computer-aided design for dna self-assembly: process and applications€¦ · ·...

1

Computer-Aided Design for DNA Self-Assembly: Process and Applications

Chris DwyerAssistant ProfessorDept. of Electrical and Computer EngineeringDept. of Computer ScienceDuke University

ICCAD 2005

Motivation

[Annotated with CNT technology, original source: George Bourianoff and ITRS, ca. 2003.]

log Length (m)

log Cost ($/gate)

log Switching time (s)

2

Outline

• DNA Basics

• Self-assembled Nanostructures– DNA Scaffolds

– DNA Guided Self-assembly

• CAD Tool Support

• Self-assembled Systems– New Constraints

– Alternative Architectures

• Conclusions

DNA Basics

• A DNA strand:– A linear array of bases (A, T, G, and C)– Directional (one end is distinct from the other)– In nature, the source of genetic information

• DNA will form a double helix:– When the bases on each strand (aligned “head-to-

toe”) are complementary: A with T, and G with C

– But only under certain “natural” environmental conditions (low) temperatures (Tm: sequence dependent) and in an ionic solution.

3

DNA Basics

• DNA hybridization is the process that forms the double helix

• Sequence and temperature control the hybridization event

∆T

DNA Basics

• A common form of the double helix (B-form) has some well-known geometric properties:– 3.4 Å per base pitch along the helix– One complete turn between every 10th and 11th base

• Flexibility: the bonds along the sugar-phosphodiester backbone of each strand can rotate– double stranded DNA has a ~50nm persistence

length (fairly rigid)– single stranded DNA has a strongly-sequence

dependent persistence length (but, it’s flexible)

4

Outline

DNA Basics






• Conclusions

Self-assembled Nanostructures

• Self-assembly is ubiquitous in nature• Generally defined as spontaneously generated order

• Thermodynamics drive the self-assembly process– we can guide the process by the choice of materials and

environmental conditions

A

B

∆TA·B

< 20 nmfeature sizes

5

DNA Scaffolds - Geometry

• The geometric properties of double strands can form specific, controlled self-assembled nanostructures:

∆T

3.4 Å

DNA Self-assembled Tiles

9 strands

Cost ($) is proportional to the total number of unique strands (& quantity)

6

DNA Scaffolds – Hierarchical Assembly

• Self-assembly can occur in hierarchies (reduces cost):– tiles (from single strands to tiles)– grids (from tiles to grids)– lattice (from grids or tiles to larger lattice)

30 nm

DNA Scaffolds - Functionalization

• Tiles can be functionalized (decorated) with nanoscale components (thus, the DNA serves as a scaffold)

• Tiles can be functionalized before OR after grid/lattice assembly

• Example chemical functionalities include:– biotin / streptavidin– DNA / nanoparticle (rods, spheres, etc.)

7


• Biotin / streptavidin (protein + active chemicals)• The DNA provides a scaffold for the protein

BELOW: AFM images of some grids functionalized with streptavidin

AFM images of a 1.4 Tb/in2 ROM (barcode)

The manufacturing scale is incredible: ~1016 grids per mL!

“Letters”: ~60nm on a side(1 experiment made ~1014 of each)

Trivia: The collection of books and manuscripts in the Library of Congress contains ~1014 letters.

A Brief Interlude About Yield

• The term “yield” is well-defined in multiple fields– Chemistry/Physics/Materials Science: extrinsically (mass)– Engineering: pass/fail (devices, circuits, systems)

• Yield in DNA self-assembly is ambiguous– Reason 1: Surface deposition is the major technique used

to assay experimental results. Substrate-to-substrate variations change the deposition rate!

– Reason 2: Partial products are common but there is no functional test (unlike with current silicon processes)

– It comes down to undefined specifications

8


• Perhaps in the future....

Crossed carbon nanotube“FET” / SBT DNA Self-assembly

+

• Nanotechnology,vol. 13, pp. 601-604, 2002.

Outline

DNA Basics

Self-assembled NanostructuresDNA Scaffolds

• DNA Guided Self-assembly




• Conclusions

9

DNA Guided Self-assembly

• Nanoparticles (rods, spheres, etc.) can be functionalized with DNA

• DNA hybridization stabilizes interactions between particles if the strands are complementary

• Sequence design and particle choice yields controlled nanostructure formation


• Example: A two particle tether

∆T

10


• An active component:– ring-gate FETs (RG-FETs) (or surrounding-gate

FETs)


• Active components for circuitry: Au – CdSe – Au (metal, semiconductor, metal or MSM) rods

500 nm wide

11


• IEEE Trans. on VLSI, vol. 12, pp. 1214-1220, 2004.

• IEEE Trans. on Nano.,2 (2): pp. 69-74, 2003.

• Nanotechnology,vol. 13, pp. 601-604, 2002.

• Perhaps in the future...– The fabrication of integrated electronic systems

Self-assembled Nanostructures

• Recap: Two Fabrication Methods– Scaffolds– DNA Guided Assemblies

Scaffolds

Nanorod assemblies

30 nm

12

Outline

DNA Basics






• Conclusions

• New technology fabric : New tool support– Goal: apply conventional circuit design approaches to

these new technologies

• First, identify a design context:– Tool flow– Layout tools– DNA sequence design

• The big picture: Moving towards full system design...

CAD Tool Support

13

Power & Timing Estimate

Device-levelDescription

SystemDescription

Synthesis Tools

ArchitecturalSimulator (custom)

BehavioralVerification

FunctionalVerification

SPICE

Layout Tools (custom) Layout

Tool Flow

Assembly Order &DNA Sequences

Assembler(custom)

Extractor(custom)

Back-annotatedCircuit

SPICE (custom)

Timing & PowerVerification

Self-assembledFabrication

LayoutOrders

Tool Flow

14

CAD Tool Support – Circuit Layout

DNA scaffold layout tool DNA-rod layout tool

• Bootstrap the automated / cell layout systems with manual layout tools and standard cell designs

CAD Tool Support – Optimized Fabrication

• The new aspects for the process tools: – DNA sequence design– Assembly orders (unique per design)

Assembly Order &DNA Sequences

Assembler(custom)

Self-assembledFabrication

LayoutOrders

15

CAD Tool Support: Wrap-up

• Current tool status:Cluster-based sequence optimizationLayout toolsCarbon-nanotube & MSM device models for a custom SPICE kernel (semi-empirical)Assembly orders / “artwork” gen. (for large circuits)

• Tool wish list: 1. yield-aware design optimizations, 2. refined (high ω) device models,3. better automated full custom support.

Outline

DNA Basics



CAD Tool Support



• Conclusions

16

Self-assembled Systems

• There are a variety of self-assembled systems– crossbars, micron-scale assemblies, biological

systems...

• The Systems Focus:

self-assembled computer architectures

New Constraints

• Self-assembly imposes:– chaos / randomness at some length scale (>1-10 µm)

• DNA hybridization imposes:– order at some length scale (< 1-5 µm)

• The two can work together but some fundamental assumptions must change:– Wire / bus interconnect

• No large-scale interconnect networks / limited local– Severe area / cost tradeoff

• Large (> 1-10 µm on a side) circuit footprints are impractical– Reliability

• The substrate can be defective• The devices can be defective

17

Alternative Architectures

• The task: Given a device technology, design a system– The new constraints prevent wholesale adoption of

conventional architectures / system designs

• Two common solutions given a defect-prone technology: – reconfigurable resources– redundant components (e.g. TMR, n-MR,

multiplexing, etc.)


• (Self-) Reconfigurability is key, however....

• The large number of simple processing nodes in a system (as many as we can assemble, ~1014 +) precludes the use of an explicit defect map

• The goal: To stitch a sufficient number of computational resources together to execute application code

18


• Four systems: Oracles, DAMP, NANA, and nSIMD:– Oracles: DNA computing with a device twist that

enables rapid (electrical) re-use of a DNA computation– DAMP: Decoupled Array Multi-processor, SIMD without

an interconnection network- embarrassingly parallel codes – only

– NANA: Nanoscale Active Network Architecture, general purpose but imbalanced due to a large communication/execution ratio- under utilized resources

– nSIMD: (nano) SIMD, similar to NANA but applies a SIMD model onto a reconfigurable network topology. Utilization is high due to a depth first network traversal.


Self-assembledComputational nodes

Self-assembledInterconnect

Defect model includes:• Rotation, position• Connectivity• Fail-stop nodes (unrealistic)

19

Wrap-up

• DNA Basics






• Conclusions

Conclusions

Shell

Arm

Core

Self-assembled device theory

20

Conclusions

Some demonstrations of self-assembled devicesThis technology is on its way...

30 nm

Conclusions

Self-assembled computer architectures and systems

– Oracles: Re-useable DNA computations– DAMP: Decoupled Array Multi-processor– NANA: Nanoscale Active Network Architecture– nSIMD: (nano) SIMD

21

Acknowledgements

Vijeta JohriVincent MaoJaidev PatwardhanConstantin Pistol

Juan BermudezLauren CohenCurt HartingJosh JohnsonJoe Tadduni

• Graduate students

• Undergraduate students

Research Sponsors

• AFRL FA8750-05-2-0018• NSF CCR-0326157• NSF EIA-9972879

22

Basic system criteria – a framework

(i) Linear signal transduction

(ii) Non-linear signal modulation by another signal

(iii) Signal amplification / restoration

(iv) Signal noise immunity

(v) Circuit patterning and interconnect

(vi) Scale of device integration

(vii) Energy consumption

(viii) Application runtime performance

10X

• DNA computing• Oracles

• ASICs, FPGAs, etc.

• Conventional serial & parallel machines• Decoupled array multi-processor (DAMP)

Assembly-time

Run-time

Temporal spectrum of Computation

IEEE Computer, vol. 38, pp. 56-64, 2005.

23

Organization & Architecture

R0* R1* R2 R3 R4ACC*

OperationB C D S

Status bitsWR

• Nanotechnology,vol. 15, 1688-1694, 2004.

• Ph.D. dissertation, Univ. of North Carolina, Chapel Hill, 2003.

• IEEE Computer, vol. 38, pp. 56-64, 2005.


Question0 Answer0

log2(n) bits. .

.

Questionn-1 Answern-1

Oracles Decoupled array multiprocessor (DAMP)

• Truth table defines binding rules

• Each tile is implemented by a self-assembled circuit

AA BB CCii CCooSS

Addition oracle example

0 0 000

0 0 110

1 1 001

0 1 100

0 1 011

1 1 111

1 0 011

1 0 100

A B SCiCo

24

0 0 11

0

1 1 00

1

0 1 01

1

1 0 01

1

LSB

MSB

A = 0 0 1 1B = 0 1 0 1

Sum = 1 0 0 0

S = 1 0 0 0

“3 + 5 = 8”


A B SCi

Co

• Each tile implemented using logic circuitry

Bit from the truth table

A B SCiCo


25



OperationB C D S

Status bitsWR

• Nanotechnology,vol. 15, 1688-1694, 2004.

• Ph.D. dissertation, Univ. of North Carolina, Chapel Hill, 2003.

• IEEE Computer, vol. 38, pp. 56-64, 2005.


Question0 Answer0

log2(n) bits. .

.

Questionn-1 Answern-1

Oracles Decoupled array multiprocessor (DAMP)

Basic system criteria

10X

Device-level simulationAND

Real-device parameter extraction

Interconnect & integration (SPICE, etc.)

Organization / architecture& application performance

(SimpleScalar, custom, etc.)

26

• Electrical behavior very similar to conventional MOSFETs

• E.g., the ring-gated FET

-1.6x10-6

-1.4x10-6

-1.2x10-6

-1.0x10-6

-8.0x10-7

-6.0x10-7

-4.0x10-7

-2.0x10-7

0.0x100

-1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0

Dra

in-t

o-so

urce

Cur

rent

Drain-to-source Voltage

Ids(Vgs= 0.00)Ids(Vgs=-0.05)Ids(Vgs=-0.10)Ids(Vgs=-0.25)

P-FET IV Curves

0.0x100

5.0x10-7

1.0x10-6

1.5x10-6

2.0x10-6

2.5x10-6

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Dra

in-t

o-so

urce

Cur

rent

Drain-to-source Voltage

Ids(Vgs=1.00)Ids(Vgs=0.90)Ids(Vgs=0.80)Ids(Vgs=0.75)

N-FET IV Curves

Case study: Silicon nanowires to DAMP

IEEE Trans. Nano, 2 (2): pp. 69-74, 2003.

Interconnect & IntegrationCase study: Silicon nanowires to DAMP

27



OperationB C D S

Status bits

WR

Decoupled array multiprocessor (DAMP)

Case study: Silicon nanowires to DAMP

COST(R1, R2) // Save last ∆X, copy F into accADDI(∆fi) // Accumulate the next intervalSTORE(R2) // Save itLOAD(Mi-1) // Load the correction factor for this intervalCOST(R0, R1) // Save the correction, load the specific Ti valueADDI(-∆Xi) // Subtract the current interval's end value (T)WAITNLT // any processor that didn't end the integration at T...SETB // all processors that DID end at T, set the B bit...RESUME // all-aboardWAITB

Application Performance (DES)Case study: Silicon nanowires to DAMP

28

0123456789

10

DAMP IBM BlueGene/L NEC EarthSimulator

SETI@Home Intel Pentium 4

Log

sear

ch ti

me

(sec

)


0

2

4

6

8

10

DAMP IBM BlueGene/L NEC EarthSimulator

SETI@Home Intel Pentium 4

Log

scal

e

Energy/op (fJ)


29

Electrolyte gated carbon nanotubes

S. Rosenblatt, Y. Yaish, J. Park, J. Gore, andP. L. McEuen, 2002.

Electrolyte gated carbon nanotubes• CMOS vs. CNT ring oscillators (per inverter)

2.91.6~55018nm‡

59.03.4929.670.1µm*

0.170.179524CNT-20nm

27.51.6730.445nm*

47.02.6928.765nm*

74.52.6617.820.13µm*

3058.8714.530.18µm†

Trans. Energy (aJ)Power (uW)fmax (GHz)Technology

† – Verified against MOSIS reference device T3AZ, Dec. 2003.‡ – ITRS 2003 prediction.* – Berkeley predictive technology models, Y. Cao et al., 2000-2002.

computer-aided design for dna self-assembly: process and applications€¦ · ·...

Documents