David Raczkowski

388 Dolores St.

San Francisco, Ca. 94110

(415) 861-2579 dbraczkowski@lbl.gov

EDUCATION

University of California, Davis

Ph. D, in Physics, June 2000

University of Michigan, Ann Arbor

B. S. in Mathematics and Physics, May 1993

RESEARCH/WORK

6/00 – pres. Lawrence Berkeley National Laboratory

Currently, I am a postdoctoral fellow in the scientific computing group with the National Energy Research Scientific Computing Center (NERSC) division. I currently maintain and develop the software package PARATEC (www.nersc.gov/projects/paratec) for use on massively parallel architecture supercomputers. It is a pseudopotential plane wave code for the computer modeling of materials. I have made many algorithmic improvements along with creating a web site.

6/98 – 6/00. Sandia National Laboratories, University of California at Davis, and Ford Motor Co.

In the summer of ’98 by working at Sandia in the advanced materials division in Albuquerque, I integrated a new algorithm that replaces diagonalization into the Gaussian-based density functional code, SEQQUEST. In California, I refined the algorithm to be linear scaling for the dominant parts with highly optimized routines to handle sparse matrices. The overall memory structure of the main program and many routines was refined to handle sparse matrices. The code was tested on silicon for systems up to 1,728 atoms on a single processor. In the summer of ’99 at Ford, I investigated yttria-stabilized zirconia (YSZ) with the intent of applying the linear scaling code to a real problem. At Davis, the accuracy of total energies has been checked stringently with silicon carbide. The accuracy of forces, relaxation geometries, and formation energies has been checked with YSZ. I also extended the algorithm to deep-level nitrogen defects in silicon carbide.

5/96 – 5/98 University of California at Davis

Investigated the viability of using of the chemical pseudopotential technique as a means for large-scale electronic structure calculations. I created a Gaussian-based density functional code in C++ for this purpose.

8/92 - 8/94 University of Michigan

Worked with a postdoc to build and develop the software for a multi-wire proportional chamber as a particle detector for the first year. In the second year, a research scientist and I designed and built a drift chamber particle detector. I developed the software, assembler and C, for the data acquisition system.

COMPUTER PROGRAMMING EXPERIENCE

F90 with MPI: The software that I currently maintain contains approximately 60,000 lines of code. It uses the MPI message passing library for handling parallelization on distributed architectures. It has proven to scale well for hundreds of processors. An example of my contributions is the diagonalization of a small submatrix on a subset of total number of processors using the library Scalapack. The distribution of memory is calculated automatically for optimal load balancing and communication efficiency.

ASSEMBLER: I wrote code for a 16-bit Motorola 68000 chip. The code read in data from an analog-to-digital converter and output the data through a serial port to a VAX workstation.

C: The C code, written on a DEC Vax workstation, read in the data from the assembler output mentioned above and manipulated the information for analysis by PAW(physics analysis workstation). This effort was focused on the acquisition of data for a high-energy spin physics experiment

C++: On a IBM Rs/6000 workstation and a DEC alpha workstation, I wrote approximately 10,000 lines of code for a Gaussian-based density functional code for molecules, i.e. no periodic boundary conditions. The code used a hierarchical class structure with each class having as a component an array of the lower class. A Cluster class was composed of atoms, made of wavefunctions, comprising contracted Gaussians, and finally primitive Gaussians. Methods were written to handle multiplication, integration, and function evaluation of the different classes. The code uses Lapack++ routines, which link to the Fortran Lapack library for the linear algebra calculations.

F77: I have added about 15,000 lines of code to the existing 45,000 lines of code. I have also made improvements and suggestions for critical speedup of the code not directly authored by me. Much of the code development involved the efficient storage and multiplication of sparse matrices, where the final matrix is sparse as well as the starting matrices. The storage scales linearly and the sparse matrix multiplication routines have also shown to scale linearly. Machine optimized BLAS routines for dense matrices were utilized as much as possible. I have also written k-space charge symmetrization routines. The code has been developed on a SGI platform and a DEC alpha platform.

TEACHING EXPERIENCE

Lecturer: I have substituted for Prof. Fong for 3 classes at the graduate quantum mechanics level. The topics have been lab and center of mass reference frames for motion and collision and basic representation and operator theory concerning Schroedinger’s equation.

Graduate reader: I was a reader for graduate quantum mechanics that included office hours.

Teaching assistant: I taught laboratory sections for all quarters of introductory physics for engineering students. I also taught laboratory sections for a new course in introductory physics designed for students of the life sciences. This included 5 hours of lab per week for 2 classes that included experiment, problem solving, lecture, and discussion.

PUBLICATIONS

D. Raczkowski,” Efficient methods for sparse matrix multiplications with resultant sparse matrices”, (in prep)

D. Raczkowski and C.Y. Fong, “The quantitative applicability of subspace optimization with localization to N deep level defects in SiC ”, (to be subittted)

D. Raczkowski, L.W. Wang, A. Canning,” Efficacy of ab-initio total energy calculations on massively parallel computers using a plane-wave basis set”, to be submitted.

D. Raczkowski, C.Y. Fong, and E.B. Stechel, " Localization in an occupied-subspace-optimization approach to electronic structure: application to yttria-stabilized zirconia ", submitted to Physical Review B.

D. Raczkowski, C.Y. Fong, P.A. Schultz, R.A. Lippert, and E.B. Stechel, "Unconstrained and constrained minimization, linear scaling, and the Grassmann manifold: theory and applications", Physical Review B, vol. 64, 155203 (2001)

D. Raczkowski, A. Canning, L.W. Wang, “Thomas-Fermi charge mixing for obtaining self-consistency in density functional calculation”, Physical Review B, vol. 64, 121101 (2001)

D. Raczkowski, Ph.D. thesis, “An Occupied Subspace Optimization for Linear Scaling in Large-Scale Ab Initio Electronic Structure Calculations”, University of California at Davis, 2000

INVITED PRESENTATIONS

5/99 Sandia National Laboratories, Albuquerque, " Linear scaling and localized nonorthogonal orbitals"

CONFERENCE PRESENTATIONS

6/01 American Physics society: division of computational physics, MIT, Boston , Ma.

" Thomas-Fermi charge mixing for obtaining self-consistency in density functional calculations

"

3/01 American Physics Society, Seattle,Washington, " Thomas-Fermi charge mixing for obtaining self-consistency in density functional calculations

"

3/00 American Physics Society, Minneapolis, Minnesota, "Effects of localization in an occupied subspace optimization approach to electronic structure for calculations of ZrO₂ and Yttria-stabilized Zirconia (YSZ)"

6/99 CECAM workshop on Wannier orbitals, Lyon, France, " Linear scaling and localized nonorthogonal orbitals "

3/99 American Physics Society, Atlanta Georgia, " An unconstrained minimization approach for use in DFT calculations"

3/98 American Physics Society, Los Angeles, Ca, "Chemical Pseudopotential Approach for electronic Structure Studies"

REFERENCES

C.Y. Fong

Department of Physics

University of California, Davis

Davis, Ca 95616

(530) 752-1792

fong@solid.ucdavis.edu

Ellen B. Stechel

Estechel@ford.com