SESSION 12 Innovations from Outside the (ISSCC's) Box.pdf

《SESSION 12 Innovations from Outside the (ISSCC's) Box.pdf》由會員分享，可在線閱讀，更多相關《SESSION 12 Innovations from Outside the (ISSCC's) Box.pdf（97頁珍藏版）》請在三個皮匠報告上搜索。

1、ISSCC 2025SESSION 12Innovations from Outside the(ISSCCs)BoxCircuits that Solve Optimization Problems by Exploiting Physics InequalitiesIEEE 2025 International Solid-State Circuits ConferenceInnovations from Outside the Box,Session 12San Francisco,8AM Tuesday,Feb.18,2025https:/doi.org/10.1073/pnas.20

2、15192117Eli Yablonovitch,Qixin Feng,Sri Vadlamani,&Patrick Xiao UC BerkeleyElectrical Engineering,Computer Science,&Physics Depts.2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 1 of 22Are equivalent to these I

3、n-Equalities principles in Physics:1.Onsagers Principle:Least entropy generation/least power dissipation2.2ndLaw of Thermodynamics(lowest Free Energy)3.The Adiabatic method/Quantum annealing 4.The Variational principle5.Principle of Least action/Least time6.First-to-gain-threshold optimization?F?maF

4、?ma?Each inequality can be used for optimization!Equalities in Physics:2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 2 of 22Computer VisionDeep Learning;Back-Propagation?co-optimize millions of matrix element

5、s 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 3 of 221.Onsagers Principle:Least entropy generation/least power dissipation2.2ndLaw of Thermodynamics(lowest Free Energy)3.The Adiabatic method/Quantum anneali

6、ng 4.The Variational principle5.Principle of Least action/Least time6.First-to-gain-threshold optimizationPhysical Laws can be expressed as equalities,but also as inequalities:The inequalities will perform optimization for free!Which of these is best for doing optimization?2025 IEEE International So

7、lid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 4 of 2223131212A possible final configurationThe actual final configurationMinimization principles in PhysicsOnsager Principle of Least Power Dissipation I2RPerfect for Least Squares Optim

8、ization 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 5 of 22?extremumThe Onsager Principle is not ThermodynamicsNear EquilibriumAnnealing/Simulated Annealing2ndLaw of Thermodynamicsentropy S extremumContinuo

9、us Power Input!Minimum Entropy GenerationMinimized Power Dissipation I2R 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 6 of 22The Onsager Relation:L.Onsager,“Reciprocal Relations in Irreversible Processes”.II

10、.Phys.Rev.38,22652279(1931).Nobel Prize in Chemistry 1968Lars Onsager 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 7 of 22resistive coupling elementsbistableelementbistableelementbistableelementbistableeleme

11、ntbistableelementbistableelementbistableelementbistableelementbistableelementSo far it just looks like another analog computer,which would demand very high analog precision.But we would compose it of bistable analog components,which would provide a digital answer:Analog machines provide digital answ

12、ers;like a Flip-Flopbistable element=nonlinear circuit,for example parametric down-convertor 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 8 of 22minimize:A.Lucas,“Ising formulations of many NP problems,”Fron

13、t.Phys.201412J12spinsi=1,1An example:The Ising problem Minimize the Energy of Interacting Magnetic Spins i 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 9 of 22(1)Boolean satisfiability in conjunctive normal

14、form(SAT)(2)0-1 Integer Programming(3)Clique(4)Set Packing(5)Vertex Cover(6)Set Covering(7)Feedback Node Set(8)Feedback Arc Set(9)Directed Hamiltonian cycle problem(DHCP)(10)Undirected Hamiltonian cycle problem(HCP)(11)SAT with at most 3 literals per clause(3-SAT)(12)Chromatic Number(13)Clique Cover

15、(14)Exact Cover(15)Hitting Set(16)Steiner Tree(17)3-Dimensional Matching(18)Knapsack(19)Job Sequencing(20)Partition(21)Max CutThe 21 NP-complete problems described by Richard Karp,all reduce to the Ising Magnet Problem:Richard M.Karp(1972).Reducibility Among Combinatorial ProblemsPlenum.pp.85103.doi

16、:10.1007/978-1-4684-2001-2_9.ISBN 978-1-4684-2003-6.2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 10 of 22V1(t)V2(t)V1(t)V2(t)spin 1spin 1spin 2spin 2ferromagnetic,J12=+1anti-ferromagnetic,J12=1noise?Power di

17、ssipation=?,?(*)Mapping holds only when the voltage amplitudes are all equal;cosine saturates for the circuit to match the Ising Hamiltonianroughly corresponds(*)to:Ising Hamiltonian=?,?,where?are the magnets 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimizati

18、on Problems by Exploiting Physics Inequalities 11 of 22noiseV(t)+capacitance modulationpump=2signalcos(t)signalsin(t)tttLC(t)=+1=1Parametric amplificationBistability Implements an Ising spinThisampli-tudeneedsto becons-trained 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits tha

19、t Solve Optimization Problems by Exploiting Physics Inequalities 12 of 22All optimization approaches are versions of physically based Minimum Entropy GenerationYoshi Yamamoto et al.2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Ph

20、ysics Inequalities 13 of 22V1(t)V2(t)Coupled electrical oscillatorsLC electrical oscillatorsWang,Tianshi,Leon Wu,and Jaijeet Roychowdhury.New Computational Results and Hardware Prototypes for Oscillator-based Ising Machines.Proceedings of the 56th Annual Design Automation Conference 2019.ACM,2019.Co

21、upled polaritoncondensatesOptical parametric oscillatorsYamamoto,Yoshihisa,et al.Coherent Ising machinesOptical neural networks operating at the quantum limit.npj Quantum Information 3.1(2017):49.Nonlinear elementJaijeet Roychowdhury et al.Yamamoto et al.All optimization approaches are Versions of p

22、hysically based Minimum Entropy GenerationShen,Yichen,et al.Deep learning with coherent nanophotonic circuits.Nature Photonics 11.7(2017):441.Silicon Photonic Optimizers 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequ

23、alities 14 of 22Oscillator phasesPhase(rad)+0/2+/2 Voltage(mV)0510105Oscillator amplitudestime(s)00.40.80.20.61.0time(s)00.40.80.20.61.015random binary weights1,+1Fully connected 32-spin graph(SPICE)2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problem

24、s by Exploiting Physics Inequalities 15 of 22The correct solution is found out of 33,000,000 possible solutionsHamiltonian or Power Dissipation15 10610 10615 106spin configuration#circuit state(from SPICE simulation)-75+25-25-500-100+50+75+10020 10625 10630 106(verified using exhaustive searchof exa

25、ct energy solutions)Dr.Patrick Xiao 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 16 of 22Bistable Parametric Oscillators 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optim

26、ization Problems by Exploiting Physics Inequalities 17 of 22SupriyoDattaBoltzmann Machine 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 18 of 22Optimization circuitby Onsagers PrincipleData Input(Fixed)Data O

27、utput(Final Steady State)varying dynamicallybut reaches a Steady State 2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 19 of 22Optimization circuitby Onsagers PrincipleData InputData OutputTrained Bits 2025 IEE

28、E International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 20 of 22Every Electrical Circuit already follows Onsagers Principle.What is different here?By explicitly taking advantage of the“mean-square”Figure-of-Merit,I2R?R,that ap

29、pears in Onsagers Principleyou obtain:(a)1000 faster time to solution(b)1000 less energy to solutionfor mean-square optimization.2025 IEEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 21 of 22Questions?Physics Success

30、fully Implements Lagrange Multiplier OptimizationPNAS 117 pp.26639-26650(2020).https:/doi.org/10.1073/pnas.2015192117Combinatorial Optimization Using The Lagrange Primal-Dual Dynamics of Parametric Oscillator NetworksPhys.Rev.Appl.21,044042(2024)https:/doi.org/10.1103/PhysRevApplied.21.044042 2025 I

31、EEE International Solid-State Circuits Conference 12.1:Circuits that Solve Optimization Problems by Exploiting Physics Inequalities 22 of 2212.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference1 of 26p-Circuits:Neither Digital nor AnalogPr

32、esenter:Supriyo Datta*Ming-Che Li*,Archisman Ghosh*,Risi Jaiswal*Lakshmi Anirudh Ghantasala*,+,Behtash Behin-Aein+,Shreyas Sen*Purdue University+Ludwig Computing12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference2 of 26Outline p-Circuits

33、:What?How?Why?Example 1:Quantum Monte Carlo Example 2:Generative AI12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference3 of 26p-CircuitsQuantum Circuitsq-bitDigital Circuits10bit12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2

34、025 IEEE International Solid-State Circuits Conference4 of 26p-CircuitsQuantum Circuitsq-bitn q-bits generate samples from distribution with 2npossibilitiesDigital Circuits10bit12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference5 of 26p-

35、CircuitsQuantum Circuitsq-bitn q-bits generate samples from distribution with 2npossibilities“Massive parallelism”12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference6 of 26p-Circuitsn p-bits generate samples from distribution with 2nposs

36、ibilitiesp-bitreal positivecomplex12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference7 of 26p-CircuitsDigital GateComparep-bitBuilding Block12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Ci

37、rcuits Conference8 of 26p-CircuitsBuilding BlockMACRNG12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference9 of 26p-CircuitsAmplitudeClocklessClockedDigitalAnalogClockless AnalogClocked DigitalClocked AnalogClockless DigitalTime12.02:p-Cir

38、cuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference10 of 26p-CircuitsAmplitudeClocklessClockedDigitalAnalogClockless AnalogClocked DigitalClocked AnalogClockless DigitalTimeCamsariSalahuddin CMOS+Zener diodes Salahuddin FPGA+sMTJCamsariBorders et a

39、l.Nature(2019)10 p-bit circuitClocked DigitalClockless AnalogCircuits with 1000s of p-bits100s to 1000s of p-bits12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference11 of 26Quantum MCMarkov Chain Monte CarloSampling AlgorithmProbability o

40、f next statedetermined by current state p-Circuits:What?How?Why?Example 1:Quantum Monte Carlo Example 2:Generative AIClocked DigitalClockless Analog12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference12 of 26Quantum MCMarkov Chain Monte C

41、arloSampling AlgorithmProbability of next statedetermined by current stateKing et al.Nature Communications 12,1-6(2021)2-D q-Bit lattice mappedto 3-D p-Bit lattice 12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference13 of 26Markov Chain M

42、onte CarloSampling AlgorithmProbability of next statedetermined by current state2-D q-Bit lattice mappedto 3-D p-Bit lattice King et al.Nat.Comm.12,1-6(2021)Quantum MC12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference14 of 262-D q-Bit l

43、attice mappedto 3-D p-Bit lattice Chowdhury et al.Commun.Phys.6,85(2023)Quantum MC12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference15 of 26Quantum MC10 J 10W x 1 sec20 J 400 W x 50 ms20 mJ 40W x 500 s Power x time=Energyis independent

44、of parallelization12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference16 of 26Quantum MC10 J 10W x 1 sec20 J 400 W x 50 ms20 mJ 40W x 500 s 1440 p-Bits x25K iterations5 inputs,1 outputElementary operationThese numbers reflect repeated app

45、lication of an elementary operationn=512.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference17 of 26Quantum MC10 J 10W x 1 sec20 J 400 W x 50 ms20 mJ 40W x 500 s 1440 p-Bits x25K iterations/1440/25K 0.25 J/1440/25K 0.5 pJ/1440/25K 0.5 nJ5 i

46、nputs,1 outputn=5Similarenergy per operationfor other problemsElementary operation12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference18 of 26Quantum MCEnergy/operationCLOCKLESSOPS/W1010n=5-5-10-15CLOCKED1015log10(E in joules)Elementary o

47、peration12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference19 of 26Generative AIElementary operationCan other problemsof interest utilizep-gates?p-Circuits:What?How?Why?Example 1:Quantum Monte Carlo Example 2:Generative AI12.02:p-Circuit

48、s:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference20 of 26Generative AItanh12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference21 of 26Generative AIp-Bittanh12.02:p-Circuits:Neither Analog

49、nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference22 of 26Generative AIUsing same weights12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference23 of 26Generative AIRe-trainedweights12.02:p-Circuits:Neither An

50、alog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference24 of 26Generative AIGenerative AIMapping a probability distribution to anotherNOT each sample12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference25 of

51、 26Summarytanh activationp-bit activationExample 2:Generative AIExample 1:Quantum MCp-Circuits:What?How?WHY?12.02:p-Circuits:Neither Analog nor Digital,Li,Datta et al.2025 IEEE International Solid-State Circuits Conference26 of 26SummaryEnergy,E per elementary operationExample 1:Quantum MCThank you

52、for your attention!My sincere thanks to the manyoutstanding students and collaboratorswho have contributed to our understandingtanh activationp-bit activationExample 2:Generative AIp-Circuits:What?How?WHY?12.3:Reversing scattering to perform deep tissue optical imaging and the current need for a sui

53、table optoelectronic solution 2025 IEEE International Solid-State Circuits Conference1 of 16Reversing scattering to perform deep tissue optical imaging and the current need for a suitable optoelectronic solutionChanghuei YangCalifornia Institute of Technology12.3:Reversing scattering to perform deep

54、 tissue optical imaging and the current need for a suitable optoelectronic solution 2025 IEEE International Solid-State Circuits Conference2 of 16Credits:DOI:10.1117/2.1201512.006243http:/kevin- scattering problem for biological tissues Tissue scattering length 100 microns Optical interactions offer

55、 wealth of information:fluorescence,Raman,absorption Can we undo scattering?12.3:Reversing scattering to perform deep tissue optical imaging and the current need for a suitable optoelectronic solution 2025 IEEE International Solid-State Circuits Conference3 of 16Focusing through scattering medium pr

56、ovides imaging capability.Thought Experiment:Looking for a fluorescent deer*through a fog.*There is such a thing in Finland.12.3:Reversing scattering to perform deep tissue optical imaging and the current need for a suitable optoelectronic solution 2025 IEEE International Solid-State Circuits Confer

57、ence4 of 16What is phase conjugation?Reverse phase of incident wavefront will time-reverse the light field.phase conjugate mirrorhttp:/www.futureworld.dk/tech/ether/phasecon/phasecon.htmhttp:/en.wikipedia.org/wiki/Image:PhaseConjugationPrinciple.en.png100 m100 m100 mphoto-refractive crystal0.5 mm ti

58、ssuephoto-refractive crystalclear agarose RecordingZ Yaqoob,D Psaltis,M Feld&C Yang,Nature Photonics Vol 2,110(2008).E McDowell,M Cui,Z Yaqoob,and C Yang,JBO 15,025004(2010)12.3:Reversing scattering to perform deep tissue optical imaging and the current need for a suitable optoelectronic solution 20

59、25 IEEE International Solid-State Circuits Conference5 of 16Time-reversal behavior vs tissue thicknessE McDowell,M Cui,Z Yaqoob,and C Yang,JBO 15,025004(2010)Shen,Y.,Liu,Y.,Ma,C.,&Wang,L.V.(2016),JBO 21,085001(2010)1.TSOPC can be observed through tissues as thick as 7 mm at 532 nm wavelength.2.Only

60、0.02%of wavefront is recorded.3.Average number of scatterings per photon is 200!Signal with sample displaced7mm chicken sample050100150200-6-5-4-3-2-10sL TSOPC AmplitudeBallistic Component01234567Thickness(mm)log(I)Phase conjugate reconstruction12.3:Reversing scattering to perform deep tissue optica

61、l imaging and the current need for a suitable optoelectronic solution 2025 IEEE International Solid-State Circuits Conference6 of 16Phase conjugation from wave point of view.=E-field wave vectorAny pair of points on the two sides of scattering medium is linked by a fixed additive phase relationship.

62、If we launch a control EM field from one side,the constituent E-fields at any point on the other side adds randomly(random walk).If we launch a time-reversed field,the phasors add constructively in-phase and reconstructs original input.12.3:Reversing scattering to perform deep tissue optical imaging

63、 and the current need for a suitable optoelectronic solution 2025 IEEE International Solid-State Circuits Conference7 of 16Digital Optical Phase Conjugation(DOPC)M Cui and C Yang,Optics Express Vol 18,3444-3455(2010).D Wang,E Zhou,J Brake,H Ruan,M Jang,C Yang,Optica 728(2015).Capabilities:Wavefront

64、update time of 5 ms.Peak-to-background ratio of 105(degrees of control)Wavefront sensing Wavefront playbackphase conjugatemirrorhttp:/en.wikipedia.org/wiki/Image:PhaseConjugationPrinciple.en.png12.3:Reversing scattering to perform deep tissue optical imaging and the current need for a suitable optoe

65、lectronic solution 2025 IEEE International Solid-State Circuits Conference8 of 16Our ability to generate a strong reconstruction signal is dependant on the number of arrows we get to play with.Right on spot(speckle-size)Off-center(Also,depends on the number of spots(speckle-size)we want to reconstru

66、ct.)Our DOPC systems typically achieves SBR 100,000.Intensity of spot(speckle-size)to background=SBR=(N-1)/4+1)N/4 A benchmark parameter:Speckle-Strength-to-Background Ratio(SBR)12.3:Reversing scattering to perform deep tissue optical imaging and the current need for a suitable optoelectronic soluti

67、on 2025 IEEE International Solid-State Circuits Conference9 of 16TRUE(Time Reversal Ultrasound Encoded)FocusingEmploy ultrasound modulation to select appropriate optical paths for optical phase conjugation.Recording PlaybackXu,X.,Liu,H.&Wang,L.V.,2011.Nature Photonics,5(3):154W Wang*,B Judkewitz*,C.

68、DiMarizo,C Yang,Nature Comm 3,928(2012).DOPChttp:/en.wikipedia.org/wiki/Image:PhaseConjugationPrinciple.en.pngReminder:12.3:Reversing scattering to perform deep tissue optical imaging and the current need for a suitable optoelectronic solution 2025 IEEE International Solid-State Circuits Conference1

69、0 of 16TRUE focusing demonstrationtissuetissueb TRUE off.c TRUE on.(Background subtracted)Conventional fluorescence imageTRUE fluorescence image50 m50 mW Wang*,B Judkewitz*,C.DiMarizo,C Yang,Deep-tissue focal fluorescence imaging with digitally time-reversed ultrasound-encoded light;Nature Comm 3,92

70、8(2012).12.3:Reversing scattering to perform deep tissue optical imaging and the current need for a suitable optoelectronic solution 2025 IEEE International Solid-State Circuits Conference11 of 16Opsin bReaChES injected into prelimbic cortexUsing an opsin well matched to our stimulation wavelength o

71、f 532 nm(bReaChEs),we demonstrate effective stimulation using TRUE focusing.US frequency=50 MHzH.Ruan*,J.Brake*,J.E.Robinson,Y.Liu,M.Jang,C.Xiao,C.Zhou,V.Gradinaru,C.Yang;Science Advances 12,eaao5520(2017)Neuronal stimulation in 1 mm thick brain slice12.3:Reversing scattering to perform deep tissue

72、optical imaging and the current need for a suitable optoelectronic solution 2025 IEEE International Solid-State Circuits Conference12 of 16Two broad classes of focusing strategiesFeedback Conjugate GuidestarGuidestar(generally faster)R Horstmeyer,H.Ruan,C.Yang,Nature Photonics 684(2015)12.3:Reversin

73、g scattering to perform deep tissue optical imaging and the current need for a suitable optoelectronic solution 2025 IEEE International Solid-State Circuits Conference13 of 16M Cui,E McDowell,and C Yang,Optics Express Vol 18,25-30(2010).M.Jang,H.Ruan,I.M.Vellekoop,B.Judkewitz,E.Chung,and C.Yang,Biom

74、ed.Opt.Express 6,72,2015.D.Wang*,E.H.Zhou*,J.Brake,H.Ruan,M.Jang and C.Yang;Optica 2,pp.728-735(2015).M.M.Qureshi*,J.Brake*,H.-J.Jeon,H.Ruan,Y.Liu,A.M.Safi,T.J.Eom,C.Yang and E.Chung;Biomedical Optics Express 8,pp.4855-64,(2017).Demonstration of OPC through intact living tissues New Zealand rabbit e

75、ar and mouse skin flap(2mm thick).OPC spotSpeckle map measured from camera at OPBrain is squishy.Living intact brain exhibits a decorrelation time of 0.4 ms at length scale of 2.5 mm.In contrast,mouse ear gives 28 ms at length scale of 2.3 mm.Intact living samples12.3:Reversing scattering to perform

76、 deep tissue optical imaging and the current need for a suitable optoelectronic solution 2025 IEEE International Solid-State Circuits Conference14 of 16Such a system will eliminate alignment issues and data transfer bottleneck.Much like MRI needed a big huge magnet to work well,time-reversal optical

77、 focusing/imaging needs a quantum jump in device development.T Laforest,A Dupret,A Verdant,F Ramaz,S Gigan,G Tessier,E Guillaume,NEWCAS,1-4(2013)Integrated Digital Optical Phase ConjugationCurrent DOPC system.Tough to align.Really tough to operate.Slow.12.3:Reversing scattering to perform deep tissu

78、e optical imaging and the current need for a suitable optoelectronic solution 2025 IEEE International Solid-State Circuits Conference15 of 16Other applications of a iDOPCApplications where scattering is a problem iDOPC can potentially help.12.3:Reversing scattering to perform deep tissue optical ima

79、ging and the current need for a suitable optoelectronic solution 2025 IEEE International Solid-State Circuits Conference16 of 16Funding Sources:Heritage Medical Research Institute fund,Rosen fund,NIH,Amgen,and other industrial sources.Current Group Members:Simon MahlerOumeng ZhangPaul LyuMingshu Lia

80、ngHaowen ZhouMax HuangSiyuan YinSteven LinZhenyu DongJosh ZhaoMaya DicksonAnne Sullivan(Lab Manager)Collaborators:Richard Cote+team(WUSTL)Magdalena Zernicka-Goetz+team(Caltech)Shawn Cao+team(Amgen)AcknowledgmentAlumni:Joshua BrakeAlbert ChungHaowen RuanJinho KimEdward ZhouMooseok JangXiaoze OuRoarke

81、 HorstmeyerJiangtao HuangfuChao HanBenjamin JudkewitzSeung Ah LeeGuoan ZhengJian RenYing Min WangSean PangMarinko SarunicLap Man LeeJigang WuXiquan CuiIvo VellekoopEmily McDowellZahid YaqoobXin HengMatthew LewRuizhi CaoCheng ShenMichelle CuaBaptiste BlochetJian XuYan Liu 2025 IEEE International Soli

82、d-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform1 of 32Skin-Inspired Electronics:an emerging sensing and computing platformZhenan BaoK.K.Lee ProfessorChemical Engineering,Material Science and Engineering,ChemistryDirector,Stanford Wearable Electron

83、ics Initiative(eWEAR)Senior Fellow,Precourt InstituteMember,ChEM-HMember,Bio-XMember,Wu Tsai Neuroscience InstituteMember,Human Performance AllianceInvestigator,Chan Zuckerberg BioHubInnovation investigator,Arc InstituteStanford University 2025 IEEE International Solid-State Circuits Conference12.4:

84、Skin-inspired electronics:an emerging sensing and computing platform2 of 32Moving precision medicine to precision healthJ.Heikenfeld et al.,Nat.Biotechnol.,Feb.2019,pp.407-419 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing plat

85、form3 of 32B.Chu,W.Burnett,J.W.Chung,Z.Bao,Nature,Sep.2017,pp.328-330Beyond smartphone and smart watch 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform4 of 32Health information from whole body:big dataFront.Chem.,Jun.201

86、9,no.399Length scalecell 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform5 of 32Skin-inspired sensing platformSensingSignal pre-processingSignal processing(computing)R.S.Johansson&J.R.Flanagan,Nat.Rev.Neurosci.,Apr.2009,

87、pp.345359A.Chortos,J.Liu,Z.Bao,Nat.Mater.,Jul.2016,pp.937950 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform6 of 32Data acquisitionComputingData conditioningSignal transmissionAmplificationAnalog-to-digitalSoftSoftRigid

88、Signal transmissionWired or wirelessAmplificationAnalog-to-digitalAnalog Signal ProcessingDigital Signal ProcessingSignal conditioningSoft sensorsElectrical,neurochemicalforce,temperatureRigidSkin-mimicking,distributed sensing(hybrid design)Localized sensing 2025 IEEE International Solid-State Circu

89、its Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform7 of 32ComfortableInvisibleImperceptibleBiocompatibleAutonomous 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform8 of 32Image from Zhe

90、ng and Bao et al.,Science,2021Other related papers:Science 2015,Nat.Mater.2016,Science 2018,Nature 2018,Science 2023e-Skin 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform9 of 32Flexible vs.skin-likeFlexible(low bending

91、stiffness)(modulus GPa-TPa)Skin-like(soft)material(low modulus kPa-MPa)Si-wafer 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform10 of 32Need new electronic materials:soft as jelly,conductive like metal,semiconductive lik

92、e silicon+2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform11 of 32Bao and Lovinger et al.1996Heeger,MacDiamid,Shirakawa 1977 Backbone twistingPolymer bendingCharge trapsRigid and crystalline 2025 IEEE International Solid

93、-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform12 of 32Nanoconfined polymer semiconductor:reduced conformational disorderJ.Xu#,S.Wang#,Z.Bao et al.,Science,Jan.2017,pp.59-64Matrix polymer:StretchableSelf-healingBiodegradableJ.Xu,H.Wu,Z.Bao,et al.,N

94、at.Mater.,Apr.2019,pp.584-601 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform13 of 32Z.Zhang,Y.Jiang,Z.Bao et al.,Nature,Mar.2022,pp.624-630Y.Jiang,Z.Zhang,Y.Wang,D.Li,Z.Bao et al.,Science,Mar.2022,pp.1411-1417 Soft ele

95、ctrode arrays at brainstem(soft high conductivity polymers)Interactive display(Enhanced luminescence soft polymer semiconductors)2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform14 of 32High-performance stretchable semico

96、nductorsMatrix polymer:StretchableSelf-healingBiodegradableHigh purity sorted semiconducting CNTs600800100012000.00.20.40.60.81.01.2M11Absorbance(a.u.)Wavelength(nm)S22T.Lei,Z.Bao et al.,Nat.Commun.,May 2019,no.216199.997%semiconductingEnhanced charge transport through nanoconfinementJ.Xu#,S.Wang#,Z

97、.Bao et al.,Science,Jan.2017,pp.59-64J.Xu,H.Wu,Z.Bao,et al.,Nat.Mater.,Apr.2019,pp.584-601 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform15 of 32Tri-layer high-k dielectric materialOTS surface modificationLow-k(15 nm)H

98、igh-k(285 nm)Total thickness 300 nmW.Wang,Z.Bao et al.,Science,May 2023,pp.735-742.High-kLow-kS 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform16 of 32Soft integrated circuitsGen 3Gen 2Gen 15 m1 m75 m 2025 IEEE Internat

99、ional Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform17 of 3250 x20 transistor array(1 mm2)D.Zhong,C.Wu,Y.Jiang,Z.Bao et al.,Nature,Mar.2024,pp.313-320High-density high mobility soft IC 2025 IEEE International Solid-State Circuits Conference12

100、.4:Skin-inspired electronics:an emerging sensing and computing platform18 of 32528-stage ring oscillator,1056 transistors1056 transistorsfOof 190 kHz with a propagation delay per stage of 0.88 s,stage switching frequency up to 1.14 MHzLSI:large-scale integrated circuit13.5 MHzCircuit-level operation

101、 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform19 of 32Detailed protocols 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform20 of 32Mannsfel

102、d,Tee,Bao et al.,Nat.Mater.,Sep.2010,pp.859-864Pyramid(micro-structured)rubber enable high sensitivity 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform21 of 32Soft Wearable Full-Hand Electronic Skin(DigiSkin)Sensor Proto

103、type(256 Pixels)Total Build Cost$5C.Xu,B.Shi,Z.Bao et al.,submittedDr.Chengyi XuBaiyu Shi 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform22 of 32 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired

104、 electronics:an emerging sensing and computing platform23 of 32Exceed human capabilityD.Zhong,C.Wu,Y.Jiang,Z.Bao et al.,Nature,Mar.2024,pp.313-320 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform24 of 32Spikes:frequency-

105、encoded sensing response through an organic transistor circuitB.Tee,Deisseroth,Bao,Science,Oct.2015,pp.313-316 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform25 of 32Gen 2 e-skin:pattern generationArtificial mechanorece

106、ptor20kPa80kPa0-1-2-3-4-5 Postsynaptic current(A)Time(s)00.050.10.150.20-1-2-3-4-5 Postsynaptic current(A)Time(s)00.050.10.150.20-1-2-3-4-5 Postsynaptic current(A)Time(s)0.20.150.10.050Y.Kim,A.Chortos,W.Xu,Y.Liu,J.Y.Oh,D.Son,.Z.Bao&T.W.Lee,Science,Jun.2018,pp.998-1003Pressuresensor#1RingoscillatorSy

107、naptictransistorPressuresensor#2-4-3-2-10 321Postsynaptic current(A)Time(s)0 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform26 of 32Neuromorphic e-skinW.Wang,Z.Bao et al.,Science,May 2023,pp.735-742 2025 IEEE Internatio

108、nal Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform27 of 32Mechanoreceptor240 units/cm2 on fingertip20-40 x 80-150 m2 each1 per 650 x650 m2Soft CMOS artificial mechanoreceptor 200,000 transistors/cm2100 x150 m2 per neuron 2025 IEEE Internation

109、al Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform28 of 32Beyond touch-sensing:high-resolution electrophysiological sensingLiu,Liu,Bao et al.,Nat.Biomed.Eng.,Jan.2019,pp.58-68Liu,Liu,Lee,Bao et al.,PNAS,Jun.2020,pp.14769-14778Heart Muscle64 ch

110、annels over 0.5 cm24x8 array,100 m electrode widthBrain Khatib,Zhao,Wei,Bao et al.,bioRxiv,2023Y.Jiang,Z.Bao et al.,Science,Mar.2022,pp.1411-1417 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform29 of 32M.Khatib,E.Zhao,S.

111、Wei,J.Dunn,Z.Bao et al.bioRxiv 2023.10.02.560482Gut probe:150 sensors,250 m x 10 cmBrain probe:300 single-unit brain recording,150 m x 10 mmSense:pressure,strain,pH,dopamine,serotonin,electrophysiology For implantable high-density sensing systems 2025 IEEE International Solid-State Circuits Conferen

112、ce12.4:Skin-inspired electronics:an emerging sensing and computing platform30 of 32Simultaneous measurement in brain and gut in behaving animalchocolateDA5-HTJ.Li,X.Chen,Z.Bao et al.,Nature,Jun.2022,pp.94-101 2025 IEEE International Solid-State Circuits Conference12.4:Skin-inspired electronics:an em

113、erging sensing and computing platform31 of 32Data acquisitionComputingData conditioningSignal transmissionAmplificationAnalog-to-digitalSoftRigidSkin-mimicking,distributed sensing-signal conditioning(hybrid design)Direct bio-interfaceEMGEEGTactile for humanoid robotNeuroStringE-skin 2025 IEEE Intern

114、ational Solid-State Circuits Conference12.4:Skin-inspired electronics:an emerging sensing and computing platform32 of 32NSF,DoD-AFSOR,ONR,AROSamsung ElectronicsCZ Biohub,Arc InstituteStanford Bio-XStanford Wu Tsai Neuroscience Institute,Human Performance AllianceStanford Wearable Electronics Initiat

115、ive(eWEAR)Karl Deisseroth(BioE),Xiaoke Chen(Biology)James Dunn(Surgery),Ada Poon(EE)Leanne Williams(Psychiatry)Todd Coleman(BioE/EE)Boris Murmann(EE),Paul George(Surgery)Bianxiao Cui(Chemistry),Sergiu Pascal(Psychiatry)Pablo Paredes(Psychiatry),Mert Pilanci(EE),Alberto Salleo(MSE),Carolyn Bertozzi(Chem)Joe DeSimone(ChemE/Radiology)