19
`
`Chemical Engineering in the
`Processing of Electronic and Optical
`Materials: A Discussion
`
`Klavs F. Jensen
`Department of Chemical Engineering
`Massachusetts Institute of Technology
`Cambridge, Massachusetts
`
`I. Introduction
`The processing of electronic and optical materials involves scientific and
`engineering concepts from a multitude of disciplines, including chemistry,
`solid-state physics, materials science, electronics, thermodynamics, chemical
`kinetics, and transport phenomena. Chemical engineers have a long history
`of solving multidisciplinary problems in other specialized fields, such as
`food processing and polymer processing, and there is a growing recogni(cid:173)
`tion of the useful contributions that chemical engineers can make to elec(cid:173)
`tronic materials processing. Within the last decade, between 15 and 30%
`of chemical engineering graduates have taken positions in electronic ma(cid:173)
`terials processing companies, and chemical engineering research on the topic
`has grown to the point where it has become recognized in the electronic
`materials community.
`The emergence of electronic materials processing, along with other spe(cid:173)
`cialized topics within the chemical engineering discipline, raises questions
`about research, teaching, and the profession in general, questions similar
`to those asked in other areas, such as biotechnology. The aim of this dis-
`
`ADVANCF.8 IN CHEMICAL ENGINEERING, VOL. Hi
`
`Copyriaht © 1991 by Acade111ic Pr-, Inc.
`All riahta of reproduction In any form reserved.
`
`395
`
`Page 1 of 18
`
`Samsung Exhibit 1008
`
`

`

`396
`
`Klavs F. Jensen
`
`cussion is to address some of these questions and not to present an exhaustive
`review of the field. A more complete overview is given in the contribution
`to this volume by Thompson [ l] and in other recent reviews [2, 3). In
`particular, the issues to be addressed are research opportunities as well as
`undergraduate and graduate teaching. The views expressed in the follow(cid:173)
`ing are those of the author and cannot do justice to the many different
`viewpoints possible in this highly interdisciplinary research area.
`
`II. Characteristics of Electronic Materials Processing
`
`Electronic materials processing is a chemical manufacturing process aimed
`at modifying materials to form microstructures with specific electronic and
`optical properties. For example, in the manufacturing of silicon-based in(cid:173)
`tegrated circuits, silicon is refined into high-purity crystals. These are sliced
`into wafers that serve as the foundation for the electronic devices. The
`subsequent process sequences involve oxidation of the wafer surface and
`deposition of semiconductors, dielectrics, and conductors interspersed by
`patterning through lithography and etching. The final microstructure is then
`cut from the wafer and enclosed in a ceramic or polymer-based package that
`provides connections to other electronic components and protects the mi(cid:173)
`crostructure from contamination and corrosion. Similar process steps are
`used in other applications of electronic materials processing, including
`production of optical coatings, solar cells, sensors, optical devices, mag(cid:173)
`netic disks, and optical storage media. While microelectronic applications
`have typically received the most attention, many other related processes
`present equally challenging chemical engineering problems.
`Regardless of the final device, electronic materials processing involves
`a large variety of chemical procedures applied to a multitude of material
`systems. In addition to conventional chemistry, electron- and photon-dri ven
`reactions play a major role through plasma- and laser-assisted processes.
`Multiple length scales are involved in the fabrication. Typical sizes of reactors
`used for depositing and removing layers are of the order of l µm and the
`substrate wafers are 15- 20 cm across. On the other hand, in electronic devices
`the typical feature size is of the order of l µm and shrinking with each new
`generation of devices. The active region in quantum well lasers is less than
`5 nm. By way of comparison, the ability to resolve a 1-µm feature on a 20-cm
`wafer corresponds to being able to observe individual houses on a map of
`the United States. The microstructures have to be reproduced uniformly across
`each wafer as well as from wafer to wafer.
`Electronic materials processing demands high -purity starting materials.
`Even minute quantities of impurities have the potential to alter or destroy
`the electronic and optical properties of a device. Unlike many other chemical
`
`Page 2 of 18
`
`

`

`Processing of Electronic and Optical Materials
`
`397
`
`processes, the costs of starti ng materials are usually insignificant in com(cid:173)
`parison to the value added during the process. Furthermore, the field changes
`rapidly, is highly competitive, and is based on innovation, unlike commodity
`chemicals, where small improvements in large-volume processes are typi (cid:173)
`cal.
`The manufacture of even simple devices may entail more than I 00 in(cid:173)
`dividual steps. However, many of the same concepts and procedures are
`invoked several times. Therefore, it is advantageous to group the process
`steps broadl y in terms of unit operations analogous to those used success(cid:173)
`fu ll y to conceptualize, analyze, design, and operate complex chemical plants
`involving a similarly large number of chemical processes and materials.
`Examples of these unit operations are listed in Table I.
`The use of chemical engineering concepts has already contributed sig(cid:173)
`ni fican tly to crystal growth r 4], thin -film forma tion [5-7], and plasma
`
`Table 1. Examples of Unit Operations in Electronic Materials
`Processing
`
`Unit Operation
`
`Bulk Crystal Growth
`
`Chemical Modifications of
`Surfaces
`
`Thin Film Fonnation
`
`Plasma Processing
`
`Lithography
`
`Semiconductor Doping
`
`Packaging
`
`Examples of Processes
`
`Czochralski
`Bridgman
`Float zone
`
`Oxidation
`Cleaning
`Etching
`
`Liquid Phase Coating
`Physical Vapor Deposition
`Chemical Vapor Deposition
`
`Etching
`Deposition
`Spin Coating
`Exposure
`Development
`
`Solid-State Diffusion
`Jon Implantation
`
`Polymer Processing
`Ceramics Processing
`Metal I ization
`
`Page 3 of 18
`
`

`

`398
`
`Klavs F. Jensen
`
`processing [8, 9]. These unit operations involve the complex blend of trans(cid:173)
`port phenomena and chemical reactions that chemical engineers have a unique
`background for understanding and controlling. The polymer processing
`aspects of lithography (10, 11] and packaging of devices [12] as well as
`the ceramics processing problems related to packaging provide additional
`opportunities for chemical engineering research.
`
`III. Research Opportunities
`
`Research in electronic materials processing must necessarily revolve around
`one general question: How do structural, electronic, and optical proper(cid:173)
`ties of a material or a device depend on the processing and how can they
`be controlled?
`The many ways in which chemical engineers can contribute to address(cid:173)
`ing this question have been described in the Amundson report [ 13] in broad
`terms. Therefore, the present discussion will focus on three examples: {l)
`organometallic vapor-phase epitaxy of compound semiconductors, (2) plasma
`processing, and (3) process control. These examples are chosen on the basis
`of the author's experience to illustrate particular research issues rather than
`to promote specific research topics. The first example explores research
`questions with clear analogies to similar problems already solved by chemical
`engineers in the areas of heterogeneous catalysis and combustion. The second
`example introduces research issues related to the presence of charged species
`and the control of microscopic features. The third case is intended to show
`how chemical engineers could use their analysis and modeling skills in
`process control of electronic materials manufacture. In addition, a discus(cid:173)
`sion of general research trends appears at the end of this section.
`
`A. Example 1: Organometallic Vapor-Phase Epitaxy
`Organometallic (also called metalorganic) vapor-phase epitaxy (MOVPE)
`is an organometallic chemical vapor deposition (MOCVD) technique used
`to grow thin, high-purity, single-crystalline films of compound semicon(cid:173)
`ductors such as GaAs, InGaAsP, ZnSe, and HgCdTe [ 14]. These films form
`the basis for a wide range of optoelectron ic devices, including solid-state
`lasers and detectors. The technique, which is illustrated schematically in
`Fig. 1, derives its name from the fact that the film constituents are trans(cid:173)
`ported as organometallic species in the gas phase to the heated growth surface,
`where the individual metal atoms are cleaved from their organic ligands and
`incorporated into the compound semiconductor lattice. For example, GaAs
`can be grown by combining trimethylgallium and arsine according to the
`overal I reaction
`
`Page 4 of 18
`
`

`

`Processing of Electronic and Optical Materials
`
`399
`
`Gas Phase Transport Phenomena
`
`® Gas Phase Reactions
`~el
`
`Transport to Surface
`
`i
`
`0
`
`Film Precursor
`
`i Redesorption of
`
`•
`
`Desorption of
`Volatile Surface
`Reaction Products
`
`01
`
`Step Growth
`
`I ®
`
`Adsorption of Film Precursor
`
`Surface Diffusion
`• '2 CXXD
`Nucleation
`and Island
`Growth
`
`Figure 1. Schematic diagram of transport phenomena and chemical reactions underlying
`organometallic vapor-phase epitaxy.
`
`( I)
`
`Ga(CH3)3(gas) + AsH3(gas) = GaAs(solid) + 3CH4(gas)
`The technique has the flexibility to grow a multitude of compound semi(cid:173)
`conductor alloys by varying the composition of the source gas.
`Transport processes govern the extent of gas-phase reactions and the access
`of the resulting film precursors to the growth interface. As the organome(cid:173)
`tallic compounds approach the hot substrate, they react to form growth
`precursors as well as undesirable species causing unintentional doping of
`the growing film. Similarly, surface reactions participate in the film growth.
`However, parasitic reaction paths may incorporate impurities into the solid
`film, in particular carbon which is derived from the organometallic precursors.
`In the worst case, the incorporation of unintentionally added optically and
`electronically active impurities wilJ render the grown semiconductor use(cid:173)
`less for device applications. This complex mixture of chemical reactions
`and transport phenomena is well known to chemical engineers in the con(cid:173)
`text of heterogeneous catalysis, combustion, and in particular catalytic
`combustion. The same modeling and experimental approaches can in principle
`be utilized to investigate MOVPE processes. In fact, the presence of well(cid:173)
`defined, high-purity source compounds and single-crystalline substrates used
`in MOVPE means that the actual process can be studied without the need
`for model systems, which have had to be invoked to gain insight into the
`more complex heterogeneous catalytic processes.
`
`Page 5 of 18
`
`

`

`400
`
`Klavs F. Jensen
`
`1. Experimental Investigations
`Gas-phase chemical kinetics give rise to challenging experimental problems.
`The reaction intennediates may be short lived and difficult to observe. Efforts
`have been made to characterize the gas-phase chemistry of a few systems,
`in particular the growth of GaAs from trimethylgallium and arsine [ 15].
`However, most of the observations have been performed ex situ and are
`affected by the sampling method used. The relatively large size of the
`molecules, the very low concentration of the species, and missing spectral
`information complicate the application of standard optical spectroscopies
`such as laser-induced fluorescence and spontaneous Raman scattering. Thus,
`the monitoring of reactive species and their chemical kinetics present op(cid:173)
`portunities for innovative experimental approaches.
`Comparatively few studies attempt to understand the surface chemistry.
`Recent photoelastically modulated reflectance spectroscopy experiments
`indicate that surface chemistry plays a major role in MOVPE [16]. This
`technique has the advantage of observing surface changes in situ under
`standard processing conditions, but it provides no information on the ac(cid:173)
`tual surface species. There is need for a concerted surface spectroscopy effort
`similar to that used in heterogeneous catalysis to unravel surface mecha(cid:173)
`nisms. The recent work by Bendt et al. [ 17] on the surface reaction mecha(cid:173)
`nisms of triisobutylaluminum on aluminum is a good example of the
`information obtainable through such an effort. Studies of industrially rel(cid:173)
`evant compound semiconductor surfaces (e.g., (100) GaAs) will be more
`difficult because these surfaces tend to decompose at elevated temperatures
`and have several possible surface reconstructions. In addition to the com(cid:173)
`mon high-vacuum electron spectroscopy techniques (e.g., XPS, AES, LEED),
`it will be necessary to develop spectroscopies for monitoring surface chem(cid:173)
`istry during actual film growth.
`An understanding of gas-phase and surface chemistry is particularly
`important to the next generation of MOVPE processes involving selective
`epitaxy [ 18] and atomic layer epitaxy (ALE) [ 19]. In the first process, the
`compound semiconductor is deposited selectively on substrate areas opened
`in a suitable masking material (e.g., Si02). This is achieved by operating
`under conditions where nucleation occurs only on the substrates. Slight
`variations in processing environment and the presence of impurities can cause
`nucleation on the mask and result in loss of selectivity.
`In ALE, growth proceeds by the deposition of a monolayer per cycle. As
`an example, in the ALE growth of GaAs, the surface is exposed to
`trimethylgallium until the surface is saturated with a monolayer of gallium
`species (the actual species is unknown). The gallium precursor is flushed
`out and arsine is introduced, which results in the formation of a layer of
`
`Page 6 of 18
`
`

`

`Processing of Electronic and Optical Materials
`
`401
`
`GaAs. The gallium precursor is reintroduced and the cycle repeated until
`the desired film thickness is reached. This is a powerful technique allow(cid:173)
`ing, in principle, molecular engineering of the active device region on a layer(cid:173)
`by-layer basis. However, with current sources the technique works only in
`a limited range of operating conditions. Clearly, an understanding of the
`surface chemistry is critical to both selective growth and ALE.
`Similar issues arise in the nucleation and growth of one material on a
`different substrate, e.g., GaAs on Si, known as heteroepitaxy. The early stages
`of film growth must be understood to minimize defect generation and to
`realize strained microstructures with unique optical and electronic proper(cid:173)
`ties. Moreover, there may be an opportunity to use surface step growth
`mechanisms to make specialized microstructures. The effect of strain on
`the chemistry is an open area. Some of the problems could be addressed
`through the use of scanning tunneling and atomic force microscopies.
`The above discussion has stressed understanding, but innovative ap(cid:173)
`proaches to MOVPE play an equally important role. The development of
`new source chemistries can bring about improvements in selective growth
`which exceed that possible through understanding conventional source
`chemistry. Similarly, novel processing ideas may circumvent existing prob(cid:173)
`lems. For example, by providing layer-by-layer growth, ALE eliminates
`uniformity problems.
`
`2. Models
`Detailed models are necessary to understand the controlling rate processes
`underlying MOVPE, to identify critical experiments, and to provide a re(cid:173)
`lationship between process performance and growth conditions. Simple plug
`flow and continuous stirred tank reactor models with simplified chemical
`kinetics (A-- B) used traditionally in chemical reaction engineering are
`useful for quick estimates, but they cannot be justified as research topics.
`Given the rapid growth in supercomputer and engineering workstation tech(cid:173)
`nology, it is possible to formulate and solve models that give an accurate
`picture of the physical process. For the purpose of this discussion, the mod(cid:173)
`eling issues can be separated into (I) gas-phase transport phenomena, (2)
`gas phase chemical kinetics, (3) surface reactions, and (4) surface growth
`modes.
`Transport phenomena in MOVPE reactors operating at atmospheric and
`reduced pressures are affected by buoyancy-driven flows caused by large
`thermal and concentration gradients [ 15]. The buoyancy-driven flows su(cid:173)
`perimpose on the main flow to yield complex mixed-convection flows, the
`study of which provides ample opportunities for research in computational
`fluid dynamics. An understanding of the origin and nature of fully three-
`
`Page 7 of 18
`
`

`

`402
`
`Klavs F. Jensen
`
`dimensional mixed convection wou1d be useful in designing MOVPE re(cid:173)
`actors giving uniform deposition rates over large substrate areas as well as
`sharp compositional transitions between adjacent layers. Besides the tech(cid:173)
`nological applications, the fluid mechanical studies would add to the gen(cid:173)
`eral knowledge base on thermal and solutal convection as well as spur the
`development of numerical techniques for solving large, nonlinear gas flow
`problems.
`Although the experimental data base for gas-phase kinetics is far from
`complete, it is still worthwhile to formulate detailed kinetic models. These
`models form a conceptual framework for understanding and evaluating
`experimental observations. In addition, sensitivity analysis of the models
`is useful in identifying the essential portions of the chemical mechanism
`for subsequent experimental studies. Examples of this approach have been
`reported for Si [20, 21] and GaAs [22, 23] deposition, but challenging
`problems remain for key semiconductor alloys such as AlGaAs and GalnAsP.
`Models of surface chemistry and growth modes will provide needed
`additional insight into MOVPE processes. Surface kinetic models can be
`used to understand the incorporation of intentionally and unintentionally
`added impurities into the growing semiconductor. This will be essential to
`predict the electrical and optical characteristics of the deposited film. Fur(cid:173)
`thermore, it may help explain the long-range ordering observed in the growth
`of compound alloy semiconductors such as AlxGa1_xAs [24). Modeling of
`the surface growth modes (e.g., nucleation and step growth) is necessary
`to understand the development of defects and develop new growth strate(cid:173)
`gies for nanoscale structures such as quantum wire devices.
`Molecular dynamics, Monte Carlo simulations, and step growth models
`aided investigation of some of these issues in the context of molecular beam
`epitaxy (MBE) [25, 26). Molecular dynamics provides detailed microscopic
`information, but current computer technology limits it to simulations of time
`scales that are six orders of magnitude smaller than those characteristic of
`actual growth rates. Monte Carlo methods allow simulations over realis(cid:173)
`tic time scales but at the expense of requiring macroscopic parameters. Step
`dynamic simulations may be a vehicle for incorporating microscopic growth
`phenomena into macroscopic reaction-transport models of MOVPE pro(cid:173)
`cesses. The extension of the above techniques from physical vapor depo(cid:173)
`sition processes such as MBE to chemical techniques raises interesting
`research problems with important technological applications.
`Since the goal of research on MOVPE is to relate processing parameters
`to the optical and electronic properties of the grown films, chemical engi(cid:173)
`neers must collaborate with researchers from other disciplines who have the
`necessary chemistry, materials science, and device background. In addition
`
`Page 8 of 18
`
`

`

`Processing of Electronic and Optical Materials
`
`403
`
`to providing needed fundamental understanding of current practice, mod(cid:173)
`els will have to be used to develop new processing equipment.
`
`B. Example 2: Plasma Processing
`Plasma processing is used extensively to deposit and, in particular, etch thin
`films. Plasma-enhanced chemical vapor deposition allows films to be formed
`under nonequilibrium conditions and relatively low process temperatures.
`Furthermore, the films have special material properties that cannot be re(cid:173)
`alized by conventional thermally driven chemical vapor deposition processes
`[8, 9]. Plasma etching (dry processing) has almost totally replaced wet etching
`since it provides control of the shape of the microscopic etch profile [27).
`Plasmas used in microelectronics processing are weakly ionized gases com(cid:173)
`posed of electrons, ions, and neutral species, and they are also referred to
`as glow discharges. They are generated by applying an external electric field
`to the process gas at low pressures (1-500 Pa). Direct-current, radio-fre(cid:173)
`quency, and microwave sources are used. Radio frequencies between 40 kHz
`and 40 MHz dominate microelectronics applications, but microwave sources
`are of increasing interest for a range of downstream processes including poly(cid:173)
`mer etching and diamond growth (28, 9].
`Plasmas contain a mixture of high-energy, "hot" electrons ( 1- 10 e V) and
`"cold" ions and neutral species ( 400 K). The high electron energy relative
`to the low neutral species temperature makes discharges useful in chemi(cid:173)
`cal processing. Inelastic collisions between the high-energy electrons and
`neutral molecules result in, among other processes, electron impact ionization
`and molecular dissociation. The created ions, electrons, and neutral species
`participate in complex gas-phase and surface reactions leading to film etching
`or deposition. Positive-ion bombardment of surfaces in contact with the
`plasma plays a key role by modifying material properties during deposi(cid:173)
`tion and giving control of microscopic etch rate profiles. A bias potential
`may be applied to the excitation electrode to increase the ion energy and
`enhance the desired effects of ion bombardment.
`Plasma deposition and etch rates are affected by a large number of pro(cid:173)
`cess parameters and physicochemical processes, illustrated schematicalJy
`in Fig. 2, making the development and operation of plasma processes dif(cid:173)
`ficult. Moreover, given a particular process chemistry, it is not obvious how
`readily accessible parameters (e.g., feed rate, pressure, power, and frequency)
`should be manipulated to obtain the desired film uniformity and material
`properties. Glow discharge physics is complex, and the chemical mecha(cid:173)
`nisms are not well known, in particular those underlying the plasma-sur(cid:173)
`face interactions. Consequently, there is considerable incentive for gaining
`
`Page 9 of 18
`
`

`

`404
`
`Klavs F. Jensen
`
`• Pumping speed
`• Reactor pressure
`•Geometrical factors
`
`Process Parameters
`Elccnical
`• Olemical species
`• Power
`• Composition
`• Frequency
`• Bias potential •Temperature
`
`SJu:fE
`• MalCrial
`• Temperarurc
`• Morphology
`
`Electrode
`
`Plasma Chemjstzy
`• Electron density and energy distribution
`• Ion density and energy distribution •--~ Oas phase transpon phenomena
`• Generation of neutral radicals
`• Gas phase chemical kinetics
`• Plasma surface interactions
`• Surface chemical kinetics
`
`Neutral Species Chemjsay
`
`Plasma Reactor Performance
`• G rowth/Etch mechanism
`• Film properties
`• Deposition/Etch uniformity
`• Microscopic feature control
`
`Figure 2. Interactions of chemical and physical phenomena underlying plasma
`processing.
`
`insight into the underlying mechanisms through concerted experimental and
`modeling efforts.
`
`I . Experimental Investigations
`The experimental aspects of neutral plasma gas-phase chemistry are very
`similar to those discussed above in conjunction with MOYPE. The devel(cid:173)
`opment of in situ diagnostics and kinetic studies are needed to unravel the
`complex free radical-dominated gas-phase chemistry. The unique aspects
`of plasma processing stem from the transport and reactions of charged
`species. Besides providing insight into the underlying fundamentals, tech-
`
`Page 10 of 18
`
`

`

`Processing of Electronic and Optical Materials
`
`405
`
`niques for measuring electric fields and electron energy distributions would
`have a considerable impact on plasma reactor modeling and design. The
`results would allow control of the electron impact reactions that form the
`basis of the subsequent neutral chemistry. Furthermore, the electric fields
`could be manipulated to achieve special material properties in addition to
`uniformity.
`Since ion-surface and neutral molecule-surface interactions are essen(cid:173)
`tial to the directional etching and the unique material properties obtained
`in plasma reactors, they are prime research topics. These interactions have
`been explored in terms of understanding both etching and deposition sys(cid:173)
`tems [9], but many questions remain to be addressed. Examples of these
`include ion-induced surface chemistry, the role of surface damage in pro(cid:173)
`moting reactions, and the relationship between ion energy and microstruc(cid:173)
`tures. As in the case of MOVPE, it will be useful to develop in situ surface
`probes besides applying the common high-vacuum surface spectroscopy tech(cid:173)
`niques. Since the advantages of plasma processing derive from the forma(cid:173)
`tion of specific micro- or nanostructures through ion-surface interactions,
`it will be important to combine ion energy measurements [29] with struc(cid:173)
`tural probes. Application of the scanning tunneling and atomic force
`microscopies is difficult but promises exciting new insight into structural
`modifications in plasmas.
`
`2. Modeling Approaches
`Modeling of gas-phase plasma chemistry may be viewed as two interwo(cid:173)
`ven problems, which are (I) to determine the electron density and energy
`distribution and (2), given the electron impact reactions, to predict the relative
`amounts and spatial distributions of neutral species. The latter problem has
`direct implications for the uniformity of the etching or deposition process.
`Since only neutral species are involved, the modeling issues are equivalent
`to those already discussed in connection with MOVPE. Because of the
`nonequilibrium nature of the discharge, the first problem presents challenging
`research questions. There are several possible avenues of attack: solution
`of the Boltzmann equation, Monte Carlo simulations of charged species
`transport, formulation of approximate fluid models, and simple equivalent
`circuit models (8, 9].
`The last two approaches show promise in engineering applications requir(cid:173)
`ing relatively simple models. The first method is the most fundamental ,
`producing the electron distribution function as the result. However, it has
`been applied only to very simplified systems. The numerical solution of the
`Boltzmann equation for realistic plasma reactor configurations raises chal(cid:173)
`lenging computational problems even for the present generation of
`supercomputers. Given a statistically significant sample and adequate cross-
`
`Page 11 of 18
`
`

`

`406
`
`Klavs F. Jensen
`
`section data, Monte Carlo simulation of the actual electron and ion trajec(cid:173)
`tories may be used to obtain a realistic representation of the plasma. However,
`the large disparities in magnitude between electron and ion transport co(cid:173)
`efficients has so far prevented self-consistent computations. Moreover,
`extensions to three dimensions are computationally intensive. On the other
`hand, Monte Carlo techniques allow inclusion of the physical phenomena
`in a straightforward manner and the new generation of high-speed parallel
`computers may have advantages in solving this type of problem.
`Models of the surface chemistry and in particular the evolution of the film
`microstructure during film growth will be useful in understanding the re(cid:173)
`lationship between processing parameters and film properties. Because of
`the microscopic scale, the modeling approaches will range from molecu(cid:173)
`lar dynamics to Monte Carlo simulations. This relatively unexplored area
`could have a significant impact on the understanding and use of plasma
`processing.
`
`C. Example 3: Process Control of Microelectronics
`Manufacturing
`The fabrication of microelectronic and photonic components involves long
`sequences of batch chemical processes. The manufacture of advanced mi(cid:173)
`crostructures can involve more than 200 process steps and take from 2 to
`6 weeks for completion. The ultimate measure of success is the performance
`of the final circuits. The devices are highly sensitive to process variations
`and are difficult, if not impossible, to repair if a particular chemical pro(cid:173)
`cess step should fail. Furthermore, because of intense competition and rapidly
`evolving technology, the development time from layout to final product must
`be short. Therefore, process control of electronic materials processing holds
`considerable interest [30, 31 ]. The process control issues involve three levels:
`• Plantwide management
`• Materials handling
`• Unit operation control
`Because of the batch wise nature of electronic materials processing and the
`long process sequences, a supervisory system is essential. This system collects
`information on the state of the system; gives status of work in progress;
`schedules work based on process priorities, product requirement, equipment
`readiness, and materials availability; and controls product and raw mate(cid:173)
`rial inventories. Materials handling concerns the physical movement of wafers
`through the fabrication line via various mechanical means [32] and is there(cid:173)
`fore primarily a mechanical engineering problem. The control of the indi(cid:173)
`vidual unit operations and the incorporation of local control functions in
`
`Page 12 of 18
`
`

`

`Processing of Electronic and Optical Materials
`
`407
`
`plantwide control are topics that chemical engineers have addressed in the
`context of chemical plants. This experience could be applied effectively to
`electronic materials processing with appropriate consideration given to the
`rapid I y changing nature of the technology.
`The equipment used in the unit operations is complex and microproces(cid:173)
`sor controlled to allow the execution of process recipes. However, advanced
`control schemes are rarely invoked. The microprocessor adjusts set points
`according to some sequence of steps defined by the equipment manufac(cid:173)
`turer or the process operator. Flows, pressures, and temperatures are regulated
`independent} y by "off-the-shelf' proportional-integral-derivative control(cid:173)
`lers, even though the control loops interact strongly. For example, fluo(cid:173)
`rine concentration, substrate temperature, reactor pressure, and plasma power
`all influence silicon etch rates and uniformity, but they are typically con(cid:173)
`trolled independent! y.
`The process control situation could be improved by development of process
`models and monitoring techniques. The models should provide an accurate
`picture of the underlying physical and chemical rate processes while be(cid:173)
`ing sufficiently simple for on-line control strategies. There is currently
`emphasis on statistically based models. However, the resulting simple
`polynomial relationship between performance variables and process param(cid:173)
`eters is only valid over a narrow operating range. Models rooted in the
`underlying physicochemical processes will allow greater flexibility and
`extrapolation to new operating conditions. The detailed models discussed
`above in connection with MOVPE and plasma processing will require more
`computational resources than are avai lable in on-line control systems.
`However, these models could be used effectively to construct reduced-order
`models superior to simple statistical relations.
`Whether or not a chemical process step has been successful is difficult
`to measure, since there are few on-line measurable electrical properties.
`For example, film thickness and grain structure of polycrystalline silicon
`can be measured after a deposition step. However, their effect on device
`performance might not show up until subsequent doping or patterning steps
`fail. Similarly, it is possible to measure etch rates on-line by laser inter(cid:173)
`ferometry, but the etch