Flexible Flex ible an and Str d Stret etch chab able le - - PowerPoint PPT Presentation
The 16th U he 16th U.S .S.-Kor orea ea For orum um on N on Nanotec anotechnolog hnology 2019.09.24 2019.09.24 Flexible Flex ible an and Str d Stret etch chab able le Organ Org anic ic Ar Artifi tificial cial Ner Nerve
Seoul National University (e-mail: twlees@snu.ac.kr)
The 16th U he 16th U.S .S.-Kor
ea For
um on N
anotechnolog hnology 2019.09.24 2019.09.24
2
3
Biological systems can inspire new generations of engineering devices. In our body, sensory, neural, and motor processing tasks are done extremely efficiently and robustly with extremely low energy consumption, in very little volumes The entire brain and body are put together with energy-efficient neurons and cells to robustly perform complex information-processing tasks. One can learn a lot of things from biology to develop efficient technologies, to learn to architect systems that can perform efficiently and reliably with unreliable devices, to build systems that automatically learn and adapt to a changing environment.
Biological mechanoreceptor (i) Pressure Nerve fiber (iv) Postsynaptic potential (ii) Receptor potential change Biological synapse (iii) Action potentials
4
Humanoid robots
5
Artificial Muscle Motor system Electronic Skin & Sensors Neuromorphic Artificial Nerves
Communications of the ACM, 2012, 55, 76-87
Bio-inspired electronics and robotics moves/senses/thinks like a human
6
Our Research Direction in Flexible Electronics (Tae-Woo Lee’s Group) Neuro-inspired Organic Artificial Sensory Nerves
7
ONW synaptic transistor that emulates a biological synapse not
biological spikes artificial spikes
biological EPSC artificial EPSC conducting line ion gel core- sheath ONW Axon Dedron
preneuron to the presynaptic membrane.
presynaptic spikes and is delivered to a postneuron.
Synaptic cleft Probe
Presynaptic Membrane
Large-scale alignment and integration of 1D materials is still a difficult challenge for large-scale circuit applications
8
60 70 80 90 0.0 2.0n 4.0n 6.0n 8.0n Post-synaptic current (A) Time (s)
EPSC
Random Accumulation Return to random
Accumulated Anions attracts a similar number of holes in the P3HT channel
STP
9
Schematic of the working mechanism of ONW ST for long-term plasticity
The spontaneous release of the trapped anions in the ONW is slow, inducing long-term memory.
LTP
10
Long-term plasticity
is a persistent increase in synaptic strength following a number of consecutive stimulations of a synapse.
11
2008 2010 2012 2014 2016 1a 1f 1p 1n 1μ
ONW ST NG ONW ST PCM RRAM Conductive bridge Ferroelectric Thin film ST
Energy consumption (J) Year
~1.23 fJ per synaptic event for individual ONW was successfully attained, which can even rival that of the biological synapses.
12
13
Artificial Central Nervous System – Brain-inspired Computing
Brain-inspired Organic Artificial Synapse
T.-W. Lee et al., Sci. Adv., 2016
Materials for CNS
Redox active polymer Electrochemical ion doping mechanism
PEDOT:PSS + PEI P3HT + [EMIM][TFSI]
T.-W. Lee et al., Sci. Adv., 2016
14
ONW ONW Synap Synaptic T tic Tra ransisto nsistor to r to mimi mimic c Periphe Periphera ral n l ner ervou vous s System System
Lee et al, Sci. Adv. 2016, 2, e1501326 Salleo et al, Nat. Mater. 2017, 16, 414
Neuromorphic computing & memory
Peripheral nervous system:
15
Artificial afferent and efferent nerves
Afferent nerve : axons
carrying sensory information from body Efferent nerve : axons
Applications of artificial nerves: robotics and prosthetics with the combination of sensors and motors
16
https://www.youtube.com/watch?v=IrYTD1xZVSs
17
Receptor potential Pressure Frequency of action potential Pressure Pressure Receptor potential Frequency of action potential
processes information
→ Biological mechanosensory system processes pressure information
18
Biological sensory nerves Artificial sensory nerves
19
Resistive Resistive pressure pressure sensor sensor mimicking mimicking mechanor mechanorecepto eceptor
0.0 30.0k 60.0k 90.0k 10k 1M 100M 10G 1T 100T Bias=-1V Bias=-3V Bias=-5V Resistance () Pressure (Pa)
Applied pressure decreases the resistance of pressure → sensors mainly by reducing contact resistances.
Biological SA-I mechanoreceptor:
pressure input intensity = 1-100 kPa
20
Biological sensory nerves Artificial sensory nerves
21
40 20
Voltage (V)
Time (ms)
VDD = -2.3 V, VLL = -4.6 V, VHH = 1 V
Oscillating frequency = 20-89 Hz Biological mechanosensory nerves: Action potential frequency range= 0.4-100 Hz
22
Receptor potential Pressure Biological mechano- receptor
(i) → (ii) (i) → (ii) → (iii)
Frequency of action potential Pressure
30 60 90
Supply voltage to ring oscillator (V) Pressure (kPa) 30 60 90 20 40 60 80 100 Frequency (Hz) Pressure (Pa)
Pressure sensitivity= 2-10 Hz kPa-1 Pressure sensitivity= 0.4-13 Hz kPa-1
Supply voltage Organic ring oscillator #1
(i) Pressures (ii) Supply voltages to ring
(iii) Oscillating voltages
Resistive pressure Sensor #1
frequency change Supply voltage change
23
Biological sensory nerves Artificial sensory nerves
∙ Vout of ring oscillators connected to the gate of synaptic transistors
24
Emulation of Signal Transmission in Biology Naturalistic Sensory and Motor Response
Ar Artificial tificial Peripher eripheral Ner al Nervous
System
Materials design for PNS
Short-term potentiation Volatile & Fast decay Low ion doping efficiency Electric double layer
Donor Donor Accep Accep tor tor
n
Donor-Acceptor polymer
Short Decay time Peak Peak×1/e
My Approaches
Neuromorphic Bioelectronics
25
Biological Synapses in the somatosensory system: Decay time of postsynaptic currents= 1.5-5 ms
Decay time =2.4 ms Peak Peak×1/e Time (ms) Drain current (A) 10 20
1 2 Gate voltage (V) Polymer semiconductors Polymer 1 P3HT Decay time (ms) 2.35 ± 1.02 299 ± 201 s Polymer 1
Decay times for postsynaptic currents (typically 2 to 3 ms)
Decay time of postsynaptic outputs reasonably short enough to receive pressure inputs at the desired frequency
26
∙ Increase of magnitude (A, B) & duration (C, D) of pressure → increase of peak postsynaptic currents ∙ Frequency of postsynaptic currents independent of duration of pressure stimulus ∙ SA-I receptors (pressure input frequencies are <~5 Hz, pressure durations are >~200 ms)
Pr Press essur ure e sensor sensor + ring + ring oscilla
tor + s + syna ynaptic ptic tr transistor ansistors
A B C D
27
30 45 60 75 90 105 |Fourier transform| Frequency (Hz) Pressure sensor#1 Pressure sensor#2 Ring
Ring
Synaptic transistor 69 Hz 91 Hz
20 kPa 80 kPa 20kPa and 80kPa Sum of (B) and (C)
Gate 1 Gate 2 Gate N Source Drain I
Syna naptic tr ptic trans ansis istor tor: : an ad an adder der
∙ Pressure signals from multiple pressure sensors are combined. ∙ Amplitude & frequency was maintained after signal integration in synaptic transistor
20kPa 80kPa 80kPa
Postsynaptic current (A) Time (s) 0.05 0.1 0.15 0.2
Postsynaptic current (A) Time (s) 0.05 0.1 0.15 0.2 Sum of (A) and (B) 20 kPa 80 kPa 20 kPa and 80 kPa
Postsynaptic current (A) Time (s) 0.05 0.1 0.15 0.2
Postsynaptic current (A) Time (s) 0.2 0.15 0.1 0.05
(A) (B)
(C)
Synapses Presynaptic neuron #1 Presynaptic neuron #2 Postsynaptic neuron
Biological nervous system
28
Pressure sensor#1 Ring
Synaptic transistor Pressure sensor#2
3 2 1 Postsynaptic current (A) Time (s) 1 2 3 4 5
Postsynaptic current (A) Time (s)
100 200 300 Smallest Victor-Purpura distance (DVP) between alphabets Pressure sensor +ring oscillator Pressure sensor +ring oscillator +synaptic transistor
100 75 50 25 (Hz)
Braille “E”
80 40 (kPa)
Pressure sensors Ring
Synaptic transistors
Movement recognition Braille reading
Biological mechanosensory nerves: Branched mechano- receptor structure
A larger DVP means more dissimilarity between two spike trains. Braille letters became more distinguishable
29
4 mm
Force gauge Reference electrode Stimulating electrode Direction
2 mm
Force Pressure Postsynaptic current
Hybrid monosynaptic reflex arc
Amplified voltage Connection to efferent nerve
20 40 60 80 10 20 30 40 50 60 Force of leg extension (mN) Pressure (kPa) 1.0 0.8 0.6 0.4 0.2 50 40 30 20 10 Time (s) Force of leg extension (mN) 20 40 Pressure (kPa)
leading to the actuation of the tibial extensor muscle in the leg
30
31
32
Optical stimulation Presynaptic impulses Postsynaptic response Neuromuscular junction Muscle contraction Presynaptic impulses Optical stimulation Postsynaptic response Artificial synapse Muscle contraction
necessary for artificial motor system
various motions.
Presynaptic membrane = Gate electrode Presynaptic potential = Gate voltage
Photosensitive protein = Photodetector
Synaptic cleft = Ion-gel electrolyte Neurotransmitter = Anion
Postsynaptic membrane = OSC NW Postsynaptic potential = Drain current Muscle fiber = Polymer actuator
Human optogenetic sensorimotor nervous system genetically- modified motor neurons This approach is promising to restore the motor function of defective neuromuscular systems
33
+ + + + + + + + ‒ ‒ ‒ ‒
‒
‒ ‒ ‒
+ + + + + +
Transimpedance circuit Polymer actuator Photodetector Synaptic transistor
10 20 30 40 50 60 1 2 3 100% strain Number of spikes (mm) 1 2 3 4 Output voltage (V)
0% strain 100% strain 100 spikes 100 spikes initial
artificial muscle fiber
34
Without artificial synapse (Constant displacement with constant voltage) Our synapse (Contraction of artificial muscle gradually increases as the fixed light pulses (action potentials) are applied repeatedly)
10 20 30 40 50 60 1 2 3 100% strain Number of spikes (mm) 1 2 3 4 Output voltage (V)
Nat., Commun., 2013, Nat., Commun., 2016
35
36
Presynaptic impulses Optical stimulation Postsynaptic response Artificial synapse Muscle contraction
Soft exosuit prosthetics Soft robotics Neurorobotics
Force Pressure Postsynaptic current Hybrid reflex arc Amplified voltage Connection to efferent nerve
Wearable Neuroprosthetics
37
Printed, Flexible Nano-Electronics & Energy Laboratory (PNEL) at SNU
Stanford University
POSTECH
Wentao Xu Yeongjun Lee