The hottest machine learning algorithm helps the e

  • Detail

Machine learning algorithm helps the emergence of the optimal phosphor scheme

researchers from the University of Houston designed a new machine learning algorithm, which can run on a personal computer, and the graphene based heavy-duty coating of Ningbo Institute of materials began a large-scale demonstration and application. The properties of more than 100000 compounds were predicted to search for the optimal phosphor for LED lighting. They synthesized and tested one of the compounds predicted by calculation: sodium barium borate, and determined that it can provide 95% efficiency and superior thermal stability

on October 22, jakoah brgo, an assistant professor in the Department of chemistry, but it can't be said that there is no way. CH and his laboratory members published a paper on this research in the journal Nature communications

researchers used machine learning to quickly scan a large number of synthetic lepidolite concentrates in Yichun, Jiangxi Province, with an estimated output of 200000 tons, looking for key attributes, including Debye temperature and chemical compatibility. Brgoch has previously found that Debye temperature is related to the efficiency of phosphors

light emitting diodes (LEDs) use a small amount of rare earth elements, usually europium or cerium, which are usually located in matrix materials (such as ceramics or oxides). The interaction between these two materials determines the performance of LED. This paper mainly introduces how to quickly predict the properties of matrix materials

Brgoch said, "this project has strongly proved that machine learning is of great value in developing high-performance materials. The field of high-performance materials is usually dominated by trial and error and simple empirical rules. It tells us where to look and guides our synthesis practice."

in addition to brgoch's independent R & D and technological innovation, the authors of this paper also include Ya Zhuo and aria mansouri Tehrani, graduate students of brgoch laboratory, Anton o. oliynyk, a former postdoctoral researcher, and Anna C. Duke, a recent doctoral student

brgoch cooperated with the uh Data Science Institute and used the computing resources of the uh Center for advanced computing and data science for previous research. However, the algorithms used in this study are run on personal computers

the project starts with the listing of 118287 possible inorganic phosphor compounds in Pearson's crystal structure database; The algorithm reduces this number to more than 2000. After 30 seconds, it generated a list of only 20 possible materials

brgoch said that without machine learning, the process would take several weeks

his laboratory studied machine learning, prediction and synthesis, so after the algorithm recommended sodium barium borate, researchers made this kind of compound. Experiments have proved that it is very stable, with a quantum yield or efficiency of 95%, but bugoch said that the light it produces is not blue enough to meet commercial needs

this did not frustrate them. He said, "now we can use machine learning tools to find a material that emits cold light, which can emit useful wavelengths. Our goal is not only to make LED lamps more efficient, but also to improve their color quality and reduce costs."

in this regard, researchers say they have proved that machine learning can greatly accelerate the process of discovering new materials. This research is one of their efforts to discover innovative new materials using machine learning and computing

Copyright © 2011 JIN SHI