A Tribute to A. Richard Newton

Rich was extremely dynamic and energetic, entrepreneurial, and, most of all, a visionary ywithout bounds and a continuing source of thought-provoking ideas.

inviting all of us to dismiss…inviting all of us to dismiss borders and explore the

unexplored.

Acknowledgements

The contributions of the following people to this t ti tl i t dpresentation are greatly appreciated:

Adam Arkin, Hugo De Man, Luca de Nardis, Simone Gambini Jay Keasling Steve LevitanSimone Gambini, Jay Keasling, Steve Levitan, Subhasish Mitra, Clark Nguyen, Korneel Rabaey, Jaijeet Roychowdhury, Alberto y, j y y,Sangiovanni-Vincentelli, Ellen Sentovich, Stan Williams, Jacob White, the sponsors of the GSRC and the FCRP program, DARPA, NSF, and … Richard Newton

Outline

Beyond MicroelectronicsBeyond MicroelectronicsKeys to Microelectronics’ SuccessSuccess in the “BeyondSuccess in the Beyond Microelectronics” ArenaTh R FlThe Reverse FlowReflections

Microelectronics in the 40’s and 50’s

First IC, Texas Instruments, 1958

Junction transistorJunction transistor, AT&T, 1950

Point-contact transistor, AT&T, 1947

The Physics Era:Search for workable devices building substrates and manufacturing strategiesSearch for workable devices, building substrates, and manufacturing strategies

Leading to Astonishing Results TodayScaling enabled integration of complexScaling enabled integration of complex systems with hundreds of millions of devices on a single die

SUN Niagara-2ISSCC 07, 500M trans.

IBM Power-6ISSCC 07, 700M trans.SSCC 0 , 00 t a s

Intel KEROM dual coreISSCC 07, 290M trans.

IBM/Sony Cell ISSCC 05, 235M trans.

The Waning Days of Moore’s Law

Limitations imposed by physicsLimitations imposed by economicsp y

may ultimately end its long run (22 nm – 16 nm – 12 nm – …?)or may not …

Beyond MicroelectronicsExciting times again Researchers intensely exploringExciting times again … Researchers intensely exploring broad range of novel components (nano)

Molecular switchesMolecular switchesJ. Heath, Caltech

MEMS disc resonatorC. Nguyen, UCB Polymer ultra-thin FET

Subramanian, UCB

Nano-optics

Plenty of others….

CNT FETsWong, Dai, Stanford

The Physics, Chemists, and Material Scientists Era:Search for workable devices, building substrates, and manufacturing strategies

Beyond MicroelectronicsExciting times again Researchers intensely exploringExciting times again … Researchers intensely exploring broad range of novel components (bio)

Enzymes

Plenty of others….

Proteins

The Biologists Era:

DNA strandsE. coli

The Biologists Era:Search for workable devices , building substrates, and manufacturing strategies

The Potential Is Huge

Advanced sensorsDisplay and interface technologiesUltra high-speed interconnectHealth monitoringDrug creation and deliveryg yCreation of new fabricsRemoval of pollutantsRemoval of pollutantsEnergy generation….

Example: Fighting MalariaNatural artemisin cost: $1/doseNatural artemisin cost: $1/doseNeed: 700 ton / year

1-3 million people die every year from malaria300-500 million people infected

as a cure

Microbial synthesis of Artemisinin

as a cure

Reduces costby factor 10!

Courtesy: J. Keasling, UCB

Example: Energy ProductionMicrobial Fuel Cells

GeobacterGeobacterSheanellaPseudomonas Brevibacillus

EnterococcusClostridiumClostridium

Insoluble electron acceptor

[REF: P. Aelterman, K. Rabaey et al., Environ. Sci. Technol 2006]

Keys to Microelectronics SuccessMoving from the lab to a (long-term) profitable production line! What it takes:

A scalable manufacturing processA scalable manufacturing processA scalable design methodologyA crisp computational model

In other words … hard core engineering

Keys to Microelectronics SuccessA l bl f t iA scalable manufacturing process

First Planar ICFairchild, 1961.

Planar transistor

12” wafer1990s.

Planar transistorFairchild, 1959.

Keys to Microelectronics SuccessA li ti

Disciplined “platform-

Application

Kernels/BenchmarksProgramming Model:Models/Estimators

Architecture Platform

based” design methodology

Cl b t tiMi hit t ( )

Architecture(s)

Clear abstractionsStandardized interfacesConstrained design

Microarchitecture(s)

Functional Blocks,InterconnectCycle-speed, power, area

Circuits Platform

Constrained design spaceComposition rulesB d il bilit fManufacturing Interface

Circuit Fabric(s)S SV V SG

S G

SSV

V

SS SSVV VV SSGG

Broad availability of intellectual property

Manufacturing Interface

Basic device & interconnectstructures

Delay, variation,SPICE models

Silicon Implementation [Courtesy: R. Newton,A. Sangiovanni-Vincentelli]

Invariable Point: Circuits PlatformBSIM5, 2006

M1 nout nin 0 0 nmos W=510n L=100n

Interfaces remain fixed – while underlying models evolve

Schihman-Hodges, 1968Other Invariable Points:Logic, Register Transfer, MicroarchitectureInterfaces remain fixed while underlying models evolve

Scalable design rules, Mead-Conway,1980.

GDS-IIContext-aware design rules,

2006.

Crisp Computational Models

An abstract machine or programming language is called Turing complete or Turing equivalent, if it has a computational power equivalent to (i.e.,

bl f l i ) i lifi d d lcapable of emulating) a simplified model of a programmable computer known as the universal Turing machine. Being equivalent to the universal Turingequivalent to the universal Turing machine essentially means being able to perform any computational task – though it does not mean being able to performit does not mean being able to perform such tasks efficiently, quickly, or easily.

Allan Turing

[Ref: Definition from Wikipedia, the free encyclopedia]

Can this Be Repeated in the “Beyond Microelectronics” Era?

Going beyond the labGoing beyond the lab …

F th l t tFor the complete story:http://www.nature.com/nature/comics/syntheticbiologycomic/index.html

Interlude - Some TerminologyA gene is a segment of nucleic acid that contains the information necessary to produce a functional RNA product in a controlled manner. pRNA acts as a messenger between DNA and the protein synthesis complexesProteins are large organic compounds made of Myoglobin proteinMyoglobinMyoglobin proteinProteins are large organic compounds made of amino acids arranged in a linear chain and joined together by peptide bonds …. The sequence of amino acids in a protein is defined by a gene and p y gencoded in the genetic code.Enzymes: proteins that catalyze (i.e. accelerate) chemical reactions.

TIM enzyme

E(scherichia). coli: is one of the main species of bacteria living in the lower intestines of mammals.

[Ref: All definitions from Wikipedia, the free encyclopedia]E. coliE. coli

Building a Biological Oscillator (Bacterial Blinker)(Bacterial Blinker)

Select a circuit architecture

A Biological

λ-cI protein

A Biological Inverter

[Ref: Elowitz & Leibler. 2000. Nature 403:335-8] [Courtesy Jay Keasling, UCB]

λ cI gene

Building a Biological Oscillator (Bacterial Blinker)(Bacterial Blinker)

Composition rulesrules

[Ref: Elowitz & Leibler. 2000. Nature 403:335-8]

Plasmid DNA string (containing 3 genes λ cI, TetR, laCI)

Building a Biological Oscillator (Bacterial Blinker)(Bacterial Blinker)

E. coli as the chassischassis (substrate)

[Ref: Elowitz & Leibler. 2000. Nature 403:335-8]

Realizing the Blinker

UT Austin TeamiGEM 2004 Competition

Using fluorescence to display th t t

[Elowitz & Leibler. 2000. Nature 403:335-8]

the output

Pioneering Synthetic Biology

Moving from ad-hoc to structured design

[Reference: Scientific American, June 2006]

Moving from ad hoc to structured design

Engineering Tomorrow’s Designs

Th ti f l bi l i l f ti d t l

Synthetic Biology

The creation of novel biological functions and tools by modifying or integrating well-characterized biological components into higher-order systemsbiological components into higher order systems using mathematical modeling to direct the construction towards the desired end product.p“Building life from the ground up” (Jay Keasling, UCB)Keynote presentation, World Congress on Industrial Biotechnology and Bioprocessing March 2007and Bioprocessing, March 2007.

Development of foundational technologies• tools for hiding information and managing complexity

t th t b d i bi ti li bl• core components that can be used in combination reliably

Microelectronics and Synthetic Biology

A. Richard Newton2nd International Conference on Synthetic Biology2nd International Conference on Synthetic BiologyBerkeley, May 2006.

[Courtesy: Berkeley Media Services]

Engineering Tomorrows DesignsScalable Manufacturing MethodologiesScalable Manufacturing Methodologies

Need for high-yield high-throughput DNA synthesis process• N t ( l ) 500 5000 b / ith t f 10 9• Nature (polymerase): 500-5000 bases/sec with error rates of 10-9• Typical bio-synthesis process: 0.003 bases/sec with 10-2 error rate!Micro-array synthesis: parallel generation of multiple bases (oligos) –

f fusing technology remeniscent of semiconductor manufacturing

1 million “dots”/cm2dots /cm2

[Images courtesy of Affymetrix]

Engineering Tomorrows DesignsScalable Manufacturing MethodologiesThe law of exponentials …

Cost for DNA sequencing and synthesis

Time to solve a protein structureTime to solve a protein structure

Engineering Tomorrow’s Design

Learning from the Microelectronics ExperienceA Disciplined Platform Design Methodology forA Disciplined Platform Design Methodology for Synthetic Biology!?

Exploration of scalable computational fabricsExploration of scalable computational fabrics Deriving useful abstractions and interfacesDeveloping modeling and characterizationDeveloping modeling and characterization environmentsAutomating the synthesis processPopulating the design space

[2007 DAC - Session 36]

A Platform-Methodology for Biology

[Courtesy: C. Myers, Univ. of Utah]

Modeling (The Device Perspective)

Biomolecule interactions involve l t t tielectrostatics

Extremely complicated 3-D geometries and coupled PDEsFast solvers for analysis (Multipole, Multigrid, PFFT)

[Courtesy: J. White, MIT]Barnase-barstar receptor-ligand complex(Picture courtsey to J. Bardhan,M. Altman, B. Tidor)

Modeling (The Circuit Perspective)Cell Inputs: Biochemicals Output: Cell Dies (or not)Cell Inputs: Biochemicals Output: Cell Dies (or not)

Common Model

( ) ( ( )) ( )dx t b

Reduction Problem

( ) ( ( )) ( )r r rdx t F x t b u t

dt= +

( ) ( )Tr ry t c x t=

Complicated network with Low order model thatComplicated network with thousands of parameters

Low order model that captures input-output

behavior [Courtesy: J. White, MIT]

Analysis: BioSPICE[Courtesy: A. Arkin, UCB]

Creating Shared LibrariesOpening the door for synthesisOpening the door for synthesis

http://parts.mit.edu/registry/index.php/Main_Page

[Courtesy: Drew Endy and Randy Rettberg, MIT]

Creating Shared Libraries

Some Hard QuestionsC l it f b t tiComplexity of abstractions

Main reason why analog VLSI never took offSame problems plague many of the “beyond micro-l t i ” l tfelectronics” platforms

○ E.g. Natural biological components have multiple interactions with other components in the cell

The opportunity: Advanced modeling model reductionThe opportunity: Advanced modeling, model reduction and analysis techniques developed over the history of microelectronics are very applicable and valuable in this domain as well

Lack of clear computational modelsStill being experimentally “discovered”Not obvious how many are needed – that is how many y y“applications domains”Is there such a thing as an equivalent to Turing completeness in life?

Engineering Tomorrows DesignsSimilar Considerations Hold for the Nano-

Electronics and Nano-mechanics ArenasA Disciplined Platform-Based Design Methodology ─ The Process

Exploration of scalable computational fabrics Deriving useful abstractions and interfacesDeveloping modeling and characterization environmentsA t ti th th iAutomating the synthesis processPopulating the design space

I ti S l bl M f t i St t iEngineering Tomorrows Designs

Innovative Scalable Manufacturing Strategies

ContactN I i tiNano-Imprinting

Self-assemblySurfaceMicro-machining

Self assembly

Example:16 kbit nanowire crossbar memory Manufactured using two successive imprint stepsMemory density: 3 5x1011 bits/cm2Memory density: 3.5x10 bits/cmCourtesy: S. Williams, HP Labs

Scalable Computational FabricsExploring logical design using carbon-nanotube (CNT) FETs

Misalignment-immuneCNT 2 NAND G tChallenges:

Misaligned CNTs, metallic CNTs

CNT 2-NAND Gate

ANY number ofANY number of misaligned CNTs

Arbitrary logic functions~10% cell-level penaltyp y13X Energy-Delay-

Product benefit vs 32 nm CMOS

Undoped region guaranteescorrect operation

[Courtesy: S. Mitra and P. Wong, Stanford, DAC 07, Paper 51.3]

correct operation

Modeling and CharacterizationExample: Mixed Technology Optical SystemsExample: Mixed-Technology Optical Systems

Empirical modelsExperimental data

Derived modelsParametric models extracted from or verified

Analytic modelsPhysics based

fitting by lower level tools

VddPin12

1416

(mW

)

2.5V

⎤⎡⎤⎡ 2 22

rW Pin

R0246810

0 5 10 15 20 25 30 35 40 45 50

Out

putP

ower

I t P ( W)

4V ⎥⎦

⎤⎢⎣

⎡−⎥

⎦

⎤⎢⎣

⎡=

)(2exp

)(),( 2

00 zW

rzW

WIzrI

)(1

)(0 sPs

ACR

RsV optic

f

f ⋅

⎟⎟⎠

⎞⎜⎜⎝

⎛+

=Input Power(mW)

[Courtesy: S. Levitan, University of Pittsburgh]

Creating a Complete PlatformExample: Mixed Technology Optical SystemsExample: Mixed-Technology Optical Systems

250 μm

Chatoyant, Univ. of Pittsburgh

The Reverse Flow of Ideas

Ideas emerging from the “Beyond Microelectronics” arenas may help to extend the span of Moore’s Law

Manufacturing processes that do not rely onManufacturing processes that do not rely on photolithographyComputational models that cope withComputational models that cope with variability and unreliability

The Reverse Flow of IdeasBorrowing from emerging technologies to advance silicon manufacturing

Using self-assembly to create low-k interconnect

[Dail Tech[Daily Tech,May 4, 2007]

The Reverse Flow of IdeasNanometer CMOS: A search for scalable and stackable abstractions that operate correctly

in the presence of large variationsat very low SNRin the presence of failures of devices and interconnectionsinterconnections

Bio inspired computing:Bio-inspired computing:Artificial neuron

Example: State-of-the-art Synchronization

Precision Timing Element(Crystal)

Intel Itanium Clock distribution[ISSCC 05]

Clock phase and skew[P. Restle, IBM]

Oscillators as Building BlocksOsc.Type

Unit Area(μXμ)

Unit Power@ 5 GHz

#/sq.mm Tot.Power

LC 300 300 >300 W 9 2 7 WLC 300x300 >300μW 9 2.7mW

MEMS 40x30 1μW 750 7.5mW

CMOS 3x3 100μW 90000 9WCMOS 3x3 100μW 90000 9W

Ring OscillatorLC OscillatorMEMS Disc Oscillator

[Courtesy: S. Gambini, UCB]

[Courtesy: C. Nguyen, UCB]

Synchronization Inspired by Biological SystemsBiological Systems

Distributed synchronization usingDistributed synchronization using only local communications and without precision timing elements

EnergyEnergy distribution

[REF: Mirollo and Strogatz, 1990] time

Synchronization in Distributed El t i S tElectronic Systems

[Courtesy: L. De Nardis, UCB/Roma]

Quick synchronization at low cost

Concluding ReflectionsWith t h ll i iliWith nanometer challenges in silicon, microelectronics design is already escaping its borders

Nano and bio components will gradually andNano and bio components will gradually and transparently infiltrate the design space

True impact and success is only possible in the presence of truly scalable designin the presence of truly scalable design platforms – the true legacy of EDASynthetic biology a star in the making. A true opportunity for the daring … !true opportunity for the daring … !

“The future is BDA”, A. Richard NewtonIt takes broad understanding to move borders - Profound changes inborders - Profound changes in educational practices are needed to enable joint exploration of micro, nano, and bio opportunities.and bio opportunities.

Exciting times again …

For your visionsFor your visionsFor your enthusiasmFor your enthusiasmFor your enthusiasmFor your enthusiasmFor your friendshipFor your friendshipThank you Rich!Thank you Rich!Thank you, Rich!Thank you, Rich!