. 1 I17 - Optical, Electrical Transport, and Me- 1 IBM Research Zurich Research Laboratory 16 Invited Contributions Understanding the properties of semiconductor- scription of complex phenomena of the type de- dielectric interfaces is of paramount importance scribed above. The methodologies are based on in the search for an alternative to silicon diox- density functional theory approaches at the mi- ide as a gate dielectric in CMOS devices. Mod- croscopic scale, on molecular dynamics simu- elling such interfaces is complex and extremely lations at the mesoscopic scale and on contin- challenging, as demonstrated by the unsatisfac- uum (computational or phenomenological) ap- tory comparison of theory and experiments so proaches at the macroscopic scale. We describe far obtained. For this reason we developed a the key ideas behind the links across the scales novel approach that combines the accuracy of and how these are implemented in speciﬁc ex- quantum calculations with the speed and ﬂex- amples. ibility of classical molecular dynamics. The physical properties of these interfaces could be Acknowledgement: This work was per- thus obtained in an unbiased way and compared formed in collaboration with Paul Maragakis, satisfactorily with a number of different experi- Nick Choly, Gang Lu, Greg Smith, Umesh ments. Waghmare and Ellad Tadmor, and supported by the Materials Research Science and Engineer- I19 - Multiscale simulations of mechanical ing Center and the Nanoscale Science and En- and electrical behavior gineering Center of Harvard University, which Kaxiras E. 1 are funded by NSF, and by a MURI grant from 1 Department of Physics and Division of Engi- AFOSR through Brown University. neering and Applied Sciences, Harvard Univer- sity, Cambridge MA 02138, U.S.A. I20 - Deformation of glasses: Bulk rheology and the molecular origins of friction A variety of physical phenomena involve multi- Robbins M.O. 1 Rottler J. 2 Muser M.H. 3 He ple length and time scales in their manifestation. G. 4 Some interesting examples of practical impor- 1 Johns Hopkins Univ., USA , 2 Princeton Univ., tance are: (a) the mechanical behavior of crys- USA , 3 Univ. Western Ontario, Canada , 4 tals and in particular the interplay of chemistry Univ. Pittsburgh, USA and mechanical stress in determining the macro- scopic brittle or ductile response of solids; (b) Simulations have been used to study deforma- the response of piezoelectic crystals to compet- tion of amorphous glasses as a function of stress ing external electric and stress ﬁelds; (c) the al- state, temperature, conﬁnement, and deforma- teration of the structure and electronic proper- tion rate and amplitude. The initial yield be- ties of macromolecular systems due to external havior in models of amorphous metals and poly- forces, as in stretched DNA nanowires or carbon mers is quite similar. The octahedral stress σy nanotubes. at which shear occurs can be described by the In these complex physical systems, the pressure-modiﬁed von Mises (PMVM) criterion changes in bonding and atomic conﬁgurations at σy = τ0 + αp, where p is the pressure. By vary- the microscopic level have profound effects on ing the potential, we show that the value of τ0 the macroscopic properties, be they of mechan- is determined by adhesive interactions between ical or electrical nature. Linking the processes molecules while α is determined by geometry. at the two extremes of the length scale spectrum The yield stress drops linearly with temperature, is the only means of achieving a deeper under- but the shear rate dependence is surprisingly in- standing of these phenomena and of ultimately sensitive to temperature. At sufﬁciently nega- being able to control them. In this presenta- tive pressures, the mode of failure changes from tion we will discuss the development of method- shear to cavitation. Chain connectivity leads ologies for simulations across disparate length to shear hardening at large strains in entangled scales with the aim of obtaining a detailed de- polymer glasses. We show that thin contaminant 17 Invited Contributions ﬁlms between bulk solids are also trapped in on a dual scale simulation scheme are presented glassy states. The yield behavior of these ﬁlms as well, covering coarse grained and atomisti- also follows the PMVM criterion and can nat- cally resolved motion for times up to the total re- urally explain the familiar macroscopic laws of laxation time of chains of up to 10 entanglement friction. In particular, we ﬁnd that the friction lengths. The work presented is done in collabo- force increases linearly with load, and that the ration with R. Everaers and C. Svaneborg (both slope is nearly independent of parameters that now at MPI for Physics of Complex Systems, are not controlled in macroscopic experiments. Dresden, Germany), G. S. Grest (Sandia Nat. Lab., USA), S. Sukumaran, A. Sivasubramanian and S. Leon. I21 - Hydrophobic attraction between protein-like solutes in water I23 - Hierarchical Modelling of Amorphous Dzubiella J. 1 Hansen J.P. 2 Polymers 1 University of Cambridge, UK , 2 University of Theodorou D.N. 1 Cambridge, UK 1 National Technical University of Athens, Greece The effective, solvent-induced forces between nanometer scale solutes in water are determined by constant pressure MD simulations. Macro- Predicting the physical properties of amorphous polymers from their chemical structure consti- scopic considerations predict a strong reduction of the hydrophobic attraction between solutestutes a serious challenge, because of the very when the latter are uniformly charged. This broad spectra of length and time scales that is conﬁrmed by the simulations of oppositely govern these properties. We will discuss some recent hierarchical simulation work aimed at charged solutes, while like-charged solutes lead meeting this challenge by (a) invoking coarse- to signiﬁcantly different behaviour between pos- graining strategies to reduce the number of de- itively and negatively charged pairs. The lat- ter exhibit the phenomenon of like-charge at-grees of freedom in the description of macro- traction, previously observed in some charge-molecules; (b) utilizing powerful connectivity- altering Monte Carlo algorithms to achieve stabilized colloidal dispersions. The hydropho- equilibration of long-chain polymer melt mod- bic interaction turns out to be very sensitive to els at all length scales; (c) mapping long dy- details of the charge pattern on the solute sur- face. namical trajectories onto mesoscopic theoreti- cal models, such as the Rouse and reptation I22 - Entangled Polymer Systems model, in order to estimate rheological proper- Kremer K. 1 ties. Measures of computational efﬁciency will 1 Max Planck Institute for Polymer Research, be discussed, and comparisons against experi- Ackermannweg 10, 55128 Mainz, Germany ment will be presented. I24 - Local Dynamics in Polymer Melts : Extensive computer simulations are used to From Individual Jumps to Diffusive Coarse- study the dynamics and rheology of entangled Grained Dynamics. polymeric systems ranging from semidilute so- lutions of ﬂexible polymers to melts of rather Ryckaert J.P. 1 Arialdi G. 2 Karatasos K. 3 stiff chains. A topological analysis is used 1 Polymer Physics, CP223, Universite Libre to predict the entanglement molecular weight de Bruxelles, Brussels, Belgium , 2 Polymer without any adjustable parameter from confor- Physics, CP223, Universite Libre de Bruxelles, mations only and shown to yield quantitatively Brussels, Belgium. , 3 Polymer Physics, CP223, correct results for the plateau modulus. Simula- Universite Libre de Bruxelles, Brussels, Bel- tions of the dynamics of speciﬁc polymers based gium. 18 Invited Contributions Local dynamics relaxation functions in polymer observed phenomena is discussed. melts which are probed by NMR and neutron 1) M. W. Deem and H. Y. Lee, “Sequence scattering experiments can be obtained from Space Localization in the Immune System Re- molecular dynamics simulations. The molecu- sponse to Vaccination and Disease,” Phys. Rev. lar processes which are responsible for the var- Lett. 91 (2003) 068101. ious relaxations can be analyzed more directly 2) J.-M. Park and M. W. Deem, “Correlations from the simulations. Our data on polyethylene in the T Cell Response to Altered Peptide Lig- (PE) indicate how coarse-graining the analysis ands,” Physica A, to appear. of local motion, a diffusive behaviour emerges at low diffusion vectors. This diffusive process I26 - Analysis of DNA-chip and Antigen-chip separates from more localized motions which data: applications to studies of cancer, stem could be at the origin of the relaxation processes cells and autoimmune diseases. detected in the amorphous domains of semi- Domany E. 1 cristalline PE and more generally may reﬂect 1 Weizmann Institute of Science, Israel the so called alpha-beta splitting in amorphous polymers. DNA chips are novel experimental tools that I25 - Random Energy Model of the Immune have revolutionized research in molecular biol- Response to Viruses, Vaccines, and Cancer ogy and generated considerable excitement. A Deem M.W. 1 single chip allows simultaneous measurement of 1 Physics and Astronomy, Rice University, 6100 the level at which thousands of genes are ex- Main Street - MS 61, Houston, TX USA pressed. A typical experiment uses a few tens of such chips, each devoted to one sample - such as Adaptive vertebrate immune system is a won- material extracted from a tumor. Hence the re- der of modern evolution. Under most circum- sults of such an experiment consist of a table, of stances, the dynamics of the immune system several thousand rows (one for each gene) and is well-matched to the dynamics of pathogen 50 - 100 columns (one for each sample). Ex- growth during a typical infection. Some tracting relevant information from such a large, pathogens, however, have evolved escape mech- complex and noisy data set requires develop- anisms that interact in subtle ways with the ment of novel methods of analysis. immune system dynamics. In addition, nega- In this talk I will brieﬂy explain how gene tive interactions the immune system, which has expression is measured by DNA chips, and evolved over 400 000 000 years, and vacci- demonstrate how we combine standard statis- nation,which has been practiced for only 200 tical analysis with novel unsupervised methods years, are possible. For example,vaccination to mine expression data obtained from various against the ﬂu can actually increase susceptibil- types of cancer. I will show that studies of gene ity to the ﬂu in the next year. As another exam- expression in human stem cells led to the dis- ple, vaccination against one of the four strains covery of a new design principle used by stem of dengue fever typically increases susceptibil- cells. ity against the other three strains. Immunodom- If time permits, I will describe a novel exper- inance also arises in the immune system con- imental tool - antigen chips - and the manner trol of nascent tumors–the immune system rec- it can be used to predict whether an individual ognizes only a small subset of the tumor speciﬁc (mouse) will or will not become diabetic. antigens, and the rest are free to grow and cause tumor growth. I27 - Driven elastic manifolds in random me- In this talk, I present a physical theory of orig- dia: Exact algorithms, new insights inal antigenic sin and immunodominance. How Krauth W. 1 localization in the immune system leads to the 1 Ecole Normale Superieure, Paris, France 19 Invited Contributions f i hi h’ i Figure 1: [I27] Schematic view of continuous-time algorithm (come to talk for details!). I present the algorithmic ideas which allow for the use of environmentally-aware self en- to compute the ultimate pinned d-dimensional ergies in soft matter simulation. manifold in a ﬁnite d+1-dimensional sample, Warren P. 1 both on a discrete lattice, and in the continuum. 1 Unilever R&D Port Sunlight I report on calculations of the roughness ex- ponents at depinning for short-range and long- Dissipative particle dynamics (DPD) can be ex- range elasticity, and of several experimentally tended so that a one-component ﬂuid of soft relevant ﬂuctuation properties. spheres exhibits vapour-liquid coexistence, with a sharp vapour-liquid interface, and a vapour I28 - Mesoscopic models for complex ﬂuids: phase of vanishingly small density [1]. This is from non-ideal to charged mixtures done by allowing the pairwise soft repulsions to Pagonabarraga I. 1 become dependent on a local density [1–3]. The 1 Universitat de Barcelona, Departament de application to capillary problems is illustrated Fisica Fonamental, Carrer Marti i Franques 1, by simulations of a pendant drop (see Figure) 08028 Barcelona, Spain [1], and of novel coarsening kinetics in vapour- liquid phase separation [4]. The model is an In this talk I will describe different strategies example of a soft-sphere liquid with a poten- to model the dynamics of complex ﬂuids in tial energy built out of local-density-dependent the presence of general geometries, paying par- one-particle self-energies. More generally, I ticular attention tot he difﬁculties that arise in argue that the use of ‘environmentally-aware’ charged ﬂuids. I will concentrate then in how many-body self-energies is simpler than the use to couple such models to analyze the dynamics of many-body pair potentials, and offers the of colloidal suspensions, and I will assess the promise to open up new directions for simula- models’ capabilities by analyzing some particu- tion of soft condensed matter. lar examples. [1] P. B. Warren, Phys. Rev. E 68, 066702 I29 - Vapor-liquid coexistence in many-body (2003). dissipative particle dynamics—a manifesto [2] I. Pagonabarraga and D. Frenkel, J. Chem. 20 Invited Contributions Phys. 115, 5015 (2001). I32 - Adaptive Kinetic Monte Carlo Method [3] S. Y. Troﬁmov et al, J. Chem. Phys. 117, for Long Time Scale Simulations 9383 (2002). Jonsson H 1 Henkleman G 2 Uberuaga B.P. 3 [4] P. B. Warren, Phys. Rev. Lett. 87, 225702 1 Univ. of Iceland, Reykjavik, Iceland , 2 Univ. (2001). of Texas, Austin TX, USA , 3 Los Alamos Labs, New Mexico, USA I30 - The Adaptive Integration Method Swendsen R.H. 1 Diffusion and chemical reactions in and on the 1 Carnegie Mellon University, USA surface of solids typically involve a climb over an energy barrier. A typical ’fast’ event on This lecture will present an adaptive Monte laboratory time-scale may have an energy bar- Carlo approach to the calculation of generalized rier of 0.5 eV and occur thousands of times free energies as a function of some parameter per second. In a direct classical dynamics sim- of interest. The Adaptive Integration Method ulation such an event is, however, impossibly (AIM), which was recently developed in collab- slow even when simple interaction potentials oration with Marc Fasnacht and John M. Rosen- are used, requiring thousands of CPU-years be- berg, uses a biasing potential based on the aver- cause of the small time step required to faith- age derivative of the Hamiltonian with respect to fully represent atomic vibrations. A different the parameter of interest to calculate the free en- approach is needed for long time scale simu- ergy and accelerate the simulation. The method lations. We present an extension of the kinetic is particularly effective in dealing with high free Monte Carlo (kMC) algorithm - Adaptive KMC energy barriers. AIM is simple and capable of - where the mechanism and rate of transitions high accuracy. It also allows a substantial im- are found within harmonic transition state the- provement in resolution over the more common ory [1]. A search method based on ﬁnding just histogram methods. Comparison calculations the lowest frequency normal mode [2] is used show that the convergence for AIM is compa- to search for saddle points on the potential en- rable to or better than the well-known Wang- ergy rim surrounding the basin corresponding Landau method for the models tested. to a given state of the system. In this way, the need for a predeﬁned event table and lattice ap- I31 - Computing time scales by milestoning proximation typically employed in KMC simu- Faradjian T. 1 Elber R 2 lations is eliminated. Application of this algo- 1 Cornell University, USA , 2 Cornell Univer- rithm to the dissociation of boron clusters in sil- sity, USA icon will be presented [3]. There, atomic forces obtained by density functional theory are used An algorithm is presented to compute time directly. Another application is multiple time scales of complex processes following pre- scale simulations of the growth of metal surfaces determined milestones along a reaction coordi- using an empirical EAM potential. A screen nate. A non-Markovian hopping mechanism is saver has been developed and used to distribute assumed and constructed from underlying mi- the computations on over a hundred PCs (see croscopic dynamics. General analytical analy- http://eon.chem.washington.edu). sis, a pedagogical example, and numerical so- lutions of the non-Markovian model are pre- [1] G. Henkelman and H. Jonsson, J. Chem. sented. No assumption is made in the theoretical Phys., 115, 9657 (2001). [2] G. Henkelman and derivation on the type of microscopic dynamics H. Jonsson, J. Chem. Phys., 111, 7010 (1999). along the reaction coordinate. However, the de- [3] Blas Uberuaga et al., Physica Status Solidi tailed calculations are for Brownian dynamics B, 233, No. 1 (2002) in which the velocities are uncorrelated in time I33 - Two dimensional model colloids: phase (but spatial memory remains). transitions, effects of external potentials and 21 Invited Contributions Figure 1: [I29] Many-body DPD simulation of a pendant drop. quantum effects ion similar to charge stabilised colloids which Binder K. 1 Lohrer M. 2 Nielaba P. 3 Sengupta are known to undergo an initial freezing, fol- S. 4 Strepp W. 5 lowed by a re- melting transition as the ampli- 1 Physics Department, University of Mainz, tude of the imposed, modulating ﬁeld produced 55099 Mainz, Germany , 2 Physics Department, by crossed laser beams is steadily increased. University of Konstanz, 78457 Konstanz, Ger- Detailed analysis of the simulation data shows many , 3 Physics Department, University of several features consistent with a recent disloca- Konstanz, 78457 Konstanz, Germany , 4 S.N. tion unbinding theory of laser induced melting. Bose National Centre for Basic Sciences, Block The differences and similarities of systems with JD, Sector III, Salt Lake, Calcutta 700098, India soft potentials (DLVO, 1/r12 , 1/r6 ) and the rela- , 5 Physics Department, University of Konstanz, tion to experimental data is analysed. Besides 78457 Konstanz, Germany these classical studies we discuss the validity of our results on atomic length scales. The nature of the melting transition for a system The modiﬁcations of the phase diagram for of hard disks with translational degrees of free- quantum hard disks with ﬁnite particle masses, dom in two spatial dimensions has been anal- obtained by path integral Monte Carlo simula- ysed by a combination of computer simulation tions indicate a new quantum melting scenario, methods and a novel ﬁnite size scaling tech- absent in the classical case. nique. The behaviour of the system is consistent I34 - Simulating Vapor-Liquid Equilibria with the predictions of the KTHNY theory. from First Principles Hard and soft disks in external periodic (light- McGrath M.J. 1 Siepmann J.I. 2 Kuo I.F.W. 3 ) ﬁelds show rich phase diagrams including Mundy C.J. 4 freezing and melting transitions when the den- 1 University of Minnesota , 2 University of Min- sity of the system is varied. Monte Carlo simu- nesota , 3 Lawrence Livermore National Labo- lations for detailed ﬁnite size scaling analyses of ratory , 4 Lawrence Livermore National Labo- various thermodynamic quantities like the order ratory parameter, its cumulants etc., have been used in order to map the phase diagram of the sys- Efﬁcient Monte Carlo algorithms are combined tem for various values of the density and the with the Quickstep energy routines of CP2K amplitude of the external potential. For hard to develop a program that allows for Monte disks we ﬁnd clear indication of a reentrant liq- Carlo simulations in the canonical, isobaric- uid phase over a signiﬁcant region of the pa- isothermal, and Gibbs ensembles using a ﬁrst- rameter space. The simulations therefore show principles description of the physical system. that the system of hard disks behaves in a fash- Conﬁgurational-bias Monte Carlo techniques 22 Invited Contributions and pre-biasing using an inexpensive approx- culated Hugoniot, including phase boundaries imate potential are employed to increase the and transformation kinetics, particularly in com- sampling efﬁciency and to reduce the frequency bination with recent developments in ultrafast of expensive ab initio energy evaluations. The X-ray diffraction measurements of shocked, ori- new Monte Carlo program has been validated by ented single crystals. Reactive potentials such as extensive comparison with molecular dynamics ReaxFF and AIREBO for organic systems offer simulations using CPMD and CP2K. Results for insight into the initial energy redistribution and the vapor-liquid coexistence curve of water us- chemical reaction events in the detonation of ing the BLYP functional will be presented. RDX, TATB, and other energetic single crystals. Using selected examples from various groups, I I35 - Large-scale (106 to 1010 -atom) classi- will try to convey the state-of-the-art in molecu- cal molecular dynamics studies of shock com- lar dynamics shock wave studies, including both pression and ﬂuid instabilities the limitations and the (largely untapped) poten- tial such studies have in contributing to model- Germann T.C. 1 Holian B.L. 2 Kadau K. 3 ing meso-scale behavior. I will also describe re- Lomdahl P.S. 4 Ravelo R. 5 cent simulations of the Rayleigh-Taylor ﬂuid in- 1 Los Alamos National Laboratory (USA) , stability in 2 and 3 dimensions, developing from 2 Los Alamos National Laboratory (USA) , 3 an intially perfect interface between a heavy Los Alamos National Laboratory (USA) , 4 Los ﬂuid on top and a light ﬂuid on bottom, with Alamos National Laboratory (USA) , 5 Univer- only thermal ﬂuctuations seeding the eventual sity of Texas - El Paso (USA) turbulent mixing. Classical non-equilibrium molecular dynamics I36 - Statistical mechanics of proteins: per- simulations have been increasingly used with spective from analytical theory, evolutionary great success over the past several years to study analysis and simulations complex dynamic processes in both solids and Shaknovich E. 1 ﬂuids. Several independent factors have con- 1 Chemistry and Chemical Biology, Harvard tributed to this growing ﬁeld. The development University, 12 Oxford, Cambridge, MA 02138 of efﬁcient parallel algorithms and the availabil- ity of inexpensive PC-based clusters have en- We will disscuss recent developments in protein abled routine large-scale simulations with 106 theory that allow a uniﬁed statistical-mechanical to 1010 atoms, to study the shock compression view of protein folding and evolution. Specif- of solids on sub-micron length scales and sub- ically we will present a mean-ﬁeld picture of nanosecond time scales. Simpliﬁed model po- protein evolution that, along with Monte-Carlo tentials, such as the Lennard-Jones pair poten- design in sequence space allows to rational- tial for close-packed solids and reactive empiri- ize most of the conservatism patterns in pro- cal bond-order (REBO) potentials for energetic tein families and folds. Further, we will present materials, have been used to explore generic novel Monte-Carlo simulations of protein fold- shock processes, including plasticity, solid-solid ing in which all heavy atoms are represented and solid-melt phase transformations, spalla- as interacting hard spheres of various sizes cor- tion, ejecta, grain boundary and polycrystal ef- responding to their van-der-Waals radii. This fects, void collapse-induced hot spots, and the model includes all degrees of freedom rele- propagation and failure of detonation waves. vant to folding - all sidechain and backbone Modern semiempirical force ﬁelds, including torsions- and uses a range of force-ﬁelds - from the embedded atom method (EAM) potential for structure-based Go potential to sequence-based simple metals and the modiﬁed EAM (MEAM) fully transferable atom-atom potentials. By potential for metals with partially covalent be- recording many folding events over over a wide havior, can be rigorously tested through the cal- range of temperatures a possible microscopi- 23 Invited Contributions cally detailed folding mechanism for three dif- I39 - Symmetries in Nature and their realiza- ferent proteins - three-helix bundle, crambin and tion on the lattice protein G is obtained. Further, using combi- Hasenfratz P. 1 nation of experimental, infoarmation, simula- 1 University of Bern, Switzerland tions and evolutionary analysis we were able to fully characterize Transition State Ensemble of Symmetries are the most important guiding a number of proteins providing detailed predic- principle in the search for the fundamental laws tions that can be further tested by experiments. of Nature. Some of the relevant symmetries are These results present a proof-of-principle for the realized on the lattice in a way which will inﬂu- possibility of a solution of protein folding prob- ence our perception of a symmetry in general. lem at an all-atom level, provided that one has a realistic all-atom potential energy function that correctly favors the native state. In a separate I40 - Going Chiral: testing new formulations development this algorithm helped to solve the of lattice QCD ”side-chain packing problem” by providing a Jansen K. 1 rigorous estimate of the number of side-chain 1 John von Neumann-Institute for Computing packing conformations that are compatible with a given backbone of a native protein. One of the most relevant questions in lattice Quantum Chromodynamics is, whether it is pos- I37 - Heavy and light quarks in a hypercubic sible to reach quark masses that correspond box to the physical values as measured in experi- Becirevic D. 1 ments.We will compare two formulations of lat- ´ 1 Universite Paris Sud, Centre dOrsay tice QCD, the chiral invariant overlap fermion and the twisted mass lattice QCD approaches. I discuss the issues related to the lattice QCD It is shown that with both kind of fermions it is studies of the phenomenologicaly relevant quan- indeed possible to reach pion masses of about tities which involve heavy and light quarks. A 200MeV. The prospects for future dynamical special attention is devoted to the problem of simulations in QCD are discussed. ﬁniteness of the size of the lattice, the effects of which may signiﬁcantly change the chiral be- I41 - Computational Cosmology havior to a number of quantities of interest. I Moore B 1 will explain how the problem can be tackled an- 1 University of Zurich, Switzerland alytically and indicate how it can be handled nu- I will review some of the grand challenge prob- merically. I will then discuss a speciﬁc example ¯ 0 ¯0 of the calculation of K 0 -K 0 and Bs -Bs mix-lems in computational cosmology, from planet formation to galaxy formation. The range of ing amplitudes and present some new results for these quantities. scales and diverse physical phenomenon ob- served in the Universe are highly challenging to I38 - A lattice perspective on weak interac- model and require sophisticated hardware and tion phenomenology software solutions. I will discuss the horizon Lellouch L. 1 project in which we plan to construct one the e 1 CNRS-Centre de Physique Th´ orique, Case worlds largest supercomputers in Switzerland 907, CNRS Luminy F-13288 Marseille Cedex based on a novel new design. 9 France I42 - High performance computing for self- I will review recent lattice QCD calculations gravitating systems and their impact on the phenomenology of weak Capuzzo Dolcetta R. 1 processes involving quarks. 1 Dep. of Physics 24 Invited Contributions The simulation of the evolution of astrophysical Rev. Lett., 90, 136401 (2003). systems (gaseous and/or stellar) is made difﬁcult by the dominant role of self-gravity. A discus- I44 - Wave function optimization for accurate sion of the various problems posed by a correct quantum Monte Carlo calculations treatment of the gravitational interaction in the Filippi C. 1 Schautz F. 2 study of stellar systems, with particular atten- 1 Instituut-Lorentz, Universiteit Leiden, The tion to stellar dynamics as a classical gravita- Netherlands , 2 Instituut-Lorentz, Universiteit tional N-body problem, is discussed, as well as Leiden, The Netherlands a scheme of their solution in the frame of high performance computing is presented. We present an efﬁcient method to optimize Jastrow-Slater wave functions for both ground I43 - Slater Determinant Quantum Monte and excited states. The approach is within the Carlo for Realistic Systems framework of the so-called energy ﬂuctuation Zhang Shiwei 1 potential method which minimizes the energy 1 College of William and Mary, Williamsburg, in an iterative fashion based on Monte Carlo VA 23187, USA sampling and a ﬁtting of the local energy ﬂuc- Accurate and robust calculations to treat elec- tuations. The orbitals are optimized together tron interactions are crucial for the reliable with the conﬁguration interaction coefﬁcients prediction of many materials properties. We through the use of additional single excitations have developed a quantum Monte Carlo (QMC) to a set of external orbitals. An improved set method [1] to interface with and improve upon of orbitals is then obtained from the natural or- more traditional electronic structure approaches bitals of this enlarged expansion. For excited for real materials. The method takes the form states, the approach is extended to treat state- of importance-sampled random walks in a space averaging for the optimization of multiple states of Slater determinants. It allows direct in- of the same symmetry. The performance of our corporation of state-of-the-art techniques from approach is illustrated with the lowest states of the very best mean-ﬁeld calculations (non-local several organic molecules, where a careful con- pseudopotentials; high quality basis sets) into struction of the wave function and the reopti- a many-body framework. Electron correlation mization of its determinantal part are required to effects are recovered by stochastic sampling of obtain accurate excitation energies within ﬁxed- mean-ﬁeld solutions (Slater determinants). The node diffusion Monte Carlo. Monte Carlo phase problem is controlled by I45 - Quantum-Classical Reactive Dynamics an approximation with a trial wave function. Kapral R. 1 Sergi A. 2 Hanna G. 3 Any one-particle basis can be used. Using a 1 Canada , 2 Canada , 3 Canada plane-wave basis and non-local pseudopoten- tials, we have applied the method to Be, Si, P, Many chemical reactions, such as proton and S atoms and dimers, molecules containing tran- electron transfer processes, have an important sition elements such as TiO, and bulk Si. Sin- quantum character and are often studied in con- gle Slater determinants from density functional densed phase or other complex environments. In theory (DFT) or Hartree-Fock calculations were such circumstances a quantum mechanical de- used as the trial wave function, with no addi- scription of the entire system is not possible. tional optimization. The calculated dissociation Instead, we describe a mixed quantum-classcial energies of molecules and the cohesive energy approach where the environment is treated clas- of bulk Si are in excellent agreement with ex- sically and the reactive degrees of freedom are periment and are comparable to or better than considered to be quantum mechanical. Reac- the best existing theoretical results. tive ﬂux correlation function expressions for the [1] Shiwei Zhang and Henry Krakauer, Phys. rate constants of non-adiabatic chemical reac- 25 Invited Contributions Relative error in computed dissociation energy measured against experimental value 0.50 present work Density functional theory (LDA) 0.40 diffusion Monte Carlo (DMC) relative error 0.30 DMC (with multi-det. twf) 0.20 0.10 0.00 -0.10 -0.20 Si2 P2 S2 TiO PbO Figure 1: [I43] Present QMC results were obtained using single determinant trial wave functions (twf) from LDA. Standard DMC results (Grossman, JCP 117, 1434 (2002); Wagner and Mitas, CPL 370, 412 (2003)) used optimized Jastrows times single- or multi-det. as twf. tions occurring in such quantum-classical sys- assignment of the measured spectral lines has tems will be presented. Methods for their sim- not been possible. ulation, which are applicable to general classi- I47 - Holistic Simulation cal condensed phase environments, will be de- scribed. The formalism and simulation schemes Sato T. 1 will be illustrated by calculations on models of 1 The Earth Simulator Center non-adiabatic reaction dynamics where the en- vironement is treated either deterministically in The Earth Simulator Project started in July of full detail or as a stochastic bath. 2002 at the Earth Simulator Center of JAM- STEC has already demonstrated several epoch- I46 - Computational spectroscopy of doped making products in the global atmosphere and Helium clusters ocean circulation simulation, the earth’s interior Moroni S. 1 dynamic simulation, nano-material simulation, 1 INFM SMC, University of Rome La Sapienza, tera-hertz oscillator design, exact lattice-QCD Italy solution, etc., that have never been obtained by the conventional computers. Those results en- Rotational spectra of various molecular species sure us that the Earth Simulator has enabled us solvated with Helium atoms are calculated by to make simulation of the whole system at once. a reptation quantum Monte Carlo algorithm. This fact is far more revolutionary and signif- Computer simulations allow us to establish a icant than one would expect. This is because relation between the dynamical properties of one has ﬁnally obtained a scientiﬁc tool to be the solvated molecule and the structure of the able to see the future evolution of environment He solvent. Our results, in excellent agree- and future design of technologies and manufac- ment with the available spectroscopic informa- turing products. Thus, the new ﬁeld of science tion, provide a more detailed microscopic pic- and technology, the revolutionary way of manu- ture than inferred from experiments. They ex- facturing, the symbiotic, safe and secured life of tend our knowledge to cluster sizes where the humans would be brought. 26 Invited Contributions However, the Earth Simulator is far too small sensory paths while ensuring that critical infor- to deal with an entire system in a true sense mation is not lost will be at the forefront of com- that is organized by inseparable numerous in- putational science. teraction forces, microscopic and macroscopic. I49 - Computer Simulation of Multiscale I come up with here a holistic simulator that Phenomena in Colloidal Liquid Crystals would be able to simulate as self-consistently as possible the evolution of an entire system. Allen M.P. 1 1 Centre for Scientiﬁc Computing, University of I48 - Beyond-Petascale Computing– The End Warwick, United Kingdom Of The Beginning Or the Beginning Of The End? This talk will summarize recent studies of in- Camp W.J. 1 Debenedictis E.P. 2 Leland R.W. homogeneous liquid crystals in the vicinity of 3 spherical and nonspherical colloidal particles, 1 Sandia National Laboratories, Albuquerque, using molecular dynamics simulation with a do- NM 87185, USA , 2 Sandia National Labora- main decomposition algorithm, on massively tories, Albuqueruqe, NM 87185 USA , 3 Sandia parallel computers. These studies employed rel- National Laboratories, Albuquerque, NM 87185 atively large systems, up to a million particles, USA and relatively long simulation runs, of order sev- As we move beyond Petascale computing, we eral million timesteps. are facing fundamental issues in computing: On We have studied a isolated spherical ob- the one hand we must deal with the physical jects, representing small colloidal particles or limits of computing machinery. In the not-too- droplets, suspended in a nematic liquid crys- distant future, computing engines may be lim- tal matrix modelled with the soft ellipsoid po- ited by the thermodynamic and heat transfer is- tential. Homeotropic boundary conditions and sues highlighted by Landauer. At the same time, strong anchoring create a hedgehog director reliability of computation will become a domi- conﬁguration on the particle surface and in its nant problem. New architectures and algorithms vicinity; we have studied the defect structure are needed that provide redundancy and other around the particle which cancels this hedgehog mechanisms to deal with ensembles of comput- defect and produces a uniform director ﬁeld at ing components that are at best probablistically large distances. We observe both ring and satel- available and software whose full correctness is lite defects, broadly consistent with the predic- imposible to check. The difﬁculties of the cou- tions of the simplest elastic theory. By study- pling between multiple length and time scales ing density and order-parameter maps, we are in computational simulations must be faced and able to examine behaviour near the particle sur- overcome. Finally, we must cope with Feyn- face, and in the disclination core region, where man’s observation that any computing engine the elastic theory is inapplicable. In a similar capable of describing a complex system is of ne- way we have studied pairs of spherical colloid cessity at least as complex as the system it pur- particles embedded in a nematic liquid crystal ports to describe. In scientiﬁc computing this host, observing the surrounding director struc- means that computer experiments may become ture, and calculating the forces on the particles as hard to prepare and interpret as the physical as functions of their separation and their orien- experiments they simulate. The questions (1) tation with respect to the nematic director. of how to prepare a description of the system Related studies of elongated colloidal parti- to be modelled; (2) of how to provide adequate cles, measuring the torque on a particle as a programming tools to characterize the solution function of tilt angle with respect to the direc- methods; and (3) of how to probe the data pro- tor, and the interactions between pairs of highly duced in a way comprehensible to ﬁnite human elongated particles, have also been carried out. 27 Invited Contributions I50 - Multiscale Computation: From Multi- grid solvers to systematic upscaling Brandt A. 1 1 Weizmann Inst. of Science, Rehovot, IL Most numerical methods for solving physical problems tend to be extremely costly, for several general reasons that will be explained. Model studies have shown that each of these reasons can in principle be removed by multiscale (e.g., multigrid) algorithms. These algorithms em- ploy separate processing at each scale of the physical space, combined with interscale intera- tive interactions, in ways which use ﬁner scales very sparingly. Having been developed ﬁrst and well known as solvers for elliptic PDEs, highly efﬁcient multiscale techniques have more recently been developed for non elliptic and time-dependent problems, and for many other types of computational tasks, including: in- verse PDE problems; highly indeﬁnite (e.g., standing wave) equations; Dirac equations in disordered gauge ﬁelds; fast computation and updating of large determinants (as needed in QCD); general fast integral transforms; inte- gral equations; astrophysics; molecular dynam- ics of macromolecules and ﬂuids; many-atom electronic structures; global and discrete-state optimization; practical network and graph prob- lems; image segmentation and recognition; to- mography (medical imaging); fast Monte-Carlo sampling in statistical physics; real-time path in- tegrals; and general, systematic methods of up- scaling (accurate numerical derivation of large- scale equations from microscopic laws), either deterministic or stochastic. The potential for fundamental achievements in physics and chem- istry will be outlined. 28