Supercomputing Institute · 2014. 1. 16. · 2 Supercomputing Institute Research Bulletin mammalian...

Volume 15, Number 3

for Digital Simulationand Advanced Computation

Supercomputing Institute

Supercomputing Institute Research Scholars . .page 51999 Undergraduate Interns . . . . . . . . . . . . . .page 8Advanced Computation of Cosmic Flows . . .page 16Flow and Transport in Porous Media . . . . . . .page 18Modeling of III-Nitride Devices . . . . . . . . . .page 20Structural Studies of Toxins . . . . . . . . . . . . .page 22

Simulations of Magnetic Structures for Digital Technology Devices . . . . . . . . . . . .page 24

Sparse ‘99 International Conference . . . . . . .page 27Colloquium Series . . . . . . . . . . . . . . . . . . . .page 31Visitors . . . . . . . . . . . . . . . . . . . . . . . . . . . .page 32Research Reports . . . . . . . . . . . . . . . . . . . . .page 35

Also in this issue

Fall 1999 Research Bulletin of the Supercomputing Institute

page 18 page 20 page 22

NSF-IGERTComputationalNeuroscience ProgramHosts First InternationalSymposium

The “First Annual ComputationalNeuroscience Symposium” was heldin the Basic Sciences and Biomedical

Engineering Building at the University ofMinnesota in Minneapolis on October 7 and8, 1999. The conference was sponsored by theUniversity of Minnesota ComputationalNeuroscience Program [which is funded by aNational Science Foundation IntegrativeGraduate Education and Research Training Jan Hondzinski (left) of the University of Minnesota Department of Neuroscience

presents her poster to Henrietta Galiana (right) of McGill University in Montreal,Canada.continued on next page

2 Supercomputing Institute Research Bulletin

mammalian inwardly rectifying potassiumchannel, refined by molecular dynamics simu-lations, and for the pure domain of a voltageactivated potassium channel. The discussionfocused on how the implications of the simu-lation studies aided the understanding ofexcitable cell membranes. Weinstein spoke ofcurrent work on the mechanisms of G proteincoupled receptors triggered by molecularrecognition and leading to signal transductionand modulation. The focus was on receptorstructure and ligand-induced transduction ofthe activation signal for a neurotransmitter.

Stephen Cannon of Harvard MedicalSchool spoke about the use of both animaland computational models used to explore thephysiological consequences of altered gating in

mutant sodium chan-nels for skeletal mus-cle excitability. A curi-ous feature of skeletalmuscle was shown tobe that symptomaticattacks could resultfrom either increasedexcitability or a loss ofexcitability.

Laurence Trussellof the OregonHearing ResearchCenter and theVollum Instituteexplained the dynam-ics of transmitterrelease and its role inshaping neuralresponses. The modelTrussell expoundedon suggested thatsustained, plateautransmission isdependent largely ona rapid recovery ofrelease sites, whiledesynchronization

The Computational NeuroscienceProgram is a crosscutting doctoral graduateprogram funded by the National ScienceFoundation Integrative Graduate Educationand Research Training Program. It combinestraining in computational and physical sci-ences with training in neuroscience.Neuroscience is a relatively new and highlyinterdisciplinary field, encompassing not onlyall the classical biological disciplines, but alsoexperimental psychology. Neuroscientists usea variety of experimental approaches tounderstand the development and function ofthe nervous system. As neuroscience matures,the need for quantitative modeling andinsights from the physical and computationalsciences grows. Scientific computation, a dis-cipline that combines state-of-the-art meth-ods in computing hardware and software tosolve challenging problems in science andengineering, has seen exceptional growth inrecent years. This progress is brought aboutin part by unprecedented gains in computertechnology and new algorithms to takeadvantage of this technology. The

Computational Neuroscience Program com-bines the University of Minnesota’s strengthsin these two interdisciplinary areas with thecomputational resources and infrastructure ofthe Supercomputing Institute to provide aunique training and research program.

Graduate Fellows pursue the Ph.D. degreewith a major in Neuroscience or ScientificComputation. They will minor in ScientificComputation or Neuroscience, respectively. Atypical fellowship is two to three years, withsupport for other years in the form of ateaching assistantship or research assistant-ship. The computational NeuroscienceProgram has four main components—Ph.D.coursework, Ph.D. research, seminars and sci-entific meetings, and teaching.

For more information about theComputational Neuroscience Program pleasevisit the program’s World Wide Web page at:

www.compneuro.umn.edu/

or contact the Program Administrator,Kathleen Clinton, at (612) 625-8424 or [email protected].

Computational Neuroscience Program

(IGERT) program grant], the NeuroscienceGraduate Program, the Scientific ComputationGraduate Program, and the SupercomputingInstitute for Digital Simulation and AdvancedComputation. Eighty participants from fivecountries attended a day and a half of talkscovering molecular mechanisms in ion chan-nels, signal transduction, neurotransmissionand receptors, computational models ofvistibular and oculomotor control, robotics andcomputer vision, and neural network models.

The conference began Thursday morningwith presentations by Mark Sansom of OxfordUniversity, England and Harel Weinstein ofMt. Sinai School of Medicine. Sansom dis-cussed a study that used a bacterial channel asthe starting point for a homology model of a

3Supercomputing Institute Research Bulletin

results from the enhancement of releaseprobability of docked vesicles.

The Thursday afternoon sessions beganwith a presentation by Dora Angelaki of theWashington University School of Medicine.Angelaki spoke on the coding of movementin inertial space. The discussion focused onthe computational problems and neuronalstrategies and concluded that, even thoughthe peripheral sensory transduction of motionis dictated and bound by the laws of physics,neural computations are centrally used by thebrain to reinterpret sensory afferent signalsand compute inertial motion.

Henrietta Galiana from McGill University,Canada followed with a discussion of systemstyle models she is helping develop for the con-

James Clark (left) of McGill University, Canada,Daniel Kersten (center) of the University ofMinnesota, and Gregory Hager (right) of JohnHopkins University.

Mitsuo Kawato (left) of the AdvancedTelecommunications Research Institute in Kyoto,Japan talks with Andrew Barto (right) of theUniversity of Massachusetts.

Mark Sansom (left) and Charlotte Capener (right)of the Laboratory of Molecular Biophysics atOxford University in Oxford, England discuss andshow off their work in a poster paper presentation.

trol of eye movements in several reflexes. Herpresentation focused on the evolution of thecontrol models used to develop this system.

James Clark from McGill University,Canada described some of the evidence for thepremotor models of spatial attention and eyemovements. One focus of his talk was on sim-ulations of a computational model of spatialattention based on the premotor theory thatreplicates a wide range of phenomena relatedto the latencies of saccadic eye movements.

The day was concluded with a keynoteaddress at the evening’s banquet by TerrenceSejnowski of The Salk Institute for BiologicalStudies and the University of California at SanDiego. Don add information on this talk . . .

The second day began with a presentationby Stephen Lisberger of the University ofCalifornia School of Medicine on the recon-struction of commands for smooth pursuit eyemovements from a population code of thedynamics of image motion. The focus was onhow a series of neural and computationalanalyses were conducted that demonstrate howimage velocity and acceleration are represented.

Gregory Hager of Johns Hopkins Universityfollowed with a discussion of the developmentof several control systems that make direct useof image information for vision-based tracking,manipulation, and navigation in three dimen-


sions. The results showed anatural hierarchy of tasks thatcould be used to determine thelevel of information any systemhas about its underlying visual-motor system.

The conference concludedwith presentations by AndrewBarto of the University ofMassachusetts and MitsuoKawato of AdvancedTelecommunications ResearchInstitute, Japan. Barto spokeof a control architecture for aneuronal model that learnedto reach using correctivemovements and how it ismotivated by the anatomyand physiology of the motorsystem. Results of simulationsdesigned to explore thedegree to which the modelmeets the computationalchallenges of controlling adynamic arm were also pre-sented. Kawato spoke of cere-bellar internal models for both robotics andcognition. Several paired forward and inversemodels were introduced and possible func-tional roles in hierarchical planning andcommunications were discussed.

In addition to these talks, eleven poster pre-sentations added to the discussions. Thespeakers and poster presentations shed lighton this intensive field and helped make theconference a success.

■

Clockwise from top left, Stephen Lisberger (left) of the University ofCalifornia at San Fransisco and Terrence Sejnowski (right) of the SalkInstitute; Stephen Cannon (left) of the Massachusetts GeneralHospital and Linda Boland (right) of the University of Minnesota;David Thomas (left) of the University of Minnesota and HarelWeinstein (right) of Mount Sinai Medical School; and Sue Kierstaedt(left) of the University of Minnesota and Dora Angelaki (right) ofthe Washington University School of Medicine.

Fifteen Years of the Car-Parrinello Method in Physics and Chemistry (March 18–19, 2000)

This workshop will honor the originators of the Car-Parinello method and review major accomplishments and cut-ting-edge developments. The Car-Parrinello method, introduced in 1985 by Roberto Car and Michele Parrinello, hasdramatically influenced the field of electronic structure calculations for solids, liquids and molecules, and initiated thefield of quantum molecular dynamics. The seminal idea was to deal directly with the total energy of the electron systemas a function of its degrees of freedom, namely the expansion coefficients for the one-electron wave functions, and treatthese degrees of freedom on the same footing with the ionic degrees of freedom. Further information can be found onthe World Wide Web at:

www.msi.umn.edu/general/Symposia/car/


Supercomputing Institute Research Scholars

The Supercomputing Institute providesgrants for research scholarships forsupercomputing research at the

University of Minnesota. This year, theSupercomputing Institute is pleased toannounce the appointment of six researchscholars. Research scholars are research associ-ates who work closely with SupercomputingInstitute principal investigators.

Olivier Mouzin, from the NationalInstitute of Applied Science in Lyon, France,is working with Professor Joan Bechtold ofthe Orthopaedic Surgery Department on acomputational model of tissue differentiationin bone that incorporates mechanical andbiological factors. The longevity of anorthopaedic implant depends on the integrityof the bone/implant interface. The timecourse of tissue differentiation surroundingan implant influences the integrity of thisinterface. The ability to computationallymodel and predict this differentiation of bonetissue and the bone/implant interface will beinfluential in focusing the next level of ration-al implant design refinement. This work isoptimizing the accuracy and predictive powerof computational models of adaptive tissuedifferentiation in bone, incorporatingmechanical and biologic factors. The bone,implant, and surrounding gap in which thetissue differentes are analyzed using biphasicfinite element modeling techniques. The roleof biologic factors, and their interaction withthe mechanical environment, is examinedthrough cytokine and growth factor networkfeedback relationships. Experimental valida-tion will be provided by data obtainedthrough pressure sensors attached to the testimplant in a coordinated experimental effort.

Yongho Kim, from Kyung-Hee University,Korea, is working with Professor DonaldTruhlar of the Chemistry Department ondeveloping techniques for improved calcula-tions on the rates of the kinds of chemical reac-tions that are important for gas-phase combus-tion. Special emphasis is being placed on meth-ods that can be used to study reactions over awide temperature range since it is often impor-tant for gaining a complete picture of combus-tion reactions and of the potential energy sur-

Gennadiy Poda and Govindan Subramanian workat the Visualization-Workstation Laboratory.

Olivier Mouzin (front) works with Dr. PascalSwider of the National Institute of Applied Sciencein Lyon, France in the Scientific Development andVisualization Laboratory at the Institute.


faces that govern them. Variational transitionstate theory with multidimensional semiclassi-cal tunneling calculations has proven to beespecially suitable for the prediction of chemi-cal reaction rates at both low and high temper-atures, and therefore, this type of dynamics cal-culation was chosen for further development.This project involves the improvement of prac-tical techniques for applying the theory to vari-ous classes of transition states, including newmethods based on multi-configuration molecu-lar mechanisms for interfacing reaction-pathdynamics calculations with electronic structuretheory, and applications to specific reactions.

Guido Kanschat, from the University ofHeidelberg, Germany, is working withProfessor Mitchell Luskin of the MathematicsDepartment on the development of datastructures for finite element research codes.The library that he is co-authoring aims atthe fast evaluation of discretization methodsincluding error estimates and adaptive gridgeneration. He is also working on adaptivemethods for parameter estimatation for par-tial differential equations.

Gennadiy Poda, from the University ofMinnesota, is working with Professor DavidFerguson on a computational analysis ofChemokine receptor structure and function—an exploration to develop novel chemokineantagonists. This study is working toward anunderstanding of the structural properties ofchemokine receptors as well as their potentialconnection to opioid receptor binding and acti-vation. This family of G protein-coupled recep-tors (GPCRs) has been identified as necessarycofactors in promoting the fusion and ultimateinfectivity of the human immunodefficiencyvirus (HIV) to host cells. These researchers areidentifying commonalities in receptor structure,function, and ligand binding modes with thelong term goal of designing potent antagoniststo the chemokine family of receptors. To evalu-ate receptor structure similarities, sequencecomparisons are performed on both chemokineand opioid receptors using both sequence align-ments and three-dimensional superpositions.Chemokine ligand modeling and docking isperformed to further isolate recognition ele-ments common to the receptors. The dockingmode(s) predicted will be compared with arecent model of peptide binding to the k-opi-oid receptor (UMSI 97/262). Finally, experi-mental ligand binding assays will be performedto evaluate predicted binding modes and com-monalities in receptor structure and function.

Eli Kostadinova Stoykova, from theUniversity of Sofia, Bulgaria, is working with

Guido Kanschat works on data structures at theSchool of Mathematics.

Eli Kostadinova Stoykova works at ShepardLaboratory.


Yongho Kim outside the Chemistry Department inSmith Hall.

Professor Uwe Kortshagen of the MechanicalEngineering Department on energy-resolvedmodeling of three-dimensional inductivelycoupled plasmas. Together, they are develop-ing a predictive, kinetic plasma model. Themodel will address inductively coupledradiofrequency plasmas in three-dimensionalgeometry. The requirement of the inclusion ofkinetic effects of plasma electrons will intro-duce a fourth dimension—the electron energy.The code will address a steady-state plasma. Itwill couple, in a self-consistent scheme, thefour-dimensional Boltzmann equation, thethree-dimensional wave equation, and thethree-dimensional Poisson equation. The codewill be developed as parallel code using themessage passing interface protocol.

Thomas Ihle, from the University ofMinnesota, is working with Professor H. TedDavis of the Chemical Engineering andMaterials Science Department on LatticeBoltzmann methods for simulating self-assembling fluids and macromolecular aggre-gates. In self-assembling fluids, correlatedmesoscopic structures exist even far fromcritical points. This structure leads to anom-alous scaling behavior in the dynamic struc-ture factor as well as anomalies in the shear

viscosity and the attenuation and dispersionof sound. In flow, the deformation of thesecorrelated domains gives rise to excess stressesthat result in rheological behavior quite dif-ferent from that of simple Newtonian fluids.Near phase boundaries, shear can lead todynamical instabilities and new structures inthese systems. This project is obtaining a bet-ter understanding ofthe dynamics of com-plex liquids and thedynamical behaviorof polymers andmembranes in sol-vent. An essentialcomponent of thisresearch is developingand implementingnew algorithms basedon generalizations ofthe lattice-Boltzmannmethod for thenumerical simulationof the equilibriumand non-equilibriumdynamics of thesesystems.

■

Thomas Ihle works in his office at theSupercomputing Institute.

Research ScholarshipsSince 1995, forty-three research schol-

ars have worked with twenty-nineUniversity of Minnesota faculty investiga-tors on supercomputing research projects.Research Scholarships are awarded inresponse to nominations and matchingfunds provided by a University ofMinnesota faculty member. Persons inter-ested in a Research Scholarship in2000–01 should contact aSupercomputing Institute principal inves-tigator in their field to discuss the possi-bility of nomination and cosponsorship.The deadline for nominations is January15, 2000, so it is well advised for prelimi-nary discussions to begin this fall.


This summer, fifteen undergraduatestudent researchers from across thecountry served ten-week internship

appointments at the SupercomputingInstitute, including three students in theComputational Neuroscience Program. Thesestudents were selected from a pool of 105

1999 Undergraduate Summer Interns

applicants. The students worked closely withfaculty advisors on many projects.

The Supercomputing Institute SummerInternship Program, currently in its ninthyear, promotes undergraduate involvement inongoing and new research in scientific com-puting, digital technology, and visualization in

the physical, medical,and social sciencesand engineering andin new software devel-opment efforts for sci-entific computing andgraphics support forsuch research with themain goal being tocarry out useful andinteresting research.This program pro-vides an opportunityfor a challenging andenriching educationalexperience for under-graduate studentsinterested in pursuinggraduate or profes-sional education andresearch in scientificcomputing and/orgraphics.

During the sum-mer, interns partici-pated in Institutesponsored tutorialsspecific to high-per-formance computing.To conclude the sum-mer, the interns pre-sented talks open tothe entire researchcommunity. These

Coordinator of the Computational Neuroscience Internship Program,Kathleen Clinton (center), speaks with the Computational Neuroscienceinterns Aaron Miller (left) and David Liebelt (right).

Charles Park (left) works with Ian Tregellis (right), a graduate student inProfessor Thomas Jones’ Astronomy research group.


talks allowed them to share their work and togain experience making scientific presenta-tions. The program allowed the students toperform research in close collaboration withfaculty investigators and their research groupsand to discuss research with faculty members,post-doctoral associates, graduate students,and other interns with similar interests.

Project Descriptions

Sarah Alfano, a Physics major at KentState University, worked with Professor J.Woods Halley of the Physics Department.Their work measured helium collisions byplacing resistive chromium source mounts onone end of a box and titanium bolometers onthe opposite end. It was expected that if thegraph of power through the chromium sourcemounts had a single peak, the graph of powerthrough the bolometer would also have a sin-gle peak. However, this was not the case.Investigations of the effects of collisions weremade to see if they could account for observedmultiple extrema. The simulation was codedin FORTRAN and has been run one thousandtimes for one hundred particles and ten thou-sand time iterations. These simulations wererun to obtain an average in hopes of suppress-ing the magnitude of statistical fluctuation asfar as possible. From the simulations ran, itwas discovered that the working scheme wasnot as straight-forward as it was hoped.

Benjamin Miles, a Biochemistry major atIndiana University, worked with ProfessorWilliam Gleason of the Laboratory Medicineand Pathology Department. Benjamin’s workhelped develop a three-dimensional model forthe Vascular Endothelial Growth Factor(VEGF) 165 isoform. Using the macromolec-ular modeling programs INSIGHT and DISCOV-ER, the known structures of VEGF werejoined, and unknown components were mod-eled. An energy minimization calculation was

then performed on the completed structuralmolecule. Once complete, PROCHECK wasused to evaluate it. In addition to the compu-tational work in developing a model of VEGF,some preliminary experiments aimed at devel-oping a quantitative VEGF assay were done.

Sara Firl, a Chemical Engineering major atthe University of Minnesota, worked withProfessor Alon McCormick of the ChemicalEngineering and Materials ScienceDepartment. Together, they studied how mol-ecules group or cluster together using varioustools such as Monte Carlo simulation. Sarahelped with a series of FORTRAN codes used toset up a simple lattice of cubic unit cells with

Sara Firl prepares for her presentation.

Thomas Miller (left) talks with Jocelyn Rodgers(right) at the program’s kickoff reception.


Anna Saputera awaits the start of a tutorial at theSupercomputing Institute.

established periodic boundary conditions. Theenergy of the unit cell was initially calculatedusing a Lennard-Jones potential. One randommolecule was moved, and the energy was cal-culated after the move occurred. The programran until a reasonable equilibrium wasreached. An energy curve and radial distribu-tion function were plotted from the finalstructure data. Further development of thisproject will study the adsorption of xylenemolecules into the pores of compounds suchas zeolites and clatharates.

Brent Grocholski, a Physics major at theUniversity of Minnesota, worked withProfessor George Wilcox of the PharmacologyDepartment. Brent and Professor Wilcoxworked on the development of improved neu-ronal simulators for sensory nerves. Brenthelped write a program that used theHodgkin-Huxley model and a fourth orderRunge-Kutta integrator to simulate neuron

action potentials. Aprogram was thenwritten to bothapproximate theJacobian for theHodgkin-Huxleyequations and per-form an LU factoriza-tion used to solve avector used inDASPK (DifferentialAlgebraic Solver withPreconditionedKrylov methods). Thepreconditioningmatrix failed toreduce the number ofiterations needed tosolve a time coursewhen the fullJacobian was used,but setting the firstrow of the Jacobian tozero reduced the

number of iterations required to solve thetime course by a significant amount.

Charles Park, a Mathematics and Physicsmajor at Yale University, worked withProfessor Thomas Jones of the AstronomyDepartment. Charles helped produce synthet-ic astronomical observations of magnetohy-drodynamical simulations of the jet flow inradio galaxies. These synthetic observations areuseful for studying the jets because they allowthe study of emissions from relativistic elec-trons in the jet and the test of reliability ofactual observations of radio galaxies and theproperties that observational astronomers infer

Summer 1999 Tutorials

64-bit Parallel MPI and Math/Numerical Libraries for the IBM SP

Code Optimization Workshop

Introduction to InsightII/Discover

Introduction to Molecular Animation

Introduction to Parallel Programming

Introduction to Parallel Programming on

the IBM SP Supercomputer

Introduction to Parallel Programming on

the SGI Origin 2000 Supercomputer

Introduction to Perl

Introduction to the POWER3

Architecture and Code Tuning Guide

Introduction to Scientific Visualization

Introduction to Shell Programming

Introduction to the Supercomputing Institute

Molecular Visualization Tools

Amy Beukelman (left) discusses her research with aresearcher at the Supercomputing Institute.


from radio telescope images. The syntheticobservations were produced by a raytracingprogram. By using standard astronomical toolsto analyze the synthetic images, results ofthree-dimensional simulations were mademore accessible to observational astronomers.The synthetic images suggested many of thequantities radio astronomers had infered fromactual observations. Other values were foundto rely noticeably on the orientation of theobject with respect to the observer.

Thomas Miller, a Chemistry andMathematics major at Texas A&M University,worked with Professor Donald Truhlar of theChemistry Department. Together with gradu-ate student Michael Hack, they implementedand tested the Army Ants (AA) method, a sta-tistically improved form of the fewest switchestrajectory surface hopping method of calculat-ing reaction dynamics for three-atom, two-state systems. A number of analysis programswere written to interpret the new output, anda Monte Carlo simulated trajectory surfacehopping program was written to optimize thecode’s various input parameters. The programswere applied to the collision of an electronical-ly excited bromine atom with a hydrogenmolecule, a problematic system with very weakcoupling between the two potential surfaces.

Jocelyn Rodgers, a Chemistry and Physicsmajor at Harvard University, worked withProfessor Truhlar on optimizing moleculargeometries via multi-level linear combinationsof electronic structure calculations. Recently,Professor Truhlar’s group has developed severalsingle-point energy methods that yield moreaccurate energies without a prohibitiveincrease in computer time. This projectexpanded these linear combination methodsto include multilevel optimization of molecu-lar geometries. A code called MULTILEVEL,written in FORTRAN90, was developed to cal-culate the energy, gradient, and Hessian (amulti-dimensional matrix of second deriva-tives) for a molecule with the multilevel meth-ods. A few molecules have been optimizedwith the methods in MULTILEVEL, and themethod showed great promise.

Christine Tratz, a Chemistry major at theUniversity of Oklahoma, worked withProfessor Truhlar and graduate student PattonFast on multi-coefficient correlation methods(MCCM) and multi-coefficient GAUSSIAN

methods (MCGx) for the efficient and accu-rate calculation of potential energy surfaces.Investigations were made on the reliability ofthese methods for types of compounds towhich they have not previously been applied.

Joseph Cooley (left) and Brent Grocholski (right)await a tutorial at the Institute.

Professor William Gleason (left) with internBenjamin Miles (right) before Benjamin’s talk.


The methods were improved by optimizingthem over a larger training set. In order to cre-ate a larger test set for the MCCM andMCGx methods, a number of GAUSSIAN jobswere run. Spreadsheets that provide the elec-tronic energies were created with the results ofthe GAUSSIAN calculations. The errors in thedifferent methods were then compared.Coefficients for each of the methods werereoptimized and modified. The meanunsigned error in the new methods wereshown to be considerably lower than in theprevious methods for all cases—before andafter reoptimization.

Aaron Miller worked with ProfessorTimothy Ebner of the Neurosurgery andPhysiology Department. Aaron’s work waspart of the Computational NeuroscienceInternship program. Aaron and ProfessorEbner performed experiments that investigat-ed whether the visuomotor pursuit trackingerror in monkeys would be a function of tar-get width and speed. A female rhesus monkeywas shown to use a two-joint manipulandumto make visually guided arm-tracking move-ments in the horizontal plane. For each suc-cessful trial, x and y position points wererecorded, smoothed, and digitally differentiat-ed to obtain speed points. Position error wascalculated as the difference between the x andy positions of the hand and target. Speed errorwas calculated as the difference between thespeed of the hand and the target. Positionerror was not found to change significantlywith changes in target size and speed. Speederror was not found to change significantlywith target size, but did increase significantlywith increasing target speed. Another experi-ment, in which the target underwent anabrupt two-fold change in speed, found anobservable transient increase in speed error.

Karis Stenback worked with ProfessorRobert Miller. Karis’ work was also part of theComputational Neuroscience Internship pro-gram. Together, these researchers worked onelucidating the mechanism by which activedendritic spikes (spike initiation) occur,observing how they propagate once they arepresent, and investigating the functional dif-ferences and attributes of active dendritic spik-ing versus active somatic spiking. Both anequivalent cylinder (EC) model, used toexhibit amacrine cell properties, and a real cellmodel, created from a trace of an on-offamacrine cell obtained during experiments,were used. Using the parameters from the ECamacrine cell model in the real cell model,alpha synapses were dispersed throughout thedendritic tree in proximal locations. Multiple

Thomas Wilson prepares his work in one of theoffices at the Supercomputing Institute.

Sekar Velu (left) and Charles Park (right) before atutorial.


voltage recording sites were placed throughoutthe dendritic tree and in the soma to elucidatewhere dendritic spikes were initiated. To lookat the functional differences and attributes ofdendritic spiking versus somatic spiking, onlythe calcium area was looked at. It was foundthat the intracellular calcium concentrationshowed a ten-fold increase when the spikemechanism was present and showed almost nochange when the spike mechanism was absent.

Amy Beukelman, a Chemistry and Biologymajor at the University of Minnesota, workedwith Professor Christopher Cramer of theChemistry Department. Together, theseresearchers studied singlet-triplet gaps andtransition states. After gas-phase calculationswere performed, calculations were carried outin solvents, as the methanol environment cho-sen more closely resembles experimental envi-ronments. Because some of the structures cho-sen were ions, it was important to look at thestructures in solution, where ionic forms ofmolecules are more favored. Single point ener-gy calculations of the optimized structureswere then performed using the AMSOL programand methanol as a solvent. The GAMESOL pro-gram was used (with methanol as a solvent) toobtain geometries of the molecules in solvent.While the triplet was found to behave classical-

ly, the singlet ion rearranged in the gas phase.The singlet-triplet gap was actually quite largeindicating that intersystem crossing betweenthe two related molecules was unlikely.

Sekar Velu, a Computer Science andChemistry major at Syracuse University,worked with Professor Douglas Ohlendorf ofthe Biochemistry, Molecular Biology, andBiophysics Department. Sekar helped build acomputer program to take two undocked pro-teins and accurately predict their dockedstructure. The project involved writing a shapecomplementarity program in C. This programutilized an earlier algorithm in order to gener-ate a correlation coefficent between two pro-tein surfaces. It was then decided that the bestapproach to take in determining the structurewas through the use of the Fourier Transformmethod. Extensive tests were run on FTDOCK

to determine the most effective parameters.The generated complexes were then testedagainst the crystallographically determinedstructure. Electrostatics were found to play adiminutive role in complex conformation, andsolvent accessibility was found to be a muchmore important factor. Shape complementari-ty was found to play an important rolebecause there appears to be an ideal value for aspecific protein-protein complex.

Sara Alfano discusses her work.

Computational Neuroscience interns Aaron Miller(left) and Karis Stenback (right) await the start ofpresentations.


Christine Tratz (left) and Maegan Harris (right), anundergraduate researcher in Professor Truhlar’sChemistry group.

Joseph Cooley, a Computer Science majorat the University of Minnesota, worked withProfessor Linda Boland of the PhysiologyDepartment. Joseph and Professor Bolandoptimized a model for ion channel gating. Amodel that simulated ion channel current flowwas programmed by allowing the model toinput biological data, performing serial andparallel simulations, and normalizing simula-tion data with biological data. Multiple simu-lations were then performed via parameterchanges, and the resultant simulation with theleast error in comparison to biological datawas chosen. Finally, multi-platform develop-ment was done. Additional work on themodel added more optimizations and features.

David Dreytser, a Chemical Engineeringand Chemistry major at the University ofMinnesota, worked with Professor DavidThomas of the Biochemistry, MolecularBiology, and Biophysics Department. Theyworked on computational simulations of elec-tron paramagnetic resonance (EPR) onPhospholamban, a 52 amino acid integralmembrane protein of the sarcoplasmic reticu-lum. The three-dimensional structure of phos-pholamban was extended into the cytoplasmicdomain. Initially, a script was written to con-vert formats, and a spin label was built. Thisspin label was attached to various binding

sights on phospholamban before moleculardynamics were performed.

Anna Saputera, a Computer Science majorat the University of Minnesota, worked withProfessor M. Germana Paterlini of the Collegeof Pharmacy. Anna and Professor Paterlini cal-culated the electrostatic free energy of interac-tion of an opioid peptide as a fuction of dis-tance and orientation from a lipid bilayer.Anna helped create a C++ program to rotateand transform the position of the peptidemolecule with respect to the lipid. She thenused MIDAS for visualization of the systembefore and after the coordinate transforma-tion. The electrostatic potential energy of themolecule/lipid systems were calculated at dif-ferent translation and orientations and withdifferent input parameters for DELPHI to checkthe performance of this program. The poten-tial energy between molecule and lipid wasthen calculated as a function of distance. Theprofile showed an atractive interaction energybetween the lipid and the peptide. The energyprofile obtained using DELPHI was checked forconvergence using different input parameters.

David Liebelt from NorthwesternUniversity worked with Professor DavidRottenberg of the Neurology and RadiologyDepartment at the VA Medical Center. ThisCody Zilverberg (left) and David Dreytser (right).


work was part of the ComputationalNeuroscience Internship Program. Davidhelped create a World Wide Web interface forthe Corner Cube Environment, which facesseveral hurdles before becoming common neu-roimaging software. The interface consists ofhtml documents displayed in an html frame-set. With JavaScript, this interface is able tostore image data and viewing preferences cho-sen by the user. With the data and preferencesstored, JavaScript generates an html form thatthe user can submit over the Internet. Theform is processed by a cgi script written inPERL. This processing involves interpreting thedata and preferences, writing data and prefer-ence files readable by the Corner CubeEnvironment, running the Corner CubeEnvironment, and displaying the output tothe World Wide Web browser.

Cody Zilverberg, a Computer Sciencemajor at St. John’s University, worked withProfessor David Lilja of the Electrical andComputer Engineering Department on per-formance visualization tools for java programs.Coding was done on the ‘Find Next’ compo-nent of a program called JaViz. JaViz visuallydisplays a tree of nodes where each node rep-resents a method call in a Java program. ‘FindNext’ contains a simple graphical user inter-face (GUI) that allows users to find the nextnode in the JaViz tree that matches certaincriteria. Because there can be many thousandsof nodes in a single tree, an algorithm neededto be developed that would read node datafrom a file, examine the data, and then (formemory purposes) discard unnecessary data.This search was implemented by using a back-ground thread that allowed the visualizer tocontinue functioning during execution.

Thomas Wilson, a Computer Science andArchitecture major at the University ofMinnesota, worked with Professor DavidYuen of the Geology and GeophysicsDepartment on a two part project. The first

part of the project consisted of a World WideWeb presentation focused on scientific visual-izations related to nuclear waste transporta-tion and processing at the Hanford,Washington nuclear waste reserve. This reportconsists of critical reviews and improvementsmade to the visualizations. The second part ofthe project involves Java. Tom helped imple-ment a generic Java applet control panel thatincluded useful features for the Java basedpV3 graphical user interface (GUI). The fea-tures included mathematical operations initi-ated by a mouse click, writing to the standardoutput, and providing an echo area in thewindow. Overall, this work found Java to be agood language to use for the new version ofpV3 because of security, functionality, porta-bility, and the GUI widgets needed.

■

Summer 2000 ProgramThe Supercomputing Institute is pleased to announce its

Undergraduate Internship Program for Summer 2000. Summerappointments will be full-time, ten-week appointments. The 2000program will run from June 12 through August 18, 2000. Thestipend is $4,000 for ten weeks. A student interested in becoming anintern must be an undergraduate student at the time of the intern-ship to be eligible and must be a citizen or permanent resident of theUnited States and its possessions.

All applications are judged competitively based on the qualifica-tions of the applicant and the availability of a suitable project.

Application forms and project lists are available on the WorldWide Web at:

www.msi.umn.edu/general/Programs/uip/uip.html

Application forms and project lists are also available by contacting:

Undergraduate Internship CoordinatorUniversity of MinnesotaSupercomputing Institute1200 Washington Avenue SouthMinneapolis, Minnesota 55415-1227

Phone: (612) 626-7620Email: [email protected]


The universe isawash withrelativistic

charged particles,mostly ions, butincluding a distinctiveelectronic compo-nent, as well. Evenunder the protectingblanket of the earth’smagnetic field andatmosphere we areconstantly bombard-ed with this natural“cosmic-rays” radia-tion. The most ener-getic particles, whichcan carry almost asmuch energy as aNolan Ryan fastball, come from well outsidethe solar system. However, the sun, and eventhe earth, get into the act of producing thelowest energy cosmic-rays, which fill thefamous van Allen Radiation Belts.

Advanced Computation of the Physics of Supersonic Cosmic Flows

The higher energy cosmic-rays are thoughtmostly to be a byproduct of strong “collision-less” shock waves in tenuous cosmic plasmas.Shocks are central features in violent cosmicexplosions, such as supernovas, but also in

much more energeticphenomena, such asquasars, and even thegravitational collapseduring formation ofgalaxies and clustersof galaxies throughoutthe history of the uni-verse itself. The last ofthese shock waves aremillions of light yearsacross and may existfor billions of years.

Research over thepast two decades hasmade it very clearthat cosmic shocks,

Simulated radio and X-ray emission from the cosmic jet flow shown above, asit would be seen by a terrestrial astronomer. The right image corresponds tosynchtrotron emission resulting from interactions betwen relativistic electronsand the local magnetic field, while the left image represents cosmic microwavebackground photons upscattered to X-ray energies by the same relativistic elec-trons.

Pictured is a three dimensional simulation of a cosmic supersonic jet. Theright image shows a volume rendering of shock waves formed inside the jetand in the cocoon it leaves behind. The left image illustrates the energeticproperties of cosmic-ray electrons in this flow. Color patterns indicate theinfluence of shocks on the electron energies.


and the cosmic-rays they expel, are funda-mental to a host of astrophysical issues. Thecosmic-rays are important because they canabsorb a significant fraction of the energybudget at a shock, while the energy distribu-tion of the cosmic-rays provides a measure ofthe shock properties and offers an incisiveprobe of the physics of the collisionless shockwaves themselves. The cosmic-rays emit lightwith distinctive properties so that they offerunique diagnostic possibilities for many cos-mic phenomena. While these features aregenerally accepted, they are poorly under-stood. Yet, their importance provides a strongincentive to develop comprehensive theories.The theoretical task is so complex, especiallyin application to real cosmic events, thatsophisticated computational methods areabsolutely essential for success.

Supercomputing Institute Fellow ProfessorThomas Jones of the Department ofAstronomy at the University of Minnesotaalong with research associate andSupercomputing Institute Research ScholarUdo Gieseler, graduate students FrancescoMiniati and Ian Tregillis, and Korean collabo-rators Professor Hyesung Kang of the PusanNational University and Professor DongsuRyu of the Chungnam National Universityhave undertaken the task of developing andapplying computational methods to overcomemany of the biggest hurdles that must beovercome along the way. They have severallinked efforts underway that have alreadyborn exciting fruit. It has been necessary topioneer computational methods to modelessential microphysics in collisionless shockedflows, high order methods to compute thedynamics of compressible, conducting fluidslaced with dynamically significant magneticfields, methods to transport relativisticcharged particles in complex conducting plas-

ma flows, and methods to compute theexpected radiative emissions from the rela-tivistic particles and simulate how the simu-lated cosmic structures would appear to anearth-bound observer. While there is stillmuch work to be done in these areas, theseresearchers are now at the exciting stagewhere they can begin to challenge existingobservational paradigms and make predic-tions about observations possible with thenext generation of space observatories.Current efforts range from developing com-putational schemes to model the complexmicrophysics that “injects” thermal chargedparticles into the cosmic-ray population andincorporating “adaptive mesh refinement”techniques for the first time into numericaltransport of cosmic-rays to simulating theobservable properties of cosmic-rays expelledin supersonic plasma jets expelled from thecenters of galaxies and cosmic-rays acceleratedin the powerful shock waves stretching acrossand around giant clusters of galaxies.

■

The simulated energy distribution of protonsbehind a collisionless shock wave showing thedevelopment of a high energy “cosmic-ray tail”above the thermal distribution.


Flow and transport phenomena inporous media are of importance in awide range of natural and industrial

processes. The flow of water containing nutri-ents or pollutants in soils and aquifiers is ofenvironmental concern. Of commercial rele-vance is the multiphase flow of oil and waterin porous and fractured rock, as well as flowand dispersion in packed beds used for cat-alytic and separation processes.

In order to model these processes, it isnecessary to understand the pore scale mech-anisms of flow and transport. This requiresan understanding of how the structure of theporous media affects the distribution, flow,and displacement of one or more fluid phasesas well as the dispersion (i.e., mixing) of onefluid in another. Recent advances in under-standing these phenomena are due, in largepart, to the refinement of experimental tech-niques such as pulsed field gradient spin-echo Nuclear Magenetic Resonance (NMR)and the development of new computationalmethods and technologies.

When two miscible fluids are brought intocontact, a transition zone develops across theinterface, and the two fluids slowly diffuse intoone another. However, if the two fluids are alsoflowing, there will be some additional mixing.This mixing is generated by a non-uniformvelocity field, which may be caused by themorphology of the medium, the flow condi-tions, or chemical and physical interactionswith the solid surfaces of the medium; thiseffect is called hydrodynamic dispersion. Therelative importance of diffusion and hydrody-namic dispersion in spreading the transitionzone is described by the Péclet number Pe =vd/Dm, where v is the mean pore velocity, d isthe average grain size of the media, and Dm isthe relevant molecular diffusion constant. The

Flow and Transport in Porous Media

larger the Péclet number, the greater the influ-ence of hydrodynamic dispersion. In most ran-dom porous media, the dominant contributionto dispersion at large Péclet numbers comesfrom random spatial variations of the velocityfield as shown in figure 1.

One approach to the analysis of pore scaleflow and transport is to solve the Navier-Stokes equation and mass-transport equa-

Figure 1: Mechanical dispersion is caused by ran-dom spatial variations of the velocity field. Particlesin close spatial proximity at t = 0 (o) are dispersedat later times (•) as they flow along divergingstream lines.

Figure 2: Color coded flow structure in a simplecubic bead pack. High-velocity flow regions arecolored yellow and white. The red regions arerecirculation zones.


tions in the pore space. An alternativeapproach has been developed by researchersat the University of Minnesota under thedirection of Supercomputing Institute FellowRegents’ Professor H. Ted Davis of theChemical Engineering and Materials ScienceDepartment and IT Dean, Robert Maier ofthe Army High Performance ComputingResearch Center, and Professor Daniel Krollof the Department of Medicinal Chemistry.These researchers are working in collabora-tion with United States Army Corps ofEngineers Waterways Experiment Stationresearchers Robert S. Bernard, Stacy E.Howington, and John F. Peters.

This approach simulates flow using the lat-tice-Boltzmann (LB) method. A discretizedversion of the Boltzmann kinetic equation issolved for the single particle distribution func-tion of the fluid. A random-walk particletracking method is then used to simulate trac-er dispersion using the LB flow fields.Computer codes for the LB method are rela-tively simple, and the method lends itselfquite naturally to parallel processing. Usingthis approach, high-resolution studies of flowand transport in random bead packs havebeen performed. The velocity distribution in acubic packing of monodisperse spheres isshown in figure 2, where it can be seen thatno-slip boundary conditions at the bead sur-faces results in most flow occuring in high-velocity channels. A comparison of simulationresults for dispersion in random porous mediawith recent NMR spectroscopy studies showagreement on transient, as well as asymptotic,dispersion rates. In particular, the results sup-port NMR findings that longitudinal disper-sion rates are significantly lower than reportedin earlier experimental literature and thatasymptotic rates are achieved earlier than pre-

viously reported. Results for the longitudinaldispersion constant, DL, as a function of thePéclet number are compared with recentNMR data in figure 3.

Related work involving the transport ofsolute in packed bed liquid chromatographyand electrochromatography is being per-formed in collaboration with Dr. Mark Schureof Rohm and Haas Company, who is current-ly an Industrial Fellow at the University ofMinnesota’s Center for InterfacialEngineering. Dr. Thomas Ihle, aSupercomputing Institute Research Scholar, isdeveloping LB methods for treating multi-phase immiscible flow and transport.

Future work will involve studies of multi-phase flow. A number of issues need to beaddressed, but one of the most importantinvolves nonaqueous-aqueous mass transfer. Inaddition to its obvious importance in advancedoil recovery processes, it is also of great interestfor environmental studies since nonaqueousliquids are a frequent source of contamination.

■

Figure 3: Comparison of simulation results (•) forthe longitudinal dispersion coefficient withNuclear Magnetic Resonance data (◊).


Certain wide band gap semiconductorsare emerging as viable candidatematerials for electronic devices that

can operate in environments and under biasconditions for which the standard semicon-ductor material, silicon, is not suitable.Among the new materials are the III-nitrogencompounds, GaN, AlN, and their alloys. Theprincipal advantage of these materials over sili-con can be traced to their electronic structurethat has (filled) valence bands separated from(empty) conduction bands by an energy gapthat is—depending on the group III ele-ment—between three and five times as largeas that of silicon. The wide band gap impliesthat large electric fields are sustainable in thematerials and the principal electrical character-istics are relatively insensitive to temperature.In addition, conduc-tion electrons can bevery mobile, allowingfor large current den-sities and rapidchanges in currents.This combination ofproperties makes III-nitride semiconduc-tors excellent candi-dates for the fabrica-tion of high-power,high-frequency, high-temperature electronicdevices with applica-tions ranging fromsatellite communica-tion systems to enginecontrol electronics.

Materials-Theory Based Modeling of III-Nitride SemiconductorDevices

III-nitride materials preparation technologyis currently in a state of very rapid develop-ment, and exploratory devices are alreadybeing made. However, device design is stillrather rudimentary. One of the main reasonsfor this relatively low level of sophistication isthe lack of predictive models needed for suc-cessful optimization. The standard devicemodels are not immediately useful becausethey rely on a large number of materialsparameters that, although well known for con-ventional materials such as silicon, are not yetavailable for the III-nitride compounds.

Supercomputing Institute Associate FellowProfessor Paul Ruden and his Modeling andSimulation group in the Electrical andComputer Engineering Department, togetherwith Professor Kevin Brennan and students at

Figure 1: Average drift velocity of electrons in wurtzite structure GaN as afunction of the applied field strength for three different temperatures asobtained from Monte Carlo simulations.


the Georgia Instituteof Technology, havedeveloped a materials-theory based devicemodeling techniquethat is particularlysuitable for the cur-rent state of III-nitride technologydevelopment. Brieflysummarized, the tech-nique starts from veryfundamental materialsproperties that areknown from parame-ter-free, first-princi-ples calculations ofthe electronic struc-ture. It then proceedsto microscopic elec-tron transport simula-tions using the MonteCarlo method. Inthese simulations, electrons are accelerated byan applied electric field and scattered throughinteractions with lattice vibrations and impu-rities. From the resulting trajectories, whichmay involve 107 scattering events, the averagedrift velocity for a given field strength is com-puted. Representative results are displayed infigure 1. Finally, based on the microscopicsimulation results, macroscopic materials char-acteristics of suitable forms are defined andincorporated into device models.

The materials-theory based device modelshave yielded predictions of breakdown volt-ages for simple GaN diodes as well as output

Figure 2: Calculated (lines) and measured (diamonds) output characteristics ofAlGaN/GaN heterostructure field effect transistor. The gate voltage variesbetween 0V (top curve) and –3V (lowest curve) in 1V steps. The largedecrease in current with increasing drain voltage arises from self-heating of thedevice. (Experimental data courtesy of Naval Research Laboratory).

performance results for relatively complexAlGaN/GaN Heterostructure Field EffectTransistors (HFETs). Temperature effects havebeen found to be very important in thesehigh-power devices due to inevitable self-heat-ing. As an example, figure 2 shows modelresults for the output characteristics of anHFET, together with experimental datarecently obtained at the Naval ResearchLaboratory. Further modeling results indicatethat even higher currents can be obtained withsimple changes in device design. These newdesigns are presently being explored.

■


S taphylococcus aureus is an ubiquitoushuman pathogen. Approximately 30%of healthy individuals are colonized by

S. aureus—typically in the nasal passages, inthe vagina, or in the perianal area. In theUnited States, S. aureus is the major cause ofnosocomial (acquired in a hospital) infectionsresulting in over 700,000 infections annually.Staphylococci can produce a variety of diseasesranging from food poisoning to toxic shocksyndrome and sepsis.

The advent of antibiotics in the early partof this century was a pivotal development inthe treatment of infectious diseases caused bygram positive bacteria such as S. aureus.Before antibiotics, mortality from gram posi-tive sepsis was in excess of 80%. The use ofpenicillin during World War II saved manylives by curing sepsis due to the infections ofwounds. However, in 1942, penicillin-resist-ant strains of S. aureus were already being

reported. In the 1950’s several serious out-breaks of staphylococcal infections by peni-cillin-resistant strains had been reported. By1997 resistance to penicillin has risen to over90% of isolates of S. aureus. A new β-lactamantibiotic, methicillin, came into clinical usein 1960. One year later, methicillin-resistantisolates were reported. Today, methicillin-resistant S. aureus (MRSA) are found in near-ly 40% of clinical isolates. The aminoglyco-side antimicrobial agent vancomycin is fre-quently the only drug available to the clini-cian for treatment of severe infections causedby MRSA strains. In 1997, the first clinicalisolates of vancomycin-resistant S. aureuswere reported. The specter of infections frommultiply-resistant strains of gram positiveorganisms looms menacingly over the futurehuman health. Recognition of this problemhas led to the establishment of a network ofinternational monitoring.

Structural Studies of Toxins Produced during Staphylococcal Infections

Path of peptide backbones of toxic shock syndrome toxin-1 (TSST-1) and exfoliative toxin A (ETA).


In order to develop novel strategies for pre-venting and treating infections by S. aureus,workers in the laboratory of SupercomputingInstitute Fellow Professor Douglas Ohlendorfare determining and analyzing the structures ofa number of proteins produced by S. aureusthat have been identified as virulence factors.Included among these proteins are a variety oftoxins, proteases, lipases, and nuclease. Bydetermining the structure of these molecules, itis possible to locate features required for bio-logical activities. Mutation of these areas canproduce molecules which can potentially beused as immunizing agents. For those enzymevirulence factors, the structure can be used todesign inhibitors which block the action ofthese molecules. This approach, for example,has allowed the production of proteaseinhibitors recently coming into use in thetreatment of AIDS.

Workers in Professor Ohlendorf ’s laborato-ry have been initially focusing on two toxinsthat have activity as superantigens. The defini-tion of a superantigen is that it functions bybinding to the foreign antigen-presentingmolecule (class II major histocompatibilitycomplex) and to a specific serotype of receptoron circulating killer T cells. In an infectionproduced by a superantigen, there is a pro-found proliferation and then death of the Tcells carrying the serotypical marker on theirsurface. Professor Ohlendorf ’s group hasdetermined the structures of toxic shock syn-

drome toxin-1 (TSST-1) and of exfoliativetoxins A and B (ETA and ETB; see figure).TSST-1 is the key causative agent of the severemultisystem disorder toxic shock syndromethat gained notoriety in the 1970’s with itsassociation with usage of superabsorbant tam-pons. ETA and ETB are key agents in staphy-lococcal scalded skin syndrome—a typicallynonfatal condition in children in which largeareas of the skin blister and peel.

TSST-1 folds into a kidney-shaped 2domain structure—one having five β strandsin a barrel and the other have a long α helixlying against a sheet of four β strands. Thisfold has been also seen in the staphylococcalenterotoxins as well as in some of the strepto-coccal pyrogenic exotoxins. ETA (as well asthe homologous ETB) also fold into twodomains. However, the central element ofeach of these domains is a six strand β barrel;the axes of the two barrels are roughly perpen-dicular to each other. This fold is well knownas that of the serine proteases. In fact, the ETshave a traditional active site, which if mutat-ed, blocks the exfoliative activities while main-taining their superantigenic properties. Themystery is how molecules with different foldsfunction as superantigens. This question isbeing studied computationally by calculatingthe congruence of molecular surfaces. Oncesimilar surfaces have been identified, mutantscan be made, expressed, and analyzed.

■


The magneticstructure ofthin magnetic

films has long been asubject of importancefor information storagemedia, sensors, andtechnological advancessuch as magneticRAM. In many appli-cations, and for mod-eling and simulations,films are composed ofsmall magnetic parti-cles that interact witheach other throughdipolar and exchangecoupling. These films can be pictured as anarray of tiny magnets whose magnetization canpoint in any direction in three dimensions.

Experimentally, the magnetic structure isstudied by magnetic force microscopy (MFM),

in which a tiny magnetized stylus is broughtinto proximity to the surface of the film andthen rastered over the film. The magnetic forcebetween the film and the stylus is then inter-preted to elucidate the magnetic structure.

Physics Professor E.Dan Dahlberg,Director of theMagnetic ForceMicroscopy Center(MMC) at theUniversity ofMinnesota, has helpeddevelop and employsensitive MFM tostudy these films andmagnetic particles.

Useful insight andinterpretation of thesemeasurements areobtained by modelingthese systems and sim-ulating their magnetic

Figure 1: A model of a 1 × 2 micron permalloy film in differing magneticstates. Saturation in the easy action direction (left), final state after relaxationin zero applied magnetic field at a temperature of 300 K (middle), and as inthe middle but at temperature 100 K (right).

Figure 2: Total magnetization as a function of applied field for the film of fig-ure 1, beginning in a saturated state at applied field +600 Oe (as in Figure 1left), and then stepping the field down to –600 Oe in steps of 10 Oe(topcurve) at reverse saturation, and finally returning the applied field of +600 Oenear the original saturated state (bottom curve).

Simulations of Submicron Magnetic Structures for DigitalTechnology Devices


structure and the evolution of that structure asa function of time, temperature, and appliedmagnetic field. Physics Professor, Charles E.Campbell, an Associate Fellow of theSupercomputing Institute and member of theMMC, graduate student Andrew B. Kunz, alsoa member of the MMC, and ProfessorDahlberg are developing and implementingsimulation methods to study these systems.They make use of both micromagnetic meth-ods, wherein the dynamical Landau-Lifshitz-Gilbert equations (LLG) are solved at zerotemperature, and Monte Carlo simulationtechniques with a Metropolis algorithm tostudy both the Monte Carlo time and temper-ature dependence of the relaxation. MonteCarlo simulations, approximated by the evolu-tion of the system as a function of MonteCarlo steps, show that the difference betweenroom temperature and zero temperature is veryimportant for some systems being simulated.

One set of simulations is for a model ofmicron size permalloy films with a thickness of20 nm that consists ofa two-dimensionalsquare mesh of grains.Each grain occupies asize of 50 × 50 × 20nm and has an associ-ated magneticmoment allowed torotate fully in threedimensions. Thepermalloy is character-ized by a uniaxialanisotropy situatedalong one of the prin-cipal axes of the filmof 5000 erg/cc—anearest neighborexchange of 1.3 10-6

erg/cm, and a satura-tion magnetization of800 emu/cc. Figure 1

(left) shows a 1 × 2 micron film whose magne-tization was initially saturated in the longdirection of the film (taken to be the easy axis)and permitted to relax in the absence of anexternal magnetic field. The final state at roomtemperature, figure 1 (middle), has a complexmagnetic structure in which the magnetiza-tions at the ends have rotated to parallel theends in order to reduce the demagnetizationenergy, and vorticity has developed in the inte-rior to lower the dipolar and exchange ener-gies. Even though there is no applied magneticfield, there is a residual net magnetization inthe direction of the original saturation. Theouter edges have preserved the original direc-tion, while the interior has a significant regionof reversed magnetization. This is a metastablestate; there is at least one configuration withlower energy which has higher symmetry, butwhich is unlikely to be reached in an experi-mental relaxation. Figure 1 (right) shows theresults of the simulation at 100K, demonstrat-ing an apparent strong dependence on temper-

Figure 3: Magnetic states at the applied fields of –60 Oe (left) and –70 Oe(right) on the demagnetization curve (top curve in Figure 2).


ature. Simulations of the same system have been

carried out as a function of magnetic field.Figure 2 is a hysteresis graph. Beginning at ahigh positive field along the long axis, the ini-tial state is the same as Figure 1 (left). Thefield is first reduced to 600 Oe, and then insteps of 10 Oe until -600 Oe, where the mag-netization is well-saturated in the oppositedirection of the initial state. The magnetizationas a function of applied field during this rever-sal is the top curve. At -600 Oe, the field isincreased in steps of 10 until it reaches 600.The lower curve is the magnetization as afunction of field as the field is increased from–600 Oe to +600 Oe. The large discontinu-ities in the hysteresis curves are a real physicaleffect, due to a sudden move from onemetastable configuration to a significantly dif-ferent configuration during a very smallchange in the field. One such shift in magneticstructure change is shown in Figure 3 for theapplied magnetic fields of -60 Oe to -70 Oe.

The magnetic structure of mesoscopic sizedparticles is a subject of growing interest at thefundamental level. MFM measurements onnickel cylinders of 100 to 300 nm diametersand 100 nm high indicate that the smallerparticles are single domain magnetic structureswhile larger diameter particles have a complexmagnetic structure. Andrew Kunz has simulat-ed such particles using the micromagneticLLG method, obtaining semiquantitativeagreement with the MFM results. Figure 4shows the magnetic structure of the top sur-face of a 200 nm cylinder where the initialmagnetic configuration was uniform satura-tion parallel to the cylinder’s axis. This isshown through false color images showing the

direction of the x, y, and z components of thetop surface, where y is the direction perpendi-cular to the page. The middle panel, showingthe y component, is very similar to the MFMimage. Since this is an LLG simulation, it is ata temperature of absolute zero.

The incorporation of finite temperatureeffects in the LLG method is a high prioritychallenge for the future of these simulations.These methods will also be applied to thestructure and stability of magnetic bits onmagnetic thin films and the more esotericproblem of spontaneous reversal of magnetismvia quantum mechanical tunneling.

■

Fig. 4: Magnetic structure of the top of Ni cylin-ders. An array of cones in the direction of magneti-zation is shown.


1999 International Conference on Preconditioning Techniquesfor Large Sparse Matrix Problems in Industrial Applications

The “1999 International Conferenceon Preconditioning Techniques forLarge Sparse Matrix Problems in

Industrial Applications” was held at theHubert H. Humphrey center in Minneapolis,Minnesota on June 10–12, 1999. The confer-ence was sponsored by the University ofMinnesota Supercomputing Institute forDigital Simulation and AdvancedComputation, the University of Waterloo, andthe Lawrence Livermore National Laboratory.Eighty-seven participants came from all overthe world—39 international participants fromAfrica, Australia, Asia, Europe, and SouthAmerica and 48 participants from the United

States and Canada. The conference was timedto coincide with the biannual “HouseholderNumerical Algebra Symposium” held inWhistler, British Columbia the week before.

The conference aimed at bringing togetherresearchers to discuss the latest developmentsin preconditioning methods and the use ofiterative procedures in scientific and engineer-ing applications. The organizing committee,Yousef Saad (University of Minnesota), DanielPierce (Boeing Company), and Wei-Pai Tang(University of Waterloo), strived for an equalarray of talks from applications and algo-rithms. This goal was reached with 15 of the32 talks falling in the applications category.

Masha Sosonkina (left) of the University ofMinnesota Duluth and Iain Duff (right) of theEuropean Centre for Research and AdvancedTraining in Scientific Computation, France.

Edmond Chow (left) of Lawrence LivermoreNational Laboratory and A. Yu. Yeremin (right) ofthe Russian Academy of Sciences.

Peter Forsyth (left)of the University of Waterloo,Canada and Martin Gutknecht (right) of ETH-Zentrum, Switzerland.

Wei-Pai Tang (left) of the University of Waterloo,Canada and Wing Lok Wan (right) of StanfordUniversity.

cism. A great deal of the progress was made infinding new inexpensive strategies for obtain-ing a good pattern for the inverse. It wasreported, for example, that taking the patternof A2 was generally sufficient to achieve theefficiency of most sophisticated techniques.

Presentations on the use of iterative methods


The opening talk, in thecategory of applications, wasgiven by Charbel Farhat fromthe University of Colorado atBoulder. His talk, “10 years ofFETI (Finite Element TearingInterconnections),” was anoverview of work done by ateam at the University ofColorado at Boulder in solv-ing a large variety of problemsin mechanics.

Peter Forsyth from theUniversity of Waterloo,Canada presented an introduc-tion to problems in computa-tional finance and discussedhis experience in dealing withthese problems. ProfessorForsyth presented some experi-ments in using iterative meth-ods for the solution of PartialDifferential Equations result-ing from the problems in com-putational finance.

Presentations on recentimprovements or applicationsto the class of approximateinverse methods were given byAlex Yeremin from theRussian Academy of Sciences,Russia, Edmond Chow fromLawrence Livermore NationalLaboratory, Michele Benzifrom Los Alamos NationalLaboratory, Jun Zhang fromthe University of Kentucky,and Thomas Huckle from theTechnical University of Munich, Germany.Overall, these talks reflected the tremendousprogress made in this area. Some of the talkspresented experiments in which these methodsare now starting to be competitive with stan-dard ILU preconditioners. A few years ago,such methods were met with much skepti-

From top left to bottom right: Venansius Baryamureeba of theUniversity of Bergen, Norway and Teresa Vespucci of the Universitádegli Studi di Bergamo, Italy; Willy Schilders of Phillips ResearchLabs, The Netherlands and Menno Verbeek and Henk van der Vorstof Utrecht University, The Netherlands; Charbel Farhat of theUniversity of Colorado, Jane Cullum of Los Alamos NationalLaboratory, and John Lewis of Boeing Computer Service; Peiru Wuof Parke-Davis Pharmaceutical Research and Yousef Saad of theUniversity of Minnesota; Andy Wathen of Oxford University, UnitedKingdom and Howard Elman of the University of Maryland; andSerge Goossens of the Katholieke Universiteit Leuven, Belgium andWim Bomhof of the University of Utrecht, The Netherlands.


in Circuit Simulation were given by WillySchilders from Philips Research Laboratories,The Netherlands, Henk van der Vorst from theUniversity of Utrecht, The Netherlands, andWim Bomhof also from the University ofUtrecht, The Netherlands. These presentationsrepresented some of the first discussions inwhich positive results have beenreported on the use of iterativetechniques in this specificapplication. The method pre-sented used one sort of inde-pendent set reordering verysuccessfully. Two other presen-tations on applications inwhich iterative methods are rel-atively new were given—one byVenansius Baryamureeba of theUniversity of Bergen, Norwayon the solution of large-scalelinear programming problemsand a second by MashaSosonkina from the Universityof Minnesota at Duluth on theuse of iterative methods forsolving problems in tire design.

A number of speakers pre-sented preconditioning tech-niques for specific applicationsor specific classes of problems.Preconditioning methods forvarious saddle-point problemsor constrained problems werepresented by Andy Wathenfrom Oxford University,United Kingdom, HowardElman from the University ofMaryland, Ilaria Perugia fromthe University of Pavia, Italy,and Matthew Knepley fromPurdue University. DanieleBertaccini from the Universityof Firenze, Italy and RaymondHonfu Chan of the ChineseUniversity of Hong Kong,

Peoples Republic of China presented precon-ditioners for P-circulant matrices and Toeplitzmatrices respectively. Uri Peskin from theTechnion-Israel Institute of Technology, Israelpresented a parallel method based onKronecker sums and products of operators foran important problem that arises in electro-

From top left to bottom right: Raymond Honfu Chan of TheChinese University of Hong Kong and Esmond Ng of LawrenceBerkeley National Laboratory; Uri Peskin of the Technion-IsraelInstitute of Technology and Michele Benzi of Los Alamos NationalLaboratory; A. Yu. Yeremin of the Russian Academy of Sciences andRobert Numrich of SGI; Herbert De Gersem of the KatholiekeUniversiteit Leuven, Belgium and Ilaria Perugia of the Univesità diPavia, Italy; Wing Lok Wan of Stanford University and RobertBridson of the University of Waterloo, Canada; and Brian Suchomelof the University of Minnesota, Maya Neytcheva of the University ofNijmegen, The Netherlands, and Lutz Grosz of The AustralianNational University.

lution approximate inverse technique.Additionally, Brian Suchomel from theUniversity of Minnesota and Lutz Grosz fromThe Australian National University, Australiadiscussed strategies using independent sets(ARMS and ILUM respectively). The meth-ods proposed by Andreas Pomp from theSwiss Federal Institute of Technology,Switzerland, which separates the unknowns intwo subsets treated differently, also fell intothis series of presentations. Esmond Ng ofLawrence Berkeley National Laboratory dis-cussed a number of variants of incompleteCholesky methods and suggested that precon-ditioning techniques should benefit fromwidely used ideas in sparse direct methods.

A number of talks were geared specificallytoward parallel implementations or parallelalgorithms. This category consisted of presen-tations by David Keyes from Old Dominion

University and ICASE at theNASA Langley ResearchCenter, Maya Neytcheva fromthe University of Nijmegen,The Netherlands, Wing-LokWan from StanfordUniversity, and SergeGoossens from the KatholiekeUniversiteit Leuven, Belgium.These presentations related tomultilevel and DomainDecomposition methods.

The conference also con-sisted of a good set of posterpresentations: twelve in thecategory of algorithms andothers and seven in the cate-gory of applications.

The talks and posters wereof high quality, and in spite ofthe rather heavy two and halfday schedule, participationand discussion were excellent.A special issue on the topic ofthe conference is currentlybeing planned.

■


optics and nanoelectronics. Herbert DeGersem from the Katholieke UniversiteitLeuven, Belgium considered Schur comple-ment type methods for solving systems arisingfrom field-circuit coupled electromagneticmodels. Martin Gander from the ViennaUniversity of Technology, Austria presented afactorization technique (in the PDE sense)specifically geared toward parabolic PDEs.Finally, methods for applications that lead todense matrices were discussed by Iain Dufffrom the European Centre for Research andAdvanced Training in ScientificComputation, France, Jussi Rahola from theCenter for Scientific Computing, Finland,and Menno Verbeek from Utrecht University,The Netherlands.

Robert Bridson from the University ofWaterloo, Canada discussed new types of pre-conditioners in his presentation on multireso-

From top left to bottom right: Matthias Bollhöfer of the ChemnitzUniversity of Technology, Germany and Thomas Huckle of theTechnical University of Munich, Germany; Charbel Farhat of theUniversity of Colorado and Willy Schilders of Philips Research Labs,The Netherlands; Matthew Knepley of Purdue University and DavidKeyes of NASA Langley Research Center; and Jussi Rahola of theCenter for Scientific Computing, Finland and Andreas Pomp of theSwiss Federal Institute of Technology.


CO

MP

UTA

TIO

NA

LB

IOLO

GY

CO

LLO

QU

IUM

SER

IES

One of thecentralquestions in

molecular biology isthe origin of the cat-alytic power ofenzymes. The rapidrevolution in struc-tural determinationand accumulation ofbiochemical informa-tion offers a uniqueopportunity of eluci-dating the detailedmechanisms of enzy-matic reactions. This,however, requires anadequate descriptionof the intermolecularinteractions that govern macroscopic biologi-cal functions. One of the successfulapproaches in computational biology is tocombine quantum mechanics with classicalforce fields, which allows the proper treat-ment of the chemical process, but still retainscomputational efficiency. In this talk,Professor Gao discussed his recent studies of

Professor Darin York (left) of the Chemistry Department at the University ofMinnesota and Professor Jiali Gao (right) at an informal lunch on the day ofthe colloquium visit.

two enzyme systems, the dephosphorylationreaction by a protein tyrosine phosphatase,and the chorismate to prephenate rearrange-ment by chorismate mutase. As a prelude tostudies of photochemical reactions in biologi-cal systems, the computational method wasillustrated by investigating the origin of theopsin shifts in bacteriohrodopsin.

■

Computational Insights on Biomolecular Interactions and Enzymatic Reactions

Professor Jiali Gao, State University of New York at Buffalo

VIS

ITOR

S


Professor Hyesung Kang of theDepartment of Earth Sciences at PusanNational University in Pusan, Korea

and Professor Dongsu Ryu of the Astronomyand Space Science Department at ChungnamNational University in Daejun, Korea visitedProfessor Thomas Jones at theSupercomputing Institute. They both workedon Simulations of Particles and Fluids inAstrophysics. This visit was an important stepin completing a collaboration that beganunder NASA support. The researchers havethe first stage under test of an innovativeAdaptive Mesh Refinement scheme for timedependent simulation of cosmic ray accelera-tion in shocks. Successful completion of thiswill open a new frontier in that field. Initialresults look excellent. Professor Kang’s visithelped refine and extend the first version ofthe code in preparation for the 26thInternational Cosmic Rays Conference meet-ing in Salt Lake City. Professor Ryu’s visit wasdevoted to refinementof a new collaborationto extend researchinto cosmologicalsimulations. ProfessorRyu has experience inthat area, whileProfessor Jones’researchers do not.The group wishes toinclude for the firsttime, the physics ofhigh-energy particleacceleration andtransport. The impor-tance of these parti-cles is now becomingquite evident.

■

A volume rendering taken from inside a numericalsimulation of the expansion of the universe show-ing the complex web of shock waves that formduring gravitational collapse of clusters of galaxies.This section is about one hundred million lightyears on a side, so an individual galaxy like theMilky Way would be a tiny dot on the same scale.

Professor Thomas Jones (left), Professor Dongsu Ryu, Professor HyesungKang, and Francesco Miniati (right) pursue their collaboration at the Institute.


VIS

ITO

RS

Professor Ying Zhang, from theDepartment of Materials Science atXiamen University in Fujian, Peoples

Republic of China, visited the University ofMinnesota to continue ongoing collaborativeresearch with Professor Roger Fosdick of theAerospace Engineering and MechanicsDepartment. These researchers have collabo-rated on problems of stress-induced coexistentphase structures from the point of view ofminimization theory. Their work involvedboth theoretical developments in nonconvexminimization (for nonconvex strain energyfunctions) and finite element computationalschemes. Motivation for this work is based onthe idea that the existence and growth ofzones of evolving microstructure plays a partnot only in the failure of materials, but also inthe design of devices and in many emergingapplications that depend on sophisticatednonlinear material behavior.

Concurrently, Professor Fosdick andProfessor Darren Mason, from theDepartment of Materials Science andMechanics at Michigan State University,began development of a thermodynamical the-ory related to their earlier work on a long-range nonlocal theory of materials. This workwas purely mechanical and included the effectof a local non-convex stored energy functionthat induced coexistent phase structures and anonlocal particle-particle interaction energythat provided for a surfacial-like energy contri-bution between phases. This exhibited definiteadvantages in modeling material behavior overthe earlier well-known gradient theoryapproach. While the original long range inter-action theory was one-dimensional and wasrestricted to the study of static states, it is cur-rently being generalized to include thermody-

namics and apply to three-dimensional con-tinua. They plan to use this generalization tostudy the time-evolution of phase structures.

Professors Fosdick, Mason and Zhang,together, worked to intensify and focus theirjoint program. Among the many importantissues is the most challenging task of develop-ing a highly efficient numerical scheme fornonconvex, spatial dependent energies inorder to search for minimal energy states andto characterize the configuration of a bodywhen many interfacial transitions are possible.Professor Zhang began to develop such ascheme during his visit. This is recognized asan important part of the program, becausedependable and efficient computational resultswill motivate a productive direction for ourfurther theoretical developments, the charac-terization of multiphase static states is nototherwise possible, and we expect that thevisualization of developing dynamical struc-tures will shed light on the importance oflocal vs. global minimizers.

■

Dr. Ying Zhang (left), Professor Roger Fosdick(center), and Darren Mason (right) meet to discusstheir work.


Dr. MurialGerbaultfrom the

Institute ofGeological andNuclear Science inLower Hutt, NewZealand visitedSupercomputingInstitute FellowProfessor David Yuenof the Geology andGeophysicsDepartment at theUniversity ofMinnesota. Dr.Gerbault’s work onfinite-element model-ing of mountainbuilding processesused ABAQUS, a pack-age available on the SP system. She workedon the phenomenon of lithospheric bucklingwithin the framework of viscoplastic rheology.She also worked with Professor Yuen’s under-graduate intern students from both theUniversity of Minnesota Department ofGeology and Geophysics program and theSupercomputing Institute program. Shehelped Liz Starin of the University of

VIS

ITOR

S

Undergraduate intern students Thomas Wilson (left) of the SupercomputingInstitute program and Liz Starin of the University of Minnesota Departmentof Geology and Geophysics program with Professor David Yuen and Dr.Murial Gerbault (right) from the Institute of Geological and Nuclear Sciencesin Lower Hutt, New Zealand

Minnesota Department of Geology andGeophysics internship program finish herwork on the effects of variable thermal con-ductivity on the thermal evolution of sedi-mentary basin. Dr. Gerbault also helpedThomas Wilson, a Supercomputing Instituteundergraduate intern, with his work onimplementing pV3, the visualization toolkitwith JAVA.

■


RES

ERA

CH

REP

OR

TS

Aerospace Engineering andMechanics

° 99/136, August 1999The Lagrange Multiplier in IncompressibleElasticity TheoryR. Fosdick and G. Royer-Carfagni

° 99/137, August 1999The Constraint of Local Injectivity in LinearElasticity TheoryR. Fosdick and G. Royer-Carfagni

° 99/176, October 1999Self-Intersection in ElasticityA. Aguiar and R. Fosdick

° 99/157, September 1999Foamy Oil Flow in Porous MediaD.D. Joseph, A.M. Kamp, R. Bai, and M.Huerta

° 99/158, September 1999Modeling Rayleigh-Taylor Instability of aSedimenting Suspension Arising in DirectNumerical SimulationR. Glowinski, T.W. Pan , and D.D. Joseph

Astronomy

° 99/149, September 1999The MHD Kelvin-Helmholtz Instability III:The Role of Sheared Magnetic Field inPlanar FlowsH. Jeong, D. Ryu, T.W. Jones, and A.Frank

° 99/141, August 1999Numerical Simulations for RadiationHydrodynamics II. Transport LimitW. Dai and P.R. Woodward

Chemical Engineering andMaterials Science

° 99/132, July 1999Ab initio Absorption Spectra of GalliumArsenide ClustersI. Vasiliev, S. Ogut, and J.R. Chelikowsky

° 99/164, October 1999Large Pairing Jahn-Teller Distortions AroundDivacancies in Crystalline SiliconS. Ogut and J.R. Chelikowsky

° 99/165, October 1999Pressure Induced Amorphization inCrystalline Silica: Soft Phonon Modes andShear Instabilities in CoesiteD. Dean, R.M. Wentzcovitch, N.R.Keskar, J.R. Chelikowsky, and N. Binggeli

° 99/145, August 1999Computational Simulations of the Growth ofCrystals from LiquidsA. Yeckel and J.J. Derby

° 99/155, September 1999Theoretical Analysis of 3D, TransientConvection and Segregation in MicrogravityBridgman Crystal GrowthA. Yeckel, V.F. de Almeida, and J.J. Derby

° 99/130, July 1999A Self-Consistent Cell Flux Expression forSimultaneous Chemotaxis and ContactGuidance in TissuesM.A. Wagle and R.T. Tranquillo

Names of University of Minnesota principal investigators appear in bold type.


RES

ERA

CH

REP

OR

TS

Chemistry

° 99/147, August 1999Perfluorocarbenes Produced by ThermalCracking. Barriers to Generation andRearrangementC.J. Cramer and M.A. Hillmyer

° 99/142, August 1999The Viscosity of Polymerically StabilizedDispersions of Spherical Colloid ParticlesE. Wajnryb and J.S. Dahler

° 99/143, August 1999The Viscosity of Electrostatically StabilizedDispersions of Spherical Colloid ParticlesE. Wajnryb and J.S. Dahler

° 99/133, July 1999Simulating Retention in Gas-LiquidChromatographyM.G. Martin, J.I. Siepmann, and M.R.Schure

° 99/134, July 1999Adiabatic Nuclear and Electronic SamplingMonte Carlo Simulations in the GibbsEnsemble: Application to Polarizable ForceFields for WaterB. Chen and J.I. Siepmann

° 99/139, August 1999A Two-Response-Time Model Based onCM2/INDO/S2 Electrostatic Potentials forthe Dielectric Polarization Component ofSolvatochromic Shifts on Vertical ExcitationEnergiesJ. Li, C.J. Cramer, and D.G. Truhlar

° 99/161, September 1999A Universal Solvation Model Based on ClassIV Charges and the Intermediate Neglect ofDifferential Overlap for SpectroscopyMolecular Orbital MethodJ. Li, T. Zhu, C.J. Cramer, and D.G.Truhlar

° 99/170, October 1999Comment on "On the Longuet-Higgins Phaseand its Relation to the Electronic Adiabatic-Diabatic Transformation Angle"B.K. Kendrick, C.A. Mead, and D.G.Truhlar

° 99/171, October 1999Energetic and Structural Features of the CH4+ O(3P) → CH3 + OH AbstractionReaction: Does Perturbation Theory from aMulticonfiguration Reference State (Finally)Provide a Balanced Treatment of TransitionStates?O. Roberto-Neto, F.B.C. Machado, andD.G. Truhlar

° 99/172, October 1999Nonequilibrium Solvation Effects for aPolyatomic Reaction in SolutionY.-Y Chuang and D.G. Truhlar

° 99/173, October 1999Perspective on "Principles for a Direct SCFApproach to LCAO-MO Ab InitioCalculations by J. Almlöf, K. Faegri, Jr., andK. KorsellD.G. Truhlar

° 99/174, October 1999Prediction of Vapor Pressures from Self-Solvation Free Energies Calculated by theSM5 Series of Universal Solvation ModelsP. Winget, G.D. Hawkins, and C.J.Cramer, and D.G. Truhlar

° 99/175, October 1999Quantum Mechanical and QuasiclassicalTrajectory Surface Hopping Studies of theElectronically Nonadiabatic Predissociation ofthe Ã State of NaH2M.D. Hack, A.W. Jasper, Y. Volobuev,D.W. Schwenke, and D.G. Truhlar


RES

ERA

CH

REP

OR

TS

Computer Science andEngineering

° 99/152, September 1999Iterative Solution of Linear Systems in the20th CenturyY. Saad and H.A. van der Vorst

Geology and Geophysics

° 99/129, July 1999Criticality of Subducting SlabsM.R. Riedel, S. Karato, and D.A. Yuen

° 99/131, July 1999Using Discrete Particles as a Natural Solverin Simulating Multiple-Scale PhenomenaW. Dzwinel, W. Alda, J. Kitowski, andD.A. Yuen

° 99/135, August 1999Thermal Evolution of Sedimentary BasinFormation with Variable ThermalConductivityL. Starin, D.A. Yuen, and S.Y. Bergeron

° 99/138, August 1999Exothermic and Endothermic ChemicalReactions Modelled with Molecular DynamicsW. Alda, D.A. Yuen, H.P. Lüthi, and J.R.Rustad

° 99/144, August 1999Feedback Effects of Variable ThermalConductivity on the Cold Downwellings inhigh Rayleigh Number ConvectionF. Dubuffet, D.A. Yuen, and T. Yanagawa

° 99/148, August 1999Transition to Turbulent Thermal Convectionbeyond Ra = 1010 Detected in NumericalSimulationsA.P. Vincent and D.A. Yuen

° 99/150, September 1999Extended-Boussinesq Thermal-ChemicalConvection with Moving Heat Sources andVariable ViscosityU. Hansen and D.A. Yuen

° 99/151, September 1999Matching Macroscopic Properties of BinaryFluids to the Interactions of DissipativeParticle DynamicsW. Dzwinel and D.A. Yuen

° 99/153, September 1999Rayleigh-Taylor Instability in the MesoscaleModelled by Dissipative Particle DynamicsW. Dzwinel and D.A. Yuen

° 99/154, September 1999A Multi-Level, Discrete Particle Model inSimulating Ordered Colloidal StructuresW. Dzwinel and D.A. Yuen

° 99/166, October 1999At What Stress Level is the Central IndianOcean Lithosphere Buckling?M. Gerbault

° 99/167, October 1999Kinetics of Diffusion-Controlled MineralReactions in the Matrix in the Kharlovo(South Siberia, Russia) Contact AureoleI.I. Likhanov, A. Ten, V.V. Reverdatto, V.A.Ananiev, and I. Memmi

Mathematics

° 99/162, October 1999Perturbations of Normally HyperbolicManifolds with Applications to the Navier-Stokes EquationsV.A. Pliss and G.R. Sell


RES

ERA

CH

REP

OR

TS

Mechanical Engineering

° 99/163, October 1999Characterization of the Near-Surface Gas-Phase Chemical Environment inAtmospheric-Pressure Plasma Chemical VaporDeposition of DiamondJ. Larson, M.T. Swihart, and S.L. Girshick

° 99/156, September 1999Effects of Tip Clearance and Rotation onThree Dimensional Flow Fields in TurbineCascadesB. Han and R.J. Goldstein

° 99/168, October 1999Large Eddy Simulation of Constant HeatFlux Turbulent Channel Flow with PropertyVariations Using a Dynamic Subgrid-ScaleModelN. Meng, L.D. Dailey, and R.H. Pletcher

Physics

° 99/140, August 1999Pauli-Villars Regularization in DLCQJ.R. Hiller

° 99/160, September 1999On the Use of Discrete Light-ConeQuantization to Compute Form FactorsJ. R. Hiller

° 99/146, August 1999Flux Lattice Melting and the Onset of Hc2FluctuationsS.W. Pierson and O.T. Valls

° 99/159, September 1999Tunneling Spectroscopy forFerromagnet/Superconductor JunctionsI. Zutic and O.T. Valls

° 99/169, October 1999Free Energy Landscape of Simple LiquidsNear the Glass TransitionC. Dasgupta and O.T. Valls


❖ Email your requests to:

[email protected](be sure to include your postal address)

❖ or you can complete this form and return it by fax:

(612) 624-8861

or mail to:

Research Reports CoordinatorSupercomputing Institute1200 Washington Avenue SouthMinneapolis, Minnesota 55415

❖ Our full bibliography is available on the World Wide Web at:

www2.msi.umn.edu/Reports/reporthead.html(some reports are available in PDF format directly from this site)

Please send me the following research reports:

UMSI __________ UMSI __________ UMSI __________

UMSI __________ UMSI __________ UMSI __________

Name _________________________________________

Address _________________________________________

_________________________________________

_________________________________________

_________________________________________

❏❏ Please check here if you would like to receive a copy of our 1999Annual Research Report

RES

ERA

CH

REP

OR

TS

Would you like to receive a copy of any of these research reports?

Supercomputing Institute1200 Washington Avenue SouthMinneapolis, MN 55415

editor/photographer/layout: Cory Hendrickson

For more information on the University of MinnesotaSupercomputing Institute, please contact Donald G.Truhlar, Director, at the address given below. You may alsoconsult the Supercomputing Institute’s World Wide Webpages: www.msi.umn.edu.

This information is available in alternative formats uponrequest by individuals with disabilities. Please send emailto [email protected] or call 612-625-1818.

The University of Minnesota is committed to the policythat all persons shall have equal access to its programs,facilities, and employment without regard to race, color,creed, religion, national origin, sex, age, marital status, dis-ability, public assistance status, veteran status, or sexualorientation.

© 1999 University of Minnesota Supercomputing Institute

Please help us keep our mailing list up to date

Check one:

❏ Add name to mailing list

❏ Remove name from mailing list

❏ Make corrections/changes as shown

Name

Institution

Address

Phone

Email

(you can also email address changes to [email protected])

Nonprofit Org.U.S. Postage

PAIDMinneapolis, MNPermit No. 155

Change Service Requested

contains a minimum of 10% postconsumer waste

Supercomputing Institute · 2014. 1. 16. · 2 Supercomputing Institute Research Bulletin mammalian...

Documents

Transcript of Supercomputing Institute · 2014. 1. 16. · 2 Supercomputing Institute Research Bulletin mammalian...