Research Objectives
Particle accelerators are playing an increasingly important role
in basic and applied science, and are enabling new accelerator-driven
technologies. But the design of next-generation accelerators,
such as linear colliders and high intensity linacs, will require
a major advance in numerical modeling capability due to extremely
stringent beam control and beam loss requirements and highly complex
three-dimensional accelerator components. The primary goal of
the Grand Challenge in Computational Accelerator Physics is to
develop a parallel modeling capability that will enable high-performance,
large-scale simulations for the design, optimization, and numerical
validation of next-generation accelerators.
Computational Approach
The Grand Challenge is focusing on two areas: electromagnetics
and beam dynamics.
The new set of tools being developed for high-resolution electromagnetics
incorporates:
- unstructured grids to better treat geometries
- new algorithms to improve accuracy and convergence
- refinement techniques to optimize computing resources
- parallel processing to simulate large problems.
Presently, there are two types of solvers in the tool set. The
first type includes a 3D eigensolver using linear and quadratic
elements; the MPI version, Omega3P, currently runs on the Cray
T3E at NERSC. The second type includes a 3D time-domain solver,
Tau3P, based on the modified Yee algorithm; this code is also
being ported to the T3E via MPI. Tau3P models power transmission
components, whereas Omega3P simulates rf cavities. Both are designed
to handle large, complex geometry meshes generated from solid
models provided by popular CAD tools such as AutoCAD and ProEngineer.
Many systems involving intense charged-particle beams can be described
by the Vlasov/Poisson equations. The IMPACT (Integrated-Map and
Particle Accelerator Tracking) code suite now under development
as part of the Grand Challenge is based on particle simulation
methods for solving the Vlasov/Poisson equations as applied to
accelerators. This code suite uses modern split-operator methods
to combine the best features of particle simulation techniques
with map-based, magnetic optics tools for simulating beam transport
in accelerators. Three parallel versions have been developed using
HPF, F90/MPI, and the POOMA framework based on C++. All codes
now run on the T3E. The charge resolution of these codes will
very soon become "real-world," i.e., the number of particles
in the simulation will be approximately the same as the actual
number of particles in a bunch.
Accomplishments
The past year has seen significant achievements, especially in
the successful parallelization of electromagnetics codes and in
further development of the parallel beam dynamics capability.
Omega3P has been used to design the Damped Detuned Structure (DDS)
for the Next Linear Collider (NLC) to within 0.01% accuracy in
frequency (Figure 1). This is accomplished on the T3E by employing
128 to 256 PEs to process several million elements in one octant
of the DDS geometry. A similar high-resolution calculation has
been performed on the radio frequency quadrupole (RFQ) for the
Spallation Neutron Source (SNS), in which an agreement with measurement
to less than a MHz in the cavity frequencies has been achieved.
In the area of beam dynamics, we have performed the largest simulations
to date for the Accelerator Production of Tritium (APT) project,
using a parallel version of the code LINAC running on the T3E.
These simulations are essential for validating designs and, in
particular, for making beam loss predictions.
|
Figure 2. Three-dimensional visualization
of the output from the LINAC3D code showing the particle density
in a mismatched beam in the presence of nonlinear accelerating
fields. The beam is moving from left to right, and the left-right
asymmetry is due to the nonlinearities. The diffuse region far
from the central core constitutes the beam halo. Understanding
and predicting beam halo is a major issue in next-generation high-current
linear accelerators.
Using the IMPACT
suite of codes, we are now performing simulations of superconducting
linacs, using a new algorithm that is more accurate than traditional
methods in regard to the treatment of rf accelerating cavities.
An example output is shown in Figure 2, which is a snapshot of
a beam that is improperly matched and has developed a pronounced
beam halo. We have also developed a parallel version of the Lie
algebraic beam dynamics code MaryLie, and using split-operator
methods, turned that code into a parallel particle-in-cell (PIC)
code that combines a high-order magnetic optics capability with
a capability to model intense beams. Finally, we have included
the effects of external noise and collisions in the particle equations
of motion using Langevin techniques.
Significance
The advanced modeling tools developed through this Grand Challenge
will allow future particle accelerators to be designed with reduced
cost and risk as well as improved reliability and efficiency.
The projects that this effort supports will have significant societal,
economic, and scientific impacts, including impacts on DOE missions
in the offices of Energy Research, Defense Programs, and Environmental
Management.
Publications
R. Ryne, S. Habib, J. Qiang, K. Ko, Z. Li, B. McCandless, W. Mi,
C. Ng, M. Saparov, V. Srinivas, Y. Sun, X. Zhan, V. Decyk, and
G. Golub, "The US DOE Grand Challenge in computational accelerator
physics," Proceedings LINAC98 (1998).
B. McCandless, Z. Li, V. Srinivas, Y. Sun, and K. Ko, "Omega3P:
A parallel eigensolver for modeling large, complex cavities,"
Proceedings ICAP98 (1998).
C.-K. Ng, B. McCandless, V. Srinivas, M. Wolf, and K. Ko, "Tau3P:
A parallel time domain solver to simulate large rf structures,"
Proceedings ICAP98 (1998).
http://gita.lanl.gov/people/salman/capgca
|
