Of all the brilliant people who have worked at Sandia in its 75 years, Jackie Chen has distinguished herself, not just through innovative work in the field of combustion, but also by defining entirely new ways to advance hers and other areas of scientific endeavor.
Jackie’s undergraduate career in Ohio State University’s mechanical engineering program was just the start of a learning journey that would include fluid mechanics, combustion and high-performance computing. Her will to innovate went so far beyond her initial training that she breathed a new specialty into existence and became a pioneer of its practice: computational simulation of turbulent reacting flows with complex chemistry.
“When I completed my doctoral study program at Stanford, focused on computational fluid dynamics, and came back here to Sandia, I had the choice at that time of moving over to the Combustion Research Facility, which was rapidly growing and in its heyday,” Jackie said. “That was one early fork in the road that felt like a big risk because I didn’t know a whole lot about combustion.”
Jackie said that her managers encouraged her to learn on the job — about several different disciplines.
New trend, big learning curve
“When I moved to the CRF, I was immediately taken under the umbrella of the Basic Energy Sciences, Gas Phase Chemical Sciences Program — and my main job was to apply first principles direct numerical simulations to study fundamental turbulence-chemistry interactions in combustion,” Jackie said. “There were many world-renowned experts in laser diagnostics at the CRF, among other things, and I was worried I wouldn’t fit in.”
She said her sponsor at the time encouraged her to incorporate detailed chemical kinetics into her modeling. She also witnessed how rapidly computing was evolving from room-sized Cray Vector supercomputers to massively parallel distributed computing. Jackie was, to put it simply, studying flames embedded in homogeneous turbulence, the so-called “flame-in-a-box,” and how air turbulence would affect it by wrinkling and strain. The computers to model those interactions were much less powerful, so she could only afford to transport a single global reaction.
“I realized that there’s no way that I could do that on my own. I needed to consult with experts on the different machine hardware, understanding what programming models to use for my reacting flow simulations,” she said. “So, I attended computer science conferences that were completely out of my area, even though I felt like a fish out of water.”
The connections she made by going places she was unfamiliar with would pay off for the rest of her career. Some colleagues at the Center for Turbulence Research at Stanford shared a piece of early computer code to help her modeling. But Jackie quickly began to realize that she might just be ahead of a wave about to crest across the entire world.
“In the early ’90s, high-performance computing was changing,” she said. “Killer micros (microprocessor machines that were able to do some of what supercomputers could do) were coming online. Large-scale Beowulf computing clusters were coming online, and so I realized that my VECTORAL code they had given me optimized for Cray Vector supercomputers was not going to work well on these distributed clusters.”
Evolving in real time
Jackie said she also realized that she lacked the expertise to convert her code to a language that would work on emerging systems built on rapidly evolving microprocessor technologies. That was when she kicked into high gear something that would become a signature for her career: connecting different interdisciplinary sciences to create something new.
“I solicited the help of computer science experts at Pittsburgh Supercomputing Center to help me with the task of creating a parallel version of the code I needed,” she said. “They basically threw out everything that was in the original code, and we modified not only the computer science infrastructure of the software using the message passing interface protocol for scalable parallelism on distributed machines, but also the underlying numerical discretization algorithms to make it more computationally efficient.”
At the time, computer scientists working with her on data analysis and visualization had no concept of flame dynamics or the physics associated with turbulence-chemistry interactions. They were able to ingest the input data but couldn’t interpret the results. So, Jackie, with her computer science colleagues, created better feature tracking tools to analyze and render flame and turbulence features embedded in a sea of computational data. The tools ultimately enabled her to understand the causality between the complex turbulent flow and flame interactions. With that nexus of understanding, Jackie was able to not only use the most powerful computing systems, but also model even more complex interactions occurring at the finest scales of the turbulence-flame coupling.
“Before direct numerical simulation and other high-fidelity simulation methods matured, researchers and developers relied on lower dimensional representations to model the behavior of a flame in highly turbulent, real-world fluctuations,” she said. “There were a lot of assumptions, and oftentimes they were inadequate. As DNS gradually matured it complemented experiments by providing benchmark data for validation of model assumptions.”
Finding power in partnerships
Soon, another hurdle arose.
“We needed to put more realism into the simulations,” Jackie said. “But supercomputers were becoming twice as powerful every couple of years, making the codes we were running inefficient, if not obsolete.”
Jackie had to find a new collaboration that could keep pace with the rapid advance of supercomputers — she had to turn constant transition into continued evolution. She turned to DOE’s Office of Advanced Scientific Computing Research. This office was at the forefront in developing high-performance computing initiatives that actively supported collaborations between applied mathematicians, computer scientists and computational scientists and engineers. The office recognized early on that such collaborations were needed to enable the computational science and engineering community to exploit rapidly evolving capabilities of supercomputers. In the aughts of the 21st century, Jackie was able to capitalize on the Scientific Discovery through Advanced Computing program by collaborating with computer scientists on scientific data management and visualization of her combustion simulations.
Following on the heels of Scientific Discovery through Advanced Computing, Jackie led one of the three application centers, the Exascale Simulations of Turbulent Combustion Center as part of the Exascale Computing Initiative, a joint partnership between Advanced Scientific Computing Research and NNSA, that laid the foundation for developing an exascale computing plan to enable science and engineering applications to make effective use of future heterogeneous supercomputers. As part of the Exascale Computing Initiative, she helped develop a process where scientific problem requirements influence computer architecture design and technology paths, and constraints on these architectures and technology paths inform formulation and design of algorithms and software. There, she worked closely with applied mathematicians developing adaptive mesh refinement algorithms for combustion solvers and computer scientists developing dynamic task-based programming systems well suited for heterogeneous computer architectures with many GPUs. She continued to tap the foundations laid by Scientific Discovery through Advanced Computing and the Exascale Computing Initiative in the Exascale Computing Project, another joint venture between the DOE Office of Science and NNSA that provided exascale-capable modeling and simulation solutions to address critical challenges in scientific discovery, energy assurance, economic competitiveness and national security.
With Advanced Scientific Computing Research sponsorship, Jackie led a team from several DOE labs to develop domain name systems tools that simulate many more decades of length and time scales, greater chemical complexity, including multiphysics, and with complex geometries.
Today, it is feasible to simulate the burning rate and emissions of alternative carbon-free fuels, such as hydrogen or ammonia, in gas turbines for dispatchable power, and flame stabilization near lean blow-off of different drop-in sustainable aviation fuels for propulsion.
From big data to big results
The simulations were capturing a lot of data, but Jackie had to teach computer scientists how to tell the programming which data were important.
“I’m able to run my calculation, but now I’ve got several hundred terabytes of raw data that I must sift through to find, to analyze, and pull out some meaningful information,” she said. “So, I went again back to my computer science colleagues at the University of California at Davis, who are experts in scientific visualization, and they specialize in methods to visualize, drill down, segment and track features of interest in large data. It just takes a person like me to define the combustion or fluids features we are looking for. They’re able to apply math constructs to help aggregate statistics about features that are buried in mountains of data, which is really cool.”
Jackie and her computing partners are producing full 3D simulations of interactions of more than 40 different chemical species at the micrometer scale — a far cry from the one-dimensional simulations she was able to produce a generation ago.
Connect the dots
Like a translator at the United Nations, Jackie learned to speak different languages to get to a point where her collaborators all began speaking the same language and surging forward in the research. But the success of this approach spread to sponsors and areas of research no one saw coming.
At one of the conferences where she was swimming with a new school of fish, she met Mike Kweon from the U.S. Army Research Laboratory. Jackie recalls that he was interested in whether these high-fidelity simulation tools could help understand ignition processes in intermittent and continuous engines for unmanned aerial systems.
“I formed a collaboration with the Army and applied many of the tools that we had developed under sponsorship from DOE to the Department of Defense problems,” she said. “We have also applied our DNS capabilities to understand differences in chemical and physical properties of sustainable aviation fuels in aero-gas turbine engines that affect how a flame stabilizes near lean-blowoff and soot particle sizes that affect contrail formation that the DOE Vehicle Technologies Office is interested in.”
The chemistry of a collaborator
In her career, Jackie has earned a lot of recognition. To name a few, she was elected a member of the National Academy of Engineering in 2018, received the Achievement Award by the Society of Women Engineers that same year, and was named a DOE Office of Science Distinguished Scientist Fellow in 2020.
At the heart of things, Jackie said modestly, she is still a fluid mechanics and combustion engineer.
“I have superficially dabbled in computer science more by osmosis through my real computer science colleagues,” she said. “It’s the same thing with algorithms, with the applied mathematicians. I learned numerical algorithms, algorithms for forward, and now also for machine learning and AI to accelerate DNS with lots of chemistry.”
Jackie takes a lot of enjoyment from that constant, new learning through collaborating with others in different fields.
“I am kind of outgoing. I’m not shy,” she said. “So, I think that helps to build relationships, especially when you’re in uncomfortable situations. I have a willingness and persistence to learn something, even if it means embarrassing myself with a colleague when I ask them to explain a concept to me for the 20th time. If you’re willing to stick your neck out there and work with them, eventually you get it.”
That mixing of different perspectives opens new possibilities and new futures for everyone.
“Oftentimes with collaborators, it’s like a lightbulb turns on in both of our heads,” she said. “We are working at this interdisciplinary divide, and if you’re willing to put in the effort, there’s a lot of low-hanging fruit. Discovering new physical understanding or applying new tools in a way that nobody else has thought of doing — that part’s fun!”
A career built by bringing people together for greater innovations is exactly the career that makes Jackie integral to Sandia’s 75 years of success.
“I think we need each other,” she said. “I feel like I don’t belong to birds of any feather. I like flying across different flocks.”