Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An electron beam lithography system: setup and characterization Busch, Alexander Anthony 1994

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1994-0548.pdf [ 2.37MB ]
JSON: 831-1.0074525.json
JSON-LD: 831-1.0074525-ld.json
RDF/XML (Pretty): 831-1.0074525-rdf.xml
RDF/JSON: 831-1.0074525-rdf.json
Turtle: 831-1.0074525-turtle.txt
N-Triples: 831-1.0074525-rdf-ntriples.txt
Original Record: 831-1.0074525-source.json
Full Text

Full Text

AN ELECTRON BEAM LITHOGRAPHY SYSTEM: SETUP AND CHARACTERIZATIONbyALEXANDER ANTHONY BUSCHB.Eng., McMaster University, 1992A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF APPLIED SCIENCEinTHE FACULTY OF GRADUATE STUDIESDEPARTMENT OF ENGINEERING PHYSICSWe accept this thesis as conformingto the required standardTHE U1’TIVERSITY OF BRITISH COLUMBIAOctober 1994© Alexander Anthony Busch, 1994In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.Department of PHVSiCSThe University of British ColumbiaVancouver, CanadaDate 16 Fg, . I4DE-6 (2/88)11ABSTRACTAn electron beam lithography system was realized by externally controlling a HitachiS-4 100 field emission scanning electron microscope with a computer. Associated facilities wereestablished for applying resist layers to substrates and for developing and etching exposedpatterns. Procedures for the system’s use were developed and optimized as it is anticipated thatmany researchers will use the system in the future.The system’s performance was characterized with importance being placed on thoseissues that impacted on the goal of achieving sub 50 nm resolution with high pattern uniformity.The results were found to depend on many parameters including the resist thickness, resistcomposition, development time, and the specific pattern that was written. Resolutions of—50 nmwith feature spacings of -200 rim were achieved in polymethylmethacrylate (PMMA) resistlayers —200 imi thick. Indications are that moving to thinner resist layers, shorter developmenttimes, and higher contrast developers will enable better resolution to be achieved in the future.Nonuniformities on the order of -10 nm over —5 p.m scales have been achieved on both line anddot array patterns. These patterns were emphasized because of their applications in distributedfeedback semiconductor lasers, and in the construction of artificial atom arrays. Systematic andrandom noise limitations were encountered that require further investigation to improve theuniformity beyond this level.Several patterning and processing concerns were also investigated to provide dataimportant for the design of devices that incorporate nanometre scale structures. The mostimportant issues are a limitation on the overall pattern size due to a fluctuating beam current, anda rotation effect that changes the orientation of the patterns on the substrate as the beam energyand working distance are varied.111ivTABLE OF CONTENTSAbstract iiTable of Contents ivList of Tables viList of Figures viiAcknowledgement ixChapter 1 Introduction 1Chapter 2 Principles of Electron Beam Lithography 62.1 Introduction 62.2 Lithographic Process 62.2.1 Initial Cleaning 62.2.2 Spin-on of Resist/Pre-Bake 82.2.3 Exposure Electron Microscope Fundamentals Electron Beam and Material Interactions 202.2.4 Development and Post-processing Development Post-Processing 252.3 Pattern Transfer 262.3.1 Etching 262.3.2 Lift-off 282.3.3 Strip-off of Resist 282.4 Summary 29Chapter 3 System Components: Description, Usage, and Characterization Techniques 313.1 Introduction 313.2 System Components 313.2.1 Exposure System Hardware Components 323.2.1.1 Hitachi S-4100 Field Emission SEM 323.2.1.2 Computer and NPGS Hardware 333.2.1.3 Interfacing of Computer and SEM 343.2.2 Exposure System Software Components 353.2.3 Other Processing Components 373.2.3.1 Photoresist Spinner and Oven 373.2.3.2 Glassware, Containers, and Manipulators 373.2.3.3 Chemical Storage and Disposal 383.2.3.4 SEM Sample Holder 393.3 Characterization Techniques 40V3.3.1 Sample Current Measurements 403.3.2 Resist Profile Measurements 413.3.3 Pattern Inspection Methods 41Chapter 4 System Characterization and Application 434.1 Introduction 434.2 Resolution and Uniformity 454.2.1 Pattern Independent Issues 474.2.1.1 Resist Thickness 474.2.1.2 Development Time 504.2.1.3 Noise From Internal Scan Circuitry 514.2.1.4 Working Distance 544.2.2 Dose Tests and Proximity Effects 564.2.2.1 Resolution 564.2.2.2 Uniformity 644.3 Pattern and Device Related Issues 714.3.1 Current Stability 714.3.2 Pattern Rotation 754.3.3 Field of View 784.4 Conclusions 79Chapter 5 Conclusions and Recommendations 805.1 Conclusions 805.2 Recommendations 815.2.1 Pre-exposure Processes 815.2.2 Lithography 825.2.3 Other Processing Recommendations 83References 85Appendix A Electron Beam Lithography Procedures 88A. 1 Pattern Design and Run File Creation 88A. 1.1 Pattern Design 90A.1.2 Run File Creation 91A.2 Procedure for the Application of Resist Coatings to Substrates 95A.3 Exposure of Resist Using the Electron Beam Lithography System 97Appendix B Further Details on Beam-Material Interactions and Resist Development 101B.1 Electron Scattering and Energy Loss 101B.2 Resist Modification Mechanisms 104B.3 Resist Development Model 108viLIST OF TABLESTable 2-1. Parameters affecting the lithographic patterning process. 30Table 3-1. Selected Hitachi S-4100 cold cathode field emission SEM specifications. 33Table 4-1. Statistical variations on the uniformity of two lOOnm spacing patterns. 68viiLIST OF FIGURESFigure 1-1. Block diagram of a single pass process for lithography pattern definition 4and transfer.Figure 2-1. Field emission scanning electron microscope column, corresponding to the 12column of the Hitachi S-4100 SEM.Figure 2-2. Thermionic and field emission of electrons. 15Figure 2-3. Process of gas adsorption on a field emission tip. 17Figure 2-4. Schematic illustration of scattering processes in the resist and substrate. 21Figure 2-5. Typical development curve used in determining contrast and sensitivity for 24a positive resist.Figure 2-6. Illustration of different etch profiles that typically occur when etching 27features in different directions on a GaAs substrate with an anisotropic etch.Figure 2-7. Illustration of resist profile needed for the lift-off process. 28Figure 3-1. Block diagram of the major components of the electron beam lithography 32design and exposure system.Figure 3-2. Details of the interface between the NPGS controls and the SEM through 35the connector box.Figure 4-1. Picture # 100929 shows 50 nm single pass lines written on a 1000 nm 46spacing. Picture # 100932 shows 60 nm dots written on a 200 nm spacing.Figure 4-2. Log-log plot of spin speed (rpm) versus resist layer thickness for two 49molecular weights of 4% PMMA (in chlorobenzene) at the centre of a1 cm by 1 cm GaAs substrate.Figure 4-3. Dot diameter versus dose for 400nm dot spacing written with microscope 53on TV SCAN MODE and SCAN MODE 4.Figure 4-4. Specimen current versus working distance. 55Figure 4-5. Feature size versus dose for various feature spacings. 57Figure 4-6. Plot of critical radius versus incident dose normalized to the dose necessary 59for an isolated dot to develop through the resist.viiiFigure 4-7. Exposure profiles from the line centre (0 nm) to a distance of 100 nm 60perpendicular to the line using the double gaussian resist exposure model(Chapter 2).Figure 4-8. Line dose versus line spacing showing the doses at which the lines first 61break through, when there are equal line and space widths, and when thepatterns are completely washed out.Figure 4-9. Exposure comparison test of 496K M.W. PMMA layer 175 nm thick, 63versus a 950K M.W. PMMA layer 225 nm thick, on GaAs substrate.Figure 4-10. Picture # 100930 was written on a 496K PMMA layer 175 nm thick, with 65an intended line spacing of 99.33 nrn. Picture # 100941 was written on a950K PMMA layer 225 mu thick, with an intended line spacing of 99.33 mu.Figure 4-11. Statistical variations in dot positions. 66Figure 4-12. Picture # 100840 - 496K PMMA, lines written on a 300 nrn spacing (20 im 70pattern). Picture # 100734 - dot patterns etched in a citric acid/hydrogenperoxide solution on a 300 nm spacing.Figure 4-13. Average specimen current versus time, attempting to hold the emission 72current constant.Figure 4-14. Specimen current versus time for constant extraction voltage. 73Figure 4-15. Working distance versus rotation angle between the screen horizontal 76and the micrometer stage horizontal motion.Figure 4-16. Picture # 100934 - Dot exposure test pattern (written from upper left to 77lower right). Picture # 100682 - part of a protractor pattern written to assistin characterizing pattern rotations in the SEM.Figure A-i. Lithographic dose test pattern. 89Figure A-2. Annotated sample run files, corresponding to those used for producing 94the patterns in Figure A-i.Figure B-i. Illustration of the process of chain scissioning leading to the reduction of 104molecular weight under the influence of ionizing radiation forpolymethylmethacrylate (PMMA).ixACKNOWLEDGEMENTFirst and foremost, I would like to thank my supervisor, Dr. Jeff Young, whose neverending assistance and insight were instrumental to the results achieved so far. I would like tothank him as well for the research experience he has given to me. I would also like to thank Dr.Tom Tiedje for the use of his equipment and facilities, as well as Dr. Mike Jackson who providedsome assistance in this regard as well. Special thanks go out to Christian Lavoie who taught mehow to use the electron microscope and to both him and Shane Johnson for the loan of somesubstrates early on to perform experiments with. Vighen Pacradouni also deserves a mention, forproof-reading portions of the thesis. As well, I would like to give a big thank-you to to ManojKanskar who arrived somewhat late in the work, but whose experience, knowledge andassistance contributed to immediate improvements in the results.1Chapter 1 IntroductionA great deal of research and development is being directed to control material structureon smaller and smaller length scales. In industry, this push is motivated by the desire to fit moredevices into a given area, allowing both higher device performance and lower cost. In addition,structures on smaller length scales show unique features that lead to new devices and allow oneto view properties that can not be seen on larger length scales. Current state of the art silicontechnology is based on 0.5 micron minimum feature sizes and much of the groundwork has beenlaid for introducing 0.3 micron devices within the next few years. Growth techniques such asmolecular beam epitaxy (MBE) and metal-organic chemical vapour deposition (MOCVD) allowthe material structure to be controlled down to -2 nm in one dimension, and the push to controlmaterial structure on similar length scales in two and three dimensions continues. Distributedfeedback (DFB) lasers employ gratings with pitches on the order of 0.25 microns, chosen to actas Bragg reflectors for the infrared light which they emit [Ref. 1]. Several new transportphenomena have been demonstrated in conductors fashioned into “wires” and “dots” withcharacteristic length scales less than 0.15 microns, where the structure of the device begins tomodify the quantum mechanical wavefunction of the conduction electrons [Ref. 2-3]. Furtherreductions in size are expected to lead to additional control over material properties and newapplications for these materials, such as the creation of structures with a photonic band gap[Ref. 4-5].Various techniques are being pursued to accomplish the goal of smaller feature sizes,including deep ultraviolet (UV) extensions of optical lithography, as well as electron beam, ionbeam and X-ray lithographic processes. Each of these techniques has its advantages and2disadvantages, and the “best” one depends on what one wishes to make and in what volumes.Optical lithography in the deep UV is attractive for large volume processes, as this technology isan extension of the UV technology already in place. X-ray lithography holds promise for givingthe same resolution capability as electron beam lithography, but it is still in the developmentstage. Deep UV and X-ray techniques also require masks that need to be manufactured byelectron-beam lithography. These considerations make UV and X-ray techniques better for highvolume processing. Ion beam lithography is attractive from the perspective that the ions do notpenetrate far into the substrate and scatter over a small volume, possibly providing betterresolution than electron beam systems. The ion beam currents available at this time are small,however, meaning that it will take much longer to write patterns with ion beams than by usingthe other methods.Electron beam based systems have the flexibility of allowing one to write the patterndirectly on the sample, although some systems are also being designed in an effort to imitate theoptical techniques and allow exposure through masks. The ability of beam-driven systems todirectly write on the sample makes them attractive for a research environment where new designsare constantly being tried and great flexibility is required of the system, while large throughput isnot. The lower throughput of the electron beam systems is due to three features: the beamoperation needs to be done in a vacuum, they are time limited by the beam currents that areavailable to expose the sample, and even if the beam current is high enough, the patterns aredefined using individual points, or pixels, which limits the speed with which the beam can bemoved across the sample by the microscope’s scanning circuitry.The flexibility and superior resolution of the electron beam process when compared to theoptical processes in use are precisely what is useful in our lab. We want to fabricate two3dimensionally textured optical waveguides to extend the basic principle behind DFB lasers, andarrays of “quantum” wires and dots (artificial atoms) with dimensions less than 50 nm. Recentresults indicate that these goals are achievable, with feature sizes at least as small as 10-25 nm on50 nm pitches having been demonstrated [Ref. 6-7]. Such structures represent new classes ofmaterials whose ground state and dynamical optical and electrical properties await discovery andexploitation. With these long term goals in mind, we began developing the electron beamlithography system described in this thesis. The majority of the thesis is devoted to describingthe system that has been developed and summarizing our efforts to characterize its performance.The electron beam lithography process is, in many ways, an extension of opticallithography. The basic single pass procedure is illustrated in Figure 1-1 and involves coating thesubstrate with a resist layer, exposing the resist, developing it, and transferring the resultingpattern to the underlying substrate. The substrate may be cleaned before spinning on the resistlayer. The sample is then baked and, ifmulti-layers of resist are needed, returned for anothercoating of resist. The sample is exposed by writing directly on the sample with the electronbeam. The resist layer is then developed and post-baked. Pattern transfer is accomplished byetching or lift-off and the resist layer is then stripped off the substrate. At this point, the samplecan be returned for further pattern definition when additional pattern transfer steps are necessary.The thesis is divided into three main chapters. Chapter 2 reviews the basic principlesinvolved in each aspect of the lithographic process. Chapter 3 describes the apparatus (hardwareand software) and the processes that were developed to implement the electron beam lithographycapabilities in a Hitachi S-4100 scanning electron microscope. Chapter 4 describes theprocedures used and the results obtained in characterizing the system’s performance. Thisprimarily consisted of determining the system parameters required to achieve sub 50 nm4Figure 1-1. Block diagram of a single pass process for lithography pattern definition andtransfer. Dashed lines and boxes indicate optional steps that may not be present in all processesdue to variation in materials and procedures.resolution and pattern uniformity on the order of 10 nm. Since it is anticipated that the systemwill be used by many researchers, detailed operations procedures are included in Appendix A.56Chapter 2 Principles of Electron Beam Lithography2.1 IntroductionThis chapter deals in detail with practical and theoretical aspects of the lithographicprocess. There are many general references available on the lithography process [Ref. 8-12],which is quite well understood on a semi-quantitative level. However, due to the complicatedmechanisms involved in electron scattering and polymer breakdown, an empirical approach isnecessary for developing a functional “system”.This chapter follows the lithographic process outlined in Figure 1-1: initial cleaning,spin-on of resist/pre-bake, exposure, development and post-processing, and pattern transferwhich includes etching, lift-off, and removal of the resist. Included in the exposure section is ageneral outline of electron microscope operation and a discussion of the electron scatteringprocesses. The chapter is summarized in a table listing the parameters identified in each section.2.2 Lithographic Process2.2.1 Initial CleaningInitial cleaning of samples depends on their prior processing history and the type ofcontaminant(s) present. For example, a particle on the surface of the sample will cause resistthickness variations during resist application, change the dose and development characteristics inits vicinity, and affect the transfer of the pattern to the substrate. This is especially evident when7one compares the relevant length scales. A particle may be several micrometres in size, while theresist thickness for nanolithography is typically about 200 nm at its largest, and pattern featurescan extend well below 100 nm in size.Preventing as many contaminants as possible from coming into contact with the sample isobviously preferable to cleaning it, since contamination from the cleaning process is alsopossible. Ensuring that a clean environment exists that minimizes airborne particles, that thewafer is thoroughly cleaned during processing steps such as etching, and keeping the time that asample is exposed to the room environment to a minimum helps in this regard. There may,however, be times when contamination is unavoidable due to processing considerations.How contaminants are dealt with depends on their composition. Contaminants can beseparated into three broad categories [Ref. 8, p.1 821: organic films, inorganic films, andparticulates. Examples of inorganic films that may be undesirable are unwanted oxide films,salts, and water stains, while particulates can come from many sources including humans,airborne particles, dirty equipment, storage containers, contaminated cleaning solutions, andprevious processing steps. Generally, these are removed by using etching techniques. On GaAs,sulphuric acid or ammonium hydroxide can be used to remove surface contaminants and oxideswithout etching the substrate. For deeper contamination some of the substrate may also need tobe etched away. This may not be a feasible option in some cases such as those involvingregrowth on a patterned substrate. Etching of the substrate may also increase the roughness ofthe surface of the sample.Organic films come from a variety of sources such as machinery oils or greases, peoplehandling the substrate, and polymer films from previous steps or contaminated chemicals. Twomethods of cleaning these types of contaminants are the use of a solvent-rinse system and/or8ashing the surface with an oxygen plasma (see Section 2.3.3). A solvent-rinse system mayinclude: an acetone rinse, possibly heating the acetone and doing multiple rinses followed by oneor more rinses in methanol to dissolve the remaining acetone, deionized water to dissolve themethanol and, finally, blowing the remaining fluid off the sample with a clean gas such asnitrogen. The purpose of multiple rinses is to ensure that the final rinse in a step is in as clean asolution as possible. Once cleaned, the sample is moved to the next step, the application of theresist coating.2.2.2 Spin-on of Resist/Pre-BakeApplication of thin, uniform coatings of resist is critical to future process steps. Unevenresist coatings could result in parts of the pattern developing while others remain undeveloped.As well, good adhesion of the resist to the substrate is required so that it can withstand theprocessing necessary to transfer the pattern to the substrate.The spin coating method is a widely used method for coating thin layers on planarsubstrates. It is very flexible in that the basic technique and apparatus can be used in theapplication of a variety of resists to a number of different substrates. It involves placing a fewdrops of the coating material in the centre of a sample and spinning the substrate at a speed from1000-10,000 rpm for a period of time long enough to reach a steady state in thickness and for theresist to be nearly dry. The substrate is then baked to remove the remaining solvents and torelieve stress that may build up in the film during the spinning process, due to shearing.When the resist is in a fluid state, particulates are more likely to stick when they strike thesurface. Since this occurs when applying the resist solution, spinning, and baking, particular care9must be taken to minimize exposure to particulates during these processes. Operating in as cleanand dust-free an environment as possible while applying and spinning the resist solution onto thesample minimizes this problem. For the pre-bake, the substrate is placed in a clean, glass-covered container.For a Newtonian fluid, where the volumetric flow rate change is zero, the thickness of thesteady state resist layer is expected to follow a relationship like [Ref. 8; 13](2-1)where o is the film thickness, u is the kinematic viscosity, is the rotation speed, and K is aconstant that depends on the volume fraction of resist to solvent and the substrate size. Thisequation does not account for effects such as the evaporation of the solution the resist isdissolved in, in which case the viscosity is no longer constant. K depends on the substrate sizebecause as this size decreases, the coating becomes thicker, holding all other parameters constant.For samples less than a centimetre wide, uneven coatings result as the area influenced by edgeeffects becomes comparable to the wafer area. The thickness behaviour becomes more complexwith low viscosity fluids at higher rates of rotation or for long times, but this regime has alsobeen solved theoretically [Ref. 13]. Since the viscosity increases with increasing solid percent,diluted solutions and higher spin-on speeds result in thinner coatings. Thinner coatings aredesirable to reduce the scattering of electrons in the resist which leads to broader developedlinewidths, as is discussed in the exposure and development sections below.The pre-bake is done at a temperature high enough to relieve the stress in the resist layerand to evaporate the remaining solvent, but low enough not to damage the coating. For polymersthis is typically done above the glass transition temperature, usually referred to as Tg, at which10point the material starts to flow. For polymethylmethacrylate (PMMA), a typical baketemperature is 175 °C (Appendix A). This is typically done for about 2 hours in an oven,although longer times on the order of 12 hours have been used [Ref. 14]. The reasoning behindusing a longer bake time is that it should relieve more stress in the layer and promote betteradhesion. If the stress releases during development or pattern transfer it could result in the resistshifting or detaching from the sample, which would destroy the desired features. Once the prebake is complete, the sample is ready for exposure.2.2.3 ExposureExposure of a lithographic pattern consists of two main processes: delivering the currentinto a focussed spot which is raster-scanned to create a pattern, and the interaction of theelectrons with the resist and substrate which cause the resist to change. This change is then usedto develop, or selectively remove, the resist coating, as discussed in the next section. Theelectron microscope will be discussed first, and this is followed by a description of themechanisms that cause the resist to change upon exposure to high energy electrons. Electron Microscope FundamentalsA schematic diagram of the four main components of a scanning electron microscope(SEM) is shown in Figure 2-1. These components are an electron source, a beam deflectionsystem, electromagnetic lenses, and apertures. The figure is more typical of a field emissionsystem such as the one used in the present work, than a thermal emission system. Thermal11emission systems generally have more condenser lenses in the colunm to reduce the larger initialspot size.The role of the source is obvious. The top, condenser, lens is used to provide ademagnified image of the source, approximately 1 nm in diameter, which is imaged by theobjective lens onto the target. The primary role of the objective lens is to allow flexibility in the“working distance” between the sample and the scan coils. Larger apertures called “spraydiaphragms” are placed throughout the system to collect stray electrons. The beam limitingapertures control the maximum angle, a, of the beam through the system. The angle a on thesample is shown in Figure 2-1(c). The apertures affect the beam aberrations through thedependence of the aberrations on a (spherical aberration a, chromatic aberration a, andastigmatism a).The following parameters all directly affect the quality of the features obtained using theSEM to expose the resist layer on a substrate: the electron energy, the spot size (of diameter d,on the sample), the average current at the sample, the current stability at the sample, and therotation angle between the beam scan directions and translation stage scan axes. Theoreticalconsiderations (Section and the information gained through the development processdescribed in Chapter 4 have led to the following optimal settings. The electron energy isparticularly important for limiting the range of forward scattered electrons (see Section the resist layer and so was always set at 30 keV, the maximum allowable energy for the S4100 SEM. For many applications the orientation of the pattern with respect to thecrystallographic planes of the sample is very important. It is therefore convenient to have thebeam axes coincide with the X,Y translation axes of the sample holder. This is accomplished bysetting the working distance to -19 mm (see Chapter 4). With the electron energy and working12Figure 2.- 1. (a) Field emission scanning electron microscope column, corresponding to thecolumn of the Hitachi S-4100 SEM: (1) Electron source, (2) Beam paths showing the effect ofsuccessive apertures in the column, (3) Apertures, (4) Condenser lens, (5) Deflection coils, (6)Objective lens, (7) Specimen. (b) Magnetic lens detail showing how the magnitude of the axialmagnetic field component, B, varies in the lens region. (c) Detail showing geometry of beamaperture angle, a, at the sample.(a)12345..—. ..—(b)65I7(c)13distance fixed, the only remaining control is through the aperture sizes and the extraction current.The apertures are usually set to the smallest available diameter to minimize the aberrations due tonon-paraxial trajectories in the lenses, as well as helping to form the smallest spot size. Thisdoes limit the current and, hence, increases the time required to expose a pattern. Together, thesmallest final aperture setting (20 p.m diameter in the S-4100 SEM) and working distance(19 mm) fix the aperture angle (-0.53 mrad (0.03°)). The extraction current is adjusted as highas possible to minimize exposure times while maintaining an acceptable period between tipflashes, after which the system must be allowed to stabilize for about 2 hours before it is againusable for lithographic purposes. An emission current of 10 pA has been found to be optimum.With these “standard” settings the average current and spot size on the sample along withthe stability of the current - both short term (seconds) and long term (hours) - are determinedprimarily by the electron source characteristics. The rest of this section therefore describes thetwo principle sources used in SEMs and compares their characteristics with regard tolithographic applications. This section also includes a brief explanation of why the beam scanaxes depend on the working distance.Electron sources are generally characterized by their brightness, 3, which has units ofAcm2sr1.For a given extraction voltage this becomes a constant of the system, expressed as== constant (2-2)where j is the current density in A•cm2 and a is the beam half angle at any image ofthe source,regardless ofspecific lens or aperture arrangements. The beam half angle at the specimen, a, isshown in Figure 2-1(c), for the image at the sample surface. Although the brightness is constantat all points in the system, apertures restrict the beam angle and lower the current available at the14sample. The relevant parameters at the sample are the final beam spot size d, the specimencurrent I and the beam angle at the specimen c, determined by the lens elements and usuallyrestricted by the final aperture in the system. Substituting the specimen values into Equation 2-2and using the definition of the current density, j, as current divided by area, we find that at thespecimen,4122 22S2 (2-3)n(d/2) na ndaS S S SFrom Equation 2-3 it is seen that the current, spot size and beam angle are not independentlyvariable. Fixing the spot size and beam angle with the column apertures and lenses limits theavailable beam current on the sample, for a given brightness.Electron sources typically operate in either one or a combination of thermal and fieldemission modes [Ref. 11, p.’7-20; 15, p.13-20]. These processes are illustrated schematically inFigure 2-2. For thermionic emission the electrons must overcome the workfunction, 9, of thematerial to be emitted. The value of 9 is affected by the crystal orientation of areas on the tipand the composition of any surface layer of atoms covering the tip which can raise or lower thework function. Thermionic emission sources use a tip with the temperature raised to just belowthe melting point of the material to give as many electrons as possible enough energy toovercome the work function. Emitted electrons are then accelerated down the column by thepotential difference between the tip and the other system components. Field emission occurswhen the potential gradient, caused by a strong applied electric field, is large. This reduces thepotential barrier width, b in Figure 2-2(2), so that tunnelling, which depends exponentially on b,can occur. Field emission occurs at energies close to the Fermi energy in the tip, below the15z—)Figure 2-2. Thermionic and field emission of electrons. (1) Thermionic emission of electrons,which must overcome the work function, 9, at the interface between the metal and vacuum, witha small electric field applied. (2) Field emission of electrons by quantum mechanical tunnellingthrough a potential barrier of width b. This is made possible by a much higher electric field inthe vicinity of the level of the workfunction. There is a much larger density of electrons available at theFermi energy than at the top of the workfunction, where thermal emission occurs. Two anodesare generally used in field emission microscopes to get fields of the required strength, on theorder of iO Vcm1. One anode is used to extract the current while the second accelerates ordecelerates the electrons to yield the required final beam energy.Since higher current corresponds to reduced pattern write times, having a high brightnessis desirable. Field emission sources have much higher brightness, on the order of 108 -1 oAcm2sr’ when compared to thermionic sources whose values are in the 1 0 106 A•cm2sr1range (both at Ez2OkeV) due to the large amount of current extracted in the tunnelling process[Ref. 8, p.69]. Another advantage of field emission tips is their smaller source sizes (z 100 iim)which means that fewer condenser lenses are needed to demagnii the source image to therequired small probe size. The main disadvantage of field emission sources is their greatertz—)‘9-jEIz t tEConduction Band4 Fermi EnergyValence BandMetal Source Vacuum(1)Metal Source Vacuum(2)16current noise when compared with thermionic sources. This is particularly relevant tolithography applications of the SEM, where the feature size written in the resist is often a verysensitive function of the dose of incident electrons.Thermionic sources, using a heated tip, have much more stable beam currents than thefield emission sources. The field emission tip current suffers from both short term fluctuationsand long term drift. The short term fluctuations are a result of diffusing surface atoms changingthe structure of the tip and are also due to changes in the number and type of atoms or moleculesadsorbed onto the tip [Ref. 11, p19; 16, p.3]. Long term downward drift is also noted, due to thebuild up of excess atoms on the tip surface which must then be cleaned off by flashing (running ahigher current through the tip for a short time). To reduce these effects and prevent damage tothe tip from ionized atoms in the large electric field around the field emission tips, they must beoperated in a much higher vacuum than the thermal tips. The life cycle of the tip is illustratedschematically in Figure 2-3, which shows the atoms building up on the tip resulting in the finalcondition when flashing is required. Some systems heat the field emission tip to an intermediatetemperature of about 1000 K in an attempt to avoid adsorption of atoms and molecules onto thetip and to reduce the noise associated with this. The result of this noise is a variation of exposuredose while writing patterns and it imposes strong limits on the uniformity of large patterns.Once emitted, the electron beam is focussed on the specimen by the lenses in the colunm.These components set the final spot size and beam aperture which, in turn, fix the currentthrough their relationship to the source brightness. The beam angle is limited at the same timeusing small apertures, as shown in Figure 2-1(a) & 2-1(c). Typical lenses are cylindricallysymmetric, weak field magnetic lenses with a bell-shaped magnetic field on the beam (z) axis asshown in Figure 2-1(b). The magnetic field is designed so that it is confined to a small volume17(a) (b) (c) (d)Figure 2-3. Process of gas adsorption on a field emission tip. (a) Condition of tip after desorbingatoms (flashing). (b) Unstable current portion as atoms adsorb onto tip. (c) More stable regimewhere atoms occasionally desorb from and adsorb onto the tip. A gradual decrease in theemission current is seen. (d) Multiple layers of atoms form and more adsorption and desorptionoccurs (nearing the point where tip needs to be flashed again) [Ref. 16, p.31.and all the components are kept at earth ground so that the radial and angular components ofvelocity are due entirely to the magnetic fields.For a weak magnetic lens the lens equation l/f=1/p+1/q is used [Ref. 11; 15] to describethe imaging system. The resulting magnification of the spot is given by M = q/p, wherefis thefocal length, p is the source image to lens distance and q is the lens to final image distance. If thedemagnification of the source image is large then p >>q and fq which implies M f/p. Underthe weak magnetic lens approximation, the electrons are assumed to have constant velocity andto not vary radially in the field region. In this case the focal length becomes [Ref. 11; 15]+ 00I e IB2(z)dz (2-4)f 8mQVcl+E/EG) J-0018where f is the focal length (m), V is the potential difference across the anode and cathode (V), Eis the beam energy (keV),E0=mc25llkeV, the axial component of magnetic field shown inFigure 2—i (b) is B(z) (T), e is the electron charge (1.602 X 1019 C), and m0 is the electron restmass (9.11 X l031 kg).The function of the first lens in the system, the condenser lens (shown in Figure 2-1) is todemagnify the initial spot image. The initial spot image occurs at a point called the “cross-over”,where the electron trajectories first cross-over each other after the electrons are emitted from thetip and are accelerated by the electric field in the tip region. The last lens in the system, theobjective lens, also has an effect on the spot size, but its main function is to focus the spot on thespecimen, or set the working distance. Decreasing the condenser lens magnetic field strengthresults in an increase of the magnification. At a set energy, increasing the working distance leadsto a larger magnification and an increase in the spot size at the specimen. At large workingdistances and low energies, the assumptions p.q and those of the weak lens assumption,constant velocity and radial distance in the magnetic field region, start to break down. Althoughthe magnetic field strength can be varied to compensate for changes in beam energy, this haslimits due to a particular microscope’s design. As the focal length of the lenses vary (especiallythe objective lens), the distance between the image plane of the condenser lens and the objectplane of the objective lens also varies. All of these effects will increase the beam spot size at thespecimen at large working distances.The apertures, Figure 2-1, are used to limit the beam angle and hence also affect the finalbeam current. Typically, the final aperture is used as the limiting aperture in the system and anaperture near the final lens sets the beam half-angle by ad/2f where d is the aperture diameterand f is the focal length of the final lens. Larger beam apertures, while they increase the beam19current, also increase the beam aberrations as the aberrations depend on the beam angle [Ref. 11-12; 15]. These aberrations ultimately limit the spot size. The major, but correctable, aberrationis due to beam astigmatism, or the beam focussing at different points along different axes in theplane of the specimen. This is corrected by using additional beam shaping elements in thecolumn. Spherical aberration, or the focussing of the beam at different points on the beam axis isinherent in the lens design and is not correctable. Also, electrons of different energy will focusto different places on the beam axis, an effect called chromatic aberration. Although small beamapertures decrease these effects, it is not possible to eliminate them entirely. A system isoptimized to minimize the effects of these uncorrectable aberrations and the result is that it is notpossible to arbitrarily increase the beam angle to get large current and small spot sizes at thesame time.A rotation of the source image is also introduced by the motion of the electrons throughthe magnetic fields. The rotation angle, 4, is given by the expression [Ref. 15]=8m0v(1+E/E)J(2-5)This rotation affects the lithographic process due to the necessity of placing the deflection systembefore or in the final lens. The scanning coils are also electromagnetic and the process ofscanning the beam across the sample changes the rotation equation from its value withoutscanning. This results in a net rotation between the screen (or lithography pattern) and themotion of the sample on the microscope stage. This becomes important if the pattern is to bewritten along a specified crystallographic direction for processes such as cleaving and etching.Since the rotation depends on the integral of the axial component of magnetic field and inversely20on the beam energy, changing the working distance or beam energy changes the rotation angle.There are thus several trade-offs to be considered when operating the microscope forlithographic purposes. The final current and spot size depend on the beam energy, condenserlens setting, working distance or objective lens setting, and, for field emission sources, there arenoise considerations as well. The optimum parameters are determined by considering theelectron scattering processes in the material. Electron-material interactions are the subject of thenext section. Electron Beam and Material InteractionsAfter impacting on the surface, the electron beam interacts with both the resist coatingand the underlying substrate. To understand the subsequent exposure and developmentprocesses, the nature of this interaction and the parameters it depends on are needed. As will bediscussed below, some of these parameters are: the impacting beam energy, the substrate andresist composition (atomic number, atomic weight, density), and the thickness of the materials.The mechanism by which the impacting electrons alter the resist is also important. The basicprocess will be presented here and more detail on the scattering and polymer breakdownmechanisms can be found in Appendix B.Upon entering the thin resist layers, the high energy electron beam spreads slightly dueprimarily to inelastic small angle (forward) scattering. This scattering is illustrated schematicallyin Figure 2-4(1). The result of this process is that energy is deposited in a small area that growsin diameter as the beam travels through the resist layer. After traversing a thin resist layer, theelectrons continue losing energy through inelastic scattering processes, but also experience21elastic scattering, which typically occurs at relatively large angles with a distribution dependingon the scattering cross-section (see Appendix B). Some of these electrons reenter the resist, atlower energy, where they give up more energy to the polymer layer over an extended range, onthe order of the penetration depth in the substrate [Ref. 10, p.61; 17, p43]. This process isillustrated in Figure 2-4(2). Regardless of the method by which energy is transferred to the resistlayer, the resulting material modifications are usually taken to depend only on the local energydensity deposited in the layer.(2)ResistSubstrateFigure 2-4. Schematic illustration of scattering processes in the resist and substrate. (1) Smallangle forward scattering. (2) Large angle scattering and backscattering from the substrate.Although the exact distribution of deposited energy is complex, a very usefulapproximation describes the radial dependence in the form [Ref. 8; 10; 12; 15; 17]:B 12EE (r)= —eabs 1+ 72 72EPenetration Depth(2-6)22where Eabs(r) is the area! energy density absorbed, B is a constant depending on the materials andinitial electron energy and flux, ( and (b are the half-widths of the forward and backscattereddistributions, and r E is the ratio of backscattered to forward scattered contributions to the totalenergy absorbed. These parameters are determined both empirically and by using numericalsimulations. The first term is due to the forward scattered primary electrons as they traverse theresist layer, while the second is of much larger range and is due to the diffuse, backscatteredelectrons. For Equation 2-6, reported values for a 0.5 jim film of PMMA on a silicon substrateare: (-O.O8 jim and (b-2 jim at 20 keV while at 50 keV the values are (-0.04 jim and (b9 jim[Ref. 18, p.43]. The 2OkeV silicon substrate values for E are in the range 0.5-1.1 [Ref. 15,p.31 2]. This illustrates the basic results of the scattering process, that higher beam energiesresult in smaller forward scattering ranges in the resist while the same conditions increase thebackscattered range by large amounts. These backscattered ranges are much larger than the sub-micron features of the pattern and lead to a proximity effect, where closely spaced patternfeatures contribute to the absorbed energy dose of other pattern elements.It is important to realize that the incident dose specified by the lithography controlsoftware is in terms of the charge incident on the resist layer (beam current times exposure time).This does not relate directly to Equation 2-6, and an additional conversion factor must beintroduced for this purpose.Thus, for an isolated dot exposure, the total energy deposited in the resist as a function ofradial distance from the exposure point can be expressed by replacing the constant B inEquation 2-8 by another constant dependent on the material parameters, beam energy and beamcurrent times exposure time. For a given exposure time, the developed spot size will be23approximately given by the radius at which the deposited dose falls below the critical value thatis needed for the resist to be developed.2.2.4 Development and Post-processing2.2.4.1 DevelopmentThe development process transfers the exposed pattern into the topology of the resist.Procedurally, it is similar to the initial cleaning; immersing in the developer for a specified time,followed by one or more rinses, and blowing the remaining fluid off the developed wafer with agas such as nitrogen. A resist and developer combination is typically characterized by twofigures, the sensitivity and contrast. Figure 2-5 illustrates both of these for a positive resist,which is one where the exposed areas are developed away. A negative resist is defined by theopposite effect: the exposed areas remain after developing. Further details on this may be foundin Appendix B.For a positive resist the sensitivity is given by the dose necessary for the resist to developall the way through to the substrate. For a negative resist it is usually defined as the point where50-70% of the resist in the exposed area is left. A resist with low sensitivity will not be good asit will take too much time to expose, while one with too high a sensitivity will have twoproblems. One is that it may be too volatile, with its chemistry changing uncontrollably betweenthe time it is applied and stripped off, and the second is that the exposure time may be so smallthat too few electrons are needed to expose it, leading to large statistical fluctuations.24Cl)el)0C—_1t)Figure 2-5. Typical development curve used in determining contrast and sensitivity for apositive resist.The contrast, y, is defined by the slope of the line in Figure 2-5 through1 D1____________= 1og— (2-7)° (iog —logD° ‘‘ pQs pos) pQswhere the symbols are defined in Figure 2-5. A larger contrast means that the developer is morediscriminating in its removal of molecular fragments and provides a sharper cut-off between thesolubility of higher and lower molecular weight polymers in the resist.The contrast and sensitivity of the resist depend on many factors and must be quoted withDopos posLog Dose25reference to specific molecular weight resists, exposure conditions, resist thickness, substrate,developer/rinse composition and times, the temperature, and any additional processingconditions, such as developing with ultrasonic vibration. Typical values for PMMA are asensitivity of 50-80 jiC/cm2and a contrast of about 2 [Ref. 10, p.96,227]. The contrast andsensitivity also depend on the distribution of molecular weights in the resist. Ideally, thisdistribution should be narrow, initially far above that which the developer will easily dissolveand, when exposed, entirely below the molecular weight where the developer will selectivelyremove the resist. Usually, the contrast and sensitivity are quoted for large exposed areas.Features that cover less area could require larger doses to develop entirely through to thesubstrate. Since longer development times will remove unexposed resist as well as the exposedregions, shorter development times are preferable. Another parameter that affects this is thedevelopment temperature, which increases the rate at which polymer is removed and reduces thenecessary exposure time.The important issue is that the doses used to define the sensitivity and contrastfundamentally relate to the absorbed energy dose described in the previous section, and not theincident dose used during exposure which is expressed in terms of the measurable chargeincident on the resist. Post-ProcessingPost-processing typically refers to a post-bake to remove any remaining solvents from thedevelopment process and to promote adhesion for subsequent pattern definition steps. However,some literature using thin layers of PMMA to create fine patterns on GaAs report that post-26baking was avoided so that the profiles were not degraded from thermal flow [Ref. 6].Degradation at temperatures as low as 50 °C was noted by these researchers. Instead, they placedthe sample in a desiccator for more than a day. A post-bake may therefore not be advisable forthin patterns with fine features. We have not noticed any effects due to post baking on largefeature patterns (greater than 150 nm) and have not attempted to post-bake on smaller featuredpatterns.2.3 Pattern Transfer2.3.1 EtchingThere are two main types of etching; wet chemical etching and dry etching. Only wetchemical etching will be discussed below and for further information on both types of etchingthere are several books available [Ref. 19-21].Wet chemical etching on GaAs involves two reactions. The first oxidizes the surface andthe second removes this oxide layer and is categorized according to whether a base or acid isused. Factors to be considered are how much the etchant attacks the masking layer, whether theetchant tends to smooth or roughen the surface as it etches, the safety and ease of use of thechemicals, and whether enhanced etching at the mask edges occurs. As well, whether the etchantis isotropic or anisotropic is important for some applications. Isotropic etches etch differentcrystal planes at the same rate, while anisotropic etches etch different crystal planes at differentrates. This is illustrated in Figure 2-6. In a diamond or zinc-blende crystal, the (111) planesgenerally have lower etch rates than other planes due to the higher density of atoms in these27planes. In GaAs these rates also depend on the direction from which they are approached; a Gaface is designated (11 1)A and an As face (11 1)B, with the (11 1)A planes typically having aslower etch rate. Further considerations within an etch system are the ratio of acid or base tooxidizer, and the overall concentration and temperature of the solution.Figure 2-6. Illustration of different etch profiles that typically occur when etching features indifferent directions on a GaAs substrate with an anisotropic etch.The typical method used is to have one container with etchant and one or more rinsesolutions. The advantage of multiple rinse solutions is that the one closest to the etchant can bedisposed of each time and refilled with fresh H20 at the end of the line. This ensures that the lastrinse is always uncontaminated.282.3.2 Lift-offLift-off involves patterning the substrate with an undercut resist profile and thenevaporating a layer of material on the substrate. The remaining resist is then stripped off, leavingthe material in a patterned shape on the substrate. This is used, for example, to pattern metalliccontacts on a substrate. The resist profile is important in this process, and the upper edge mustbe narrower than the lower edge in order for the metal to deposit in a way that will allow theresist to be removed. The required profile is illustrated in Figure 2-7 and shows the areas ofresist that are still exposed where the resist stripper can access and dissolve the resist and “liftoff the remaining metal.Figure 2-7. Illustration of resist profile neededfor the lift-off process.2.3.3 Strip-off of ResistRemoving the resist is typically done with a solvent, such as acetone for PMMA, that theresist will dissolve in. A similar procedure to the initial cleaning procedure is used. Anothermethod, called ashing, involves etching the surface with an oxygen plasma. This has theMetal29advantage that it leaves no residue, which the solvent process may leave on the surface. Adisadvantage is that it does not remove metallic ions well, which the wet stripping process does.However, a plasma stripping process can be followed by a wet stripping process to removemetals [Ref. 22].2.4 SummaryA large number of parameters have been identified in this chapter as having an effect onthe final pattern quality. These parameters are summarized in Table 2-1.30Process ParametersInitial Cleaning Contaminant TypeSpin-on Coating Rotation SpeedSubstrate Size/ShapeCoating CompositionCoating PropertiesPre-Bake Resist Properties (Tg)Bake TemperatureBake TimeExposure Accelerating VoltageBeam Spot SizeBeam CurrentCurrent StabilitySubstrate CompositionSubstrate ThicknessResist Properties/ThicknessResist Contrast/SensitivityPattern GeometryVacuum ConditionsDevelopment Developer PropertiesExposureDevelopment TimeTemperatureAgitation During ProcessPost-Processing Bake TemperatureBake TimeDevelopment ProcessResist PropertiesEtching Resist CompositionSubstrate CompositionEtchant CompositionTemperatureTimeStrip-off Resist Resist CompositionStripper PropertiesAgitationTemperatureTimeTable 2-1. Parameters affecting the lithographic patterning process.31Chapter 3 System Components: Description, Usage, and Characterization Techniques3.1 IntroductionThis chapter, together with Appendix A, describes the configuration and usage of all thehardware and software components of the lithography system. The components, discussed inSection 3.2, include a photoresist spinner used to apply the resist coatings onto the substrates, anoven to bake the photoresist, the chemicals and related storage and handling equipment, and theexposure system. A block diagram of the exposure system appears in Figure 3-1 below. Theexposure system consists of a field emission scanning electron microscope (SEM); a computerwith an interface board equipped with digital-to-analog (DAC) and analog-to-digital (ADC)converters; an interface between the two; and software to design and control the exposureprocess, called the Nanometer Pattern Generation System (NPGS). The methods used tocharacterize the specimen current stability, the resist thicknesses, and the topography of thedeveloped resist layers are described in Section 3.2. Section 3.3 describes the characterizationtechniques used to study the system and its application to electron beam lithography. Morespecific procedures for the lithographic process can be found in Appendix A.3.2 System ComponentsThe following subsections describe the exposure hardware and software and the samplerelated hardware.32Figure 3-1. Block diagram of the major components of the electron beam lithography design andexposure system. The programs are: DesignCAD (used to design the patterns), MRF (used tocreate the run files), NPGS, PG, and AL (used to write the patterns). More detailed explanationsof the software can be found in Section 3.2.2 and Appendix A.3.2.1 Exposure System Hardware ComponentsThe exposure hardware system consists of a S-4100 Field Emission SEM, a computer torun the software and control the exposure system, and an interface between the two. Each will bedescribed below, beginning with the electron microscope. Hitachi S-4100 Field Emission SEMThe electron microscope used in this study is a cold cathode field emission SEM, modelS-4100 from Hitachi. The microscope was modified by interrupting the signal between themicroscop&s internal XY scan circuitry and its electron beam column x-y scan controls, as willbe described in the interface section below. Operating instructions can be found in the operation33manual [Ref. 16]. Proper adjustment of astigmatism and fine focus are particularly important forattempting to obtain the best resolution with the lithography system.A summary of the microscope’s characteristics is provided in Table 3-1. Note that all ofthese features, together with the image shift, astigmatism correction, and coarse/fine focuscontrols are set by the user and not the lithographic software described below. Use of themicroscope’s controls during the pattern exposure should therefore be avoided except whereabsolutely necessary.Performance Magnification 20 to 300,000XResolution 1.5 nm (Vace = 30kV, WD = 5mm)Electron Optics Accelerating Voltage (Vace) 0.5 to 30 keV (100 V steps)Parameters Emission Current (Temm) 5 to 20 JIA (1 tA steps)Working Distance (V.TD) 5 to 35 mm (1 mm steps)Specimen Stage X, Y Directions 0 to 25 mm (2 urn steps)Z Direction (WD) 5 to 30 mm (2 im steps)Tilt -5° to 45° (0.18° steps)Rotation 360° (0.5° steps)Table 3-1. Selected Hitachi S-4100 cold cathode field emission SEM specifications. From theS-4100 operating manual [Ref. 16, p.5, 2-17, 2-33]. Computer and NPGS HardwareThe computer used to control the pattern exposure is a 80486 50 MHz IBM-ATcompatible running the MS-DOS operating system and equipped with a Data TranslationDT-2823 multi-function board. This board is equipped with 16 digital I/O lines, an internal34clock, two 16-bit analog-to-digital conversion (ADC) inputs and two 16-bit digital-to-analogconversion (DAC) outputs. The board is capable of storing output data for both DACs and thenusing this data to change both DAC output levels at the same time. This allows the X,Y rastercontrol for the microscope to have both channels simultaneously updated. The digital I/O linesare not used at this time, but could be used in conjunction with the NPGS software to controladditional hardware such as a motorized sample stage. There is also a card installed in thecomputer that can be used in the future if a beam blanker is installed in the SEM, and an adapterto connect the DT-2823 board to three BNC connectors for the X,Y scan outputs and the videosignal inputs. This adapter is also equipped with circuitry to allow adjustments of the outputvoltage ranges from the DT-2823 DACs. Interfacing of Computer and SEMThe electron microscope was easily modified to allow external control of the electronbeam scan controls. These changes are illustrated in Figure 3-2. The modifications involvedplacing a double-pole double-throw (DPDT) switch between the microscope internal scancontrols and the magnification circuitry of the microscope, which was accomplished at connectorCN-60 [Ref. 23, drawing 6. SG/VA P.C.B. (2/6)), which was available at the rear of themicroscope. BNC cables were run to allow the switch to be attached to the front panel of themicroscope. BNC cable connections were then made between the NPGS adapter box describedabove and this connector box on the rear of the microscope. The microscope provides a VIDEOOUT port on its rear panel, and this was connected directly to the NPGS adapte?s VIDEO INconnector. Finally, the computer case was grounded to the microscope rear panel.35Figure 3-2. Details of the interface between the NPGS controls and the SEM through theconnector box.The range of signal levels required at the magnification input is ±10 V. The board wascalibrated for this range using the procedure outlined in the NPGS installation guide [Ref. 24],and then the NPGS system was calibrated using the resolution standard supplied with the NPGSsystem.3.2.2 Exposure System Software ComponentsThis subsection describes the components of the NPGS software system and explainstheir relation to each other. A comprehensive description of how to use the system is containedin Appendix A.The Nanometer Pattern Generation System (NPGS) software, from J.C. NabityLithography Systems, consists of three main components. The patterns are designed using acomputer aided design (CAD) package called DesignCAD, and the files are saved in an ASCIIformat readable by the NPGS programs. The exposure conditions are then set in a programcalled Make Run File (MRF). Finally, the actual exposure is done with the NPGS program and36the two programs it calls. One program must be run to generate the pattern (PG) and the othermay be run to do alignment (AL) with respect to previously written markers.DesignCAD is a two-dimensional CAD program. The use of this program is outlined inits accompanying manuals, along with further instructions in the NPGS manual [Ref. 25; 26].The pattern is designed using absolute units of rim. Different exposures are indicated by the useof different colours in the pattern design, up to a maximum of 16. As well, different drawinglayers allow more parameters to be specified. All pattern features are defined or broken downinto line segments, whose width is specified in the CAD program. When this file is complete,the MRF program is used to specify the run parameters. The number and names of the patternsto be written, along with exposure information for each pattern is specified. Since patterns arebroken into individual exposure points, called pixels, the pixel spacings need to be specified inthe program. As well, the magnification at which the pattern is to be written is set in thisprogram.The pattern exposure is accomplished by running the NPGS or PG software. The NPGSsoftware has the advantage that it calls either PG to write the patterns or AL to align patterns asnecessary, but the PG and AL programs can also be run independently. If the run file created byMRF contains both patterns to be written and ones marked for alignment purposes then theNPGS program must be used. The NPGS and PG programs allow for an adjustment parameter tobe entered at run time, to adjust the exposure time. This is used to account for the differencebetween the current entered in the run file created by MRF and the actual measured beam currentat the time of exposure. The software also allows for the possibility of automatic stage controlsand digital control of the microscope parameters, although the S-4100 is not equipped with thesefeatures.373.2.3 Other Processing Components3.2.3.1 Photoresist Spinner and OvenA photo-resist spinner was purchased for applying the resist to substrates. It is a modelEC-1O1D photo-resist spinner manufactured by Headway Research Inc. The spinner is capableof rotation speeds from 500 to 10,000 rpm. It works by holding the substrate onto a chuck usinga vacuum and then accelerating the substrate up to a set speed. The spinning then continues for aset time and then stops, after which the vacuum to the chuck is disengaged. A pump was used toprovide the vacuum source needed for the spinner to hold the substrate on the chuck (19-25 “Hg).The vacuum chuck that the substrate sits on needs to be slightly smaller than the substratesurface. Chucks were designed and built to handle 5, 10, and 25 mm substrates using stainlesssteel and Teflon (due to concerns about the chemicals in the resists and those in the cleaningprocesses) and have a hole in their tops to provide the vacuum connection to the spinner shaft. AVWR Model 1300U oven capable of temperature ranges from 30-200 °C was used, for pre andpost bake procedures. It has a warm up period of about 1-2 hours and a temperature stability of+5 °C. Glassware, Containers, and ManipulatorsAs described in Chapter 2, the PMMA resist is above its glass temperature during thebake procedure. Accordingly, glass petri dishes were used to hold and cover the substrates toreduce the chance of particulates coming into contact with it during this procedure. The38substrates are then transferred to plastic wafer containers and, as an additional precaution againstUV radiation that can expose the resist, aluminum foil is used to cover this container. Glassbeakers and stainless steel tweezers are used to handle the substrate during the cleaning anddevelopment procedure, described in Appendix A, due to the nature of the solvents used in thisstudy. The photoresist was handled using disposable pipets, in order to reduce the likelihood ofcontaminating the equipment used during this procedure. The resist is used directly from theoriginal container, as it was found that the solvent rapidly evaporates from solutions in smallercontainers leading to thicker resist layers. These pipets were disposed of in a ‘sharps” containerdesigned for contaminated glass and other sharp objects. A source of dry nitrogen gas to blowfluids off the substrate after processing is also necessary. Chemical Storage and DisposalThe chemicals used for most of the procedures in this study are flammable liquids:acetone, methanol, isopropyl-alcohol (IPA) and methyl-isobutyl-ketone (MIBK). As well, thereis the resist solution, containing polymethylmethacrylate (PMMA) and chlorobenzene. Thefacilities where these chemicals are used require good ventilation and personal protectiveequipment must be used when handling these chemicals including gloves, lab coat, and an acidsolvent vapour mask, as appropriate. These are stored in a flammable liquids cabinet anddisposed, due to their compatibility, in the same waste container. The only exception is thewaste photoresist, which requires a separate container due to the chlorinated solvent(chlorobenzene) that it contains.393.2.3.4 SEM Sample HolderThe SEM sample holder has four purposes. It must hold the sample securely, allow fortranslations of the sample, allow the beam current to be measured, and provide space for astandard where the beam astigmatism can be corrected and fine focussing carried out on a samplewith known features. The sample holder used in this work consisted of 8 holes for holding up to4 stubs at a time around a central Faraday Cup. Samples to be patterned are attached toaluminum stubs. Initially this was done with double sided adhesive tabs, but metal clips are nowavailable for this purpose. The metal clips are preferable as the adhesive often sticks to thesample after it is removed. These stubs are then attached with screws to the sample holder. Agold standard, attached to a separate stub, is also inserted in the holder. The Faraday Cup is ametal cup (13.38 mm deep and 3.38 mm in diameter) with a 0.225 mm hole drilled in its topsurface to allow the beam to enter and has a small solid angle (2.22 X 1 0 sr) that an electronfrom the bottom of the cup can scatter into and escape from the cup. The current measured isthen proportional to the true current, with a smaller solid angle trapping more of the beam andproviding a current value closer to the true beam current. Experience with the above holder ledto the design of two other specimen holders that hold the samples flat on their surface with theastigmatism correction standard at the same level, and they also incorporate a Faraday Cup.They are designed to use metal clips to hold the sample in place and will make it easier to alignthe substrates so that they are perpendicular to one of the stage motion directions. They are alsocapable of holding a larger variety of sample sizes. They were not, however, ready in time toassist with this work.403.3 Characterization Techniques3.3.1 Sample Current MeasurementsIt is critical that the current at the sample be known before running the NPGS program, asthe exposures are based on the beam current specified in the run files. For sample currentmeasurements, the Faraday Cup described in Section is used. The sample stage in theSEM is not grounded. The current in the beam at the sample surface is measured by attaching ameter to a BNC connector on the outside of the chamber where the stage is normally grounded tothe chamber exterior. A Keithley Model 617 Electrometer, capable of measurements down to0.1 fA, was used to measure the beam current. The connections were found to be a major sourceof noise and, as a result of this, a connector and cable were purchased that are designed formeasuring currents in this range. The signals were then found to be much more stable againstmotion by the operator and the background noise in the measured signal was reduced fromfluctuations in the region of 100-200 fA to a level of 30-50 fA. The GPIB port on theElectrometer was used during the characterization of the beam current over time, under thecontrol of the same computer used for lithography. A program was written to acquire a currentreading every 2.5 seconds and to write this data to a file along with other information on themicroscop&s operating conditions. A 2.5 second time period was chosen after examining theElectromete?s operating manual and is due to its data collection and conversion method. Notethat this GPIB connection cannot be used when patterns are being written, as this compute?sresources are used entirely by the lithographic system. Should this become desirable, anothercomputer would have to be used to simultaneously write patterns and acquire current data.413.3.2 Resist Profile MeasurementsThe resist thickness was measured using a DekTak Profilometer. This equipment scans atip in one direction across a surface and measures the changes in tip height while scanning. It iscapable of measuring height changes down to a few tenths of a nanometre with a horizontalresolution on the order of the tip radius (about 12 microns). The machine uses an opticalmicroscope to position the tip, and it is difficult to find patterns with features less than about 20microns across, leading to the necessity of exposing larger areas (about 100 microns across) inorder to make this machine useful. Resist profile measurements were conducted by exposing theresist, with a dose well above its sensitivity, over an area of about 100 jim2. The resist is thendeveloped and inspected under an optical microscope to ensure it has developed through. TheDekTak tip is run over one of these surfaces and the difference between the heights of thedeveloped and undeveloped areas is then measured. Optical methods that use the refractiveindex of the resist and optical interference to find the resist thickness are discussed as arecommendation for the future (Chapter 5), as these would have the advantage of not requiringthe resist to be exposed before the thickness is measured.3.3.3 Pattern Inspection MethodsQuantitative measures of the developed resist topography played a central role indeveloping the overall system. Two methods are used for resist inspection, optical methods andthe electron microscope. A Nomarski optical microscope, with a resolution of approximately0.5 jim is used to initially examine the patterns. This provided a quick method of examining42patterns immediately after development thus providing the option of longer development times ifthe pattern was not visible. It should be noted, however, that it is not always clear whether apattern has developed all the way through to the substrate and so this method is only used as arough guide. The colour of the features can be compared with the beam dump area or an areawhere the substrate is known to be showing through to gauge whether the pattern is through theresist layer or not. For pattern features smaller than -0.5 im evidence of a pattern may be seen,but it may not be clear that the pattern is through. In both cases, however, the patterns need to beviewed in the electron microscope to quantify their profiles.The electron microscope, with its superior resolution, was used for the bulk of the patterninspections. Photographs taken with this instrument played an important role in this process andit is anticipated that the ability to directly record the images onto a computer file would alloweven more flexibility (see Recommendations, Chapter 5). Viewing patterns with the electronmicroscope exposes them more, so redeveloping after viewing them can not be done. Lessdamage is done to the patterns by viewing at an accelerating voltage of 1.0 keV as compared to30 keV, but the resolution is also poorer at the lower voltage. The breakdown of the resist leadsto two effects. One is that the area viewed becomes a negative resist, making it more difficult tostrip the resist off. The second is that the exposed areas can break down to leave a thin carbonlayer on the surface that is resistant to etching [Ref. 15, p.29]. Both of these effects have beenseen and an example is shown in the system characterization and application chapter, Chapter 4.43Chapter 4 System Characterization and Application4.1 IntroductionThis chapter explains how the system was characterized and how processes weredeveloped to achieve resolutions of less than or equal to 50 nm with uniformity on the order of10 nrn. Issues associated with the need to write large area patterns for DFB lasers are specificallyaddressed.In Chapter 2, several parameters were presented that affect the lithography process (seeTable 2-1). Due to the large number of parameters and the lack of consistent information in theliterature, it was necessary to fix some of these and then vary others to determine optimumconditions for pattern generation. Some of these parameters were fixed due to research needs orequipment limitations, some due to their relation to other processing steps, and others weredetermined after reviewing the available data on the chosen resist materials.One of the fixed parameters was the substrate, gallium arsenide (GaAs), chosen becauseof its importance in opto-electronic applications. The resist, PMMA, was selected for its highresolution, based on information in the literature [Ref. 8-101 and discussions with otherresearchers. The development/rinse system, MIBKJIPA, was adopted after discussions with theresist manufacturer and from the need to get the lithographic system up and working quickly.There were two choices for the electron beam energy. One was to use a low beam energy toreduce the backscattered electron region to as small an area as possible. The other was to use ahigh beam energy to produce the smallest spot size and penetrate as deep into the substrate aspossible, to reduce the forward scattering as much as possible and lower the amount of electrons44that backscatter into the resist. The drawback to high beam energy, as discussed earlier, is thatthis backscattered area is much larger at high beam energy than at low energy. The latter choicewas made due to the extra difficulties expected with low beam energy (very thin resists, largerspot sizes in the SEM at low beam energy), after reviewing the literature [Ref. 27-30]. Thisresulted in using high beam energy (30.0 keV), large condenser lens settings (15), and thesmallest final aperture (20 tm) to get the smallest beam size impacting on the resist layer and thedeepest penetration into the substrate.After initial trials with generic test patterns that included many types of sub-patterns, itwas recognized that the optimum conditions were so pattern specific that each pattern had to betreated separately. Two patterns were therefore adopted for further optimization; 1 -D arrays oflines for use in the fabrication of DFB laser structures and 2-D arrays of dots to produce quantumdot arrays. For the lines in DFB lasers, this requires sub 100 nm structures with excellentuniformity and spacings of 250-300 nm. The arrays needed to be more than 100 im in size to besuccessfully cleaved into lasers with the available equipment and the uniformity should beexcellent over this entire range so that specific pattern nonuniformity can be introduced into thedesigns if necessary. For the dot arrays, sub 50 rim features at sub 100 nm spacings would bedesirable, with excellent pattern uniformity.The parameters that were varied during this investigation were the development time,resist thickness, resist molecular weight (M.W.), incident dose, working distance, field of view,and array pitch. The results of the investigation of these parameters are discussed below. Theresolution and uniformity of patterns are discussed first, by addressing the issues common toboth and then presenting the results of the dose and the proximity effect tests. Pattern resolutionsof -50 nm were reproducibly achieved, and 40 nm features were obtained with less consistency.45Pattern uniformity varied due to proximity effects on the edges of patterns. As well, systematicand random variations of feature size and position, on the order of 10 nrn, were found within thepatterns. It was found that a combination of thin resist layers and short development timesprovided the most consistent results with the best resolution and uniformity. Finally, the resultsof overall process investigations are shown and their relevance to the production of devices isdiscussed.4.2 Resolution and UniformityThe results presented below were achieved by optimising several of the aboveparameters. These were the development time, resist M.W. and thickness, incident doses, andpattern spacing variations (proximity effect). The process also involved conductinginvestigations into the noise present in the system, both systematic and random. These effortsresulted in features sizes of 5 0-70 mu being obtained in both line and dot patterns with spacingsof 200 nm and above. Results for 60 nm dots on a 200 nn’i spacing and 50 nrn lines on a 1000nrn spacing are shown in Figure 4-1. These illustrate that the resolution is close to the bestreported for similar conditions [Ref. 101, but that the uniformity can be improved. This isespecially evident in the dot patterns, where the dot size can be seen to change over the area ofthe pattern. (Note that the larger features in the lower centre region of the picture are damagecaused from a region viewed at high magnification, which will be discussed later in the chapter.)Preliminary results show some 40 nm features and results on 100 mu spacings, but moreexperimentation and characterization needs to be done for these results to be consistent. The 100nm results also show some systematic nonuniformities which will be discussed in Section 4-1. Picture # 100929 shows 50 nm single pass lines written on a 1000 nm spacing.Picture # 100932 shows 60 nm dots written on a 200 nm spacing. Both patterns were written on175 mn 496K PMMA layers (the dark regions) and developed in the MIBKIIPA/H20system for90/10/10 seconds.47These results indicate that going to even thinner resists and shorter development times wouldresult in smaller feature sizes than those achieved, but experiments would need to be conductedto confirm this and to optimize the process under those changed conditions.The following discussion is separated into two subsections. The first deals with issuesthat affect the resolution and uniformity of features regardless of the specific patterns being used.The second section quantitatively assesses the resolution and uniformity obtained for 1D and 2Darrays as a function of the array pitch.4.2.1 Pattern Independent IssuesThe resolution and uniformity depend to a large degree on the same parameters, once thefixed parameters are chosen based on the criteria discussed above. The important variableparameters are the resist thickness, incident (exposure) dose, and development time. There isalso an important noise source that was traced to the microscope circuitry that has an impact onboth the resolution and uniformity. Finally, the effect of working distance will be brieflydiscussed. Resist ThicknessInitial patterns written with 496K M.W. PMMA resist layers spun on at 2500 rpm wereinconsistent and of poor resolution ( 150 nm feature sizes). Results improved immediatelywhen the resist was spun on at 5000 and 7000 rpm. A systematic study of the resist thicknessdependence on spin-speed was therefore carried out. Resist layers of both 950K M.W. and 496K48M.W. 4% PMMA were spun on to 1 cm by 1 cm GaAs substrates and baked at 170 °C for 30minutes. The central region of the wafers was exposed by adjusting the final aperture of theSEM to its largest setting and letting the microscope do a scan on its slowest mode (MODE 4)for two minutes, exposing a block approximately 100 jim square. This was calculated to give anincident dose of approximately 100-150 jiC/cm2,large enough to expose the areas completely.The beam was then switched to external control while being moved to the next area to preventlarge area exposures in the areas between patterns which would have caused the resist to thin inthese areas during development. A pattern in the form of a 2 mm by 2 mm square was generatedwith exposed blocks in each corner and in the middle of the square. The developed patterns werethen measured using a Dek-Tak Profilometer and the mean and standard deviations for the 5patterns on each wafer were calculated. The results are plotted in Figure 4-2. The 496K M.W.PMMA solution gave thinner resist layers than the 950K M.W. PMMA solutions, as expectedsince the viscosity of its solution is smaller than that of the 950K M.W. PMMA solution. The496K M.W. PMMA layers had minimum thicknesses of about 150 nm as compared to about200 nm for the layers from the 950K M.W. PMMA resist solutions. Since the maximum spinspeed of the equipment is 10 krpm, to achieve even thinner layers would require that the solutionbe diluted to less than 4% PMMA. The theoretical considerations presented in Chapter 2 givean expected 2/3 dependence of resist film thickness on spin speed. Fits to the data in Figure 4-2gave exponents of -0.49 and -0.58 for the 950K M.W. and 496K M.W. 4% PMMA curves,respectively. Since the simple theoretical treatment in Chapter 2 does not include evaporation ofthe film solvent, which would change the viscosity of the solution, or non-Newtonian fluids, andassumes no time dependence, it could be expected that this may not be entirely accurate todescribe the actual resist layer thickness profiles.49SID10001001000° 95OKPMMA4%“ 496K PMMA 4%Spin Speed (rpm)10000Figure 4-2. Log-log plot of spin speed (rpm) versus resist layer thickness for two molecularweights of 4% PMMA (in chlorobenzene) at the centre of a 1 cm by 1 cm GaAs substrate. Filmswere baked for 30 minutes at 170 °C. The solid lines are best fit curves to Equation 2-1,allowing the power to vary from its theoretical value of -2/3. This results in a power lawdependence on the spin speed of -0.49 for the 496K M.W. PMMA and -0.58 for the 950K M.W.PMMA.50The uniformity of the resist layer is also expected to have an effect on the lithographyresults. A nonuniform layer could develop through in some areas and not in others or may resultin the same exposure and development conditions producing a pattern at some times and not atothers. Although the resist layer profiles were not plotted for the whole wafer it was notedoptically that layers spun on at higher speeds looked more uniform as the colour variation acrossthe wafer was much less at higher spin speeds. A cleaning procedure involving a process ofacetone/methanol/water followed by blowing off the wafer with dry nitrogen was used. Thisprocedure, along with another nitrogen blow off after mounting the wafer (while it was spinning)was found to reduce the number of visible particulates left on the surface. Spinning on full twoinch wafers was also attempted. With these it was found that removing the wafer from itspackage and immediately mounting it on the spinner chuck, followed by a spinning blow off withnitrogen before applying the resist, gave the best results, with very few visible defects. Inaddition, smaller wafer pieces ( 7 mm) were found to be less uniform than larger ones during thespinning process. Optical inspections indicated that this was due to the edge effects covering aproportionally larger area of the wafer than in the case of the larger substrates. Finally, it wasfound that more consistent layers were achieved when using the resist from its original container,rather than from smaller storage containers. Development TimeAn ideal developer would remove all of the resist exposed beyond a critical level, leavingthe rest untouched. No real developer is ideal, and the sloped transition in Figure 2-5schematically illustrates the fact that there is a range of doses over which the developer is51discriminating. As a consequence, features can be obtained at relatively low doses by extendingthe development time. Theoretically, this is a way of minimizing the exposure time for largepatterns where current fluctuations can start affecting the uniformity. However, it was found thatlonger development times yielded poor resolution (minimum reproducible feature sizes withtolerable uniformity for 5 minute development times were -200 nm), and consequently alsolimited the minimum achievable pitch of line arrays. The features obtained with longdevelopment times also tended to have blurred edges. The results are qualitatively consistentwith those expected using the developer in a low-contrast regime. The quantitative conclusion isthat reductions in exposure time cannot be achieved while maintaining the resolution anduniformity required for DFB laser gratings. The majority of the work was subsequently carriedout using development times of 90 s, and corresponding doses (-8 fC/dot for isolated dots and0.6 nC/cm for lines) that utilize the higher contrast qualities of the developer. Noise From Internal Scan CircuitryA source of noise affecting both the pattern resolution and uniformity was encounteredand eventually traced to the microscope circuitry. It was found that pattern features written withthe microscope scan control in the TV SCAN MODE (where the microscope scans full screens inthe shortest time) made writing features on 100 nm spacings very difficult. When these patternscame through at all, their features were not well defined when they were viewed with the SEM.As well, patterns showed variations in long lines with a period that was traced to a 60 Hzvariation in the exposure conditions. When the 100 nm patterns were written with the SEM inSCAN MODE 4 (the microscope’s slowest full screen scan mode), they became well defined,52even when the patterns were overexposed and breaking apart. As a result of this, patterns weresystematically written to investigate the extent of this effect. Figure 4-3 shows a comparison oftwo identical dot patterns written with the microscope on the TV SCAN MODE and onMODE 4. It shows that the feature size is smaller for the MODE 4 patterns than for the TVSCAN MODE patterns. They both show through the resist at close to the same dose, but the sizeof the dots written using the TV SCAN MODE increase more rapidly, the difference between thedot sizes for the two modes being about 20 nm initially and increasing to about 40-50 nm. Onlythe features for a 400 nm dot spacing are shown in the figure, but a similar trend was found atother spacings as well.This has an impact on the resolution, as small dose changes just above the breakthroughthreshold cause greater changes in the feature size when using the TV SCAN MODE. It thushas an effect on the latitude in the exposure dose, giving a larger latitude in the MODE 4 patternswhere the dot size has only increased by a factor of 1.4 with a doubling of dose from 10 to 20 fCwhile the dots written with the TV SCAN MODE increase by a factor of 2.3 times their originalsize in the same dose range. As the line spacing is decreased, both the better resolution and extraexposure latitude become very important, as will be discussed in Section 4.2.2. During thepattern writing, the scan circuits in the microscope are not supposed to be coupled to the beamcontrol circuitry that the lithography system drives. However, the scan MODE 4 circuitry hasbeen designed with a low pass filter that removes 60 Hz noise while the TV SCAN MODE has ahigher cut-off. This would provide an explanation for the difference in the patterns that isobserved, given that there is some extraneous coupling. Although this would not be expected toaffect the uniformity of very large features (i.e., at very large doses), the goal of sub 50 nmfeatures on spacings of 100 nm or smaller is affected, since this source of noise affects the ability53‘ • • • • • ••1O0 -Q-.o’.• - - - o- - - d50_____TV Fast Scan Mode0’““ ScanMode40 ‘ I II10 15 20 25 30Incident Dot Dose (fC)Figure 4-3. Dot diameter versus dose for 400nm dot spacing written with microscope on TVSCAN MODE and SCAN MODE 4. These patterns were written on a 225 nm 950K PMMAlayer on GaAs, and were developed using MIBKJIPA/H2Ofor 90/10/10 seconds.54to write patterns features on small scales. It is not known at this time whether this problem alsohas an effect on the beam current as well. Working DistanceThe SEM working distance is also a parameter that was of concern. To reduce therotation effect mentioned in Chapter 2, and discussed more in Section 4.3, it was preferable towork at a larger working distance of 15-17 mm. No effects on resolution and uniformity wereseen during this study due to changing the working distance. This is likely because the beamspot size, although larger at longer working distances, is still only on the order of 10 nm at aworking distance of 15 mm. Since this is less than the smallest resolved features (40 nm)obtained at even the smallest working distance, it is not expected to be a limiting factor for theresolution and uniformity at this point. Another concern was the beam current. Although thiswould not be expected to change as the working distance is changed because the final aperture inthe S-4100 is before the objective lens, it needed to be confirmed. The results of this study arepresented in Figure 4-4. The working distance was adjusted while attempting to keep theemission current constant. The plot shows that the beam current remained fairly constant duringthis time, even though the extraction voltage changed from 4.8 to 4.9 keV when the emissioncurrent was adjusted to keep it at 10 1iA.In summary, a combination of thin resist layers ( 200 nm), larger exposure incident doses(0.6 nC/cm for isolated lines and 10-15 fC/dot for dots) and shorter development times ( 90 s)gave the best results in terms of resolution, uniformity, and consistency. In addition, thesefactors were found not to depend on the working distance at the resolutions achieved so far. An555.45.1a)L)a)I4.54.20 5 10 15 20 25 30 35Working Distance (mm)Figure 4-4. Specimen current versus working distance. Distance was varied in 5 mm incrementsfrom 20 mm down to 5mm and then from 30 mm down to 20 mm. The extraction voltagejumped for the last three data points due to the need to adjust the emission current at this point tokeep it at 10 ptA, and the black line separates the these two groups of points. The specimencurrent was measured by the method outlined in Chapter 3. Microscope conditions:Accelerating voltage = 30.0 keV, emission current = 10 piA, aperture settings 4, condenser lenssetting = 15, magnification = 100 000X.I I I I I IVet=4•9kV.....Vet=4•8kV56important feature affecting both resolution and uniformity was noise in the beam position andcurrent and improved results were obtained by writing patterns with the SEM in SCANMODE 4. Further studies need to be done, however, to confirm whether there is any lowfrequency AC noise on the beam current. The next section presents the results of studiesconducted once the above parameters had been identified and analyzed.4.2.2 Dose Tests and Proximity Effects4.2.2.1 ResolutionA generic dose test cannot be conducted, due to the proximity effect: the basic patterndesign and dimensions must be specified when quoting any doses. In an attempt to achieve thefinest resolution and uniformity possible, line and dot patterns were written on GaAs withvarying exposures to investigate the dependence of the feature size on the exposure dose. Lineand dot spacings were also varied to observe the dependence on the proximity effect.Figure 4-5 summarizes the dot and line sizes obtained for a number of different arrayspacings as a function of the incident dose. The proximity effect is manifest for spacings below-500 nm, where the array spacing has a large influence on the results. The trends are clear andexpected due to the increasing influence of adjacent exposures when the feature spacing becomescomparable to the backscattered electron range (see Equation 2-6 of Section As thespacing decreases, the features start breaking through at lower dose, and the feature size increasesat a fixed dose.Incident Line Dose (nC/cm)10 20Incident Dot Dose (fC)Figure 4-5. Feature size versus dose for various feature spacings. The patterns were written on a225nm 950K PMMA layer on a GaAs substrate and were developed in MIBKIIPA/ H20 for90/10/10 seconds. Patterns were written with the microscope on the MODE 4 setting and at1 000X magnification. (a) Single pass lines, (b) Dots on a single-pixel square grid.57I • I • I • I • I • ILine Spacing-400 nm 1000 nm600nmI I • I__0.8300I200t)1000-0.4150I1/100I° 5001.0 1.20.6 1.4 1.6I I I I II I I IGo p.tJ0000•6d•0000 Spacing (nm)--0-- 200--0’- 300400 (b)_I I I I I • • I I I i0 3058These results can be understood on a semi-quantitative level by calculating the depositeddose profiles using the “two gaussian” approximation described in Section Assuming ahigh-contrast developer, the feature size can be estimated by finding the dot radius at which thedeposited dose exceeds a critical threshold value. Results of such a simulation are shown inFigure 4-6. These were obtained with ( = 25 nm, (b = 1000 mu, and 1E = 4 on a fixed field ofview, 2 jim by 2 jim. A range of line exposure profiles are shown in Figure 4-7, correspondingto 6 jim long single-pass lines at different spacings, exposed with the same beam parameters as inFigure 4-6. In Figure 4-7, increasing dose corresponds to scaling the amplitude by a constantmultiplicative factor, and the feature size corresponds to the position at which the exposure curvecrosses a critical level (a horizontal line). Clearly the features tend to break through a fixedhorizontal threshold at lower doses as the spacing decreases. In addition, the factor by which thedose can be increased from the break through value to that where the whole pattern is exposed(when the “background” level goes above the threshold), decreases rapidly at smaller spaces.This is why the slopes of the curves in Figure 4-6 increase substantially at smaller spacings.The results, when combined with the current stability measurements reported in Section4.3.1, illustrate the limitations that may be expected when writing large patterns. A quantitativediscussion of this issue is included in Section 4.3.1, but clearly to achieve - 10% uniformity infeature size at 200 mu spacings, the current must be stable to within - 10% if one works at a dosesafely above threshold.Figure 4-8 illustrates this point in a different way by showing the points where lines firststart to come through, where there are equal line and space widths, and where the pattern featuresare washed out due to overexposure. This is done for the incident line dose as a function of theline spacing. Although these patterns were written with the microscope on the TV SCAN5950401302O1000.0 2.5Normalized Incident DoseFigure 4-6. Plot of critical radius versus incident dose normalized to the dose necessary for anisolated dot to develop through the resist. Patterns are fixed area dot exposures using thedouble gaussian resist exposure model (Chapter 2). (. = 25 nrn, Cb = 1000 nm, and 1E = 4.The numbers represent dot-spacings in mn and all plots take into account the proximity effectby summing the contributions from neighbouring features.0.5 1.0 1.5 2.0600.500.450.400.350.30CIDC0. 100Figure 4-7. Exposure profiles from the line centre (0 nm) to a distance of 100 nm perpendicularto the line using the double gaussian resist exposure model (Chapter 2). Model parameters:= 25 nm’ (b = 1000 nm, 1E = 4. The proximity effect is accounted for by summing thecontributions from the nearest features (20 for the 200 nm spacing and 4 for the 1000 nmspacing). The horizontal line represents a fixed threshold at which feature spacings start to showthrough.20 40 60 80Distance From Line Centre (nm)611.21.0I0.8C)0.6C)C) 300 600Line Spacing (nm)Figure 4-8. Line dose versus line spacing showing the doses at which the lines first breakthrough, when there are equal line and space widths, and when the patterns are completelywashed out. Patterns were written on a 225 nm 496K PMMA layer on GaAs (n+ Si doped) withthe microscope on TV SCAN MODE, and they were developed using MIBK!IPA/1E120for5/5/0.25 minutes.I • I • I • I • IEqual Lines and SpacesPattern Washed OutI • I • I •Pattern Not Through100 200 400 50062MODE, similar effects with a wider dose latitude are expected for those written on SCANMODE 4. At a line spacing of 100 nm, the exposure goes from patterns that are not through theresist layer to ones that are overexposed completely in a dose range of less than 0.025 nC/cm.Since variations in the current of only a few percent can easily take the exposure out of thisrange, it makes it much more difficult to expose patterns at these small spacings.The results of this study are that a limit of 50 nrn has been reached for line and dotpatterns at doses of 0.5 nC/cm for the lines and 5 fC for the dots on a spacing of 200 nm in resistof -200 mn thickness. This may be improved by going to thinner resist layers, as the followingresult suggests.Normally, one would expect that the resolution would be better for a higher molecularweight (M.W.) resist when compared to a lower M.W. resist. It is expected that for the sameresist layer thickness and dose the lower M.W. resist will have its M.W. distribution shiftedlower than the distribution of the higher M. W. resist in the patterned area. Since lower molecularweights are more soluble than higher ones in the developer, the pattern features should be largerin the lower M.W. resist patten. From Chapter 2, a thinner layer would have less forwardscattering in the resist layer and so feature sizes would be expected to improve if all otherparameters are held constant. Figure 4-9 shows a comparison of dot and line patterns on a 496KM.W. PMMA layer 175 nm thick with the same patterns on a 950K M.W. PMMA layer 225 nmthick. Both patterns were developed for the same amount of time (MIBKIIPAIH2Ofor 90/10/10seconds). The dot patterns are both on 400 nm spacings and the line patterns on 300 rimspacings. Although the 496K M.W. PMMA layer dots develop through at a lower dose, aswould be expected, the feature sizes in this thinner layer follow closely the dot diameters in the950K M.W. PMMA layer once it starts to develop through. A similar trend is seen in the line630 496K PMMA, 175 nm thick0 95OKPMMA,225nmthick5 10 - - 15 20 25 30 35Incident Dot Dose (fC)—I • Q —0 496K PMMA, 175 nm thick0 950K PMMA, 225 nm thick 0000000- 0• 000(b)00• I I I0.5 1.0 1.5Incident Line Dose (nC/cm)Figure 4-9. Exposure comparison test of 496K M.W. PMMA 175 nm versus 950K M.W,PMMA 225 nm layers, on GaAs substrate. (a) Dot pattern, 400 nm spacing. (b) Line Pattern,300 nm spacing. Both patterns were developed in MIBKi’IPA!H20for 90/10/10 s.I I I I I I I I I I I • • I I I I I I I I I I • • •160_____________________________0140120 0 0ODD100 CD80 DOD00(a)60 0 0000• 0040 • • • • I I I I I I a I • I . . . . I a i i i • i I0300_ ____ ____ ____,250____ ___E200I1501005064patterns, although here the variation in the starting dose is much less. Thus, thinning the 496KM.W. PMMA compensates for its intrinsic resolution limitations (compared to those of 950KM.W. PMMA). Further thinning is expected to result in further improvements in the resolution. UniformityUniformity is crucial in many applications of finely patterned structures. Variations inthe line/space ratio (duty cycle) of DFB gratings leads to inhomogeneous coupling along thelength of the grating that can degrade the linearity and linewidth of the laser. Advantages in laserthreshold reduction that might be realized by the increased density of states in quantum dotactive layers will be lost if different dots have different energy levels due to inhomogeneities.Uniformity must be quantified on different lateral length scales. Here we limit ourselvesto scales from the inter-feature spacing up to 5 jim, limited by the resolution of the SEMphotographs from which the measurements were taken.The uniformity was quantified on length scales up to 2 jim from the photographs ofFigure 4-10 and the corresponding data presented in Figure 4-11. Table 4-1 summarizes the datafrom the studies of these patterns. The two patterns have 100 nni line spacings and are written,with respect to the photographs, from left to right and from the top to the bottom of the pictures.They show two types of uniformity problems: variations in the size of the dots and variations inthe spacing.Measurements of samples on the photograph were made for 5 rows and 10 columns ofhorizontal and vertical spacings on both samples, as well as vertical and horizontal measurementson the sizes of the dots in the same region. The averages and standard deviations were calculated65-. -100941 30.0kvFigure 4-10. Picture # 100930 was written on a 496K PMMA layer 175 nm thick, with anintended line spacing of 99.33 nm, an exposure time per point of 157.2 jis and an incident doseof 2.0 fC. Picture # 100941 was written on a 950K PMMA layer 225 nm thick, with an intendedline spacing of 99.33 nm, an exposure time per point of 156.4 is and an incident dose of 2.5 fC.Both patterns are 20 lim squares and were developed using MIBKIIPA/H20for 90/10/10 s.2066rj 150-e-e0I I I IA7—_V-Ez — — — — — — —ELzo 100941 - Horizonta-±1— 100941 - Vertical100930-Horizonta— 100930 - VerticalI I I I I(a)1 2 3 4 5Row NumberIHorizontal VariationsI I II10E00-20I‘2010E0(00-20I I I I I I I I I II I I- Vertical Variations1 2 3 4 5 6 7 8 9 10Column NumberFigure 4-11. Statistical variations in dot positions. (a) Column standard deviations from meanspacing corresponding to the samples shown in Figure 4-9. (b) Horizontal variations frommean position for dots from 950K PMMA sample (100941) in Figure 4-9. (c) Verticalvariations in dot position for same sample as in (b). For (b) and (c), different symbolscorrespond to data points from different rows in the pattern.67and the results are presented in Figure 4-11 and Table 4-1. Table 4-1 shows that the deviationsin the dots sizes are on the order of 4-7 nm. This could be due both to variations in current as thepattern is writing and/or small scale noise in the position of the beam during exposure. It couldalso be due to variations in resist thickness over small distances and requires further investigationfor the cause to be conclusively identified.Figure 4-1 1 shows the variations from the mean for both the horizontal and verticaldirections plotted against the column number for the 950K M.W. sample, and the standarddeviations for each row on both samples. The horizontal and vertical deviations show strongcorrelations between rows, indicating that these are systematic errors. The plot of standarddeviation versus row number shows an average of about 10-12 nm, also listed in Table 4-1. Thesample set is small, however, and differences between them would be more significant with theuse of larger data sets. However, at the field of view that these patterns were written, 98.63 jim,the smallest voltage resolution from the controlling computer’s digital-to-analog board gives asmallest pixel spacing of 4.5 mu when the nonlinearity of the DAC is considered. Thisnonlinearity results in voltage outputs that can vary slightly from the intended value,corresponding to a loss in resolution from the board of about 1.5 bits. This should be a stableoutput error, which is consistent with the systematic trend. Noise problems due to other circuitryin the computer can also lead to further degradation of the signal [Ref. 31, p.614-615], reducingthe resolution to about 14-bits, or 10 mu at this magnification, consistent with the results in Table4-1. This leads to the conclusion that these deviations may be due to errors from the D/A systemand/or degradation of the signal from the computer, and this will require more investigation.Even assuming no voltage errors in the DAC output, there is a limit of 16-bits resolution from theDAC which, at 1000X magnification (98.63 jim field of view) represents a limit of 1.5 mu.68Reduced variations may be possible by writing the same patterns at higher magnification.If the variations are due to the NPGS system, which is likely, the microscope should scale themwhen the magnification is increased and the systematic variations should therefore be reduced.Random fluctuations, which from Figure 4-11 appear to be on the order of 5 nm rms, would notbe subject to the same scaling factor if they enter the system after the point at which the interfaceto the microscope is reached. Thus, it is possible that more precise positioning may need to bedone at higher magnifications - leading to smaller overall pattern sizes.Resist Dot Size H/W Spacing Pattern 0H Pattern o,Layer (nm) (nm) (nm) (nm) (nm)950K 39±4 1.0±0.1 94±8 8 11496K 52±7 1.0±0.1 90± 10 14 14Table 4-1. Statistical variations on the uniformity of two lOOnm spacing patterns. 496K datafrom Figure 4-9, picture 100930. 950K data from Figure 4-9 picture 100941. The patternstandard deviations are over the whole data set. H stands for horizontal and V for vertical.Larger scales out to -5 im were measured in a dot pattern on a 200 nm spacing written atthe same time as those in Figure 4-10. The deviations in pattern feature size and spacing werealso measured to be - 10 nm. Pattern feature deviations on the line pattern in Picture 100480 ofFigure 4-12 were measured to be - 10-20 nm.Two other features are important considerations for the uniformity. These are theconsequences of the proximity effect and random defects. Examples of these are shown inFigure 4-12. The first picture shows the edge of a pattern. The proximity effect can be seen atthe left of the pattern, as the resist lines are thicker (about 190 nm wide) and the pattern is notquite through to the substrate (shown by the dark areas). As one moves into the pattern (to the69right in the picture), the resist lines become thinner (about 130 nm wide). This is due to the linesin the centre of the pattern having more neighbouring lines contributing to their absorbed dosethrough the proximity effect than lines at the edge of the pattern. If extremely high uniformity isrequired then this effect needs to be compensated by giving the outside regions higher exposures.The second picture in Figure 4-12 has some missing pattern features. It shows a patternof dots that were written and etched in GaAs. The etch was a citric acid/hydrogen peroxide etchconsisting of 10 ml 50% (mass percent) citric acid, 1 ml 30% H20,and 33 ml of deionizedwater. It was etched for 2 minutes at 23 °C. Note that significant numbers of the dots, whichappeared to be fine when viewing the resist under the SEM, did not etch through to the substrate.This is especially apparent at the edges and is most likely due to the proximity effect in theseareas in a similar manner to the lines discussed above, as test patterns on the same wafer with alower incident dose had even larger numbers of missing dots on the edge of the pattern.Dots in the middle of the pattern are also missing. At present, the source of this is notknown, but there are three likely causes. One is that the writing system is missing these dots, thesecond could be due to fluctuations in the resist thickness, and the third could be contaminantsthat are reducing the incident dose in these areas. A method of investigating this in the future isto write multiple versions of an identical dot pattern. Random fluctuations due to resist thicknessvariations and contaminants should result in dot features shifting from pattern to pattern, whilesystematic variations would be expected to remain. Finally, note the two black rectanglessurrounded by non-etched flat areas. These are where the pattern was viewed to characterize thedot sizes in the resist before etching. This viewing modified the resist in this area, reducing itssolubility so it was not removed by the stripper, indicating that it received so much exposuredose in this region that the resist became negative there. The areas surrounding these rectangles70Figure 4-12. Picture # 100480 - 496K PMMA, lines written on a 300 nm spacing (20 pmpattern). The dark lines on the left edge of the pattern are not all the way through to the substrateand are caused by damage during the viewing process. Picture # 100734 - dot patterns etched ina citric acid/hydrogen peroxide solution on a 300 nm spacing. Note that this sample is tilted at22.5° with the top side of the picture away from the viewer.71that did not etch are likely to have suffered from the breakdown of the organic resist locally,leaving a carbon layer that would resist the etch.4.3 Pattern and Device Related IssuesAs well as resolution and uniformity, there are several concerns that relate more stronglyto the pattern itself and further processing issues. The main areas of concern are the behaviour ofthe beam current over time, the pattern size that can be written, and the orientation of the patternon the substrate.4.3.1 Current StabilityThe stability of the beam current over long time scales becomes important as the patternsize increases. Drift in the current causes parts of the pattern to be exposed while others are notand will result in poorer uniformity as discussed previously. A study of the beam specimencurrent as a function of time was conducted and the results are presented in Figures 4-13 and4-14. In Figure 4-13, the emission current was held constant by adjusting it whenever it changedfrom 10 pA to a value of 9 1iA. This was done from the point where the tip was flashed to thepoint where the microscope shut off the beam and indicated that the tip needed flashing again.The graph shows the value of the extraction voltage as well. It indicates that the early stages aremore unstable and require frequent increases in the extraction voltage to keep the emissioncurrent stable. It is thus better to write later in the cycle when the beam is more stable. Thevariation in the beam current is large, which can lead to a wide variation in exposed pattern1211‘—‘10c,)32Time(hours)J._,)5.04.5> C4.03.5Figure4-13.Averagespecimencurrent versustime,attemptingtoholdtheemissioncurrent constant.3datapointsweretakenevery30sandaveraged.ThisdatadidnotusetheprogramdescribedinChapter 3.Microscopesettings:Acceleratingvoltage=30.0keV,workingdistance=15mm,emissioncurrent =10iA, aperturesettings=4,condenserlenssetting=15, magnification=300000X.01234567813 12 11 0 6 5 4 3 2 1 00.0Time(hours)Figure4-14.Specimencurrentversustimeforconstant extractionvoltage.Datapointsweretakenevery2.5susingtheprogramdescribedinChapter3.MicroscopesettingsarethesameasforFigure4.—.174features due to the varying dose that the fluctuating current causes. Due to this variation, anothermethod was sought to try to get a more stable signal.In the second method the extraction voltage was held constant, letting the emissioncurrent drop over time. The data was taken for 3.8 keV, then the extraction current was adjustedto bring it back up to the set value (10 1iA). At this point the extraction voltage moved upwardsand another run was done, with the extraction voltage now at 4.8 keV. This was repeated oncemore with the extraction voltage now having a value of 5.1 keV. This data is shown inFigure 4-14. The data indicates that writing in the latter stages of the tip cycle provides a morestable long term signal. However, the short term current becomes more unstable as the extractionvoltage increases. The colunm was then baked and data taken after this had less short term noisethan the data in Figure 4-13, but had the same long term current drift characteristics.The maximum write time from Figure 4-14 is about 1 hour when remaining within 20%of the start current. As the current drops, the pattern features would receive less dose, whichwould cause variation in the feature size. A pattern consisting of a 180 urn square array of200 nm lines with centre-to-centre spacings of 20 nm, an incident line dose of 0.6 nC/cm, and astarting current specified in the lithography run files of 10 pA has an exposure time of 16.2minutes. Referring to Figure 4-14, if one is writing in the later part of the cycle at an extractionvoltage of 4.8 keV then the specimen current would have dropped by -10% in this time.Referring back to Figure 4-5, fluctuations in the beam current of 10% would cause patternfeatures to change in size by -10%. If the pattern was written with the extraction voltage at5.1 keV the shorter term fluctuations during this time are worse and are - 10%. One way toattempt to compensate for this in the pattern design could be to write pattern features such assingle pass lines with multiple passes of the beam over the same lines. For example, an array of75vertical lines could be written first left to right and then right to left in an attempt to even out thedose. This would, however, present other possible problems such as drift of the pattern elements,i.e., the lines may not write over each other with each pass through the pattern. The dataindicates that it would be better to write large patterns with closely spaced features with a morestable beam source, such as a thermal emission tip, so that the current would remain constantover longer time scales. The trade-offs are that the field emission sources provide smaller spotsizes and larger beam currents than the thermal sources, as discussed in Chapter Pattern RotationAs discussed in Chapter 2, there will be an angle between the motion of the sample,accomplished with the micrometer stage, and the direction the pattern is written. This effectdepends on the magnetic fields in the objective lens and scanning coils that the scanning beampasses through, and so the rotation effect depends on the working distance. Since cleavingoccurs preferentially in the (011)and (01 1)directions on (001)GaAs substrates, this rotationbecomes important when attempting to create structures that must be oriented with respect tocleaving planes. In addition, if etching is anisotropic the pattern may again need to be writtenwith a certain orientation on the substrate. Accordingly, the difference between screen and stagemotion was characterized for future work in this area. This data is shown in Figure 4-15 where,at an accelerating voltage of 30.0 keV, the zero point was found to be 19 mm, while at 5 mm theangle was -23°.The first photograph in Figure 4-16 consists of a pattern that was written at 15 mm76Working Distance (mm)Figure 4-15. Working distance versus rotation angle between the screen horizontal and themicrometer stage horizontal motion. Measurements were taken by off the screen and from SEMphotographs, and were taken by translating a fixed feature from the bottom left of the screen tothe right of the screen and measuring the angle between the screen horizontal and the stagetranslation line found from connecting the point on the right of the screen to the original positionat the bottom left of the screen. The inset indicates the direction of the screen and stage axes andhow the angle between them is defined.302520‘;;‘ 15‘—i 1050-5-10I Io\ Screen0Stage Micrometer Directions00000000o_____000008I I I0 From Screen MeasurementsFrom Photographs0 5 10 15 20 25 3077Figure 4-16. Picture # 100934 - Dot exposure test pattern (written from upper left to lowerright). The line across the picture is from translating between patterns and the large white circleis the beam dump. Note that there is an angle between the stage translation direction and thepattern horizontal (-3°). Picture # 100682 - part of a protractor pattern written to assist incharacterizing pattern rotations in the SEM.78working distance and shows the angle between the pattern axes and translation axes. It alsoshows the beam dump where the beam sat before and after the pattern was written and the linefrom translating between patterns. As well, Figure 4-16 shows part of a 3600 protractor patternthat can be used to help characterize the rotation angle. It is translated horizontally across thescreen and the angle can then be read off the protractor. As mentioned previously, early workwas accomplished at shorter working distances of about 8 mm while later work was conducted at15 mm working distance in an attempt to reduce the angle of rotation between the screen andstage axes. No fundamental resolution problems have been noted at this time, but the beam spotsize at 30 keV and with the apertures on their smallest settings would be about 10 nm or less at15 mm. As a result, it is not expected that this would be the fundamental limiting factor at thistime, since the minimum features are currently on the order of 50 nnt4.3.3 Field of ViewAs well as rotation, the total field of view available is an important consideration. Forease of processing, especially when having to cleave around a pattern, as large a field of view aspossible is preferred. At a magnification of 1000X, the field of view is 98.63 rim. As discussedin the first part of this chapter, the systematic uniformity problem in the patterns is likely due tothe lithography patterning system’s digital-to-analog board and the transfer of this signal to themicroscope. This nonuniformity in the pattern spacing would limit the field-of-view that can beused. This would place limits on the usefulness of large patterns. For applications like DFBlasers and quantum dot arrays, for example, the nonuniformity would eventually impact on theperformance of the devices. This would places limits on the sizes of such devices, and the only79way to increase the device size would be to write patterns and align them with respect to eachother. This would be much easier with a beam blanker in the system, in which case the exposedbeam dump areas and translation axis lines would not exist. As shown in Figure 4-16, this lackof a beam blanker means that before and after the pattern is written the beam must sit in a placeoff the pattern which has been found to expose an area up to about 6 p.m in diameter. These extraexposure features must be considered when designing the patterns as further processing stepssuch as etching and metal lift-off will have an effect in these areas as well as in the patternedarea.4.4 ConclusionsPatterns with consistent feature resolutions down to -5O nrn have been achieved on bothline and dot patterns at feature spacings down to -2OO nm, on a field of view of 98.63 p.m. Theproximity effect along with developer sensitivity and contrast currently limit the resolution andstudies indicate that using thinner resist layers and shorter development times would give evenbetter resolution. As well, other development systems with higher contrast could be investigated[Ref. 14]. Systematic pattern uniformity variations on the order of 10 nrn rms and randomvariations of 5 nn-i rms have been found in the patterns. A source for the systematic variationshas been tentatively identified and more work is needed in this area to overcome this problem.Limitations on patterns due to beam current stability have been discussed, although more work isneeded to characterize low frequency AC noise on the beam current. As well, limitations onpatterns from the pattern rotation and field of view have been studied.80Chapter 5 Conclusions and Recommendations5.1 ConclusionsAn electron beam lithography system was implemented using a Hitachi S-4100 SEM andthe NPGS electron beam lithography exposure system. Associated facilities were developed toapply a variety of resist layers to substrates and to develop and etch the exposed patterns.Procedures for this system’s use were developed and optimized as it is anticipated that manyresearchers will use the system in the future.The system’s performance was characterized with importance being placed on thoseissues that impacted on the goal of achieving sub 50 nm resolution with high pattern uniformity.The results were found to depend on many parameters including the resist thickness, resistcomposition, development time, and the specific pattern that was written. Resolutions of -50 nmon feature spacings of -200 nm were achieved in 496K M.W. and 950K M.W. PMMA resistlayers -200 inn thick. Resolutions reported in the literature of about 10 nm on sub 50 nmspacings with 20 inn resist layers indicate that moving to thinner resist layers, shorterdevelopment times, and higher contrast developers will enable better resolution to be achieved[Ref. 6-7]. However, since further processing steps such as etching can also etch the resistlayers, this may limit the utility of very thin resist layers. Pattern uniformity on the order of 10nm, viewed over -5 im scales, has been achieved on both line and dot patterns, chosen for theirusefulness in the laboratory’s research, including the construction and characterization of DFBlasers and arrays of quantum dots. Systematic and random noise limitations were encounteredthat require further investigation to improve the uniformity beyond this level.815.2 Recommendations5.2.1 Pre-exposure ProcessesRecommendations for the pre-exposure processes fall into three areas: the processenvironment, the resist layer attributes, and characterization of the resist layer attributes. Sinceparticles on the substrate were found to be a major concern with these thin resist layers a moredust free environment than is currently used is desirable. This needs to be balanced against thenecessity of also having a well-ventilated area and one possibility is to place the spinner in a boxwith a fan attached for exhausting the fumes and with a filter on the air intake. Alternatively, afume hood could be used, although then the air cleanliness relies on the room air being clean.The entire procedure could then be accomplished in this cleaner area. If necessary, the ovencould also be placed within such an enclosure. Concerns with this idea are ensuring that thefumes would not present a safety hazard with the exhaust equipment and/or oven. Ideally allcleaning, spinning and baking would be accomplished in the same enclosure to minimize theexposure, possibly requiring remote handling.As thinner films were found to be a method of achieving better resolution it would bedesirable to have the ability to spin on layers as thin as 20 nm and characterize them. Dilutedresist solutions could be used to provide thinner resist layers. As well, there is some requirementfor multiple layer resists to provide finer features for lift-off processes. For both of these, itwould be useful to be able to optically characterize the thin film thicknesses in a quick manner inthe laboratory without needing to go through the exposure and development steps that were donein this work to characterize the film thicknesses [Ref. 32]. By monitoring the resist layer82thickness under identical spin and bake conditions, this would also allow monitoring of the resistsolutions over time to identify changes that may occur - thickening of the solutions, for example.With an x-y stage, it would also allow one to examine the spatial variation of thickness over asubstrate.5.2.2 LithographyFurther work in the lithography area can be divided into three main areas: hardwareaspects, theoretical investigations, and further characterization. As well, one important featurethat would assist all these areas would be the ability to capture images from the SEM. Theability to capture many images easily would allow more detailed characterization of theresolution, uniformity, and topography of imaged features over a much larger area than iscurrently possible with photographs from limited areas of the samples.Hardware aspects involve changes, additions, and further characterizations of the system.One thing to consider is re-calibrating the NPGS system with a higher resolution standard thanthat which was originally used, although this may require that a standard be designed andmanufactured using the lithography system if a suitable one cannot be found. A useful additionto the system would be a beam blanker. This would allow the alignment of patterns to beaccomplished without the beam dumps and stage translation exposure lines that now occur in thepatterns and would be helpful in larger scale work using fine features that may need to be writtenat higher magnifications. Further characterization of the system involves two areas; noisecharacterization and vacuum characterization. Noise characterization of the beam current andposition in the column is continuing and may require further shielding of the column. As well, it83would be useful to have a more accurate idea of the composition of gases in the specimenchamber and the pressure in this area to assess if the vacuum level plays any role in theresolution and uniformity of the lithographic process.Theoretical investigations could consider the lithographic process in more detail andattempts could be made to improve the models and predict the process with more accuracy.Experimental studies could be conducted in conjunction with this theoretical process.Further characterization could include several areas. Lithographic tests on much thinnerresist layers would, from the above work, be expected to increase the resolution - perhaps at thecost of further processing steps, however. Other development systems could be investigated tocompare the sensitivity and contrast of these systems to the one used in this work. As well, otherresist types could be investigated (examples: negative resist, inorganic layers).Further experimental work could be conducted on thin substrates, on the order of100-200 nm, which are expected to have a small proximity effect due to reduced backscatteringfrom the substrate. This should enable much finer resolution to be obtained. Accomplishing thiswould require greater uniformity in the patterns and would require more investigation of thelimitations, both systematic and random, imposed by the uniformity variations seen in this workand attributed, in the systematic case, to the NPGS system hardware.5.2.3 Other Processing RecommendationsOnly preliminary wet chemical etching work was accomplished for this work, and muchmore could be done here. Examples are the characterization of the effect of pattern feature sizeon the etch rate, to more fully investigate the effect of etchants on the resist layer, to investigate84the development of etch profiles and, possibly, to investigate the use of an intermediate layer toact as an etch mask under the resist. As well, only the citric acid-hydrogen peroxide etch systemwas used and other systems could be investigated.Although not used in this study, other pattern techniques could also be investigated, suchas dry etching and lift-off processes, for defining patterns useful in further studies.85References1. Griesinger, U.A., C. Kaden, N. Lichtenstein, and others, “Investigations of ArtificalNanostructures and Lithography Techniques with a Scanning Probe Microscope”, J. Vac.Sci. Tech. B 11(6), 2441 (1993).2. Davis, L., K.K. Ko, W.-Q. Li, and others, “Photoluminescence and electro-opticproperties of small (25-35 nm diameter) quantum boxes”, Appl. Phys. Lett. 62(22), 2766(1993).3. Greus, Ch., A. Forchel, J. Straka, K. Pieger, and M. Emmerling, “High quantumefficiency InGaAs/GaAs quantum wires defined by selective wet etching”, J. Vac. Sci.Tech. B 9(6), 2882 (1991).4. Meade, Robert D., A. Devenyi, J.D. Joannopoulos, and others, “Novel applications ofphotonic band gap materials: Los-loss bends and high Q cavities”, J. Appl. Phys. 75(9),4753 (1994).5. Smith, D.R., R. Dalichaouch, N. Kroll and others, “Photonic band structure anddefects inone and two dimensions”, J. Opt. Soc. Am. B 10(2), 314 (1993).6. Katoh, T., Y. Nagamune, G.P. Li, S. Fukatsu, Y. Shiraki, and R. Ito, “Fabrication ofultrafine gratings on GaAs by electron beam lithogrpahy and two-step wet chemicaletching”, Appl. Phys. Lett. 57(12), 1212 (1990).7. Fischer, P.B., and S.Y. Chou, “10 nm electron beam lithography and sub-SO nm overlayusing a modified scanning electron microscope”, Appl. Phys. Lett. 62(23), 2989 (1993).8. Thompson, L.F., C.G. Willson, M.J. Bowden, editors, Introduction to microlithography.theory, materials, andprocessing, ACS Symposium Series, M. Joan Comstock, serieseditor, American Chemical Society (1983).9. Jones, R.G., guest editor, Advanced Materials for Optics and Electronics, Special Issue,4(2) 53-178 (1994). NOTE: This is a special issue on microlithography, from achemistry viewpoint.10. Brewer, George Raymond, Electron-beam technology in microelectronicfabrication,Academic Press, London (1980).11. Valiev, Kamil A., The Physics ofSubmicron Lithography, Plenum Press (1992).12. Brodie, Ivor and Julius J.Muray, The Physics ofMicro/Nano-Fabrication, Plenum Press(1992).8613. Washo, B.D., “Rheology and Modelling of the Spin Coating Process”, IBM Journal ofResearch and Development, 21(2), 190 (1977).14. Fischer, P.B. and S.Y. Chou, “lOnm electron beam lithography and sub-50 nm overlayusing a modified scanning electron microscope”, Appi. Phys. Left. 62(23), 2989 (1993).15. Reimer, Ludwig, Scanning Electron Microscopy (Springer Series in Optical Sciences,Vol. 45), Springer-Verlag, Berlin (1985).16. instruction Manualfor Model S-4100 Field Emission Scanning Electron Microscope,Hitachi, Ltd. (1991).17. Kyser, D.F.and N.S. Viswanathan, “Monte Carlo simulation of spatially distributedbeams in electron-beam lithography”, J. Vac. Sci. Tech. 12(6), 1305 (1975).18. Kelly, Michael J., and Claude Weisbuch, The Physics and Fabrication ofMicrostructuresand Microdevices, Springer-Verlag, Berlin (1986).19. Williams, Ralph, Modern GaAs processing thechiques -- 2nd Ed., Artech House, Inc.,New York (1990).20. Runyan, W.R., and Kenneth E. Bean, Semiconductor integrated circuit processingtechnology, Addison-Wesley Publishing Company, Inc., New York (1990).21. Vossen, J.L. and W. Kern, editors, Thin Film Processes, Academic Press, New York(1978).22. Van Zant, Peter, Microchip Fabrication: A practical guide to semiconductor processing,2nd Ed., McGraw-Hill Publishing Co., New York (1990).23. Model S-4000/S-4100 Field Emission Scanning Electron Microscope SchematicDiagrams, Hitachi, Ltd. (1989).24. Nabity, J.C., Nanometer Pattern Generation System Installation Guide, Version 7.3, J.C.Nabity Lithography Systems, Bozeman, Montana, USA (May 1993).25. DesignCAD 2-D Reference Manuals, Version 6, American Small Business Computers,Pryor, OK, USA, 1992.26. Nabity, J.C., Nanometer Pattern Generation System, Version 7.4, J.C. NabityLithography Systems, Bozeman, Montana, USA (March 1994).27. Lee, Y.-H., N. Browning, N. Maluf, G. Owen, R.F.W. Pease, “Low Voltage Alternativefor Electron Beam Lithography”, J. Vac. Sci. Tech. B 10(6) 3094 (1992).8728. Peterson, P.A., Z.J. Radzimski, S.A. Schwalm, P.E. Russell, “Low-voltage Electron BeamLithography”, J. Vac. Sci. Tech. B, 10(6) 3088 (1992).29. Stark, T.J., T.M. Mayer, D.P. Griffis, P.E. Russell, “Effects of Electron Energy inNanometer Scale Lithography”, Proc. SPIE 1924, 126 (1993).30. Lee, Y.-H., N. Browning, R.F.W. Pease, “E-beam Lithography at Low Voltages”,Proc. SPIE 1671, 155 (1992).31. Horowitz, Paul, Winfield Hill, The Art ofElectronics, 2nd Edition, Cambridge UniversityPress, London (1989).32. Hirvi, K., T. Makela, J. Pekola, M. Paalanen, “Economical device for measuringthickness of a thin polymer film”, Rev. Sci. Instr., 65(8), 2735 (1994).33. Newbury, D.E., Joseph I. Goldstein, and others, Advanced Scanning Electron Microscopyand X-ray Microanalysis, Plenum Press, New York (1986).34. Holt, D.B. and D.C. Joy, editors, SEMMicrocharacterization ofSemiconductors,Academic Press, Ltd., London (1989).35. Moore, J.A. and Jin 0. Choi, “Degradation of Poly(methyl methacrylate): Deep UV, Xray, Electron-Beam, and Proton-Beam Irradiation”, Radiation Effects on Polymers, ACSSymposium Series No.4 75, American Chemical Society, 1991 (156-192).36. Reiser, Arnost, Photoreactive Polymers: the science and technology ofresists, JohnWiley and Sons, Ltd., New York (1989).37. Greeneich, James S., “Time evolution of developed contours in poly-(methylmethylacrylate) electron resist”, J. Appi. Phys, 45(12), 5264 (1974).38. Greeneich, J.S. and T. Van Duzer, “Model for Exposure of Electron-Sensitive Resists”, J.Vac. Sci. Tech. 10(6), 1056 (1973).39. Shimizu, R. and T. Ikuta, “Experimental and theoretical study of energy dissipationprofiles of keV electrons in polymethylmethacrylate”, J. Appi. Phys. 46(4), 1581 (1975).88Appendix A Electron Beam Lithography ProceduresThis appendix describes procedures for designing lithographic patterns and using theelectron beam lithography system that has been described in Chapter 3 to create those patterns ina resist layer. The pattern design, using the DesignCAD program, is discussed in Section A. 1.1followed by a discussion on the creation of the lithography exposure run files using the MRFprogram in Section A. 1.2. The procedure for applying resist coatings to substrates is thencovered in Section A.2. Finally, a procedure for using the NPGS System and Hitachi S-4100SEM to write the lithography patterns is given in Section A.3. All of these procedures assumethat the person following them has read and understood the relevant operating manuals for boththe equipment and the program software.A.1 Pattern Design and MRF File CreationIn this section, the design process will be illustrated by using a typical exposure pattern.This pattern and its associated run files were used to generate the test exposures for most of thedata presented in Chapter 4.In general, the pattern characteristics will be dictated by the particular application. Here,the “application” was largely to optimize the system parameters for writing one-dimensional linearrays and two-dimensional dot arrays. This presented two main considerations for the design ofthe patterns: the line/dot patterns themselves and a sequence of parameter increments. The testgrid used for these parameter characterization studies is shown in Figure A-i. The patternconsists of 20 im by 20 urn sub patterns separated by 4 urn. The crosses in the pattern consist of89+ + +y+ +xFigure A-i. Lithographic dose test pattern.100 nm wide lines with a high exposure specified: their purpose was to be easily visible whentrying to locate the pattern for inspection at low magnification where, depending on the testpattern, it may be very difficult to find the pattern itself. Each sub pattern consisted of an arrayof lines or dots with a specified linewidth and pitch, along with an exposure that varied betweensub patterns.There are two separate steps involved in generating the set of files needed so that theNPGS program can write the patterns. The first step involves using DesignCAD to specify theabsolute dimensions and relative exposures for all the feature elements that make up the pattern,and the second step involves generating a run file that specifies the absolute doses, beam current,and the pixel spacings used when filling the feature elements. Each of these two steps isdescribed separately below.90A.1.1 Pattern DesignPattern generation consists of designing individual pattern elements, arranging them intoa complete pattern, and associating them with different colours and/or layer numbers, each withseparate parameters specified in the run file. Each layer of the pattern is composed of a series offeature elements (eg. individual lines or dots, or filled boxes), and they are exposed in the orderthat they are created with DesignCAD, although this order can be changed within DesignCAD.The simplest way to generate the 16 separate arrays in the dose test pattern of Figure A-i is todefine each as a 20 pm by 20 im filled box with a dose (colour) that depends on its position inthe 4 X 4 array. The actual lines or dots within each box are then controlled in the run file wherethe pixel step sizes used to fill the boxes, zx and y, are specified by the line spacing distanceand centre-to-centre distance respectively. Each array thus has the same line or dot spacings, anda different dose. A more general way of generating line arrays with variable linewidths is todefine each line as a separate feature element.No beam dump is used in this pattern, but if necessary it would have been placed in theupper right corner of the pattern where the NPGS program default beam dump point is located. Itshould be noted that this requires that the X-Y control voltages be set using the DACO programincluded for this purpose to +10 V on both settings which is the standard practice here due to thelack of a beam blanker (see exposure procedure 9.1 in Section A.3).Once the pattern is drawn, the elements are given different colours corresponding todifferent exposures. For example, each box in Figure A-i is given a different colour.Each layer is like a separate drawing in that it can have both the spacings, themagnification, and the exposure doses (from the colours) set individually when the run file is91created (Section A. 1.2). For example, since the crosses of Figure A-i need different spacingsand exposures than the test patterns, they are defined in layer 2 of the drawing, and are placed inpositions corresponding to locations between the filled boxes in layer 1.Which of the particular features above is used depends on the individual pattern, but itshould always be remembered that the pattern drawn sets the limits on the highest magnificationthat can be used to expose the pattern since the pattern must fit entirely into the field of view tobe accurately reproduced. Through the number of pattern colours, the pattern also sets themaximum number of separate exposure doses that can be specified in the pattern. At this point,when the pattern design is completed, the run file should be created.A.1.2 Run File CreationThe run file reads the previously created pattern(s) and sets the conditions for exposingthe features specified in those patterns. First, the number of patterns (up to 16) and their namesare specified. The individual patterns can then be repeated up to 20 times with a separate offsetfor each one, allowing a complicated pattern to be repeated without creating a very large patternfile. Once all the patterns are specified, the exposure conditions for each pattern are entered. Asample of portions of a run file are shown in Figure A-2. These correspond to run files used tocreate the patterns that are the basis of most of the data in Chapter 4. The pattern usedcorresponds to the one shown in Figure A-i.The exposure conditions within a pattern are defined by setting the magnification and thebeam current. The magnification is determined by the size of the pattern, with relation to thepattern’s origin. Once these are set, the other features can be specified. The centre-to-centre and92line-spacing distances are set by considering the pattern features. The minimum spacing isdetermined by the 16-bit resolution on the DAC board in the computer and the magnificationspecified earlier. This means that the minimum spacing is given by the field-of-view divided by216. Any spacings set will also be rounded to an integer multiple of this value. For lines, thecentre-to-centre spacing is made small enough to allow a smooth transition from point to point,while the line-spacing is set to 1/4 of the final desired linewidth. For single pass lines, only thecentre-to-centre distance is of concern and the line spacing in a filled polygon structure definesthe line pitch. As an example, for the filled square patterns discussed previously, the centre-to-centre spacings were set to 20 nm. In these patterns from Figure A-i, the line spacings were setto be the spacing required between each single pass line, from 100 to 1000 nm. For the dots, thetwo spacings were made the same allowing a square grid of dots to be generated. The spacingsare set independently in layer 2 for the cross patterns as is shown in Figure A-2.Using the data entered above, the exposures can be specified. Different exposure dosesare entered for each colour used in the pattern file. Either an incident dose or an exposure timecan be set, and the other one is calculated using the beam current which was previously entered.Two types of dose may be set, line and area doses. Both use the charge deposited, calculated bymultiplying the exposure time by the beam current. For line doses, the charge deposited isdivided by only the centre-to-centre distance and the units are in nC/cm. For area doses, bothspacings are used and the units are in jiC/cm2. For dots, the spacings define the pattern and arenot directly relevant to the exposure. As such, for single dots the exposure time is the parameterthat is specified and the relevant incident dose was found to be in the fC range. The minimumvalue of the exposure times for systems without a beam blanker is 10 ps and exposure dosesbelow the fC range are only possible if the beam current is reduced from the nominal value of9310 pA. For the individual exposures, the relevant values to enter are either known from previousexperiments or must be estimated. If no other data is available the sensitivity for the developerbeing used should be the starting value. In a test run, most of the exposures should be in thisestimated region, with some exposures significantly above and below the sensitivity value toensure that some pattern features will come through if the exposure settings are not correct.Once these exposures are tried, the results are fed back in to narrow the range and find theoptimal dose in the next pattern written. The exposures can be scaled at run time if the beamcurrent is higher than the one specified in the run files, so an exact value is not needed whendesigning the run file. A good value to take is 5 pA, since the beam current was found to varybetween 1 and 10 pA during experiments. The net result of this exposure scaling is, however, toleave the dose unchanged since the exposure scaling factor is calculated by dividing the file’sbeam current by the measured beam current.In the run file, different pattern numbers or layers can be used to vary one or both of thespacings and the exposures. In the files corresponding to Figures A-i and A-2, there are 16exposures in each pattern written, and ten patterns in a run file. Thus, one run file was used tocreate a dose test for a specific geometry and then vary the geometric spacing ten times to createtest patterns where the line or dot spacings changed from 100 to 1000 nm. There is a great dealof flexibility in what can be accomplished in a combination of patterns and run files and planningahead before creating these files is crucial to maximize their utility for a particular application.94Run FileName. ab408000.rf6Pause only for ‘p option’ (y,n) noHow many patterns to write’ 10Pattern Name #01 ab408000How many times to repeat pattern’ 1-Pattern Location- 0,0Pattern Name #02- ab408000Flow many times to repeat pattern?-Pattern Location- 0,0Pattern Name #10- ab408000How many times to repeat pattern’ 1-Pattern Location- 0,0Pattern File: ab408000 (1 of 10)Pattern File: ab408000 (10 of 10)Layer #01- (w,p,c,s) wOrigin Offset (x,y)- (im,im) 0,0Magnification- 1000Center-to-Center Distance- (A) 105.4Line Spacing- (A) 9993.6Microscope Configuration # 1.000Measured Beam Current- (pA) 5.0Color #1 (white) (jisec) 21.1-Line Dose: (nC/cm) 0.100Color #2 (red) (i.ssec) 42.2-Line Dose- (nC/cm) 0.200Color #3 (green) (i.tsec) 63.2-Line Dose- (nC/cm) 0.3 00Color #4 (blue) (ssec) 84.3-Line Dose- (nC/cm) 0.400Color #5 (brown) (gsec) 105.4-Line Dose- (nC/cm) 0.500Color #6 (magenta) (psec) 126.5-Line Dose (nC/cm) 0.600Color #7 (cyan) (ssec) 147.6-Line Dose (nC/cm) 0.700Color #8 (dark gray) (psec) 168.6-Line Dose (nC/cm) 0.800Color #9 (light gray) (jisec) 189.7-Line Dose (nC/cm) 0.900Color #10 (light red) (isec) 210.8-Line Dose (nC/cm) 1.000Color #11 (light green) (isec) 231.9-Line Dose (nC/cm) 1.100Color #12 (light blue) (psec) 253.0-Line Dose: (nC/cm) 1.200Color #13 (yellow) (iisec) 274.0-Line Dose (nC/cm) 1.300Color #14 (light magenta) (isec) 295.1-Line Dose (nC/cm) 1.400Color #15 (light cyan) (isec) 316.2-Line Dose (nC/cm) 1.500Color #16 (white) (.tsec) 421.6-Line Dose (nC/cm) 2.000Layer #02 (w,p,c,s) wOrigin Offset (x,y) (sm,pm) 0,0Magnification 1000Center-to-Center Distance- (A) 255.9Line Spacing- (A) 255.9Microscope Configuration # 1.000Measured Beam Current- (pA) 5.0Color#I (white) Qisec) 130.9-Area Dose (pC/cm2)100.000Run File Name ab408001 .rf6Pause only for p option? (y,n) noHow many patterns to write? 10Pattern Name #01 ab408000Pattern File: ab408000 (2 of 10)Layer #01: (w,p,c,s) wOrigin Offset (x,y) (pm,pm) 0,0Magnification 1000Center-to-Center Distance- (A) 2001.7Line Spacing- (A) 2001.7Microscope Configuration # 1.000Measured Beam Current- (pA) 5.0Color #1 (white) (psec) 200.2-Line Dose (nC/cm) 0.050Color #2 (red) (p5cc) 400.3-Line Dose (nC/cm) 0.100Color #3 (green) (psec) 600.5-Line Dose (nC/cm) 0.150Color #4 (blue) (psec) 800.7-Line Dose (nC/cm) 0.200Color #5 (brown) (psec) 1000.8-Line Dose (nC/cm) 0.250Color #6 (magenta) (isec) 1201.0-Line Dose (nC/cm) 0.300Color #7 (cyan) (psec) 1401.2-Line Dose (nC/cm) 0.350Color #8 (dark gray) (p5cc) 1601.4-Line Dose (nC/cm) 0.400Color #9 (light gray) (iisec) 1801.5-Line Dose: (nC/cm) 0.450Color #10 (light red) (iisec) 2001.7-Line Dose (nC/cm) 0.500Color #11 (light green) (p5cc) 2201.9-Line Dose (nC/cm) 0.550Color #12 (light blue) (p5cc) 2398.0-Line Dose (nC/cm) 0.599Color #13 (yellow) (p.sec) 2598.2-Line Dose (nC/cm) 0.649Color #14 (light magenta) (jisec) 2798.4-Line Dose (nC/cm) 0.699Color #15 (light cyan) (psec) 2998.5-Line Dose (nC/cm) 0.749Color #16 (white) (p5cc) 3198.7-Line Dose (nC/cm) 0.799Figure A-2. Annotated sample run files, corresponding to those used for producing the patternsin Figure A-i. Similar files were used to produce the data and pictures in Figures 4-i, 4-5, &4-10.95A.2 Procedure for the Application of Resist Coatings to Substrates(1) Pre-heat the oven to the pre-bake temperature. For PMMA, this must be 160-180 °C.175 °C was always used in this work to allow the temperature drop to remain above the160 °C lower limit when the sample was inserted into the oven. Note that the warm upperiod takes approximately 1.5 hours. Monitor the temperature using the thermometer onthe top of the oven.(2) Assemble the equipment. This includes: cotton tipped applicators, 4 glass beakers,absorbent pads (to cover the work area), glass dishes to place the substrate in (the numberof these depends on the amount of substrates and their size - more than one can be placedin a dish, but they should be separated by at least 1 cm from each other), appropriate sizedchucks, aluminum foil, disposable pipets and bulbs (2 or 3), gloves, wipes for cleaningthe glassware, and personal protective equipment - lab coat, goggles, and acid-solventfilter mask (as appropriate). Chemicals and related equipment include: acetone,methanol, deionized (DI) water, waste solvent container, resist, resist waste container,glass (sharps) container.(3) Turn on the spinner and allow 5 minutes for the electronics to warm up.(4) Replace the aluminum foil over the bowl of the spinner, if necessary. Clean the glasscontainers used to bake the substrates, if necessary. This may involve an initial cleaningand scrubbing with a cleaner designed for chemistry glassware. The container may thenbe rinsed with acetone and methanol if this is deemed necessary, and wiped dry.(5) Inspect the chuck to insure that the 0-ring is present and not damaged. Attach the chuckto the spinner by aligning the chuck’s screw with the flat part of the shaft and pushing thechuck onto the shaft. It should fit on smoothly. Tighten the screw until it stops going in -do not tighten it beyond this point.(6) Put on the protective equipment. Turn on the pump.(7) Clean the top of the chuck with acetone using a cotton tipped applicator. These should beonly used once, so discard the applicator or break it after use so it won’t be accidentlyused again.(8) Spin on procedure(8.1) Clean the sample, if necessary, with acetone, methanol, DI water, and blow offwith nitrogen. Spray the sample using a squeeze bottle while holding it withtweezers over a beaker. Note that if using a new packaged clean wafer then thiscleaning step may be skipped.(8.2) Place the substrate on the spinner chuck. Centre the piece on the chuck.96(8.3) Set the speed on the spinner and spin the wafer, resetting the speed if necessary.Use the nitrogen gas used in step 8.1 to blow off the wafer while spinning. Stopthe rotation using the rear part of the spin control pedal.(8.4) Apply the resist with the disposable pipet. Cover the wafer entirely. Whenremoving the resist from its container, do it quickly and replace the capimmediately to prevent the resist from thickening. Place the first drop in thewaste container, as the solution the resist is dissolved in evaporates quicklyleaving a more concentrated solution in the bottom of the pipet that may create athicker or less uniform resist layer. If you will not be using the rest of the solutionin the pipet immediately dispose of it using the chlorinated waste container.Similarly, it is recommended that a new pipet be used for each application ofresist for the same reason, unless the process is done quickly.(8.5) Spin the resist and substrate for 30 to 40 seconds, by which time the colour of theresist layer on the sample should be uniform and have reached a steady state.(8.6) Remove the wafer and place it in a glass container. Cover the container.(8.7) Clean the holder with acetone and a cotton tipped applicator. Use a newapplicator and break or discard it after use.(9) Repeat step 8 until all the wafers have been coated. If necessary, replace the chuckfollowing steps 5 & 7. If replacing the chuck, a new pipet should be used in step 8 as thetime spent will be too long and the resist will thicken in the pipet affecting thereproducibility of the process.(10) Place the glass container(s) containing the substrates into the oven and bake them. Therecommended bake time for PMMA is 2 hours. Longer times may be used if there areproblems with further processing steps that result in the resist lifting off the substrate.Occasionally monitor the temperature of the oven to ensure it reaches the temperature instep 1 and remains within the limits specified there.(11) Turn off the pump and spinner.(12) Clean the chuck with acetone and a cotton applicator. Remove the chuck from thespinner and place it in its storage container. Remove and replace the aluminum foil in thespinner bowl. Discard the old foil.(13) Clean and store the equipment and chemicals. Put waste chemicals and used pipets intotheir proper disposal containers.(14) When the bake is finished, remove the glass containers from the oven and allow them tocool. Turn off the oven. When the containers have cooled, remove the substrates andplace them in storage containers. Cover the storage containers with foil.97A.3 Exposure of Resist Using the Electron Beam Lithography SystemAt this point, the procedures outlined in A. 1 and A.2 should already have been carried out. Asubstrate that is coated with a pre-baked resist layer should be ready to have a pattern written onit. As well, the NPGS run file(s) and their accompanying pattern(s) should be ready to run withthe NPGS program.(1) Both 12 and 24 hours before using the microscope to write patterns, the cold trap on themicroscope should be filled with liquid nitrogen. As well, fill the cold trap beforestarting to use it to write the patterns.(2) Sample Preparation:(2.1) If necessary, cleave the substrate into a piece of the correct size for furtherprocessing. This is accomplished on GaAs by marking the edge where the cleaveis wanted with a diamond pencil and applying pressure on this scratch. Unless ithits a defect, the wafer will cleave in one of two directions on a (001) surface,along either the (011) direction or the (011) direction.(2.2) A scribe can be used to mark the edge of the wafer where the pattern is to bewritten. The advantage of this method is that the pattern will cleave along thismark, if cleaving through the pattern is desired. This is the recommended method.Alternatively, the NPGS manual recommends marking every mm along the edgeof the substrate where the pattern will be written with a pencil, by scrapinggraphite from the pencil which serves as both a reference and something to focuson at the edge of the wafer. Note that this method should not be used if it willcontaminate further processes. Neither method should be used if the pattern is tobe written so close to the edge that the mark would interfere with the exposure.(2.3) The substrate should now be attached to the holder, ensuring that it lies flat.Although double sided adhesive tabs can be used to attach the wafers to theirholders, a better way is to use metal clips. The substrate should be mounted in thevicinity of both the Faraday Cup (used to measure the beam current) and a goldstandard (used to correct astigmatism in the beam). Handle the samples andholders with gloved hands only, to avoid depositing oils and other contaminantson them.(3) Turn on the microscope, check and record the pressures on the ion pumps (IP1 to 1P3),and insert the sample and holder into the specimen exchange chamber following themicroscope operating instructions. While pumping down the exchange chamber, checkand set all the microscope parameters (accelerating voltage, working distance, emissioncurrent, condenser lens setting, and both aperture settings). Also, load the pattern and runfiles into the computer in the appropriate subdirectory, if this has not already been done.(4) Ensure the raster rotation, dynamic focus, and tilt compensation features are switched off.98(5) Before turning on the beam, use the micrometer control knobs to position the holder sothat the sample to be exposed is not under where the beam will be when it is turned on.With the initial micrometer stage settings, the beam is in the centre of the holder.Following the microscope operating instructions, flash the beam and turn it on. Go tolow magnification and, taking care not to expose the sample, move the holder until thegold standard is in view. Focus using the Z-micrometer stage and the fine focus controlknob only, and follow the procedure for aligning the beam and correcting theastigmatism in it. You should be able to see clear features on the order of a few nm insize when the astigmatism has been corrected. Do NOT use the coarse control as thischanges the working distance. Take your time as this is a critical step for the lithographicprocess and you also need to wait for the beam current to reach a more stable period in itscycle (see the discussion in Chapter 4 for further details).(6) Measure the beam current using the Faraday Cup and picoammeter (see Section 3.3.1 ofChapter 3). Record the micrometer X and Y locations for the Faraday Cup.(7) Move to the edge of the sample where you wish to write, being careful not to expose toomuch of the sample, and focus on the edge of the sample using high magnification.Again, use the Z-micrometer and fine focus control knob only. The edge of the substratewill be at an angle to the screen, and must be aligned so that it is parallel to one of themicrometer stage directions. Do this by translating the sample and, using the rotationknob on the stage, adjust its rotation until translating it over distances of a few micronscauses very little change in the screen position of the sample edge. Record the locationof the position on the edge where you wish to write the pattern. Move back to theFaraday Cup and record the current value again. If the current is not varying quicklyduring steps 6 and 7, then you are ready to write.(8) Record the current and move back to the location where you wish to write the pattern,using the location recorded in step 7. If the current is not the same as the one specified inthe run file(s), you can change the run file(s) or use the “e” parameter in the NPGS andPG programs to scale the exposure time as discussed in the NPGS instruction manual.The value of “e” is given by dividing the run file current value by the measured value.For example, if the run file has a current of 5 pA and the measured current is 10 pA thene = 0.5. The scaling parameter is a faster method and should be used, since the beamcurrent is varying with time and this method uses the least amount of time.(9) Exposure:(9.1) Switch the microscope X-Y Scan Control Switch from internal to external control.Set the X-Y position to the upper right of the pattern using the DACO program.Move to the position where you wish to write the pattern using the X and Ymicrometer knobs, keeping in mind that the beam is not off and will expose theresist during this movement and also when the beam sits at the pattern positionbefore writing the pattern.99(9.2) Keep a separate record of each exposure run. Record the pattern position andwrite the pattern using the run file, recording the start and stop times. Rememberthat the pattern location is an offset of the location on the edge of the sample thatwas recorded in step 7. Monitor and record the current through the sample whilewriting the pattern along with the time (for example, after each pattern in a runfile is written). Note that the current will change once a pattern starts writing andit has been found that it is more stable and consistent to record it while the writingis in progress. Also monitor the extraction voltage and emission current. Asdiscussed in Chapter 4, the most stable current was found to occur by notadjusting the emission current while writing patterns.(9.3) When finished writing the patterns, move away from the pattern and off the waferusing the X-Y micrometer controls and then switch back to internal scan control.Since there is no beam blanker installed in the system, make sure that the electronbeam does not cross through previously written patterns during this procedure.Record the specimen current, using the Faraday Cup, and the time.If writing multiple run files or patterns that take a long time to write, move off the waferand back to the Faraday Cup between pattern files to check the current, recording the startand stop times of this movement and the time the current measurement was taken, alongwith the beam current value.(10) When finished, shut the beam off and remove the sample following the microscope!soperating instructions.(11) Remove the sample from the holder and place it back in its container. The sample is nowready to be developed.(12) Development Procedure:(12.1) Set up the solutions in a row so that you can move in one direction whiledeveloping. Fill two 50 ml beakers with about 40 ml of developer (methylisobutyl-ketone (MIBK) in this work), 40 ml of rinse (isopropyl alcohol (IPA) inthis work), and, in a 500 ml beaker, about 300 ml or more of deionized water.Place the nozzle to blow off the samples near the water rinse beaker. Place atimer where you can monitor it during the procedure.(12.2) Develop the sample by immersing and agitating it with tweezers (the self-holdingkind are recommended) in first the developer (90s), followed immediately by therinse (1 Os) and then the water (1 Os). It is recommended that these times be usedto start with, and once good, consistent results are obtained consideration may begiven to changing them. At the last step, keep the sample immersed in the waterwhile getting the nozzle ready. Remove the sample and immediately blow thewater beads off the surface on both sides. Place the sample in its storagecontainer.100(12.3) The sample can now be inspected under an optical microscope, to see if thepatterns are through. Note that this will not work for features smaller than themicroscope’s resolution, but it can be a useful tool to inspect the sample and redevelop it if necessary.(12.4) Dispose of the used developer and rinse in the waste solvent container. Rinse andwipe dry all the beakers used and place them in their storage locations.(13) The sample can now be inspected in the SEM to view the pattern, but keep in mind thatthis will expose the pattern further and may cause additional problems if furtherprocessing is required, as discussed in Chapter 4. For storage, avoid humid areas andkeep in mind that light can cause photo-degradation of the resist.101Appendix B Further Details on Beam-Material Interactions and Resist DevelopmentThis appendix describes more details on electron beam-material interactions and theeffect they have on polymer resists, with special attention paid to the resist used in this work,PMMA. The electron beam lithography models commonly encountered in the literature are alsodiscussed. Electron beam-material scattering and energy loss will be discussed first, followed bya discussion on the effects of the impacting electrons on the resist and, finally, the consequencesfor resist development.B.1 Electron Scattering and Energy LossTo characterize the distribution of energy absorbed in the resist layer, both the scatteringand energy loss of the electrons as they travel through the resist and substrate must be described.There are two types of scattering: elastic scattering which involves changes in direction but verysmall to no loss in energy, and inelastic scattering where the electrons lose energy, and may alsoundergo small changes in direction. These can be incorporated into a model of the scattering inthe resist layer which simplifies the overall process into forward scattering or back scattering[Ref. 8; 10; 15]. Forward scattering events are those that occur as the electrons in the beaminitially traverse the resist layer. Scattering in this regime mainly has the effect of broadening theincident beam over a small area, as was discussed in Chapter 2. Backscattered electrons arethose that re-enter the resist after undergoing a combination of energy loss and large anglescattering in the substrate. As mentioned in Chapter 2, this results in an interaction area in theresist that is much larger than the one due to forward scattering.102In simulations, the scattering cross sections for the electron are used to determinescattering angles and the lengths the electron travels between collisions. A conunon method is touse the elastic scattering cross section to find the scattering angle and an inelastic expression fordetermining the energy loss. In more general terms, there are cross sections corresponding toelastic as well as inelastic processes. For elastic events, the screened Rutherford cross sectioncan be used for elements of low to intermediate atomic number (eg, up to Ge, Z=32) andintermediate beam energies of about 20-5 0 keV. For more accurate results and outside of thisrange, exact Mott cross sections are used [Ref. 15; 33; 34]. Some of the inelastic effects that canbe considered are: the generation of slow secondary electrons (0-50 eV from conductionelectrons) and fast secondary electrons (>50 eV from more tightly bound electrons), inner shellionizations, and multiple electron excitations (bremsstrahlung radiation and plasma scattering).To be thorough all these effects would need to be included in a model to determine the totalscattering cross-section and mean free path. However, the screened Rutherford scattering cross-section can be used to approximately calculate the elastic scattering cross-section [Ref. see 33,p.20; 15, p.8 1 for greater detail]. The differential screened Rutherford cross section is given bye4Z2 dOdo (6) =_________________________16 (4 t € B)2 [si112(O/2) + 02/4]2 (B—i)6/4 - 8 = 3•4x 1O3Z/E (B is in keV)where Z is the atomic number, 6 the scattered angle, E the electron energy, e is the electroncharge,€ the permittivity of vacuum, and 8 is called the screening parameter. The total elasticcross section, ef, is found by integrating over all the angles from 0 to 180° using dO = 2tsinOdO.Similar methods are followed for other relevant cross sections that need to be calculated. As the103electron energy increases, the need for relativistic corrections needs to be examined. ForEquation B-i, relativistic corrections need to be added at energies higher than about 50 keV.Using the cross-sections, the mean free path is found from1 1 ci a..“uT i J1 ,,hA. (B-2)cwhere AT is the total mean free path, A are the individual process path lengths, and the othervalues are as defined previously. cj is the mass fraction used for materials that involve more thanone element in their composition, where l is the atomic fraction of the jth element with atomicmass A. By integrating Equation B-i to an angie O and dividing by the total cross-section e/,one can get a probability that an electron will scatter from 00 to @. For simulations, this iscombined with the mean free path ATto find the angle and distance that an electron scatters to[Ref. 33-34]. The remaining piece needed is an expression for energy loss.To account for energy loss, the continuous slowing down approximation (Bethe energyloss equation) can be used [Ref. 8; 10; 12; 15; 33-34]. This equation is an approximation thatincludes the effects of energy loss from all inelastic processes and gives an energy loss per unitlength of, in its nonrelativistic form,= -7.85 x c.31n 1.166 (keV/cm)dx E AE .11 (B-3)(9.’76z + 58.8Z1°’)x 1O (keV)104where the only additional undefmed symbol is J1 which is the mean ionization potential for anelement with atomic number Z1. This is used to record the loss of energy as the electron movesthrough materials in a simulated trajectory and can give, with a large number of simulatedtrajectories, an idea of the amount of energy deposited in the resist layer and substrate.Modelling using the above expressions makes attempts to correct for the proximity effecttoo time consuming and so more analytical models have also been developed, such as the“double gaussian” model presented in Chapter 2. However, the more complicated models useexpressions such as those above in Monte Carlo simulations in their attempts to understand thelithography process. These models usually involve thousands of simulated impacting electrons,and follow their trajectories through the resist and substrate, recording the energy deposited inthe resist layer. If, as was discussed in Chapter 2, the changes in the resist are taken to beentirely due to the absorbed energy density, the changes to the resist and its development maythen be defined using equations of the sort described in the next two sections. The incident dosecan then be calculated if enough simulated trajectories are used that a fairly good average of theenergy deposited in the resist is built up.B.2 Resist Modification MechanismsThe mechanisms causing changes in the molecular weight of resists are described in thissection, and an expression for the change in molecular weight as a function of the absorbedenergy density (or absorbed dose) is presented.When passing through the resist and substrate, the electrons loose energy by inducing105ionizations within bonds of the molecules and, at lower levels of energy transfer, non ionizingelectronic and lattice excitations within the material. Both of these can induce changes in thematerial and can also result in the emission of secondary electrons from the ionization process.These secondary electrons can also cause ionizations and excitations. Electrons recombiningwith the ionized molecules, or cations, can also cause excited states within the molecules. Theinitial beam energies are much higher than the bond energies and thus do not tend to selectivelyexcite bonds as in photochemistry [Ref. 35; 9, p.139-153]. The bond energies in polymethylmethacrylate (PMMA), for example, are in the range 3.5-4.5 eV [Ref. 35], while typical values ofionisation energies for atoms and molecules are in the range 8-50 eV [Ref. 36, p.36]. Theexcitations can also result in the emission of photons or phonons. In organic polymers used forelectron beam lithography two important overall reactions follow from these ionizations andexcitations.After the molecule is ionized, if the ensuing reactions lead to the formation of bondselsewhere on the polymer or combination reactions between polymers then a cross-linkednetwork of bonds could form and result in the polymer being less soluble, which leads to theformation of a negative resist. Alternatively, if the reactions lead to the breaking up of thepolymer into smaller fragments, called chain scissioning, then the process leads to the formationof a positive resist. Both of these types of reactions usually proceed simultaneously and,although scissioning dominates in positive resists at lower doses, higher doses lead to negativeresists in most organic polymers. This is because the scissioning is limited by the number ofavailable bonds, while the cross-linking proceeds continuously and links the scission fragmentsas well as the unbroken fragments. In addition to the above reactions gaseous molecularfragments can also be produced. Eventually, at high enough doses, a cross-linked, carbonaceous,106Cl-f3 CH3 CH3• —CH2--C-—CHC- c——-—CO CO COOCH3 QCH3 OCH3Ionizing KRadiationCH3 CH3 CH3——CH2C-—C-C- ------• COOCH3 OCH3CH3 CH3 Cl-I3 A eo CO + .OCHCH—C CH2— + • C OCH CO2 + • Cl-I3CO‘CH COOCH32OCH3yH COOCH3Figure B-i. Illustration of the process of chain scissioning leading to the reduction ofmolecular weight under the influence of ionizing radiation for polymethylmethacrylate(PMMA) [Ref. 35; 36, p.269,302-316]. • denotes the highly reactive unpaired bonds,or radicals in the process.polymerized substance is formed on the surface due to the increasing numbers of carbon-carbondouble bonds and large fractions of the non-carbon elements leave the material as gaseousfragments [Ref. 15, p.124.].Polymethylmethacrylate (PMMA), the resist used in this work, is a positive electron-beam resist and the process leading to chain scissioning in PMMA is illustrated in Figure B-i.The initial impacting electron ionizes the molecule, leaving a highly reactive unpaired bond onthe polymer (called a radical), a radical fragment of the molecule, CO(OCH3),and an electronthat can ionize other bonds if it has sufficient energy. The CO(OCH3)fragment breaks down into107gases and other radicals. The gaseous products formed can escape from the substrate, leavinggaps that assist the solvent in removing the molecular fragments during the development process.The scissioning molecule breaks into two fragments, one of which is a radical. The remainingradical fragment reacts with an electron in a process called hydrogen abstraction to form a doublecarbon bond which stabilizes the end of the second molecular fragment. The hydrogen atom lostfrom the molecule during this process can then react with the radicals formed from theCO(OCH3)fragment. In addition to the scission process illustrated, the process of hydrogenabstraction can occur in PMMA without chain scissioning, forming a carbon double bond withno break in the polymer chain. This is the process responsible for lowering the scissioningefficiency of PMMA [Ref. 35].The result of these processes in resists is usually characterized by radiation chemicalyields described in terms of the number of specified events per 100 eV of energy absorbed: G(x)for cross-linking and G(s) for chain scission events. The final molecular weight is given by theexpression [Ref. 37; 9]± = + [G(s) - G(x)}D B 4Mt,. lOOpN ( -)where Mf is the fragmented molecular weight and M is the initial molecular weight, both inamu (gmol’). D is the absorbed dose in eVcm3,p the resist density in gcm3 and Na isAvogadro’s number. Note that if scissioning is the only process to occur the G(x) values aretaken to be zero. Since resists contain a distribution of molecular weights, this expression wouldneed to be modified to account for this in a simulation of the lithography process.108B.3 Resist Development ModelAs was discussed in Chapter 2, the developer acts on the resist material with reducedmolecular weight. To simulate this, a rate equation incorporating a dependence on the averagefinal molecular rate, Mf, of the fragments produced by the electron beam-resist interactions of theformR = R0+(M)A (B-5)is often used [Ref. 9, p.168; 17]. In this expression, R is the solubility rate of molecularfragments with molecular weight Mf, and R0, A, and B are parameters that must be empiricallydetermined. When used with expressions for the energy deposited within the resist layer and theresulting change in the molecular weight of the resist, the solubility of fragments in an area ofresist that have received a specified incident exposure dose can be modelled. From this, theamount of resist removed as a function of time can be modelled and estimates of the finaldeveloped pattern features can be made.Several people have investigated electron-beam lithography models [Ref. 17; 37-39], butthe usefulness of many of the models in predicting electron beam lithography processes is not yetclear, due to both the complicated nature of the reaction mechanisms and the large numericalsimulations that must be carried out to model these processes.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items