Memristive devices for neuromorphic computation Luís Guerra IFIMUP-IN (Material Physics Institute of the University of Porto – Nanoscience and Nanotechnology.

Memristive devices for neuromorphic computation

Luís GuerraIFIMUP-IN (Material Physics Institute of the University of

Porto – Nanoscience and Nanotechnology Institute)

New Challenges in the European Area: Young Scientist’s 1st International Baku Forum

23rd of May, 2013

Outline

• The Memristor• Applications• Neuromorphic Computation• Fabrication• Results• Willshaw Network• Conclusions

Theorized in 1971[1], physically achieved in 2008[2]:- Two-terminal passive circuit element;- Resistance depends on the history of applied voltage or current;- Self-crossing, pinched hysteretic I-V loop, frequency dependent.

The Memristor

From [2]: D. B. Strukov, G. S. Snider, D. R. Stewart, and R. S. Williams, Nature 453, 80 (2008).

From: Y. V. Pershin and M. Di Ventra, Advances in Physics 60, 145–227 (2011)

𝜔1≫𝜔2≫𝜔3

[1] Chua, L. Memristor - The Missing Circuit Element. IEEE Transactions On Circuit Theory CT-18, 507–519 (1971).

ApplicationsResistive Random Access Memories (ReRAM)- Non-volatile, reversible resistive switching;- High-speed and high ON/OFF ratio;- High-density;- Possibly multi-level;

Neuromorphic computation – “the use of very-large-scale integration (VLSI) systems, containing electronic analog circuits, to mimic neuro-biological architectures present in the nervous system”

- Uncanny resemblance to biological synapses.

HPToshibaSandisk

SamsungPanasonic

From: Mead, C. Neuromorphic electronic systems. Proceedings of the IEEE 78, 1629–1636 (1990).

Neuromorphic ComputationEven the simplest brain is superior to a super computer,the secret: ARCHITECTURE!

From: Versace, M. & Chandler, B. The brain of a new machine. Spectrum, IEEE (2010).

Human brain:- 106 neurons / cm2

- 1010 synapses / cm2

- 2 mW / cm2

Total power consumption: 20 Watts

Memristors:- Cheap- Power efficient- Small

FabricationTwo-terminal resistance switches, typically a thin-film metal-insulator-metal (MIM) stack:- Ion-beam for film deposition;- Optical litography for microfrabrication.

Metals:Ag, Al, Cu, Pt, Ru, Ti.Insulator:

HfO2

Device area:1 – 100 μm2

150 μm2

From: Strukov, D. B. & Kohlstedt, H. Resistive switching phenomena in thin films: Materials, devices, and applications. MRS Bulletin 37, 108–114 (2012).

-3 -2 -1 0 1 2 3-15

-10

-5

0

5

10

15

20

Cu

rre

nt (

mA

)

Voltage (V)

1 2

1

2

Results

-2 -1 0 1 2-20

-15

-10

-5

0

5

10

15

20

Cu

rre

nt (

mA

)

Voltage (V)

1 2 3 4 5 6 7 8 9 10

- Bipolar switching;- SET (HRS to LRS) and RESET (LRS to HRS) processes;- SET current compliance;- Loss of hysteresis with consecutive loops.

Device area: 9 μm2

Results

-5 0 5 10 15 20-70

-60

-50

-40

-30

-20

-10

0

10

20

Cu

rre

nt (

mA

)

Voltage (V)

1 2 3 4 5 6 7

-5 0 5 10 15 20

-90

-80

-70

-60

-50

-40

-30

-20

-10

0

10

Cu

rre

nt (

mA

)

Voltage (V)

Inset showing SETs in detail

0 2 4 6 8 10 12 14 16 18 20 221E-9

1E-8

1E-7

1E-6

1E-5

1E-4

1E-3

0.01

0.1

1

10

100

1000

Cur

rent

(m

A)

Voltage (V)

- Bipolar switching;- SET current compliance;- High reset current / high Vset variability;

Device area: 1 μm2

Willshaw Network

Associative memory mapping an input vector into an output vector via a matrix of binary synapses (memristors);

Nanodevices have high defect rates Work around them!

Study of Stuck-at-0 (OFF) and Stuck-at-1 (ON) defects.

Capacity and robustness to noise can be improved by adjusting the current readout threshold, according to the type of predominant defect.

ConclusionsMemristor open possibilities for applications in:- ReRAM and Neuromorphic computation, among others.

Key features of memristors:- Resemblance to biological synapses;- High scalability, below 10 nm;- CMOS compatible;- Fast, non-volatile, electrical switching;- Low power consumption;- Cheap.

Thank you for your attention

Acknowledgments:J. Ventura, C. Dias, P. Aguiar, J. Pereira, S. Freitas, P. P. Freitas