SLIDE 1
NRP SYMPOSIUM PRESENTATION
SPMS04 Li Jiang Rong River Valley High School
Spin Orbit Torque-Based Synaptic Devices in an Artificial Neural Network
2
NRP SYMPOSIUM PRESENTATION SPMS04 Li Jiang Rong River Valley High - - PowerPoint PPT Presentation
NRP SYMPOSIUM PRESENTATION SPMS04 Li Jiang Rong River Valley High - - PowerPoint PPT Presentation
NRP SYMPOSIUM PRESENTATION SPMS04 Li Jiang Rong River Valley High School Spin Orbit Torque-Based Synaptic Devices in an Artificial Neural Network 2 1. RATIONALE Significance of Project SMALLER devices with HIGHER data density
SLIDE 2
SLIDE 3
1. RATIONALE
Significance of Project
SLIDE 4
SMALLER devices with HIGHER data density
4
8-Bit Data
1 1 1 1 1 1 1 1 00000000
Conventional Electronics Spintronics
VS
SLIDE 5
INCREASING power consumption
5
Transistors Power
SLIDE 6
2. AIMS
Objectives of Project
6
SLIDE 7
Device Size Dependence
Device Size
7
(100 nm – 800 nm)
- No. of Resistance States
SLIDE 8
Hopfield Network
8
Character Recognition Simulation of SOT devices
SLIDE 9
3. METHODOLOGY
Experimental Procedures
9
(Device Size Dependence)
SLIDE 10
10 1 2
SOT Devices
(Device Size Dependence)
SLIDE 11
11
Magnetron Sputtering
1
Electron Beam Lithography Ion Milling
(Device Size Dependence)
SLIDE 12
12
Magnetron Sputtering
Si/SiO2 Substrate
1
[ [
[
[
Ta [Co/Pt]4 Ta
Electron Beam Lithography Ion Milling
(Device Size Dependence)
SLIDE 13
13
Electron Beam Exposure
1
Electron Beam Lithography
Substrate
Positive Resist
Substrate
Negative Resist
Substrate Substrate Substrate Substrate
Magnetron Sputtering Ion Milling
(Device Size Dependence)
SLIDE 14
14
Removing material to a desired depth (17.2 nm)
1
Ion Milling
Substrate
Magnetron Sputtering Electron Beam Lithography
(Device Size Dependence)
SLIDE 15
15 2
Computer with LabVIEW programme
Keithley 2400 source meter to send Iwrite pulses
Sample Electromagnets Motor with 1.8° rotation step size
Field-Induced Current-Induced
&
Arduino to control motor
(Device Size Dependence)
SLIDE 16
16 2
Field-Induced Current-Induced
&
(Device Size Dependence)
SLIDE 17
4. RESULTS ANALYSIS
Discussion of Project
17
(Device Size Dependence)
SLIDE 18
18
Field-Induced
- 10
- 8
- 6
- 4
- 2
- 1
(
)93.6 91.8 90.0 88.2 86.4 RH () H (kOe)
- 8
- 6
- 4
- 2
(
)93.6 91.8 90.0 88.2 86.4 RH () H (kOe)
f) h)
- 8
- 6
- 4
- 2
(
- )
93.6 91.8 90.0 88.2 86.4 H (kOe) RH ()
- 8
- 6
- 4
- 2
- 2
- 1
(
- )
93.6 91.8 90.0 88.2 86.4 RH () H (kOe)
a) b)
- 8
- 6
- 4
- 2
- 1
(
- )
93.6 91.8 88.2 86.4 84.6 RH () H (kOe)
c)
- 8
- 6
- 4
- 2
(
)93.6 91.8 90.0 88.2 86.4 RH () H (kOe)
d)
- 8
- 6
- 4
- 2
(
)93.6 91.8 90.0 88.2 86.4 RH () H (kOe)
e)
- 8
- 6
- 4
- 2
(
)93.6 91.8 90.0 88.2 86.4 RH () H (kOe)
g)
(Device Size Dependence)
SLIDE 19
19
Current-Induced
- 1.5
- 1.0
- 0.5
0.0 0.5 1.0 1.5
- 0.8
- 0.6
- 0.4
- 0.2
0.0 0.2 0.4 0.6 0.8
RH () J (10
7 A/cm 2)- 1.5
- 1.0
- 0.5
0.0 0.5 1.0 1.5
- 0.6
- 0.4
- 0.2
0.0 0.2 0.4 0.6
RH () J (10
7 A/cm 2)a) b)
- 1.5
- 1.0
- 0.5
0.0 0.5 1.0 1.5
- 0.8
- 0.6
- 0.4
- 0.2
0.0 0.2 0.4 0.6 0.8
RH () J (10
7 A/cm 2)- 1.5
- 1.0
- 0.5
0.0 0.5 1.0 1.5
- 0.8
- 0.6
- 0.4
- 0.2
0.0 0.2 0.4 0.6 0.8 1.0
RH () J (10
7 A/cm 2)- 1.5
- 1.0
- 0.5
0.0 0.5 1.0 1.5
- 0.2
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
J (10
7 A/cm 2)RH ()
- 1.5
- 1.0
- 0.5
0.0 0.5 1.0 1.5 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
RH () J (10
7 A/cm 2)e) f) g) h)
- 1.5
- 1.0
- 0.5
0.0 0.5 1.0 1.5
- 0.2
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
RH () J (10
7 A/cm 2)- 1.5
- 1.0
- 0.5
0.0 0.5 1.0 1.5
- 3.6
- 3.4
- 3.2
- 3.0
- 2.8
- 2.6
- 2.4
RH () J (10
7 A/cm 2)c) d)
(Device Size Dependence)
SLIDE 20
3. METHODOLOGY
Experimental Procedures
20
(Hopfield Network)
SLIDE 21
21
1
5 10 15 20 25 30 50 100 150 200 250 300 350 400 450
Number of Synapses Number of Neurons
300 synapses needed in a 5x5 block pattern (Hopfield Network)
SLIDE 22
22
2
SOT devices as synaptic weights (Hopfield Network)
SLIDE 23
23
3
SOT devices as synaptic weights
- 15
- 10
- 5
5 10 15
- 1
1 2
RH () J (10
7 A/cm 2)
- 12
- 10
- 8
- 6
- 4
- 2
2 4 6 8 10 12
(a.u.)
(Hopfield Network)
SLIDE 24
4. RESULTS ANALYSIS
Discussion of Project
24
(Hopfield Network)
SLIDE 25
25
Hopfield Network
3
Character recognition of letters ‘R’, ‘V’, ‘H’ and ‘S’ (Hopfield Network)
SLIDE 26
26
Hopfield Network
3
Character recognition of letters ‘R’, ‘V’, ‘H’ and ‘S’ (Hopfield Network)
SLIDE 27
27
5. CONCLUSION
Future Implications of Project
SLIDE 28
- Engineer devices with specific switching characteristics
- Work towards energy-efficient brain-inspired computing
28
SLIDE 29
References
29 1. Kurenkov, C. Zhang, S. DuttaGupta, S. Fukami, H. Ohno. (March, 2017). Device-size dependence of field-free spin-orbit torque induced magnetization switching in antiferromagnet/ferromagnet structures. Applied Physics Letters, 110, 092410. 2. William A. Borders, Hisanao Akima et al. (December, 2016). Analogue spin-orbit torque device for artificial-neural-network-based associative memory operation.Applied Physics Express 10, 013007.
SLIDE 30
30