Plasmas are very hot gases that make up about 99% of all visible
matter in space. Because they are so hot, the atoms inside these gases
break into their components, leaving positively charged ions and
electrons that move independently.
Safronova is an expert in computing the properties of atoms. Atoms
have distinct levels of energy, and when an electron jumps from one
energy state to another, it emits a particle of light called a photon.
How fast this jump takes place is called the transition rate, which
changes depending on the energy levels that are involved in the process.
Analyzing this light is the main way astronomers learn about stars
and galaxies in the cosmos. But these experimentalists need help from
theoretical physicists like Safronova, who can calculate where to look
for these transitions. To do this, theorists use atomic-scale models and
then point experimentalists in the right direction.
Safronova explained that in the problem they were working on, a
theoretical calculation of a key property of two very bright, visible
transitions of iron called 3C and 3D (which occur in most hot
astrophysical plasmas) strongly disagreed with experimental results. The
property in question showed how strong one transition was relative to
the other.
The discrepancy had previously been attributed to theory not being
accurate enough. Yet, no matter how theoretical physicists like
Safronova modified the advanced calculations, the predicted transition
rate didn’t change in any meaningful way. This led the theorists to
remain certain their calculations were correct, but it still bothered
them.
So, Safronova and her collaborators at other institutions worked to
improve their methods for calculating this transition rate using
supercomputers, including UD’s Caviness and DARWIN high-performance
parallel computing systems. Key to this process was Charles Cheung, a
postdoctoral researcher in Safronova’s group who earned both bachelor’s
and doctoral degrees in physics at UD in 2016 and 2021. Cheung is
credited with writing the new parallel version of the atomic code that
helped solve this long-standing discrepancy.
“Charles was instrumental in terms of theory for making this work happen,” said Safronova.
Cheung recalled the laborious process of running code on a single
computer for an earlier paper and waiting weeks for the computer to
return a single number, then repeating the calculation over and over
with different parameters. The power of high-performance parallel
computing, harnessed by Cheung’s new parallel code, greatly speeded up
this process on the current work.
“For that first paper it took me months of runtime just to figure out
the numbers I needed. Now, I can rerun that original two-week
calculation and have an answer in 15 minutes,” said Cheung, the paper’s
second author. “I can run problems that are over 100 times larger, too,
to explore exciting new systems in atomic physics that are of extreme
interest to experimental groups for building quantum sensors.”
The new version of the code allowed the theoretical team to make much
larger calculations than previously possible while reducing numerical
errors. Cheung’s first version of parallel code scaled with about 50%
efficiency, but with additional refinements, today main parts of the
code perform with 99 to 100% efficiency.
“With our new code, we were able to put uncertainty numbers on our
predictions,” he continued. According to Safronova, this ability to put
precision on theoretical predictions in such complicated systems is new,
and it provided even greater confidence that the team’s atomic-scale
models were correct.
Meanwhile, Safronova’s experimentalist collaborators decided to redo
the experiment at PETRA III, a German synchrotron light facility at the
DESY laboratory outside Hamburg. The improved experimental precision
allowed the research team to obtain data on the tails of the spectral
lines they observed. The new measurements confirmed that the theoretical predictions were correct.
The researchers recently published their findings in Physical Review Letters.