D. J. Ehrlich. J. G. Black, M. Rothschild, and S. W. Pang

In Situ Technologies forSemiconductor Patterning

In situ processing methods simplifY fabrication of semiconductor devices and,furthermore, eliminate exposure of surfaces to air, liquids, or human operators. Theprincipal methods for in situ patterning of semiconductors are reviewed. Two methods,laser-direct writing and excimer-Iaser projection, appear to be promising candidates foradvanced semiconductor fabrication. Ion-beam techniques will have important specialapplications.

In situ processing of microelectronics maypermit the fabrication of semiconductor devicesby dramatically simpler and more precisemethods than those of current technology [1).These processing methods are based on new andpotentially revolutionary nonlithographictechnologies, and rely heavily on the stilldeveloping area of beam-controlled techniquesfor microfabrication.

The underlying principle of in situprocessing isthe replacement of the current indirectsemiconductor patterning process ofphotoresist pattern-transfer sequences withdirect-photon-, electron-, and ion-beam­controlled techniques that produce microscopicreactions on the semiconductor. This approach,in principle, eliminates the usual solvent and airexposure, which degrade Critical electronicsurfaces in resist processing.

In the most advanced application. anelectronic device could be completely processedwithin a single vacuum chamber, in a mannersimilar to molecular-beam epitaxy technology. Asecond application, already partly realized, isreal-time fabrication (again withoutphotoresists), using "closed-loop" construction,of highly precise microdevices. This approachstreamlines conception, design, and fabricationsequences and allows more efficient realizationof new device ideas. Current fabricationtechnology is ponderously slow and inordinatelyexpensive for prototyping, and thereforeimpedes the development of new device andcircuit designs. To address these limitations, in

The Lincoln Laboratory Journal. Volume 1, Number 2 (1988)

situ processing combines existing and futurebeam technologies with compatible nonbeammethods - to achieve totalvacuum and/or real­time fabrication.

New semiconductor patterning technologiesoffer possibilities ranging from development of a0.25-llm optical lithography to an entirelynonlithographic laser-direct-writing technologyfor post-fabrication (final-step) processing andcircuit prototyping. The former could provide asimpler, faster, and less expensive alternative tox-ray, electron, or ion-beam lithography fornext-generation VLSI. The latter could be usedfor high-accuracy trimming, wafer-scaleintegration, and interactive circuit design.

Studies of laser-controlled microchemicalreactions for deposition, etching, and doping ofelectronic materials began in the early 1980s.Since that time, an increasing number ofresearch groups have become interested and anincreasing breadth for these processingmethods has been demonstrated. A variety oftechnology targets have been identified.

Beam patterning is probably the most criticalarea of development for the new technology.The principal alternatives to classicallithography are focused-beam-writing methods,for low-volume operations where turnaroundtime orprocess localization is at a premium, andin situ flood-beam prin.ting methods for high­volume production. Because the applicationsand the critical issues are qUite different, weconsider these two classes of methodsseparately.

187

Ehrlich et aI. - In Situ Technologies for Semiconductor Patterning

Focused-Beam-Writing Methods

The most mature of the beam-writingmethods. in terms of instrumentation. iselectron-beam technology. A source and/orspace-charge-limited flux of 1018 quanta/cm2 scan be readily delivered to a substrate.Commercial scanning-electron-beam instru­ments can achieve <10-nm spot sizes. canregister to about 50-nm precision over largefields, and can scan at multiple-MHz pixel rates.A particular strength of this technology is thesophistication of the software control that hasbeen developed for lithography. Althoughresearch [2,3) in direct processing (deposition)extends from the 1950s. applications are limitedby the small particle momentum and low crosssections of gas-. adsorbed-, and solid-phasereactions. Therefore. reactive electron-beamprocesses generally cannot utilize the high scanrates that are achievable.

The constraint of cross sections is generallygreater at the typical high energy (20 to 100 keY)of most electron-beam columns. Currentresearch [4.5) has begun to stress lower-energyprimary beams and reactions stimulated bysecondary electrons. where cross sections aremuch greater. At the moment applications ofelectron-beam reactions remain in the researchenvironment.

An alternative direct-writing technology,based on surface reactions stimulated with UV/visible lasers [6,7], has been Vigorouslydeveloped since the late 1970s. An almostarbitrarily high quantum flux (> 1024 quanta/cm2 s) can be achieved. Spot sizes are limited bydiffraction to -0.5 /lm, although nonlinearreactions permit practical writing of featuresless than 0.25/lm. Intense research activity overthe last five years has led to the demonstrationof a large variety of deposition. etching. anddoping reactions.

Laser-direct writing employs a scannedmicrometer or submicrometer reaction forserial. but real-time, modification of electronicstructures. Although a conceptually radicaldeparture from planar fabrication technology,direct writing is intended as a new tool toenhance this existing technology. Such in situ

188

writing offers solutions to some pereniallyawkward problems for lithography.

Commercial instrumentation, however. lagslaboratory systems significantly. Althoughlaser-writing instrumentation has beendemonstrated to be capable of matching orexceeding (8) electron-beam technology in pixelrate. the registration and imaging capabilities ofoptical systems are fundamentally more limited.This disadvantage is offset. for reasons to bediscussed below, by reaction speed, becausegood laser reactions are many orders ofmagnitude faster than alternative beam-writingtechniques. A few applications are now inmanufactUring practice. with several others inearly stages of commercialization. Table 1 listsseveral of these applications.

A third alternative. focused-ion-beamtechnology. shares many of the properties ofelectron-beam technology. and. in part becauseof this, instrumentation has been developedvery rapidly. Achievable fluxes with brightsources (eg, Ga+, Si++) are _10 18 quanta/cm2 s in200-to-50-nm-diameter spots. The greatermomentum of ions relative to electrons providesgreater substrate/chemical actiVity. Inaddition. direct-write ion implantation is anattractive potential application. Many articlesdescribe the status of this technology (9).

Focused-ion-beam deposition has beenapplied commercially to photomask repair. Inaddition. several laboratories have attachedfocused ion beams (FIB) to molecular-beamepitaxy systems to achieve in situ three­dimensional patterning.

Molecular-Flux-Limited Rate

Efficient beam-writing reactions operating inthe gas phase are often limited. not by the arrivalrate of beam quanta. but by the molecular fluxof reactants onto and away from the surface inthe vapor. This molecular flux depends onambient pressure and. at high pressures. on thesize of the reaction volume. A simple treatmentconsidering diffusive transport of reactants isgiven in Ref. 6. Figure 1 shows the steady-staterate calculated for a reaction efficiency of 0.1(one out of ten incident molecules react at the

The Lincoln Laboratory Journal. Volume 1. Number 2 (l988)

Ehrlich et aI. - In Situ Technologies Jor Semiconductor Patterning

Table 1. Representative Laser-Direct-Write Applications and Processes

Application

Gate-Array Structuring

Patterned Etch of Si

Mask Repair

Patterned Etch of AI

Two-Level IC Metallization

Microwave Circuit Tuning

Nonplanar Self-DevelopedLithography for HybridInterconnect

Process

p+-polysilicon depositionby 488-nm pyrolysis of

SiHiB2H6

488-nm-driven photochemicalthermal etch by CI2

Cd deposition by 257-nmphotolysis of Cd(CH3)2;Cr deposition by pyrolysisof Cr(CO)6

Thermally driven etch byHN03/H3PO4/K2Cr207solution

Pyrolytic W deposition fromWF6' ablative via formation

Pyrolytic W deposition fromWF6' ablative via formation

CI2 etch of amorphous Si layer

Reference

7(b)

7(c)

7(d), 7(e)

7(f)

7(g)

7(h)

7(i)

, .

surface) and for various beam spot sizes. Atlarge (lOOO-~m) spot size, boundary-layereffects become dominant, restricting reactantarrival and departure rates. The size of thereaction zone limits rate enhancement, whichsmaller zones experience with increasingpressure (ie, above a few Torr). This is also theusual pressure/rate limit of large-areareactions in conventional processing.

At smaller, eg, I-11m beam size, rates increasenearly linearly with pressure to pressures ofseveral atmospheres. This full enhancement isrealizable in laser deposition, where pressurescaling frequently permits rate enhancementsof 103 over conventional large-area chemicalvapor deposition in the same chemical systems.Pressure scaling in electron-beam and ion­beam systems is restricted by the requirement ofhigh vacuum at the beam source. Nonetheless,with differential vacuum pumping and localized

The Lincoln Laboratory Journal. Volume 1. Number 2 (1988)

injection of the reactant, it has been possible toreach effective molecular pressures of -1 Torr atthe reaction zone for focused-ion-beamdeposition. This distinction is important inlocating the strategic areas of application of thetwo technologies.

Comparison of Focused-Ion-Beamand Focused-Laser Technologies

At this time, the principal writing alternativesfor in situ processing are the focused-ion-beamand laser-direct-write technologies. Keycharacteristics of each are listed in Table 2. TheFIB has advantages in the sophistication ofinstrumentation and in resolution, both ofwhich carry over for processing operations andalso for imaging and registration. Lasertechnology has advantages in quantum energy(2 to 7 eV), which is close to typical chemical-

189

Ehrlich et aI. - In Situ Technologies Jor Semiconductor Patterning

Reaction Zone Radius:

1 J.Lmbond strengths (permitting chemicalselectivity), and in the magnitude of the beamflux that can be delivered in a relatively smallspot. The greatest advantage of the laser isprobably the transparency of (high-pressure)gases to photons, which permits processdiversity and speed.

In practice. both technologies have carved outdifferent niches. Figure 2 illustrates the overlapin microchemical processing. Limitations inthroughput prevent both technologies, in theirsimple forms. from economic mass productionof silicon chips. However. nonlinear andmonolayer surface-modification reactions couldextend the use of both technologies. Laser­writing techniques may enter a regime oflimited-purpose mass chip production afterconsiderable development. The likely areas ofapplication for direct-writing techniques aredevice research. mask repair, trimming.prototyping. and packaging of high-densitydigital and optoelectronic systems. Both laserand ion-beam methods are likely to be used forcircuit surgery (ie, debugging and analysis). Asnoted first at the Optoelectronics Joint ResearchLaboratory [101. FIB have a natural synergy withmolecular-beam epitaxy machines.

1001.0 10

Pressure (Torr)

0.1

Fig. 1 - Results of a general analysis of process ratelimitations for focused-beam microchemical techniques inthe high-rate mass-transport-limited regime. The modeling(from Ref. 6) takes into account mass transport limitationsby gas-phase diffusion. The transport limit is calculated for1,OOO-j1.m, 100-pm, 10-pm and 1-pm beam diameters anda O. 1 surface-reaction efficiency. In all cases, boundary­layer effects pin rates as reactions are scaled to higherpressure. However, smallerbeam diameters permitscalingto much higher rates. Focused-ion-beam (FIB) reactionsare limited currently to -1-Torr effective pressures by theefficiency ofdifferentialpumping schemes. Laser reactionsare limited to about one atmosphere in current reaction celldesigns.

co·u 1013C'IJQ)

c:

NC/) 1017 -----------E Laser (1 J.Lm Spot)o"­~ 1016::JoQ)

-0 1015~

Table 2. Beam Characteristics for Writing Applications

(I) Focused Ion Beam (II) Laser Beam (Photomechanical)

Resolution: -0.05 11m

Quantum: 30-200 keV

Flux: _1018 quantalcm2 s

Special Strengths

Sophisticated Scanning

Large Particle Momentum

Functions

Implantation

Etching, Deposition

Imaging and Registration

Resolution: 0.2-1 11m

Quantum: 2-7 eV

Flux: _1024 - 1027 quantalcm2 s

Special Strengths

Selective Photochemistry

High Speed

Functions

Deposition

Etching

Shallow Doping

190 The Lincoln Laboratory Journal. Volume 1. Number 2 (1988)

Ehrlich et aI. - In Situ Technologies for Semiconductor Patterning

Packaging

Metallization

FIB FLUX LIMIT

ICPrototype

IC MassProduction LASER REACTIONS

TRANSPORT LIMIT~

r-------;;;-~---­

IIIII

Device Trimming,Selected-AreaCustomization

Mask ~epair

FIBTRANSPORT

LIMIT

Vl"'­(/)

Q)xa:"CQ)Q)Co

CJ)

(/)(/)Q)Uo...

a.."CQ)...:lC"Q)a:

10-1

Pixel Size (j.Lm)

Fig. 2 - Current focused-beam microchemical processing applications in electronics, classified by required resolution andpixel-writing rate. The blue line circumscribes the region currently addressed by FIB; the red line marks a similar regionaddressed by focused lasers. The molecular-reaction flux limit in each case is indicated. A falloff in pixel rate at large pixelsize for FIB processing results from the finite total beam flux available from liquid-metal ion sources. FIB technology hasimportant resolution advantages but is limited by the effects illustrated in Fig. 1 to lower process rates. Both technologies arepotentially extendable, by a factor of three to five, to smaller pixel size using nonlinear reactions (already demonstrated forfocused lasers). Andby using surface-modification processes, FIB and focused-laser technologies can be extended upwardin rate by several orders of magnitude.

The Lincoln Laboratory Journal, Volume 1. Number 2 (l988J 191

Ehrlich et aI. - In Situ Technologies for Semiconductor Patterning

Printing Technologies

Pattern printing methods employ eitherextended-field-of-view projection techniques orproximity masks. These printing methods havebeen applied by using photons, ions, andelectrons in all conceivable combinations. At themoment, the most vigorously developed - andalso the most suitable for in situ processing ­are optical projection and (for high resolution)ion-beam proximity/projection techniques.Ions are favored over electrons (among otherreasons) for having very little proximity effect;the more limited lateral range of the ionexposure leads to sharper features.

Optical projection technology will becomevastly more capable of in situ processing whendeep-UV excimer lasers, which offer resolutionand depth-of-field advantages, are incorporatedas radiation sources [11). The configurationoriginally used in Lincoln Laboratoryexperiments is shown in Fig. 3. Deep-UV (193­or 157-nm) light pulses, 20 ns in duration,ilhiminate a nearly conventional photomask

that is imaged at reduction onto the wafer. Atypical fluence at the wafer is -30 to 300 mJ /cm2 (1-to-1O-MW/cm2 peak intensity). Thisfluence is sufficient to induce many forms ofnonlinear solid or solid/vapor reactionsincluding, for example, ablation of thin filmsfrom the wafer. As shown in Fig. 4,extraordinarily high resolution (the highest ofany optical printing technique) can be achieved.Theoretical throughputs equal those of currentmass-production methods, but instrumen­tation and supporting alignment/overlaytechnologies to exploit the resolution remain tobe developed.

Excimer-Iaser projection is intended as a newlithography for mass production. The far-UV orvacuum-UV output of a small excimer laser isused in a geometry similar to that of aconventional reduction stepper. Excimerprojection is potentially important both as anovel resist and as a resistless patterningtechnology. A variety of new reactions based oninterfacial chemistry, defect injection, or solidtransformations are induced in extended-area

~-----_..._-,'-=-------

----------------~----

jwafer

III

tIII IIJ

vapor/Cell

SchwarzchildReflector(NA 0.5)

\

-~------Jj

\36X

TransmissionReticle

(Cr on Quartz/LiF)

DivergingCondenser

Excimer-LaserBeam

(193 nm, 157 nm)

~ ------ ---

Fig. 3 - Configuration for the original submicrometer-resolution excimer-projection printing experiments (Ref. 11). Pulsed193-nm or 157-nm light from an excimer laser illuminates a transmission mask in much the same geometry as optical step­and-repeat machines. Reflective optics are used to reduce the image of the reticle onto the wafer. The wafer is mounted ina vapor cell to control the ambient environment.

192 The Lincoln Laboratory Journal. Volume 1. Number 2 (1988)

Ehrlich et aI. - In Situ Technologies for Semiconductor Patterning

Fig. 4 - Scanning- electron micrograph of O. 13-jlm lines andspaces in diamond-like carbon resist on GaAs. The 150­nm-thick resist was deposited in a radio-frequency plasmain butane gas at 10-mTorr pressure, in a geometry similarto that of a reactive-ion-etching system. Exposure was inair, with one 193-nm excimer-Iaser pulse; the f1uence wasabout 0.13 J/cm2. The f1uence is near threshold for ablation,and its effect is blistering of the carbon, which permitssubsequent transfer into the GaAs substrate by reactive ionetching. The unexposed carbon serves as the etch mask.

patterns. It is anticipated that, based on itspotential resolution, speed, and cost ofimplementation, this technology could becomea viable 0.25-!-lm production technology.Although less radical a departure in philosophyfrom current methods, the surface interactionsused for excimer processing are often moreradically new than direct-writing reactions.

An alternative printing technology with yethigher resolution is the masked-ion-beammethod [12-14]. As with optical projectiontechnology, the theoretical throughput of themasked-ion-beam technology is sufficient formass production. The configuration and atypical result are indicated in Fig. 5. A collimatedflux of accelerated (50-to-100-keV) ions (forexample, protons) are directed through a stencil[12,13] or channeling [14] mask onto asubstrate. The source is typically a conventionalion implanter, which will deliver a 1013_to_10 14_

cm-2 dose, acceptable for driving many reactiveprocesses in a few seconds or less. A significantadvantage for in situprocessing, relative to otherhigh-resolution (eg, x-ray) methods, is the

The Lincoln Laboratory Journal. Volume 1. Number 2 (1988)

allowance of a large (> 100- !-lm) mask/wafer gapwithout loss of linewidth control. As with x-raylithography, practical mask technology is anactive and important area of research. Recently,efforts to develop a reduction ion-projectiontechnology employing stencil masks haveincreased [IS]. This technology could relaxaspects of the thermal-loading and mask­fabrication (feature size) constraints of masked­ion-beam methods and could allow macroscopicseparation of the mask and wafer.

Processing Issues

Laser-direct patterning methods fordeposition, etching, and doping of differentelectronic materials are currently available formore than 100 different surface reactions. Fordirect writing these have been reviewed recently[6]; many techniques based on photochemical orthermal reactions in vapors or liqUids have beendeveloped. A linewidth of 0.2 !-lm, well below theRayleigh limit, can be achieved in well­controlled reactions [16].

The available methods developed aresummarized in Fig. 6, which locates genericclasses of laser techniques as they now standwith regard to rates and induced localtemperature. Since Virtually all surfacereactions are accelerated by heating, the trade­off between rate requirements and maximumtemperature often determines the choice ofmethod for an application. The photochemicalmethods on the left side of the figure, forexample, are better for processing membranesor fragile semiconductor substrates; the speedof higher-temperature thermal methods on theright makes them preferable for less sensitiveapplications.

Steady-state rates for laser reactions are farhigher than they are for the closely relatedtraditional counterparts [6]. For example,pyrolytic silicon chemical vapor deposition(CVD), which in a traditional furnace typicallyhas a maximum useful rate of less than 10-2

!-lm/s, produced excellent electronic-grade poly­Si at a rate greater than 102 !-lm/s in laser-directwriting [6]. For this reason, laser-direct writinghas been applied to such applications as circuit

193

Ehrlich et aI. - In Situ Technologies for Semiconductor Patterning

Excimer Projection with andwithout Resists

prototyping, mask repair, and in situ design ortrimming.

For projection microchemistry, which uses ashort-pulse excimer source rather than thecontinuous-wave laser sources of direct writing,reactions are just beginning to be developed(17). Some of the methods of direct writingshould be immediately transferable.

Among the most useful near-term processesfor projection microchemistry may be reactionsbased on solid transformations in thin films,which are illustrated by the Alia cermet systemin Fig. 7. In this process. a thin film. typicallyabout 100 nm thick, of a metastable Aliacomposite is deposited by evaporation of Alunder a controlled oxygen ambient. The mm isexposed to single 20-nsArF-laser pulses. whichrenders it dry- or wet-developable. The highsensitivity and contrast of such rugged etch­resistant mms permits efficient patterning andgives a resolution ofless than 0.51lm (17). Thusthe great strength of excimer microchemicalmethods may be in lithography-liketechnologies, which are likely to be relativelystraightforward extensions of current opticalprojection techniques.

Ion SourceH+/Ar+

, , Ion= = - Accelerator= == = • 10-100 kV, •

~ L8J-s.M.SSSeparator

c::::J c=I Beamc::::J c=I- Scanner

~ -- -Stencil MaskGap

T -ResistSubstrate

(a)

1....~1----1 p.m-----'.-f

Sub-0.1-p.m Resolution

(b)

Fig. 5 - (a) Schematic representation of the LincolnLaboratory masked-ion-beam lithography exposurestation. A beam of ions (H+ or Ar) of about 20 to 100 ke Vis generatedandmass-selectedbyan electromagnet. A setof scanning plates is used to rock the beam angle whenneeded. The ion flux has a beam divergence ofless than 0.5mrad and a current density of0.2 p.A1crnZ, with a better than99% uniformity over 1 in2 . The stencil mask hastransmission holes in a 1-p.m-thick SiNx membrane and isseparated from the sample byan intervening spacer to forma gap ranging from 25 to 300 p.m. (b) Scanning-electronmicrograph of BO-nm lines and spaces exposed in 300 nmofpolymethylmethacrylate (PMMA). A 50-ke Vproton beamand a total dose of2 x 10'3 ions/crnZ were used with a 1-p.m­thick SiNx stencil mask at a 25-p.m gap. T~e exposure timewas 10 s, followed by 30-s development In a 40% methyl­isobutyl ketone/60% isopropyl-alcohol solution.

The advantages of excimer lasers for more orless traditional resist exposure have beendiscussed (18). Now, however. new inorganicresist and resistless processes are underdevelopment.

Apart from the implicit attractiveness of ashort wavelength and an easily obtained high­UV flux. excimer projection offers two keycapabilities. One is very high contrast (.2:10).because it is not limited to the relatively narrowconfmes of traditional organic resists. Andrelated to this property is the second keyproperty, sensitivity enhancement (ie. positivenonreciprocity), which accompanies high-dose­rate exposure in many microchemical reactivesystems (17).

Many systems undergo orders of magnitude­sensitivity enhancements (reduced requiredtotal dose) as the dose rate is increased by

194 The Lincoln Laboratory Journal. Volume 1. Number 2 (1988)

Ehrlich et aI. - In Situ Technologies Jor Semiconductor Patterning

Temperature Rise (K)

10 100 1000

"

PHOTOCHEMICAL

PHOTOCHEMICAUTHERMAL

THERMAL

I I II I II I I

GaAs/CH3Br--1 I Photoetching

.I......- InP/CH3Br

Si/CI2

Photodeposition

AI,Ti,Cd,Zn,W,GeI

ElectrochemicalEtching

GaAs, CdSI

Photo-Polymerization I

PMMA II

-- I II I PhotochemicalI I DopingI II I InP, SiI II II I MicroalloyingI I -AI, Zn-lI

r I 1 II Activated II Etching I

IIAI II II I

I I PyrolyticI II I Deposition

I I f--- Si-----lI I I! I

Fig. 6 - Classification of laser-induced microreactions by their induced surface temperature rise. The photochemicalreactions operate at very low temperatures; the thermal processes offer much higher reaction rates.

The Lincoln Laboratory Journal. Volume 1, Number 2 (1988) 195

Ehrlich et aI. - In Situ Technologies Jor Semiconductor Patterning

increasing the incident pulse energy. Single­pulse exposure, an advantage for exposure-tooldesign simplicity, can be obtained in manysystems. An example ofO.4-~mlines and spacesin 1.5-~m-thick polymethylmethacrylate(PMMA), obtained with ArF projection, is shownin Fig. 8.

Circuit Prototyping byMicrochemical Direct Writing

0.4 JLrnLines 'j

10 JLrn

The long lead time required to create masksand to process ICs conventionally makes

Fig. 7 - Scanning-electron micrographs demonstratingthe excellent submicrometer resolution obtainable byprojection patterning ofan AI/O cermet film at 193 nm. One120-mJ/cm2 pulse was used to induce solid-statetransformations in the 30-nm-thick film. Development wasperformed by wet etching in nitric acid/phosphoric acidfollowed by reactive ion etching in CHF3 .

196

Fig. 8 - Projection patterning of an organic bilayer. A 140­nm-thick film of AZ 1350 J, which served as the imaginglayer, was ablated in one pulse (80 mJ/cm 2) of 193-nmradiation. The underlying planarizing layer was 1.5-llm­thick PMMA. Following patterning of the imaging layer, thePMMA was exposed to UV-Iamp flood illumination and theexposed areas were wet-developed. The photograph is ascanning-electron micrograph of O.4-llm lines and spaces.Some damage is done to the fragile structure by the heavyelectron dose used in scanning-electron microscopy.

interactive modification of circuits in a testenvironment a desirable capability. A series oflaser-microchemical techniques was developedfor restructuring a semicustom integratedcircuit under test [7a). Two such laser-activatedprocesses, one for local aluminum etching, theother for poly-Si deposition, were used in theinitial application. The first process (used toeliminate existing connections) is a chemicaletching process, which is thermally activated bya tightly focused laser beam. The second process(used to add connections to the existing circuit)is laser-activated CVD of boron-doped poly-Si.

Using these two tools, a designer can debugand optimize a circuit after conventionalfabrication. Many tasks can be performed thatpreviously required multiple passes at design orcomputer simulation. Circuits can be modifiedwith design corrections, they can be optimizedfor performance, redundant sections can beadded or deleted to improve yield, problemsections can be isolated and connected tobonding pads for testing, and, ultimately, smallquantities ofsemicustom circuits can be createdwithout masks or conventional processing. Theapproach used to demonstrate these

The Lincoln Laboratory JournaL Volume 1, Number 2 (1988)

Ehrlich et aI. - In Situ Technologies Jor Semiconductor Patterning

First-LevelLaser-Deposited Tungsten

Covered With1-J.lm Insu lator

~Second-Level

Laser-Deposited TungstenCrossing Over,

No ContactWith First Metal

Fig. 9 - The laser-direct-writing process applied to silicon integrated circuits. A focused laser spot causes a local chemicalvapor deposition (CVD) reaction, which, as the spot is scanned across the chip, defines a metal interconnection line. Newprocesses also permit multiple-level restructuring of circuits.

advantages for prototyping of CMOS gate arraysis shown in Fig. 9. Greater detail can be found inRefs. 7a and 7g.

Mask Repair by Photodeposition

In the last several years, techniques have beendeveloped for the repair of opaque and, moreimportantly, clear defects in optical photomasksby direct-write etching and deposition [7d].These methods are now well-established aslaboratory techniques and in a commercialproduct available from Quantronix (Smithtown,NY). Figure 10 illustrates the recent extension of

The Lincoln Laboratory Journal. Volume 1, Number 2 (1988)

similar methods to repair pinholes in x-raymasks.

UV-laser photodeposition of lead was adaptedto the repair of clear defects in fully fabricatedpolyimide-membrane x-ray masks [19]. Thespatial resolution of the process was exploredwith conventional x-ray exposures and wasfound to be adequate for one-step repair of 1-~mdefects, with good edge definition and withoutdamage to the fragile mask membrane. Smallerfeatures should be possible by combiningphotodeposition and an ion-milling step thatuses an FIB, a process developed by Micrion(Beverly, MA).

197

Ehrlich et aI. - In Situ Technologies Jor Semiconductor Patterning

In Situ Optimization and Trimmingby Microchemical Writing

Fig. 10 - Repair of an x-ray photomask by laserphotodeposition, a typical application of laserphotochemical technology. For this figure, a 257-nmfrequency-doubled AI'" laser deposited lead by photolysisof adsorbed tetramethyllead. X-ray exposures before (a)and after (b) repair are shown. The mask substrate is a 1­11m-thick polyimide membrane.

Several groups. including researchers atLincoln Laboratory and Lawrence LivermoreLaboratories. are developing laser­microchemical procedures for wafer-scaleintegration (21). Laser-microchemicaltechniques may offer great processingversatilityfor interconnection of complex systems.Therefore. the techniques illustrated in Fig. 9are being adapted to wafer-scale integrationapplications. In addition. a fast direct­deposition linking process has been developedfor high-density circuit structuring.

Recent Lincoln Laboratory work hasdemonstrated wafer-scale systems withcompleXities of 104 and 105 programmedswitches. by laser (22) and electron-beam (23)switching methods. respectively. Work is inprogress to combine these two beamtechnologies in more complex wafer-scalesystems.

Wafer-Scale Integration

direct-writing techniques have used physicalablation to trim. for example. poly-Si resistorson analog circuits and devices (20). Theincreased resolution and greater range ofmaterial processing of microchemicaltechniques have greatly increased the precisionofthese operations. In situoptimization offragilesubstrates is also now possible.

Figure 11 illustrates one such example: the insitu trimming of a reflective-array compressoron lithium niobate. To avoid perturbation ofdevice operation during trimming. the phaseand amplitude performance of the device mustbe adjusted in a subtle manner. This process isdesigned around low-temperature photo­chemical etching of Mo using Clz vapor.

-.\1 J.Lml.-After Deposition

-.1 1 J.L mI.-Before Deposition

(b)

(a)

Recently. direct-writing techniques have beendesigned for the optimization of expensivediscrete devices used in radar. microwave. oroptical-signal-processing applications.Because of substrate-to-substrate variation.process variability. and limits to the precision ofsimulation. such devices are typicallyincompletely optimized with multistepprocessing. To obtain higher performance.

Conclusions

Patterning is one of the key strategic issues forin situ processing. This field of research isevolVing rapidly and much progress isanticipated in the next few years. We havepointed out the strengths. stemming fromfundamental considerations. of FIB for highresolution and of lasers for high processing

198 The Lincoln Laboratory Journal. Volume 1. Number 2 (l988J

LiNbOSSubstrate

--~-

Cermet

ModifiedCermet

InputlOutputTransducers

Ehrlich et aI. - In Situ TechnologiesJor Semiconductor Patterning

488.0 - nm Beam~Reactive Ambienl)-

EtchedGrating

Fig. 11-ln situ trimming of a reflective array compressor. A laser-activated Clz etch of Mo thin films finely tailors the phaseand amplitude response of the device over a large bandwidth. The etching process provides a feature resolution of0.25 pm,affording extremely precise control of device performance.

___ -The biItcolrt-taburnroryJOi:ifuaTIolumeTNumber 2 (1988) 199

Ehrlich et aI. - In Situ Technologies for Semiconductor Patterning

rates. Whether resolution or processing rate isthe more dominant concern depends on thespecific beam-wrtting application.

Acknowledgments

The authors thank J.N. Randall for earlycontributions in masked-ion-beam lithography,and C.A. Bukowski, S.P. Doran, J.H.C.Sedlacek, and T.J. Pack for their importanttechnical contributions to the Lincoln

200

Laboratory research efforts in the areasdiscussed. The authors also thank V.S. Dolat,W.S. Graber, T.M. Lyszczarz, and J .Y. Tsao fortheir contributions to the work described and formany stimulating conversations.

This work was supported by the DefenseAdvanced Research Projects Agency, and theDepartment of the Air Force, in part under aspecific program sponsored by the Air ForceOffice of Scientific Research.

The Lincoln Laboratory Journal, Volume 1, Number 2 (1988)

Ehrlich et aI. - In Situ Technologies jor Semiconductor Patterning

References1. A synopsis of current technology appears in Proceedingsoj the NATO Workshop on Emerging Technologiesjor In SituProcessing, NATO ASI Series 139, ed. by D.J. Ehrlich andV.T. Nguyen, Dordrecht. Netherlands (1988).2. G. Miillenstedt and R Speidel, "ElektronenoptischerMikroschreiber unter ElektronenmikroskopischerArbeitskontrolle," Phys. Blatter 16, 192 (1960).3. A.N. Broers, J.M.E. Harper. and W.W. Molzen, "250-ALinewidths with PMMA Electron Resist," App!. Phys. Lett.33. 392 (1978).4. T.H. Newman, K.E. Williams, and RF.W. Pease, "HighResolution Patterning System with a Single Bore ObjectiveLens," J. Vac. Sci. Techno!. B 5.88 (1987).5. RR Kunz and T.M. Mayer, "Catalytic Growth RateEnhancement of Electron Beam Deposited Iron Films,"App!. Phys. Lett. 50. 962 (1987).6. D.J. Ehrlich and J.Y. Tsao, "A Review of Laser­Microchemical Processing," J. Vac. Sci. Techno!. B 1. 969(1983); Laser Microjabrication: Thin Film Processes andLithography, ed. by D.J. Ehrlich and J.Y. Tsao, AcademicPress, New York (to be published).7. (a) D.J. Ehrlich. J.Y. Tsao, D.J. Silversmith, J.H.C.Sedlacek, RW. Mountain, and W.S. Graber, "LaserMicrochemical Techniques for Reversible Restructuring ofGate-Array Prototype Circuits," IEEE Electron Device Lett.EDL-5. 32 (1984); (b) T.F. Deutsch, D.J. Ehrlich, D.D.Rathman, D.J. Silversmith, and RM. Osgood, "ElectricalProperties of Laser Chemically Doped Silicon," App!. Phys.Lett. 39.825 (1981); (c) D.J. Ehrlich, RM. Osgood, and T.F.Deutsch, "Laser Chemical Technique for Rapid DirectWriting of Surface Relief in Silicon," App!. Phys. Lett. 38,1018 (1981); (d) D.J. Ehrlich, RM. Osgood, D.J.Silversmith, and T.F. Deutsch, "One-Step Repair ofTransparent Defects in Hard-Surface PhotolithographicMasks via Laser Photodeposition," IEEE Electron DeviceLett. EDL-1. 101 (1980); (e) M.M. Oprysko, M.W. Beranek,D.E. Ewbank, and AC. Titus, "Repair of Clear PhotomaskDefects by Laser-Pyrolytic Deposition," SemiconductorInternational, January 1986, pg 90; (f) J.Y. Tsao and D.J.Ehrlich, "Laser-Controlled Chemical Etching ofAluminum," App!. Phys. Lett. 43, 146 (1983); (g) J.G. Black,S.P. Doran, M. Rothschild, and D.J. Ehrlich,"Supplemental Multilevel Interconnects by Laser DirectWriting: Application to GaAs Digital Integrated Circuits,"App!. Phys. Lett. 50, 1016 (1987); (h) C.L. Chen, J.G. Black,L.J. Mahoney, S.P. Doran, W.E. Courtney, D.J. Ehrlich, andRA Murphy, "GaAs Monolithic Microwave IntegratedCircuit Trimming Using Laser-Direct-Written TungstenMicrostrip Lines," Electron. Lett. 23. 402 (1987); (i) D.B.Tuckerman, "Laser-Patterned Interconnect for Thin-FilmHybrid Wafer-Scale Circuits," IEEE Electron Device Lett.EDL-8. 540 (1987).8. G.A Bums and J.A. Schoeffel, "Performance Evaluationof the ATEQ CORE-2000 Scanning Laser Reticle Writer,"Optical Microlithography VI, Proc. SPIE 772. 269 (1987).

The Lincoln Laboratory Journal. Volume 1. Number 2 (1988) .

9. See, for example, articles in J. Vac. Sci. Techno!. B 6. No.3 (1988).10. E.Miyauchi and H. Hashimoto, "Application of FocusedIon Beam Technology to Maskless Ion Implantation in aMolecular Beam Epitaxy Grown GaAs or AlGaAs EpitaxialLayer for Three-Dimensional Pattern Doping CrystalGrowth," J. Vac. Sci. Techno!. A 4.933 (1986).11. M. Rothschild and D.J. Ehrlich, "A Review of ExcimerLaser Projection Lithography," J. Vac. Sci. Techno!. B 6. 1(1988).12. J.N. Randall, D.C. Flanders, N.P. Economou, J.P.Donnelly, and E.I. Bromley, "Masked Ion Beam ResistExposure Using Grid Support Stencil Masks," J. Vac. Sci.Techno!. B 3. 58 (1985).13. S.W. Pang, T.M. Lyszczarz, C.L. Chen, J.P. Donnelly,and J.N. Randall, "Masked Ion Beam Lithography forSubmicrometer-Gate-Length Transistors," J. Vac. Sci.Techno!. B 5. 215 (1987).14. C.W. Siayman, J.L. Bartelt, and C.M. McKenna,"Masked Ion Beam Lithography for Submicrometer DeviceFabrication," Submicron Lithography, Proc. SPIE 333. 168(1982).15. G. Stengl, H. Liischner, W. Maurer, and P. Wolf, "IonProjection Lithography Machine IPLM-Ol: A New Tool forSub-0.5-Micron Modification of Materials," J. Vac. Sci.Techno!. B 4, 194 (1986).16. D.J. Ehrlich and J.Y. Tsao, "Nonreciprocal Laser­Microchemical Processing: Spatial Resolution Limits andDemonstration of 0.2~m Linewidths," App!. Phys. Lett. 44.267 (1984).17. H. Yokoyama, F. Uesugi, S. Kishida, and K. Washio,"Photothermal Effect Contribution on Film QualityImprovement in Excimer-Laser Induced Metal CVD," App!.Phys. A 37, 25 (1985); D.J. Ehrlich, J.Y. Tsao, and C.O.Bozler, J. Vac. Sci. Techno!. B 3. 1 (1985).18. S. Rice and K. Jain, "Reciprocity Behavior ofPhotoresists in Excimer Laser Lithography," IEEE Trans.Electron Devices ED-31. 1 (1984).19. J.N. Randall, D.J. Ehrlich, and J.Y. Tsao, "Repair of X­Ray Lithography Masks Using UV-Laser Photodeposition,"J. Vac. Sci. Techno!. B 3,262 (1985).20. K. Jain, "Laser Applications in SemiconductorMicrolithography," Lasers and Applications 2. 49 (Sept.1983).21. B.M. McWilliams, I.P. Herman, F. Mitlitsky, RA Hyde,and L.L. Wood, "Wafer-Scale Laser Pantography: Fab­rication of n-Metal-Oxide-Semiconductor Transistors andSmall-Scale Integrated Circuits by Direct-Write Laser­Induced Pyrolytic Reactions," App!. Phys. Lett. 43. 946(1983).22. J.I. Raffel, AH. Anderson, G.H. Chapman, K.H. Konkle,B. Mathur, AM. Soares, and P.W. Wyatt, "A Wafer-ScaleDigital Integrator Using Restructurable VLSI," IEEE Trans.Electron Devices ED-32, 479 (1985).23. D.C. Shaver, RW. Mountain, and D.J. Silversmith,"Electron-Beam Programmable 128k-Bit Wafer-ScaleEPROM," IEEE Electron Device Lett. EDL-4. 153 (1983).

201

Ehrlich et aI. - In Situ Technologies Jor Semiconductor Patterning

DANlEL J. EHRLICH re­ceived a BS degree inphysics and a PhD degree inoptics from the Universi­ty of Rochester. He joinedLincoln Laboratory in 1977,and is now Leader of the

Submicrometer Technology Group. Dan manages diverseactivities in microfabrication and microelectronicsapplications.

MORDECHAI ROTH­SCHILD is a project man­ager in the SubmicrometerTechnology Group. His re­search is centered on laserprocessing and lithogra­phy. Mordechai came to

Lincoln Laboratory from the University of SouthernCalifornia four years ago. He received a BS in physics fromBar-Ilan University and a PhD in optics from the Universityof Rochester.

202

JERRY BLACK is a staffmember in the Submicro­meter Technology Group.He came to Lincoln Lab­oratory from GTE Labora­tories four years ago. Jerry'swork is primarily focused

in the area of laser processing of microelectronic devices.He received a BA in physics and biophysics from theUniversity of Pennsylvania. and MA and PhD degrees inphysics from Harvard University. Jerry recently returnedfrom a Visiting Faculty position at Griffith University. Bris­bane, Australia.

STELLA W. PANG developsnew techniques in dryetching and ion-beamlithography. She hasworked for Lincoln Lab­oratory for seven years andis a staff member in the

Submicrometer Technology Group. Stella received a BSEEfrom Brown University and a PhD in electrical engineeringfrom Princeton University.

The Lincoln Laboratory Journal. Volume 1. Number 2 (1988)