NEC and Tohoku University Succeed in AI-Based New Material Development

February 12, 2018

TOKYO, Feb 12, 2018 — NEC Corporation and Tohoku University applied new technologies developed by NEC, which use AI to predict the characteristics of unknown materials, to the joint development of cutting edge thermoelectric conversion technology known as a thermoelectric (TE) device(1) using spin current(2), and achieved 100 times better thermoelectric conversion efficiency over the course of approximately 1 year.

These new technologies consist of a group of AI technologies for material development which incorporate the various types of knowledge required for material development, and technology for the batch acquisition of the large amounts of material data that AI needs to learn material properties.

The AI technologies for material development use heterogeneous mixture learning technology(3) and a number of machine learning technologies specific to material development(4), (5) independently developed by NEC. These technologies were combined with technology for the batch acquisition of material data, which enabled the acquisition and evaluation of data regarding more than 1,000 types of materials with different compositions data and resulted in much more accurate AI learning.

NEC and Tohoku University applied a development technique combining these technologies to the development of a spin-current thermoelectric conversion device, and demonstrated that it is possible to enhance thermoelectric conversion efficiency by designing an actual material in line with AI-derived new material design guidelines.

In the future, NEC and Tohoku University will develop more sophisticated AI-based technologies for predicting the physical properties of new materials and conduct further research and development aimed at the practical application of spin-current thermoelectric conversion device technology and the realization of IoT devices that continue working for years without a power supply.

Background

In recent years, cases in which machine learning technologies and analysis technologies based on informatics are applied to large amounts of data to extract hidden information and predict future trends are becoming increasingly common in a wide range of fields.

In the realm of natural sciences, there has long been research into the use of informatics in the search for new substances and materials and, especially in the biomedical, pharmaceutical and chemical fields, advances in so-called combinatorial technologies(6), which involve acquiring data through comprehensive investigation of search objects, have resulted in the extensive use of informatics in projects such as the human genome project.

Similarly, in fields that deal with solid state materials such as metals, semiconductors and oxides, techniques that make full use of the advantages of machine learning and analysis technologies from the viewpoint of shortening the duration of research and development and reducing costs are attracting attention in recent years as materials informatics (MI). Harnessing knowledge gained over many years of basic research in the natural sciences field using informatics, NEC began conducting research into analysis technologies specifically for solid state materials data more than five years ago. However, it was challenging because it took time and effort to produce and evaluate data in this field, making it difficult to obtain enough groups of data of sufficient quality required for the application of MI.

The initiatives of NEC and Tohoku University have now resulted in NEC’s independent development of AI technologies for material development that effectively analyze material data, and technology for the batch acquisition of data about many different types of solid state materials. NEC and Tohoku University have also applied a material development cycle incorporating these two types of technologies to the development of thermoelectric conversion materials using spin currents and considerably shortened the duration of material development.

Features of the new technologies

1. Development of AI technologies for material development
A material development cycle using MI handles an enormous amount of material data and it is, therefore, inevitable that the data processing and classification and the data evaluation that was previously painstakingly performed by specialist researchers can no longer keep up. Also, in the development process, the amount of material candidates is generally greater than the amount of data that can be obtained, and selection of the process for searching these candidates needs to be performed more efficiently than in the past. To address these issues, NEC developed the following AI technologies for material development which correspond to the various characteristics of actual material data.

  • AI technology for combinatorial data analysis: Machine learning technology for processing and classifying large amounts of data acquired using combinatorial techniques at high speed. Knowledge of physics/materials science is partially incorporated into existing machine learning algorithms to realize high precision, high speed data processing(4).
  • AI technology for physical property modeling: Machine learning technology for evaluating large amounts of material data. Inductive modeling of physical properties using heterogeneous mixture learning with high readability (white-box) and high precision prediction enables researchers and AI to increase their understanding of data in concert. This technology plays a very important part in the extraction of the material parameter candidates that characterize the physical properties model.(3)
  • AI technology for materials screening: Machine learning technology for efficiently searching for the target material from among a large amount of materials candidates. On the selection (screening) of materials with reference to the material parameters that characterize the physical properties model, the technology performs high-speed searches of ultra-multi-dimensional systems, which was difficult with existing Bayesian Optimization(7). This is achieved through the application of branch-and-bound type algorithm that predict moves into the future based on Combinatorial Game Theory.(5)

2. Establishment of combinatorial data acquisition platform (batch evaluation and acquisition of large amounts of data) 

NEC and Tohoku University developed a combinatorial data acquisition platform that combines theoretical calculation data acquired through material property simulation with test data acquired through proprietary thin film materials preparation/characteristics evaluation technologies, to create enough data groups of sufficient quality.

In one example of composition-spread materials shown in the photograph, thin film material with more than 1,000 different types of composition was produced on a single substrate in one film preparation process. Electrode pads, etc. according to the purpose of measurement are patterned on the evaluation sample and experimental data can be acquired efficiently through an automatic evaluation system, which was also developed independently. The quality and amount of data is vastly improved compared to previous techniques used to produce and evaluate a single material in one experiment.

In the material physical property simulation, various calculation techniques are used according to the purpose, for example, first-principles calculations, and theoretical calculations are performed to ensure correspondence with experimental data. Various types of physical property parameter group, ranging from general physical properties such as electric resistance and thermal conductivity to detailed physical properties relating to electronic state are acquired as theoretical calculation data.

3. Application to the development of a spin-current thermoelectric conversion device

A spin-current thermoelectric conversion device recovers waste heat that exists widely in society by converting it into electricity and will, therefore, enable the countless IoT devices that will be installed in the future to continue working for many years even without a power source.

The main issue with a spin-current thermoelectric conversion device is that the thermoelectric conversion efficiency performance has still not reached a practically applicable level. When the recently established material development cycle combining a combinatorial data acquisition platform and AI technologies for materials development was applied on a trial basis to the search for a new platinum (Pt) alloy, this led to various discoveries, including that the Pt alloy was a magnet and the importance of the Pt itself contained in the alloy being spin polarized.

Evaluation of the characteristics of a new alloy CoPtN that is designed to enhance spin polarization of Pt according to AI-derived knowledge confirmed that thermoelectric conversion efficiency around 100 times higher than that of previous Pt alloy was obtained. This level was also much closer to the output level of commercially available thermoelectric conversion elements that use semiconductors. It was also demonstrated that it was possible to significantly shorten development time to around 1 year.

“We will continue to further expand the search for materials using AI in the future, focusing on further improvement in the thermoelectric conversion efficiency of spin-current thermoelectric conversion elements and the development of new low-cost materials,” said Eiji Saitoh, Professor, Tohoku University.

“We aim for the early pranctical application of spin-current thermoelectric conversion elements as power source technology for IoT devices that will continue to work for years without a battery or other power source and as low cost, high performance electronic cooling technology,” said Soichi Tsumura, General Manager, NEC IoT Devices Research Laboratories.

These results were achieved as part of the Exploratory Research for Advanced Technology (ERATO) “SAITOH Spin Quantum Rectification Project” (Research Director: Eiji Saitoh, Professor, Tohoku University; Research Period: 2014 – 2020 fiscal year) of the Japan Science and Technology Agency (JST) and the Promoting Individual Research to Nurture the Seeds of Future Innovation and Organizing Unique, Innovative Network (PRESTO) “Development of Basic Technologies for Advanced Materials Informatics through Comprehensive Integration among Theoretical, Experimental, Computational and Data-Centric Sciences” (Researcher: Yuma Iwasaki, NEC Corporation; Research Period: 2017 -2021 fiscal year) of JST.

NEC plans to exhibit and showcase these research findings at nano tech 2018 – The 17th International Nanotechnology Exhibition & Conference from February 14 (Wed)-16 (Fri), 2018 at Tokyo Big Sight, Japan.

(1) A thermoelectric device is a device which converts thermal energy directly into electricity and vice versa.
(2) The spin current is a flow of a magnetic property of an electron, so-called “spin.”
(3) http://jpn.nec.com/ai/analyze/pattern.html
(4) “Comparison of dissimilarity measures for cluster analysis of X-ray diffraction data from combinatorial libraries” Y. Iwasaki et al., nature partner journal Computational Materials 3, 4 (2017).
(5) Proceedings of the 78th JSAP Autumn Meeting 5p-C18-4 (2017) R. Sawada et al.
(6) The combinatorial approach involves comprehensively investigating predicted combinations to gain an understanding of the overall trend of search candidates.
(7) Bayesian Optimization algorithms are stochastic optimization techniques that are used to search for the maximum or minimum value based on observed facts.

About NEC Corporation

NEC Corporation is a leader in the integration of IT and network technologies that benefit businesses and people around the world. By providing a combination of products and solutions that cross utilize the company’s experience and global resources, NEC’s advanced technologies meet the complex and ever-changing needs of its customers. NEC brings more than 100 years of expertise in technological innovation to empower people, businesses and society. For more information, visit NEC at http://www.nec.com.


Source: NEC Corporation

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industy updates delivered to you every week!

Fluid HPC: How Extreme-Scale Computing Should Respond to Meltdown and Spectre

February 15, 2018

The Meltdown and Spectre vulnerabilities are proving difficult to fix, and initial experiments suggest security patches will cause significant performance penalties to HPC applications. Even as these patches are rolled o Read more…

By Pete Beckman

Intel Touts Silicon Spin Qubits for Quantum Computing

February 14, 2018

Debate around what makes a good qubit and how best to manufacture them is a sprawling topic. There are many insistent voices favoring one or another approach. Referencing a paper published today in Nature, Intel has offe Read more…

By John Russell

Brookhaven Ramps Up Computing for National Security Effort

February 14, 2018

Last week, Dan Coats, the director of Director of National Intelligence for the U.S., warned the Senate Intelligence Committee that Russia was likely to meddle in the 2018 mid-term U.S. elections, much as it stands accused of doing in the 2016 Presidential election. Read more…

By John Russell

HPE Extreme Performance Solutions

Safeguard Your HPC Environment with the World’s Most Secure Industry Standard Servers

Today’s organizations operate in an environment with ever-evolving threats, and in order to protect themselves they must continuously bolster their security strategy. Hewlett Packard Enterprise (HPE) and Intel® are addressing modern security challenges with the world’s most secure industry standard servers powered by the latest generation of Intel® Xeon® Scalable processors. Read more…

AI Cloud Competition Heats Up: Google’s TPUs, Amazon Building AI Chip

February 12, 2018

Competition in the white hot AI (and public cloud) market pits Google against Amazon this week, with Google offering AI hardware on its cloud platform intended to make it easier, faster and cheaper to train and run machi Read more…

By Doug Black

Fluid HPC: How Extreme-Scale Computing Should Respond to Meltdown and Spectre

February 15, 2018

The Meltdown and Spectre vulnerabilities are proving difficult to fix, and initial experiments suggest security patches will cause significant performance penal Read more…

By Pete Beckman

Brookhaven Ramps Up Computing for National Security Effort

February 14, 2018

Last week, Dan Coats, the director of Director of National Intelligence for the U.S., warned the Senate Intelligence Committee that Russia was likely to meddle in the 2018 mid-term U.S. elections, much as it stands accused of doing in the 2016 Presidential election. Read more…

By John Russell

AI Cloud Competition Heats Up: Google’s TPUs, Amazon Building AI Chip

February 12, 2018

Competition in the white hot AI (and public cloud) market pits Google against Amazon this week, with Google offering AI hardware on its cloud platform intended Read more…

By Doug Black

Russian Nuclear Engineers Caught Cryptomining on Lab Supercomputer

February 12, 2018

Nuclear scientists working at the All-Russian Research Institute of Experimental Physics (RFNC-VNIIEF) have been arrested for using lab supercomputing resources to mine crypto-currency, according to a report in Russia’s Interfax News Agency. Read more…

By Tiffany Trader

The Food Industry’s Next Journey — from Mars to Exascale

February 12, 2018

Global food producer and one of the world's leading chocolate companies Mars Inc. has a unique perspective on the impact that exascale computing will have on the food industry. Read more…

By Scott Gibson, Oak Ridge National Laboratory

Singularity HPC Container Start-Up – Sylabs – Emerges from Stealth

February 8, 2018

The driving force behind Singularity, the popular HPC container technology, is bringing the open source platform to the enterprise with the launch of a new vent Read more…

By George Leopold

Dell EMC Debuts PowerEdge Servers with AMD EPYC Chips

February 6, 2018

AMD notched another EPYC processor win today with Dell EMC’s introduction of three PowerEdge servers (R6415, R7415, and R7425) based on the EPYC 7000-series p Read more…

By John Russell

‘Next Generation’ Universe Simulation Is Most Advanced Yet

February 5, 2018

The research group that gave us the most detailed time-lapse simulation of the universe’s evolution in 2014, spanning 13.8 billion years of cosmic evolution, is back in the spotlight with an even more advanced cosmological model that is providing new insights into how black holes influence the distribution of dark matter, how heavy elements are produced and distributed, and where magnetic fields originate. Read more…

By Tiffany Trader

Inventor Claims to Have Solved Floating Point Error Problem

January 17, 2018

"The decades-old floating point error problem has been solved," proclaims a press release from inventor Alan Jorgensen. The computer scientist has filed for and Read more…

By Tiffany Trader

Japan Unveils Quantum Neural Network

November 22, 2017

The U.S. and China are leading the race toward productive quantum computing, but it's early enough that ultimate leadership is still something of an open questi Read more…

By Tiffany Trader

AMD Showcases Growing Portfolio of EPYC and Radeon-based Systems at SC17

November 13, 2017

AMD’s charge back into HPC and the datacenter is on full display at SC17. Having launched the EPYC processor line in June along with its MI25 GPU the focus he Read more…

By John Russell

Researchers Measure Impact of ‘Meltdown’ and ‘Spectre’ Patches on HPC Workloads

January 17, 2018

Computer scientists from the Center for Computational Research, State University of New York (SUNY), University at Buffalo have examined the effect of Meltdown Read more…

By Tiffany Trader

Nvidia Responds to Google TPU Benchmarking

April 10, 2017

Nvidia highlights strengths of its newest GPU silicon in response to Google's report on the performance and energy advantages of its custom tensor processor. Read more…

By Tiffany Trader

IBM Begins Power9 Rollout with Backing from DOE, Google

December 6, 2017

After over a year of buildup, IBM is unveiling its first Power9 system based on the same architecture as the Department of Energy CORAL supercomputers, Summit a Read more…

By Tiffany Trader

Fast Forward: Five HPC Predictions for 2018

December 21, 2017

What’s on your list of high (and low) lights for 2017? Volta 100’s arrival on the heels of the P100? Appearance, albeit late in the year, of IBM’s Power9? Read more…

By John Russell

Russian Nuclear Engineers Caught Cryptomining on Lab Supercomputer

February 12, 2018

Nuclear scientists working at the All-Russian Research Institute of Experimental Physics (RFNC-VNIIEF) have been arrested for using lab supercomputing resources to mine crypto-currency, according to a report in Russia’s Interfax News Agency. Read more…

By Tiffany Trader

Leading Solution Providers

Chip Flaws ‘Meltdown’ and ‘Spectre’ Loom Large

January 4, 2018

The HPC and wider tech community have been abuzz this week over the discovery of critical design flaws that impact virtually all contemporary microprocessors. T Read more…

By Tiffany Trader

Perspective: What Really Happened at SC17?

November 22, 2017

SC is over. Now comes the myriad of follow-ups. Inboxes are filled with templated emails from vendors and other exhibitors hoping to win a place in the post-SC thinking of booth visitors. Attendees of tutorials, workshops and other technical sessions will be inundated with requests for feedback. Read more…

By Andrew Jones

How Meltdown and Spectre Patches Will Affect HPC Workloads

January 10, 2018

There have been claims that the fixes for the Meltdown and Spectre security vulnerabilities, named the KPTI (aka KAISER) patches, are going to affect applicatio Read more…

By Rosemary Francis

GlobalFoundries, Ayar Labs Team Up to Commercialize Optical I/O

December 4, 2017

GlobalFoundries (GF) and Ayar Labs, a startup focused on using light, instead of electricity, to transfer data between chips, today announced they've entered in Read more…

By Tiffany Trader

Tensors Come of Age: Why the AI Revolution Will Help HPC

November 13, 2017

Thirty years ago, parallel computing was coming of age. A bitter battle began between stalwart vector computing supporters and advocates of various approaches to parallel computing. IBM skeptic Alan Karp, reacting to announcements of nCUBE’s 1024-microprocessor system and Thinking Machines’ 65,536-element array, made a public $100 wager that no one could get a parallel speedup of over 200 on real HPC workloads. Read more…

By John Gustafson & Lenore Mullin

Flipping the Flops and Reading the Top500 Tea Leaves

November 13, 2017

The 50th edition of the Top500 list, the biannual publication of the world’s fastest supercomputers based on public Linpack benchmarking results, was released Read more…

By Tiffany Trader

V100 Good but not Great on Select Deep Learning Aps, Says Xcelerit

November 27, 2017

Wringing optimum performance from hardware to accelerate deep learning applications is a challenge that often depends on the specific application in use. A benc Read more…

By John Russell

2017 Gordon Bell Prize Finalists Named

October 23, 2017

The three finalists for this year’s Gordon Bell Prize in High Performance Computing have been announced. They include two papers on projects run on China’s Read more…

By John Russell

  • arrow
  • Click Here for More Headlines
  • arrow
Share This