JANUARY/FEBRUARY 2001 17

turning to single, electron-conductingmolecules or small groups of metalatoms. The effort has received supportfrom the Defense Advanced ResearchProjects Agency in the United Statesand from the European Commissionabroad.

The limit of Moore’s LawIncreased interest in nanoelectronics,

particularly in its subfield of molecularelectronics (or moletronics), derivesfrom the electronics industry’s rapid ap-proach to the physical size limits ofsemiconductor devices. This trendpoints to the likely end of Moore’s Law(which states that the number of tran-sistors that you can fabricate as part ofa silicon-based integrated circuit dou-bles every 18 to 24 months).

For several decades, the electronicsindustry has managed to follow theSemiconductor Industry AssociationRoadmap, shrinking the linear dimen-sions of transistors and their connect-ing wires by a factor of 0.7 every threeyears, says James Ellenbogen of theNanosytems Group at the MITRECorporation, McLean, Virginia. Thisplanned improvement fueled “incred-ible investment and wealth,” he says.By 1999, the industry was fabricating180 nm lines, producing transistors

750 nm on a side. Within 10 to 15years, however, the industry will reachthe physical limits of solid-state tech-nology, known as “the Wall,” becauseof fabrication limitations imposed by

semiconductor devices’ quantum me-chanical behavior, Ellenbogen says. Atthat point, semiconductor perfor-mance improvement will still be possi-ble, but not on a planned basis. Addi-tionally, reaching the physical limits ofsemiconductor electronics has causedfabrication equipment costs to rise. To-day, a silicon fabrication facility costsin the neighborhood of US $2.5 bil-lion, and estimates predict this will riseto $25 billion by 2010, says JamesTour, a chemist involved in molecular

F O C U S

BEYOND THE WALL: COMPUTING WITH MOLECULESBy David I. Lewin

SILICON-BASED TRANSISTORS AND CIRCUITS ARE JUST TOO

BIG, SAY A SMALL GROUP OF RESEARCHERS IN THE NEW FIELD

OF NANOELECTRONICS. TO CREATE ELECTRONIC DEVICES FAR

SMALLER THAN THOSE CURRENTLY AVAILABLE, RESEARCHERS ARE

Figure 1. A molecular electronic switch. (Photo: J.M. Seminario, J. Tour, and A.G. Zacaria.)

18 COMPUTING IN SCIENCE & ENGINEERING

F O C U S

electronics research at Rice University,Houston, Texas. As insulator layers intransistors approach a thickness ofthree atomic diameters, it becomes eas-ier for electrons to leak away, says Tour,Chao Professor of Chemistry at Rice.Furthermore, at such small scales, sili-con’s electronic band structure beginsto break down; the bands separate, in-terfering with electron flow.

From the microscale to thenanoscale

For some 25 years, researchers havebeen looking at a number of nanoscalealternatives to silicon semiconduc-tors—some an extension of semicon-ductor technology, others an out-growth of polymer chemistry. Singlemolecules that act as wires or switcheshave an advantage over microelec-tronic devices in that you can makethem identically in extremely largenumbers.

“The challenge over the last decadehas been to find molecules that can naturally replace solid-state devices and continue miniaturization,” Ellenbo-gen says. Once researchers identify can-didate molecules, a number of researchgroups —including collaborations be-tween Hewlett-Packard and UCLA,Yale and Rice Universities, and in theEuropean Union—have been attempt-ing to assemble molecular-scale switchesand wires into prototype computingstructures. Some researchers have evensuggested DNA as a wire, although in-vestigators reported last year in Naturethat the nucleic acid probably wouldn’twork. Since 1996, scientists in the USand abroad have succeeded in demon-strating such switches and wires, he notes.

Molecular electronics brings to-gether electrical engineers, computerscientists, and chemists in what isclearly an interdisciplinary field. Therole of chemistry, says Tour, is to iden-

tify and synthesize the molecular de-vices. The research’s goal is to find asingle molecule that can act as switch,gate, or memory.

“The reason organic molecules areunder consideration is that [chemistshave] a long history of controlled or-ganic reactions,” says Tour. In mostcases, the researchers are investigat-ing the use of organic structures orcarbon nanotubes to create the elec-

tronic switches or wires. Some of theproposed molecules, such as Tour’smolecular wire made of acetylene-linked aromatic groups, can conductelectrons.

Reversible statesOther candidate molecules, rather

than holding electrical charges as sili-con devices do, can be placed reversiblyin one of two oxidation–reductionstates. But the key to organic molecules’usefulness in molecular electronics,DARPA program manager WilliamWarren suggests, will be the fact thatsuch molecules self-assemble. “We’renot forcing molecules into compromis-ing positions,” he says. Self-assemblywill permit rapid fabrication of elec-tronic devices with switch densities farhigher than that possible for silicon-based semiconductors.

“People are making real moleculardevices today,” says Mark Reed, chair-

man of the Electrical Engineering De-partment at Yale University in NewHaven, Connecticut. Although devicesmight currently consist of one or a fewmolecules, an explosion of results hasemerged from molecular electronicslabs over the past two years. “The fieldis evolving at a very rapid rate,” Reedsays, “but it’s too early to say whichtechnologies will win.”

Reed and Tour have successfullydemonstrated that a certain class of or-ganic molecule can act as a reversibleswitch, and that another molecule canact as a memory device by storing elec-trons as needed. However, in 1999,Hewlett-Packard Labs in Palo Alto,California, and the University of Cali-fornia at Los Angeles announced thatthey had built an electronic switch us-ing an organic chemical called rotaxane.This group. including Phil Kuekes(Hewlett-Packard) and James Heath(UCLA), linked a number of molecularswitches to form a version of a logicalAND gate—one of digital computing’sbuilding blocks.

In the summer of 1999, Heath andhis colleagues reported in Science theuse of molecules for AND and ORgates in conjunction with semiconduc-tor-technology wires; once switched,however, the molecules were not re-versible. Last year (August 2000), alsoin Science, the group reported the de-velopment of a switch using a mole-cule that is electronically reversible.Also last year, the Hewlett-Packardgroup reported self-assembly of chem-ically created wires on a silicon surface.

Improved memory retentionIn addition to increased density, mol-

ecular electronic devices present otheradvantages over silicon. Stability is onearea in which moletronic devices mayoutperform silicon-based integratedcircuits, Tour says. Additionally, mole-

The field is evolving at a

very rapid rate, but it’s

too early to say which

technologies will win.

JANUARY/FEBRUARY 2001 19

cular-based memory devices look toprovide markedly better times for hold-ing memory in the absence of power—10 minutes in certain devices under de-velopment, and occasionally up to 8 to10 hours. “My hope is that we will soonsee memory retention for as long as 10years,” Tour says. In comparison, a typ-ical silicon electronic memory deviceneeds to be refreshed 10,000 times persecond.

Unlike some other fields of nanoelec-tronics research, molecular electronicsis investigated—and the devices func-

tion—at room temperature, Kuekessays. Although researchers have not yetattempted to hermetically seal the de-vices, production devices will probablyrequire seals to protect the organiccomponents from oxygen’s destructiveeffects.

Quantum cellular automataNot all nanoelectronics approaches

involve polymers, however. At the Uni-versity of Notre Dame, Craig Lent andhis colleagues have been working toencode bits in nanoscale structures that

do not carry current, but have twocharge states.

“The basic thing that we’re after is tomake a quantum cellular automata cellbased on a single molecule,” Lent says.Other groups, including James Tourand Mark Reed, and Philip Kuekes andJames Heath, have worked toward us-ing single, current-carrying moleculesas the basis of molecular switches.

Until he and his colleagues identify asuitable molecule, Lent is using smallmetal dots with two extra electrons,which act like charged molecules. Se-

Figure 2. A still microscope image of a nanopatterned assembly. A voltage pulse was used to blow holes in an insulating self-assembled monolayer and, in solution, molecular wires (the white spots) were inserted in the predefined holes that formed. The molecules (wires) were made by Jim Tour’s group. His students and Mark Reed’s students spent a week together workingout the technique and getting the images.

20 COMPUTING IN SCIENCE & ENGINEERING

F O C U S

ries of dots can act as wires, propagat-ing the charge state, and can even pro-duce power gain, as in traditional elec-tronics. Input and output are hard, saysLent, “and output is the hardest,” be-cause charge states and not flows ofelectrons are involved. The metal dotsystem is a good model. Researchers arelearning to make the dots smaller, andhave to work at close to absolute zero(70 K), although they foresee gettingthe devices to operate at 20 K. “Thereare niche applications in which thatwould be useful,” Lent says.

Quantum cellular automata replacesthe current switch—fundamental to thesolid-state devices underlying moderndigital computing—with charge re-arrangement, something that worksnaturally at the molecular level. Onepossibility, in addition to the metal dots,would be a biological molecule involvedin photosynthesis, which forms a dipolewhen exposed to light. Such a moleculecould be used for coupling to QCA el-ements, Lent suggests. “To understand[what’s needed] we have to developgreater capacity to model molecules ona computer,” he says. Such computa-tional chemistry simulation makes uppart of the nanoelectronics effort atNotre Dame. Lent cautions that it willbe a long time, however, before QCAwill form the basis of real products.

Massive parallelism at thenanoscale

As the various nanoelectronics re-search groups pursue the develop-ment of their alternative switches, de-signs for molecular computing de-vices have been put forward. In part,the DARPA program, which aims tospur the development of a completelymolecular-based computer, is pushingthis effort

The Hewlett-Packard Laboratory-UCLA research group has drawn on an

earlier attempt at massively parallel, de-fect-tolerant computing that used sili-con chips to inspire its plans for a molecular-scale nanoprocessor, a proj-ect that has DARPA funding. The sili-con-based computer, called Teramac,consisted of 864 field-programmablegate array chips with 10 times the num-ber of configuration bits per gate thanin commercial chips. Using a largenumber of interconnections, Hewlett-Packard researchers were able to builda reconfigurable, refrigerator-sized ma-chine that used software to route ap-proximately 200 identified chip defects.

To get really massive parallelism, saysKuekes, you need to go to the chemists.Instead of fabricating hundreds of semi-conductor switches, chemists can makethe molecular switches by organic syn-thesis techniques that produce the mol-ecules by the mole—a trillion trillionmolecules at a time. “This would bemore switch molecules,” says Ellenbo-gen, “than all the transistors that havebeen made in the last 50 years.”

If you choose the right molecule, itselectrical properties can let it bothhold a bit of memory and serve as aswitch. However, says Kuekes, “thedefect tolerance is critical.” The chem-ical yields are less than 99%, so thecomputing architecture needs to bedefect-tolerant. Although a single de-fective transistor can make a semicon-ductor chip containing more than amillion transistors unusable, Kuekesexpects that self-assembled molecularswitch arrays can be trained to bypassdefects.

“In the course of the next year, we willbe working to produce a 100-nano-meter-square region of molecularswitches . . . that can form the equiva-lent of a 16-bit microprocessor chipfrom 1970,” says Kuekes. “This is thestate that silicon devices were in 30 yearsago, but on an incredibly smaller scale.”

Designing nanoelectroniccomputers

Other groups, especially Ellenbo-gen’s Nanosystems Group are attempt-ing to devise and model computer ar-chitectures based on nanoelectronics.The group has explored how you canuse molecular electronics to designlogic devices and an adder. However,the MITRE group aims to initially en-hance existing silicon technology byimplementing adjunct molecular de-vices on a solid-state chip. Such hybridtechnology, says Ellenbogen, wouldprime the pump for the spread of mol-ecular electronics technology.

“Eventually, smaller, simpler comput-ers [based on molecular electronics] willbe widespread,” says Ellenbogen. Healso predicts that pinpoint computerswith the power of a 1980s-era PC willcome into widespread use. “Nanoscaledevices will come to imbue all kinds ofstructures with computational capacity,”he says.

However, even with the foreseenadvantages for molecular-based elec-tronic devices, silicon chips won’t dis-appear from the marketplace in thenear future. The existing microelec-tronics technology offers a big advan-tage in that it can produce a 1,000- to10,000-fold signal gain. For this rea-son alone, Tour says, “I don’t think weare going to see CMOS go awayrapidly.”

David I. Lewin is contributing editor to Com-

puting in Science & Engineering, IEEE Intelligent

Systems, and IEEE Concurrency magazines. Pre-

viously, he served as contributing editor for

Computers in Physics. He writes on science,

medicine, and technology from Silver Spring,

Maryland. Contact him at dlewin@capaccess.org.