The images etched in a positive photoresist layer by means of polychromatic exposures and all-reflective projection printing methods can be described by a model presented elsewhere. For that model, the lateral intensity distributions of the optical image need be computed only at a single effective exposure wavelength that is compatible with a properly focused resist image profile. Although the illumination in the optical system we used was partially coherent, we successfully used the effective enhanced numerical aperture for the system, and calculated the lateral intensity distribution in the optical image on the assumption of incoherent illumination. In this study, we deposited a layer of Al over a Si substrate wafer which we then coated with a layer of quartz and a layer of positive photoresist. We calculated the dynamic exposure response of the photoresist film by using the modulation transfer function of a defocused perfect lens in order to simulate the resist images from zero to a few Rayleigh units of defocus. Simulated and experimentally determined resist image profiles and linewidths were compared and shown to be in good agreement. The results of this comparison lead us to believe that defocusing at the time of our study was caused by a tilt of the wafer along the scan axis.
A four phase transparent gate CCD has been fabricated using 1:1 projection printing on all photo-lithographic processes which required resist resolution in the 2-4 µm range on many of the masking levels. The photolithographic processes and techniques used will be described and presented.
The economics of circuit manufacture require die size to be minimized to maintain yield potential. As circuits increase in complexity, one way of achieving die size reduction is to reduce linewidths. Average production circuit geometries of 4 µm in 1977 are forecast to be reduced to 2 to 3 µm in 1981. Presently available systems will be able to meet the requirements for these smaller geometries. As these geometries get smaller, tighter tolerances will be required in all process parameters and some procedures will become obsolete. Improvements in optical projection will be made and shorter wavelengths of the exposing illumination will be used to obtain increased resolution. This will introduce a new set of conditions which must be addressed. Finally, as the industry looks to 1982 and beyond, submicrometer geometries will receive more and more attention as will X-ray and E-beam lithography.
A mathematical model for partially coherent imaging in projection printing is presented using Fourier transform methods. Some interesting two-dimensional examples are offered which illustrate the dramatic effects of partial coherence in the mask illumination on the edge contours in the image. Optical simulation techniques are discussed with possible application to system evaluation and design.
Increased packing densities require smaller geometries. For memory circuits the trend is towards 2 microns and less. Available and suitable imaging techniques for wafer production are limited. Photolithographic step-and-repeat imaging at reduction ratios greater than one to one appears to be a most promising technique. Its successful application demands a clear understanding of wafer process parameters and system imaging interaction, appropriate expertise for integration of opto-mechanical - electronics and control functions, component selection, and design trade-offs. A major performance factor is the illumination system. This paper will provide a detailed description of illumination system criteria and implementation. A new illuminator, an extension of a proven design, makes possible exposure energy levels in excess of 300 milliwatts/cm2 of great uniformity over a 10 mm x 10 mm area. Resulting exposure times are fast enough to make direct stepping on the wafer practical. Images stepped on wafers in both negative and positive photoresists will be discussed. Data obtained by photo-optical and SEM means demonstrate the potential usefulness in terms of resolution, topology, resist thickness and type, of photolithographic direct wafer exposure. Standing wave effects and negative resists exposed without nitrogen purge will be discussed.
In order to accurately achieve VLSI micro-lithography for integrated circuit applications, with high throughput, new optical printing systems are required. The manufacturer of a projection alignment system must contend with and solve numerous problems in optical, electronic, and mechanical technologies. This paper describes a new projection aligner, and the solutions which were employed to achieve the desired goals. Test results are given where applicable. Included is the basic optical concept, methods of achieving VLSI precision alignment, operator-machine interaction, and the high intensity exposure system. In addition, for VLSI imaging,an automatic focusing system has been designed which monitors the focus of the wafer through the actual optical path. A system of automatic wafer alignment will also be described.
The exposure of an x-ray resist by radiation from laser-heated plasmas was recently demonstrated. Single-shot submicrosecond exposures with a very favorable x-ray spectrum are possible. In order to reduce the cost of a laser-plasma x-ray lithography system, it is desirable to maximize the intensity in the soft (1 to about 3 keV) range. The x-ray output of laser-plasmas depends on laser pulse parameters (wavelength, pulse shape and energy), the focal conditions, and the target composition and geometry. Laser-plasma x-ray characteristics and their sensitivity to experimental parameters are reviewed in this paper. Presently available information indicates that a Nd:glass laser having pulse width in or near the 1-10 nsec range with at least 500 J of energy should be adequate for practical single-shot x-ray lithography.
The proximity effect in electron-beam lithography describes enhanced resist exposure due to electron scattering in the resist and backscattering from the substrate. Since good edge definition requires high resist contrast, the proximity effect can substantially alter developed pattern shapes and fidelities. This effect increases as pattern sizes decrease and becomes rather severe for submicrometer geometries. We have explored methods of compensating for the proximity effect in submicrometer patterns exposed on a vector-scan exposure system. This paper discusses two approaches which can be utilized to process figure specifications prior to exposure time to compensate for the proximity effect. We also show how these approaches fit into an overall program of pattern specification and exposure.
A two step fabrication process was developed which has produced previously unobtainable high frequency surface acoustic wave (SAW) devices. In addition, the design-to-test cycle time has been shortened significantly to allow an effective interactive design procedure. The SAW structure places some of the most stringent precision requirements on current electron-beam lithography since finger placement errors are directly related to phase errors in the electrical performance of bandpass or pulse compression filters. A form of double precision is implemented in the interface software to enable successful patterning of the monotonic variation in line width from less than 0.4 µm to more than 0.9 µm required for a 1 to 2 GHz pulse compressor. Compensation for proximity effects due to the electron beam profile was also implemented. Alignment accuracy within a field is controlled to within ±250 A and field placement is accomplished via a laser interferometer controlled stage. Several alternative processes, including direct slice writing, reverse lift off, and X-ray lithography are compared. Once the E-beam master is generated, large area contact replication is achieved using a modified conformable mask printer. This process has extended the range of SAW device performance beyond 2 GHz in a fundamental mode which represents a significant advancement in microfabrication. Three day turnaround from design to packaged devices was demonstrated using this technique.
Owing to restrictions determined by the laws of diffraction, electronic and X-ray systems are gradually replacing optical objectives in microelectronics for imaging. It is, however, interesting to analyze the means by which the performances of optical solutions may be increased by working with ultraviolet rays at shorter wavelengths, enhancing the chances of producing the optical objectives at lower costs than the electronic and X-ray objectives. Ultraviolet objectives of short wavelengths need catadioptrical systems. Substances presently used for Transmission are unsuitable for calculations of combinations in high resolution ; they are in fact, limited in Transmission in the ultraviolet of Lyman. The possibility of using a catadioptrical objective obtained by a holographic recording of a fringes system on a concave spherical support was announced by our company in 1975. Since then we have demonstrated how such systems can produce aplanatic images. (Variations configurations of this type have been calculated.) Today's known solutions for the problem give a resolution of 0.5 µ on a limited field. But new solutions involving several catadioptric holograms are in progress. We have manufactured such objectives and we show tests of resolution in problems of mask making in microelectronics.
This paper will describe the development of two ultra-high resolution, 1/10x reduction lenses corrected for the 365nm. mercury line. Designed for adaptation to current step and repeat cameras, they are suitable for use in the production of SAW filters, microwave circuits, bubble memories and LSI devices requiring sub-micron line geometry. The paper will discuss the problems involved in designing a 0.40 N.A. lens at 365nm. which has both high transmission and diffraction-limited performance over a 4.2mm. field. Special equipment and techniques are also required during the optical and mechanical manufacturing phase in order to produce a lens which will retain the high level of image quality inherent in the design. Testing of these lenses presents another challenge because critical visual inspection is not possible at 365nm. Image quality was verified by measuring the Optical Transfer Function using a very precise knife-edge scanner. Curves showing the measured OTF out to 1400 cycles/mm. will be presented. If available, photographic tests and/or examples of circuitry produced by these lenses will be shown.
Good evening and welcome to our panel discussion this evening, on the subject of "1 to 3 μm Lithography: How?" 1 to 3 micrometers is a range we have been talking about for years. Back, in 1963 when I first got into the industry, people were talking about doing one micron geometries. Personally, I was a little bit awed by this, because I had never been able to make them. I didn't know anyone personally who had ever made them. But everybody was talking about it. Well, suddenly we find ourselves in the situation that we honestly have to start making geometries in the 1 to 3 micrometer range. We are going to have to start doing it in large volumes, economically, soon. And that is the subject we would like to talk about tonight. For our panelists this evening, we have Peter Moller, the Director of Marketing Semiconductor Products, Electro-Optical Divi-sion of Perkin-Elmer Corporation; Bill Tobey, the Director of Marketing of GCA Corporation, Burlington, the manufacturer of DW Mann Products; Norm Austin, the Vice President and General Manager of the Industrial Products Division of ETEC Corporation; Sam Harrell who is Vice President and General Manager of Cobilt Division of Computervision. Tonight's program will start with introductory remarks by each of the panelists. Then I would like to invite all of the members of the audience to participate by asking questions and making comments. Remember, we are all in this thing together. Which is a bit frightening. Believe it or not, these guys are up here to make and sell products. For the most part they are going to make the things that we want to buy, or at least they are going to try to. So it is up to us to tell them what we want; to ask the questions that need to be answered and make suggestions that need to be made. So I guess what I am saying is that"Now is your chance." You have got them all together. They can't tell you how much better their equipment is than the other guy's, because the other guy is sitting right beside him. (I'm not trying to get dirty, I am just trying to get honest.) Anyway, with that as an in-troduction we'll go ahead and let the panelists start out with their in-troductory remarks. They are going to be talking about where they think the industry is going and what their companies can do to help. Then I expect very active participation from the audience. After all, you are the guys who have to do it, and you can't do it without help. Now the members of the panel represent the companies that make the imaging equipment. Not represented on the panel are the people who do the processing and who make the materials. To fill those two voids I would like participation from the audience; those of you who have been involved in Plasma Etching, Ion Beam Milling, the production of photo resist or electron resist products. We can't have everybody on the panel, so I would like very heavy participation in those areas from the audience. So why don't we start off with Peter Moller?
As geometry sizes decrease and mask sizes increase, the importance of high speed automatic mask inspection is increasingly evident. High speed scanning of large surfaces, while recording the location of artifacts less than 1/10 the size of a hair in less time than it takes to talk about it, presents some formidable problems for the optical architecture of the system.
The semiconductor industry is presently striving to increase the information density as well as the die size of most new integrated circuits. This goal, coupled with the simul-taneous shift towards projection and proximity printing, has dramatically increased the need for more precise manufacturing and quality control techniques. Both projection and proximity printers suffer from a very limited depth of focus. This focal length "budget" varies from printer to printer, depending upon the basic system design, as well as any residual image or alignment errors. The ability of these optical systems to reproduce high spatial frequency information is intimately tied to the flatness of the photomask reticle and silicon wafer upon which the information is to be printed. Specifically, this paper will deal with the quantitative analysis of surface flatness and how it relates to the flatness measurement of masks and wafers. Classical as well as contemporary surface testing interferometers will be described in order to exemplify and explain the observed fringe patterns. Fringe analysis methods will be described, including an example of the most advanced computerized techniques. This paper concludes that surface testing interferometry will become the "work horse" of flatness measurement in the semiconductor-micro-lithography industry, just as it has in the optical industry.
Four line-width measuring systems were tested for variations in the measured width of identical resolution targets as a function of the clear-to-dark (C/D) ratio of the surrounding field. Test masks were made of resolution bars in a checkerboard field. The C/D ratio of this field was varied on each mask to create eleven samples per mask with C/D ratios from 0% to 100%. Photocomposition was used to ensure that the C/D ratio of the field did not disturb the dimensions of the resolution bars during manufacture. Thus, any dimensional variations from sample to sample were due to effects of the C/D ratio on the measurement process. This was verified in a scanning electron microscope. Masks produced in both chromium oxide and iron oxide were measured on an image-shearing system, a photometric system, and on two different TV-type systems. One TV-type system was affected by the C/D ratio. This system was tested further to determine the cause of the problem. A method of minimizing the problem was determined.
An optical scanning microscope system for accurate measurement of linewidth on wafers is described. The development of this system in both theory and experiment parallels the treatment of transmitted light measurements on see-through photomasks as previously described. Threshold equations for determining edge location have been developed which include corrections for contrast and phase. In reflected light, the wavelength dependence of these parameters due to thin film interference requires a much narrower spectral bandwidth. In addition, other light losses dictate the use of laser illumination. Control of the spatial coherence of the laser is discussed. A comparison of theoretical and experimental results is given along with a comparison of reflected and transmitted light measurements.
Integrated circuit processes for producing horizontal geometries can be effectively monitored by precise linewidth measurements in both x and y directions. This paper describes a noncontact dimensional measurement system that is fully automatic and has a precision of ±0.025 µm for symmetric lines from 1 to 3 µm wide--e.g., 5-kÅ thermal oxide over silicon structure. Based on the analysis of diffraction patterns, the system operates on a wide variety of well defined line structures. By suitable choice of target, this system can also be applied to the measurement of mask-wafer overlays. The performance of the system for large numbers of samples is described and compared to the classical linear input. The signals are the same for structures having the same linewidth but produced by different processes. For some thick semiconductor structures, the edges defining the pattern boundaries have significant width themselves. Methods for measuring such struc-tures precisely by analysis of diffraction patterns are described.
Submicron photoresist imaging technology is discussed in terms of its use in the production of semicon-ductor devices. Photoresist processing techniques necessary for one micron work are reviewed, and optimization techniques presented. Recent developments in semiconductor technology that contribute to sub-micron photolithography are covered, and all major process fabrication steps are reviewed. SEM's showing one micron range images in positive photoresist, along with parameters used to generate them, are presented. Finally, recommendations are made for establishing a one micron range photoresist imaging process.
A new technique for generating registered patterns on opposite sides of a substrate has been developed. Rather than introducing the patterns sequentially and registering them with an infrared or dual viewing microscope, the photomasks for the two surfaces are pre-aligned and assembled into a hinged configuration referred to as an "alligator" mask. Substrates can then be inserted into the "alligator" mask and exposed on both sides without on-line alignment. Since a registered set of exposed patterns now exist, subsequent processing can be performed simultaneously on both surfaces. "Alligator" masks have been utilized for silicon wafer diffusions at Western Electric Company, Allentown, Pa. since July, 1976 and produced devices with top-to-bottom registrations better than ±10pm. Device registration is a function of initial mask alignment and hinge fatigue (fatigue occurs only in direction orthogonal to hinge axis). With improved mask fabrication, better top-to-bottom registration should be possible. An average mask lifetime of 766 wafers has been observed with mask wear being the primary failure mechanism. The exposure system, designed and built by Western Electric Company, is capable of exposing two 1.5" wafers on both sides with a beam divergence of <1° and a beam uniformity of 10%.
The exposure of certain positive photoresists has been shown by Dill and co-workers to be modelable in terms of a local inhibitor concentration which results in a local development rate. The development process is assumed to be a surface etching reaction in which the surface velocity is the local development rate. The characterization of resists for the purpose of line-edge profile simulation therefore involves the measurement of development rate for controlled exposure dose profiles. In this paper a technique is described in which the resist thickness is continuously plotted during development. The low frequency capacitance is measured using a conducting substrate as one plate, and the highly conductive developer as the other plate of the capacitor. The inverse capacitance, proportional to the composite resist-oxide thickness, is obtained using an analog divider. Examples of standing wave effects in Shipley Az 1350 resists, and resist development rate modification using chlorobenzene are presented.
Yield maps provide a powerful tool for the location and analysis of mask defects. The routine use of yield maps has not been widely accepted due to the requirement for involved statistical computations. Empirically justified approximations are derived for simple, quick yield map analysis. The physical significance of these approximations is discussed both graphically and heuristically.
The adhesion of chromium films to various glass substrates has been studied in various glass composition. 16 types of glasses were prepared by substituting B2O3 for SiO2, Al203 for CaO, and RO and R2O for CaO respectively in an almino-silicate glass. The adhesion of chromium films is evaluated by the so-called "scratch test" and observation of pinholes generated by ultrasonic agitation. It was found that a) with increased content of B2O3, pinhole density, ρ, increased and critical load Wc, at which film was stripped from substrate by scratching, decreased, b) with substitution of RO for CaO, ρ became larger in the order of CaO < MgO, ZnO < SrO, BaO, PbO, and Wc decreased in the order of MgO > CaO > ZnO > SrO, BaO, PbO, c) with R2O3 substitution, ρ decreased in the order of K2O, Na2O, Li2O. Thus, it is clear that B2O3, BaO, SrO, PbO, K2O and Na2O in glass contribute to poor adhesion of chromium films. It was also fauna that HOYA LE-30 glass had excellent adhesion to exceed an ordinary alumino-boro-silicate glass.