Condensed Matter Physics Seminar

Speaker: Jeehwan Kim, Massachusetts Institute of Technology

Title: Material challenges and opportunities in next generation electronics: from non-silicon electronics to artificial neural networks

Refreshments available at 3:45 pm.

Date: Mon, 08 Oct 2018, 4:10 pm – 5:10 pm
Type: Seminar
Location: 1400 BPS Bldg.

Abstract:
The current electronics industry has been complete­ly domina­ted by Si-based devices due to its excep­tional­ly low materi­als cost. How­ever, demand for non-Si elec­tronics is be­coming sub­stantial­ly high because cur­rent/next genera­tion elec­tronics re­quires novel func­tional­ities that can never be achieved by Si-based materi­als. Un­fortu­nate­ly, the extreme­ly high cost of non-Si semi­conduc­tor materi­als prohib­its the pro­gress in this field. I will dis­cuss about my group's efforts to ad­dress these issues. Our team has recent­ly con­ceived a new crys­talline growth con­cept, termed as “remote epi­taxy”, which can copy/paste crys­talline informa­tion from the wafer remote­ly through graphene, thus gen­erating single-crys­talline films on graphene [1-2]. These single-crystal­line films can be easi­ly released from the slip­pery graphene sur­face, and the graph­ene-coat­ed sub­strates can be re­used in­finite­ly to gen­erate single-crys­talline films. There­fore, the remote epi­taxy tech­nique can pro­duce expen­sive non-Si semi­conductor films with un­preceden­ted cost ef­ficien­cy while allow­ing ad­dition­al flex­ible de­vice func­tional­ity re­quired for current ubi­quitous elec­tronics.

Lastly, I will discuss about an ultimate alter­native com­puting solu­tion that does not follow the con­ventional von Neuman method. As Moore's law ap­proaches its phys­ical lim­its, brain-inspired neuro­morphic com­puting has recent­ly emerged as a promising alterna­tive because of its com­patibil­ity with AI. In the neuro­morphic com­puting sys­tem, re­sistive ran­dom access memory (RRAM) can be used as an arti­ficial synapse for weight ele­ments in neural net­work al­gorithms. RRAM typ­ically util­izes a defec­tive amor­phous sol­id as a switch­ing medium. How­ever, due to the ran­dom na­ture of amor­phous phase, it has been chal­lenging to precise­ly con­trol weights in arti­ficial synap­ses, thus result­ing in poor learn­ing accura­cy. Our team recent­ly demon­strated single-crys­talline-based arti­ficial synap­ses that show pre­cise con­trol of synap­tic weights, promis­ing super­ior on­line learn­ing ac­curacy of 95.1%—a key step paving the way towards post von Neumann com­puting [3]. I will dis­cuss about how we de­sign the materi­als and de­vices for this new neuro­morphic hard­ware.

References:
[1] Y. Kim et al., and J. Kim, “Remote epitaxy through graphene: Role of under­lying sub­strates on van der Waals epi­taxy,” Nature, Vol. 544, 340–343 (2017)
[2] Wei Kong et al., and J. Kim, “Polarity governs atomic inter­action through two-dimen­sional materials,” Nature Materials (2018) In Print
[3] S. Choi et al., and J. Kim, “SiGe epi­taxial memory for neuro­morphic com­puting with repro­ducible high per­formance based on engineered dis­loca­tions,” Nature Materials Vol. 17, 335–340 (2018)

Speaker Bio:
Professor Jeehwan Kim is an Associate Professor of Massa­chusetts Insti­tute of Tech­nology in the Mech­anical engineer­ing and Materi­als Science and Engineer­ing. He is a Prin­cipal Investi­gator in Research Labor­atory of Elec­tronics at MIT. Prof. Kim's group focuses on innova­tion in nano­technology for next genera­tion com­puting and elec­tronics. Before join­ing MIT in 2015, he was a Research Staff Member at IBM T.J. Watson Research Cen­ter in York­town Heights, NY since 2008. Many of his patents have been li­censed for com­mercializa­tion. Prof. Kim is a recipi­ent of 20 IBM high value inven­tion achieve­ment awards. In 2012, he was ap­pointed a “Mas­ter Inven­tor” of IBM in recog­nition of his ac­tive intellec­tual prop­erty genera­tion and com­mercializa­tion of his re­search. He is an inventor of 200 issued/pendi­ng US patents and an author of 40 arti­cles in jour­nals. He received his B.S. from Hongik University, his M.S. from Seoul National Uni­versity, and his Ph.D. from UCLA in 2008, all of them in Materia­ls Science.