1 Hermann Kohlstedt Nanoelektronik Technische Fakultät Christian-Albrechts-Universität zu Kiel Memristive Devices in Analog Neuromorphic Circuits NanoNetwork Workshop_Bergen_June 2013
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Hermann Kohlstedt
NanoelektronikTechnische Fakultät
Christian-Albrechts-Universität zu Kiel
Memristive Devices in Analog Neuromorphic Circuits
NanoNetwork Workshop_Bergen_June 2013
2
V = 1000 cm3
ab
c
Achip 1 cm2
N 2 104 chips
Ptotal 106 W 1 MW (!)
Ntransistors 2 104 1010 2 1014
PMOSFET 5 nW
A Brain replaced by Computer Chips
approx. number of synapses
Pbrain 25 W
Psynapse 250 fW
3
Contents
• Introduction
• Neurobiology – A few Milestones
• Neuromorphic Electronics
• Two examples: Pavlov`s Dog and an Amoeba
• A memresistive Flash cell
• Summary
4
Computer - Brain
Computer: Arithmetic operation
Brain: Pattern Recognition / Associative Memory
2376492 = 1541,5875
Vacation:
Computing Gap
Introduction
5
DendriteSynapse
AxonSoma
Neurons for information processing
Analog VLSI and Neural SystemsCarver Mead, Addison‐Wesley 1989, p. 44
M. Mahowald, R.Rodney Douglas, A silicon neuron, Nature 1991
Bible of analog VLSI for Neural Circuits: A survey of Bio‐Inspired and other alternative ArchitecturesD. Hammertrom in:Nanotechnology, Vol. 4, Ch. 10, p. 252Wiley, 2008, ed. by R. Waser
Data spikes
Introduction
Review: G. Indiveri et al. Neuromorphic silicon neuron circuits,frontiers in Neuroscience 5, article 73 (2011).
6
Pulse duration:3, 5 ms (in electronics: 60 ns)
Signal speed ‐ along the axon:100 m/s(in electroncis 2.4 x 108 m/s)
Spikes – the information unitsIntroduction
7
Neurobiology – A few Milestones
In Search of Memory, Eric R. Kandel, W. W. Norton & Company, New York 2006.S. R. Cajal , La fine structure des Centres Nerveux, Proc. R. Soc. London (B) 1894 , 55 , 444
Santiago Ramón y Cajal
In other words: He suggest already that something like a synaptic cleft must exist! (in 1890!!)
Cajal: Learning means, that the synaptic interconnection are not fixed. They adjust in correspondence to the input signals from the environment.
8
Donald O. Hebb.
“The memory in brains is distributed overthe whole “system” but certain regions store different aspects!”
Neurobiology – A few Milestones
Hebbs learning rule: When an Axon of cell A excites cell B and repeatedly or persistently takes part in it's firing,some growth process of metabolic changes take place in one or both cells. Thus, that's efficiency is increased!
"Cells that fire together, wire together."
D. O. Hebb , The Organization of Behavior , John Wiley , New York 1949 .
9
Neurobiology – A few Milestones
T. V. P. Bliss and T. LØmo, Long‐Lasting Potentiation of Synaptic Transmission in the Dentate Area of the anaesthetized Rabbit Following Stimulation of the Perforant Path, J. Physiol. 232, 331 (1973).
From Molecules to Networks, Ed. John H. Byran and James L. Roberts, Academic Press 2009:J. H. Byren et al., Learning and Memory Basic Mechanisms: , Chap 19 p. 541
Hippocampal Brain Slice
Long Term Depression (LTD)Long Term Potentiation (LTP)
10
What means learning in biological Systems?
Three Levels
1BehaviorPsychology
Implicit learningExplicit learning
2Networks
Architecture
3Nerve CellsBiochemistry
Automatic in quality:habituation, sensitization,classical conditioning
Conscious or declarative:Recall people, places, facts, and events etc.
11
In Search of MemoryEric R. KandelW. W. Norton Company 2006
A reductionistic Principle
Aplysia California: a Snail
To bridge the Gap between Behavior and Cell Biology
E. R. Kandel, Science 294, 1030 (2001).
12
In Search of Memory, Eric R. Kandel, W. W. Norton & Company, New York 2006.
A reductionistic Principle
13
L. O. Chua,Memristor – the missing circuit element, IEEE Trans. Circuit Theory 18, 507 (1971).See also: materials today Dec. 2011 Memory matters and MRS Bulletin, Resistive switching phenomena in thin films, Feb. 2012
Leon Chua`s Memristor
I
V
14
Which memristive Device should I use?
NanoionicsR. Waser , R. Dittmann , G. Staikov , K. SzotAdv. Mater. 2009
Ferroelectric Tunnel JunctionsAndrè Chanthbouala, et al.Nature Nanotechnology 2012
You have the choice – a few Examples
Ionics and Tunnel BarriersD. S. Jeong et al. Solid‐State Electronics 63, 1 (2011)
Ti‐Oxide
Spin Transfer Torque Devices
MgO
P. Krzysteczko et al., Adv. Mater. 2012
NanoinonicsD. B. Strukov, G. S. Snider, D. R. Stewart, R. S. Williams, Nature 2008, 453, 80.
Reviews: Doo Seok Jeong et al. Rep. Prog. Phys. 75 (2012)S. D. Ha and S. Ramanathan, JAP 110 (2011)
15
Memristive Devices for Neuromorphic Systems
Sung Hyun Jo et al., Nano Lett. 10, 1297‐1301 (2010).
• synaptic plasticity: spike timing dependent plasticity (STDP)
Memristive devices as artifical synapses
T. Ohno et al., Nature Materials 10, 591–595 (2011).
• precondition of learning: long term potentiation
16
Image No. 0030628Image No. 0030628Credit:Credit:The Granger Collection, The Granger Collection, NYC NYC —— All rights reserved.All rights reserved.
Pavlov`s Dog: Classical Conditioning
17
• Experiment to understand implicit learning in biological systems.
before conditioning
after conditioning
IVAN PETROVICH PAVLOV (1905)
Experimental Psychology and Psychopathology in Animals, Vol. 1 p. 47‐60, Ivan P. Pavlov, Lectures on Conditioned Reflexes, International Pub., New York 1928
Pavlov`s Dog: Classical ConditioningAssociative Learning
18
• Voltage dividercomprizing a memristive device
• Electrical circuit layout: single memristive device implemented in an analogue circuitry
Adder Comparator
+
-
+
-
Reference set-point: Threshold Vcth
RM VM
OP1
OP2R1
Alertness Vout
Unconditional stimulus (UCS)
Conditional stimulus (CS)
Neural mediating circuit for associative learning
Vcth
Vmth
M. Ziegler, et al., Advanced Functional Materials, 22, 2744 (2012)/ExperimentalO. Bichler et al. Neural Computation 25, 549 (2013)/ExperimentalY. V. Pershin and M. Di Ventra, Neural Networks 23 (2010)/ Emulator
19
• circuit with threshold voltages for the comparator and mem device
Vbell < Vcth & Vfood> Vcth
Vbell + Vfood > Vpmth (before conditioning) & Vbell > Vcth (after conditioning)
Implicit learning
20
• Pt/SiO2/Ge0.3Se0.7/Cu memristive device in voltage divider
• synaptic potentiation via transition LRS to HRS
• effective threshold voltage of the device
Device requirements
-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6
-0.5
0.0
0.5C
urre
nt (m
A)
Voltage (V)
0.47 kΩ
Vnmth
= -0.18V
Vpmth
= 0.33V1
2
3
4
R. Soni et al., J. Appl Phys. 110, 054509 (2011).
21
Learning behavior in biological systems
Unicellular organism, able to solve mazes
Interesting candidate to study basic cognitive functions
T. Ueda, Hokkaido University
Amoeba: Physarum polycephalum
• Anticipation to enviormental changes for periodic repetition
22
Amoeba Anticipate Periodic EventsT. Saigusa, A. Tero, T. Nakagaki, Y. Kuramoto, Phys. Rev. Lett, 100, 018101 (2008)
Temperature
favorable
unfavorable
Humidity
Biological experiment:
anticipated events
23
R = 1 Ω L = 2 H C = 1 F
resonant circuit
R L
C
A memristive circuit model to mimic an amoebaY. V. Pershin, S. La Fontaine and M. Di Ventra, Phys. Rev. E 80, 021926, 2009.
Simulation: Electronic Emulatormemristive device
Problems for experimental implementation:
Circuit parameter
Using a real memristive device
24Y. V. Pershin and M. Di Ventra, Adv. Phys. 60 (2011) and references therein
Periodic input pattern needed for learning
Amoebae anticipation
Simulation
25
Electronic circuit
R = 100 Ω L = 100mH
C = 50 nF
R = 10 kΩ
memristive device
I Output current
26
High off‐resistance for an ideal LC circuit
Change to on‐state requires a threshold voltage
High Reset voltage in respect to set voltage
Requirements for the memristive device
TiO2‐x
Al
Ag
1.
27
Experimental implementation:M. Ziegler, et al. An electronic implementation of amoeba anticipationApplied Physics A (2013)
anticipated events
28
Amoeba anticipation Periodic EventsT. Saigusa, A. Tero, T. Nakagaki, Y. Kuramoto, Phys. Rev. Lett, 100, 018101 (2008)
Better anticipation to environmental changes for periodic repetition
29
Response (µA
)
Amoeba anticipation
Non-periodicpattern
Periodicpattern at resonancefrequency
Vin (V)
Vin (V)
30
Which memristive Device should I use?
NanoionicsR. Waser , R. Dittmann , G. Staikov , K. SzotAdv. Mater. 2009
Ferroelectric Tunnel JunctionsAndrè Chanthbouala, et al.Nature Nanotechnology 2012
You have the choice – a few Examples
D. S. Jeong et al. Solid‐State Electronics 63, 1 (2011)
Ti‐Oxide
Spin Transfer Torque Devices
MgO
P. Krzysteczko et al., Adv. Mater. 2012
NanoinonicsD. B. Strukov, G. S. Snider, D. R. Stewart, R. S. Williams, Nature 2008, 453, 80.
Reviews: Doo Seok Jeong et al. Rep. Prog. Phys. 75 (2012)S. D. Ha and S. Ramanathan, JAP 110 (2011)
31
Floating Gate Transistor as Memristive Device?
• Memristive operation mode of a single EEPROM cell
• Reduction to a two‐terminal device: simultaneous read/write
H. C. Card and W.R. Moore, Electronic Letters 25, 805 (1989).C. Diorio, P. Hasler, B.A. Mimich, and C. A. Mead, IEEE Trans. on Elec. Dev. 43, 1972 (1996).
What about:
Three terminal devices:WriteReadErase
32
A two‐terminal MemFlash‐cell
M. Ziegler, et al., Appl Phys. Lett. 101, 263504 (2012).
• Reduction to a two‐terminal device: simultaneous read/write
33
MemFlash
34
•Memristive devices with improved performance:Yield, parameter spread, retention, etc.
• System architecture:Mixed signal circuits including memristive devices
•Which neurobiological schemes are essential ?Long Term Potentiation, Spike Time Dependent Plasticity, Feedback Loops, Coding, Encoding etc.
How large is the benefit of memristive Devices for Neuromorphic Electronics?
Doo Seok Jeong et al. Towards artificial and synapses: a material point of viewRSC Advances 3, 3169 (2013).
35
Thanks to …
Martin Ziegler, Mirko Hansen, Christoph Riggert, Rohit Soniand Marina Ignatov
Thorsten Bartsch
Wolfgang Krautschneider, Dietmar Schröder
Karlheinz Ochs, Thomas Mussenbrock
Doo Seok Jeong
AG Nanoelektronik 2012
Paul Meuffels
Financial support from Schleswig‐Holsteins Landesgraduiertenförderung is gratefully acknowledged.
36
…and Axel as Pavlov`s Dog,…
…my daughter Nora for painting her dog
Related Documents
