It was long thought that 1μ represented the limit of conventional optical lithography, and that the so-called "advanced lithographics", i.e. x-ray, E beam, would be required to do sub-micron work. It is now evident that conventional optics can be pushed below the 1μ barrier, but can sub-micron geometries be realized in a practical production environment? Or do they still belong in the realm of the laboratory? "Trade-offs" can be made in reduction ratio's, wavelength corrections, N.A.'s, field sizes, and etendue (optical efficiency), but can they make optical sub-micron lithography a viable production alternative? These questions are addressed and images from some of today's most advanced lenses are examined.
The dependency of photoresist line width control on the photoresist thickness variations has been quantified for three different microelectronics wafer surfaces. The exposures were made with a step-and-repeat projection camera which is designed to image with both the Hg-g and Hg-h spectral lines to the wafer surface. A filter was placed in the light path so that resist exposures with only the Hg-g line could also be studied. The effect of camera defocus and nominal photoresist thicknesses of 800 and 1600 nm was included. The expected result of the line width variation periodicity being equal to 1/2 of the exposing wavelength was found. However, also found is an unexpected periodicity which could be described as λR/6
The demand for higher yields in IC fabrication, economic pressure to realize the potential of projection aligners, the requirement of defect-free reticles for water steppers, and the need to reduce mask handling on the production line have generated. considerable interest in pellicle protection. of IC masks and reticles. In this paper, the mask protection problem is examined with emphasis given to the relevant optical requirements of pellicles for various projection. systems.
This paper describes a two-dimensional optical encoder that is used to control the x-y stepping stage of a modern state-of-the-art wafer stepper. Stage location is determined by measuring the position of a quartz grid plate that is attached to the stage. Accuracy and resolution of measurement are maintained by varying the null position of the grid. Operational principles and actual performance are discussed.
Automatic alignment by laser beam scanning method is known as a useful technique for both contact and proximity printing, and is now put in practical use. This method utilizes a laser spot as an optical probe which detects the information of a mask and a wafer simultaneously. Laser beam forms a small spot on the objects, and scans them at a constant velocity by a rotating polygonal mirror and F-θ optical systems. The alignment marks on the objects are placed along the scanning line. In order to increase the signal-to-noise ratio of the object information, a part of the scattered light from the marks is selectively detected by optical filtering method. As a result, high detectability of wafers information and highly accurate automatic alignment can be achieved. This laser spot scanning method is easily applied to the projection printing aligners too, such as lens or mirror system. It has an advantage to make full use of the characteristics of the projection system. For example, the interference phenomena between a mask and a wa-fer can be eliminated by the use of a quarter-wave plate or phase-shift mirrors in combination with the polarization characteristics of a laser. Cosequently, stable output signals are obtained, and the alignment accuracy is improved to 0.3μm (2σ).
The technological window for X-ray lithography is examined. What seems to narrow the technological window for X-ray lithography and remove the economical advantage is not a symptom of X-ray lithography but of the conventional means of generating X-rays. The type of source needed is an intense, pulsed X-ray source of suitable wavelength range. Although laser generated and other types of plasma generated, pulsed X-ray sources are being investigated, a plasma focus device has been ignored as a possible intense, pulsed, soft X-ray source which can satisfy the requirements of X-ray lithography. Here, means of generating an intense X-ray pulse from a plasma focus device which is suitable for lithographic needs is analyzed. Estimates of energy conversion efficiencies, spectrum of radiation, duration, and intensity of the radiation pulse are presented. Optimization, scaling considerations, and engineering solutions to anticipated difficulties are discussed.
A compact, efficient, high-brightness pulsed plasma x-ray source has been demonstrated. The source has a conversion efficiency, from electrical input to usable x-ray energy, of greater than one percent. X rays are emitted from a cylindrical volume approximately 1 mm in diameter and 10-mm long. Various wavelengths can be obtained by the choice of appro-priate plasma materials. Submicrometer features have been replicated using x-rays in the 12 A region to expose a wafer, coated with FBM resist (~ 50X PMMA sensitivity), through a polyimide/gold x-ray mask. The exposure was made through a 25-μm-thick beryllium filter with the wafer 30 cm from the source. Twenty x-ray pulses were required to fully expose the resist. For this experiment the available electrical pulse power supply was limited to one pulse every few minutes. However, the x-ray source itself is capable of much faster pulse rates, and electrical driver technology to power the x-ray source at one or more pulses per second is within present state-of-the-art.
The electron beam exposure systems (EBES) currently used for photomask fabrication have been designed to operate to positional accuracies of ∓ 0.031 micron over an address range of 0.20 to 1.00 micron. The tuning, calibration and characterization of these systems has required an exacting and time consuming series of tests. A standard test array, MARKET A15, has been developed which enables full system characterization with only a thirty minute master generation time'and four hours analysis. System parameters such as linewidth control, resolution, registration, absolute accuracy and pattern scale accuracy can be determined from a single plate. The tests are composed of a series of exposures designed for analysis under an optical microscope along with a second group requiring reinsertion of the plate into EBES for "self-analysis". This array can be used as a basis for tuning, monitoring and multiple systems calibration of EBES or EBES-like rastor scan systems, A discussion of tests and techniques demonstrates how several EBES type systems in Western Electric and Bell Laboratories have been brought into mutual compatibility.
This study describes the results of feature size distribution, pattern location accuracy and level to level registration error on chrominum master masks fabricated by EeBES-40. This system has the capability of high speed electron beam blanking at 40MHz, the capacity for large size masks (with 6 inch mask cassette), and the automatic cassette handling system. OEBR-100(PGMA), as the electron beam negative resist, is used for 5 inch and 6 inch chrominum masks. The chrominum etching process is used for both wet and dry plasma technology. Test patterns and 64 K bit memory TEG, as the practical pattern, are used in this study. More than 40 measurements are taken, uniformly distributed over 96 to 112mm square, and the feature size distribution is measured by a laser interferometer X-Y measuring system. Pattern location accuracy and level to level registration error are obtained using EeBES-40 quality assurance programs called MARKET/PLOTMARKET. This program operates by scanning over the resist image of the test pattern, utilizing the normal fiducial mark location hardware. The followinc results are obtained; (1) Feature size distribution within 6 inch mask : ∓0.1 μm (2) Level-to-level registration error2 : less than 0.1 pm High quality masks with about 0.02 defects/cm2 , and rapid throughput of 6 hr./10 masks using the auto-matic 10-cassette handling system are obtained.
X-ray masks being fragile and expensive, we elaborate a duplication method giving replicas exactly iden-tical or complementary. Using the X-ray transparency of X-ray mask substrates, we make in a simultaneous exposure two replicas of a master-mask, their polarity being determined by that of the two used resists. A careful calculation ensure a correct exposure of the two resist layers by synchrotron radiation. Examples are given for 2 positive or 1 positive 1 negative resists. The polarity inversion of PMMA is also used.
Two commercially available automatic mask inspection machines, the KLA and the NJS, were compared for sensitivity to both clear and opaque defects. The sensitivity comparison showed the KLA 100 at 2 units to be similar to the NJS 2MD1 at 60X. The KLA 100 at 1 unit is similar to the NJS 5MD23 at 1.5μm. At each sensitivity range, roughly comparable percentages of false calls were encountered for each machine. The matter of false calls (on false reports of a defect at a location) is not clear-cut. Machine adjustment with regard to light balance, focus and registration of the machine to the mask are extremely significant in this regard. Additionally, as design rules shrink, the presence of features on polygons that are near the size of the defect being sought create severe problems with regard to false calls. The correlation of visual sampling results to automatic inspection results showed two things: (1) individual plate sample results correlate poorly to the plate quality as it is determined by automatic inspection, (2) the average quality of the plates produced in a week shows a correlation of the automatic and the visual results that is much better than that of the individual plates. The mathematics of sampling has been examined. It has been folind that the visual quality vs. automatic inspection results can be constructed from models which predict the experimentally observed pattern.
As VLSI technology pushes toward micrometer range geometries, wafer flatness becomes critical to direct step on wafer projection printing. This paper describes a system that utilizes the automatic focusing capability of a wafer stepper to achieve in situ measurement of wafer flatness. A depth gauge attached to the lens column conveys to the system a voltage analog of column height during exposure of each die. The system then computes locations of isoplanar contour lines on the wafer and plots a topographic map on an X-Y recorder. Any vertical interval between contour lines can be specified between 0.5 and 2 micrometers. A close correlation has been found between interferometric images of wafer flatness and topographic maps produced by the system while attached to a wafer stepper. The chief advantage of the wafer topographic map over the interferometric image is its ability to capture flatness measurements during actual exposure. Also beneficial are tick marks on all contour lines which indicate direction of increasing wafer height. Application of the system will enhance prediction of photolithographic yield, identification of process steps causing wafer deformation and effectiveness of process quality control.
Quantitative determination of sub-micrometer linewidths in semiconductor devices and masks is demonstrated utilizing an approach employing complementary experimental measurements and theoretical modeling. Experimental measurements are performed utilizing the Microlength Calibrating Electron Probe (MCEP), a new facility at the National Bureau of Standards consisting of a scanning electron microscope modified to incorporate a scanning stage and laser-interferometer position measurement system. Automated data acquisition and analysis for the MCEP are achieved through interfacing to a laboratory minicomputer. Theoretical modeling based on Monte Carlo calculations provides a basis for selection of the position in the experimentally measured backscatter electron intensity profile that corresponds to the actual material line edge. A measurement on a photomask is shown which illustrates the utility of the MCEP facility and the Monte Carlo modeling calculations for accurate measurement of sub-micrometer linewidths.
Photoresist linewidth variations are particularly difficult to control in the vicinity of steps where resist thickness changes occur. These variations are primarily due to the interference phenomena which affect the coupling of energy into the resist and vertical standing wave intensity contrast in the resist. In this paper the user oriented program SAMPLE is used to quantitatively explore linewidth control in the vicinity of steps as a function of exposure, resist, and substrate parameters. Nominally, projection printing of 1.0 um line and space patterns in 1.0 urn of AZ13.50J resist crossing 1.0 urn steps on silicon, oxide and aluminum substrates is considered. The linewidth variation, resist profile shape and linewidth control sensitivity are characterized by the behavior of the two extreme cases. These are the maximum linewidth which occurs at the largest resist thickness on the bottom of the step corresponding to a destructive interference node and the minimum linewidth which occurs at the smallest resist thickness on the top of the step corresponding to a constructive interference node. The significance of the aerial image contrast and substrate reflectivity are explored. The possibility of using increased optical absorption and post exposure bake procedures to improve process control are considered. The use of multiple wavelength exposures with 6 raction limited optics is shown to significantly improve both linewidth control and the resist profile shape. The degree to which a refractive optical system can realize this improvement can be limited by the extent to which the change in focal error with wavelength reduces the contrast. This limit is particularly important in assessing the potential advantages of multiple wavelength systems, because the performance of such lenses is expected to be limited by chromatic aberration effects.
The purpose of this paper is to discuss the practical trade-offs of electron beam direct-write and optical projection techniques in terms of resolution and linewidth control for device geometries in the one micrometer feature region. Overlay, field size, throughput and other economic issues will not be addressed. It is generally understood that the overlay accuracy of E-beam direct-write is the bestl among the various lithographic approaches.
Edge location in very high-resolution images may be affected by the statistical distribution of discrete grains that collectively form the opaque regions. An ideal case is considered in which a one-dimensional edge, represented by a step probability density function, is characterized by the location and contrast (magnitude relative to a background level) of the step discontinuity. This function is randomly sampled at a number of points, and a maximum likelihood criterion and a Monte Carlo technique are used to estimate the edge location from the samples. The root mean square deviation of this estimate from the actual edge location is obtained as a function of the edge contrast, the actual edge location, and the number of samples. This simulation is applicable to a class of single straight-edge images for which the number of samples is interpreted as the number of grains in the image and for which the edge orientation is known. 4-reNAyin -inn
Linewidth control using a tri-level resist system on wafers with topology is investigated. An absorbing dye is incorporated in the bottom layer to improve the usable resolution. Resist patterns of 1 μm and 0.75 μm over underlying geometries are demonstrated using a projection aligner. The advantages of a multilevel system are investigated using an exposure and development simulation program for optical lithography. The relative contributions of planarization and reflection suppression are discussed.
High resolution optical lithography demands a sophisticated photoresist technology. Recent work in this arfa has concentrated on multilevel resists. An example is a two layer resist (TLR) technique which uses a thin top layer of positive resist over a thick layer of PMMA. The pattern formed in the thin top layer serves as a mask for a deep UV blanket exposure of the thick PMMA bottom layer. Previously, no deep UV source existed which was well suited to this application. This paper describes the development of such a UV source and its application to high resolution optical lithography. The TLR technique requires a UV source with output in the 200-230nm range and essentially no output below 200nm or between 230 and 300nm. A specially designed low pressure Cd arc lamp has been constructed for this application. The lamp is unique in the way the output is maximized in the 215-230nm range. The lamp is very efficient, converting 10% of its input into output at these wavelengths. Using the new lamp, resist (PMMA) structures of 0.5μ width having 3:1 vertical:horizontal aspect ratios have been fabricated. Such patterns have been used as etch masks for CC14 plasma etching of 0.5μ patterns in polysilicon. The large aspect ratio of the resist patterns was important in preventing resist degradation (during etching) from affecting the final polysilicon geometry.
Two of the most challenging problems confronting processes aimed at 2μm pitch minimum geometries are: imaging on highly reflective,grainy metals and the subsequent anisotropic plasma etch of aluminum or aluminum alloys. Metal lift off processes can eliminate these difficulties altogether by reversing the normal procedure for metalization. "Metal Lift Off" refers to processes in which the metal or metal alloy layer is deposited over the photoresist image and then the unwanted metal is removed by dissolving away the photoresist and lifting off the excess metal by suitable mechanical means (Fig. 1).
The degradation of PMMA, irradiated with a Philips deep-UV source, was studied by means of gel permeation chromatography (GPC). A steep increase of the dissolution rate in MIBK was found with only a small decrease in molweight. This result supports Ouano's hypothesis of pore formation in PMMA upon E-beam irradiation, due to chemical degradation of the polymer. The sensitivity and contrast value of PMMA and PMIPK was determined, using 1 min. dip development in2MIBK. The minimum deerUV dose (200-300 nm.) needed to clear the film ranged from 750 mJ/cm2 for PMMA to 65 mJ/cm2 for sensitized PMIPK, with contrast values between 2.5 and 3.7. The plasma etch resistance of PMMA increased five-fold when the resist was mixed with an aromatic polymer, without loss of resolution or contrast. The sensitivity, contrast, resolution and plasma etch resistance of PMIPK based resist could be made as good as or better than that of existing positive photoresists.
Positive and negative photoresists are contrasted. Twelve parameters are listed with evaluation as to which photoresist gives superior results. Emphasized are resolution and dimension targeting. Reasons are given stating why positive gives better resolution and why dimension targeting is used.
Waycoat WX-118 is a new positive photoresist characterized by high photospeed and good resolution. WX-118 is highly responsive to Deep UV radiation (270-320 nm) as well as Conventional UV (>320 nm). Details are presented on this resist in regard to: 1) absorption spectra, 2) film thickness vs. spin speed, and 3) optimum softbake and hardbake conditions. WX-118 and Hunt's HPR-204 positive resists are compared in conventional UV exposure, both in contact and 1:1 projection. SEMs are presented to illustrate the superior resolution capabilities of WX-118 in a 1:1 projection system. The Deep UV data was obtained on a UV-3 Micralign 1:1 projection system where AZ-2400 was used as a comparison. SEMs are used to present information on the unique resist sidewalls obtained with WX-118 exposed on a UV-3 Micralign projection aligner.
Recently there has been a tremendous increase in positive photoresist technology to manufacture fine line semiconductor devices. However capabilities of current positive resist can not satisfy LSI production requirements, especially on linewidth control. In this paper new positive photoresists, AZ-1450J, HPR-204 and OFPR-800 were studied to minimize linewidth precision of MOS dynamic RAMs using 1:1 scanning projection systems. In order to optimize resist linewidth characteristics of new materials, i.e. exposure sensitivity, striation, chemical stability, wet developing, resistance for new dry etching, removal and impurity, were evaluated for each material. And the availability was confirmed by application to the test device for MOS dynamic RAM. This is including 128 X 10 transistor arrays with various channel length and possible to measure the threshold voltage imbalance of transistors by scanning of a decoder circuit. Consequently it was found that the channel length variation decreased within 0.1 1111 ( 3o-'). and the threshold voltage variation was ∓ 50 mV ( 3σ ) or less for 2 μm line using AZ-1450J and OFPR-800.
A commercially available positive photoresist, Polychrome 129 SF, has been evaluated as both a positive electron beam resist and photoresist. The photoresist sensitivity, measured uniquely for a given developer based upon image dimensional control, is determined to be less than that of AZ 1350, under as nearly equivalent conditions as possible. The E-beam resist sensitivity is lower than that of PMMA reference, but the resist possesses good resolution. Near vertical edge wall profiles are obtained for 1-micrometer lines, but no undercutting is ever achieved regardless of electron charge density magnitude. The resist is capable of submicron line and space E-beam resolution, and 0.50-0.75 micrometer wide isolated line patterns can be routinely achieved.
The production of today's 2-3 micron line widths with the expectation of continuing yield improvement can only be accomplished by the use of automated wafer processing equipment. Data is presented on a new positive system that combines the high resolution advantages of positive photoresists with automated in-line puddle development. In the puddle development process the imaged wafer is developed on an automated wafer processing track, with WX 108 developer. Process constraints of film thickness loss and time of development were established to allow investigation of the following variables; softbake conditions, exposure energy, and developer temperature. It was found that a 50% variation in developer dwell time caused less than a 4% developed image deviation from the mask.
The curing or prebaking of photoresist-coated substrates using microwave radiation has several characteristics that are useful when processing LSI circuits. Some properties of microwave energy, the process results obtained with its utilization, and the advantages for photoresist baking will be presented.
In response to the need for moderate cost, totally dedicated plasma etch systems a Technics Planar Etch II plasma reactor was evaluated for the etching of micron-range geometries in 5000Å of polysilicon, phosphorous doped to 25-30Ω /0. Initial characterization included the use of a vacuum throttle in conjunction with different vacuum pumps and wafer support mechanisms to determine the most optimum gas flow conditions for uniform clearing of un-patterned wafers. A variety of gases were then employed for the etching of one, two and three micron lines and spaces using KTI Type II positive photoresist in all applications. Each etchant was operated at various pressures and RF power levels to determine absolute etch rates and sensitivities of rates to power and pressure fluctuations. Underlying oxide etch rate, rate of photoresist attack, uniformity in clearing, etched-edge profile, degree of undercut and cleanliness of exposed surfaces following etch were all monitored. It was concluded that none of the etchant gases utilized in itself lead to a set of conditions adequate for maintaining one micron geometries with no loss of underlying oxide, but rather a dual gas sequential operation was established which satisfies all requirements by utilizing an initial gas that etches uniformly with minimal undercut followed by a second etchant for clean, uniform clearing with no attack of the underlying gate oxide. The Technics Planar Etch II is conducive to etching 111 geometries in 5000 Å polysilicon films on a routine basis. Batch size is not large, eight 2" wafers/run, but the system is in a price range compatible with multi-unit, dedicated installations.
Solvent and plasma cleaning of RF sputtered, DC sputtered, and vacuum evaporated chromium-chrome oxide films on soda-lime glass substrates was investigated. One-cycle cleaning was evaluated using KLA automatic inspection. Long-term (eleven-cycle) effects on critical dimension, optical density and reflectivity by each process were measured. Deposition type had no effect on cleaning efficiency, although each did show long term effects on linewidth and optical parameters. RF sputtered chrome reflectivity increased drastically with all types of cleaning. Chrome lines grew, as did the optical density. DC sputtered chrome demonstrated a decrease in density in some cleaning methods, while vacuum evaporated films varied in linewidth.