By Leda Zimmerman, MIT Department of Nuclear Science and Engineering
January 30, 2022
” Architecture is both a technical and innovative field, where you try to enhance features you desire for certain type of functionality, like the size of a building, or the layout of different rooms in a house,” she states. Andrejevics work in artificial intelligence resembles that of designers, she thinks: “We begin with an empty website– a mathematical design that has random criteria– and our objective is to train this design, called a neural network, to have the functionality we want.”
Andrejevic is a doctoral advisee of Mingda Li, an assistant teacher in the Department of Nuclear Science and Engineering. As a research assistant in Lis Quantum Measurement Group, she is training her machine-learning designs to hunt for brand-new and useful traits in materials. Her deal with the lab has landed in such major journals as Nature Communications, Advanced Science, Physical Review Letters, and Nano Letters.
MIT doctoral candidate Nina Andrejević (best) has actually developed with her twin sibling Jovana (left), a PhD candidate at Harvard University, a method for screening product samples to predict the presence of topological attributes that is much faster and more versatile than other approaches. Credit: Gretchen Ertl
One area of special interest to her group is that of topological materials. “These products are an exotic stage of matter that can transfer electrons on the surface area without energy loss,” she states. “This makes them highly intriguing for making more energy-efficient technologies.”
With her sis Jovana, a doctoral candidate in applied physics at Harvard University, Andrejevic has actually developed a technique for testing material samples to forecast the existence of topological qualities that is quicker and more flexible than other approaches.
If the supreme objective is “producing better-performing, energy-saving technologies,” she says, “we should initially understand which products make great candidates for these applications, whichs something our research study can assist confirm.”
Teaming up
The seeds for this research were planted more than a year ago. “My sibling and I always stated it would be cool to do a job together, and when Mingda suggested this study of topological products, it occurred to me that we could make this an official partnership,” states Andrejevic. The sisters are more comparable than a lot of twins, she keeps in mind, sharing numerous scholastic interests. “Being a twin is a substantial part of my life and we work together well, helping each other in locations we do not comprehend.”
Andrejevics dissertation work, which encompasses numerous tasks, utilizes specialized spectroscopic techniques and information analysis, reinforced by artificial intelligence, which can discover patterns in vast amounts of data more efficiently than even the most high-throughput computer systems.
When she graduates this winter, Nina Andrejević will head to Argonne National Laboratory, where she prepares to concentrate on developing physics-informed neural networks. Credit: Gretchen Ertl
” The unifying thread among all my jobs is this concept of trying to speed up or enhance our understanding when using these characterization tools, and to consequently obtain better information than we can with more approximate or standard models,” she states. The twins research on topological materials functions as a case in point.
In order to tease out novel and possibly useful residential or commercial properties of products, researchers must interrogate them at the atomic and quantum scales. Neutron and photon spectroscopic methods can help catch formerly unknown structures and characteristics, and identify how heat, electric or magnetic fields, and mechanical stress affect materials at the Lilliputian level. The laws governing this world, where materials do not act as they may at the macro-scale, are those of quantum mechanics.
Present speculative approaches to recognizing topological products are challenging technically and inexact, potentially leaving out viable candidates. The radiation information it offers a signature unique to the sampled product.
” We desired to establish a neural network that could determine geography from a products XAS signature, a far more available measurement than that of other approaches,” states Andrejevic. “This would hopefully permit us to screen a much broader category of possible topological materials.”
Over months, the researchers fed their neural network information from 2 databases: one included products theoretically predicted to be topological, and the other contained X-ray absorption information for a broad series of products. “When effectively trained, the design ought to function as tool where it reads new XAS signatures it hasnt seen before, and tells if you if the material that produced the spectrum is topological,” Andrejevic explains.
The research duos technique has shown promising outcomes, which they have currently released in a preprint, “Machine learning spectral indications of topology.” “For me, the excitement with these machine-learning projects is seeing some underlying patterns and having the ability to comprehend those in terms of physical amounts,” says Andrejevic.
Moving towards materials research studies
After a course in nanoscience and nanoengineering, she joined a research group imaging products at the atomic scale. “This experience moved me closer to the field of materials science.”
Maker learning, critical to Andrejevics doctoral work, will be central to her life after MIT. When she finishes this winter, she heads straight for Argonne National Laboratory, where she has actually won a distinguished Maria Goeppert Mayer Fellowship, granted “worldwide to impressive doctoral researchers and engineers who are at early points in promising careers.” “Well be attempting to design physics-informed neural networks, with a concentrate on quantum materials,” she states.
This will indicate stating bye-bye to her sis, from whom she has never ever been separated for long. “It will be extremely various,” states Andrejevic. But, she adds, “I do hope that Jovana and I will collaborate more in the future, no matter the range!”
“These materials are an exotic phase of matter that can transfer electrons on the surface without energy loss,” she says. “My sister and I constantly stated it would be cool to do a job together, and when Mingda recommended this study of topological materials, it occurred to me that we could make this an official partnership,” states Andrejevic. Existing experimental approaches to identifying topological materials are challenging technically and inexact, potentially omitting practical candidates. “Well be trying to create physics-informed neural networks, with a focus on quantum materials,” she states.
Over time, these interests assembled into an academic course that shares some characteristics with the household profession, according to Andrejevic, a doctoral candidate in products science and engineering at MIT.
Doctoral candidate Nina Andrejevic integrates spectroscopy and artificial intelligence methods to determine valuable and novel properties in matter.
Born into a family of architects, Nina Andrejevic enjoyed creating illustrations of her house and other structures while a kid in Serbia. She and her twin sis shared this enthusiasm, in addition to a hunger for math and science. With time, these interests assembled into a scholarly course that shares some attributes with the family occupation, according to Andrejevic, a doctoral candidate in materials science and engineering at MIT.
