SiGe epitaxial memory for neuromorphic computing with reproducible high performance based on engineered dislocations

Shinhyun Choi, Scott H. Tan, Zefan Li, Yunjo Kim, Chanyeol Choi, Pai Yu Chen, Hanwool Yeon, Shimeng Yu, Jeehwan Kim

Research output: Contribution to journalArticle

97 Citations (Scopus)

Abstract

Although several types of architecture combining memory cells and transistors have been used to demonstrate artificial synaptic arrays, they usually present limited scalability and high power consumption. Transistor-free analog switching devices may overcome these limitations, yet the typical switching process they rely on - formation of filaments in an amorphous medium - is not easily controlled and hence hampers the spatial and temporal reproducibility of the performance. Here, we demonstrate analog resistive switching devices that possess desired characteristics for neuromorphic computing networks with minimal performance variations using a single-crystalline SiGe layer epitaxially grown on Si as a switching medium. Such epitaxial random access memories utilize threading dislocations in SiGe to confine metal filaments in a defined, one-dimensional channel. This confinement results in drastically enhanced switching uniformity and long retention/high endurance with a high analog on/off ratio. Simulations using the MNIST handwritten recognition data set prove that epitaxial random access memories can operate with an online learning accuracy of 95.1%.

Original languageEnglish (US)
Pages (from-to)335-340
Number of pages6
JournalNature Materials
Volume17
Issue number4
DOIs
StatePublished - Apr 1 2018

Fingerprint

Data storage equipment
random access memory
analogs
filaments
Transistors
transistors
Memory architecture
endurance
learning
Scalability
Durability
Electric power utilization
Metals
Crystalline materials
cells
metals
simulation

ASJC Scopus subject areas

  • Chemistry(all)
  • Materials Science(all)
  • Condensed Matter Physics
  • Mechanics of Materials
  • Mechanical Engineering

Cite this

SiGe epitaxial memory for neuromorphic computing with reproducible high performance based on engineered dislocations. / Choi, Shinhyun; Tan, Scott H.; Li, Zefan; Kim, Yunjo; Choi, Chanyeol; Chen, Pai Yu; Yeon, Hanwool; Yu, Shimeng; Kim, Jeehwan.

In: Nature Materials, Vol. 17, No. 4, 01.04.2018, p. 335-340.

Research output: Contribution to journalArticle

Choi, S, Tan, SH, Li, Z, Kim, Y, Choi, C, Chen, PY, Yeon, H, Yu, S & Kim, J 2018, 'SiGe epitaxial memory for neuromorphic computing with reproducible high performance based on engineered dislocations', Nature Materials, vol. 17, no. 4, pp. 335-340. https://doi.org/10.1038/s41563-017-0001-5
Choi, Shinhyun ; Tan, Scott H. ; Li, Zefan ; Kim, Yunjo ; Choi, Chanyeol ; Chen, Pai Yu ; Yeon, Hanwool ; Yu, Shimeng ; Kim, Jeehwan. / SiGe epitaxial memory for neuromorphic computing with reproducible high performance based on engineered dislocations. In: Nature Materials. 2018 ; Vol. 17, No. 4. pp. 335-340.
@article{c30a52423ab14e6c83bcac6aa8cae784,
title = "SiGe epitaxial memory for neuromorphic computing with reproducible high performance based on engineered dislocations",
abstract = "Although several types of architecture combining memory cells and transistors have been used to demonstrate artificial synaptic arrays, they usually present limited scalability and high power consumption. Transistor-free analog switching devices may overcome these limitations, yet the typical switching process they rely on - formation of filaments in an amorphous medium - is not easily controlled and hence hampers the spatial and temporal reproducibility of the performance. Here, we demonstrate analog resistive switching devices that possess desired characteristics for neuromorphic computing networks with minimal performance variations using a single-crystalline SiGe layer epitaxially grown on Si as a switching medium. Such epitaxial random access memories utilize threading dislocations in SiGe to confine metal filaments in a defined, one-dimensional channel. This confinement results in drastically enhanced switching uniformity and long retention/high endurance with a high analog on/off ratio. Simulations using the MNIST handwritten recognition data set prove that epitaxial random access memories can operate with an online learning accuracy of 95.1{\%}.",
author = "Shinhyun Choi and Tan, {Scott H.} and Zefan Li and Yunjo Kim and Chanyeol Choi and Chen, {Pai Yu} and Hanwool Yeon and Shimeng Yu and Jeehwan Kim",
year = "2018",
month = "4",
day = "1",
doi = "10.1038/s41563-017-0001-5",
language = "English (US)",
volume = "17",
pages = "335--340",
journal = "Nature Materials",
issn = "1476-1122",
publisher = "Nature Publishing Group",
number = "4",

}

TY - JOUR

T1 - SiGe epitaxial memory for neuromorphic computing with reproducible high performance based on engineered dislocations

AU - Choi, Shinhyun

AU - Tan, Scott H.

AU - Li, Zefan

AU - Kim, Yunjo

AU - Choi, Chanyeol

AU - Chen, Pai Yu

AU - Yeon, Hanwool

AU - Yu, Shimeng

AU - Kim, Jeehwan

PY - 2018/4/1

Y1 - 2018/4/1

N2 - Although several types of architecture combining memory cells and transistors have been used to demonstrate artificial synaptic arrays, they usually present limited scalability and high power consumption. Transistor-free analog switching devices may overcome these limitations, yet the typical switching process they rely on - formation of filaments in an amorphous medium - is not easily controlled and hence hampers the spatial and temporal reproducibility of the performance. Here, we demonstrate analog resistive switching devices that possess desired characteristics for neuromorphic computing networks with minimal performance variations using a single-crystalline SiGe layer epitaxially grown on Si as a switching medium. Such epitaxial random access memories utilize threading dislocations in SiGe to confine metal filaments in a defined, one-dimensional channel. This confinement results in drastically enhanced switching uniformity and long retention/high endurance with a high analog on/off ratio. Simulations using the MNIST handwritten recognition data set prove that epitaxial random access memories can operate with an online learning accuracy of 95.1%.

AB - Although several types of architecture combining memory cells and transistors have been used to demonstrate artificial synaptic arrays, they usually present limited scalability and high power consumption. Transistor-free analog switching devices may overcome these limitations, yet the typical switching process they rely on - formation of filaments in an amorphous medium - is not easily controlled and hence hampers the spatial and temporal reproducibility of the performance. Here, we demonstrate analog resistive switching devices that possess desired characteristics for neuromorphic computing networks with minimal performance variations using a single-crystalline SiGe layer epitaxially grown on Si as a switching medium. Such epitaxial random access memories utilize threading dislocations in SiGe to confine metal filaments in a defined, one-dimensional channel. This confinement results in drastically enhanced switching uniformity and long retention/high endurance with a high analog on/off ratio. Simulations using the MNIST handwritten recognition data set prove that epitaxial random access memories can operate with an online learning accuracy of 95.1%.

UR - http://www.scopus.com/inward/record.url?scp=85040791877&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85040791877&partnerID=8YFLogxK

U2 - 10.1038/s41563-017-0001-5

DO - 10.1038/s41563-017-0001-5

M3 - Article

C2 - 29358642

AN - SCOPUS:85040791877

VL - 17

SP - 335

EP - 340

JO - Nature Materials

JF - Nature Materials

SN - 1476-1122

IS - 4

ER -