Volume 2, Number 1 " December 1988

The Research Laboratory of Electronics at the Massachusetts Institute of Technology

MATERIALS RESEARCH: Meeting the Challenge of Microelectronics Technology

Since the discovery of the transis- tor at Bell Laboratories in 1947, scien- tists have been challenged by the need to produce materials for highly minia- turized and increasingly fast micro- electronic components. Driven by the technology of the information age, sci- entists have examined the unique properties of semiconductors, funda- mentally exploited new combinations of these novel materials, and devel- oped innovative design techniques to miniaturize electronic components. At the heart of this research is the ever- shrinking and exceedingly complex in- tegrated circuit microchip. In addition, the development of microscaled tran- sistors and other electronic compo- nents has resulted in systems with larg- er and faster capabilities, particularly for information processing and high- speed communication. I Every year since 1960, the number of circuit components on the most ad- Professor Henry I. Smith e.p loins the development of an alignment ststemJbr x-ray vanced microchip has nearly doubled. nanolithography that should be capable of 100-angstrom precision. Under his direction. Today, over 4 million transistors can be RLE's Submicron Structures Laboratory has pioneered new technologies in submicron onto one packed neatly computer structuresfabrication and explored deep -submicron MOSPETs and the exciting new memory or DRAM random (dynamic field of quantum-effect electronics (see related article on page 7.). Working with Prqfes- access memory) chip. The dramatic sor Dimitri Antoniadis and graduate student Ghavam Shahidi, dx. have discovered that miniaturization of electronic compo- the deleterious hot- electron effects seen in silicon MOSFETs with channel lengths below nents over the last forty years has raised 0.25 microns actually decrease at linewidths below 0.15 microns. This is apparent/)) questions about the physical limits of because the source-drain transit timesare shorter than electron energy-relaxation time. the materials, devices, and systems in- The quantum-effect electronicsgroup (Professors Dimi/ri Antoniadis,Jesus del Alamo, volved. Such limitations in materials Clifton Ponstad, Marc Kastnei; Leslie Kolodziejski, Patrick Lee, Terry Orlando, and science are fundamental Henry Smith.) has demonstrated several novel quantum transportphenomena which imposed by can occur in sub-0.1-micrometerstructures. It is hoped that their basic research into such quantum phenomena may eventual!)' lead to more advanced computation and (continued on pg. 2) electronic systems.

MATERIALS RESEARCI I (continued)

Director's Message physical science and the composition, structure, and behavior of the materials used. In the 1960s, the smallest feature size of an electronic circuit was 30 mi- Electronic and optical materi crons. In contrast, today's scientists als form the basis upon which work with electronic struc- structures, devices, circuits, and typically tures one micron and below systems are built. They are the measuring (submicron). By the mid-1990s, scien- starting point for much of RLE's tists anticipate that the smallest features search in electronics and optics, will be approximately 0.1 micron. and increasingly they are engi- Materials research seeks to ad- neered products that are designed dress these limitations through differ- to meet stringent and innovative ent fields of requirements of the smallest and investigation: processing and fabrication of silicon and com- fastest systems. In earlier days, nat- fundamental ural elements like silicon were pound semiconductors; materials pro- doped in hulk to form the starting principles underlying cessing effects; the growth and material for electronic integrated characterization of and me- circuits. But now, much of RLE re- crystalline tallic thin films for microelectronic ap- search focuses on synthetic materi- plications; experiments with novel sili- als, using novel techniques of hand con heterostructure formation gap engineering to provide en- epitaxy; in compound semiconductors; and the hanced carrier mobility and desir- control of thin-film crystalline struc- able optical properties. tures on amorphous substrates. The achievement of high-qual- ity, defect-free electronic materials (continued on pg. 3) is the focus of several projects aimed at the controlled micro- structural evolution of thin films. In this way, device-quality material can be formed on a substrate, and Proj ssor Jonat/)an Allen, Director, grain boundaries in conducting SHORT CIRCUITS Researcb Electronics. wires can be oriented to drastically Laboratory of reduce the effects of electromigra- The staff of currenLc tion. One of RLE's major new em- would like to phases is the epitaxial growth of dope and deposit materials at the note the follow- novel materials and devices involv- submicron scale in the planar di- corrections to ing Ill-V and II-VI heterostructures, mensions, leading to new device ing the issue: using both molecular beam and configurations and techniques for June 1988 chemical beam techniques. These surface defect repair. RLE is also "Faculty Profile," page 11: Dr. Charles expanded experimental programs heavily involved in the study of su- V. "Chuck" Shank of AT&T Bell Labo- support the synthesis of materials perconducting materials, which ratories in Holmdel, NewJersev, was that involve direct control of have much sci- single recently generated mistakenly referred to as Chuck atomic layers as the material is entific excitement. Chang. grown, yielding unat- In all of these areas of experi- previously of at RLE," 16: In tainable structures and dramatical- mental materials, structures, and 'Ilistory Optics page the 1945 photograph depicting Al- ly improved devices. devices, RLE leads the develop- bert G. Hill andJerrold R. Zacharias While epitaxy builds up a ma- ment of new fabrication tech- examining a microwave waveguide, terial in one dimension, high-reso- niques, driven by the needs of Albert McCavour Clogston (left in lution lithography permits submi- electronic and optical systems. En- photograph) was misidentified as cron control of structure and gineered control at atomic dimen- George Briggs Collins. device planar dimensions, provid- sions provides the means to syn- 'UPDA'I'E: ing the world's smallest FET tran- thesize exciting performance that Communications," page sistor. Focused ion beam tech- is breathing new life into high- 20: Author Donald George Baltus niques also furnish the ability to speed miniature systems. was misnamed [)avid. Thanks to our eagle-eyed readers who called our attention to these errors.

Principal Research Scientist Dr. John Alelngailis uses focused ion beamsfor patterned depositionfrom ad- sorbed gas molecules andfor patterned implanta- tion or lithography. The unique capabilities ffo- casedion beam implantation have been usedatMIT tofabricate tunable (iunn diodes and to expose re- sistfeatures down to 0.05 micron linewidths. With the apparatus shown in thephotograph, goldfilms have been deposited in linewidths down to 0.1 mi- cron. The focused ion beam column is mounted on an ultrahigh vacuum chamber, and will be usedfor microdeposition ofreactive metals such as tungsten or aluminum. It will also he used to develop tech- Professor G4fton C. Fonstad examines a heterostructure specimen grown by niquesfor x-ray lithography mask repair andfor in- molecular beam epitaxy for use in lacer diode fabrication. He has successfully tegrated circuit restructuring and repair. In addi- grown state-of-the-art multiple quantum well heterostructures in two materi- tion to deposition, both of these applications require als systems; gallium-aluminum-arsenide/gallium arsenide and indium-gal- material removal, or micromilling, with submicro- lium-aluminum-arsentde/indium phosphide. In collaboration with institute meter resolution. Grooves 0. / micron wide and 0.2 Professor Jlerinann A. Haus, he has incorporated heterostructures in wave- micron deep have been milled with ivell-defined, guide devicesfor high-speed, guided wave optical circuits. Professor Fonstad steep sidewalls. The deposition andremoval ofhigh has also demonstrated the growth of strained-layer heterostructures on (iii) atomic number materials; such as or arsenide, and the in gold tungsten gallium was first to observe the internal electricfields these with submicrometer resolution andhigh aspect ra- layers. The growth ofstrained layers on (111)-oriented substrates willfacilitate tio, is essential to x-ray lithography mask repair, highly efficient optical modulators for integrated optical circuits; and will en- while good electrical conductivity is requiredfor cir- hance laser diode performance. cuit repair.

An integrated circuit microchip contains the integrated-circuit pattern tronic properties than the bulk silicon. primarily consists of transistors, resis- exposed by electromagnetic radiation Thus, a semiconductor's electrical tors, capacitors, and various intercon- or charged particles. Selected areas on properties can be changed and closely nections finely mapped out and fabri- the photoresist are etched away, leav- controlled using doping techniques. cated on a single semiconductor ing a circuit pattern on the remaining Because silicon's few valence electrons crystal. Microchips are made by slicing photo res ist. are entirely consumed by covalent a cylinder of single-crystal silicon into Next, the silicon is provided with bonds in the element's lattice, the addi- smooth, ultrathin wafers. The wafers charge carriers by doping, a technique tion of a dopant with more valence are heated in the presence of oxygen, used to inject the silicon with impurity electrons will result in extra valence which results in the growth of a layer of elements (charged ions or dopant electrons that are free to conduct elec- silicon dioxide on the wafers' surface. atoms). Doping transforms specific tric current. Conversely, when using a The surface is then covered with a poly- areas underneath a wafer's surface so mer coating called photoresist, that that the areas will have different elec- (continued on pg. 4)

MATERIALS RESEARCH (continued)

dopant with fewer valence electrons than the semiconductor, positively charged carriers (or "holes") occur, which can move when an electric field is applied to the material. Common dopants are arsenic (with an extra va- lence electron than silicon) and boron (with one less valence electron than silicon). Ion bombardment from doping can damage the silicon crystal, so it is slowly annealed, or heated, to remove defects. Simultaneously, another layer of silicon dioxide is deposited to pro- vide an insulating layer and to protect the wafer from high electric fields. The wafers are further processed by depos- iting metals or alloys by a variety of methods: vacuum evaporation, sput- tering, or chemical-vapor deposition. The wafers are sliced into many chips, Doe alec! of PvoJesso- John M. Graybea/'s is cjuantooi lrcoispoil in loi.v-chmen- each as many as 4 million research containing sional disordered systems. Examples of/ow-dimensional systems include two-dimen- transistors and then mounted per chip, sional electron gas trapped on the inter/bce in silicon,11OSEETs, and quasi-one-dtmen- in a ceramic or plastic package. (See sional systems in MOSFETs with narrow gates (afei.c hundred angstroms wide.). Studies diagram on page 5, "How a computer of devices in low-dimensional systems have led to the discovery of inportantfiuctu- chip is made".) ations in resistance in both metallic and insulating regimes. Professor Grayheal contin- ues to small device in the Chips with microscopic bulk and explore transport phenomena quantum regime by observing resonant and the basic states in narrow, dis- surface defects will not operate proper- tunnelling examining features of edge ordered systems. This willfurther understanding of the behavior of individual quantum ly. The larger the chip, the more likely states in microelectronic devices. Here, he is with it will contain defects or deviations photographed thin-film sputtering equipment used in his research on the synthesis of high-temperature superconducting from the exact periodicity of the crystal films. array (dislocations, stacking faults, and grain boundaries). Because good chip yield is inversely related to the chip's size, a chip usually measures less than one square centimeter in size. Scien- tronic elements. This is known as migration problems. tists are currently exploring new meth- electromigration, or current-induced ods to grow larger, defect-free silicon diffusion. The direct cause of electro- crystals and to enable the creation of migration is the increased current den- increasingly dense integrated circuit sity experienced as integrated circuits The active devices commonly used elements on a single-crystal chip. become smaller, Although the total cur- in today's electronic systems are either Since external wiring increases the rent in a miniaturized wire might be bipolar or field-effect transistors cost of a chip considerably, another low, its current density (measured in (FET5). Although bipolar transistors are challenge is to pack as many intercon- terms of the amount of current per unit fundamental to large computer CPUs nects on a single chip as possible. But, of cross-sectional area) can be large. (central processing units) and are fast- problems arise when attempting to mi- Once the high current density in the in- er performers than FETs, the FETs form croscale electronic devices. For exam- terconnect's depleted region causes the basis for the computer's memory ple, the smaller the transistor, the thin- movement of aluminum atoms to the and are also used in logic operations ner its silicon dioxide insulator low current-density regions, the phe- for small to mid-sized computers. becomes. As a result, electrons on a nomenon will result in interconnect The most, common element in transistor's gate can leak through or failure. Connector material actually both bipolar transistors and EEl's is "tunnel" into the silicon substrate be- gets pulled along the wire, and can semiconducting silicon (Si). SemicOn- low, thus affecting the transistor's leave voids in the interconnect. Cop- doctors are solid crystalline materials performance. per-aluminum alloys have been used in whose electrical conductivity lies be- In addition, as a result of shrinking interconnect material to slow the tween that of a conductor and an insu- the size of interconnections, unwanted electromigration process. But, with the lator. They are the basis, or substrate, motion of component material can oc- increasing density of circuit elements, on and within which a microcircuit is cur in the aluminum wires, or intercon- other novel materials and techniques fabricated. Silicon possesses many nects, that join the chip's various dee- are being investigated to solve eleetro- beneficial properties that make it ap-

propriate for use in integrated circuits. As an abundant, naturally occurring element, it can he formed into near- perfect crystals inexpensively. Silicon's band gap (1.12 electron volts), or the energy difference between its valence and conduction electrons, enables it to maintain semiconducting properties over a wide range of temperatures. When interacting with silicon or other semiconductor materials, elec- trons exhibit unusual behavior. In semiconductor materials, electrons travel as if their mass was much smaller than electrons travelling in free space. L For example, in silicon, an electron HowOW a travels as if its mass was one-fifth the ef- computer fective mass of a free electron. Beyond chip is made silicon, researchers have already begun Building one layer on top of another to investigate other suitable scm icon- ducting materials. Gallium arsenide (GaAs) is a frequent alternative to sili- A chipch~ starts- out asa. a- single' segment out of ozid hurrhundredsredsir.on . round, .1-8RICIncInch sllcon,I'. wafer, con, since its carrier mobility enables 0which is coated with a layer of silicon dioxide, an insulating material on which minute electrical faster circuit switching times (about silicon circuits will be formed. two-and-a-half times faster than sili- con's). One drawback to gallium arsen- A thin film of plastic material called ide is its lower thermal conductivity. I9 photoresist, which is sensitive to light (or X-rays), is coated on the oxide layer while When fabricating smaller devices, the the wafer is spun rapidly to spread it evenly. device's switching speed is limited by oxide the substrate's ability to conduct heat silicon away from the device.

Gallium arsenide is a compound beam of visible light, ultraviolet light or semiconductor made of elements X-raysrays is projected onto the coated wafer up Wthrough a photomssk, a thin sheet similar to a from the table's III and photographic negative, on which the image of periodic group the desired circuitry has been drawn. The group V. The periodic table highlights photoresist, like a photographic print, ac- quires the image of the complex circuit. the similar properties of various chemi- cal elements, which are arranged in or- der of atomic number or in weight hardened photoeestst horizontal rows (periods) and vertical exposed phoioresist is developed in 0Thechemical baths and baked, which softens columns (groups). Circuits based on the areas bearing the outline of the circuit. semiconductor materials from periodic A table columns Ill and V are capable of higher clock rates because electrons Acids and solvents strip away the exposed photoresist and the Insulating silicon dI move faster in Ill-V compounds than in 0 de below' It, leaving the circuit pattern em- bedded in the remaining photaresist and the silicon, a group IV element. Other silicon base below. high-speed Ill-V materials include: in- dium phosphide, gallium-aluminum-

arsenide, and indium-gallium- The remainder of the photoresist Is clean- ed sway, leaving the circuit image in the arsenide-phosphide. insulating oxide and silicon base. In comple.- Once the beneficial of chips, successive layers of Impurities such as properties boron and arsenic are added selectively to con- materials and the advantages of micro- struct transistors in the circuitry. The process of masking. layering chemicals and stripping scaling are combined, faster devices is repeated many times. can be successfully fabricated, The MESFET pin gridchemical Impurities, called dopants, are add- (metal-semiconductor FET), 115Y , 1 ed to form negative and positive conducting created by substituting gallium arse- - II' I areas. Metal and doped polysilicon are added I to form pathways among individual transistors nide for silicon, is faster than a MOS- and other components. Lasers then slice the wafer Into individual which individual- FET. The MODFET chips, are (modulation-doped ly mounted on plastic casings such as this pin N FET) is a device that is ,I grid array. The chip Is Inserted Into a printed quantum-well circuit board along with other electronic parts. even faster than a MESFET. A layer of So'r InLet Olose nail 0. sa5alas aluminum-gallium-arsenide is deposit- Corp. dlrgrar, / ed on an undoped gallium arsenide

Reprinted courtesy of The Bostott Globe

(continued on pg. 6)

MATERIALS RE, SEARCI I (continued)

substrate to form a single-crystal lattice known as a superlattice. The physical properties of these layers provide a "quantum well" to trap electrons from the aluminum-gallium-arsenide. The fast electron conduction in a MODFET is attributed to the absence of dopants in the substrate, which increases its mean free path. Mean free path is the distance an electron or hole can travel before scattering from an atom. The longer the path, the higher the mobil- ity. In future technologies, it may be possible to construct both vertical and horizontal quantum wells so that elec- trons can travel between cross- sectional "squares" to perform digital operations. By combining elements from groups III and V, custom-made elec- tronic materials can be produced for various microelectronic devices. One approach to building these devices is heternepitaxy. By layering thin films of different materials on a sub- crystalline Professor (7w-I V Thompson checks a zone melter that was constructed in his laboratory. strate, can a heteroepitaxy provide The apparatus has enabled the correlation of solidification conditions, interface mor- a multilayer superlattice, periodic array phologies, and crystalline defect structures at the liquid-solid intetface during zone of alternate made of two layers melting recrystallization of thin films for the first time. Professor Thompson c research semiconductors. The techniques used involves inicrostructural evolution during polycrystalline and epitaxial thin-film pro- to grow heteroepitaxial films include: cessing, control and modtjication ofstructures through post-deposition processing, and the determination microstructural on and other liquid-phase epitaxy, where cooling of of effects reliability film properties. He has a model that a heated solution of desired elements produced general permits quantitativepredictions ofgrain growth rate and final grain sizes in polycrystalline silicon films. Detailed correlations have been occurs on a substrate; chemical vapor made between grain sizes, grain size distributions, and grain orientation distributions deposition, which exposes the sub- with statistics from electromigration-induced failures of alum inum-based intercon- strate to heated gaseous elements or nects. Currently, he is working on techniques to make eery large-grain films that will he and molecular beam compounds; epi- resistant to electromigration. taxy, which targets heated molecule beams or atoms at a substrate in an ul- trahigh vacuum. Researchers are not restricted to naturally occurring Ill-V compounds. from heated effusion cells, or cruci- fine Although a material's structure and detail that provides a close look at bles. These beams are aimed at a these materials. A electron properties determine its performance high-voltage substrate on a level and behavior, structural configu- single-crystal tempera- microscope provides sub-100-ang- ture-controlled substrate holder in an strom resolution, while the rations can he changed through closely scanning vacuum chamber. The result- controlled processing methods. Crystal ultrahigh electron microscope affords a three- is oriented or controlled growing techniques, such as molecular ing epitaxy dimensional view at close to 100-ang- with each successive beam epitaxy (MBE), can combine the growth, layer strom resolution. Materials scientists the lattice orientation of also use to examine the desired properties of many chosen ele- resembling spectroscopy the beneath it. Finished films are ments. MBE techniques have been re- layer photon interaction of a specific wave- flat to within one atom in ferred to as "spray painting with atoms" uniformly length (x-ray, ultraviolet, visible, or in- by Bell Laboratories, where it was in- depth. frared) with a material's electrons, and the from that vented. Molecular beam epitaxy can resulting spectrum build custom-made semiconductor interaction. Since crystals by depositing different Several analytical techniques are light's long wavelength is not suitable for fine semiconductor compounds in evenly used to characterize microstructures etching extremely features onto a silicon wafer's deposited and alternating atomic lay- and their properties (such as lattice photo- ers, or films. Beams of a chosen materi- periodicity and other crystallographic al's atoms or molecules are emitted information), and to achieve extremely (continued on pg. 8)

RLE'S SUBMICRON STRUCTURES LABORATORY: The Size of Things to Come

The Submicron Structures Laboratory (SSL) at RLE was estab- Cell lished in 1978 to develop advanced C,,c,,l Ca,le, '_ techniques for fabricating submi- cron structures, and to pursue nov- el research applications of such structures, from micro- SAW Dem ranging c.". 1573 -- Synapse electronics to - en) cortex x-ray astronomy. Meoy ccli. 985 Since its inception, Professor Hen- ry I. Smith has directed SSL's pio- DDRAM Fb,non 1,0, neering efforts in micro- and nano-

fabrication, deep-submicron and n- F EMT \.

devices, (c ms ,.thali quantum-effect crystalline I T) thin films on amorphous sub- IL strates, and periodic structures for Figure 1 Figure 2 x-ray optics and spectroscopy. C In its research on x-ray and

holographic nanolithography, the 100XF.eoot-.f X,.y hill SSL has been the leading laborato- -7 ALE 3.41 ry in the world, and has exploited }..oeoi Holes in film its to achieve LI" [1 AtE3 DNA unique technologies it l,,mw,) mecute significant firsts in microelectron- ics and quantum-effect electronics. 11 - Silicon MOSFETs with channels Lc,,1h,,el,Ol Cell membrane Multloyer film patch shorter than 100 nanometers have Boo,, proemso,) by MBE been fabricated. The velocity over- shoot and reduced hot-electron ef- L__Ii s,A fects observed in these devices are ,o - path 120 leading to new concepts for minia- m"

turization of electronics. Studies of II .So,,th Ml C) ix .1.1 quantum mechanical transport in Crystal MOSFETs with channels as narrow Figure 3 Figure 4 as 30 nanometers, and in gallium arsenide MODFETs with fine- 1 the human hair in cross-section with a period grating gates (linewidths Figure compares swface-acoustic-wave device, a cell, a one-micron scale marker, a bacterium, anda smaller than 100 nanometers), are one-megabit memory neuron from the cerebral cortex. uncovering unanticipated new the human hair quantum phenomena which, in the Figure 2 is a ten-time magnification ofFigure 1, which again shows in cross-section, a bacterium, and a one-micron scale marker. A human chromo- future, may form the basis for ad- some and a (the interconnection between neurons) are shown next to vanced electronic and computa- synapse MIT's work on gold diffraction gratings for x-ray astronomy and a MOSI"FT device tional systems. used to stud quantum effects in electron transport. In space provided by the Mi- 1. The dimensions a transistor crosystems Technology Research Figure is a 100 times magnification of Figure of made at MIT smallest MOSFET in the world) are shown with the Submi- Laboratory (MIT Building 39), SSL (the along cron Structures Lab's work on reactive ion etching and x-ray nanolithographv. The encompasses 1,000 square feet of cell mein brane "elemental switch" in and the cross- class 10 clean rooms, and another patch (the biological systems, section ofa "quantum wire" in GaA lAs/GaAs, only 350 angstroms wide, are Shown 1,000 square feet of class 10,000 to scale. The latter used to stud)' phenomena in a one-dimensional clean is quantum space. electron conductor.

Figure 4 magnifies Figure 1 one thousand times. The holes in AIF3 were made 4), These pictures illustrate the ex- Mike Isaacson at cornell using an extremelyfine 5-angstrom electron beam. The small size of electronic de- tremely inuftilaj'erfilm shows the control of layer thickness achievable with molecular beam vices and other structures achieved epitaxt'. The cell membrane patch and the DIVA molecule are drawn to scale. The in the Submicron Structures Labo- dislocation relates to work done at MIT on controlling the location of dislocation ratory at MIT (and elsewhere). in silicon, perhaps leading to somefuture applications.

? IA'l'ERIAI S RESEARCH (continued)

resist, new techniques are being devel- oped. One alternative is high-energy electron beams, because their wave- length is smaller than an atom's diame- ter. Another technique used to etch small structures on semi- ex-tremelyis A conducting chips x-ray lithography. number of sources of soft x-rays can be used, including electron bombard- ment, plasma sources, and synchro- Irons. Synchrotrons produce very- short-wavelength x-rays by accelerating charged particles. During acceleration, the particles emit large amounts of electromagnetic radiation, and the re- sulting radiation can be "tuned" by reg- ulating particle acceleration. High-en- erg), synchrotron x-ray radiation is also an alternative to conventional radiation sources in spectroscopy since its high- er radiation intensity is more parallel 1'ro/'.csor,Jesiis del Alamo (center) and,tlIT students Sandeep Bald (left) and We/id collimated), and extremely (or precise Azzam examine device performance using a semiconductorparameter analyze,-. Their measurements can be obtained. Synch- e.\pertments measure the characteristics ofhete,-ostructure FETs inadefro,n indium-gal- rotron x-ravs can also produce high- lium-arsenide/indiu,n-alumjnum-arsenjde semiconductor materials. Professor del Alamo's resolution x-ray diffraction maps that research involves high-p erfor.inance semiconductor devices for microwave and telecommunications. detail important information on a mate- optical rial's properties and its behavior.

The development of new materials and novel processing technologies rely on a deep theoretical understanding of materials and the development of more powerful analytical tools to studs' them. Although silicon remains the most in- dustriallv important semiconducting material, researchers are considering other options such as cryogenic super- conductive integrated circuits, thin films of semiconductors on insulator materials, and substrates created by dif- ferent materials through molecular- beam epitaxv. In developing new microelec- tronic devices, researchers are explor- ing the benefits of a three-dimensional chip, where the circuit components are stacked in layers instead of solely on the chip's surface. Because most elec- tron transport occurs at the silicon wa- fer's surface, scientists are attempting to build circuits in layers with conven- tional such as deposition techniques, l'ro/i'ssor Leslie A. Kolodziejski's re,ceu,-c/, isfocu.sed on t/,efrthrication ojsemiconcluctor vacuum evaporation or chemical-vapor lasers based on li-VI compounds, which contributes to other collaborations in RLE in- deposition, so that the device will ex- volving quantum-effect devices and specialpwpose soild-state optical lasers. She uses tend above and below the chip's actual chemical heatn epita.vy tofˆthricate and characterize II-VI and ill- V heterojunctions, the surface. boundaries between two dUjerent semiconductor materials. This a'orkfiinhe,c understanding of the epitaxialproce,cses involved in using coherent and incoherent light.

FACULTY PROFILE

In some cases, they weren't terribly ele- FACULTY nint, but they worked. Later, we got involved in making a pc of SAW device called the reflec- PROFILE: Lye-array-compressor. This had cx- Lremely difficult specifications. It had to I. \y )1k at frequencies above one giga- Henry Smith hcrtz, and that meant electrodes with features below one micron. The elec- From 1.968-1980, I. Smith Henry t cudes had to be parallel to a very tight worked on surface-acoustic-wave tolerance. In addition to I/O elec- (SAW) devices and the devel- pioneered trodes, we had to make a grating in a opment of techniques forfabricating lithium niobate substrate, where the subnzicron structures at MIT's Lincoln grating's periodicity varied in a precise Laboratory. In 1,977, he began thepro- way from one end of the device to the cess the Submicron of establishing other. It was a complicated device, and Structures at M17 and Laboratory way beyond what had been done be- joined the facultyfull-time in 1980- At fore. So, Dick set out esso) llenr)~ 1. Smith Williamson and! his research includes Proj present, nanofab- to solve the problems, one at a time. rication, -submicron tlOSFF7:c, deep Eventually, we had great success. electronics in sub-100- which was a in quantum-effect group, starting program In the process of doing those pro- nanometerstructures, on surface-wave devices for radar crystalfilms signal jects, I developed new techniques such and amorphous non-lattice-matching processing. as conformable photomask lithog- substrates, and elementsfor diffractive raphy, techniques for ion etching, and and " Can describe the transition x-ray optics spectroscopy. Professor you some analytical techniques. We also de- Smith and his are from your research in SAW devices colleagues responsi- veloped an in-house electron-beam several innovations and in- to sub- blefor key developing techniques for lithography capability at Lincoln. This ventions in submicron structure tech- micron structure fabrication? enabled me to do exploratory work in nology and applications: conformable In all devices, the pacing element is c-beam lithography. Some of this was photomask lithography, x-ray lithog- fabrication. It's easy to go to the black- done with a graduate student, Richard raphy, SAW devices, in- reflective-array board and dream up a new device Ilawryluk (Ph.D. '74), who investigated teiferometric alignment, graphoepi- you'd like to make, or a certain perfor- electron backscattering effects. During tcy, stallization, zone-inciting recrj mance level you'd like to achieve. That this time, I got the idea for x-ray lithog- subboundary entrainment, surface-en- is never the pacing element. What de- raphy. It was difficult to juggle all these sub-100- ergy-driven grain growth, termines whether or not you can do it projects simultaneously, so Dave nanometer silicon MOSFETs, and sur- is if you can fabricate it. I liked to solve Spears, who was in another group at MO1)FETs in face uperlattice gallium problems and get things done, so I Lincoln, came over to help with x-ray arsenide. gravitated to where the problem was, lithography development. That came to and that was in microfabrication. fruition in 1971. After x-ray lithography In Ernie Stern's group, I devel- and the reflective-array-compressor de- *How did become interested in you oped fabrication techniques specifical- vice became successful, I was able to surface-acoustic-wave (SAW) de- ly for surface-wave devices. This in- expand the impact of microfabrication vices? volved making electrodes with very techniques. I started working on acoustic-wave de- fine dimensions, and doing it reliably In Ernie Stern's group, we didn't vices when I was in the Air Force in the on a variety of substrates. It had to he have a charter to do research on micro- early '60s, These devices used bulk clone reliably because, in the early fabrication, or SAW propagation for waves, which didn't intersect the sur- days, we may have had only one piece that matter. We were committed to face except when you launched or de- of some exotic, expensive material that making signal processing devices for tected them. During the time I taught at was about IA-inch wide by 3/4-inch long. radars in the field. So, we didn't have Boston College (1966-68), Dick White That was it! We had to do whatever pro- the luxury of time to investigate things at Berkeley invented the surface-wave cessing was necessary to make a SAW thoroughly, You had the feeling that transducer. That changed the whole device with the material available. It you were hanging on by your finger- field because acoustic devices could had to work the first time. In addition, it nails; that tomorrow, things wouldn't then be made in which the wave propa- had to have features beyond state-of- work because you didn't understand gated entirely on the surface. When I the-art semiconductor device research, them thoroughly. I always had an in- came to Lincoln Laboratory in 1968, I and beyond what industrial equipment tense desire to investigate questions was in the microelectronics group, but was able to achieve. So, I developed more deeply. One way to do that was to I interacted with Ernie Stern on SAW fabrication techniques specifically suit- have graduate students. It was unortho- devices. Later that year, I joined Stern's ed to high-performance SAW devices. dox at the time, but I always had at least

F A C U L T Y P R 0 F 11, E

one student working with me. At that me to provide an on-campus site for apart. You've got to decide which one time, their salaries were low, so it the laboratory, but we couldn't pin you're going to jump onto!" They gave didn't make much of a dent in the bud- down a specific location. Paul Gray as- me the option of being full-time either get, and they didn't detract from the sured the NSF that he'd find appropri- at Lincoln or on campus, but not part- mainline program. ate space on campus. But, this wasn't time at both places. I decided to come enough. Cornell promised to construct on campus because that's where my in- " flow did theSubmicron Structures a new building for the national lab, and terests really were and because the lab- it was awarded to them. Laboratory get started? oratory at Lincoln was staffed with fully It was probably the best thing that competent people who didn't need me The major event occurred about 1975. could have happened because it shook anymore to do their thing. The campus Jay Harris, from the National Science up man)' people at MIT. After the NSF environment gives one much greater Foundation (NSF), came by to talk announced the center was going to freedom. Basically, you can do what- about how several areas in engineering Cornell, Paul Gray wrote me to say that ever you can raise the money to do. and science (electronics, electrooptics, we should do this anyway. Shortly after But, more importantly, it's rewarding to and what was then called integrated op- that, the director's office at Lincoln work with graduate students, and the tics) were being inhibited because of asked me to set up a laboratory dedi- discipline of teaching has an enormous difficulties in fabricating submicron cated to submicron structures there, impact on how you approach science structures. He talked about establishing with money that they'd raise internally. and research. a national laboratory dedicated to sub- Paul Penfield (then Associate Head of micron structures techniques, so that the Electrical Engineering and Comput- "Did submicron structures people without facilities could go there your er Science Department) and Peter work stem research to make the devices they wanted. That from previous Wolff (then Director of ELE) asked me atRLE? was the original concept. I was skepti- if I would consider coming to campus cal because it wasn't just a matter of It was new, but there's some as an and set a lab- interesting the adjunct professor up having right facilities, it was having history. The very first work in electron oratory. Their offer was intriguing. I al- the know-how. Knowledge is much beam lithography was done by Dudley ways enjoyed working with graduate harder to generate than equipment or Buck at RLE in the late 'SOs. He died students, and had accomplished a lot central facilities. As a result of jay's ini- prematurely, and his research also died through them. I felt that we'd be doing tiative, the NSF held workshops across at MIT. Chuck Crawford, an electrical something new by conducting graduate the country to evaluate the desirability continued work research at the forefront of modern engineering professor, of establishing a national laboratory. in charged particle beams and electron fabrication technology, and that we Because of those workshops, the lab's optics. But, Chuck didn't do microfabri- could pursue novel applications of sub_ theme was shifted to cation, and we never interacted at Lin- proposed empha- micron structures, some of which had size research on submicron structures coln. The submicron work actually come to mind as a result of writing the as well as the resource concept. In started at Lincoln in 1968, and was mo- proposal. The idea of setting up a labo- 1976, the NSF circulated a request for tivated by the need for SAW devices. Al- ratory not only on campus, but also at proposals to establish a national re- though it was recognized that the tech- Lincoln, specifically devoted to re- search and resource for submi- would be facility search that we could never do before nology someday important cron structures. I decided to submit a for electronic devices, that was not the was attractive. So, I set about to do for a center at Lincoln Labora- function. proposal both. driving At the time, I didn't have tory. any sig- With from we nificant connections with the MIT cam- help many people, established Lincoln's Submicron Tech- *In a recent Boston Globe article, pus. I received encouragement and were described as "the nology Program. It was operational by you father" help from Al McWhorter, who was head What has been late 1977. And, we did the same thing of x-ray lithography. of the Solid-State Division at Lincoln; role in the devel- on campus.John Melngailis, from Lin- your technology's from people in the director's office at and what the use coln's SAW device group, joined me opment suggested Lincoln; and from Paul Gray who was and came to four a week, of x-rays in photolithography? then MIT's Chancellor. campus days while a continuing at Lincoln one day I invented a technique called conform- Proposal writing can he a highly week. the of able where, us- creative That's when I started During early days setting photomask lithography activity. the I was on one up laboratory, campus ing ultraviolet light and evanescent talking with people on campus. I sub- day a week, and four days a week at Lin- coupling to a material, mitted the proposal, and a site visit fol- photosensitive coln. That worked well, and we got we could resolution what lowed in 1977. Later, I found out get beyond April both laboratories and going, producing anybody thought possible. We used that our proposal was technically the exciting results. that technique to make SAW devices best. The problem was that we pro- In 1980, as both programs were with submicron lines, extreme posed locating the laboratory at Lin- growing quickly, Don Maclellan in the linewidth control, and what we refer to coln. Although no one said it in so director's office at Lincoln said to me, as enormous latitude. I many words, there was no way that the today process "You're standing on two chairs, and used to brag that we had a 300% expo- NSF was going to establish a national they're sliding further and further sure latitude at 0.4 micron linewidths, laboratory at Lincoln. The NSF urged

FAG U L '1' Y P R 0 F I L E

certainly looked feasible. Dave Spears, who was in a different group at Lincoln, was willing to work on the project. He devoted full-time to the issues involved with the source and the mask. By that time, we knew how to pattern the ab- sorber with an c-beam, but the technol- ogy' had to he developed to make an x- ray lithography mask. The obvious choices for an absorber were high-den- sity materials with high atomic num- bers. Gold seemed like the ideal mate- rial, and it still is-gold or tungsten. When Dave came on hoard, we soon agreed that aluminum was a good source (with an x-ray line at 8.34 ang- stroms), and we used silicon mem- branes for the mask material. Although there have been different approaches to x-ray lithography over the years, IBM's multihundred million dollar pro- gram at East Fishkill, New York, still uses patterned gold on silicon mem- branes as masks (which we patented), and wavelengths between R and 10 ang- From left, graduate students William GhuandAnthony Yen, and Profi?ssorSnhitb examine stroms. a mask usedfor x-ray nanolithography. Prqfessor Smith holds a silicon wqfer (lower

right). " Are there advantages in moving from optical to x-ray lithography?

Yes-process latitude, yield, and which was an enormous advantage. An orders of magnitude higher than in dif- linewidth control. In setting up a scan- individual SAW substrate might cost fraction analysis or medical x-rays. ning electron-beam lithography sys- $1,000, so we couldn't experiment with (There is a semantic problem with tem, I found that we could make what- process parameters; we had to make it x-rays. The same terminology is used ever patterns we wanted with the work the first time. High yield and pro- for wavelengths ranging from 0.1 to electron beam, no matter how fine the cess latitude were the driving func- 100 angstroms. But, the absorption of linewidth. But, there were disadvan- tions. x-ray's goes as the third power of the tages with the electron beam related to One day, I was visited by lain Ma- wavelength. So, the behavior between electron scattering and a limited pro- son from University College in London. what we call "hard" x-rays and "soft" x- cess latitude. Also, the process was ex- We discussed conformable photomask rays can he as different as between mi- tremely slow. I found that it was better lithography, and how far I could push crowaves and ultraviolet, for example.) to use the electron beam to make a it. After he left, I got the idea to use an I calculated the optimal wave- photomask, and then replicate it. If you x-ray regime phenomenon known as length for x-ray lithography at about 10 wanted finer features, it was appropri- absorption edges. That is, I thought of angstroms. But, good sources for .10- ate to make a mask with the electron using a wavelength such that a material angstrom x-rays did not exist. In fact, beam and then replicate it. with x-ray-s. whose absorption edge was at a shorter very little work had been done at that One advantage of x-rays is that we wavelength could be used as a transmit- wavelength. I figured out what could can combine many different patterns ter membrane, and a material whose possibly be used as a mask and what on one mask, and then expose it all at edge was at a longer wavelength could could be used as an absorber. In the once, We can make an x-ray mask that be used as an absorber. At the time, x- first experiment, I used low-energy combines techniques of electron-beam rays were mainly used in medicine and bremsstrahlung x-ray's from a copper lithography, photolithography, and ho- for diffraction analysis of crystal struc- target. I experimented with the resist lographic lithography. It's similar to tures. Both use penetrating, short wave- we used for electron-beam lithography combining text from a word processor length x-rays. So, I sat down and quick- and found it was also sensitive to x-rays. and photographs from a camera, ar- ly calculated that the x-ray wavelength I even got a chemist to synthesize a new ranging them on paper, and producing one would need for lithography was resist for me that contained mercury, one unified page on a photocopy ma- very different (approximately 10 ang- which also worked. chine. The electron beam is like a word stroms). It was a much softer x-ray, be- I didn't have much time to devote processor, and the x-ray is like a print- cause the absorptivity would have to be to the project, but x-ray lithograph)' ing press or photocopy machine.

F A C U L 'I' Y P R 0 F I L E

.Isx-ray lithography practical to same as that of an x-ray mask (about 2 mass produce memory chips? microns). The wing's membrane must be able to fold when not flying, and go We can get fine linewidths with other through metamorphosis, rainstorms, techniques such as c-beam and even and predator pursuit. So it's pretty optics. But, x-ray lithography will ulti- strong. The intuition that membranes niately he cheaper because one can get are terrible to work with is understand- fine linewidths together with extreme- able, but, there is no problem in devel- ly precise Iinewidth control and broad oping an engineering technology so we process latitude. It also gives a much can handle these membranes. higher yield because certain defects don't print. For example, in the early " a current days of our research, we observed that Are you excited about dust particles were largely transparent project? to x-ravs, so they didn't print. Our first The most exciting things we're doing published paper on x-ray lithography these days are in the areas of quantum- shows a mask with many defects that effect electronics and very-short-chan- didn't print. nel MOSFET devices, fit the latter, Scanning electron micrograph the re- IBM is pursuing x-ray lithography of we've led the way by showing that ordi- suits of x-ray nanolithography done at because it will ultimately he much MOS transistors can be made with MIT The micrograph shows 300-ang- nary use a synchrotron, so it channel cheaper. They strom (30-nanometer,)-wide lines ofpoiy- lengths below 1,000 ang- tends to be an installation. expensive methyl methacrylate (PAL1IA,) exposed stroms. We've demonstrated that the ef- When completed, the IBM East Fishkill, and developed on a silicon substrate. Ex- fective electron velocity can be much New York installation will cost several posure was done by graduate studentAn- higher than the bulk saturation velocity IBM and oth- hundred million dollars. thorn' Yen, using the carbon K x-ray at (we call this velocity overshoot). We ers see x-ray lithograph)' as a way to 4.5 nanometers. also found that short-channel devices keel) up with foreign competition. In have reduced hot-electron effects, there are four or five Japan, major pro- x-ray from deep-ultraviolet lithography which could have an important impact in twelve grams x-ray lithography, as an approach to commercial integrat- on how small transistors will be scaled. beam lines at Tsukuha, and several ed circuit production isn't linewidth- We are interested in making very small companies experimenting internally it's process latitude. devices and studying theirphysics. We with synchrotron radiation for x-ray One thing that makes people shy aren't simply trying to break a switch- In lithography. this country, there's away from x-ray lithography, particular- ing speed record. That might be inter- one beam line for industrial 0111)' ly managers or planners, is the fact that esting to do, but it's not our focus. work-the IBM at Brookhaven, facility the mask is a pattern of an absorber on (When I say "we," I mean the graduate New York. in addition to Today, syn- a thin membrane. They're scared that if students who've done the work, Profes- chrotron-based x-ray lithography, one they poke the membrane, it's going to sor Dimitri Antoniadis, and myself.) I of the hottest is the topics laser-pro- break. Well, it will break if you poke it! see our role as exploring the impact of duced plasma work done at Hampshire But, for their thickness, membranes are deep-submicron and nanometer tech- Instruments. I consult there, and feel extremely strong. If you were to ana- nology on transistor devices. So, we confident that the laser-produced plas- 1)-xe what makes a material strong, and have made devices, demonstrated ve- mas will re- enable manufacturers to look at the radius of curvature that a locity overshoot in them, determined their with an x- place optical steppers material can sustain, you'd find that sonic of the problems, and confirmed but ray stepper at about the same price, things get stronger as they get thinner. that there is a reduction in hot-electron with much yield. higher For example, if you bend glass, it effects. But, we can't do everything. breaks. But, if glass is formed into a Now, HIM has picked up this thrust and very fine fiber, you can wrap it up like is making very short-channel devices, " Can ultraviolet light be used in fishing line. As a material gets thinner, confirming our velocity overshoot photolithography? it can sustain a smaller radius of curva- measurements, and getting a lot of Most people are betting on ultraviolet ture; it can bend without breaking. The mileage out of making circuits that way. but, in it's wishful light, my opinion, best example is fiberglass, and the Another area that we're currently thinking. The finest optical projection same is true for a membrane. researching will almost certainly have a is conducted in our lithography being Recently, it occurred to me that na- major impact beyond the transistor age. laboratory. We have done finer ture should have discovered this by This is the investigation of quantum-ef- linewiclths than by optical projection evolution. So, I decided to see if this feet electronics as a potential replace- anyone-] 400-angstrom linewidths. It was the case. A dragonfly made the mis- ment for transistor-based electronics, is to certainly possible produce 1,500- take of landing near me, and devoted Even though we can make sub-1,000- and Iinewidths 2,000-angstrom by opti- his wings to science, We looked at the angstrom transistors, that approach to cal a projection. But, it's not nianufac- dragonfly's wings in an electron micro- electronics is likely to run out of gas in What turing technology. distinguishes scope. Their thickness is about the the foreseeable future. The nice aspect

FAG U LTY PROFILE

- about quantum-effect electronics is the " What are the limits to quantum highly interdisciplinary work. It in- effect-electronics technology? close volves interaction with Professors As in the early days of surface-wave de- Dirnitri Antoniadis, Terry Orlando, Clif 200 nm vices, the pacing element is still fabrica- Fonstad, andJesus del Alamo in the tion. That's the thing that will always Electrical Engineering and Computer slow you down. Today, quantum-effect Science Professors Marc Department; electronics is complicated by the fact Kastner and Patrick Lee in the Physics that not only do we need good two-di- Department; and people at MIT's Fran- mensional patterning, etching, and de- cis Bitter National Laboratory. It Magnet ago$$* position techniques, but the substrate also addresses basic issues of materials are also more challenging. Some recent physics. We are generally talking about Ill-V theories Professor Pat- 16400406 con-denseddeveloped by-matter materials compound semiconductor rick Lee on electron are be- transport made in a multilaver configuration. tested in the devices we make. We ing 41000-* That involves molecular beam epitaxy. don't know how to use quantum-effect Progress in this area requires the high- electronics in systems yet, but we are level expertise of people like Profes- the surface at scratching by looking An c/cc/iou i uuu/crogu ciph n/a 2,000- sors Clif Fonstad andJesus del Alamo, how behave. they ang-strom-period (600-angstrom linewidth,) and new faculty member Leslie metalgrid on a AIGaAs/GaAs substrate. Kolodziejski. They and their students " What is the re- The modulates the seen progress ofyour grid potential by must be at the forefront of making the a two-dimensional electron at the search in quantum-effect electron- gas high-mobility, low-effective-mass mate- ics? AlGaAs-GaAs interface. As a result, the rials which are the building blocks for wave nature of electron transport is We are at a "fun" where we are quantum-effect electronics. Quantum- stage manfesfed in bacl.tdiffraction and nega- effect electronics combines the best of making devices that involve extremely tive differential tran.sconductance, a molecular beam and nanofabri- high technology. This is done for, what quantum-mechanical effect observedfor epitaxy cation I calf, academic fun and games." When thefirst time at MIT technologies. In electronics, we make quantum-effect devices, we're quantum-effect usually looking for a particular effect. we're really talking about nanofabrica- Then, we make measurements in the tion because we use the fact that when The Grail," if electrons are confined to a small devices to see what happens. We always "Holy you behave as waves. see something unexpected. It's an area will, is to get very strong enough space, they of research that couldn't be more excit- There's a trade-off between tempera- quantum effects at tempera- ture and The finer the di- ing. So, we're finding many surprises dimensions. mensions we can make, the the about quantum effects in condensed- tures such as 77 degrees higher matte structures and devices. We're Kelvin. This means that we temperature at which we can observe effects. The Grail," if not alone in doing this. There are many quantum 'Holy need an engineering tech- will, is to very quantum-effect programs around the you get strong quantum effects at such as 77 de- world because many people see quan- nology able to fabricate temperatures tum-effect electronics eventually re- structures and with grees Kelvin. This means that we need reliably an able to fab- placing transistor-based electronics. engineering technology control at dimen- ricate structures and with Our unique approaches, especially in precise reliably pre- cise control at dimensions below x-ray nanolithography, give us some sions below 500 angstroms. 500 That's the special advantages. That's the challenge! angstroms. challenge! A big question is how are we going to make useful systems. It's exciting to .1 understand that have "cus- see quantum effects, but what can they you tomers" at RLE do for computation? I think that quan- means other than those used in con- foryour veryfine andprecise tum-effect electronics will play a major ventional computers (which are built gratings. role in future computer systems, but as a collection of interconnected Peter Wolff's vision and insight as to the these systems will look very different switches, where each switch is a transis- importance of bringing submicron from today's. Beyond that statement, I tor). People have proposed doing com- structures technology to campus is ex- don't know what I'm talking about. putation where one cell communicates emplified perfectly in an interaction Right now, all we can do is make vague only with its nearest neighbors. Al- that we had recently with Professor Da- "motherhood" statements and say that a though I am beyond my field in saying vid Pritchard and his student David neuron works by communicating be- this, we probably will not build a Keith. They were investigating atomic tween localized patches on a cell mem- switch-based computer with quantum- interferometers that utilize atom dif- brane surface, and acknowledge that effect electronics. There are alternative fraction offof standing-light waves, and we can do useful computation by ways to do it. considered doing it with physical grat-

FACULTY PROFILE

" ings. They came to see us about build- that a crystal's external geometry af- What is needed forfurtherprog- ing a reflection grating, Mark Schatten- fects its orientation. In other words, the ress in submicron structures and burg of the Center for Space Research, surface energy is affected by the exter- microelectronics? who works in the Submicron Struc- nal geometry, and surface energy can What is really needed are young faculty tures Lab on the development of dif- make one orientation more favorable members willing to work in this area. fraction for than gratings x-ray astronomy, another. Additional facilities are also needed for suggested that the same used For several Carl and I have gratings years, molecular beam epitaxy. This is a very for x-ray astronomy could be used in an been looking at surface-energy-driven crucial need since two new faculty atom interferometer. installed a thin They grain growth. By taking extremely members, del Alamo and Kolodziejski, grating and it worked! This was an films (about 100 angstroms thick, so don't have this equipment. In addition, amazing experiment because it demon- that a large percentage of the atoms [ap- we don't have an electron-beam lithog- strated that quantum mechanics really 2%] is at the or proximately right top raphy system. For now, we've decided does work for large compound parti- bottom surfaces), and manipulating the not to have one because it's a large cap- cles like atoms. The we in If things study surface, we can alter surface energy. ital investment, and we can use other textbooks about quantum mechanics we put appropriate indentations in a facilities to make x-ray masks. Eventual- (for example, the particle's wave func- surface, we can modi' surface planar ly, we will need our own c-beam capa- tion goes through all the slits simulta- energy. That can influence phenomena bilities. But, more than anything else, neously) were demonstrated nicely in like grain growth. Both energetics and we need young top-quality people this experiment. It was fun to become kinetics must he taken into account. We working in this area. involved in a peripheral way with Dave are currently working on an extension " Pritchard's interferometer. It was a nice of the graphoepitaxy effect, which I'm Do you have advice for someone application of nanotechnology, and very excited about. It's so far out that I contemplating a career in submi- we're excited when this happens. haven't written a proposal yet because cron technology? The Submicron Lab isn't a no would touch it until it super sponsor Come and do a postdoc here! We will machine shop where a user writes up a works. always attract the best students because work order and we to meet his try this is where the action is. Most of our needs. We see ourselves as collabora- work is done by students working long tors, because the of a development hours in the lab. It's not done on a structure needed a is a by researcher blackboard. But, we need to invent new highly interactive thing. Dave Prit- ways to do graduate research, For ex- chard's case is a good example. What If we look at the problems ample, most graduate students are he thought he wanted for a grating was that Michelson had, quite jointly supervised because no single not what was needed. He had an idea of professor has enough expertise to ad- what he wanted, and asked if we could often they were mechanical vise on all aspects of the problem. In make it. At first, it sounded like a major that most problems people fact, most students don't have enough project. But, it turned out what we had wouldn't have had the pa- time in the normal Ph.D. program to already developed was more appropri- tience to solve. Or, learn everything needed to successfully ate-a transmission grating rather than they pursue quantum-effect research. That's a reflection grating. So, you really must considered the to problem why we frequently have two students interact in order to come up with the be beneath their If working together on one project. They best solution. dignity. tend to where one you take that approach, you develop specialties student, for example, might do all the .Are to build mono- miss a lot of you working opportunities. molecular beam epitaxy, and the other silicon in order to make crystalline student will do other parts of the task. If your own substrates? we take two students and define a proj- Not exactly, I think you are referring to ect, they tend to quickly settle into what materials work in collaboration with they prefer to do. A great misunder- of academic research in the Professor Carl Thompson. The idea is Carl and I also worked on zone- standing "outside" world is that a can to explore new ways of producing use- melting recrystallization. By taking professor tell astudent what to do. ful crystalline films. Graphoepitaxy is polycrystalline materials (like silicon) My experience is that can't do that. You have to an example of this. A few years ago, we on an amorphous substrate, melting you find out what the student likes to do, demonstrated that if we put a very fine them, and then freezing them in a con- and with that. If tell a student to grating in a surface, we could influence trolled way, we got the orientations we go you do he or she doesn't want how crystals grew on that surface. In wanted. This was a collaboration with something to do, it won't done. fact, Dale Flanders, Mike Geis, and I Mike Geis at Lincoln. It didn't always in- get achieved oriented crystal growth on an volve suhmicron structures, but it fre- One ofyour colleagues described amorphous substrate, which was un- quently involved surface patterning. you as an "unabashed technolo- heard of at the time. The graphoepitaxy We found that we could pattern the sur- gist. "How would you like to he re- effect, if viewed more broadly, means face to influence the freezing process. membered?

WE 1,141 1, lIt 1,(II

I would say it another way. I'd like to he remembered as someone who respect- ed technology, because technology is the route to success. But, I'd like to think that my interests didn't end with technologyperse. I think that some people in the history of science were successful because they focused on the pacing element, or the limiting thing. If it happened to be the technology, they went ahead and did it. I could cite Al- bert Michelson. He attacked the tech- nology of precise measurement and in- vented the Michelson interferometer. You could call him an unabashed tech- nologist, if you like. But, I see Michel- son as someone who vigorously at- tacked the problem at hand. When he Dr BernardF Burke, William A. M. Originally, the lDRS system was not in- started his work, no one could have Burden Professor of Astrophysics in the tended for use as an astronomical tele- predicted that Michelson's instruments Department of Physics (left), and RI,E scope, since it was built to relay signals would lead to the Michelson-Morley Sponsored Research Staff John W. Bar- from satellites. In the new application, experiment. Although that wasn't what rett have received the NASA Group the TDRS is pointed at various quasars, caused Einstein to create the theory of Achievement Award for extending with precise frequency control and re- relativity, it was certainly the crucial ex- very-long-baseline interferometry cording at White Sands, New Mexico. periment that laid the foundation for techniques into space applications. Other radio telescopes, in Japan and relativity. I'm not trying to put submi- Their work was cited for expert plan- Australia, simultaneously observe the cron structures in the same category, ning and execution of the Tracking and quasars. These Far East stations are nec- but here is an example of what was be- Data Relay Satellite (TDRS) Very Long essary because the TDRS station is in ing pursued was clearly a technology. Baseline Interferomcter (VLBI) geosynchronous orbit over the Atlantic Thephysics of interferometers was well demonstrations in 1986 and 1987. Ocean. In the late 1960s, RLE's Radio understood, but the technology was These demonstrations produced the Astronomy Group pioneered \TLBI de- pursued because that was the problem world's first astronomical space- velopments and demonstrated the exis- at hand. The impact of developing in- ground VLBI observations, and proved tence in space of compact, very intense terferometer technology has been the technologies for future space-V[.Bl natural masers less than 0.01 of an arc- enormous in astronomy, optics, and missions. The Vt.BI technique, which second in angular size. The Rumford nearly all other areas of science. allows construction of radio antennas Prize recognized this work in 1971. To- I'd like to be remembered as as large as earth, involves the use of day, the new VI.BI applications permit someone who respected technology separate radio telescopes synchronized construction of a radio telescope effec- and solved the problem at hand. When in phase to record signals on magnetic tively larger than twice the earth's di- I say, respect for technology, I mean that tape for computer processing. ameter. some of the problems we have to solve are very nitty-gritty: how do you make material A stick to surface B? These are Quasar nasty little problems. If we look at the problems that Michelson had, quite of- Haystack ten they were mechanical problems N. TDRS that most people wouldn't have had the patience to solve. Or, they considered the problem to be beneath their digni- ty. If you take that approach, you miss a lot of opportunities. Respecting tech- nology involves not taking the attitude that a particular problem is something you can pass on to someone else, or that it's beneath to solve. If your dignity -- I73O-3Q - it's a to solve it. - problem, you've got -- -- Islo_OB, '[hen, it eventually pays dividends.

Schematic illustration of the TDRS VLBI experiment.

, ;+ h Mfl 0 MC tall taUt I In tiUlWöt 0

Dr Jerome YLettvin, Professor of Biol- A has retired from the Institute and Dr. M Assistant Profes- Dr A NihatBerker, Professor of Phys- ogy, John Graybeal, has assumed a faculty position at sor in Physics, has joined RLE's Surfaces ics, was named recipient of the highest Rutgers University. Professor Lettvin and Interfaces Group. His research fo- scientific honor awarded by the Turk- has been associated with RLE since cuses on low-temperature condensed ish government. The TUBITAK Science 1951. Since that time, he and his col- matter physics, and emphasizes super- Award is given annually b' the Turkish have carried out conductivity and novel artificial materi- Scientific and Technical Research leagues distinguished research on the bioelectrical processes al structures. His investigations also in- Foundation, and encompasses all fields involved in cognition and sensory per- clude the consequences of disorder on of science and engineering. Professor ception in living He is widely superconducti'ity, the physics of low- Berkers award cited his important sci- systems. for his excellent work on dimensional and entilic contributions, at an internation- recognized systems, supercon- vortex motion. al level, in the fields of solid-state and vision and pattern recognition pub- ducting lished in the 1959 landmark paper, statistical physics "What the Frog's Eye Tells the Frog's Brain." Professor Lettvin will be main- taining an affiliation with BLE.

Dr Simon (2J Mochrie('84 Ph.D.), As- sistant Professor of Physics, has joined Dr Sow-Hsin Chen, Professor of Nucle- RLE's Surfaces and Interfaces Group. ar Engineering, has received the Alex- Dr. SrinivasDevada.s, Assistant Profes- Ilis research involves the stability of ander Von Humboldt U. S. Senior Sci- sor of Electrical Engineering and Com- both metal and semiconductor sur- entist Award for outstanding contri- puter Science, has joined RLE from the faces. Before coming to MIT in 1987, he butions in the investigation of t.Jniversity of California at Berkeley, initiated a research program in the interfacial and aggregational phenom- where he recently completed his Ph.D. structure and stability of clean metal ena in amphiphile/water/oil systems in high-level CAD synthesis and testing. surfaces at AT&T Bell Laboratories. This using small-angle neutron scattering He has made significant contributions research has led to a greater under- and photon correlation spectroscopy. to datapath synthesis and logic synthe- standing of the microscopic structure Professor Chen spent six months visit- sis and verification. Currently in RLE's of Au (100) reconstruction, and the dis- at several West ing physics departments Circuits and Systems Group, he is in- covery of two new phase transitions German academic institutions in con- vestigating synthesis from logic to the (rotational and two-dimensional melt- nection with this award. gate level. ing) on that surface.

TENURE GRANTED Congratulations to threefaculty members atRU? who were granted tenure eJjécliveJuly 1988:

Dr. Sylvia T (7eyer, Class of 1943 Career Development Associate p Professor of Chemistry. BA. ('74) I lope College and Ph.D. ('79) Uni- versity of California at Berkeley. Before joining the MIT faculty in 1981, Professor Ceyer was a post- doctoral fellow at the National Bu- reau of Standards. She has re- ceived the 1981 Dreyfus Award, the 1987 Flarold E. Edgerton Award, the 1988 Baker Award, and the Dr Carol V Espy-Wilson has been ap- 1988 Young Scholar Award from pointed Research Associate in ELE's the AAUW Educational Foundation. Speech Communication Group. Work- She also holds an Arthur P. Sloan ing with Professor Kenneth A. Stevens, Foundation Fellowship and a Ca- she was previously a postdoctoral fel- mille and Henry Dreyfus Teacher- low conducting research in feature- Scholar grant. based word recognition systems. Dr. Espy-Wilson developed a perceptual experiment using synthetic speech to better understand the primary and sec- ondary features needed to distinguish acoustically similar sounds. She re- ceived her B.S.E.E. from Stanford Uni- Dr. KeithA Nelson, Associate Pro- versity in 1979, and SM. ('81) and Ph.D. fessor of Chemistry. B.S. ('76) and ('88) from MIT. (Photo bj'Mark Wilson.) Ph.D. ('81) Stanford University. Professor Nelson came to MIT in 1982 after a postdoctoral fellow- ship at UCLA. In 1985, he was se- lected as a National Science Foun- dation Presidential Young Investigator, and was awarded an Alfred P. Sloan Fellowship in 1987. He recently received the 1988 Cob- lentz Award for outstanding work in spectroscopy.

. zoqyh~ Dr. Carl V Thompson, Associate Dr. P_ Glass was Re- James appointed Professor of Electronic Materials in search Scientist in lily's Speech the Department of Materials Sci- Communication Group. Since 1985, he ence and Engineering. S.B. ('76) has worked with Dr. Victor W. Zue in MIT, and SM. ('77) and Ph.D. ('82) the area of man-machine communica- Harvard University'. Professor tion, and developed the LAMINAR and Thompson came to MIT in 1982 as APACHIE softwares for LISP machine an IBM postdoctoral fellow in RLE. workstations. LAMINAR is an interactive In 1983, he became an Assistant used to from facility synthesize speech Professor, and was Mitsui Assistant different vocal-tract configurations. Professor from 1985-87. He is APACHIE, also an interactive facility, is widely known for his fundamental used to perform acoustic-phonetic research in thin-film processing of Dr. Glass received analysis speech. and the control of thin-film micro- his SR. in from Carleton lJniversi- 1982 structures for electronic devices. tv and SM. ('85) and Ph.D. ('88) from MIT.

BACK TO THE FUTURE: Professor Dudley A. Buck (1927-1959)

Born in San Francisco in 1927, Dudley A. Buck received his B.S. degree in electrical engineering at the University of Washington in 1948, and spent two years as a communications officer with the U.S. Navy. He entered MIT in 1950, and served as a research assistant on the Servomechanisms Laboratory's Project Whirlwind while working to- wards his master's degree, which he re- ceived in 1952. He served as an instruc- tor in the Department of Electrical Engineering until 1958, when he re- ceived his Sc.D. and was appointed As- sistant Professor. He was also associat- ed with both RLE and Lincoln Laboratory. During his nine years at MIT, Pro- fessor Buck made outstanding contri- butions to the field of low-temperature physics. Tie was most recognized for the development of the cryotron, a su- perconductive, magnetically controlled gating device that was hailed as a revo- lutionary component for miniaturizing the room-sized computers of the '50s. A "Dudley Buck was one of the most imaginativepersons to do research technical "The Su- in paper, Cryotron-A RLE His work on computer elements and photolithography was perconductive Computer Compound," years ahead the times.. of was recognized with the BrowderJ. -MIT President Emeritus Jerome B. Wiesner Thompson Memorial Prize of the Insti- tute of Radio Engineers in 1957, an "7'he day is rapidly drawing near when digital computers will no longer be made by award that was made to a scientist un- assembling thousands of individually manufactured parts into plug-in assemblies der 30 years of age who presented the and then completing their interconnection with back-panel wiring. An alternative to year's most outstanding paper. Dudley this method is one in which an entire computer or a largepart of a computer is made Buck also won an honorable mention in a single process. Vacuum deposition of electrodes onto blocks ofpure silicon or in Eta Kappa Nu's selection of the Out- germanium and the subsequent diffusion of the electrode material into the block to standing Young Electrical Engineer of form junctions is a most promising method. The successful development of this meth- 1958. During his last two years, he od would allow large numbers of transistors and all their interconnecting wiring to sought to carry miniaturization even be made in one operation. Vacuum deposition of magnetic materials and conduc- further by attempting to make cross- tors to form coincident-current magnetic-core memory planes is a secondpromising film cryotrons with dimensions only a method that will allow an entire memory to be made in one operation. The vacuum few millionths of an inch. deposition of superconductive switching and memo, circuits is a third method that While his professional accom- will make possible the printing ofan entire computer. The authorsfeel sure that the plishments were many and varied, his most significant milestone in computer component technology will be the an- loyalty to MIT and the thoroughness of nouncement by one or more firms, in perhaps 2 years, that all of the technicaiprob- his teaching were equally outstanding. lems of building a printed system have been solved, and that one of their engineers In addition to his research, he repre- with his vacuum system can make a digital computer in an hour...." sented the MIT Admissions Office on high school visits, and exhibited what "An Approach to Microminiature Printed Systems" by Dudley A. BUCk and Kenneth R one called "a Shoulders, Proceedings of the Eastern joint Computer Conference. colleague contagious quality of optimism, enthusiasm, and In December 1958, Dudley Buck presented the above paper outlining a scheme to just plain joy about each man's work." form a thin film upon which would be placed a resist by electron-beam polymeriza- Because of his strong convictions about tion of siloxane vapors. Then, a selective removal process would be carried out b the importance of education, and his means of a gaseous etchant that would leave the desired superconductor as an cle- deep interest in youth, he was elected ment in cryotron fabrication. From the time of this paper until his death, Dudley chairman of the Wilmington School Buck worked on various aspects of vapor deposition of refractory metals as good Committee. He was also a former scout- superconductive films, in the first step of a lithographic technique. master and a lay speaker in the Wil-

mington Methodist Church. Dudley Buck's dedication arid cre- ativity were sources of admiration to his students and colleagues. Before his untimely death in 1959, he possessed all the attributes of greatness and had already achieved much in a short time.

First came the vacuum tube, then the transistor. The cryotron was destined to spark an- other revolution in electronics. Above the vapors of liquid helium, Dudley Buck compares afragile, bulky vacuum tube with his cryotron. (Photo Gjon Miii)

The cr)'olron u.'as man 'sfirst practical use of superconductivity-the ability of some metals to conduct current with no resis- tance at low (be- extremely temperatures Electron micrograpi) oja inoiiixienion film etched into 0.00 1-inch s(lital-"shows Dudley low Fahrenheit). In the -420 degrees Buck's prop osed process to produce etched wiring on a 0.1-micron scale involving the hand its inventor, is the of incredibly selective removal ofa thin film. It dfferedfrom conventional systems of the day since the small will into a thim- crvotron (100 fit process was carried out in a vacuum system, and electrons or ions replaced light as a ble). Its was based on the operation effects means to control the deposition of the resist. of magneticfields on superconductivity at liquid helium temperatures. (Photos courtesy ofMIT Historical Collections)

Jones, MIT Center for Advanced Engi- UPDATE: UPDATE: neering Study, Room 9-234, Cam- RLE Collegium Communications bridge, MA 02139 (617) 253-7444. Publications The ELF Communications group wel- Collegium Membership comes inquiries regarding BLE The RLE Collegium was established in The following new publications are research and publications. 1987 to promote innovative relation- available: Contact: ships between the Laboratory and busi- Barbara Passero ness " The 316-page Progress Report No. 130 organizations. The goal of RLE's Communications Officer is to increase conimunica- reports research results from all RIFt Collegium Research Laboratory of Electronics research groups during 1987 and in- tion between RLE researchers and in- Room 36-412 dustrial in electronics cludes an extensive bibliography, in- professionals Massachusetts Institute of Technology and related fields. dex of research project staff, and an RLE Cambridge, MA 02139 members have the personnel roster. Collegiuni op- (617) 253-2566 to close affiliations portunity develop " A companion publication, the RIP with the Lahoratorv's research staff, and Publications Update, lists abstracts of can access results quickly emerging reports published by RLE from January and scientific directions. Collegiuns 1987 throughJune 1988. Reports are benefits include access to a wide range listed by subject area: general interest, of educational video publications, pro- circuits and systems, digital signal pro- RI.E disclosures, semi- grams, patent cessing, image processing, materials nars, laboratory visits, and an on-line and fabrication, optics and devices, calendar of events. physics of thermonuclear plasmas, and The RLF Collegium membership speech communication. An author in- fee is $20,000 annually. Members of dex is also included. MIT's Industrial Liaison Program can RLE currents elect to transfer 25% of their ILP mem- RLE currents is a biannual publication bership fee to the RLE Collegium. After Videotapes of the Research Laboratory of an initial one-year membership, a ELF, in cooperation with MIT's Center Electronics at the Massachusetts three-year commitment will be re- for Advanced Engineering Study and Institute of Technology. (luired. Membership benefits are sup- the Industrial Liaison Program, has vid- Jonathanported by the Collegium fee. In addi- eotaped research presentations from Allen Editor-in-chief tion, these funds will encourage new two symposia, Speech Communication ,John F. Cook Photography research initiatives and build new labo- and Processing (December 198), and ratory facilities within RLE. Future Directions in Electronics: 40th Everett Design ...... Design For more information on the ELE Annit'ersary Symposium RIP (Octo- of Dorothy A. rleischer StaffWriter Collegium, please contact RE head- ber 1986). These color videotapes can and Editor . quarters or the Industrial Liaison Pro- be purchased or rented. For further in- Passeco. Production gram at MIT. formation, contact B. narbaraJ. please Carolyn and Circulation

Donna Maria Ticchi ...... Managing Editor

HenrsJ. Zin'tmermann Advice and Perspective

The staff of currents would like to thank Professor Henry I. Smith for his technical guidance during the preparation of this issue.

Inquiries may be addressed to RLE currents, 36-412, Research Laboratory of Electronics, Massachusetts Institute

of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139. (ISSN 1040-20121

RLE ProTector Jacob White (standing)presents his research on custom integrated circuits and to RIP members computer-aided design Collegiumn from Pitney Bowes during a recent on- 1955, siassacltuseits instituteof technology. campus briefing. (Photo by Dorothy A. Fleischer) All rights resenedworldwide.

20