An MIT team enlists AI to facilitate the detection of an interesting material phenomenon that could lead to electronics with no energy dissipation.
Superconductors have long been considered the dominant approach to realizing electronics without resistivity. In the past decade, a new family of quantum materials, “topological materials”, have offered an alternative but promising means for achieving electronics without energy dissipation (or loss).
Compared to superconductors, topological materials offer some advantages, such as robustness against disturbance. To achieve dissipation-free electronic states, a major route is the so-called “magnetic proximity effect”, which occurs when magnetism slightly penetrates the surface of a topological material. However, observing the proximity effect has been challenging.
Image credits: Geralt via Pixabay, free license
The problem, according to Zhantao Chen, a mechanical engineering PhD student at MIT, “is that the signal people are looking for would indicate that the presence of this effect is usually too weak to be detected conclusively by conventional methods.” Is.”
That’s why a team of scientists — based at MIT, Pennsylvania State University, and the National Institute of Standards and Technology — decided to try a nontraditional approach, which ended up yielding surprisingly good results. .
what’s under and between the layers
For the past few years, researchers have relied on a technique called polarized neutron reflectometry (PNR) to investigate the depth-dependent magnetic composition of multilayered materials, as well as to observe phenomena such as the magnetic proximity effect. In a PNR, two polarized neutron beams with opposing spins are reflected from the sample and collected at a detector.
“If a neutron encounters a magnetic flux, such as that found inside a magnetic material, which has the opposite orientation, it will change its spin state, resulting in different signals measured from the spin up and the neutron beam. can spin,” explains Nina Andrjevic, PhD in materials science and engineering. Consequently, the proximity effect can be detected if a thin layer of a normally non-magnetic material – placed immediately adjacent to a magnetic material – is shown to become magnetic.
But the effect is very subtle, extending to depths of only 1 nanometer, and ambiguities and challenges can arise when it comes to interpreting experimental results. “By bringing machine learning into our methodology, we hope to get a clearer picture of what is happening,” said Norman C., in the Department of Nuclear Science and Engineering. Rasmussen Career Development Professor, who was leading the research team, noted. That hope was indeed born, and the team’s findings were published in a paper Applied Physics Review,
The researchers investigated a topological insulator – a material that is electrically insulating in its interior but can conduct electric current at the surface. They chose to focus on a layered material system consisting of the topological insulator bismuth selenide (Bi.)2I know3) interfered with the ferromagnetic insulator europium sulfide (EuS).
with2I know3 In itself, is a non-magnetic material, so the magnetic EUS layer dominates the difference between the signals measured by the two polarized neutron beams. However, with the help of machine learning, the researchers were able to identify and quantify another contribution to the PNR signal – the induced magnetization in Bi2I know3 at the interface with the adjacent EUS layer.
“Machine learning methods are highly effective at deriving underlying patterns from complex data, making it possible to understand subtle effects such as proximity magnetism in PNR measurements,” says Andrjevic.
When the PNR signal is first fed into a machine learning model, it is highly complex. The model is able to simplify this signal so that the proximity effect increases and thus becomes more specific. Using this pared-down representation of the PNR signal, the model can then quantify the induced magnetization – indicating whether or not a magnetic proximity effect is observed – along with other characteristics of the material system, such as thickness, density and Roughness component layers.
AI. better see through
“We have reduced the ambiguity that has arisen in previous analyses, thanks to a doubling in resolution obtained using machine learning-assisted approaches,” say graduate researchers Leon Fan and Henry Heiberger, who participated in this study.
This means that they can sense physical properties on a length scale of 0.5 nm, half the typical spatial range of the proximity effect. It’s like writing on a blackboard from 20 feet away and not being able to form any words. But if you can cut that distance in half, you might be able to read the whole thing.
The data analysis process can also be significantly accelerated through reliance on machine learning. “In the old days, you could spend weeks messing with all the parameters until you could get a simulated curve to fit the experimental curve,” Lee says. “It may take many tries because the same [PNR] The signal can correspond to different combinations of parameters.”
“The neural network gives you an answer right away,” Chen says. “No more guesswork. No more trial and error.” For this reason, the framework has been installed in some reflectometry beamlines to support the analysis of a wide variety of materials.
Some outside observers have praised the new study—the first to evaluate the effectiveness of machine learning in identifying proximity effects, and one of the first machine-learning-based packages to be used for PNR data analysis. . “The work by Andrjevic et al. provides an alternative route for capturing fine details in PNR data, showing how high resolution can be achieved consistently,” said Kang L. Wang, Distinguished Professor and Raytheon Chair, Electrical Engineering at the University of California, Los Angeles.
“This is really an exciting progress,” commented Chris Leighton, Distinguished McKnight University Professor at the University of Minnesota. “Their new machine learning approach can not only speed up this process but squeeze even more material information out of the available data.”
The MIT-led group is already considering expanding the scope of its investigation. “The magnetic proximity effect isn’t the only weak effect we care about,” Andrjevic says. “The machine learning framework we developed is easily transferable to a wide variety of problems, such as the superconducting proximity effect, which is of great interest in the field of quantum computing.”
Written by Steve Nadis
Source: Massachusetts Institute of Technology
