Using birefringence monitoring, several experimental fused silicas are tested for 193-nm compaction durability. f N ( I J2 't 10.1 All samples exhibit densification that can be described by the equation: (Lip/p )u = 1C · l--7 · - · _.!!_J where 10 10 't 'tis the pulse length ('t0 = lns), Np is the pulse count, K is a constant, and I is the 193-nm energy density (I., = 1 mJ/cm2). The extracted value of K varies from 84ppB to 660ppB for experimental (1995-1996) fused silicas, as much as a factor of two improvement compared to (1990-1994) grades. The role of irradiation geometry in compaction is also investigated using finite element simulations. The net optical path difference formed for a given level of damage is found to increase by approximately 25% when the damage radii is increased from 30% to 70% of the total sample diameter in a relatively thin optical element. Keywords: damage, fused silica, two-photon, densification, compaction, lithography, 193-nm, stress-birefringence

Currently it is very popular to use off-axis illumination technique for higher resolution with wider depth of focus. However there are several problems in the technique, one of which is deterioration of image quality induced by the non- uniform effective source distribution. If the intensity distribution on the illumination aperture lacks of spatial symmetry, each diffraction order beam impinging on the wafer surface has angularly asymmetric distribution. This makes the optical system have pattern size dependent telecentricity error. For a simple line or grouped lines it gives rise to only the pattern displacement with defocus which can hardly be detected unless there are any reference. But the periodic island type patterns which have discrete features and multiple pitch components in one direction can be bent and deformed asymmetrically with defocus. Asymmetric imaging for island type patterns gives rise to also the pattern CD asymmetry with defocus. We present schematic explanation of the effects of non-uniform effective source and the simulation result. We also investigated the phenomena in a high density DRAM cell active layer of 460 nm minimum pitch and characterized it by various approach.

A new super resolution technique with wide applicability and high resolution enhancement has been developed. Various optical properties of the new optics have been analyzed by some optical simulations. Using a spontaneously optimized source shape for the mask pattern itself leads to better applicability with the new optics. From experimental verification, the DOF of the new optics is 1.5 times wider than the modified illumination. An alignment accuracy of 0.1 micrometers (on mask) is acceptable between the first and second mask. According to simulation results, the optical proximity effect is reduced by the new optics. These features offer significant merits to photolithography technology. Therefore, we estimate the optical proximity effect by optical image measurements and simulations. From the simulation result, we can use the third source aperture to control the optical proximity effects. For conventional illumination, a 0.35 micrometers L/S image is formed by two beam interference. For the new optics, the same pattern is imaged by four beam interference because the UV light is diffracted twice by the mask pattern. While the optical proximity correction system was being developed, it was confirmed that the pattern size fluctuation is sufficiently smaller for four or more beam interference. The experimental results of an x-z image profile for 0.35 micrometers , 0.4 micrometers or 0.5 micrometers line and any space are consistent with simulation results.

A novel method to reduce a mask error effect in alternative phase shift lithography is proposed, which is derived from the analysis of the pupil plane images generated by masks with and without error. The analysis shown that zeroth order diffraction light, which usually shows zero intensity for precisely formed mask with the same areas for phased and non-phased openings, is generated by mask error both in phase and transmission. So, applying a pupil filter that cuts off rays passing through the central part of pupil is considered to be effective for the reduction of mask error effect. Impacts of the pupil filtering on imaging were evaluated by the numerical calculations of aerial images for various masks and results revealed that by applying a circular pupil filter with same radius as that of non- diffracted illumination image, tolerance in phase and transmission error for L/S patterns enlarged up to 20 degree and 50%, respectively.

The prerequisite to successful optical proximity correction is an in depth understanding of the relevant parameters leading to patterning inaccuracies. The work presented in this paper is based on a test chip specifically designed to investigate sources of 1D line width errors, deemed the most critical for optical proximity correction. Data are presented relating line width errors to pitch and pattern density and highlighting the complex interactions between these two main contributors of line width error. The goal of this paper is not to derive a mathematical model for line width variation in etched polysilicon line structures but merely to qualitatively bound the complex interaction of pitch and pattern density based line width variation. The results of this investigation indicate that pattern density has a very significant effect on line width and that lithography, not just reactive ion etch, is significantly impacted by pattern density.

DUV scanning exposure systems have been steadily gaining market acceptance for the past five years, and soon, all major suppliers will offer 248-nm scanning tools. One of the major reasons for the emergence of this technology has been the purported improvement in critical dimension (CD) uniformity across the scanned field versus what can be realized in a full field stepper. Using high precision electrical resistance CD metrology, we have characterized the across field CD control capability of several DUV scanning tools and DUV steppers. Analysis is carried out through focus for multiple linetypes representing various orientations and nearest-neighbor proximities. Where possible, different NA/(sigma) combinations are examined as well. Surprisingly good full field sub-0.20 micrometers CD control is obtained even for 0.50 NA, and higher NA allows for non zero process latitude at 0.14 micrometers geometries. While it was initially anticipated that 193 nm ArF lithography would be required for 0.18 micrometers technology manufacturing, it has become apparent that 248 nm lithography will be employed for these groundrules, particularly for logic applications with predominantly semi-isolated features.

The use of I-Line exposure wavelength for manufacturing at and beyond 0.35 micrometers presents many challenges in manufacturing. The lack of resolution, depth of focus, exposure latitude, and iso/dense offsets have caused some to switch from I-Line to DUV. With our installed I-Line base we felt it necessary to implement techniques to extend our tool life, reduce manufacturing costs while improving manufacturing margins. The results of the differential modification techniques were used to reduce the effects of topography, density, and low k lens issues. The differences seen between the binary and phase shift plates show the advantage of phase shifting below 0.35 (mu) manufacturing. We have been able to demonstrate between critical dimension (CD) control using phase shift mask with dense iso compensation over a standard binary reticle. The data shows improved CD control across the stepper field, wafer, and overall lot distribution. The impact of this work was improved speed performance. It also allowed us to move the CD's to smaller dimension because of the better control without increasing fallout due to electrical parametric roll-off.

Germanium based ARL (Anti Reflective Layer) having high conformality over topography and removable during resist strip process was developed. Its various characteristics were investigated. The ARLs were composed with GeN, and (Ge,Si)N_x and fabricated by reactive RF sputtering. The optical constants of Ge based materials were measured and the ARL performance for DUV lithography was obtained. Since the GEN_x is dissolved in water during resist develop process, it can not be used. Therefore, silicon was added to solve this problem. Thin film characteristics of (Ge,Si)N_x compound were analyzed using XRD, XPS, AES, SEM. The ARL performance was confirmed by resist patterning. Because the (Ge,Si)N_x material is removable by H₂SO₄ strip, yet most of current inorganic ARL is not, it has advantages for process simplicity.

The use of attenuated phase-shift masking materials is being considered as one of the key resolution enhancement techniques for sub-0.18 micrometers lithography. In addition to proper optical performance, films for use at DUV and 193 nm wavelengths require suitable plasma etch characteristics and stability at mask exposure levels of pulse excimer lasers. This may limit practical materials to those which allow suitably volatile etch bi-products and possess stable stoichiometric composition. Several film families have been produced which can deliver a (pi) phase shift and 4 - 15% transmission within a single films thickness including Al/AlN, MoSiO, TaSiO, Zr/ZrN, Si_xN_y, and TaN/Si₃N₄. Of these materials, TaSiO, Si_xN_y, and TaN/Si₃N₄, allow for adequate plasma etch performance with selectivity to fused silica and resist. Excimer laser 193 nm exposure at fluences corresponding to mask exposure levels show some degree of optical degradation of materials prone to oxidation.

To enhance depth of focus (DOF) for isolated line patterns, we have developed a new assistant pattern method. In this method, opaque additional patterns are placed beside the attenuated phase-shift main pattern. DOF enhancement effects of this and conventional assistant pattern method were evaluated by means of a KrF excimer exposure tool with variable NA (0.45, 0.50 and 0.55). Using the new method with off-axis illumination, we obtained 1.5 micrometers DOF for 0.25 micrometers isolated line pattern, much wider than that achieved in the conventional method (0.9 micrometers ). Furthermore, we confirmed that this new method was effective not only for improving the exposure dose latitude but also for reducing the optical proximity effect.

The application of an attenuated PSM for printing contact hole layers is one way to extend the useful lifetimes of an I-line stepper. The development of such a process needs to consider more than just the traditional parameters. Selecting the optimum mask CD bias, stepper NA, sigma, exposure, and focus, is even more critical than for the process using conventional binary masks due to sidelobe printing with attenuated PSM. Using an electrical CD measurement technique, this study was able to evaluated process latitude at various stepper and mask bias settings, at one point and at sixteen points per field. One of the resists tested showed 0.85 micrometers depth of focus and > 20% exposure latitude for 0.30 micrometers contacts on 0.75 micrometers pitch. The overlapping process window at sixteen points per field is much smaller than that at only one point per field, justifying the need for such an evaluation. Full wafer CD mapping allowed us to quantify across wafer and across field CD variation. Though the electrical measurements exhibited excellent correlation with SEM measurements (R² equals 0.994), it is still necessary to verify a chosen process under SEM inspection to evaluate the resist profile for any artifacts such as sidelobe printing.

There are an increasing number of issues confronting lithography engineers in modern wafer fabs. Of these problems, yield loss due to reticle defects has received less attention as compared to the shrinking process window of advanced lithographic requirements. Wafer fabs also have concentrated primarily on equipment and people as major sources of yield loss. The requirement of increased focus on reticle defects is examined. The constraints of the current manufacturing capability of mask shops is driving the need for a better method to link the actual lithographic manufacturing process to the reticle defect analysis. This paper will propose a method for integrating the reticle inspection and wafer process as a method for advanced reticle disposition and specification generation. By linking the fab wafer process to the reticle defect inspection a more complete picture of the impact of having reticle defects can be assessed.

In this paper, the results of an NA-sigma optimization study are reported, carried out experimentally for an advanced ASML PAS5500/300 deep-UV stepper. The work has been primarily focused on a 0.25 and sub-0.25 micrometers gate layer in a logic CMOS process. A positive and negative tone resist process have been compared in terms of CD control and line- end shortening. Dry etch effects and across-field behavior has been taken into account. Furthermore the contact level of the 0.25 micrometers process have been optimized. Effects of layer dependent NA-sigma settings on overlay have been studied.

The traditional lithographic approach employed by the semiconductor industry has been to pursue use of advanced prototype optical exposure tools and resists. The benefits of doing so have been: (1) The lithographic process that is used in development more closely resembles the process that will in fact be used to manufacture the chip. (2) The cost of low K₁ imaging (phase-masks, off-axis illumination, and surface imaging resist) can be avoided. However with the introduction of 1Gb-dynamic random access memory (DRAM) development, a paradigm shift is being experienced within the optical lithographic community. With 1Gb-DRAMs, the minimum feature size falls irreversibly below the optical wavelength used to image the feature. Such a situation will make low K₁ factor imaging unavoidable. With 175 nm groundrules typical for first generation 1G-DRAMs, K₁ factors near 0.4 will be common with 0.5 as an upper limit on advanced systems currently in development irrespective of optical wavelength. This paper will cover the selection process, experimental data, and problems encountered in defining and integrating the lithographic process used to support the critical mask levels on 1Gb-DRAM development. Factors considered include: resist, masks, and illuminations via both simulation and experiment. The simulations were conducted with both internal and externally developed software. The experimental data to be reviewed was generated using an experimental 0.6 NA KrF step and scan system provided by Nikon. The resist used is commercially available from the Shipley corporation.

An application of alternating phase-shifting mask to 0.16 micrometers logic gate patterns is studied. A double exposure method using positive resist in KrF excimer laser lithography is applied to obtain random gate patterns. To optimize the exposure conditions, proximity effects for 0.16 micrometers line patterns under various combinations of NA and (sigma) are examined using an aerial image simulator. To control the linewidth in +/- 10% CD, mask bias according to the space width and the mask pattern with two adjacent apertures which transfers an isolated line are applied. By using these techniques, 0.16 micrometers logic gate patterns including SRAM are demonstrated.

Decreasing critical dimensions and pitch at metal and hole levels requires the use of Deep-UV processing. Associated with this processing are issues of resist-substrate interactions, small process margins, and increasing aspect ratios forcing the modification of etch processes. In overcoming these issues at metal, we evaluate resist profiles on TiN, oxide hard mask, organic BARC and Ti metal. In all cases we see an improved profile when compared to resist on TiN. Electrical performance is compared in terms of linewidth, process linearity, and isolated to dense line bias for TiN, oxide hard mask, and Ti. Post etch electrical CD repeatability results are presented for TiN, Ti, and oxide. Cross section micrographs demonstrate consistent metal post etch profiles regardless of the metal stack. Results show the process is scaleable to 0.36 micrometers pitch metal. For hole patterning we see a major obstacle is achieving the required resolution capability with sufficient process latitude. We investigate the process margin for 0.25 micrometers contacts on BPSG substrates and the effect of pitch on the contact dimensions. Evaluation of the contacts after etch shows vertical sidewalls. The ability to extend the process to 0.20 micrometers is also explored. The results presented in this discussion demonstrate some of the issues and requirements for Deep-UV processing. With this technology we must explore the use of interfacial layers to overcome resist interactions with substrates and consider what might be non-standard processing to achieve desired results.

In previous work we combined fast aerial image simulation with a closed-loop Optical Proximity Correction (OPC) control system to generate pre-compensated mask geometries which account for pattern transfer distortion effects at small feature sizes. We also presented the variable- threshold resist (VTR) model in which an image-dependent threshold is used to calculate linewidths directly from the image intensity. The model parameters can be determined by `tuning' the model with linewidth measurements from chosen sample sites on the wafer. In this paper, we present verify our OPC approach experimentally by showing after etch SEM wafers of corrected and uncorrected designs. In doing so, we show that (1) OPC can eliminate bridging effects in uncorrected designs, (2) VTR model is fairly insensitive to process variations and (3) mask writing effects are important and cannot be ignored.

Proximity correction systems require an accurate, fast way to predict how a pattern configuration will transfer to the wafer. In this paper we present an efficient method for modeling the pattern transfer process based on Dennis Gabor's `theory of communication'. This method is based on a `convolution form' where any 2D transfer process can be modeled with a set of linear, 2D spatial filters, even when the transfer process is non-linear. We will show that this form is a general case from which other well-known process simulation models can be derived. Furthermore, we will demonstrate that the convolution form can be used to model observed phenomena, even when the physical mechanisms involved are unknown.

This paper explores through numerical simulation and experiment the role of partial coherent illumination on reflective notching in optical projection printing over a nonplanar substrate. The intensity in the photoresist when patterning a gate line over a LOCOS generated active area well covered in polysilicon was simulated with the rigorous electromagnetic simulation program TEMPEST-PCD which includes partial coherence through a decomposition method. The simulation results are compared to experimental SEM pictures of developed photoresist lines for patterning a gate over a trench with varying moat widths when using a DUV 248 nm stepper with a NA of 0.5 for a (sigma) of 0.3 and 0.6. Variations in linewidth of both plus and minus 20% are observed due to the interactions with the underlying topography. The simulations give physical insight into trends seen in the experimental critical dimension (CD) measurements. Simulations indicate that most of the effects observed are due to lateral specular reflection from the underlying structure aiding in the dissolution of the resist near the feature. These trends seen in simulation and in experiment indicate that the topography has a similar impact on the CD regardless of the coherence.

Clearly, all lithography simulation tools comprise models that depend on certain measures of the lithography process as input. In fact, it can be said that these models are only as good as their input. Therefore, the tuning of these models to fit the particular conditions of a certain process is a subject worthy of investigation. In this work, we extend previous efforts to generate `tuned' parameters for the lithography process models within PROLITH/2 by including the evaluation of resist cross-sections under a variety of conditions in our analysis, and by more closely examining the sources of factory-specific tuning problems. Specifically, we identify and quantify error sources related to film thickness measurements, utilize the so-called `Poor- Man's DRM' technique to tune PROLITH's resist and development parameters, and compare the resulting simulations to both swing curves and resist cross-sections of various sizes on multiple substrates. The merit of this tuning approach is evaluated based on these comparisons. We conclude that optimization of simulator parameters is critical for accurate resist profile prediction and that, once optimized, the model provides quantitatively predictive results over a wide range of experimental conditions.

The optical proximity effect becomes significant near the practical resolution limit of photolithography, depending on the wavelength and numerical aperture of the stepper. Recently, VLSI design rules have almost reached their limits. Larger ICs cannot be designed and be manufactured without using a lithographic DRC (design rule check) tool or an OPC (optical proximity correction) tool. Therefore, it has become necessary to develop a technology which can accurately predict resist features from the designed circuit layout. We studied deviation in both the line width and the length due to proximity effect and investigated the phenomena. Also we developed a technology which can accurately predict the behavior of the proximity effect from an aerial image. This technology is based on a simple threshold model. We optimized the calculations for an aerial image and the threshold of intensity in order to predict deviations in the line width and length. We also considered the profile of an aerial image to predict the critical point. The calculations for an aerial image and threshold which we optimized in this manner can be used to predict 2D patterns.

The variation in critical dimension (i.e., CD) as a function of exposure field position is an important error component in total linewidth control. In this paper, we describe method for uncovering root causes of the across field CD variation in a lithographic projection system. For example, using a special reticle artifact we form images of the exposure system source distribution at various points in the lens field. These field dependent source distributions provide important clues into the variation of partial coherence, dose and numerical aperture around the exposure field. We also present a novel technique for inferring the residual projection lens aberration signature from CD measurements in photoresist. Tests to understand the role of stray light in across field CD variation are also investigated. Exposure system specifications consistent with small variation in across field CD are provided on the basis of these test vehicles.

This paper shows, that as resolution is pushed into regions below 0.6 (lambda) /NA, understanding the effects of wavefront aberrations is crucial to producing stepper systems that can meet end-user requirements. We show how aberrations can affect the choice of optimum NA and partial coherence for a given reticle object when considering critical dimension uniformity and depth of focus. The ability to measure the complete wavefront and extract meaningful full-field aberration data is shown using an advanced through-the-lens interferometer that operates at the wavelength and bandwidth of the lithographic radiation. The impact of aberrations an image quality criteria is shown through a sensitivity analysis using an imaging approximation model that represents various image criteria as a weighted sum of aberration coefficients. The validity and use of such a model is shown by correlation to full- field experimental measurements.

Overlay errors of fine patterns using modified illumination are experimentally measured and analyzed. Overlaid `vernier' patterns which fine lines and holes are overlaid on standard line patterns are used to measure the overlay errors. The experimental results using a KrF excimer laser stepper with various illumination conditions show that the overlay errors depend on the illumination coherence factor (sigma) and the pattern feature. These overlay errors consist of magnification and offset and these errors depend on the pattern feature. The offset error between line pattern and hole pattern are considerably large for device fabrication. The magnification component originates from coma aberration of the projection lens. We will confirm and estimate the effect of coma aberration on the overlay error by optical simulation. The origin of the offset error is considered to be coma aberration due to lens element decentering. This `decentering' coma aberration causes the offset error and the effect of `decentering' coma aberration has the feature dependence.

Exposure tool optimization in process development today extends beyond the classic concepts of exposure and focus setting. The lithographer must understand and tune the system for critical feature performance using variables such as Numerical Aperture (NA), Partial Coherence (PC), and critical level tool matching. In a previous study, the authors demonstrated that the phase-shift focal plane monitor (PSFM) accurately measures focal plane variations when appropriate calibrations are employed. That paper also described the development of a model for classic aberrations such as Astigmatism, Field Curvature and Coma. The model considered geometrical aberrations (Seidel) with radial symmetry across image plane as being the primary contributor to CD variation across stepper image plane. The publication correlated image plane focal results to an approximation of the stepper's Critical Dimension (CD) behavior for a matrix of NA and PC settings. In this study, we continue the analysis of the focus budget and CD uniformity using two generations of optical steppers in a 0.35 micrometers process. The analysis first addresses questions involving the use of the PSFM including the variation of calibration across the exposure field and the advantages of using field center or full field calibrations. We describe a method of easily measuring the uniformity of NA and PC across the exposure field. These new tools are then applied as an aid in lens image field characterization and tool-to-tool matching. The information gathered is then applied to measure image aberrations and predict CD variation across the image under various conditions of focus. The predictions are validated by a comparison against CD uniformity as measured by a commercial Scanning Electron Microscope. Present work confirmed previous work and recent assumptions that Zernike diffraction theory of aberration is most appropriate for current stepper lenses with local image plane focal variations across entire field being the major contributor to field CD variations.

We study the optical proximity effect and its correction using empirically derived models for DUV lithography taking into account random process variations. The sensitivity of corrected configurations to different sources of process variation (exposure dose, defocus) is evaluated. For correction at a centered condition (optimum dose, zero defocus), problems may arise in ill-conditioned areas (inside corners of T-shape features, butting line-ends, etc.), when going away from the best focus and/or exposure dose, within the exposure/defocus window. Correction for sharp corners (aggressive correction) shows a stronger sensitivity to defocus than less corner sharpening (conservative correction). Furthermore, we study what types of design configurations tend to print poorly with process variations and investigate alternative correction optimization schemes that stabilize the printing performance in such areas. Various optimization alternatives to improve performance within the process window are evaluated.

In this study, the optical proximity effect (OPE) of positive and negative tone line patterns is compared under a variety of exposure conditions. 0.25 micrometers lines with various pitch sizes were printed by a KrF stepper, and the CD variation as a function of pitch was evaluated. We found that the OPE was suppressed significantly in negative patterns under various conditions. The effect of resolution enhancement techniques on the OPE is also investigated. We found that negative patterning with the combination of off- axis illumination and attenuated phase-shift masks not only improved DOF but also gave a small OPE.

Optical proximity correction (OPC) was applied to alternating phase shift masks to improve printed resist pattern fidelity. Mask patterns were modified with jog type corrections. DRAM cell patterns were exposed by using a 0.55 NA, 0.36/0.55 (sigma) , KrF excimer laser stepper onto 0.5 micrometers thick chemically amplified negative resist. With 0.55 (sigma) , OPC was effective and printed resist pattern was very close to designed one. However, with 0.36 (sigma) , large pattern deformation was observed due to coma aberration.

An optical proximity correction algorithm based on statistical methodology is developed. The response surface function (RSF) for the CD in the lithographic process is extended by introducing variables for the mask pattern size. The values of process parameters and mask pattern size are concurrently optimized by using the RSF. This methodology allows design for manufacturability, considering error distributions of process parameters such as focus position and exposure dose. The algorithm is applied to a DRAM cell pattern. The result indicates the annular illumination with larger coherency than that of the conventional illumination improve the CD limited yield.

The implementation of a simple, semi-empirical resist model into an OPC algorithm, which up to now uses aerial image simulation, is described. The model assumes that the main component of proximity effects comes from the aerial image. It uses two pattern density functions to describe the shift in edge placement due to resist and etching processes. Besides the parameters for the aerial image (numerical aperture, coherence, wavelength, lens aberrations, defocus, etc.), the model needs only four additional parameters. The model is tested using resist simulation and electrical linewidth measurement data from fully processed testwafers. For linewidths of 350 nm and larger, printed with i-line lithography into a standard i-line resist, the OPC algorithm with the implemented model reduces proximity effects to less than 10 nm. A similar performance is indicated by preliminary data of electrical linewidth measurements.

The matching of wafer steppers is accomplished typically by patterning two successive layers, using different steppers of interest for each layer, and measuring the overlay at many points in the exposure field. Matching is considered to be optimized when some metric, such the sum-of-squares of overlay errors, is minimized over all measured points within the field. This is to be contrasted to the situation which arises during the in-line measurement of overlay errors in production, where a far more limited sampling of points is involved. There are several consequences to limited sampling. Adjustable intrafield overlay components, such as magnification, may appear to vary up to several parts-per- million as a consequence of varying chip size. These variations are substantially larger than the normal variations of these components for fixed field sizes, and so have significant consequences for the application of statistical methodologies to the control of overlay components. The width of the distribution of overlay errors across the field may typically increase between 10 to 20 nm (3(sigma) ), with even larger increases in mean shifts, all varying with field size. Reticles may also introduce similar variations, both random and systematic. Reticle beam-writer errors lead to systematic intrafield errors, particularly asymmetric field magnification and field skew. Steppers may compensate for these systematic reticle errors, and step- and-scan systems are more effective at this compensation than step-and-repeat machines. For steppers which have process dependent alignment, this compensation must be determined on products, which leads back to the problems associated with limited sampling. Correction for the overlay errors induced by limited sampling may be accomplished by look-up tables incorporated into the overlay analysis software. For each pair of steppers and each sampling plan, corrections can be applied at each measurement point in order to bring the full field and limited sampling plans into consistency. This will lead to a true minimization of overlay on product and will stabilize statistical process control of overlay components.

The feasibility of manufacturing 280 nm gates for ASIC technology using i-line lithography is examined. Off-axis illumination, sub-resolution assist features and proximity effect bias corrections were considered. The experiments were performed with a reticle designed to evaluate the effects of line pitch, bias and field uniformity on the feature dimensions. Results show that dense and isolated features were found to print at about the same linewidth under all three illumination conditions. However, deviations as large as 40 nm were found at intermediate pitches, implying that some form of optical proximity correction is needed to maintain critical dimension (CD) control for a mask pattern with varying feature densities. Sub-resolution assist lines adjacent to isolated 280 nm lines significantly improved the apparent wall angle of the features compared to true isolated features. The use of these features comes at a cost; the sub-resolution features can be printed under certain conditions and could possibly lead to device failure. Multi-dimensional matrices of CD measurements with varying dose, focus, bias and pitch, when displayed in an appropriate manner, are being used to identify the relative advantages of different illumination conditions. Off-axis illumination offers a large depth of focus for all pitches if proximity effect biasing is applied. Conventional illumination with biasing can improve exposure latitude.

As the device performance requirements tighten to improve speed distributions, the speed binning caused by across the wafer critical dimension (CD) variation will have significant impact on manufacturing performance yields. Overall speed performance yields are impacted by wafer to wafer and across wafer poly CD and poly profile variation caused at photo and etch. With the use of our final technique, called `Exposure Compensation', we were able to compensate for CD variation seen at final FI CD and electrical test. Implementation of exposure compensation process gave an effective two week control that exceeded the goal of 90 nm total electrical distribution. The improved control at FI CD can clearly be seen from the results of electrical measurements and transistor performance. While the techniques adequately improve site to site variation, further work is still required to improve across field variation.

Critical level photolithography optimization for 0.35 μm design rules faces new challenges threatening the process control and jeopardizing manufacturing yields. To address these challenges we have adopted a methodology combining photolithography simulation and limited proof-of-principle DOE. It allowed us to analyze the process control requirements and to establish photolithography strategies resulting in optimum manufacturing performance. As manufacturing CD's continue to shrink towards and beyond the wave length at which photolithography is performed, the lithography process control becomes limited by optical proximity effects. The proximity effects modify the geometry of the reticle features printed in the resist and thus they impact both CD control and overlay performance. The proximity effects can be highly non-linear and thus difficult to quantify unless sophisticated analytical methods are adopted. Another challenge faced by IC manufacturing rose with the advent of illumination options [l, 2]. In particular, off-axis illumination options offered with commercial exposure tools expand the exposure capabilities of steppers. Inevitably, these new illuminators complicate optimization of photolithography process. If the optimization task is to be executed in a purely empirical fashion, the optimization resources requirements become prohibitive. To simplify the photolithography optimization cycle, we have developed methodology combining process modeling [3] and limited proof-of-principle testing; the modeling was done with Nikon Corporation's proprietary simulator. Our methodology was employed in photolithography process optimization of a set of critical levels used in manufacture of a DRAM chip. Photolithography modeling in support of the process optimization resulted in a set of illumination and projection optics alternatives. The optimization of illumination strategies involved conventional and off-axis exposure options. The modeling identified illumination tradeoffs in terms of process depth of focus, DOF, exposure latitude EL, and stepper throughput. The results of the analysis suggested off-axis illumination as a basis for a robust exposure process. The results of the modeling served as a basis for a reduced design-of-experiment, DOE, quantifying the CD characteristics of the levels. We present the results of photolithography optimization of a chosen critical level exposed with an i-line stepper, and discuss the merits of the methodology. The discussion presents examples of optimization solutions. This report reviews the results of the modeling, including aerial image analysis and resist simulation. The report also reviews the results of the proof-ofprinciple metrology. We compare the modeling and the metrology and draw conclusions on the quality of the models' predictions. We interpret the model results in terms of CD characteristics of the critical level features exposed and developed in the resist. The aerial image analysis and resist simulation allowed us to reduce the optimization space. This, in tum, reduced the resource required to conduct the limited, proof-of-principle verification. We stress the advantage of using the methodology combining process modeling together with limited, proof-of-principle metrology to simplify and to shorten the exposure optimization process.

There are many commercially available deep-UV resists today that show performance gains above Apex, which has been used in 0.35 micrometers production for some time. In addition there are inorganic antireflection layer (ARL) schemes that can now be used with these newer resists, which were difficult to use with Apex due to substrate sensitivity problems leading to footing. We will require the benefits that these newer deep-UV resists and inorganic antireflection coatings can provide as critical gate dimensions fall below the 0.25 micrometers regime. A key benefit of the use of inorganic layers for antireflection control is the ability to maintain conformality, thus avoiding critical dimension (CD) changes that occur as organic antireflection layers are coated over topography. In this paper we evaluated two of the leading edge positive results in a 0.25 micrometers pilot process on ASML 5500/90 and 5500/300 steppers. We examined basic process characteristics such as depth-of-focus (DOF), exposure latitude, dense/isolated bias, post-exposure delay stability, CD swing effects, and substrate sensitivity. This study shows that the benefits of a new resist on an inorganic ARL are reduced swing effects, decreased post- exposure-bake sensitivity, and improved delay stability, all of which result in improved CD control. A slight decrease in DOF and exposure latitude is observed, however the resultant process latitude is sufficient for sub 0.25 micrometers lithography. Results of other tests as well as data off of the ASML PAS5500/300 are also presented.

The evolution of microlithography to 0.25 micrometers and below has driven the need for performance enhancements in several critical areas. Among these are imaging, illumination, and overlay. This paper briefly reviews Micrascan^TM III system concepts. The main body of the paper presents system level performance and discusses the key subsystems which enable 0.25 micron imaging and 55 nm overlay. Autocal, Micrascan^TM III's image, reticle and wafer position reference subsystem is discussed with respect to functionality and performance with a pulsed illumination source. Micrascan^TM III illuminator performance, including automated off axis illumination module are presented. Performance of the magnetically levitated Monostage and its interaction with overlay and imaging is discussed. System performance with respect to resolution, image quality and overlay on product levels is presented and analyzed.

To meet the high productivity standards, set by current top- end stepper systems, the use of excimer laser sources and high scanning speeds are essential. This paper reports on a new Step & Scan system capable of exposing 26 X 33 mm fields, using a 248 nm DUV-lens with a variable Numerical Aperture of 0.40 to 0.63. The system is equipped with an advanced AERIAL^TM illuminator which allows the user to choose coherence and illumination modes on a job-by-job basis. The double telecentric lens is equipped with lens manipulators to allow on-site aberration control. Results are presented on dynamic image distortion, field flatness and dynamic imaging performance. Performance of the overlay accuracy and dose accuracy at high scanning speeds proves that Step & Scan technology is now developed to a level suitable for use in high volume sub 0.25 micrometers manufacturing.

The faculties of current alignment sensors for CMP processes are studied. The alignment sensors are LSA, LIA and FIA, and all of them are currently equipped on Nikon's NSR series. The results showed that those sensors have enough faculties for present CMP processes. For much progressed CMP technologies, we have experimentally made a new alignment sensor and examined it. The new sensor is based on the `Phase Contrast Microscope' and is made by modifying the FIA optics. By using the new sensor, a mark with only 20 nm step height can be clearly detected with higher accuracy than conventional FIA, and the sufficient possibilities of the new concept alignment sensor for the `progressed CMP process' are confirmed.

As the most critical semiconductor device geometries shrink down to the quarter micron order, requirements for overlay accuracy also become increasingly critical in the actual semiconductor manufacturing process. Factors in overlay error (especially, alignment error) originate in the interaction of process and tool. It is therefore necessary to improve alignment accuracy from both the process and the tool sides. In an effort to solve this as a tool supplier, we at Canon must minimize tool factors to reduce alignment errors caused by the interaction of process and tool factors. We though that we needed some evaluation criteria with such interaction take into account, and redefined the concepts of tool induced shift and wafer induced shift as a criterion. This paper introduces these new concepts and discusses validity of the criteria showing experimental results of alignment accuracy implementing the idea in the real process.

Lithography steppers currently use a combination of mechanical stages to achieve control of a wafer location in six degrees of freedom. In these stages, the wafer is carried on a fine stage which provides six degrees of freedom control with approximately 100 micrometers travel. This fine stage is mounted on a coarse mechanical stage which provides X-Y positioning with approximately 200 mm travel. The fine stage is typically comprised of multiple piezo- actuators and/or voice coil drives which are used to position a flexure mounted platen.. These mechanical stages generally suffer from poor dynamics, a result of the compound flexures used. For these reasons photolithography equipment manufacturers are researching alternate non- contact methods of precise positioning. This research has centered on replacing long travel degrees of freedom with air bearings which carry a fine stage capable of roll, pitch, and vertical displacement (Z). These air bearing/mechanical stages have improved resolution and stability in addition to shorter setting times. An alternate approach is the use of a magnetically levitated (Mag- Lev^TM) stage that provides six degrees of freedom control without mechanical contact. This type of stage is ideal for clean room use where particle generation from mechanical friction is a major source of contamination. Mag- Lev stages are also mechanically simple, therefore easier and cheaper to fabricate and more reliable than flexure stages. Previous papers have reported the performance of the fine set Mag-Lev stage in a laboratory lithography experiment. Integrated Solutions has chosen to develop a full Mag-Lev stage in three phases. The first Mag-Lev implementation replace all fine mechanical elements with a single suspended structure. This stage will have displacements of 300 micrometers in X,Y,Z and milliradians of rotation in the remaining three degrees of freedom. The stage will be carried over the surface of the wafer by stacked air bearings. This hybrid Mag-Lev/air bearing stage enjoys the benefits of non-contact bearings and will significantly improve stepper performance in both throughput and overlay. The second phase replaces one degree of air bearings with a six degree of freedom Mag-Lev stage with one axis long travel. This geometry eliminates one redundant degree of freedom and five uncontrolled rigid body modes of an air bearing axis. The final phase of development is the implementation of a six degree of freedom planar stage that will consist of a single moving part with two axis long travel (in excess of 200 mm). This stage geometry represents the tightest possible dynamic structure with no redundant or uncontrolled degrees of freedom. Very high bandwidths are possible with Mag-Lev stages which work to reduce settling times and improve disturbance rejection. This paper will focus on the development of two stages; a fine set stage as in phase 1, and a single degree of freedom long travel stage as in phase 2. Step and settle and position stability data for both stages will be presented along with the expected improvement in wafer throughput. In addition to the prototype stages future concepts for a phase three planar stage will be presented with expected performance goals.

The ArF excimer laser light source will extend the optical lithography to below 0.18 micrometers design rules. Still under discussion is the most effective layout of the stepper or scanner imaging optics. The decision for an all-fused-silica lens, an achromatic lens using CaF₂, or a catadioptric imaging system has great impact on the laser-bandwidth requirement. In addition, the potential performance and perspective of the laser must be considered in the system layout of the production stepper or step and scan tool. Cost effective operation of such a lithography tool requires a 193 nm excimer laser with high power and repetition rates in the order of 1 kHz or higher. Precise dose control of the exposure demands high repetition rate and excellent stability of the laser output energy. We have developed an ArF laser which can be operated with up to 800 Hz repetition rate based on NovaTube^TM technology. Optimized materials and discharge configurations have been used to achieve laser tube lifetimes above 10⁹ pulses. Up to 4 X 10⁹ pulses tube lifetime have been achieved in beta-site tests. Gas lifetime of several 10⁷ laser pulses is obtained. A solid-state switch has been adapted for the reliable and cost efficient excitation of the 193 nm lithography laser. To achieve laser output of different bandwidths various resonator arrangements have been investigated. The paper gives an overview of the currently achievable power levels at different bandwidths and discusses future trends.

The performance of argon fluoride excimer lasers is an important issue in determining the practical feasibility of 193-nm exposure systems. This paper presents a summary of the experience gained at MIT Lincoln Laboratory regarding the long-term performance of 193-nm lasers, used under conditions similar to those expected in production-type lithographic systems.

Advances in technology now make possible solid-state 193 nm lasers. Solid-state can operate with pulse repetition rates of > 10 kHz, minimizing peak-power damage to stepper optics. Furthermore, solid-state lasers are potentially more reliable and could have lower operating costs than ArF excimer lasers. Achievement of spectral linewidths < 0.1 pm for use with refractive lens systems is straightforward in solid-state laser systems. Ultraviolet solid-state laser technology is much less mature than excimer laser technology; so while there is far to go, there is much more potential for rapid progress in solid-state lasers than in excimer lasers. Aculight has begun a program to develop a multiwatt, 10 kHz solid-state 193 nm laser. Although efficient conversion from 1064 nm to 193 nm is easier for high peak power pulses, minimization of lens damage requires low peak power. Eventual goals for the technology are to achieve output powers 10 - 20 W at > 20 kHz repetition rate in > 10 ns pulses, limiting peak powers to < 200 kW. High pulse repetition rates will permit excellent dose control, and facilitate decoherence of the high-coherence beam from the solid-state laser system.

Highly narrowed Argon Fluoride excimer laser for practical refractive exposure system with high NA and wide field size lenses is developed. The laser realizes 0.6 W at 400 Hz. The spectral bandwidth is less than 0.75 pm, the stability of central wavelength is within +/- 0.25 pm and the concentration of energy within 3- pm band is more than 95%. The gas lifetime is more than 1 X 10⁷ pulses without gas purifier. By the one gas life performance test and the short term performance test, we confirm this laser is useful for the development of microlithography process.

We have developed an ArF excimer laser exposure system that combines a monochromatic quartz projection lens system and a spectrally-narrowed ArF excimer laser. The bandwidth of the laser was narrowed to 0.7 pm by using three prisms and an etalon. We fabricated 0.12 micrometers lines-and-spaces using an alternating phase shifting mask and a top surface imaging resist.

Non-linear multiple exposure (NOLMEX) method for forming a fine line and space pattern is described. The combination of the resist with non-linear photosensitivity and multiple exposures allows us to exceed the optical cut-off frequency, which has never been realized in the field of photolithography. In this paper, a new approach for NOLMEX method utilizing an ordinary resist instead of the non- linear resist is introduced. In the experimental results, 0.19 micrometers L/S pattern was resolved with a commercial i-line stepper using ordinary illumination, which indicates the potential of NOLMEX method.

We conducted studies of some optical imaging systems having super resolution pupil functions for 0.18 micrometers to 0.1 micrometers lithography as well as for high resolution inspections. The optical systems with these aperture stops have high transmissivity to illumination light beam if appropriate illumination optical systems are used, and have large depth of focus. When this optical imaging method is applied to optical inspection systems, the resolutions of these optical systems can be doubled. The theory of this new optical imaging method is discussed along with the theory and design of illumination optical system. The imaging method will be easy to implement, and can be used for steppers, direct writing lithography, and confocal microscopes.

Modified illumination scheme such as off-axis illumination or low partial coherence illumination in combination with phase shift mask is very reliable technique even to the manufacturing stage. In spite of its benefit of drastic improvements in resolution and depth of focus, it has several well known drawbacks. Those are pattern structure dependency, increased proximity effect and significant degradation of illumination power. Furthermore, several studies are reported about the distortion-like registration error between each illumination aperture while various apertures are mix-used. However, the origin of this distortion-like overlay errors is still uncertain and just considered as having some relationship with lens aberration probably, coma. In this study, we investigated the nonlinear character of the registration errors and try to find out the implicit relations of its to the lens aberration character through experiments and simulations.

In this paper, the mask critical dimension (CD) control budget for the 0.18 micrometers generation design rules are assessed via experimental based data. The lithography tool used for wafer exposure is 193 nm microstepper with NA of 0.6. The resist process used is the top surface imaging (TSI) resist process. In the study, the process CD variation due to mask CD error, mask CD error combined with process variation such as focus and exposure, mask CD error combined with low k_l imaging limitation, and mask CD error combined with the characteristics of a given resist process, such as feature dependent silylation, are investigated. It is found that for the lines with a relative larger pitch (e.g., pitch is approximately 2X the line width), the resist CD responds to the mask CD linearly down to a 0.16 micrometers feature size. However, unlike the most single layer resist processes, the slope of the line, i.e., (Delta) CD(wafer)/(Delta) CD(mask), in the above mentioned TSI process is not equal to 1. Their values are between 1.2 - 1.7, depending upon the pitch. This means that for a given mask CD error, the corresponding wafer CD error will be 1.2 - 1.7 times larger. In other words, the mask CD control budget will not scale the same as that of wafer CD control budget when the design rule scales. In fact, the mask CD control budget should be at least 1.2 - 1.7 times tighter than that of the wafer CD budget scaling. In the case of contacts, the resist CD linearity only extends to the 0.375 micrometers feature size with a slope approximately equals to 1. For the contacts smaller than 0.375 micrometers , the slopes are greater than 1 depending upon the exact mask contact sizes. For the contacts smaller than 0.25 micrometers , the slopes are between 2 - 3, depending upon the pitch. In addition, it is found that the defocus increases the slope value significantly for the small contacts sizes. The slopes for the case of lines, however, are found much less sensitive to the defocus conditions (defocus equals +/- 0.2 micrometers approximately +/- 0.3 micrometers ). Our assessment showed that mask CD control budget (4X) for 0.18 micrometers design rules are necessary to be in a range of +/- 8 nm approximately +/- 12 nm (4X), assuming that +/- 25% of total wafer CD budget which is +/- 10% of the target CD is given to the mask.

The print bias between isolated and dense resist pattern (I- D bias) which is originated from the optical proximity effect, is one of the major issues for the processing of randomly arrayed and complicatedly structured logic devices designed with the smaller features than ever. In this study, the characteristics of I-D bias which is related to not only optical parameters but also resist processing parameters were investigated quantitatively through the way of experiment and simulation. Examined processing parameters were as follows: numerical aperture and partial coherence factor ((sigma) ) as the exposure parameters; the ratio of line to space width (duty ratio) as the mask parameter; resist thickness, prebake temperature and post exposure back temperature as the resist processing parameters. The experiments were reviewed in simulation tools such as Prolith2+ and SOLID-C. And the resist patterns were acquired using DUV exposure tool with various kinds of resist.

Pattern collapse, CD (critical dimension) non-uniformity, and poor pattern fidelity induced by the multi-interference effect of DUV process are currently becoming significant issues to an actual process implementation. In the micro- device fabrication of DRAM, capacitor storage node patterning designed with a rectangular shape of sub- quartermicron spacewidth is critically involved in these issues dependent on the process scheme of capacitor formation. In this study, we have evaluated the effect of barrier layers such as thin poly film, inorganic and organic BARC to minimize a thin film multi-interference effect. In addition, illumination source effect during the exposure process and process polarity effect of a positive and negative tone process were also experimented to settle down process control stability.

The effects of increasing bottom-side anti-reflective coating (BARC) thickness on the CD distribution within a device are presented. In conjunction with the increasing BARC thickness, reductions in the photoresist thickness are shown to be beneficial. Significant reductions in CD variability and increases in the depth of focus versus CD spread are achieved with increased BARC and reduced photoresist thickness. Although significant improvements are seen with a thicker BARC film for the photolithography process, the importance of optimizing the etch process for the thicker films is shown. The effects of CD distribution on important electrical device parameters are also presented.

A new efficient inorganic anti-reflective coating (ARC) layer which refers to silicon oxynitride (SiOxNy) is evaluated and applied to 0.35 micrometers and 0.3 micrometers technologies using i-line lithography. This new bottom ARC layer can be applied to various substrates such as poly silicon, Wsix and Al by controlling the deposition conditions to change n&k values and film thickness, not only for i-line lithography but also for DUV lithography. Hereby we present the study of this new inorganic bottom ARC layer on poly silicon for i-line technology. The results show that it can enhance the process window by 30% for 0.35 micrometers 0.3 micrometers devices. The resolution limit of i-line lithography can be pushing down to 0.25 micrometers by reducing thin film interference in photoresist. Swing curve effect is reduced by more than 5X as compared to standard resist process. CD uniformity for 0.35 micrometers device is improved significantly across 8 inch wafers using bottom ARC. Notching and necking effects are totally eliminated as compared to standard top ARC process with 18% notching and necking effects. Optical proximity effects are reduced by 30%. No impact on overlay is found using SiON bottom ARC. Etching this SiON ARC layer is investigated and it is found that reasonable good etch profile can be achieved. Thin film uniformity variation study shows that +/- 50 A film thickness variation can be tolerated with +/- 5% CD variation. Thermal stability study of the ARC layer shows no peeling or bubble found at high temperature of 850 degree(s)C. No removal of this ARC layer is needed. In addition, the further study also shows that straight resist profile can be obtained by fine tuning the ARC layer thickness. As compared to organic bottom ARC, SiOxNy bottom ARC has the advantages of low cost of ownership, less defect density and shorter cycle time is needed. Prolith/2 simulation tool is used for the study.

The line width variations with various illumination types and sublayers are studied for deep UV light source and developed 0.25 micrometers isolated line pattern. The line width variation comes mainly from the reflectivity variation caused by multiple thin film interference effect. Compared to the conventional normal illumination, off-axis illumination shows bigger variation due to larger reflectance and worse aerial image around the best focus. The swing curves of several sublayers show different swing ratios and they depend on the optical properties and the thickness of sublayers. Antireflection coating (ARC) should be used in order to control the line width and the desired variation of less than 0.05 micrometers can be obtained with 0.19 micrometers thick ARC whose refractive index is 1.663 + 0.2 i. We also find that it will provide better latitude and line width control as well as more throughput when we used the resist thickness that will give minimum line width in the swing curve.

The difficulty of using focused ion beam (FIB) tools for embedded phase shift mask repair is the removal of gallium ion stain. Conventional anti-staining techniques, such as post-repair plasma etch or XeF₂ assisted FIB sputtering, is not applicable to embedded phase shift masks because of the phase shifting error introduced by additional substrate etch. A novel techniques, focused ion beam biased repair, has been developed to eliminate the post-repair residual defects caused by gallium ion staining. The biased repair involves sputter removal of an opaque defect to an area larger than its original size. The reduction of light transmission due to gallium stain is then effectively compensated by the enlarged repair. The concept of introducing a bias is also extended for clear defect repair, which makes it possible to repair a clear defect by using simple FIB induced carbon deposition with no need to create a proper transmission and phase shift. The new technique has been confirmed by both computer simulations and experimental repairs of MoSi and CrON embedded phase shift masks.

Several phase shift mask designs such as subresolution type, subresolution alternating type, outrigger type, and rim type for contact hole patterning were theoretically and experimentally evaluated in terms of their lithography performance. The various designed layouts which have the phase regions of manifold dimensions were simulated with lower coherent factor ((sigma) equals 0.3) and higher coherent factor ((sigma) equals 0.62). The detailed evaluation was made in terms of the log slope of the log image at nominal feature edge and side lobe effect of the intensity profile.

We have developed an edge-phase-shifting (EPS) lithography for the fabrication of sub 0.3 micrometers T-gates using a 5X i-line stepper with a 0.4 numerical aperture lens. Two exposures have to be accurately aligned on each other within 150 nm. The first exposure uses EPS lithography and defines the gate length in a negative resist. The second exposure defines the cross-section of the T-gate. Using this lithography, hetero structure field effect transistors can be fabricated.

A modern pattern generator needs very high data-handling capacity, partly because the number of features per mask is growing, partly because OPC and other sophisticated lithographic enhancements increase the complexity. MICRONIC makes a wide range of laser pattern generators for display, reticles and other electronics applications, and therefore needs a hardware/software datapath that is configurable over a wide range of requirements and cost levels. An extendible multiprocessor architecture offers this flexibility: possibility for multiprocessing for very high power, and at the same time ability to configure small systems for low-end applications. Preliminary benchmarks show a speed increase over previous systems by 10 to 60 times for a single processor.

As the integration density in the manufacturing of IC's increases and tighter design rules are implemented, the accuracy of overlay in the photolithography process is becoming all the more important. Consequently, investigation and characterization of the accuracy (as well as precision) of the overlay measurement are being required to insure that the overlay metrology tool qualifies for the new technologies. In this paper, we analyze the relationship between wafer substrate types and the respective characteristic overlay measurement errors associated with them. We compare results using different illumination wavelengths in the overlay tool. We define the Wafer Induced Shift (WIS) contribution to the measurement accuracy error and introduce a metric for it. And we analyze the relationship between the TIS and WIS. We show that the wavelength of the optical overlay measurement (OL, for short) should be fitted separately to each of the various kinds of overlay measurement targets, or marks. Buried type targets, covered by transparent materials, for which the measurement is made through the transparent interlayer, benefit from wavelength selection due to the possible improvement in contrast associated with an interference effect. Open type targets, on the other hand, in which the measurement signal is collected off the substrate shape itself, do not benefit from that. There is another problem to consider, however. Such targets may show a considerable WIS, even thought TIS and precision are good both before and after etch. The WIS error can be as high as 100 nm. This could create serious OL problem and it means that TIS and precision are not enough to characterize the quality of measurement, and WIS is equally important, it not more so. We look into the causes of WIS in the substrate shape, overlay mark design and fabrication method. We show that with careful selection of illumination wavelength, some layers produce no WIS. On the other layers WIS remains a problem and other methods are still required to control it.

A method of stepper set-up, maintenance, and tracking is described that is intended to match the overlay performance of multiple-model steppers in a multiple technology manufacturing line. The philosophy of this method is to determine, track and control systematic errors associated with overlay, and to measure and characterize the non- systematic or random errors to determine the base line performance capability of the stepper tool set. A description of the stepper set-up procedure, the data collection and analysis routines, and data from these procedures will be presented. A matrix of baseline stepper- to-stepper performance will be shown, along with additional analysis of non-systematic errors that were found to be not very `random'. An example of line performance after stepper set-up will be shown, along with a discussion on differences between set-up analysis and production lot measurements.

Patterning photoresists on reflective topography such as aluminum is one of the more difficult problems in device manufacturing. Interference effects caused by reflected light from the substrate/photoresist interface and surface topography result in coupling of additional energy into the film. This leads to linewidth variation known as reflective notching which severely impacts process latitude and increases critical dimension variation. For many years, suppliers approached the problem by adding dyes that absorb in the actinic region to create a larger non-bleachable absorption. In recent years, strongly absorbing intermediate layers or ARC's, both organic and inorganic, have seen widespread implementation to control reflective notching. However, if a fab is not equipped to accommodate the required ARC process, the processing can be very time consuming, cumbersome and costly. This study was undertaken to determine if a non-ARC, i-line photoresist process could be developed to reduce or eliminate aluminum reflective notching and accompanying critical dimension variation. This study was designed to screen, identify, and characterize various resist chemistries. Based on the screening characterization, a final, cost effective resist chemistry without ARC was selected, fully characterized and transferred into production. The selected material is currently being used in a high volume 0.60 micrometers CMOS, 200 mm wafer manufacturing process.

This paper discusses an extension of the fast multipole method to electromagnetic scattering from doubly periodic, multilayer wafer topography. The novelty of our approach lies in the use of a pseudo-periodic translation operator which can be computed efficiently using fast Fourier transform. Results obtained using the rigorous boundary conditions for dielectric surfaces are compared with those obtained using the approximate impedance boundary condition. The latter is shown to give good results for the type of topography usually encountered in lithography simulation. Results of reflective-notching simulation using the IBC method are presented.

We have developed a 3D practical lithography simulation system and realized the practical parameter optimization. A very fast 3D profile computation is performed by using the FFT algorithm in aerial images and the post exposure bake. The optimization system is developed by using object- oriented technology in order to reduce the computation time. The resulted computational time is less than one tenth of the conventional system. The total parameter optimization time is also reduced.

Several commercial photoresists are characterized with respect to their changes of the real part of the refractive index during the bleaching process. A finite difference beam propagation algorithm is used to evaluate the impact of these refractive index changes on the lithographic process. It is shown, that the refractive index changes result in modified process windows, side wall angles, swing curves and iso/dense behavior.

As the critical dimension (CD) of optical lithography processes continues to decrease, the process latitude also decreases and CD control becomes more difficult. As this trend continues, lithography engineers will find that they require improved process optimization methods which take into account the random and systematic errors that are inherent in any manufacturing process. This paper shows the methodology of such an optimization method. Lithography simulation and analysis software, combined with experimental process error distributions, are used to perform optimizations of numerical aperture and partial coherence, as well as the selection of the best OPC pattern for a given mask.

In state of the art IC design and manufacturing certain lithography layers have unique requirements. Latitudes and tolerances that apply to contacts and polysilicon gates are tight for such critical layers. Industry experts are discussing the most cost effective ways to use feature- oriented equipment and materials already developed for these layers. Such requirements introduce new dimensions into the traditionally challenging task for the photolithography engineer when considering various combinations of multiple factors to optimize and control the process. In addition, he/she faces a rapidly increasing cost of experiments, limited time and scarce access to equipment to conduct them. All the reasons presented above support simulation as an ideal method to satisfy these demands. However lithography engineers may be easily dissatisfied with a simulation tool when discovering disagreement between the simulation and experimental data. The problem is that several parameters used in photolithography simulation are very process specific. Calibration, i.e. matching experimental and simulation data using a specific set of procedures allows one to effectively use the simulation tool. We present results of a simulation based approach to optimize photolithography processes for sub-0.5 micron contact windows. Our approach consists of: (1) 3D simulation to explore different lithographic options, (2) calibration to a range of process conditions with extensive use of specifically developed optimization techniques. The choice of a 3D simulator is essential because of 3D nature of the problem of contact window design. We use DEPICT 4.1. This program performs fast aerial image simulation as presented before. For 3D exposure the program uses an extension to three-dimensions of the high numerical aperture model combined with Fast Fourier Transforms for maximum performance and accuracy. We use Kim (U.C. Berkeley) model and the fast marching Level Set method respectively for the calculation of resist development rates and resist surface movement during development process. Calibration efforts were aimed at matching experimental results on contact windows obtained after exposure of a binary mask. Additionally, simulation was applied to conduct quantitative analysis of PSM design capabilities, optical proximity correction, and stepper parameter optimization. Extensive experiments covered exposure (ASML 5500/100D stepper), pre- and post-exposure bake and development (2.38% TMAH, puddle process) of JSR IX725D2G and TOK iP3500 photoresists films on 200 mm test wafers. `Aquatar' was used as top antireflective coating, SEM pictures of developed patterns were analyzed and compared with simulation results for different values of defocus, exposure energies, numerical aperture and partial coherence.

A 3D planar pseudo-vector approach to modeling resist exposure has been implemented for optical projection lithography. The approach allows for high numeric aperture effects revealed by others in the resist latent images; comparisons made show its fine agreement with rigorous vector modeling up to NA value of 0.7. Fast Fourier transform technique has been applied to reduce mask and resist (including postbake) modeling runtime; the effects of finite accuracy approximations of the mask areas and mask periodicity have been investigated and compensated. The overall performance of the exposure model has been optimized to be 10X faster than for a similar vertical propagation model. A modified `cell' algorithm has also been used to model resist development; this resulted in more accurate resist profiles for the same latent image accuracy, providing additional runtime savings to resist profile simulations.

A method for simulating exposure, post exposure bake and development process is described and applied for analyzing the problem of reflective stepped substrate. The method is based on the numerical solver of Helmholtz equation and the rigorous boundary condition which considers the entire system consisted of projection optics, photoresist and substrate. Diffusion equation and the ray tracing model are used for simulating the post exposure bake and the development process, respectively. The developed photoresist pattern on the stepped reflective substrate topography are obtained, and it shows irregular resist profiles due to the reflection of light at the step edge.

The problem of accurately describing the topography of a mask arose in x-ray lithography, when the absorber structures have dimensions of several hundred wavelengths and may have topography. However, the application of phase shifting structures and a refinement of the image formation calculation even in optical masks has recently raised the same issue for the case of optical lithography. This paper reports the design and implementation of Topography Description Model (TDM), a rule-based 3D topography generator that allows the user to describe the exposure system, such as a system comprising of mask, gap, resist and wafer. The paper illustrates some applications of TDM to real-world studies, such as studying the effect of absorber side-wall slope on the final image and optimizing XRL masks.

Reflective notching from projection printing a narrow feature over a reflective topography was simulated with the 3D electro-magnetic simulation program TEMPEST. Various topographical and optical parameters were varied to determine their effect on the reflective notching and to gain insight into the mechanism of reflective notching. It was determined that corner angles in the topography and anti-reflection coating use are important parameters while resist surface angle and polarization are not.

In this paper four radically different methods for solving scattering of light off of mainly lameller grating type structures are compared. The methods are: (1) the analytic waveguide method (Metrologia), (2) the integral equation system with parametrization and sampling along the profile (IESMP) developed at BIFO, (3) the integral equation method with discretization along the x-axis (PC-Grate 2), and (4) the rigorous coupled-wave analysis (RCWA). The results show that the analytic waveguide method and the integral equation method with parametrization (IESMP) agree with extremely high accuracy in all cases studied for both polarization states TE and TM. For the special case of near field slit simulation the same high agreement is found between Metrologia and a FEM method. It is also found that the agreement of RCWA and PC-Grate 2 with either of the other two methods is only fair, particularly for TM mode on gratings made of good conductors. Using an improvement, the RCWA results become significantly better and correspond with the results of IESMP and Metrologia.

Signamization is an automated process to optimize the NA and (sigma) settings of a stepper for a given combination of critical features and CD tolerances, basing on the E-D Forest methodology. It has been applied to a variety of feature combinations using disk and annular illuminations, binary intensity mask and phase-shifting mask. They are based on pre-simulated aerial images. This paper deals with signamization of resist images derived from aerial image data. The aerial image is converted to resist image analytically in real time using the lumped parameter model, which make any pre-simulated aerial images; such as those for binary intensity masks, attenuated phase shifting masks, alternating phase shifting masks, disk illumination. Nikon ring illumination, and Canon ring illumination, usable for any resist system characterized by resist thickness, absorption coefficient (alpha) , and contrast (gamma) , without having to simulate for each particular resist system. The signamization results show that in addition to the optimum NA and (sigma) , there is a (gamma) threshold above which further improvement is unrewarding, a higher (alpha) can improve imaging performance, and that the resist thickness plays an important role only when the reflection swing ratio is large or when it is reduced significantly. When this happens, together with (alpha) becoming very small and (gamma) very large, and resist image approaches that aerial image, showing the validity of the lumped parameter model for signamization and the computer programming.

In this paper we report on the method of measuring the magnitude of scattered light, flare in a stepper lens and its effect on the overall process latitude. By measuring the required energy to clear large square patterns, it was possible to calculate the amount of scattered light in a lens system. When the size of a square pattern is larger than 10 micrometers , the magnitude of scattered light saturates to be independent on the size. But it gradually increases as the clear to dark area ratio of the mask increases. Simulation result shows that scattered light degrade the MTF of an aerial image and when it becomes larger than 2%, the process latitude start to decrease. Measuring scattered light with different mask opening ratio reveal that it exceed 2% when the mask opening ratio is around 70%. There was no significant difference in scattered light between different DUV scanner tools when the patterns were larger than 10 micrometers , but some difference was observed at the smaller patterns. Overall process latitude were measured with masks with different opening ratio. The exposure latitude decreased as mask opening ratio increase for positive photoresist, depth of focus was not significantly affected at increase of 20 to 50%. But we can observe 0.2 micrometers of focus margin decrease for mask open ratio change from 50 to 80% for both 260 and 240 nm patterns. So if we assume that there is no difference in process performances between positive and negative photoresists, using a negative photoresist will give more process latitude.

The most critical issues for lithographers, resolution, depth of focus, exposure latitude and proximity effect, have been frequently discussed. In addition to those issues, lens aberration effects such as astigmatism, coma, field curvature and distortion have been emphasized in view of process stabilization. Deep UV lenses have undergone improvements in the field curvature, distortion, and other aberrations to the point that they can be implemented for near quarter micron patterning. KrF excimer laser stepper and scanner have made Deep UV technology as a practical choice for lithographers to develop their own DRAM processes. In order to adopt process development beyond the region of ultimate resolution limit, precise lens aberration analyses are necessary. KrF excimer laser stepper (NA equals 0.55), used for 0.18 micrometers design rule patterning, is examined. Several types of real device pattern, such as word and bit line, active, storage node, contact hole patterns, are tested under various illumination conditions (conventional, various partial coherence, off-axis illumination). These results are compared with simulation results.

To get wide lithography latitudes in ULSI fabrication, an optical proximity correction system is being widely used. We previously demonstrated that the optical proximity effect is highly dependent on beam interference conditions. By using an aperture with a spindle shaped opaque region and a controlling interference beam number optimized for imaging, we can obtain a high correction accuracy of less than +/- 0.01 micrometers for all kinds of pattern. To put the optical proximity correction into practical use, we must fabricate the corrected mask either by an EB or a laser writing system. But during mask writing, there is another problematic proximity effect. The optical proximity effect caused by mask fabrication error is becoming a serious problem. In this paper, we estimate the optical proximity effect caused by mask fabrication error. For EB writing, the mask feature size of 0.35 micrometers line changes dramatically in a space less than 0.8 micrometers in size; this is not tolerable. For a large pitch pattern, modified illumination reduces the DOF to 0 micrometers . Otherwise, laser writing stably fabricates a mask feature size for a 0.35 micrometers line, and the modified illumination reduces the optical proximity effect. This resist feature fluctuation is binary, so, correcting the mask pattern is easy. Although, it was wrongly thought that for larger pitch pattern, the DOF was reduced by the modified illumination, the DOF reduction actually came from the combination of the two proximity effects. Using an accurate mask produced by a laser writer, we do not observe any DOF reduction in modified illumination. Moreover, this has led to development of an optical proximity correction system with EB proximity correction.

Different 1995 - 1996 grade experimental fused silica samples were evaluated for their resistance to UV-induced compaction at 193 nm under elevated sample temperature conditions. Stress induced birefringence was used as a sensitive compaction monitor. We found that compaction rate decreases with increasing sample temperature. Isochronal annealing experiments were performed on two different sets of pre-damaged fused silica samples. Annealing of compaction was observed at temperatures as low as 200 degree(s)C, and an activation energy of 0.1 eV was found.

Air turbulence affects the performance of the Helium-Neon interferometer used to control the wafer stage of stepper or step-and-scan lithography systems. In this paper, we describe characterization and reduction of the major error sources in an Air Turbulence Compensated Interferometer designed to address this problem.

A small field ArF excimer laser based exposure tool has been designed and fabricated for the 193 nm lithography process research and exposure tool development. The projection optics based upon Schwartzchild concept has a specification of 3 mm field diameter, X5 reduction ratio, and 0.5 NA. The exposure tool uses an unnarrowed ArF excimer laser as a light source, and uses a fly's eye homogenizer to produce a reticle illumination uniformity of < +/- 5% RMS/pulse. The results of preliminary exposure with PMMA resist coincided with simulation and expectation, and advanced imaging tests carrying out for various resists. In this paper, we report the detailed system parameters and characterization data of the small field ArF excimer laser exposure tool and some of advances in 193 nm lithography that have been achieved with the system.

Chip makers are gearing up for 686-class micro-processors and 64 Mb production. These require producing < 0.30 micrometers critical features. They are also focusing on 256 Mb DRAMs that require below 0.25 micrometers design rules. Therefore, the shift from mercury lamp based i-line to 248 nm Deep-Ultra- Violet (DUV) steppers has begun in earnest. Current KrF laser models from several suppliers satisfy the optical requirements necessary for pilot production. However, as the move from DUV lithography R&D environment to production occurs, the laser's manufacturability, uptime requirements and cost-of-operation (CoO) become more demanding. The chip maker requires reliability, availability and maintainability data from laser manufacturers to support uptime and CoO estimates. Often, such data are required long before the lasers actually go into production.

Stable and efficient UV-laser systems require stable and efficient optics which can only be realized by applying high performance, low loss and highly damage resistant thin film interference coatings. The interaction between excimer laser photons and optical coatings is determined by combined effects of high repetition rates and high energy densities. Optical stability investigations on oxide and fluoride thin films have been performed to estimate ns-laser-induced- damage-thresholds. By using thermal mirage technique both the detection of the damage onset after illumination with KrF excimer laser (248 nm, 20 ns) and the exploitation of the increase of the signal per unit fluence for the interpretation of the origins of radiation damage were possible. Damage sensitive defects were identified to origin both from residual gas and evaporation sources. By optimizing electron beam evaporation technology the defect density was drastically reduced.

In ArF excimer laser lithography, catadioptric systems are widely used as projection optics. These systems consist of a mirror, a polarizing beam splitter, quarter wave plates and corrector lenses. Combination of a polarizing beam splitter and a quarter wave plate is used to separate the forward and backward beam and to remove the central obscuration. The life time of the ArF excimer laser optics is limited by the well-known defect, the compaction of fused silica. The compaction is proportional to the square of energy density, and the life time is given by the inverse of compaction. If we make the image space working distance longer, the energy density becomes smaller and the life time becomes longer. We have designed two examples of catadioptric systems and analyzed their optical performances and life times. Also we have reviewed the relationships between the working distance and the size of optical component, however, there is a strong relationship between both of them. The longer working distance makes the optical component larger. It has the significant meaning in optics design, because the big sized optical material, which can be used at 193 nm, is very rare. Especially, the material for prism of polarizing beam splitter is hard to attain. Hence there should be some compromises between the working distance and the size of optical components.

A new concept based on a Fabry-Perot interferometer for the generation of nearly nondiffracting Bessel beams is described experimentally and theoretically, and proposed for potential applications in microlithography such as the fabrication of small isolated patterns. It was demonstrated that the depth of focus can be increased by a factor of about 2, and simultaneously the transverse resolution improved by a factor of 1.6, when using this technique to image contact holes. The theoretical curves show very good agreement with the measured intensity distribution. The properties of simultaneous imaging of two contact holes were also investigated. It was shown experimentally that even in the most critical case (when the first diffraction rings overlap), undesirable interference effects between the adjacent contact holes can be eliminated by means of a phase shifting technique.

In order to study the feasibility of submicron lithography using second-harmonic light (SHL) at wavelength of 255.3 nm of copper vapor laser (CVL), a 1:1 catadioptric projection lens and an illumination system are designed and constructed. A resolution of 0.7 micrometers line/space patterns is produced in AZ1350J resist, which reveals that the SHL of CVL can be used as an illumination source in submicron lithography.

We have designed and constructed an excimer laser lithography system at 248.3 nm. The system consists of a KrF excimer laser illuminator and a 1:1 broadband catadioptric projection lens with a half circle image field. Resolutions of 0.6 micrometers L/S and 0.5 micrometers L/S are obtained in AZ1350J resist and PMMA resist respectively.

One of the most crucial emerging challenges in lithography is achieving rapid and accurate alignment under a wide variety of conditions brought about by the different overlying films occluding the marks. The problem is exacerbated by planarizing processes such as chemical mechanical polishing that reduce the topographical contrast used to view the marks and by distortion of the wafer and of the stage. Thus an effective learning process is needed to rapidly acquire the best possible positional information from an array of the marks across the wafer. In this paper, a neural network model for global alignment is described. This method has significant advantages including robustness to measurement noises and other random disturbances. Stage distortion can also be easily included in the model, resulting in more accuracy in the presence of a known repeatable distortion. The process can learn to identify and ignore any alignment marks that yield significantly erratic signals. Different approaches for training of the system will be discussed. A few common distortion functions will also be used to test the model. Preliminary simulation results show very accurate alignment in the presence of noise and distortion. In addition the simulations show that stage distortion has no effect on the results. In other words, increasing stage distortion amplitude will not change the positional accuracy in a given time.

In this paper, a new automatic model-based Optical Proximity Correction (OPC) approach is reported. It combines a fast aerial image solver and a physically based empirical resist model. The fast aerial image calculation is achieved by using the algorithm of the coherent decomposition of the partially coherent optical systems. The first order resist model of Brunner is extended to accommodate intensity log- slope variations. The modified model can be calibrated by either experimental data or resist process simulations. In this paper, resist simulations using Depict are selected to fit the model parameters. Defocus effects in OPC are also discussed. A new scheme that combines edge biasing and the addition of sub-resolution features is shown to greatly improve the process latitude.

In this paper, the printability of mask defects at 0.18 micrometers design rule is studied via 193 nm top surface imaging (TSI) resist process. The mask defect printability is determined by taking into account the wafer process critical dimension (CD) variability. In the experiment, an 193 nm microstepper with 0.6 NA imaging lens was used to expose the programmed defect mask. The resist CD response to the mask defect area is measured under the different process conditions, i.e., different exposure dose or focus. It is found that similar to a single layer 248 nm DUV resist case that has been studied before, the TSI resist CD responds to the mask defect area linearly for the small mask defects. From such a set of CD-defect response lines, the allowable mask defect requirement is assessed via the statistical explanation of the printable mask defect size which is tied to the wafer process specifications and the actual wafer process CD controllability. Its printability is further verified by the process window measurements. Our results showed that for 193 nm exposure and TSI resist process, a mask defect as small as 0.13 micrometers (4X) is printable at 0.18 micrometers design rule. This printable mask defect size is approximately independent of mask bias in a linear CD region, i.e., a region in which the resist CD responds to the mask CD linearly. Although the above experimental results are obtained based on a TSI process, many characteristics of mask defect behavior can be generalized to the other resist processes. Based on the statistical defect printability analysis method that we have developed, the printable mask defect size can always be re-defined without additional data collection when the process control or the process specification changes.

In this paper, the principle of two typical i-line ((lambda) equals 365 nm) and excimer laser ((lambda) equals 248 nm) uniform illumination optical systems based on Cohler illumination is reviewed, the principle and method of the calculation and simulation of intensity distribution of the illumination system for sub-micron photolithography consisting of aspherical surfaces, non-co-axial surfaces and lens array are put forward, that is ray-tracing based on the Snell theorem and spot diagram. At the same time, a program CALOSD made by us is described, two typical results of calculation and simulation of intensity distribution of illumination optical systems are given.

We describe a method for optical proximity correction (OPC) based on the principle of aerial image matching. Three basic, sub-resolution elements are used: scattering bars, anti-scattering bars, and serifs. We examine the effects of adjusting the sizes and placements of all three elements, and report the improvements achieved, in CD uniformity, printing fidelity, resolution, and depth of focus. We describe the use of a vertex-based `geometry engine' to correct very large microprocessor-style random logic chips of up to 20 million transistors. Our experience producing over 200 full-chip i-line OPC reticles, and the extensibility of the method to deep ultraviolet wavelengths, are presented.

Through the use of phase shift techniques, focus errors have been demonstrated to result in easily measurable overlay shifts in printed resist patterns. Using box-in-box with phase shifter design, patterns are printed on wafers and measured on standard overlay equipment. Results are compared to more conventional methods of focus detection. Details include measurement and calibration methodology, focus, and focus tilt results. Additionally, SPC in IBM's ASTC Fab is demonstrated with the Phase Shift Focus Monitor.

This paper details a method that was used to evaluate and characterize autofocus uniformity and repeatability issues which were exhibited on ASML 5500/100 Stepper systems at Advanced Micro Devices in Austin, Texas. Equally significant is the process by which these issues were resolved through a cooperative effort with ASM Lithography, which is also discussed. Data shown in this paper was collected using the Phase Shift Focus Monitor Reticle, produced by Benchmark Technologies. The coactive relationship between ASML and AMD was integral in identifying the cause of this long-standing condition, as well as providing and implementing the required solution.

The compaction (densification) of fused silica under low fluence (< 1 mJ/cm²/pulse), long term (hundreds of millions of pulses) 193 nm irradiation has been studied. With the use of a finite element analysis, the unconstrained densification, (delta) (rho) /(rho) , is extracted from the experimentally determined wavefront distortion. We find that the densification of silica in the low fluence exposure regime corresponds to what is predicted from the behavior described by: (delta) (rho) /(rho) equals 0.000117 (NI²)^0.53 where N equals number of pulses, I equals intensity (mJ/cm²/pulse). Results are presented of high and low intensity studies that establish the utility of the above form as a predictive tool for densification in fused silica.

`Wavefront Engineering' is the discipline of producing an exposure pattern, adequate for delineating resist at high yield, in spite of the limitations of the imaging technology. Although this discipline relies on century-old optical insights and despite decade-past experiments demonstrating dramatic improvements in resolution and process-window, this field has languished in comparison to traditional approaches, such as exposing with shorter wavelength radiation or larger numerical apertures. Further progress in NA and (lambda) will soon be limited by physical and materials considerations, necessitating other schemes for decreasing the critical dimensions of volume-production devices. Today, with 193 nm systems delayed and non-optical approaches confronting infrastructure and economic barriers, the semiconductor industry is trying to adopt such wavefront engineering techniques as Off-Axis Illumination, Optical Proximity Correction and Phase-Shifting Masks. CAD/CAM methods similar to those applied to optimize lenses and chips now are being applied to optimize the exposure-dose pattern itself.

The evolution of optical lithography is traced back more than 30 years to its beginnings with contact printing. As the complexity of integrated circuits increased, the intolerance for defects drove the industry to projection printing. Projection printing was introduced in the early 1970's by imaging the full wafer at 1:1 magnification. The rapid increase in wafer sizes was accommodated by annular field scanning using 1:1 imaging mirror systems. Decreased linewidths and tighter overlay budgets combined with larger wafers created huge difficulties for the mask maker which weren't relived until the introduction of reduction step- and-repeat printing of small blocks of chips in the late 1970's. Further demands for smaller linewidths and larger chips have driven optical lithography to shorter wavelengths and to scanning the chip in a step-and-scan printing mode. Future advancements in lithography will likely combine novel scanning techniques with further reductions in wavelength.

This paper provides a short history of the development of resist materials. We trace the development of resists from the very beginnings of photography in the early 1800's to today's efforts to develop 193 nm resist materials.