Laser physicists are tantalizingly close to building a device that delivers highly focused beams of tumor-killing ionized particles and packing it in a compact unit that would fit in a doctor’s office. OncoRay, a research center in Dresden, Germany, is designing what would be the first prototype compact particle therapy units.
In a few femtoseconds, the laser devices excite electrons in a foil a few microns thick, accumulating laser energy and then transferring it to surrounding atoms, creating ions that deliver radiation to a target. Researchers know that’s what happens on a large scale. The underlying physics, however, occurs in the particles that constitute matter’s building blocks – a scale at which relativistic phenomena can and do affect the outcome.
Physicist Michael Bussmann of the German research laboratory Helmholtz-Zentrum Dresden-Rossendorf (HZDR) hopes to clarify those fine details. The team he leads is creating three-dimensional physics simulations of the particle acceleration process. His computational radiation physics group recently began running the simulations on Titan, a Cray XK7 supercomputer at the Oak Ridge Leadership Computing Facility, using 57 million processor hours allocated through the U.S. Department of Energy’s INCITE program.
“A lot of interesting physics is happening,” Bussmann says. The problem is that researchers can’t always reliably repeat experiments because multiple parameters affecting the outcome constantly change. “It is vital to compare experimental results to realistic simulations. We need computers like Titan to do realistic simulations because we are already at the edge of what computers can do right now.”
The simulations examine what happens when the laser’s high-energy electromagnetic fields rip matter apart, producing negatively charged electrons and positively charged ions. Separating electrons from ions creates the acceleration fields necessary to give energy to the ions. Long before striking its intended target, the laser pulse enters a plasma state in which the charged particles rapidly interact with electric and magnetic fields in complex ways that can cause instabilities under certain conditions. Bussmann’s team is modeling a series of experimental conditions, including varying properties of the laser pulse and the irradiated target’s atomic composition.
‘Please steal this code.’
“What we would really like to do is control the whole process,” he says. “As a patient, you would not want to be on the other end of that ion dose that is trying to cure you if there is no control over the dosing. Control, understanding and reproducibility are important for the medical applications.”
Electromagnetic radiation’s ability to melt away cancerous lesions was clear over a century ago, from the first medical applications of X-rays. But the excitement radiation therapy generated was quickly tempered by its indiscriminate destruction of healthy tissue. The central challenge since that time has been to kill cancerous cells while sparing healthy ones.
Heavier particles like the positively charged ions physics accelerators create have shown promise in delivering intense radiation to tumors while minimizing the radiation scattered to nearby tissue. This ion beam therapy, however, has fallen short of its potential. One serious issue: Producing ions energetic enough to hit tissues deep within the body with therapeutic radiation doses requires a big, expensive device.
At HZDR, Bussmann is collaborating with experimental scientists to demonstrate that a laser-accelerated ion beam can kill cancer cells in the laboratory, as they reported in the New Journal of Physics. At this point, though, the efficiency of laser energy transfer into ion beam energy is too poor to reach deep tissue tumors. Most of the electric energy powering the laser dissipates as heat. At a total efficiency of less than 1 percent, there is a lot of room for improvement. Bussmann’s group thinks that using modern diode-pump lasers will boost the efficiency by at least a factor of 10.
“Diode lasers are smaller and hopefully easier to control, so we can more easily reproduce experimental conditions,” he says.
The Titan simulations play a crucial role. The group wants to resolve the dynamics based on the fundamental physics – the motion of electrons and ions in ultra-strong fields – while generating a complete picture of the overall acceleration process. They think optimizing the laser parameters and experimenting with the form, structure and composition of the laser targets will help boost the ion beam acceleration.
The group’s ability to do these large-scale 3-D physics simulations hinges on an innovative code developed by a group of undergraduate students at Dresden University working with the HZDR physicists. PIConGPU (particle-in-cell on graphics processing units) does the heavy lifting by using massively parallel GPU computer chips to calculate the relative movement of particles within the electromagnetic field. The key, Bussmann says, is that the data stays on the GPU, reducing time-consuming communication with central processing units. The group coined the term “super cells” to describe the technique, which runs calculations concurrently, each on its own GPU. The code recently helped Titan set a record for greatest number of calculations (7.1 petaflops, or quadrillion math operations per second) using a GPU-centric particle-in-cell code.
The open-source code is versatile, Bussmann says, and can be used for atomic fusion, astrophysics or any other high-energy plasma physics simulation. The HZDR computational radiation physics unit also has used the code to compute particle emissions from plasma jets that radiate from the centers of stars and black holes. The group is eager to have others try it out. “Please steal this code,” Bussmann says.
The group has completed two runs on Titan under the INCITE allocation. Each simulated a levitating target developed experimentally by Jörg Schreiber’s group at Ludwig-Maximilians University in Munich. A levitating target means no arm must hold the target in place, so all laser energy really goes into the target. Researchers hope a spherical target comparable in size to the laser focal spot will be best to drive most of the laser energy into creating a strong charge-separation field. They’re comparing the Titan simulation results with experiment results.
“We have seen some very general effects that we have also seen in experimental work such as light scattering and possible instabilities growing,” Bussmann says. “We are trying to avoid being too quick in our analysis, but it really looks exciting right now.”