A significant departure from traditional ways of viewing mask quality will be necessary in order to have a viable mask manufacturing technology for sub - 0.5um optical lithography. It will be necessary to evaluate mask performance in the context of the overall IC manufacturing process. This paper will look at the limitations of present state-of-the-art mask specifications and suggest that a new approach to mask verification is needed. Instead of making apriori and somewhat arbitrary decisions about what masks need to look like, we must have a method of determining which mask errors are of concern and which are not. Ultimately, we can exploit our mask technology to compensate for errors created by our IC manufacturing process.

To realize higher CD controls of submicrometer devices, the submicrometer pattern corrections were investigated in optical reduction steppers considering the primary residual aberrations. The optical pattern fidelities on the reduction pattern transfer were estimated at first using the three-dimensional photoresist image simulator RESPROT (Resist Process Three-Dimensional Simulator), which is examined the Seidel's primary aberrations, i.e. spherical aberration, astigmatism, field curvature, distortion and coma. From RESPROT calculations it was known that astigmatism affects pattern shape depending on image height, coma and distortion make position shifts in exposure field, and image contrast is influenced by field curvature. These results were reflected to device design rules, process latitude enhancements and lens manufacturings. To use premature high NA g-line lenses and minimize the diffraction limit for submicrometer area, reticle pattern corrections are very useful for sub-space patterns writing in contact holes, "tailoring" of W/L for MOS-gate patterns, and sub-field position control for distortion correction on ER writing. For almost of this investigation, 5-10nm order corrections were required from original design on wafer. In order to make good use of higher NA lenses, focus latitude enhancement are required because field curvature control is very severe. For these demand, multi-step superpositions of focus and exposure, FLEX -method, is useful to enhance the depth of focus effectively. These simulated and examined results are covenient to realize submicrometer devices on advantageous optical lithography using more shorter wavelength, i.e. i-line or excirner laser steppers.

Experimental results of a high-resolution holographic imaging system using wave-front conjugation are demonstrated. The hologram is created from the information on a master mask that is imaged with a simple single-element lens. The aberrations introduced by the lens are eliminated because the rays retrace the path back-through the lens. The theoretical resolution in the fully coherent imaging system is λ / 2 NA = 0.61 μm ( λ = 488 nm, NA ~0.4). Sub-micrometer resolution of the reconstructed aerial image is experimentally observed over the entire 1 cm diameter field.

Carbonized novolac photoresist has been demonstrated as a patching material for clear defects in optical masks. The processing is simple and uses commonly available equipment. The patches, though not as tough as chrome, adhere well and survive mask cleaning. Mask materials appear to survive heating up to 1000°C, but distortion testing has only been done up to 500°C. Adhesion of the carbonized resist begins to degrade above 700°C processing temperature.

A key challenge in integrated circuit manufacture is to transfer images onto silicon wafers without introducing defects. Image qualification (IQ) is the technique used to assure that no repeating defects are transferred from the reticle to the lx wafer image. Increasing design complexities and smaller feature sizes make automated inspection a necessity. The most common IQ technique involves imaging a metal-coated glass wafer (aluminum 1 or chrome2) and automatically inspecting it using traditional mask and reticle inspection equipment. This method requires additional processing steps (metal etching and stripping) not commonly employed in a wafer fab area to prepare develop-inspect samples. A new technique is now available which more closely matches standard wafer fab processing and eliminates the need for metal etching by allowing direct automatic inspection of developed resist images on a quartz wafer (ROQ) substrate. In this paper, the ROQ equipment and process are discussed and inspection results are compared to conventional metal film results. The concepts of defect fidelity function and effective defect capture rates are used for comparison between the two techniques4.

Design, mechanization and performance of a through-the-lens focus control system now in use by the mask-making facility at the IBM Essex Junction facility on their most advanced step-and-repeat cameras are discussed.

Raster scan lithography systems, such as scanned-laser and most E-beam mask writers, produce images through a mosaic of discrete picture elements (pixels). Image qualities of the printed mask (or wafer) are governed by the interplay of several printing variables, including size and shape of the writing spot, pitch and orientation of the pixel grid, relative intensities of pixels, and exposure characteristics of the resist. We will review the theoretical foundations of raster imaging and show how these variables affect several key measures of lithographic image quality, including minimum feature size, edge placement resolution and accuracy, dimensional uniformity, and edge roughness. We will present three quality enhancement techniques and compare their performances to that provided by a basic raster printing scheme where pixels of binary (on or off) values are printed on a cartesian grid. The first technique involves rotating the printing grid 45 degrees to the main axis of the data coordinate system. We will demonstrate that, for lithographic images where most edges are parallel to the data axes, this grid provides 41% more addressable edge positions than a non-rotated grid with the same pixel density. The second technique, adapted from computer-graphics "antialiasing" applications, involves modulating the intensity of pixels along the edges of features to finely control the shape of the aerial image. This provides a vernier mechanism for the placement of exposed edges between grid locations and results in finer effective addressability and smoother edges. Third, we will review how multiple pass printing (a.k.a. vote-taking) reduces random errors, and show how it also reduces systematic errors when certain printing parameters are alternated between passes. Finally, we will present a single printing strategy in which all three techniques are combined to yield high accuracy, high-resolution images with economic use of printing pixels.

With short wavelength light sources, VLSI chips with less than 0.5μm line width can be produced. In such case, alignment accuracy must be 0.1μm or less. Mechanical positioning accuracies of the stages, such as a wafer stage or a mask alignment stage, govern the total overlay accuracy. This report presents a mechanical improvement for the mask alignment stage, such as improvement in X,Y axes motion accuracy and realizing Z and tilt motion controls. A prototype compact and high precision alignment stage has been manufactured. In order to improve the X,Y motion accuracy, the stage has friction drive mechanisms and has a measuring method with analog/digital conversion of linear scale signals. In the experimental results obtained from the X,Y stage, less than 0.025μm motion resolution has been obtained.' No backlash, no stick-slip, and no hysteresis have been observed.

In wafer steppers, the alignment of an exposure field to the reticle being imaged is known to affect the success or failure of that field's circuit(s). Because of this relationship between alignment accuracy and device yield, much emphasis is placed on obtaining and consistently maintaining an alignment accuracy within tight design rules. In current wafer steppers, alignment options can be placed in two primary categories: Global Alignment techniques and Site-by-Site Alignment techniques. During Global Alignment, selected areas of the wafer undergo alignment. The grid of all exposure field positions is generated according to the information obtained during the selected alignments. Step-and-Repeat exposure of all fields is then performed according to this grid. In Site-by-Site Alignment (a.k.a. Field-by-Field Alignment), the stepper performs a step/align/expose sequence on each exposure field until all fields on the wafer are printed. While Site-by-Site alignment can result in greater overlay accuracy, Global Alignment is primarily used on all but the most critical of alignment levels. This is due to the much higher throughput that Global Alignment allows. Nikon has developed and implemented an original technique that achieves a throughput nearing that of most standard Global Alignment methods, while maintaining an overlay accuracy consistent with a Site-by-Site alignment scheme. This method, called Enhanced Global Alignment (EGA), utilizes the Laser Step Alignment (LSA) system of a Nikon Step-and-Repeat (NSR) system to measure the offsets of several selected fields. The NSR's computer constructs a mathematical model of the wafer's exposure field grid according to the measured offsets. The model includes six components that may contribute to overlay error: 1)Translation in X 2)Translation in Y 3) Scaling in X 4)Scaling in Y 5) Wafer Rotation 6) Pattern Orthogonality Error All exposure field addresses are determined during the EGA. Thus, there is no further alignment once the Step-and-Repeat exposure sequence begins. This sequence is performed on each wafer. The wafer model constructed by EGA relies heavily on the offsets measured by LSA. While the LSA system has a wide dynamic range (over which no variation in offset measurement accuracy is seen), the alignment mark topography must be constructed such that it can be detected by the LSA system. Since different processes may yield differing alignment topographies, an alignment mark size found to be optimum for one device manufacturer may not be best suited to another manufacturer with a different process. Optimization studies have focused on mark design variables that affect the LSA, while monitoring the process variables encountered and their effects. Discussions of these variables, including current recommendations, will be detailed.

In order for VLSI circuits to function properly, the masking layers used in the fabrication of those devices must overlay each other to within the manufacturing tolerance incorporated in the circuit design. The capabilities of the alignment tools used in the masking process determine the overlay tolerances to which circuits can be designed. It is therefore of considerable importance that these capabilities be well characterized. Underestimation of the overlay accuracy results in unnecessarily large devices, resulting in poor utilization of wafer area and possible degradation of device performance. Overestimation will result in significant yield loss because of the failure to conform to the tolerances of the design rules. The proper methodology for determining the overlay capabilities of wafer steppers, the most commonly used alignment tool for the production of VLSI circuits, is the subject of this paper. Because cost-effective manufacturing process technology has been the driving force of VLSI, the impact on productivity is a primary consideration in all discussions. Manufacturers of alignment tools advertise the capabilities of their equipment. It is notable that no manufacturer currently characterizes his aligners in a manner consistent with the requirements of producing very large integrated circuits, as will be discussed. This has resulted in the situation in which the evaluation and comparison of the capabilities of alignment tools require the attention of a lithography specialist. Unfortunately, lithographic capabilities must be known by many other people, particularly the circuit designers and the managers responsible for the financial consequences of the high prices of modern alignment tools. All too frequently, the designer or manager is confronted with contradictory data, one set coming from his lithography specialist, and the other coming from a sales representative of an equipment manufacturer. Since the latter generally attempts to make his merchandise appear as attractive as possible, the lithographer is frequently placed in the position of having to explain subtle issues in order to justify his decisions. It is the purpose of this paper to provide that explanation.

The impact of wafer stepper overlay errors on device yields and design rules are studied. First, the classical Lynch model for normally distributed sizing and overlay errors is reformulated for orthogonal geometries. Then the distribution of overlay errors in the linear Perloff model describing global alignment is derived. Finally, a Monte Carlo program, OVS, for simulating stepper overlay errors is introduced. OVS is used to determine the impact of individual component errors, such as those due to lens distortion or to mask making, on the overall distribution of errors.

This paper presents a general analysis of condenser aberrations in projection systems that employ Kohler illumination. We first analyze condenser aberrations in terms of Hopkins' theory of partially coherent image formation, and then discuss practical consequences for lithography. The intensity in an aerial image depends jointly on the object, the imaging lens, and the mode of illumination. The angular distribution of illumination isequivalent to the so-called effective source, and is determined by the actual source and the illumination optics. In order that image formation be the same for all points on the mask, the relationship between the entrance pupil of the imaging lens and the effective source should be the same for all points in the field. Condenser aberrations (including defocus) cause the effective source to vary with position in the field. Thus, it is appropriate to refer to the "local effective source". Changes with field position in the effective source cause variations in the imaging of a particular pattern. One phenomenon, which is well known, is that of lateral image shift with wafer defocus, caused by a lateral shift in the effective source due to condenser aberrations. This is usually analyzed geometrically, by considering a tilt, relative to the principal ray, of the illumination direction at each mask point. According to this geometrical analysis, the image translates linearly with defocus, whereas the more accurate analysis employing the shifted effective source shows a shift that is different and that is nonlinear. There are joint effects involving aberrations of both the imaging lens and condenser. For example, if the the imaging lens has coma, the image suffers a radial expansion that can be compensated by wafer and condenser defocus. In general, the plane of optimum image placement can be displaced axially from the plane of optimum resolution and the paraxial focal plane. Condenser aberration effects are entirely accounted for with the standard formulations for partially coherent imaging, so long as the correct effective source is used. Thus, these effects can be modeled with any partially coherent imaging program that permits an arbitrary effective source.

In the age of submicron optical lithography, focus has become a critical process parameter. Each decrease in minimum feature size is accompanied by a corresponding decrease in depth-of-focus (DOF). Sources of focus errors, such as wafer warpage, topography, and the thickness of the photoresist, however, are not being reduced in proportion to the DOF. Thus, the effects of focus on the practical resolution capabilities of a lithographic tool are becoming increasingly important.

Electromagnetic diffraction theory is applied to obtain a rigorous and comprehensive description of the imaging and exposure process in a projection optical system imaging a one-dimensional, periodic object in a planar layer of photoresist. The method is applicable to high numerical aperture and thick-photoresist systems, and accounts for the exposure dependent absorption characteristics of positive photoresists. It is used with the development simulator in SAMPLE to simulate the physical profile of the developed image. Theoretical and experimental results are given, which show asymmetrical variation of the developed image with focus. This asymmetry is found to depend on photoresist thickness, and the dependence is shown to be incompatible with the usual approximation of normal ray propagation in the photoresist.

Diffraction by small open lines in chrome (slits) has been measured and the data compared to scalar diffraction theory. The measurement system consists of a chopped HeNe laser illumination, a mask held by a translational and rotational stage, and a photomultiplier tube on a revolving arm connected to a lock-in amplifier under the control of an IBM PC XT computer. Results for slits down to a wavelength in size show transmission nearly proportional to slit size and only slight polarization effects. Initial plots of total transmitted power divided by slit size showed a decrease for small slits. However, correcting for mask bias and/or edge roughness shows agreement with rigorous electromagnetic theory.

The question as to how accurately small object features can be reproduced in optical microlithography does not have a simple answer. It depends not only on the dimensions of the feature, but also on whether it is a line or a space, whether there are other features nearby (the proximity effect), on the resist thickness, and on whether the features are to have the dimensions of the ideal optical image or are "biased". This paper explores these topics by modeling the imaging, exposure, and development steps. In order to discover the significant dependencies we first investigate the simple model of a resist layer deposited onto a non-reflecting substrate. It shows that an isolated resist line has approximately the correct dimensions in all sizes when the imaging is done with partially coherent light. The isolated spaces, however, show deviations from the design size, which are caused by diffraction effects. For spaces (0.6 - 1.0) λ/NA wide the light intensity in the center of the space is larger than the intensity of the incident light causing the resist to develop through faster than in a very large area For spaces smaller than 0.6 λ/NA the light intensity in the image drops rapidly and the spaces can no longer be reproduced. The dependence on coherence of the light, on resist thickness, and on the degree of focus are also investigated. On real sur-faces of silicon, Si02, and aluminum reflections cause interference effects and dramatic variations in exposure with resist thickness. After these effects are taken into account, the resolution is much poorer (0.9 λ/NA in the case of silicon, worse for aluminum). Biasing of the masks does not improve the resolution capability. On the other hand, a post-bake of the re-sist pattern can cause a dramatic improvement of the imaging quality and can increase the resolution to where it is comparable to that obtained on the non-reflective substrate.

A novel approach to printing large area integrated circuits (having a surface area larger than the nominal field and submicron geometries) through joining two or more stepped image fields was tested for feasibility. A "superfield" of 30 mm by 30 mm was printed which combined three exposures of a nominal 30 mm by 10 mm field on a lx stepper. Tapering the illumination across the common border region between the over-lapping "subfields" blended one into another. This technique also minimized the effects of misalignment and uneven exposure between fields and preserved the minimum resolution of the lens system in all areas of the stitched superfield.

Two-dimensional image simulation with SPLAT and resist profile dissolution simulation with SAMPLE are used to explore the suppression of large defects with voting in projection printing. The parameters chosen are for the Ultratech 1000 stepper due to the availability of experimental data with programmed defects. The results are stated in terms of feature and defect sizes in λ/NA where λ, is the wavelength and NA is the numerical aperture. Models based on 1) full resist dissolution simulation, 2) 30% clear field intensity of simulated images and 3) simple algebraic approximations are used. The case of using three votes on transparent defects placed at various locations in line patterns is studied in detail. An excellent fit between simulation and experimental results was obtained. It was also found that defect suppression with 3:1 voting is always a factor of 3 for small defects, and improves for larger defects except both when the partial coherence is less than 0.5 and when defects are within 0.2 λ/NA of a neighboring feature. Voting may be used to overcome mask defects which are so large that they have bridged patterns on the wafer. For 10% linewidth variation, the use of voting raises the critical defect size from 0.24 to 0.5 λ/NA. The effects of process bias and overlay are also explored.

Positive photoresists with actinic dyes are used rather commonly on highly reflective topographies. Non-actinic dyes either by themselves or in combination with actinic dyes are recently becoming available in positive photoresist systems for level-to-level alignment purposes. It is known that actinic dyes decrease the photospeed of a given positive photoresist system, however their effect on the dissolution rate of the exposed/unexposed resist is more complicated. For example, it has been shown that hydrophobic dyes may, through surface effects, decrease the ability of the developer to wet the surface of the exposed resist, thereby decreasing the apparent photospeed of the resist. The above effects, caused by the presence of the dye in the resist, can make establishing a given process for these dyed resists extremely difficult. However, it can be shown that by using Development Rate Monitor (DRM) techniques, a process for dyed photoresists, with acceptable process latitude, can be more easily established. In this study, viable techniques are investigated via DRM methods. Some of the process parameters that have been considered in this study are: 1. Unexposed film thickness loss, contrast (gamma), critical dimensions (CD) and photospeed fluctuation(s) as a function of developer type/concentration. 2. Using the PROSIM©1 approach, relevant values from the above are used to consider their impact on the contrast of a given dyed resist system. 3. The effect of both actinic and non-actinic dyes, together and separately, in the same resist, is considered using the approaches described above. Any apparent synergism is compared and its effect on overall process latitude is reviewed. The results obtained for the DRM modeling of dyed resist processes in this study demonstrate the validity of the DRM procedure(s) and its relevance to day-to-day production requirements.

As optical lithography moves into the area of submicron geometries, an increased understanding of lithographic process parameters is required. One area which requires careful examination is the interaction of the photoresist with specific spectral components of the actinic light. This will allow the photoresist performance to be separated from the performance of the imaging optics. An accurate indication of photoresist performance requires the measurement of photoresist response to a specific wavelength and energy dose. The spectrosensitometer reported on here was developed to address this need.

The need for photolithographic equipment to automatically align with improved accuracy is imperative, particularly considering the submicron design rules of megabit devices. By the addition of a non-actinic dye to a conventional positive photoresist (EZ-dye to EPA914, for example) and the utilization of a 1:1 Ultratech stepper, an improvement in alignment signal quality and the process window for optimized alignment operation on non-optimized targets is reported. The results obtained with the same resist without the dye are used as a baseline to further characterize the influence of the dye on target recognition and yield in actual production. The technique/mechanism of the dye interaction with the dark field technology utilized by the Ultratech stepper is described in detail. For example, it can be shown that with the incorporation of a dye that absorbs light at a given wavelength (for the Ultratech stepper this is found to be about 550 nm), the detection signal is optimized and the characteristic alignment peak is more easily recognized because of its increased sharpness and intensity. It follows from these observations that level-to-level alignment is improved.

New PCM (Portable Conformable Mask) method for excimer laser lithography has been developed. It consists of RD2000N (Hitachi Chemical Co., Ltd) as the top imaging layer, and PMGI (Poly-dimethyl glutarimide, Shipley Co., Ltd) as the bottom planarizing layer. In order to evaluate several kinds of deep UV resists and to search for a suitable resist process, we constructed an excimer laser exposure system, which is composed of a KrF excimer laser emitting at 248 nm with a bandwidth of 0.005 nm, and all quartz projection optics with a numerical aperture of 0.30, and a reduction ratio of 1/5. The excimer laser can produce about 100 mJ per pulse in the narrow bandwidth region. Spectral narrowing was accomplished by inserting etalons within the laser optical cavity. The reticle pattern was transferred to the wafer through the reduction lens. The energy density was about 15 mJ/cm2 at the wafer. From the evaluations of resist materials using the exposure system, it has been shown that RD2000N has high sensitivity and high resolution capability at the wavelength of KrF excimer laser. In the experiments of the PCM method, 0.5 μm thick RD2000N top layer was exposed and developed with an aqueous developer. RD2000N has very strong absorption in deep UV region, so can function as an adequate contact mask for the PCM method. After the top layer was delineated, that pattern was transferred to 1.0 μm thick PMGI as the bottom layer with deep UV flood exposure and subsequent developing of PMGI. With the PCM method, we have achieved 0.5 μm line and space resist patterns with aspect ratio of 3. The method is free from RIE treatment and thought to be more practical than a typical tri-layer resist which needs oxygen RIE for the bottom layer patterning.

A novel way to obtain tapered profile in photoresist is proposed. Opening at the mask level near the principal features, some slits smaller than the resolution limit of the projection optic at a distance also smaller than it, we can obtain a controlled sloped profile in photoresist. We have already applied this method to open tapered contacts and vias without using the thermal-reflow tecnique of BPSG optimizing the configuration of the subresolved slits. There is the experimental evidence that it is retained the possibility to open contacts up to the diffraction limit of the projection optics. Our method is Patent Pending .

Microwave energized bulbs are deep UV sources, which are well suited to applications requiring high light intensities of within the 200-260 nm wavelength range. This deep UV source has to date only been used either for transferring the photoresist patterns from the imaging layer to the planarizing layer of a PCM bilevel structure, or for photostabilizing the patterns delineated in a novolac based photoresist. So far, nobody has made use of the inherent resolution capability of the microwave powered deep UV light beam. The subhalfmicronic lithographic process described in this paper associates the resolution capability of the FUSION ILLUMINATOR 100 system with the contact performances of a K. SUSS mask aligner. Vacuum contact printing reaches the 0.3-0.5 )m resolution range for exposure times reduced to less than one minute. This paper presents and discusses results obtained by vacuum contact printing over W or Si02 layers.

Proximity Printing is a simple, high throughput non-contact method for wafer exposure. Image quality of such a system is limited by the diffraction of light at the edges of object. This causes exposure variation in resist and hence image distortion. Proximity printing process has been simulated for lithography purpose. Computations were done for an array of apertures. Aperture separation and aperture to substrate gap have been varied to see the effect on image intensity distributions. Some experiments were done using Cobilt aligner in proximity mode. Microphotographs of resist image were taken & compared with the 3D Plot of computed results. Good agreement is observed.

A novel technique is proposed to induce a positive tone, lift-off stencil in AZ 5214-E resist. The technique requires a flood exposure to be applied to a film of resist which induces a solubility gradient through the film. The solubility gradient is then reversed by a post exposure bake and is retained after image-wise exposure. The technique is demonstrated with results of image-wise exposures, by contact lithography, from 0.9 and 0.5 micron spacewidths in dark-field chrome masks. Resists profiles are displayed as a function of flood exposure, image-wise exposure and developer normality. Spacewidths in the resist film are plotted as a function of image-wise exposure energy and Gallium Arsenide MESFET structures with 1.0 micron gate-lengths are fabricated using the new process.

The limitations of subhalf-micrometer optical lithography in depth of focus, field size, and overlay accuracy, are discussed. The existing depth of focus budget as well as those for the near and the ultimate future are given. The optical depth of focus is explored in the point of view of microlithographers and means to overcome the depth of focus limit are suggested. Field size requirement of projected IC products is compared with the capability of known and projected optical systems followed with a discussion on methods to overcome the field size limit. The configuration of imaging optics is characterized and discussed by 1X versus reduction systems, refractive versus catadioptric and reflective systems, step-and-repeat versus step-and-scan systems. Overlay accuracy is discussed in terms of off-axis versus through-the-lens alignment, bright-field versus dark-field, actinic versus nonactinic alignment, followed with an overlay budget outlining the specification of the contributors to alignment errors. If all speculations are fulfilled, it is feasible to consider 0.18 μm resolution with 0.07 μm overlay.

A high N.A. i-line lens and a field-by-field leveling system have been developed. Resolution of the lens is better than 0.65 μm. The leveling system has achieved ±0.3 uμ/field on an experimental basis.

Using the i-line wavelength for production wafer steppers offers an improved combination of high resolution and large depth of focus compared to existing g-line steppers. A number of practical issues associated with i-line steppers include possible lens related absorption effects, lenses with insufficient field size or resolution, availability of i-line resist processes, accurate through the lens alignment, and multiple machine matching performance with low distortion lenses. Experimental results are presented in this paper on each of these issues. Data is presented for a multiple number of i-line steppers equipped with the Zeiss 10-78-58 lens.. Support of half-micron design rules is demonstrated. Resolution over the image field to 0.5 μm with a large depth of focus is shown and overlay performance to less than 150 nm is demonstrated.

This paper describes the development of g-line projection lenses as shown by an investigation of the decreasing value of focus depth due to lens production errors. It presents a study of the decreasing value and forecasts what the depth of focus will be for each linewidth in optical lithography in the next several years. Next we report the image performance of a new lens with a high numerical aperture that has been developed recently. This is the 5x reduction lens with a field size of 15 mm square (21.2 mm diameter) and a numerical aperture that is switchable between 0.48 and 0.43. The practical resolution is 0.7 μm and distortion is less than 0.1 μm over the image field. The depth of focus is in the range of 1.5 μm. In addition, we report the potential of this lens for achieving a resolution of less than 0.6 μm.

Alternative routes to 0.5μm lithography made possible by new developments in lenses, excimer lasers and resists are discussed. Steppers using excimer laser illumination with wavelength of 248nm provide 0.5μm lithography. Advanced processing techniques such as image reversal and contrast enhancement materials, make possible dramatic increases in resolution. New higher numerical aperture i-line lenses coupled with these processes also provide 0.5μm lithography. Experimental results illustrate the possibilities and trade-offs of alternate technologies.

The submicron resolution capabilities, depth of focus, critical dimension control, and process latitude of a 1:1 0.40 NA lens using some of the newer high contrast resists were studied. Over the 3.5 cm2 field size the lens resolved 0.6 microns with a 0.8 to 1.6μ depth of focus. The depth of focus for 0.7μ features was found to range between 1.6 to 2.0μ. Resolution of 0.5μ with a 1.0μ depth of focus was achieved over a limited field size. The critical dimension measurements show that through focus 98% of the data stays within ±10% of the measured feature size. This critical dimension control and the improvement in resolution to 0.6 microns across the field indicates that the practical resolution of the lens can be better predicted by using 0.7 λ/NA. The application of this lower practical linewidth size to space-bandwidth-product was demonstrated.

A new i-line projection aligner, the LD-5010i, has been developed and has two primary features : good patterning ability and good overlay accuracy. In this paper, performance of the i-line projection system and the characteristic alignment method, called the two-wavelength detection, are described.

The demands of the semiconductor industry for super high-pattern density with ever-reducing feature sizes, places requirements on semiconductor equipment manufacturers to produce lithographic tools of increasingly challenging specifications. Currently, the tool of choice for submicron production lithography is the 5X stepper system. The pressure to increase the performance of this type of system is causing lens manufacturers to evaluate optical systems that have higher Na values operating at shorter wavelengths than ever before. Different equipment manufacturers offer many combinations of x , Na, and image field size. They are all balancing these variables to produce a specification that their own lens fabricators have the manufacturing skills to build. This paper briefly discusses the design and testing of a lens, and the matching of the photoresist and its processing to the lens operating at I-line, in order to maintain and replicate the high-quality submicron image.

A photolithography for 0.5um feature size was studied. For this experiment a g-line Stepper with a lox lens of 0.6 numerical aperture (N.A.) and 5mm square field size was used. Profiles of both a single layer resist and a trilayer resist were investigated by caking SEM photographs. This process was applied to the fabrication of silicon microwave bipolar transistors. In conventional photoresist processes, 0.45um line and space were resolvable for the photoresist thickness of 1.2um, and the cross sectional profile of the resist was rectangular. The focus depth for 0.5um pattern was ±0.2~0.4um and depended on resist. The 0.5um patterns were uniformly resolved in the field. A 0.5um pattern with 0.5um step height could be resolved. The rectangularity of the resist cross section was strongly dependent on resist properties. The trilayer resist with the feature size of 0.4um line and space was achieved. We Found that the focus depth of the trilayer resist process was ±0.6um and twice as much as that of the conventional single layer resist process. This is because even poor profile of top layer resist was useful for making good rectangular profile of bottom layer resist. As a result, more focus margin could be obtained for the trilayer resist. The same result of focus depth was obtained in the case of 0.5um step height. On 1.0um step height, though 0.5um line and space could be separated, it could not be applied for practical use due to the fluctuation of demensions. This is because bottom layer resist could not form 1 flat surface. Contrast Enhancement Lithography (CEL) process was applied, but no improvement of resolution was obtained. A silicon microwave transistor with minimum size of 0.5um was fabricated using this Stepper. The trilayer resist was applied for the metallization lithography. For 0.5um rule lithography except metallization, the conventional single layer resist process was ased and we made the surface of the wafers as flat as possible. The yield of 0.5um pattern variation less than +0.05um was 60% in a wafer, however some patterns were not resolved. The poor patterns were formed due to the poor flatness of the wafers. Correlation between the flatness of the wafers and the pattern resolution was obtained. Required specification of Local Thickness Variation (LTV) was estimated for single layer and trilayer resist processes, respectively.

Recent advances in lens design have pushed optical lithography well into the submicron domain which was once considered to belong to X-ray or E-beam lithography. Realisation of submicron design rules of a 1 Mbit memory has been possible at PHILIPS using an i-line lens (Zeiss 10-78-48) incorporated in internally developed stepper (Sire-3) using a single layer resist technology. This paper will give a schematic description of the stepper and describe the lens characteristics of the two in house systems. The stability of the steppers will be illustrated by means of system parameters which were monitored during a prolonged period. The resist process was characterised on accelerated pathfinder lots which were used to detect lithography related problems in an earlier phase and to determine machine and process latitudes. The results of this characterisation and implementation activity will be reported. Special attention was given to focus determination which proved to be very critical due to the lens characteristics and large chip-size. Finally the results obtained on the 1M SRAM device with 0.7 ium minimum geometry will be presented. Based on this work it appears that with proper planarization procedures for minimizing the topography problems, a 0.7 μm design rule is practical using single layer resist on this stepper.

For the last several years controversy has centered around a successor to optical g-line step and repeat lithography. Diverse proponents have variously advanced e-beam, x-ray, ion beam, excimer, and optical exposure tools. At the same time several alternative resist schemes have been proposed. Multilayer, contrast enhancement, inorganic, and image reversal resists are some of the candidates. This paper presents the analysis of the selection and development of a lithography unit process designed to satisfy the needs of a high performance multilevel interconnect CMOS process at Hewlett Packard Corporation. This effort represents a unique opportunity because HP recognized the need to determine the capabilities of viable lithography technologies before embarking on a definition of the other parts of the VLSI process.

The Stepper Image Monitor(SIM) system tests the resolution and overlay performance of complete stepper systems by measuring the aerial image intensity profile at many wafer locations and focus offsets. Overlay error vectors can be measured with a precision of .02μm and best focus can be determined to a precision of O.lμm without exposing test wafers. New data will be presented on lens heating effects, along with a simple model. Imagery of high NA steppers will be explored by observing image profiles of various sub-micron lines. A new poly-on-oxide process to fabricate permanent SIM reference wafers will be described. This paper will describe other recent developments which have made SIM techniques broadly applicable to many practical problems of stepper setup and characterization.

Image quality criteria derived from the point spread function (PSF) are useful for testing lithographic lenses. The Strehl ratio (SR) is a particularly simple and powerful measure of lens quality which is readily accessible from measurements of the PSF. The SR correlates well with traditional figures of merit such as the mean square wavefront aberration and the modulation transfer function. For object dimensions on the order of X/NA, the SR is a direct measure of the field dependent variation of exposure dose caused by lens aberrations. In this sense, the SR connects the lens quality to the final pattern definition in the image recording medium. Experimental aspects of the PSF measurement are emphasized. The lens test bench used to measure the value of the SR through focus and with field position is described. Data showing the PSF, and the field and focus dependence of the SR, are presented for high resolution Hg I-line and G-line lenses.

The complete characterization of a resist imaging process is critical to it's success in a production environment. AZ-5214 resist in image reversal mode offers a single layer process that appears to be viable for submicron processing. It does, however exhibit some unique problems that had to be solved during process development. These are: 1) an extreme sensitivity to processing delays, 2) undercut profiles on reflective metal surfaces and 3) different dry etch characteristics. In this development effort the key advantages of image reversal were: reduced proximity effects, higher resolution, vertical resist profiles, as well as greater exposure latitude. AZ-5214 used in image reversal mode is very sensitive to any change in delay time during the processing sequence. In each case increased delay time causes growth in linewidth. This change in dimension is most significant right after the exposure and bake steps. The problem can be minimized by employing a stabilization time after softbake and exposure. Integrating the reversal bake, flood expose, and develop sequence in a single track is also necessary for consistent dimensional control. Negative tone processing of AZ-5214 resist exhibits undercut profiles at the resist metal interface. Metals with different reflectivity such as aluminum, moly, and Ti/W all demonstrate this problem. The use of anti-reflection layers provide vertical profiles at the expense of increased process complexity. Layers that have shown good results are PE/CVD nitride and PE/CVD oxide. The process characterization data includes: exposure and focus latitude for a range of feature sizes, proximity effect data for dense and isolated features, linewidth uniformity after etch, and dwell time latitude. Also included are the results of daily particle counts over a four month period.

Device design criteria of half micrometer linewidths have driven optical lithography to extend its imaging wavelengths into the deep ultraviolet region. Excimer lasers are the only source that produces the high spectral brightness required by projection lenses at these wavelengths. A spectral line narrowed KrF (2484 Angstrom) excimer laser and a new illuminator have been developed and integrated with current DSW Wafer Stepper® technology. The integration of these new features require the design of control subsystems to monitor and control parameters such as wavelength and temperature. Wavelength must be tuned and stabilized to optimize the resolution and distortion performance of the reduction lens. The effects of temperature and bandwidth are also important to lens performance criteria, for example astigmatism, distortion, and contrast. Acceptance results for an excimer stepper will be discussed with special emphasis on resolution capability, illumination uniformity, focus stability, and dose repeatability.

A new KrF excimer laser stepper system has been developed for sub-half-micron VLSI's device fabrication. The optical system has a monochromatic 5x reduction-projective lens with a large field area (15mm) and a high numerical aperture (NA:0.36). The excimer laser with intracavity etalon, which is compacted by the seperation of a blower system from the laser head, can produce a practically acceptable power of 40mJ/pulse. The mean value of spectral band width was 0.008 nmFWHM and its dispersion was 2.5% of it. This laser indicated excellent properties and stabilities. From the simulation of the Optical Transfer Function (OTF) of the reduction lens, we have found that the lens of NA=0.36 is most suitable for the spectral band width of 0.008 nmFWHM. We have performed the evaluation experiment of the reduction lens by evaluating the resolution and profile of fabricated resist (MP2400) patterns. As a result, the fine pattern of MP2400 down to 0.35 μm lines-and-spaces was successfully fabricated in the center field of this lens. Therefore, we have a good agreement of actual resolution limit with the theoretical resolution limit. Moreover, We evaluated of resists using this KrF excimer laser stepper system. Evaluated resists were MP2400, HPR1182, RD2000N, TPR101 and SNR, where SNR was used as a top layer resist of bi-layer resist (BLR). As a result, we have found that the highest resolution resist is MP2400. In conclusion, the newly developed KrF excimer laser stepper has been confirmed as an effective technology to fabricate sub-half-micron design rule devices : 16MbitDRAMs.

Excimer laser lithography will extend optical techniques to deep-uv wavelengths and enable linewidths at 0.5μm or below to be printed routinely. The main problems concern the development of suitable lenses, resist technology and alignment systems. This paper addresses mainly the first problem, from the viewpoint of 1:1 Wynne-Dyson optics. Two lenses are discussed, one operating at λ = 249nm and the other at λ = 193nm which is currently undergoing evaluation.

Reticles (masks on enlarged scale) are needed for optical pattern transfer in the production of integrated semiconductor circuits. In order to meet present requirements for 5X reticles only a direct writing technique is feasible. This means direct exposing of photoresist either with light or an electron beam. Many of todays highly dense reticles require some 10 5 to 10 6 discrete exposures when generated with an optical pattern generator. Optical pattern generators normally use mercury arc lamps to expose positive photoresist, which in turn need 200 milliseconds for each of these discrete exposures, thus requiring to stop the table at every exposure position ("stop and go" mode). This results in running times of several days per reticle. Therefore most reticles are nowadays being manufactured with very expensive e-beam machines. In the early 80's we started the first experiments to expose photoresist with an excimer laser. In order to obtain the maximum gain in speed, the goal was to operate with only one excimer laser pulse per exposure, so that a fast "flash on the fly" operation with an optical pattern generator became true. Equipping a conventional optical pattern generator with an excimer laser as the light source, it has become possible to expose substrates coated with standard photoresist in the "flash on the fly" mode with only 13 nanoseconds per flash. So the thruput could be increased up to 25 times in comparison to a pattern generator equipped with a mercury lamp. A comparison of both operation modes will show that an immense increase of speed is possible, even when a ten years old M3600 pattern generator is used. This system is in function now with very high reliability since more than three years in our IC development line.

This paper will discuss the problems associated with excimer laser photo-lithography -the combination of a KrF narrow band width excimer laser (non-injection locked type) with a large field fused silica monochromatic reduction lens. An excimer laser with a KrF narrow bandwidth, in combination with a large field monochromatic lens which is appropriate for use with such laser, have been developed and tested. The system's resolution capability has been confirmed at 0.4 um L/S with MP2400 resist. The laser has been designed so as to be installed and maintained in a clean room environment as well as to have a very narrow spectrum line. A very narrow band-width beam, down to 0.003nm, has been attained through a stable resonator with more than 20mJ pulse energy. The ultra-compact laser head (300mm x 545mm x 1100mm) contains a small laser discharge unit (182mm x 156mm x 584mm), and no amplifier because the oscillator is highly efficient in spite of the narrow line emission. Maintenance is much easier in the clean room environment. Users can replace the discharge unit as easily as they would change Hg-lamp, only taking twenty minutes, and while they clean the window and check the electrodes of the removed unit, the laser can be operated with the easily installed replacement -already passivated discharge unit. The laser head unit is separated from a gas circulating unit and trigger pulse circuit - vibration, heat, EMI noise and particle generation. Therefore, it can be installed even in the thermal clean chamber of a stepper. The N.A. (numerical aperture) of the monochromatic lens is 0.36 and the field size is 15mm x 15mm. In fact, three kinds of lenses with N.A.s of 0.4, 0.35 and 0.3 respectively, were designed and individually evaluated for their OTF's and defocus's dependence on the light source's spectral width, and also their co-relationship. In parallel, simulations on the relationship between each lens' chromatic aberration and laser spectral width were completed and such results have been integrated with the laser performance data ie. output power v.s. spectral line width, and its interactions from speckle noises/background emmissions per line. The system is capable of achieving a resolution preformance of 0.4 microns over a field measuring 15mm x 15mm, and it is very reliable for practical applications since it employs the combination of 5-7/1000nm spectral bandwidth excimer laser with N.A. 0.36 lens, which has been optimized from the preceding simulations and experimental data. When the N.A. of a lens is being determined it is necessary to take into consideration the spectral width of the light source because in the case of a narrow bandwidth laser, it would be difficult to satisfy two contradictory factor6 - speckle noises/background emissions and narrow bandwidths. An extremely narrow bandwidth laser (less than 3/1000nm) would be much less reliable as it would demand sophisticated configuration and adjustments. A monochromatic lens with a high N.A. would struggle in its efforts to locate such a very narrow linewidth industrial laser source with easy maintenance and compact size in the clean room environment.

Measurements of pulse-to-pulse energy fluctuation statistics were made at several wave-lengths for a commercial discharge pumped excimer laser, under a variety of operating conditions. Various approaches to achieving accurate exposure dose control in microlithographic exposure tools using noisy pulsed sources are discussed. A pulse-by-pulse active control technique for static field lithography is described in detail, together with measured performance data. Alternative techniques suitable for one dimensional scanning slit field exposure tools are presented.

Interference patterns are produced by point light sources originating from an optical integrator. Influence of laser light coherence on interference pattern contrast is examined analytically and experimentally. The contrast is independent of temporal coherence and dependent on spatial coherence. The modulus of the complex coherence factor remains between 0.35 and 0.55 for the number of transverse modes between 30 and 240. The interference pattern can be eliminated by rotating the optical integrator. Resolution dependence on spectral bandwidth is experimentally examined using an NA=0.42 non-achromatic projection lens. Dependence of lens resolution on spectral bandwidth is greater than that estimated from MTF curve.

A single stage excimer laser for UV microlithography has been developed. It emits 2 Watt average power at about 248.4 nm with a bandwidth of 0.5 cm -1. A novel computer-controlled feedback circuitry provides for automatic stabilization of both average power and integrated bandwidth. In addition, a computer-controlled frequency stabilization unit makes that new laser ready for industrial use as light source in deep UV microlithography.

This paper discusses the different factors which need to be addressed in the development of excimer lasers for use in microlithography. Included is a brief description of the basics of excimer laser technology and a short history of the use of excimer lasers in microlithography. Emphasis is placed on the differences between excimer lasers designed for use in microlithography and conventional excimer lasers. The paper concludes with a discussion of maintenance issues and operating costs.

Cr masks used for conventional optical lithography were evaluated for their damage under excimer laser irradiation at 248 nm and 308 nm. The damage of Cr films on quartz ranged from erosion of pattern edges to total ablation depending on the fluence. At low fluences, cumulative stressing of the Cr films by the laser pulses leads to development of fine cracks. Difference in damage threshold at 248 nm and 308 nm was observed.

Excimer laser projection lithography is expected to become a widely used technique, capable of high throughputs at resolution well below 0.5 μm. In this paper we demonstrate excimer projection patterning with 0.13-μm resolution, and we address issues related to laser engineering and optical materials, which are encountered in the design of practical excimer projection systems.

A compact small field 20:1 reduction optical imaging system has been developed to provide a stable platform for printing submicron lithographic images at 248nm in resist without the cost and complexity of production equipment. The target applications are those in the areas of process development for resists in UV and deep UV and R&D laboratories and universities doing image science. In all of these applications, the normal issues of production lithography such as throughput, field size, automated substrate handling, etc. are unimportant. A simple bench top lithography system without these production oriented constraints is described.

A primary thrust of optical lithography is the use of shorter wavelengths of light in order to produce smaller geometries for VLSI device production. Excimer lasers have been identified as likely candidates to extend optical lithography into the 0.4-0.3 micron regime. This paper describes an excimer laser optical approach for the projection imaging of sub-micron geometries. An excimer laser-based imaging and ablation system is described in detail along with initial results obtained with resist materials. The optical system employs inter-changeable lenses, with varying numerical apertures, reduction ratios, and field sizes. The data presented shows how excimer laser energy at the 248nm and 193nm wavelengths can produce 0.4-0.5um resolution using direct photoablative decomposition of the resist layer. Attenuation of the beam allows the system to be used for conventional latent image exposure and wet development with deep UV resists. Small-field chrome-on-quartz reticles at 15x and 36x are used to project device structures, resolution patterns and alignment marks. SEM photos show the resolution capabilities across the effective field of the system. Patterns as small as 0.3um are resolvable using this system.