Filling cavities virtually

Building a particle accelerator is no small undertaking.

These huge closed-loop devices can run for miles underground. Inside, scientists generate beams of some of the smallest particles in the universe, accelerate them to nearly the speed of light, then smash them into each other. Physicists observe these collisions, hoping to unlock secrets of the universe – including what happened in the “Big Bang” that’s believed to have created everything.

The whole works can be thrown off by something as basic as the shape of the cavities that make up the tunnel. That’s why scientists are using powerful computers to simulate accelerator cavity or cell shapes. Simulation lets them test different configurations without the time and expense of actually building them.

The Department of Energy’s Interoperable Technologies for Advanced Petascale Simulations (ITAPS) project has helped one DOE facility – the Stanford Linear Accelerator Center or SLAC – run simulations that doubled the effectiveness of one of its particle smashers, the second generation electron-positron collider, known as PEP-II.

ITAPS is part of the DOE’s Scientific Discovery through Advanced Computing (SciDAC) program. SciDAC sponsors development of the hardware and software supercomputers use to advance DOE’s research programs, including accelerator design.  ITAPS is designed to create a “toolbox” of software solutions other researchers can use for DOE projects.

“ITAPS … is a collaborative effort including six DOE laboratories and three universities,” says its head, Lori Diachin of DOE’s Lawrence Livermore National Laboratory. “It brings together a large team of experts in this area to start marching toward a common goal.”

The cavity walls’ shape can be tuned to achieve the right frequencies – and computer modeling can help.

Diachin’s ITAPS group is largely focused on improving mesh generation – a method to digitally divide the regions of a system being modeled into a grid – and Adaptive Mesh Refinement (AMR), a technique to focus the grid on regions of most interest to researchers or on regions where there’s the most activity.  It’s like focusing a camera on the subject while the less important background is left fuzzy.

Diachin says her ITAPS group is “about creating the next generation of advanced mesh generation technologies – getting very high quality meshes, and inserting them into application codes” – including one for accelerator cell design.

“We work with a number of different application teams, but probably the most prominent has been the accelerator design team at the Stanford Linear Accelerator Center,” Diachin says. Besides Stanford, researchers from Rensselear Polytechnic Institute and DOE’s Sandia and Argonne national laboratories are collaborating with Diachin’s group on the project.

Stanford “has been quite successful on the computational end of modeling cavities,” says Tim Tautges, a computational scientist affiliated with Argonne who’s generating some of the tools used for meshes.

The goal was to increase PEP-II’s “beam luminosity” – the number of particles that go through the accelerator successfully. That’s important because “The more luminosity, the higher the probability of getting the low-probability interactions physicists are looking for, which is the name of the game,” Tautges says.

Increasing beam luminosity largely depends on minimizing the wake fields generated by packets of electrons as they speed through accelerator cavities like that of PEP-II.  Wake fields – electromagnetic waves – are similar to the wake a speeding boat creates, making things rough for the boats that follow. Electrons have the same problem as they cross the wake fields of electron packets that precede them.

“If too many bunches of particles get sufficiently knocked off their path, they get accelerated into the wall of the cavity, and that’s very bad,” Tautges says. “That will heat the cavity up and shut the whole thing down.”

Additionally, the more packets of electrons pumped through the cavity, the stronger the resulting wake fields.

Since the cavities resonate at certain frequencies, like a violin string, “nailing the acceleration frequency with high accuracy” can overcome this propensity, Tautges adds.  “Just as important is to damp out other frequencies, because it’s the other frequencies that excite the wake fields and cause these packets to go off path.”

The cavity walls’ shape can be tuned to achieve the right frequencies – and computer modeling can help, Tautges says.

“A common need in all of this is to be able to simulate those cavities with high accuracy, and also to simulate them with high-fidelity geometric representation because the shape is so important to the results,” he adds. “That’s the goal.”

Collaborating with Diachin’s ITAPS group, the SLAC team ran computer simulations that enabled it “to double the luminosity of their current accelerator design, which directly increases the likelihood of scientists observing the interactions that the accelerator is built to observe,” Tautges says.

Diachin adds: “The main point is that you have mathematicians, computer scientists, and application physicists all working together toward a common goal.  It is the nature of these interdisciplinary teams that SciDAC has fostered. The accelerator program is one of the real success stories of that model. Without these collaborations there’s a lot SLAC would not have been able to do in terms of simulations with the fidelity they would like.”