NERSCPowering Scientific Discovery Since 1974

Shedding New Light on Luminous Blue Variable Stars

3D Simulations Disperse Some of the Mystery Surrounding These Massive Stars

December 7, 2018

By Jennifer Huber

Three-dimensional (3D) simulations run at two of the U.S. Department of Energy’s national laboratory supercomputing facilities and the National Aeronautics and Space Administration (NASA) have provided new insights into the behavior of a unique class of celestial bodies known as luminous blue variables (LBVs) - rare, massive stars that can shine up to a million times brighter than the Sun.

Astrophysicists are intrigued by LBVs because their luminosity and size dramatically fluctuate on a timescale of months. They also periodically undergo giant eruptions, violently ejecting gaseous material into space. Although scientists have long observed the variability of LBVs, the physical processes causing their behavior are still largely unknown. According to Yan-Fei Jiang, an astrophysicist at UC Santa Barbara’s Kavli Institute for Theoretical Physics, the traditional one-dimensional (1D) models of star structure are inadequate for studying LBVs.

luminous blue variable star

A snapshot from a simulation of the churning gas that blankets a star 80 times the sun's mass. Intense light from the star's core pushes against helium-rich pockets in the star's exterior, launching material outward in spectacular geyser-like eruptions. The solid colors denote radiation intensity, with bluer colors representing regions of larger intensity. The translucent purplish colors represent the gas density, with lighter colors denoting denser regions. Image: Joseph Insley, Argonne Leadership Computing Facility


“This special class of massive stars cycles between two phases: a quiescent phase when they’re not doing anything interesting, and an outburst phase when they suddenly become bigger and brighter and then eject their outer envelope,” said Jiang. “People have been seeing this for years, but 1D, spherically-symmetric models can’t determine what is going on in this complex situation.”

Instead, Jiang is leading an effort to run first-principles, 3D simulations to understand the physics behind LBV outbursts — using large-scale computing facilities provided by Lawrence Berkeley National Laboratory’s National Energy Research Scientific Computing Center (NERSC), the Argonne Leadership Computing Facility (ALCF), and NASA. NERSC and ALCF are DOE Office of Science User Facilities.

Physics Revealed by 3D

In a study published in Nature, Jiang and colleagues from UC Santa Barbara, UC Berkeley, and Princeton University ran three 3D simulations to study three different LBV configurations. All the simulations included convection, the action when a warm gas or liquid rises while its cold counterpart sinks. For instance, convection causes hot water at the bottom of a pot on a stove to rise up to the top surface. It also causes gas from a star’s hot core to push to its outer layers.

During the outburst phase, the new 3D simulations predict that convection causes a massive star’s radius to irregularly oscillate and its brightness to vary by 10 to 30 percent on a timescale of just a few days — in agreement with current observations.

“Convection causes the star to expand significantly to a much larger size than predicted by our 1D model without convection. As the star expands, its outer layers become cooler and more opaque,” Jiang said.

Opacity describes how a gas interacts with photons. The researchers discovered that the helium opacity in the star’s outer envelope doubles during the outburst phase, making it more difficult for photons to escape. This leads the star to reach an effective temperature of about 9,000 degrees Kelvin (16,000 degrees Fahrenheit) and triggers the ejection of mass.

“The radiation force is basically a product of the opacity and the fixed luminosity coming from the star’s core. When the helium opacity doubles, this increases the radiation force that is pushing material out until it overcomes the gravitational force that is pulling the material in,” said Jiang. “The star then generates a strong stellar wind, blowing away its outer envelope.”

Massive Simulations Required

Massive stars require massive and expensive 3D simulations, according to Jiang. So he and his colleagues needed all the computing resources available to them, including about 15 million CPU hours at NERSC, 60 million CPU hours at ALCF, and 10 million CPU hours at NASA. In addition, NERSC played a special role in the project.

“The Cori supercomputer at NERSC was essential to us in the beginning because it is very flexible,” Jiang said. “We did all of the earlier exploration at NERSC, figuring out the right parameters to use and submissions to do. We also got a lot of support from the NERSC team to speed up our input/output and solve problems.”

In addition to spending about 5 million CPU hours at NERSC on the early phase of the project, Jiang’s team used another 10 million CPU hours running part of the 3D simulations.

“We used NERSC to run half of one of the 3D simulations described in the Nature paper and the other half was run at NASA. Our other two simulations were run at Argonne, which has very different machines,” said Jiang. “These are quite expensive simulations, because even half a run takes a lot of time.”

Even so, Jiang believes that 3D simulations are worth the expense because illuminating the fundamental processes behind LBV outbursts is critical to many areas of astrophysics — including understanding the evolution of these massive stars that become black holes when they die, as well as understanding how their stellar winds and supernova explosions affect galaxies.

Jiang also used NERSC for earlier studies, and his collaboration is already running follow-up 3D simulations based on their latest results. These new simulations incorporate additional parameters — including the LBV star’s rotation and metallicity — varying the value of one of these parameters per run. For example, the speed from rotation is larger at the star’s equator than at its poles. The same is true on Earth, which is one of the reasons NASA launches rockets from Florida and California near the equator.

“A massive star has a strong rotation, which is very different at the poles and the equator. So rotation is expected to affect the symmetry of the mass loss rate,” said Jiang.

The team is also exploring metallicity, which in astrophysics refers to any element heavier than helium.

“Metallicity is important because it affects opacity. In our previous simulations, we assumed a constant metallicity, but massive stars can have very different metallicities,” said Jiang. “So we need to explore the parameter space to see how the structure of the stars change with metallicity. We’re currently running a simulation with one metallicity at NERSC, another at Argonne, and a third at NASA. Each set of calculations will take about three months to run.”

Meanwhile, Jiang and his colleagues already have new 2018 data to analyze. And they have a lot more simulations planned due to their recent allocation awards from INCITE, NERSC, and NASA.

“We need to do a lot more simulations to understand the physics of these special massive stars, and I think NERSC will be very helpful for this purpose,” he said.

As a Ph.D. physicist and research scientist at Berkeley Lab, Jennifer gained extensive experience in medical imaging and technical writing. She is now a freelance science writer and science-writing instructor. You can reach her at:


About NERSC and Berkeley Lab
The National Energy Research Scientific Computing Center (NERSC) is a U.S. Department of Energy Office of Science User Facility that serves as the primary high-performance computing center for scientific research sponsored by the Office of Science. Located at Lawrence Berkeley National Laboratory, the NERSC Center serves more than 6,000 scientists at national laboratories and universities researching a wide range of problems in combustion, climate modeling, fusion energy, materials science, physics, chemistry, computational biology, and other disciplines. Berkeley Lab is a DOE national laboratory located in Berkeley, California. It conducts unclassified scientific research and is managed by the University of California for the U.S. DOE Office of Science. »Learn more about computing sciences at Berkeley Lab.