KR20070020410A - Methods for exposing patterns and emulating masks in optical maskless lithography - Google Patents

Abstract

FIELD OF THE INVENTION The present invention relates to optical maskless lithography (OML). In particular, the present invention relates to providing a recognizable relationship between a mask and a phase shift mask technique in an OML.

Optical Maskless Lithography, Phase Shift Masks

Description

METHODS FOR EXPOSING PATTERNS AND EMULATING MASKS IN OPTICAL MASKLESS LITHOGRAPHY}

FIELD OF THE INVENTION The present invention relates to optical maskless lithography (OML). In particular, the present invention relates to providing a recognizable relationship between a mask and a phase shift mask technique in an OML.

For a general background on the type of phase shift mask technology similar to the invention disclosed herein, http://www.reed-electronics.com/semiconductor/index.asp?layout=articlePrint&articleID=CA319210 dated March 12, 2004. Reference is made to an article entitled "Application of Advanced Phase-Shift Masks" by Wilhelm Maurer, which is accessible from.

Moore's law ensures that computer power increases exponentially with lower prices. This dynamic growth in processing power would have led to the idea that semiconductor device manufacturing could be an adventurous business, such as a wild cat in oil. The exact opposite is true. Fabrication device manufacturing is a conservative business because fabrication batches are very expensive and the fabrication process is sensitive to very small mistakes. Qualification cycles for new equipment and retrofitting standards and old equipment are time consuming and demanding. Even small changes are widely detected until they are released to the product.

The co-assigned application has described an SLM-based system, many of which have overlapping inventions and are suitably configured to fabricate masks. Additional work was done to adapt SLM technology to the chip's direct write. However, the acceptance cycle is the most difficult.

Opportunities arise for introducing SLM-based systems that emulate patterns generated with conventional reticles, including reticles with phase shift patterns and OPC features. Creating a direct pattern from the SLM with a small and understandable difference in the pattern generated by the mask has the potential to improve user confidence and speed acceptance of the new system. It is also possible to provide a direct record of compacting large batches by conventional reticles on the direct path from small batch prototypes. More desirably, it may provide components and systems that are more easily configured and controlled and are more flexible and transparent.

Summary of the Invention

FIELD OF THE INVENTION The present invention relates to optical maskless lithography (OML). In particular, the present invention relates to providing OML with a recognizable relationship to mask and phase shift mask techniques. Specific aspects of the invention are described in the claims, the description and the drawings.

Brief description of the drawings

1 is a schematic diagram of an optical maskless lithography (OML) image generation system.

2 shows that the wafer is scanned at a constant rate with a short pulse length used in the OML in microsteps per stamp.

3 is a schematic diagram of an OML system architecture.

4 shows the result of the calibration measurement of the mirror.

5 shows a preliminary optical design of the projection optical device.

6 shows an example of a mirror tilt arrangement.

7 shows the resulting general process window of resist for 60 nm lines at 130, 200, 400, 600 and 1200 nm pitch.

8 is two aerial images of 60 nm contact holes.

9 illustrates one-dimensional digital filtering of a bitmap.

10 shows a scanner in a cascade and similar functionality of a maskless tool and a mask writer.

11A and 11B show on-machine simulation of corners from a reticle.

12A and 12B show processing simulations of double dipole decomposition.

The three graphs of FIG. 13 show ED windows using a 6% attenuated PSM reticle, on grid SLM image and off grid SLM image.

14A and 14B show near-field wavefronts from the transmission reticle and the micro mirror SLM.

Figure 15 illustrates one way of looking at the function of the diffractive SLM image as 2D modulation and filtering.

16 shows a cross row layout used in some SLMs.

17A-17D show the phase modulation mirror type and trajectory in the complex plane.

18A-18D illustrate various data paths.

19 shows a schematic diagram of a method of printing a 35 nm transistor gate through a phase edge and trim mask using SLMs and reticles.

20 shows the printing of chromeless phase lithography semi-isolated lines with 65 nm and 45 nm line widths.

21A-21D graphically show a lookup table (LUT) calculated to implement digital filtering (Figs. 22-34 are not used).

22A and 22B show an embodiment of an off grid correction filter in accordance with the present invention.

Figure 23 shows the resulting LUT function for gray pixels and dark pixels.

24A and 24B show the improvement calculated due to the novel off grid filter.

27A and 27B show another embodiment of an off grid correction filter in accordance with the present invention.

28A shows an SLM with an amplitude transmissive modulated pixel.

28B shows the ideal pattern from the binary mask.

29 shows a lookup table for pixels on the edge and pixels outside the edge.

30 shows a performance comparison of an illumination table and a grid filter (off grid filter).

31A shows the uncompensated edge of the pattern.

31B shows the compensated edges of the pattern.

32A shows an SLM with amplitude transmission modulated pixels.

32B shows the ideal pattern from the binary mask.

details

DETAILED DESCRIPTION The following detailed description refers to the drawings. Preferred embodiments are described to illustrate the invention, but are not intended to limit the scope thereof, which scope is defined by the claims. Those skilled in the art will recognize various equivalent changes in the following description.

Introduction

For low-volume execution, optical maskless lithography (OML) provides an attractive alternative for mask based lithography due to ever-increasing reticle costs. Foundries and ASIC fabs know that reticles are a significant part of the increase in manufacturing costs, especially in the production of small series. Optical maskless lithography provides a cost-effective alternative while maintaining a process compatible with existing fab technologies.

Optical maskless scanners with wavelengths of 193 nm and 0.93 NA can be achieved for resolutions compatible with 65 nm nodes. A throughput of 5 wph (300 nm) is preferred.

Spatial light modulators (SLM) and data path technology developed by Micronic for the SIGMA line of mask-writers provide computer-controlled reticles with imaging and optical properties similar to conventional reticles. The proposed optical maskless scanner combines a series of multiple SLMs with ASML TWINSCAN platform and uses 193 nm technology to ensure optical process transparency in the fab. The reticle stage and infrastructure are replaced by an image generation subsystem consisting of a set of SLMs and a data delivery system capable of delivering nearly 250 GPixels / sec. The newly designed optical train has a maximum NA of 0.93 and is compatible with ASML's TWINSCAN series of conventional lithography scanners, including support for all illumination modes available in conventional scanners.

Maskless lithography approaches require high data volumes. Unlike electron beams, optical maskless lithography has no inherent physical throughput limitations. SLM pattern generation techniques are well suited for throughput scaling. The conversion path from the input file to the image pattern of the resist through rasterizer and SLM can be done in parallel by using multiple SLMs simultaneously. For the random pattern the problem is not trivial, but the nature of the repeated scanner field on the wafer simplifies the problem.

The great commonality of imaging techniques between optical maskless scanners and conventional scanners is that they are expected to exhibit the same level of imaging performance in both their types and systems. The image generation process employs existing enhancement techniques (eg, OPC) from mask-based lithography to facilitate the transition from maskless to mask-based mass production as manufacturing ramps up. The table below shows a preliminary system specification for one embodiment of an OML tool.

     Parameter specification

PO interface

     PO aperture number 0.7 to 0.93

     PO magnification 267 *

     Usable Depth of Field (uDOF) ± 0.1 μm

     Pixel size 30 nm at wafer side

Throughput

     300 mm wafers; 125 exposures, 16 × 32mm, 30mJ / ㎠ dose 5 wph

     200 mm wafers; 58 exposure, 16 × 32mm, 30mJ / ㎠ dose 10 wph

As a reticle  Micro mirror  use

Optical maskless lithography seeks to combine conventional (ie mask-based) photolithography scanners with a fixed array of multiple micromechanical SLMs used to generate mask patterns in real time instead of reticles.

1 provides a schematic diagram of an optical maskless image generation system. Aspects of the SLM pattern generator are disclosed in the references cited above. The workpiece to be exposed is processed at stage 112. The position of the stage is controlled by an accurate positioning device, such as a pair of interferometers 113.

The workpiece may be a mask with one layer of resist or another exposure sensing material or may be an integrated circuit with one layer of resist or another exposure sensing material for direct writing. In the first direction, the stage moves continuously. In another direction, typically in a direction perpendicular to the first direction, the stage moves slowly or stepwise so that several rows of stamps are exposed on the workpiece. In this embodiment, a flash command 108 is received at a pulsed excimer laser source 107 that generates a laser pulse. This laser pulse may be in the deep ultraviolet (DUV) or extreme ultraviolet (EUV) spectral range. This laser pulse is converted into illumination light 106 by a beam conditioner or homogenizer.

Beam splitter 105 sends at least some illumination light to SLM 104. The pulse is short, such as only 20 ns in length, so that any stage movement is frozen during flash. SLM 104 is responsive to data stream 101, and data stream 101 is processed by pattern rasterizer 102. In one configuration, the SLM has 2048 × 512 mirrors, each 16 μm × 16 μm, and has a projected image of 80 nm × 80 nm. In another configuration, the SLM has a mirror of 8 μm × 8 μm with a much smaller projected image. It includes CMOS analog memory with micromechanical mirrors formed at 0.5 microns on top of each storage node.

The electrostatic force between the storage node and the mirror drives the mirror. The device operates in diffraction mode, not specular reflection, and deflects the mirror by only a quarter of the wavelength (62 nm at 248 nm or 48 nm at 193 nm) and is fully off from fully on. Going to The mirror drives on, off and 63 intermediate values to create a fine address grid. Patterns are stitched together from millions of images of SLM chips. Flashing and stitching proceeds at a rate of 1000 to 4000 stamps per second. To reduce stitching and other errors, patterns are written 2-4 times with offset grids and fields. In addition, the fields are mixed along the edges.

The mirrors are individually calibrated. A CCD camera sensitive to excimer light is placed in the optical path at a position equivalent to the image under the last lens. The SLM mirror is driven through a known continuous voltage and the response is measured by the camera. A calibration function is determined for each mirror and used for real-time correction of gray-scale data during writing. Rasterized into a gray scale image, where the gray level corresponds to the dose level of each pixel in the four lighting passes in a vector format pattern. This image can then be processed using image processing. The final step is to convert the voltage to the SLM into an image to drive. Image processing functions are performed in real time using programmable logic. Through various steps disclosed in the related patent application, the rasterizer pattern data is converted into the value 103 used to drive the SLM 104.

In this configuration, the SLM is a diffraction mode micromirror device. Various micrometer devices have been disclosed in the art. In another configuration, the illumination light can be delivered through a micro shutter device such as an LCD array or a microchemical shutter.

OML uses a series of SLMs based on an extension of the 1 MPixel SLM technology used in Micronic's SIGMA mask writer. The SLM is irradiated by an excimer laser source pulsed through an optical system in front of the SLM, which projects a reduced image of the SLM onto the wafer. In the OML tool, each SLM pixel is an 8 μm × 8 μm tilt mirror. When all the mirrors are flat (ie relaxed), the SLM surface acts as a mirror and reflects all the light specularly through the projection optics. This corresponds to a specific area of the corresponding reticle. When the mirror is fully tilted, the surface is not flat and light is lost by diffraction out of the stop of the projection optics; Thus, dark areas are created on the wafer. The intermediate tilt position will reflect some of the light to the projection optics, ie a gray region is created.

The SLM chip consists of CMOS circuits that are functionally similar to the CMOS circuits of reflective LCD devices and the circuits of computer TFT screens. Pixel cells include storage capacitors and transistors to allow storage nodes to be charged and isolated from analog voltages. The pixels are addressed in turn during the loading of a new frame by regular matrix addressing, ie by scanning all columns and rows and applying an analog voltage to each. The area is divided into a large number of load zones that are scanned simultaneously, so that the entire chip is reloaded within 250 msec.

In the pixel cell, the storage node is connected to the bottom electrode of the mirror. The electrostatic force pulls the mirror and tilts it. The exact angle is determined by the balance between the analog voltage and the stiffness of the flex hinge, ie the device has an analog action and the loaded voltage can control the tilt in a very small increment. Actual resolution is limited by the DAC providing the drive voltage.

Intuitively, the tilting mirror appears to produce an image image on the wafer. Image images are known to produce artifacts when scanned through a focus range. In this case, however, the small size mirror provides a high spatial frequency for the image information. Thus, practically all the image information is removed by the finite aperture 110 of the projection lenses 109 to 111. (Finite aperture may also be referred to as a Fourier stop). The result is an image on the wafer surface that is completely amplitude modulated and therefore acts in the same way as the image from the reticle. In particular, since the rows of mirrors on the SLM are tilted across the direction, there is no telecentric (ie transverse movement of the line through the focus) effect.

In modern bitmap based mask writers, the grid generated by the pixels is subdivided by gray scaling. While not necessarily intuitive, the fact that the diffractive micromirror can be driven to produce similar virtual grid functions has been demonstrated by numerous simulations and in practice by the SIGMA mask writer. The rasterizer outputs 64 levels of pixel values depending on the area of the pixel covered by the feature to be printed, and the pixel values are converted to mirror tilt angles. The resulting virtual address grid is 30/64 nm in a single pass. In two passes the grid can be further subdivided into 30/128 nm = 0.23 nm. This is small enough to actually make the system "gridless". Any input grid set to 1.0, 1.25, 0.5 or 0.25 rounds closest to 0.23 nm. The maximum round off error is 0.12 nm and the round off errors are evenly distributed. The resulting distribution of critical dimension (CD) uniformity is 0.28 nm (3σ) negligible. Also, grid snapping or aliasing effects cannot be observed.

SLM-based image generation systems replace reticle stages and reticle handlers with related metrology, electronics, and software. The pattern is printed onto the wafer by synchronizing the loading of the image data into the mirror array with the firing of the laser pulses and positioning the wafer stage. Naturally, the mirror array forms a fixed projection grid. Gray scaling is used to control line width and line position in increments of sub nanometers. This can be accomplished by placing the pixel in an intermediate state between "off" and "on" so that only a portion of the light is transmitted. In order to obtain good pattern fidelity and position, the size of the pixel projected onto the wafer should be approximately half of the minimum CD. With 8 μm × 8 μm pixels, the projection system reduces pixels by a factor of 200 to 300 times. The ultimate stamp size is limited by the maximum size of the lens element close to the SLM.

To achieve high throughput, the OML tool delivers a full dose (ie energy per unit area) at only 2 pulses per stamp, compared to 30 to 50 pulses of a conventional lithography scanner. Because of the small field size, the actual laser power is quite low. The data path can achieve partial compensation for pulse-to-pulse, but lasers with very good pulse-to-pulse energy stability help to meet dose control requirements.

While the wafer is being scanned at a constant speed, short pulse lengths are used in the OML, making microstepping per stamp more similar to the system as shown in FIG. Stitching quality is therefore a significant performance issue, and interlayer overlay and intralayer alignment are extremely important. In the figure, the pattern data of die 205 is divided into stripes 210. The strip can be printed by a series of SLMs. The stripe can be divided into microstripes 220 corresponding to printing by the SLMs 232 in the array. SLMs in array 230 are loaded with data. Loading of the SLM 232 to manufacture the micro shots 242, 246, 248 and the microstripes 220 begins with the ideal pattern data 242. Calibration, correction and overlap adjustments are applied 243 and producing data 244 is sent to the SLM. The wafer is printed by controlling the order 250 of stamps and stripes across all SLMs in the array.

OML  Subsystem Overview

Design decisions for OML correlate with throughput and CD uniformity. Throughput is determined by pixel size, number of pixels in one flash and SLM frame rate, while resolution is primarily influenced by pixel size and optical design. Binary parameters include the number of pixels per SLM, stage speed, data flow, and the like.

Integrating an optical maskless scanner on an existing ASML TWINSCAN platform means retrofitting some subsystems. Most notably, the reticle stage (including interferometer) and the reticle handler are removed from the system. These reticle modules support the use of all necessary data path drive electronics and SLMs, replacing them with multiple SLM array (MSA) modules consisting of multiple SLMs in predefined patterns with pattern processing software needed to dynamically form the required mask patterns. do. In addition, lasers, illumination systems and projection optics are specifically designed to meet the unique optical requirements of the OML.

Thus, changes in form and function of the main system will typically affect other subsystems, even though they are small. For example, because the exposure of the resist is performed only in two laser shots, the dose control must be changed, and the synchronization must be changed to adjust the activity of the SLM instead of the reticle stage.

3 schematically shows the degree of change between a major module into a system architecture and a system and a conventional ASML TWINSCAN, and distinguishes items that are unique to OML tools as well as items that require functional and / or structural changes. Many parts of the architecture can be reused with image generation systems and optical paths. The image generation 310 system is adapted from the SIGMA results. Since the SIGMA results are used as a single SLM, the multi-SLM array is entirely new. Functional and / or structural changes to the SIGMA results are shown for the residual subsystem of image generation.

Image generation subsystem

The image generation subsystem defines the core functionality of the optical maskless scanner and consists of the SLM unit, drive electronics and data path. Structurally, it is very similar to the SIGMA mask writer's image generation subsystem, although it is expanded to accommodate even higher throughput as well as specific improvements to the resulting image fit and overlay. The SLM is a VLSI MOEM array of reflective tilting mirrors, each of which can be combined to induce a phase change by modulating the reflected intensity and producing a geometric 2D pattern such as its circuit or portion. Since each mirror is several microns in size, it is essential to use a strong reduction projector to reduce the size of the pixels on the wafer to print important features. Specifications for one embodiment of SLM and drive electronics are provided in the table below.

parameter  Specification

     Mirror size 8 μm × 8 μm

     Array size 2048 × 5120

     Frame rate ≥4 kHz

     Drive voltage <10V

     Number of analog levels (calibration) 64

Ideally one could pack the entire target side of a PO with a single large array of mirrors, but such devices go beyond current MEMS technology. Thus, there is a need to use a series of multiple SLMs in parallel to provide the required number of pixels to achieve the desired throughput. Pixels from different SLMs of a multi-SLM array (MSA) are stitched together to form a tight image on the wafer surface using a combination of motion control and gray scaling techniques. The wafer stage continues to move and stitch together separate SLM images while printing with a set of overlapping pixels along the edges between the SLMs. The layout is configured to complete the movement of the pattern with two overlapping laser pulses. Replacing the stamp and pixel grids between the pulses averages the residual grid and SLM artifacts, thereby reducing any appearance of the grid and SLM chip structure.

The requirement for mirror-to-mirror uniformity is higher than can only be achieved by strict manufacturing tolerances. Minor differences in each mirror result from changes in film thickness, changes in critical dimensions of the curved hinges, and the like. The response of each pixel at the displacement angle to the induced voltage must be calibrated and corrected with the calibration map applied to the bitmap data based on a shot-by-shot basis. Gray scaling for stitching as well as compensation for any bad pixels are embodied in this map. OML tools calibrate SLM in-situ to accommodate long-term drift of SLM pixels. Due to the fact that large volumes of pixels and projection images of pixels are sub-resolution, calibration is accomplished by observing groups of pixels and allowing the groups to provide uniform intensity at varying intensity levels. 4 shows the results of calibration of the mirror in the SLM of SIGMA 7100. It is the processing image in flat gray of 8x8 array (64 pixels) of SLM before and after calibration. The leveling effect of the calibration is obvious.

Data path

Together with analog-driven electronics, the data path transfers data to the MSA at an expected data rate of about 250 GPixels / sec. The steps for converting the pattern data into an SLM image to be printed are as follows:

! Pattern entry: At the start of the run, the user will upload a mask file (eg GDSII or OASIS) to an optical maskless scanner that contains all the patterns for the die to be printed. The rasterizer is OPC correction of the input data stream, optimized to produce optical images from the SLM as close as possible to the actual reticle. Sub-resolution OPC features are represented accurately by the SLM, and the image created on the wafer is almost identical to the image from the reticle. Alternatively, OPC correction may be introduced to the data stream in real time.

! Fracturing: Prior to execution, the pattern data is divided into fragments corresponding to the multi SLM array layout and continued through writing and stitching methods for reproducing the pattern on the wafer. This data is split to create small overlapping borders on each side and to cause the split image to be stitched during exposure.

! Rasterization : During execution, the proper image segmentation for each SLM is converted to a bitmap of pixel values representing the image. The rasterization step also includes the application of corrections and individual mirror calibrations to ensure the proper processing of the ideal image on the pixel grid and the proper image fit on the physical device while maintaining the proper feature size and position.

! Data write: The rasterized pattern for each SLM is delivered to the SLM simultaneously with the laser and wafer stages so that the pattern is established on the SLM during the laser flash of the appropriate pulse.

If extremely high data flow rates and complex patterns are reproduced, data integrity is an extremely important aspect of the data path. During software development, regression tests can be used to compare against the output of previous versions.

A second aspect of data integrity is to avoid bit errors in storage and transmission of large data volumes and large data volumes. This is done by standard methods, and since most of the data paths operate in asynchronous mode, errors are detected before the error does any damage. In most cases, the correct data can be retransmitted or regenerated. The system can be configured to flag all errors and to specify actions to be taken on specific kinds of errors (eg, interruption of work, die interruption, die autocalibration or potentially broken die marks in log files).

Finally, high capacity of the data path is achieved through the use of high level parallel electronic architectures. The disadvantage of parallel systems is the statistically high risk of module failure. Therefore, particular attention should be paid to module diagnostics to detect any hardware problems early. With this principle and prevention, data paths do not contribute much to yield loss.

light

The illumination system for the direct recording of the scanner (320 in FIG. 3) is very different from that for the scanner and varies considerably from the illumination system used in SIGMA. Since only a small portion of the total optical field has active pixels, the illumination system should be designed to only illuminate the active pixels of the target field. Modifications to 2-pulse printing affect the laser requirements for OML. The power requirement is about one tenth of that of a conventional scanner, and the first reason is the large decrease in field size and the relatively low throughput. The repetition rate of the laser matches the refresh rate of the SLM. 4kHz laser can be used. Pulse-to-pulse stability of 1% 3σ is useful, which is about 10 times better than conventional lithography lasers using average pulses of 30-50 pulses for dose uniformity. Alternatively, additional pulses can be used to deliver a larger average dose and can be set to the correct dose error from the previous pass. These alternatives can improve dose control, which reduces throughput.

Laser pulse timing errors (eg, jitter) can impact overlay performance. In conventional scanners, the wafer and reticle stages are run in synchronization, and laser timing and pulse length do not have a significant effect on the pattern position. In optical maskless lithography, the SLM array is "still" during exposure, that is, the image scans at the speed of the wafer stage. For a wafer stage speed of 300 mm / sec order, 30 nsec laser timing jitter produces a 9 nm position error, which is not acceptable for some applications. This damaging effect is constant for a certain wafer stage speed, and so there is no need to worry about the overlay, but the duration of the pulse causes damage to the image. In addition, the effect of damage from the relatively short pulse duration of X / Y asymmetry is easily corrected in the data path.

The table below summarizes the preferred laser characteristics.

 Parameter specification

     Wavelength 193.368 nm

     Bandwidth 10 pm

     Static Range 193.33 to 193.45 nm

     Rep rate (max) ≥4 kHz

     Power ≥5W

     Pulse energy ≤10 mJ

     Pulse length ≤20 ns

     Pulse Energy Stability <1% 3σ

     Pulse jitter <5 nsec

Dose measurements use sensors in the illumination system to track the intensity of each pulse. Since dropped pulses or large pulse-to-pulse stability can have a significant impact on tool performance, power tracking with such detectors is useful in OML scanners that average for several pulses. Dropped pulses can be easily detected by combining the detector with synchronization so that each sink pulse has a corresponding energy reading and the tool software easily identifies the effective detector reading for each pulse. The 193 nm illumination energy detector used in ASML scanners tracks energy per pulse. These detectors are calibrated between the wafers with energy detectors on the wafer stage, which in turn is referenced periodically to a global standard with a removable master detector.

The illumination optical design concept is based on a multi-array design that provides pupil and field confinement with multiple capacitors to provide illumination homogeneity. This concept allows the PML to generate the same illumination profile and sigma settings as conventional scanners. Advantages of a multi-SLM array design include the following.

! Field Defining-This design takes into account field-defining elements (FDEs), so that only the active mirror portion of the SLM in the multi-SLM array is illuminated. Since only a small portion of the optical field region for a multi-SLM array contains active pixels, this is necessary to improve the stray light characteristics of the system and to consider low power.

! Pupil Polarization Support- First, for scalability to future lithography generations, multi-SLM array designs consider pupil polarization to enhance certain features in ultra high-NA systems.

Projection optics

Among the projection optics 320 subsystems, Calibration Optics & Metrology are very different from the subsystems used in SIGMA. The catadioptric design form with beamsplitting cube 526 has been identified as a useful design for OML because of its optical scalability to 65 nm nodes as well as potential scalability to next generation requirements. This design reduces the amount of glass used and does not require a significant amount of CaF ₂ . A preliminary optical design for the projection optics is shown in FIG. 5. Illumination system 520, multi-SLM array 512, projection optics 530, and wafer stage 540 are shown.

Multi- SLM  Array

Mechanical mounting and electrical and optical packaging of each SLM are part of the design of the multi-SLM array. Since accurate control of the space between the active portions of the SLM is necessary to achieve proper stitching between the images of the individual SLMs, the packaging must be designed to accommodate the desired SLM layout.

The extension of SLM technology to print directly on the wafer presents a unique challenge. System details on throughput require up to 60 MPixels per laser flash to be printed, along with the requirements to provide two-pulse printing. In 4kHz operation, assuming that each SLM consists of a series of 2048 x 5120 active mirrors, six SLMs are required for the target surface of the projection optics. The limitation on the maximum feasible lens diameter before the SLM is that the packaging and spacing requirements to ensure proper stitching of the individual SLM images during printing affect the layout of the SLM of the optical field.

The construction of multiple SLMs to meet optical, packaging and service issues presents optical, electrical and mechanical tradeoffs. The electrical design also supports data rates in excess of 250 GPixels / sec to write data at 4 kHz refresh rate for each SLM. Since current SLM designs do not include on-board digital / analog conversion, each SLM is driven by an analog signal. Thus, each SLM requires up to 1000 DACs and up to 2000 coax electrical wires to drive the amplifiers and amplifiers next to the chip. In addition to the required data rate, the full volume of the connection increases heat dissipation and reliability issues.

Imaging  Performance simulation

Complex simulation packages such as KLA Tencor's Prolith 1.7, Solid-C v.6.2 from Sigma-C and LithoCruiser from ASML MaskTools are commercially available to perform performance simulations for conventional lithography. These tools are not currently integrated with the rasterization module or the ability to handle the SLM imaging characteristics of OML. For the analysis of imaging performance, commercial simulations were run from the shell of Matlab, which provided a missing function. It is desirable to develop a more user friendly simulation infrastructure.

OML imaging performance simulators with core on-engine analysis engines are promising. In combination, the custom Matlab script rasterizes any 2-D pattern (e.g., lines, contacts, SRAM cells, etc.) into an array of pixel tilts and lists the rasterized images through the pixel grid and through the pulses. Each grid / pulse sequence analyzed by the core Prolith engine is provided for the calculations necessary to do so. The tilted mirror is typically divided into 10 or more regions, each of which is flat and has a phase that is averaged with respect to the corresponding region of the tilting mirror. It has been found that more than seven regions provide a good approximation of flat mirrors with linearly varying phases. For this task the mirrors were ideally assumed to be flat, with uniform 100% reflectivity, no reflections from the slits between them, and to operate exactly with the deflection determined by the data path. Each grid / pulse sequence is recombined and analyzed to predict the final performance of the pattern under given lighting and PO conditions. The simulation was intended to be realistic and was performed with high-NA, vector unpolarized light. Most of the results are based on machining simulations. Where the developed resist profile is shown, the resist model used is the best estimate model for TOK6063, which was used for other work at the ASML Technical Development Center in Tempe. The optics are assumed to be ideal and there are no aberrations and errors in setting the lighting device. Preliminary results from the OML imaging performance simulator show good correlation in critical dimensions, contrast and NILS when compared to similar 6% attenuated phase shift masks (Att-PSM).

6 shows an example of a mirror tilt configuration to produce a 60 nm CD line pattern 610 with scatter bars OPCs 622, 624. The applied algorithm converts line 610 and scatter bars 622, 624 to mirror tilt settings. The figure shows two rows 632, 634 of the mirror 636. Gray shades show local phase changes as a result of mirror tilt.

The mirror tilt depends on the position of the feature with respect to the SLM pixel grid. In the second pass, the feature position with respect to the grid is changed. The sum of the two passes is more symmetrical than is apparent from FIG. 7. The figure shows a typical process window resulting in resistance to 60 nm lines at 130, 200, 400, 600 and 1200 nm pitch. The exposure tolerance (EL) depends on the depth of focus (DOF) with a 9.1% exposure tolerance at the best focus. The 8% exposure tolerance corresponds to 0.085 micron DOF.

8 shows two machined phases of 60 nm contact holes of 130 nm pitch at two different positions for the projected SLM pixel grid. The photo above shows zero features at the first grid position and the photo below shows a 20 nm shift with respect to the mirror grid. Images were generated with Diagonal Quasar (0.97 / 08,15 ⁰ ) 0.93 NA. As can be concluded from these images, the pixel grid effect can be reduced by applying the digital filter and rasterizing algorithm described below.

Rasterizing  Algorithm evolution

Rasterization of non-adjacent imaging systems is simple in principle: overlay a desired pattern with a pixel grid and assign each pixel a gray value, which is the fraction of pixels covered with a feature (assuming it is an exposed feature). This is called an area bitmap because every pixel value represents an area. This rasterization is also useful in laser scanning pattern generators (PGs) and particle beams PG. If the pixel is not small compared to the diffraction limit spot of the optical system, nonlinear correction may need to be applied. The nonlinear function may be corrected for nonlinearity in a modulator such as a sonic optical modulator.

However, this approach does not print correctly for partially coherent light. The area bitmap must be transformed from the array of modulators to the intensity bitmap of the desired intensity by a nonlinear function. The nonlinear function can be calculated from the first principle or can be measured in a dedicated experiment. Nonlinear functions are called lighting tables. In rasterization using an illumination table, the SLM system print corrects the CD of the pattern to approximately k1 = 0.5. This works for a mask writer, but for maskless tool printing lines up to a k1 value of about 0.2, the illumination table approach is not appropriate.

Close analysis shows that the image log slope depends on the position of the edge relative to the grid, even if the line width is corrected by the illumination table method. At the edge located at the grid position, the SLM prints very close to the ideal amplitude mask because one pixel is completely on and the other pixel on the edge is completely off. However, if an edge is displayed with half the pixels, then the edge will be present between the grid positions, and there will be an intermediate pixel between clear and dark with the median. This functions as a low pass filter. The result is that the image is lowpass filtered at off grid locations and not at on grids. The effect of change across the grid is compensated by double or quad pass printing with a typical loss of edge sharpness of about 10%. This may be acceptable for a mask writer writing four passes, but may cause undesirable results for maskless lithography or only two passes. In wafer lithography, features are printed close to the resolution limit and CD linearity may be compromised. The feature is almost lost and reinstated by the high contrast of the resist process. In this imaging scheme, the image log slope that changes with the grid causes unwanted CD changes. The CD to grid can be calibrated for one pitch but will fail for other pitches or feature type. More sophisticated rasterization is needed.

9 illustrates one-dimensional digital filtering of a bitmap. The nearest row 910 is a row rasterized bitmap. After row 910, 920 is shown a filtered bitmap to enhance the off grid edge. The second from the back 930 is a filter that enhances all edges, and the row 940 that follows is a combination of rows 920 and 930: a filter that removes the grid and simultaneously enhances all edges. Negative values indicated by black shadows (eg 943) do not exist in normal image processing. Here they operate with mirror tilts that produce negative complex amplitudes.

The solution developed for change through the grid is a grid filter, a digital filter operating on an area bitmap. Digital filtering can do a lot, but its function is, among other things, to remove the visibility of the grid in the image. To do so, kernel derivation is associated with the area bitmap, but only with neighbors of pixels having an intermediate pixel value, a "gray" pixel (eg, 912). It should be described as a convolution in a close sense. The nearest neighbor on the dark side is darkened (923), and the bright one on the bright side is brightened (921). How dark and bright depends not only on the intermediate pixel values, but also on the lighting mode. Many parameters are calculated for the actual optical settings. These parameters are extensive and easily calculated. However, instead of using the nearest neighbor, you can extend the filter to the second neighbor or use a much larger kernel. Tuning the algorithm changes the tradeoff between pixel size and image quality. Features below two pixels can be printed with good fit. Resolution is limited by the optical system. When the parameter is tuned, the edge sensitivity of the edge of the on grid or off grid is the same. The algorithm seems surprisingly insensitive to the type of feature. One setting seems to work almost correctly for most features, and can be explained by the fact that the filter only makes small corrections to the pattern.

21A-21F show graphs of lookup tables (LUTs) calculated to perform digital filtering. LUTs are calculated for a particular optical setting (wavelength, fixture, aperture number), pixel characteristics (size and amount of "negative black" used), and the number of neighbors included in the filter. They are individual for each pixel and represent the change in gray level at each pixel as a function of the gray level of the intermediate pixel. It is possible to combine the bright pixel LUT and the dark pixel LUT into one single LUT whose entry is not the gray pixel value, but the difference in gray value between the pixel to be compensated and the edge pixel. For example, the LUT has a wavelength of 193 nm, aperture numerical value of 0.93, fixture sigma of 0.6, 0.8 or 0.99, pixel size of 30 nm and 0, -6% (corresponding to -.245 amplitude) or -1 (180 ° phase) May be calculated for the reflectivity at the dark of the phase shifted region). LUT 21a was used in the calculations of FIGS. 11 and 12. LUT 21b was used in the calculation of FIG. LUT 21c was used in the calculation of FIG. The LUTs 21d and 21f represent cases where gray pixels, dark pixels or bright pixels remain constant and neighboring pixels are adjusted.

The LUT is calculated by comparing the Fourier transforms of the one-dimensional edges represented by the white, gray and black SLM mirrors and the same edges represented by the ideal reticle / mask. The difference in the Fourier transform is minimized by varying the gray values of some pixels near the edges. The number of pixels involved in the minimization will affect the shape of the LUT curve. The difference in the Fourier transform is minimized to NA / λ * (1 + σ) for all spatial frequencies, and σ is the partial coherence factor in the illuminator. This is repeated as the edges step over one pixel and begin to align with respect to the pixel grid and then pass the entire pixel until it is aligned with the grid again. The edge position corresponds to the area coverage value (0 to 1), which is an edge pixel gray value. For nonzero transmission in the dark region, for example 6% phase shifting, the gray value used as the entry for the LUT is simply scaled linearly in the amplitude reflectance range (between -sqrt (0.06) and 1). Area coverage (0 to 1). The adjustment range of all pixels increases with increasing sigma, increasing NA, increasing pixel size, increasing transmittance in the shifter / dark area, and decreasing wavelength.

22A and 22B show an embodiment of an off grid correction filter. This off grid filter operates during rasterization on the area bitmap and detects and increases gray pixels and lowers dark pixel neighbors to negative black. The pixel value is changed into two lookup tables, one for gray pixels and one for dark pixels, and is pre-calculated prior to exposure. FIG. 22A shows an uncompensated edge comprising dark pixels P2, gray pixels P1 and light pixels from left. The uncompensated gray value of pixel P1 determines the compensated gray values P1 * and P2 * according to P1 * = LUT1 (P1) and P2 * = LUT2 (P2), where LUT1 and LUT2 are two different lookup tables. After compensation, in FIG. 22B, the gray level of the compensated gray pixel P1 * increased and the gray level of the compensated dark pixel P2 * fell below gray level zero.

In this embodiment, the LUT is calculated with infinite edges moving for one pixel in n steps using, for example, the equal MATLAB linspace function. For each nominal edge position (corresponding to area coverage), the position at the reference level and the image log slope are compared with when the edge is on grid. The reference level is determined for the pattern of the on grid. LUT is calculated repeatedly. The initial value for the LUT is:

LUT1 (1: n, 1) = linspace (0,1, n)

LUT1 (1: n, 2) = linspace (0,1, n)

LUT2 (1: n, 1) = linspace (0,1, n)

LUT2 (1: n, 2) = a * x ^ 2-a * x, x = linspace (0,1, n),

Where a = 0.217 * 4, i.e. the largest negative black * 4 or else

The LUT is applied to pixels P1 and P2 according to the following.

P1 * = LUT1 (P1,2)

P2 * = LUT2 (P1,2)

The machining phase is then calculated. At each n steps the correction condition is calculated as follows for the position and the ILS.

Corr_pos = nominal_position / real_position

Corr_ILS = ILS_reference / ILS_real

Depending on the location or whether the ILS is optimized, either LUT1 or LUT2 is updated.

LUT1_new (P1,2) = LUT1 (P1,2) * Corr_pos

LUT2_new (P1,2) = LUT2 (P1,2) * Corr_ILS

Once one convergence criterion is achieved, the LUT is repeated from applying the pixels P1 and P2 and optimized for the rest until both conditions are achieved.

23 shows the LUT function of the result. LUT1 for P1 is the best line in the graph. LUT2 for P2 almost reaches the maximum negative black amplitude that can be achieved by the tilting micromirror.

Improved calculations due to this embodiment of the off grid filter are shown in FIGS. 24A and 24B, 25A and 25B, and 26A and 26B. Some of the parameters used to calculate this result include 90 nm high density L / S; Annular illumination 0.7 / 0.9; 2 nm mesh lead: 30 nm pixel size; 13 pupil mesh points; And NA 0.92925925925926. 24A and 24B show placement error versus grid shift. In FIG. 24A, the minimum placement error of zero corresponds to a grid shift of 0, 15 or 30 nm and uses an illumination table LUT. With the off grid correction filter of this embodiment, there is almost no placement error regardless of the grid shift over the range of 0 to 30 nm. 25A and 25B, the contrast obtained between opposite sides of the intended boundary between dark and light is again graphed 25a for grid shift for the illumination table LUT and again for the off grid correction filter of this embodiment. It is shown by the graph 25b. Finally, the normalized image log slope was plotted against the grid shift for illumination table LUT 26a and the off grid correction filter of this embodiment 26b. Those skilled in the art will appreciate that normalized image log slopes are normalized to feature size and tend to be proportional to the exposure width. Changing the parameters to 60 nm high density L / S and 15 pupil mesh points changes the shape of some curves in this figure, but generally confirms the performance of this embodiment of an off grid filter.

The operation of another embodiment is shown in Figs. 27A and 27B. This version of the off grid filter operates directly on the area bitmap and replaces the lighting table LUT. During operation, an edge is detected and the edge pixel and two neighboring pixels are changed. The pixel value changes with three lookup tables, the lookup table for each pixel. The lookup table is precomputed prior to exposure. In Figures 27A and 27B, P1 (gray pixels), P2 (dark pixels), and P3 (light pixels) are graphed for their area bitmap gray level, i.e. area coverage. P1, the uncompensated gray value of pixel 1, determines the compensated gray values P1 *, P2 * and P3 * according to the following.

P1 * = LUT1 (P1)

P2 * = LUT2 (P1)

P3 * = LUT3 (P1)

Where LUT1, LUT2 and LUT3 are three different lookup tables.

The LUT is calculated by substantially minimizing the difference between the Fourier transform (FT) from the SLM and the complete binary or phase shifting mask on top of the projection optics pupil.

Edge offset correction filters that substantially minimize the difference between the Fourier transform from projecting the emission from the SLM's aligned pixels and the complete binary mask or phase shifting mask on the projection optics pupil are one, two, three or more. It may also be performed using pixels.

28A shows the SLM with features having a width w (1 + gl) to the left, where w is pixel width and gl is in range [0,1]. The pixels are modeled with amplitude transmission, which can have negative values. a, b and c are parameters used to minimize the diffraction pattern difference compared to the ideal case. Complementary FIG. 28B shows an ideal pattern from a binary mask. The feature has the same width, w * (1 + gl) as in the SLM case. The real and imaginary parts of the difference in FT, FT_SLM (fx, a, b, c, gl) -FT_ideal (fx, gl) ranges in the range [-NA (1 + σ) / λ, NA (1 + σ) / [lambda]] is the respective value of g minimized for all fx. NA is the numerical aperture of the projection optics, and sigma is the degree of interference of illumination.

FT_SLM = w * sinc (w * fx) * (1 + a + (gl + b) * exp (-i * 2 * π * w * fx) + c * exp (-i * 4 * π * wfx))

FT_ideal = w * sinc (w * fx) + gl * w * sinc (gl * w * fx) * exp (-i * π * w * fx (1 + gl))

F_min = (FT_SLM-FT_ideal) / (w * sinc (w * fx) =

= a +

+ b * exp (-i * 2 * π * w * fx) +

+ c * exp (-i * 4 * π * w * fx) +

+ gl * exp (-i * 2 * π * w * fx)-

-gl * sinc (gl * w * fx) sinc (w * fx) * exp (-i * π * w * fx (1 + gl))

The equation system can be written back in matrix form, A * x = h. The overlapping linear equation system A (fx) * [a, b, c] = h (fx, gl) can be solved by least squares.

In FIG. 29, the calculation of the result is graphically plotted for λ = 193 nm, w = 30 nm, NA = 0.93, σ = 0.96. Bottom line shows LUT2 = c. The middle line shows LUT1 = b.

In FIG. 30, the application of this embodiment of the grid filter was performed in high density lines and spaces with half pitch 60 nm. The result is a smaller CD range, smaller PE range, higher contrast, smaller contrast range, higher NILS and smaller NILS range than the lighting table LUT.

This embodiment of the grid filter can be extended to include a binary mask as well as a phase shifting mask including weak phase shifting and strong phase shifting (chromeless phase lithography (CPL)). 31 and 32 show in the same manner as FIGS. 27 and 28, wherein the ideal pattern from the reference reticle and the SLM both have features with width w * (1 + gl), where w is pixel width, g is in the range [0,1] and gld is equal to gl * (1-d) + d, ie gld is equal to g scaled to the range [d, 1]. In this case, the transmission outside the region of the feature is not zero, but instead the amplitude has a magnitude d, which can range from -1 to any value lower than the transmission in the bright region. Thus, it can be 0 as in a binary mask, -1 to 0 as in a phase shift mask, or -1 as in CPL. The corresponding equations representing the Fourier transform, complete phase shifting reticle, and the difference to be minimized of the SLM are:

FT_SLM = w * sinc (w * fx) * (1 + a-d + (gld + bd) * exp (-i * 2 * π * w * fx) + c * exp (-i * 4 * π * w * fx)) + d * δ (fx)

FT_ideal = (1-d) * w * sinc (w * fx) + (1-d) * gl * w * sinc (gl * w * fx) * exp (-i * π * w * fx (1 + gl )) + d * δ (fx)

F_min = (FT_SLM-FT_ideal) / (w * sinc (w * fx) =

= a +

+ b * exp (-i * 2 * π * w * fx) +

+ c * exp (-i * 4 * π * w * fx) +

+ (gld-d) * exp (-i * 2 * π * w * fx)-

(1-d) * gl * sinc (gl * w * fx) / sinc (w * fx) * exp (-i * π * fx (1 + gl)),

Where δ (fx) is the dirac delta function.

As mentioned above, the above equations apply to binary, weak phase shifting and strong phase shifting (CPL). When the SLM and grid filter are used to emulate the performance of the cross aperture phase shifting mask (AAPSM), the above equation cannot be applied directly. For AAPSM, bright areas in masks with opposite phases must be processed separately, and the resulting pixel values are added together. Areas with zero phase can simply be processed from the binary mask along with the surrounding dark areas, and corresponding settings should be used. Bright areas with 180 ° phase can be treated the same as from binary masks with surrounding dark areas, but simply have negative transmission.

In FIG. 29, the calculation of the results is graphically shown for λ = 193 nm, w = 30 nm, NA = 0.93, σ = 0.96 and d-√ (0.06) =-0.245. The value of d corresponds to a 6% attenuated phase shift mask. The top line shows LUT3 = a. The bottom line shows LUT2 = c. The middle line shows LUT1 = b. The value P1 of the uncompensated edge pixel is in the range [d, 1].

Built-in Reticle

10 shows step by step similar functions of the maskless tool 1030 and the mask writer 1010 and the scanner 1020. Images that may be obvious should be taken as the meaning of the text. From the input side, the tool is a mask writer 1032. From the output tool is a scanner 1036. The input and output interfaces are almost identical to the input and output interfaces of machines that are not built in.

Embedded mask writer 1032 transforms the data to produce embedded reticle 1035, which is an image generated by the SLM, not the SLM surface itself. Most of the OPC characteristics of the system start with the optical projection system and the illumination system. To correct them, analyze images in the range of 1 micron or larger. Some additional OPC effects come from mask 1015, most importantly corner rounding, CD errors are coupled to feature size, pitch and polarity, and to the density effect of the mask process. The effect of the 3D electromagnetic boundary conditions of the actual reticle may achieve printing. Clear areas look small, edges are polarized and the transmission of thin lines is affected by the EMF effect.

SLM produces the same image as the mask, but does not have the error because of the digital image and the large reduction. Electromagnetic 3D effects have shown that they do not enter the reproduction of mirrors larger than 2 microns. The characteristic effects of SLMs and data paths are combined in a finite size of grids and mirrors. They can be corrected by the nearest neighbor operation in the bitmap. In fact, the grid filter removes one of the most prominent system features, the pixel feature of the image, and neutralizes the data into an SLM image.

11A and 11B show simulated machining images of corners from several grid locations (solid lines) with reticles having corner radii (dashed lines) of 0, 40, 80, 160, 200, 240, 320 nm and applied grid filters. Indicates. FIG. 11B is an enlarged view of FIG. 11A. The system output shown in FIG. 11 represents the characteristics of the embedded reticle. The entire system is compared to printing characteristics from reticles having simulated and known characteristics, including rasterization with mirrors. 11A shows a simulated corner on a wafer printed from a reticle and SLM with a known corner radius. 11B is an enlarged view of a corner. The dashed lines consist of reticles with varying corner radii, and the solid lines consist of SLMs with patterns located at different grid positions. SLM prints are shown with corner pullbacks that are within 1 nm of the ideal reticle, while the state-of-the-art physical reticle from a VSB mask writer with a corner radius of 80 nm has the disadvantage of about 1 nm large.

The built-in reticle 1035 of this system 1030 can be seen as an ideal representation of the input data without grid effects and significant resolution loss.

off -Axis lighting

The lighting modes of the scanner may overlap in the maskless tool. Because of the differences in optical layout, with maskless with much smaller etendue, there are differences in how off-axis illumination is implemented, but the same illumination pattern will provide the same image characteristics, axicon, diffraction Produced by an element or other means.

In general, grid filters work best with conventional lighting. With conventional lighting, including non extreme annular schemes, one setting appears to be correct for all features. With dipole illumination, the grid filter is slightly affected by the pitch. The dipole filter works perfectly only for certain features and other features provide CD errors through a grid of less than 1 nm in a single pass. However, this is what dual pass printing is designed to suppress, so there should be no measurable effect on the resist.

OPC  transparency

It is useful to match maskless scanner OPCs with scanners that use regular reticles, but is this possible? The following two examples show that full transparency can be achieved with minimal transparency with a complete reticle opposite of the actual reticle.

12A and 12B show processing simulations of double dipole decomposition using reticle (a) and SLM (b). 12 shows simulated double dipole decomposition with extreme off-axis illumination. Data is naturally decomposed into intersections between overlapping vertical and horizontal sections to compensate for line-end shortening. 12A shows a separate image of the horizontal and vertical components and an image superimposed with a binary reticle. 12B shows the same component exposed using SLM (tilting mirror, single pass, grid filter). These images are indistinguishable. Both would require some OPC added during decomposition, but the same OPC correction would apply in both cases.

The following example shows the exposure width window for a semi-isolated line with a scatter bar. The line is 50 nm, 1.67 pixels and the scatter bars are 20 nm or 0.67 pixels wide. The three graphs of FIG. 13 show an EL window using a 6% attenuated PSM reticle (FIG. 13A), an on-grid SLM image (FIG. 13B) and an off-grid SLM image (FIG. 13C). The EL window is not exactly the same, which may be due to the tuning of the grid filter. As will be further shown, providing a larger window can be easily set. However, in this case, even if the OPC number is smaller than the pixel, the embodiment must ensure that the SLM has the same OPC characteristics as the mask.

Will the result be bad in the resist? As we deal with more complex systems, some error will remain. However, the grid filter eliminates deviations through the grid placement of the log slopes and thus the resist wall angle. Since only the first diffraction order from the reticle / SLM pattern is performed on the resist for the conventional pattern, there is little freedom of image to make a big change once the CD and log slopes are fixed. Therefore, the resist result should match the processing result.

SLM seems to be made to match an ideal reticle that closely matches the conditions of the OPC characteristics. Real reticles are not ideal; The difference observed in the OPC model is due to the physical reticle and not the SLM. It may be possible to detune the maskless tool, but if the difference is considered important, the most practical would be to use two OPC models for the actual reticle and SLM. Compare the differences in the corner pullbacks of FIG. 11 for display. Since it is useful for the image to be close and the difference well understood, successful printing on the maskless tool will give confidence that the device will successfully print to the reticle after the deterministic process of pattern conversion.

Phase Shifting mirror

Micro mechanical mirrors have high and uniform reflectivity. The adjacent fields of the mask and the SLM certainly look different. 14A and 14B show adjacent field wave fronts from the transmission reticle (a) and the micro mirror SLM (b). Incoming waves are not shown. Adjacent field wavefronts from the SLM are optically processed to produce intensity variations in the image plane. How does SLM produce high contrast images that are nearly identical to those from masks?

15 illustrates one method of viewing the functionality of a diffractive SLM image with 2D modulation and filtering. Mask 1512 produces a flower pattern with sidebands around illumination beam 1514. All information about the mask pattern is contained in the phase and amplitude of these sidebands. The opening 1516 cuts the center of the diffraction pattern, ie the low pass filters the sidebands and takes the squared modulus of the complex amplitude of the image plane 1518 and then the image is formed. The system can be analyzed with straightforward fourier optics. The top and bottom rows of FIG. 15 compare the SLM based imaging system with image formation on the mask base.

Diffraction SLM 1522 has a surface structure that produces a sideband 1524 that is far outside of opening 1526. Radio engineers say that SLM provides carrier frequencies, actually some two-dimensional space carriers. If all mirrors operate for dark, it means that the zero order illumination beam is extinguished. There is no light passing through the aperture, and the image on the wafer is dark. When the surface structure is modulated by the pattern, the 0-order reappears, but is surrounded by sidebands with information about the pattern. Opening 1526 cuts sideband 1524 and image 1528 is formed. This image 1528 is identical to the image from the mask 1518 because the side bands are the same. As shown in the figure, there is a sideband around each of the carrier frequencies. Different adjacent fields (FIG. 14) may provide the same image 1518, 1528 because of the amplitude belonging to the carrier frequency with sidebands.

16 illustrates a cross row layout used for some SLMs. The tilt angle is increased by a factor of about 50 times, making the tilt recognizable. SLM modulation is a Fourier stop, or phase modulation that is converted into amplitude modulation in the aperture of the projection system. This conversion is not generic or automatic with reference to FIG. 16 and is the result of the planned mirror design, mirror size and layout of the tilt pattern. The detected image is identical to the image from the mask if the following two conditions are met: The sidebands must be symmetrical and the carrier must be high enough to avoid contamination at the opening by the sidebands around the carrier frequency. The image may not contain any phase information, ie the complex amplitude must be real at all points. This is most easily understood given the typical phase shift reticle. There is no phase shift reticle with phases other than 0 ° and 180 °, and the phase angle is precisely specified and closely monitored. The endurance to phase shift is different in SLM and PSM reticles and is generally quite complex, but follows the same general rule: there should be no large imaginary part of complex amplitudes in the image. If there is a phase difference in the image plane, there will be instability through the focus, ie the edge will move through the focus and the CD and / or overlay will be adversely affected.

Other conditions of no contamination at the opening stop by the sidebands from the carrier can be easily satisfied by making the mirror small, but there is a significant adverse effect on throughput. If the disappearance of the zero order occurs and the carrier is far enough from the opening, the method of diffracting light by wrinkling the surface is not important. Pistons, tilting mirrors or sinusoidal height modulation are all possible. For a tilting mirror, a layout with rows with mirrors tilting left and right provides a large clean area around the zero order and reduces contamination. The diffraction pattern is shown in FIG.

Piston stand Tilting mirror

With the tilting mirror, the behavior of the actual value can be guaranteed by symmetry. It has been proposed that piston mirrors provide higher contrast and image log slopes. It is indeed easy to set up an embodiment with a chromeless mask and a piston mirror operating with an on grid feature that provides the high contrast of a chromeless mask. Emulating a chromeless mask with a 1D feature off grid is less easy, but more likely. In this case, an intermediate mirror value should be used, which is not practical for the piston. The group of mirrors can be optimized to cancel the phase effect. It is more difficult to maintain the CD and log slopes at their design values in a typical 2D pattern while satisfying the phase offset conditions, and some mirror sizes are not possible.

The balance of phase with the piston mirror is not automatic like the tilting mirror, but must be explicitly controlled by the rasterizer. With a row of mirrors tilting in the cross direction, the mirror layout of FIG. 16 may be rasterized as if the mirror were an actual amplitude modulator, ie each mirror was essentially rasterized based on local data only for the same pixel. Can be. This simplifies the data path architecture and rasterization can be performed with explicit algorithms suitable for execution on the pipelined DSP architecture implemented in the FPGA. Piston mirrors require a more sophisticated rasterizing architecture, while at the same time a larger number of mirrors and smaller sizes. If you want to develop the higher contrast of a phase shifting mirror, do you have to accept this complexity? The ext section will provide a simpler phase shifting solution.

Phase Shifting Tilt mirror

17A-17D show trajectory and phase modulation mirror type in the complex plane. The panel shows (a) a flat tilting mirror, (b) a piston mirror, (c) a flat tilting mirror with dark centers, and (d) a tilting mirror with a phase step. The tilting mirror can be modulated to provide the strong phase shifting shown in FIGS. 17C and 17D. The automatic balance of the phases is maintained with simple mechanics, and all points on the real axis of the complex plane are accessible to a single mirror. The same SLM can be used for binary mode, attenuation mode, high transmission attenuation mode, 3-tone mode, cross aperture mode, phase edge mode and CPL mode. The only disadvantage is the loss of about twice the brightness.

The complex amplitude R is calculated as follows.

Where S is the surface of the mirror, r (x, y) is the local complex reflection coefficient, λ is the wavelength and h (x, y) is the local height. For the piston mirror, the complex amplitude is calculated as follows.

That is, the phase factor is multiplied by a constant integrated reflectance R ₀ .

The four mirror types and ballistics of the integrated complex reflectivity coefficient R are performed in the complex plane when they are operated. 17A shows a current mirror used in a sigma mask writer. R goes from 1 + 0i at 0 ° deflection of the mirror edge to -0.2 + 0i at 257 °. Negative amplitude can be used to emulate the attenuated PSM, or can be used more black than black in grid filters. The theoretical value of -0.2 is too small at the same time in the grid filter and in the negative black / damped PSM mode, but the actual device has a more negative amplitude, typically -0.3. This value can be further changed depending on the mirror design.

17B shows the piston mirror. Complex reflections follow the perimeter of the unit circle and are very bright at all phase values. Any point in the circle, ie the values of black and gray, can be accessed through a combination of two or more mirrors.

17C shows a tilting mirror having a center along an axis darkened by antireflective coating, light scattering microstructure or cut-out. When the area is removed from the center, the amount of negative amplitude increases, while the flat state reflection decreases, making the trajectory more symmetrical, but at the same time smaller. However, it is possible to scale the figures due to the change in the illumination energy, so the values 0.5 + 0i and -0.5 + 0i can be used as clears and shifters. In contrast to regular scanners, maskless tools have burnable laser energy. 20 times lower throughput means that the energy required for the wafer plane is 20 times smaller.

17D shows a step mirror that is a different tilting phase shift device. It has a phase step at the surface of λ / 4, corresponding to a 180 ° phase difference upon reflection. Since half of the area is shifted 180 °, it is dark when flat. Tilting to the right brightens the reflectivity by about 50%. Tilting to the left brightens it, but has a 180 ° phase difference. This mirror has some advantages, which are also useful for binary printing. In particular, it is very dark when not active, and its properties simplify the stitching of successive SLM images. Complex amplitude is symmetric around the origin of the complex plane, which is a desirable characteristic in CPL, phase edge and AAPSM modes. When a step mirror is used for a three-tone pattern, the value can be arbitrarily selected, for example, along the real axis of the CPL or the three-tone high transmission attenuation mode with a shifter less than -1 + i0.

Phase Shifting  Data path

18A-18D show various data paths: (a) an ideal data path with a proprietary data format that embodies many area types in the same file, (b) a modulated data path compatible with current infrastructure ( c) a modulated data path that can be generalized to any number of layers (d), a modulated version of c in which the layers are not rasterized individually. In contrast to the data paths shown in Figs. 18A and 18B, patterns other than binary and attenuated PSM have more than two levels in the pattern input data. As phase shift reticles are manufactured today, the input data describes two separate layers for the mask writer, chrome and shifter layers as separate files using a standard format such as GDSII or OASIS. It is particularly desirable to be compatible with this input detail, since the layers are not functionally identical. Since the reticle is masked by chrome, the shifter is typically larger in the data than intended for the reticle. 18C shows the data paths for reading two files, rasterizing them separately and combining the bitmaps into one single bitmap for SLM. Grid filters may be used in this single bitmap. 18B shows how the same two layers can be used to subdivide four region types rather than three region types using a dedicated data format. In general, a pattern having N tones can be described by the same type of custom format.

Data paths with dual rasterizers will undoubtedly be more expensive to design. Referring to FIG. 18D, it may or may not be advantageous to develop a new rasterizer that directly converts two or more standard input layers into a multivalue bitmap. More research has to be done in this area to explore the most useful rasterizers for phase shifting and multi-tone images.

Phase Shifting Example

19 schematically illustrates a method of printing a 35 nm transistor gate having a phase edge and a trim mask using SLM (a) and reticle (b). The resist profile of the 35 nm line printed using the phase edge generated by SLM is shown in (c). As can be seen in the magnification of (c), some positions relative to the grid are overlaid. FIG. 19 illustrates a simulated example of a phase edge using the step mirror of FIG. 9D and a grid filter modulated to operate in the phase shifting domain. 19A and 19B show in schematic form how a transistor gate structure is formed by a trim mask and a phase edge having either a regular reticle 19b or an SLM 19a. For SLMs and reticles, 0 and 180 represent the phase of reflected / transmitted light, while the gray level represents the tilted black SLM mirror. The simulation of FIG. 19C shows the resist profile for phase edges for positions of 0, 5, 10, 15, 20, 25 and 30 nm in relation to the mirror grid. Depending on the 65 nm node, the target CD is 35 nm. Surprisingly, it is noteworthy that lines that are not wider than a single pixel can be moved through the grid with changes that are only visible in the simulation results when they are highly consistent and highly magnified in their original state. The explanation for the surprising lack of such grid effects is a combination of the fact that grid filters and lines are reduced by overexposure.

The following example is shown in Figure 20: CPL semi-isolated lines with 65 nm and 45 nm line widths. In CPL, the shifter lines on the reticle, which may be provided with chromium on top to reduce permeability, are printed in polar sigma. The result is a thin dark line with good contrast and insensitivity to CD masking. The two embodiments shown here show that CD and contrast can be maintained as the line moves through the grid. Simulation of the step mirror of d was used. The result is again a line with neither CD nor contrast dependence in the position relative to the grid.

Phase for ASIC Shifting Enable

Manufacturing a PSM reticle will never be simple or easy. Repairing the shifter will always be difficult, and durability will always be tight due to the optical power of phase shifting. The stronger the phase shifting, the harder it is to manufacture. The difficulty of manufacturing a phase shifting maskless tool is another kind. The difficulty is the difficulty of development, and when there is a functional system, it will record phase shifting patterns as easily as binary patterns. Why does the CPL use binary patterns for gates when they are not more expensive and take less time?

Even if the assumption that phase shifting and binary numbers consume the same time and the same cost is not entirely true, the availability of phase shifting maskless tools will still tilt the field for more aggressive designs and processes. Even if it proves that the phase edge poly layers need to be written in four passes instead of two, the ASIC industry will save time and provide better performance than FPGAs.

Can phase shifting typical change to reticle-based production? Probably not, and probably even more economical to continue performing phase shifted layers despite quite large volumes. With functional tools, doors can have higher performance in smaller volumes and movement to the reticle will be performed separately based on economics and detailed planning. Perhaps a maskless-only process would be appropriate.

Pixel size

Phase shifting embodiments all use the same pixel size, 30 nm. This size was preselected (although a little conservative) as appropriate for binary and attenuated imaging modes. Features printed with phase shifting are smaller, but contrary to expectations, the simulations reported here indicate that 30 nm is still adequate. Clearly, due to grid filters, this result is still preliminary. Although disadvantageous to throughput, it is always possible to make printing better and more ideal by using smaller pixels.

Litho -neutrality, Litho -Matches and Litho -plus

The maskless tool can be executed in a different relationship than a regular scanner. In the process used in typical and subsequent fabrication without always going to the mask, the transparency between the mask and the maskless may not be an issue. This can be called litho-neutral, as opposed to situation-lithography-matching where consensus is needed.

Resource-matching is a difficult concept. Maximizing the processing window and processing proximity effects by the OPC software is simpler and simpler than matching OPCs of different machines. This is known from mask lithography, in which the same optical settings on two scanners of different types often do not print the same because of residual aberration and incomplete matching of the illuminator. It is important that a maskless tool is neither better nor worse than a scanner for transparency. Worse imaging causes unnecessary rework, but better image quality in maskless tools may cause the article to fail when switched to mask-based manufacturing.

By the same principle, the term litho-plus may refer to a situation where printing performance in a maskless tool has a higher priority than matching. The two most obvious ways to achieve litho-plus are digital processing of the bitmap and pattern decomposition into partial patterns printed with different optical settings. Dual pass and multiple pass printing on the stepper are partially hampered by overlay accuracy and partly hampered by the cost of the double reticle. The tradeoffs for maskless tools are different: overlays are very good because there is no need to realign wafers and reticles and there is no fixed cost of double reticles. Throughput, on the other hand, is inversely proportional to the number of passes. Notwithstanding this, it may be possible and useful to resolve patterns into logic and memory, high density and isolation features or different pitches, as well as x and y lines. An embodiment is a high density contact hole array that can be printed on negative resist by crossing high contrast lines formed by phase shifting. Polarization is another reason for resolution, especially for NA above 1.00.

Grid filters have been described as a way to achieve uniformity across the grid, but other kernels, typically induction furnaces, can be added. If the induction kernel is applied to the entire pattern all edges will be sharp-agrated, if desired it will provide a boost for thin lines and small features. The tradeoff is that the processed edges still have to fit the dynamic range of the pixel, which generates more digital noise and requires more laser energy. By adding a small induction term to all edges, the ED window of FIG. 13 can be made larger for the SLM. Induction terms on all edges can be added to the grid filter with minor changes to the architecture of FIG. 9.

summary

Rasterization from the Sigma mask writer to hold the CD across the grid was performed one more step and retained both the CD and log slopes across the off grid position of the feature. SLM images are very close to the image of the ideal image of the data without loss of visible grid and resolution. This makes the use of pixels very effective, and the three simulations of the paper still show lines smaller than two pixel widths that print well.

Compared to the image from the physical reticle, the image from the SLM is not worse and rather better because a large number of steps have been eliminated in the physical reticle limited resolution and precision. At the top of this, a digital filter can be used to increase the contrast than is possible with a physical reticle.

The target of the baseline design is to match the image characteristics of mask-based scanners for seamless transition and seamless mix-and-match of the design between mask lithography and maskless lithography.

These results indicate that a maskless tool can be configured to provide the same OPC model as a mask based scanner. After the grid filter removes the influence of the defined pixel size, the OPC characteristics are entirely determined by the projection optics and the lighting conditions.

Maskless tools have the best utility in the most difficult layers and are typically phase shifted and highly OPC. The tilting mirror used in Micronics' mask writer was modulated to provide strong phase shifting characteristics. It can be operated to produce any complex amplitude of -1 + 0i to + 1 + 0i on the real axis, while maintaining simple mechanical properties. Although the current data path can only rasterize the described patterns in two levels, a modulated data path capable of performing three-tone and multi-tone patterns has been described.

The illustrated embodiment, all related to baseline optics targeted for a 65 nm design node (λ = 193 nm, NA = 0.93 dry, and 30 nm projection pixel size), is shown at 50 nm (attPSM), 45 nm (CPL, dipole). ) And semi-isolated lines of 35 nm (phase edges). These are considered the most difficult cases for maskless tools because the line width is less than or equal to two mirrors and the grid filter is less effective for dipole lighting than for less extreme lighting modes.

The final conclusion is that the scale of all CD errors in this paper is negligible compared to the CD error budget. This result, which should be confirmed by more experiments and additional simulations, does not mean that there are no CD errors, and that they must come from different sources.

Some specific embodiments

The invention may be practiced as a method or a device adapted to carry out the method. The invention may be a fabrication technique, such as media impressed with logic for performing maskless emulation of the phase shifting method and the generation of OPC features.

One embodiment includes providing a spatial light modulator (SLM) having at least one mirror having a complex reflection coefficient with a negative real part and an adjacent mirror having a complex reflection coefficient with a positive real part. That's how. The method further includes illuminating the SLM with a partially coherent beam and converting vector data to drive the SLM. The vector input data includes more than two beam relaying states and is used in one or more lithographic image enhancement methods used with reticles. These methods of lithographic image enhancement are selected from the group of chromeless phase lithography (CPL), phase edge lithography, cross aperture (Levinson type) lithography, three-tone lithography or high transmission attenuation lithography. More than two beam relaying states may include either fully on and fully off plus gray areas or phase shift areas, as described in vector data prior to rasterizing.

Another aspect of the first embodiment is defined by the SLM using one or more mirrors oriented such that one or more pattern edges have complex reflection coefficients with negative real parts, and emulate one or more lithographic image enhancement methods.

Continuation of further embodiments includes emulating a particular method of lithographic image enhancement. One of these embodiments is a method of forming a lithographic pattern in an image plane on a workpiece using a spatial light modulator, having one or more mirrors with complex reflection coefficients with negative real parts, and using partially coherent light. Illuminating the SLM with the partially coherent light. The method includes driving a mirror having a complex reflection coefficient with negative real parts up to the phase edge in contrast to one or more adjacent mirrors and projecting partially coherent light onto the image plane through a finite opening from the SLM. It includes more.

Another such embodiment is a method of forming a lithographic pattern in an image plane on a workpiece using a spatial light modulator, having at least one mirror having a complex reflection coefficient with negative real parts, and using partial coherent light, Illuminating the SLM with coherent light. The method involves driving a mirror having a complex reflection coefficient with a negative real part to emulate phase interference between rows of a chromeless phase shift mask and projecting partially coherent light onto the image plane through a finite aperture from the SLM. Steps.

Another embodiment is a method of forming a lithographic pattern in an image plane on a workpiece using a spatial light modulator, having one or more mirrors with complex reflection coefficients with negative real parts, and using partial coherent light, partial interference Illuminating the SLM with sex light. The method further includes driving a mirror having a complex reflection coefficient with a negative real part to emulate a cross aperture phase shifting mask and projecting partial coherent light onto the image plane through a finite aperture from the SLM. do.

Another embodiment is a method of forming a lithographic pattern in an image plane on a workpiece using a spatial light modulator, having at least one mirror having a complex reflection coefficient with negative real parts, and using partial coherent light, Illuminating the SLM with coherent light. The method further comprises driving a mirror having a complex reflection coefficient with a negative real part to emulate a three-tone phase shifting mask and projecting partially coherent light onto the image plane through a finite aperture from the SLM. Include.

A related embodiment is a method of forming a lithographic pattern in an image plane on a workpiece using a spatial light modulator, having at least one mirror having a complex reflection coefficient with negative real parts, and using partial coherent light, and partial interference Illuminating the SLM with sex light. The method includes driving a mirror having a complex reflection coefficient with a negative real part to emulate a high transmission attenuated phase shifting mask and projecting partial coherent light onto the image plane through a finite opening from the SLM. .

Another disclosed embodiment is a method of exposing a lithographic pattern comprising providing a spatial light modulator having at least one mirror having a complex reflecting component with a negative real part and an adjacent mirror having a complex reflecting coefficient having a positive real part. . The method includes illuminating the SLM with a partially coherent beam and transforming vector input data to drive the SLM. Vector input data including OPC features or decompositions were used for lithographic image enhancement used with the reticle. The OPC feature or decomposition is any of a group of scatter bars, serifs, OPC jogs or double dipole decomposition.

Continuation of related embodiments includes the emulation of OPC features or decomposition used with reticles. One related embodiment is a method of forming a lithographic pattern in an image plane on a workpiece using a spatial light modulator, having at least one mirror having a complex reflection coefficient with negative real parts, and using partial coherent light, Illuminating the SLM with a coherent illumination source. The method further includes driving the mirror to emulate one or more subprinting resolution scatter bars and projecting partial coherent light onto the image plane through a finite opening from the SLM.

Another related embodiment is a method of forming a lithographic pattern on an image plane on a workpiece using a spatial light modulator, having at least one mirror having a complex reflection coefficient with negative real parts, and using partial coherent light, Illuminating the SLM with a partially coherent illumination source. The method further includes driving the mirror to emulate a subprinting resolution serif and projecting partially coherent light onto the image plane through a finite opening from the SLM.

Another embodiment is a method of forming a lithographic pattern in an image plane on a workpiece using a spatial light modulator, having one or more mirrors with complex reflection coefficients with negative real parts, and using partial coherent light, partial interference Illuminating the SLM with the castle illumination source. The method further includes driving the mirror to produce a jogging alignment pattern enhanced by the phase difference between adjacent mirrors of the SLM and projecting partial coherent light onto the image plane through a finite opening from the SLM.

Yet another embodiment is a method of forming a lithographic pattern in an image plane on a workpiece using a spatial light modulator, having at least one mirror having a complex reflection coefficient with negative real parts, and using partial coherent light, Illuminating the SLM with a coherent illumination source. The method further includes driving the mirror to emulate double exposure dipole resolution resolution enhancement using multiple exposures of the SLM and projecting partially coherent light onto the image plane through a finite opening from the SLM.

Although the present invention has been described with reference to the above-described preferred embodiments and examples, it will be understood that these examples are intended to be illustrative rather than limiting. Computer-assisted processing is included in the embodiments described above. Accordingly, the present invention provides a method for emulating mask-based lithography using a phase shifting SLM, a system comprising logic and resources for performing emulation of mask-based lithography using a phase shifting SLM, a phase shifting SLM Impressed media with logic to perform emulation of mask-based lithography, masked using phase shifting SLM or data stream impressed with logic to perform emulation of mask-based lithography using phase shifting SLM It may be embodied as a computer-accessible service that performs computer-assisted emulation of based lithography. It will be understood that modifications and combinations can be readily made by those skilled in the art within the spirit of the invention and the following claims.

Claims (12)

Providing a spatial light modulator (SLM) having at least one mirror having a complex reflection coefficient having a negative real part and an adjacent mirror having a complex reflection coefficient having a positive real part; Illuminating the SLM with a partially coherent beam; Converting vector input data to drive the SLM, The vector input data is, Chromeless phase lithography (CPL), Phase edge lithography, Cross-opening (levinson type) lithography, 3-tone lithography, or A method of lithographic pattern exposure in a group of high transmission attenuated lithography, used in one or more lithographic image enhancement methods used with reticles. The method of claim 1, Wherein the one or more pattern edges are defined by an SLM using one or more mirrors oriented to have complex reflection coefficients with negative real parts, and emulate one or more lithographic image enhancement methods. A method of forming a lithographic pattern on an image plane on a workpiece using a spatial light modulator (SLM), having at least one mirror having a complex reflection coefficient with negative real parts, and using a partially coherent light source, Illuminating the SLM with the partially coherent light; Driving a mirror having a complex reflection coefficient with a negative real part to one or more adjacent mirrors to a phase edge; And And projecting partially coherent light from the SLM through an finite aperture onto an image plane. A method of forming a lithographic pattern on an image plane on a workpiece using a spatial light modulator (SLM), having at least one mirror having a complex reflection coefficient with negative real parts, and using a partially coherent light source, Illuminating the SLM with the partially coherent light; Driving a mirror having a complex reflection coefficient with a negative real part to emulate phase interference between rows of a chromeless phase shift mask; And And projecting partially coherent light from the SLM through an finite aperture onto an image plane. A method of forming a lithographic pattern on an image plane on a workpiece using a spatial light modulator (SLM), having at least one mirror having a complex reflection coefficient with a negative real part, and using a partially coherent light source, Illuminating the SLM with the partially coherent light; Driving a mirror having a complex reflection coefficient with a negative real part to emulate a cross aperture phase shifting mask; And And projecting partially coherent light from the SLM through a finite aperture onto an image plane. A method of forming a lithographic pattern on an image plane on a workpiece using a spatial light modulator (SLM), having at least one mirror having a complex reflection coefficient with negative real parts, and using a partially coherent light source, Illuminating the SLM with the partially coherent light; Driving a mirror having a complex reflection coefficient with a negative real part to emulate a three-tone phase shifting mask; And And projecting partially coherent light from the SLM through a finite aperture onto an image plane. A method of forming a lithographic pattern on an image plane on a workpiece using a spatial light modulator (SLM), having at least one mirror having a complex reflection coefficient with a negative real part, and using a partially coherent light source, Illuminating the SLM with the partially coherent light; Driving a mirror having a complex reflection coefficient with a negative real part to emulate a high transmission attenuated phase shifting mask; And And projecting partially coherent light from the SLM through an finite aperture onto an image plane. Providing a spatial light modulator (SLM) having at least one mirror having a complex reflection coefficient having a negative real part and an adjacent mirror having a complex reflection coefficient having a positive real part; Illuminating the SLM with a partially coherent beam; Converting vector input data to drive the SLM, The vector input data is Scatter Bar, Serif, OPC jog, or A method of lithographic pattern exposure, comprising OPC features or decomposition, used in a lithographic image enhancement method used with a reticle, among a group of dipole decomposition. A method of forming a lithographic pattern on an image plane on a workpiece using a spatial light modulator (SLM), having at least one mirror having a complex reflection coefficient with a negative real part, and using a partially coherent light source, Illuminating the SLM with the partially coherent light; Driving a mirror having a complex reflection coefficient with a negative real part to emulate one or more subprinting resolution scatterbars; And And projecting partially coherent light from the SLM through a finite aperture onto an image plane. A method of forming a lithographic pattern on an image plane on a workpiece using a spatial light modulator (SLM), having at least one mirror having a complex reflection coefficient with a negative real part, and using a partially coherent light source, Illuminating the SLM with the partially coherent light; Driving a mirror having a complex reflection coefficient with a negative real part to emulate one or more subprinting resolution serifs; And And projecting partially coherent light from the SLM through a finite aperture onto an image plane. A method of forming a lithographic pattern on an image plane on a workpiece using a spatial light modulator (SLM), having at least one mirror having a complex reflection coefficient with a negative real part, and using a partially coherent light source, Illuminating the SLM with the partially coherent light; Driving a mirror having a complex reflection coefficient with a negative real part to produce a jog in a line pattern enhanced by a phase difference between adjacent mirrors of the SLM; And And projecting partially coherent light from the SLM through a finite aperture onto an image plane. A method of forming a lithographic pattern on an image plane on a workpiece using a spatial light modulator (SLM), having at least one mirror having a complex reflection coefficient with a negative real part, and using a partially coherent light source, Illuminating the SLM with the partially coherent light; Driving a mirror having a complex reflection coefficient with a negative real part to emulate double exposure dipole resolution resolution enhancement using multiple exposures of the SLM; And And projecting partially coherent light from the SLM through a finite aperture onto an image plane.

Images