US 6,498,685 B1Grant

Maskless, microlens EUV lithography system

Issue Date:2002-12-24

•33 Claims

•36 Drawing Sheets

Abstract

A maskless, extreme ultraviolet (EUV) lithography system uses microlens arrays to focus EUV radiation (at an operating wavelength of 11.3 nm) onto diffraction-limited (58-nm FWHM) focused spots on a wafer printing surface. The focus spots are intensity-modulated by means of microshutter modulators and are raster-scanned across a wafer surface to create a digitally synthesized exposure image. The system uses a two-stage microlens configuration to achieve both a high fill factor and acceptable transmission efficiency. EUV illumination is supplied by a 6 kHz xenon plasma source, and the illumination optics comprise an aspheric condenser mirror, a spherical collimating mirror, and two sets of flat, terraced fold mirrors that partition the illumination into separate illumination fields covering individual microlens arrays. (The system has no projection optics, because the image modulator elements are integrated with the microlens arrays.) The printing throughput is estimated to be 62 (300-mm) wafers per hour (assuming a resist exposure threshold of 20 mJ/cm²), and print resolution is estimated at 70 nanometers for mixed positive- and negative-tone patterns (at k₁=0.6).

Metadata

Inventor

Kenneth C. Johnson

Application Information

Application Number:US 09/478,233

Filing Date:2000-01-04

Priority Date:1999-01-11

Art Unit:7

Classifications

IPC:

G02B 2710B41J 1514B41J 2700

Field of Search:

2504922250492222504921359619359626355 44355 45355 55347239347241347244378 34378 35378 84378 85

Patent Drawings (36 sheets)

Description

[0002] This application claims priority from the following provisional applications, the disclosures of which are incorporated by reference:

[0003] “Conformal Imaging Microlens Array”, provisional application 60/115,451, filed Jan. 11, 1999.

[0004] “Conformal Imaging Microlens Array with High Fill Factor”, provisional application 60/119,403, filed Feb. 1, 1999.

[0005] “Spatially modulated microlens array for EUV maskless lithography”, provisional application 60/116,074, filed Jan. 15, 1999.

[0006] “High-Resolution EUV Microlens Design”, provisional application 60/119,655, filed Feb. 11, 1999.

[0007] “EUV Microlens System Design Outline”, provisional application 60/123,527, filed Mar. 8, 1999.

[0008] “Improved grating modulator design for EUV maskless lithography”, provisional application 60/124,140, filed Mar. 12, 1999.

[0009] “ETV Microlens System Design Outline”, provisional application 60/124,422, filed Mar. 15, 1999.

[0010] “EUV Modulator Design Concepts”, provisional application 60/125,487, filed Mar. 22, 1999.

[0011] “EUV Microlens System Design Variations and Refinements”, provisional application 60/133,450, filed May 11, 1999.

[0012] “Collection Optics Design for EUV Microlens Lithography”, provisional application 60/135,636, filed May 24, 1999.

[0013] “EUV Microlens Design Options”, provisional application 60/136,925, filed Jun. 1, 1999.

[0014] “Moiré Alignment Tracking System for EUV Microlens Lithography”, provisional application 60/137,309, filed Jun. 3, 1999.

[0015] “Phase-Measuring Microlens Microscopy”, provisional application 60/139,002, filed Jun. 14, 1999.

[0016] “EUV Microlens Fabrication Options”, provisional application 60/143,470, filed Jul. 12, 1999.

[0017] “Twice-improved grating modulator design for EUV maskless lithography”, provisional application 60/151,461, filed Aug. 30, 1999.

[0018] “Maskless EUV Lithography System Employing Transmission Micro-Optics”, provisional application 60/162,684, filed Nov. 1, 1999.

BACKGROUND OF THE INVENTION

[0019] This invention relates primarily to microlithography, and more specifically to extreme ultraviolet (EUV) lithography. The invention is applicable to semiconductor lithography and lithographic patterning of microstructures such as micromechanical systems, micro-optics, porous membranes, etc.

[0020] EUV lithography is a leading candidate technology for next-generation semiconductor lithography, which will require sub-100-mn printing resolution. An industry consortium (the EUV Limited Liability Company) is focusing development efforts on an engineering approach that is basically an evolutionary extension of LW and DUV projection lithography, the main difference being that the EUV system requires all-reflective optics due to the lack of EUV-transmitting lens materials (Ref. 1). The basic system design has nine reflective surfaces between the EUV source and the wafer, including four condenser mirrors, the photomask, and four projection camera mirrors. Seven of the reflectors (including the photomask) comprise multilayer coatings that operate at near-normal incidence and are optimized for peak reflectance at a wavelength of 13.4 nm.

[0021] Mirror coatings are one of the key enabling technologies that have made EUV lithography possible. Presently, multilayer Mo/Si (molybdenum/silicon) multilayer mirrors routinely achieve peak reflectance efficiencies of 67.5% at the 13.4-nm operating wavelength, with a spectral reflectance bandwidth of 0.56-nm FWHM (Refs. 2, 3). Nevertheless, the lithography system's optical efficiency is significantly limited by the compound reflectance losses of seven multilayer mirrors operating in series.

[0022] The mirror surface figure tolerances scale in proportion to the operating wavelength, which is an order of magnitude smaller than DUV wavelengths at the 13.4-nm EUV wavelength. Furthermore, tolerance requirements for mirrors are generally much more stringent than lenses. Thus, the four EUV projection camera mirrors which image the mask onto the wafer must be fabricated to extremely tight (i.e., atomic-scale) absolute figure tolerances over large aperture areas.

[0023] The system uses multilayer reflective masks, which operate at 4:1 reduction. Mask fabrication presents significant technical challenges due to factors such as the short operating wavelength and small feature dimensions, lack of EUV pellicle materials, and very stringent registration requirements. EUV masks are expected to cost approximately $50,000 at 100-nm print resolution (Ref. 4).

[0024] Two research groups at the University of California at Berkeley and Stanford University are currently investigating maskless EUV lithography approaches (Refs. 5, 6). Their basic approach is to replace the mask with a spatial light modulator (SLM) comprising a programmable array of micromirrors whose reflectance properties can be varied by modulating the mirrors' tilt or translational positions. Although these are maskless systems, they nevertheless require EUV projection optics to image the micromirror array onto the wafer. Due to minimum size limits of the SLM pixel elements, the projection optics must operate at a very high reduction ratio, (e.g., 20 to 40× demagnification), which precludes the use of current-generation (4×) EUV projection optics.

[0025] Another limitation of micromirror systems is that the SLM must operate at a very high frame rate (e.g., multi-megahertz) in order to achieve requisite throughput requirements. (The system's total data throughput requirement is at least 10 bits per second.) This requires a continuous EUV source such as a synchrotron, and precludes the use of a pulsed EUV source such as a high-power, laser-produced plasma (LPP) source that is currently being developed by Sandia National Laboratories for EUV lithography (Refs. 7, 8).

[0026] Several recent patents and publications describe a variety of other SLM-based maskless lithography design concepts, some of which are applicable to EUV (Refs. 9-18). Most of these systems employ projection optics, but one of these inventions, a Fresnel zone plate microlens system (Refs. 16-18), eliminates the need for projection optics as well as the mask. The system is designed primarily for x-ray lithography at a wavelength of 4.5 nm, but the basic design principle could be applied equally well to EUV. The device comprises an array of zone plate microlenses that focus an incident x-ray (or alternatively, EUV) beam onto an array of diffraction-limited points on a wafer surface. Each microlens transmission is modulated by a micro-actuated shutter as the wafer is raster-scanned across the focal point array to build up a synthesized, high-resolution image.

[0027] Since the zone plate system does not use projection optics, it could conceivably use a very large number of zone plate lenses in multiple, large-area arrays, all operating in parallel. Based on printing throughput requirements, the required modulation frame rate would scale in inverse proportion to the number of lenses, so with a sufficiently large number of lenses the system could conceivably use an LPP EUV source operating at a moderate frame rate (e.g., 6 kHz, Ref. 8). However, the proposed method for manufacturing the zone plates lenses (spatial-phase-locked e-beam lithography) is not very practical for large-quantity production, and there are also other complicating factors that preclude the use of an LPP source with the zone plate system.

[0028] The plasma source in the LPP has a fairly large spatial extent (approximately 200 μm diameter, Ref. 7), and since each focus point on the wafer is a diffraction-limited image of the source, optical resolution would be limited by the source's geometric image size. Furthermore, the LPP would be used in conjunction with EUV mirrors that have a 2% combined spectral bandwidth, and the zone plate's chromatic dispersion within this band could also significantly limit focus resolution. The source's geometric image size on the wafer and the image's chromatic spread should preferably both be significantly smaller than the diffraction-limited image of an ideal monochromatic point source. This condition could be achieved, in principle, by using very small microlenses with short focal lengths; but the zone plate lens elements must be sufficiently large to accommodate the shutter mechanism, data paths, and supporting framework, while also maintaining an acceptably high aperture fill factor. Due to this aperture size limitation, a practical zone plate system would probably perform poorly with an LPP, and a more costly synchrotron source would thus be required.

[0029] Another limitation of zone plate lithography is that the printing resolution and contrast can be significantly degraded by extraneous diffracted orders. Continuous-profile, refractive microlenses such as those shown in Ref. 13 would not have this limitation, but for EUV application such microlenses would be either too highly absorbing or too small to be of practical utility when used in the mode described in Ref. 13.

SUMMARY OF THE INVENTION

[0030] The present invention overcomes the limitations of prior-art microlens printing systems by using multiple microlenses in series to focus exposure illumination onto a printing surface such as a semiconductor wafer. In a preferred embodiment of the invention, the microlenses are used in pairs, each pair consisting of first- and second-stage microlenses, respectively designated as lens L₁and lens L₂, wherein L₁focuses illuminating radiation onto L₂, and L₂thence focuses the radiation to a focal point on the printing surface. Each such microlens pair forms a printer pixel, and multiple such pixels are combined to form a printhead, which focuses the radiation onto multiple focal points on the printing surface. A modulator mechanism modulates the focal points' exposure intensity levels, and as the points are modulated, the printhead is scanned relative to the printing surface (either the surface or the printhead, or both, can be moved for the scanning) to build up a synthesized, high-resolution exposure image on the print surface.

[0031] In general, a printer pixel can comprise a sequence of two or more microlenses L₁, L₂, . . . L_N(N being an integer greater than 1), wherein each lens L_m(1≦m<N) focuses the illuminating radiation onto the next lens L_m+1 in the sequence, and the last lens L_Nfocuses the radiation onto a focal point on the printing surface. Typically L₁is a comparatively large element with low optical power, and the succeeding lenses are progressively smaller and have progressively higher optical power.

[0032] The modulator mechanism preferably comprises modulator elements that are incorporated as components of the printer pixels, with each modulator element providing independent, digital control of the exposure level at a corresponding focal point. However, in some embodiments a single modulator could be used to modulate the radiation before it is conveyed to the printhead. This option would be useful for applications that do not require independently modulated pixels, for example, in the fabrication of hole arrays for microfiltration membranes. The preferred embodiments use binary-state (i.e., ON/OFF) modulators, but gray scale (continuous-level) modulation may also be employed.

[0033] The disclosed embodiments relate primarily to the ETV lithography application, although the disclosed design methods and configurations are also applicable or adaptable to other wavelength ranges such as DUV. In a preferred embodiment, molybdenum microlenses are used to focus an incident EUV beam (at a wavelength of 11.3 nm) onto an array of diffraction-limited (58-nm FWHM) focal points on a wafer surface. A modulator mechanism, such as an SLM comprising microshutters proximate to the microlens apertures, modulates the beam intensity at each focal point. As the points are modulated, the focal point array is raster-scanned across the wafer to build up a digitally synthesized, high-resolution image.

[0034] The microlenses are preferably continuous-profile, refractive elements. A refractive EUV lens with sufficient optical power to achieve the requisite focus spot resolution (<100-nm spot size) would have to be very small—on the order of 1 μm diameter or less; otherwise the lens would be too thick to allow acceptable EUV transmittance. A single array of microlenses and modulators of this size could not be practicably manufactured without incurring unacceptable fill factor losses. But this limitation is circumvented by using a multi-stage microlens system. In a two-stage embodiment, comparatively large (5-μm diameter), low-power (NA=0.015) first-stage lenses focus EUV illumination onto much smaller (1-μm diameter) second-stage lenses which have much higher optical power (NA≧0.1). Each first-stage lens's diffraction-limited focused spot overfills the 1-μm aperture of a corresponding second-stage lens; and the second-stage lens focuses the illumination down to a much smaller (58-nm FWHM) focus spot on the wafer. This design approach makes it possible to use a sparsely distributed array of second-stage lens elements without incurring significant fill factor losses. The sparse distribution simplifies fabrication processes, and the wide spacing between second-stage lens apertures provides accommodation for structural support and for the SLM data paths and actuators.

[0035] The disclosed two-stage embodiments of the invention have a number of advantages over prior-art EUV lithography systems, including some or all of the following: (1) The need for EUV photomasks is eliminated. (2) By integrating the SLM with the microlens arrays, the need for image projection optics is eliminated. (3) The optical tolerance requirements for the microlenses are very moderate compared to all-reflective EUV projection optics. (4) Proximate image points are exposed sequentially—not simultaneously—so they do not interact coherently, and the printed image would thus be devoid of the kind of coherent proximity effects that are exhibited by projection systems. (5) The second-stage lens apertures function as spatial filters that eliminate stray radiation and scatter from sources such as flare in the EUV illumination optics. (6) The EUV illumination source is imaged onto the second-stage microlens apertures, not onto the wafer; so the source size and spatial energy distribution do not significantly affect optical resolution (although these factors do have a minor influence on optical efficiency). (7) In contrast to zone plate lenses, the refractive lens elements could produce very clean and highly resolved focus spots free of spurious diffraction orders and scatter, and exhibiting negligible chromatic dispersion.

[0036] The wafer exposure process is commonly referred to as “printing”, and the combination of each associated first- and second-stage (or more generally, first- through N-th stage) microlenses and their associated modulator component is referred to herein as a “printer pixel”. (In some alternative embodiments, the SLM need not be integrated with the microlens arrays and the focus spot intensities need not be controlled by separate modulator elements. In this case the pixels would not include the modulator components.) The pixels are assembled into arrays termed “printheads”; and the printheads are typically assembled into larger arrays, termed “wafer print modules”, each of which covers and exposes a full 300-mm wafer. The complete exposure system (or “printer”) typically comprises multiple wafer print modules covering separate wafers, which are all supplied EUV illumination from a single EUV source.

[0037] In a specific embodiment, each printhead comprises approximately 1.3×10⁷pixels distributed over a 20-mm square area. In addition to housing the microlenses and SLM components, each printhead can be equipped with peripheral microsensors that sense its position relative to an alignment pattern on the wafer, such as a periodic tracking pattern formed in the wafer scribe lines. Each printhead can also be provided with its own micro-positioning actuators that dynamically maintain focus, tilt, and overlay alignment in response to the position sensor feedback from a large number of microsensors. This design approach has two advantages relative to more conventional “through the lens” or “off axis” alignment sensors employed in projection lithography systems (Ref. 19, Chap. 5). First, the effect of nonsystematic errors such as noise in the tracking signals is minimized by using a large number of microsensors in the alignment tracking system. Second, by combining the printing microlenses, positioning sensors, and positioning actuators in a single, compact unit in close proximity to the wafer, tolerance stack-up in the tracking servo is minimized and very accurate and responsive tracking control can be achieved.

[0038] Thermal expansion differences between the printhead and the wafer could result in significant effective magnification errors and overlay misregistration. For example, a 0.1° C. temperature difference could induce a 5-nm overlay error (which is most of the alignment tolerance budget). Magnification errors can be substantially eliminated by forming each second-stage microlens on an individually controllable micromechanical actuator which makes nanometer-scale lateral positioning adjustments in response to the tracking sensor feedback. The microlens actuators could also be used to correct a variety of other critical error factors. (For example, the allowable tolerance range on tilt and coma in the second-stage microlenses can be more than doubled by applying a calibrated position offset to each actuator.)

[0039] The system is designed for use with a high-power (1700 W), xenon LPP source operating at a 6 kHz repetition rate (Ref. 8). The system's total data throughput requirement (>10¹²bits per sec) and the low (6 kHz) frame rate necessitate the use of a very large number of printer pixels (>10⁸). Furthermore, an even higher number of pixels is needed if the system is designed to optimize optical efficiency. The microlens efficiency is affected by the source collimation, and efficiency would be significantly compromised if the EUV illumination's angular spread were much more than 1 mrad. Given that the source diameter is close to 200 μm (Ref. 7), this implies that the collimation optics should preferably have a minimum effective focal length of at least 200 mm (i.e., 200 μm/0.001 rad). With a focal length this large, acceptable optical efficiency can only be achieved by collecting radiation over a very large area; and hence a very large number of microlenses is used to efficiently utilize the source.

[0040] The high microlens number requirement is achieved by distributing the ETV illumination over a large number of printheads covering multiple wafers. The printer comprises a total of 360 printheads (i.e., 4.6×10⁹pixels), which are divided among eight wafer print modules, each module comprising 45 printheads distributed over a 300-mm wafer. Eight wafers are simultaneously exposed with a single LPP illumination source.

[0041] This “massively parallel” design approach is feasible because the SLM is not imaged through projection optics. Aside from considerations of LPP source compatibility, this approach has the advantage that the SLM frame rate and tracking speed, which scale in inverse proportion to the number of pixels, would be very moderate. (The micromechanical and tracking servo design requirements for a 6 kHz system would be much simpler than alternative multi-megahertz SLM systems.) The SLM power consumption and radiative heat gain are also minimized with this approach. Additionally, the large number of printheads (360) makes the data flow easier to handle. The system's total data rate for pattern generation is 2.8×10¹³bits per sec. This translates to 7.7×10¹⁰bits per sec per printhead (i.e., 2.8×10¹³/360), which can be managed, for example, with several fiber optic data lines and integrated demultiplexers. The data storage and data processing hardware would make up a significant portion of the system cost, but this cost can be minimized by supplying a common data stream to all eight modules.

[0042] The EUI illumination encounters four mirror surfaces between the source and the printheads. These include a deep, aspheric condenser element, a shallow, spherical collimator element, and two sets of flat, terraced fold mirrors that partition the collimated beam into individual rectangular illumination fields on the printheads. The condenser and collimator mirrors are multilayer Mo/Be (molybdenum/beryllium) mirrors optimized for peak reflectivity at a wavelength of 11.3 nm, whereas the fold mirrors operate at glancing incidence and can therefore use a much simpler broadband reflective coating (such as a ruthenium film).

[0043] The Mo/Be mirrors have a reflectance bandwidth of 0.27-nm FWHM per mirror, about half that of Mo/Si mirrors optimized for 13.4 nm (Ref. 2). Nevertheless, the total reflected EUV power is much higher with Mo/Be mirrors because the xenon LPP source's emission spectrum has a sharp peak near 11 nm (Ref. 7). The narrow bandwidth of Mo/Be mirrors is problematic with EUV projection systems because it is difficult to accurately match spectral reflectance peaks between seven mirrors (Ref. 3). But with just two multilayer mirrors the maskless system's matching tolerance is much less critical, and it is easier to take advantage of the higher source emission at 11 nm.

[0044] The system's printing resolution is estimated at 70 nm for mixed positive- and negative-tone images. Printing throughput is estimated at 62 300-mm wafers per hour, based on an assumed resist solubility threshold of 20 mJ/cm². (The maximum attainable flood-exposure level is 80 mJ/cm².) Projection EUV system designs typically assume a more stringent exposure sensitivity of 10 mJ/cm², based on 100-nm print resolution and a 13.5-nm exposure wavelength. But the microlens system is designed for higher resolution (70-nm feature size) and operates at a shorter (11.3-nm) exposure wavelength, making the print quality more susceptible to line edge roughness induced by shot noise (Ref. 20). The higher 20-mJ/cm²exposure threshold would counterbalance the system's higher shot noise sensitivity. (If a 10-mJ/cm²exposure threshold were assumed, throughput could be doubled.)

[0045] The microlens elements can be manufactured with nanometer-scale profile control by adapting shadow-mask deposition and etching processes that are used, for example, in shadow-mask molecular beam epitaxy (MBE), Refs. 21-24. Interference lithography methods can also be used to scale up the shadow-mask processes for high-volume production. Since the refractive index of molybdenum is very close to one at EUV wavelengths, the optical surface tolerances are comparatively loose in relation to mirror optics. The estimated figure tolerances are 4 nm RMS for the first-stage (5-μm diameter) element and 2 nm RMS for the second-stage (1-μm diameter) element, assuming that a calibrated lateral position offset is applied to the second-stage elements. (By comparison, EUV projection mirrors are required to meet a 0.25-nm RMS figure tolerance across aperture dimensions exceeding 100 mm, Ref. 1.) Additionally, the microlens system is very tolerant of isolated pixel defects. For example, an isolated nonfunctional pixel (i.e., a pixel that is permanently stuck in either the ON or OFF state) would only induce an estimated 1% uncorrected dimension shift on a 70-nm feature.

[0046] A further understanding of the nature and advantages of the present invention may be realized by reference to the remaining portions of the specification and the drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

[0047] FIG. 1 is a cross-sectional schematic view of a printer pixel comprising two EUV microlenses and a modulator;

[0048] FIG. 2 is a cross-sectional schematic view of a printhead comprising multiple pixels;

[0049] FIG. 3 is a plan view of a printhead comprising a printer pixel array and two position sensor arrays;

[0050] FIG. 4 is a cross-sectional schematic view of a wafer print module comprising multiple printheads;

[0051] FIG. 5 is a plan view of a wafer print module and wafer;

[0052] FIG. 6 is a cross-sectional view of the EUV collection optics' condenser and collimator mirrors;

[0053] FIG. 7 illustrates the optical geometry of a “refractive conjugate surface”;

[0054] FIG. 8 illustrates the relationship between optical surface positioning error and resulting optical phase error;

[0055] FIG. 9 is a cross-sectional view of a first-stage EUV microlens;

[0056] FIG. 10 is a cross-sectional view of a second-stage EUV microlens;

[0057] FIG. 11 is a wafer-plane amplitude profile plot for the focused beam produced by a second-stage EUV microlens;

[0058] FIG. 12 is a wafer-plane, peak-normalized intensity plot for the focused beam produced by a second-stage EUV microlens, in comparison to that of an ideal, unapodized, circular-aperture lens;

[0059] FIG. 13 is a tabular summary of the lens design parameters;

[0060] FIGS. 14A and 14B illustrate an alternative phase Fresnel lens design for the first-stage microlens;

[0061] FIG. 15 illustrates an alternative triplet design for the second-stage microlens;

[0062] FIG. 16 illustrates an alternative pixel configuration comprising three microlenses;

[0063] FIG. 17 illustrates a micromechanical actuator design for adjusting a second-stage microlens's lateral position;

[0064] FIGS. 18A-18F illustrate a method for fabricating three-element second-stage lenses;

[0065] FIG. 19 illustrates a modulator mechanism comprising a micromechanical shutter;

[0066] FIGS. 20A and 20B illustrate an alternative modulator mechanism comprising two proximate diffraction gratings, one of which is micromechanically actuated;

[0067] FIG. 21 illustrates an EUV illumination system comprising a condenser mirror, collimator mirror, and two sets of fold mirrors;

[0068] FIG. 22 illustrates the collimator mirror's aperture geometry;

[0069] FIG. 23 shows dimensional details of the first fold mirror set, in the projected view of FIG. 22;

[0070] FIG. 24 shows the first fold mirror set, from a perspective looking into the reflected beams;

[0071] FIG. 25 shows the fold mirrors, wafer print module, and wafer, from a perspective looking at the second set of fold mirrors edge-on;

[0072] FIG. 26 illustrates the aperture geometry of the second fold mirror set, from a perspective looking into the beams reflected by the second set;

[0073] FIG. 27 illustrates the illumination system's optical clearances around the EUV plasma source;

[0074] FIG. 28 illustrates the series reflectivity spectra for the two fold mirror sets (assuming bare ruthenium coatings), and also illustrates series reflectivity spectra of the condenser and collimator mirrors for two different incidence angles;

[0075] FIG. 29 illustrates design construction details for the condenser and collimator mirrors;

[0076] FIG. 30 illustrates the EUV plasma source's angular emission spectrum;

[0077] FIG. 31 illustrates the EUV plasma source's on-axis spectral radiant intensity;

[0078] FIG. 32 illustrates an incremental solid angular range bounding a set of rays emitted from the plasma source;

[0079] FIG. 33 illustrates an annular portion of the collimator's reflected beam corresponding to the solid angular range of FIG. 32;

[0080] FIGS. 34A and 34B illustrate a simple sequential dot row exposure pattern in which each pixel exposes a contiguous sequence of dots;

[0081] FIGS. 35A and 35B illustrate an interleaved dot exposure pattern;

[0082] FIGS. 36A and 36B illustrate a variant of the interleaved scan technique in which a printhead comprising a pixel row with a finite number of pixels is iteratively stepped and scanned to cover a dot row of indefinite extent;

[0083] FIG. 37 illustrates exposure gaps in the “stitch zone” at the boundary of a scan field;

[0084] FIG. 38 illustrates the positional relationship between two adjacent printheads;

[0085] FIG. 39 illustrates a printhead and scan path for the “lens-scan” scanning method;

[0086] FIG. 40 illustrates the image dot positional coordinates;

[0087] FIG. 41 illustrates the printer pixel positional coordinates;

[0088] FIG. 42 illustrates a scan field covered by a printhead;

[0089] FIG. 43 illustrates four juxtaposed scan fields covered by a single printhead in separate scans;

[0090] FIG. 44 illustrates four proximate scan fields covered by separate printheads in a single scan;

[0091] FIG. 45 summarizes the lens-scan design parameters;

[0092] FIGS. 46A and 46B illustrate a transitionless scan method;

[0093] FIG. 47 illustrates a “wafer-scan” scanning method comprising a transitionless scan;

[0094] FIG. 48 illustrates the pixel layout design used in the wafer-scan method;

[0095] FIG. 49 shows a magnified portion of FIG. 48, illustrating the wafer-scan design parameters;

[0096] FIG. 50 summarizes the wafer-scan design parameters;

[0097] FIG. 51 illustrates a radial “point exposure profile” for a single, isolated image dot, and a cross-sectional “line exposure profile” of an isolated dot row;

[0098] FIG. 52 illustrates a dot pattern exposing a positive-tone disk image;

[0099] FIG. 53 illustrates a dot pattern exposing a negative-tone disk image;

[0100] FIG. 54 illustrates a positive-tone line image;

[0101] FIG. 55 illustrates a negative-tone line image;

[0102] FIG. 56 shows cross-sectional exposure intensity profiles for the disk images of FIGS. 52 and 53;

[0103] FIG. 57 shows cross-sectional exposure intensity profiles for the line images of FIGS. 54 and 55;

[0104] FIGS. 58A and 58B illustrate the printing effect of a defective pixel;

[0105] FIG. 59 illustrates the sensitivity of the image size to defocus of the first-stage microlens, for the positive- and negative-tone disk images of FIGS. 52, 53;

[0106] FIG. 60 illustrates the sensitivity of the image size to defocus of the second-stage microlens, for the positive- and negative-tone disk images of FIGS. 52, 53;

[0107] FIG. 61 is a cross-sectional view of an exposure test pattern that would be used to characterize second-stage lens centration errors, pixel defects, and alignment and stitching errors; and

[0108] FIG. 62 is a block diagram illustrating the control system's main architectural components and interactions.

DESCRIPTION OF SPECIFIC EMBODIMENTS

§1 System Overview

[0109] Following is a general overview of the system design concept, in a preferred embodiment. Later sections provide more detail on the system architecture, operational principles, methods of manufacture, and alternative embodiments.

[0110] FIG. 1 is a cross-sectional, schematic illustration of a printer pixel 101. The pixel comprises a first-stage microlens L₁, a second-stage microlens L₂, and a modulator 102, which may be a digitally controlled, micromechanical shutter 103a. The lenses are preferably composed of molybdenum. Although the lenses function as focusing elements, each lens has a meniscus form (i.e., it is thinner at the center than at the edge) because the refractive index of molybdenum is less than 1 at EUV wavelengths. A collimated incident EUV beam 104 is focused by lens L₁toward a point 105 at the center of lens L₂'s aperture, but the diffraction-limited focus spot (illustrated by the amplitude profile 106) actually overfills the lens aperture. (In this design the illumination source's geometric image at point 105 is much smaller than the diffraction-limited spot size; although the geometric image could dominate the diffraction limit in alternative embodiments.) Lens L₂further condenses the beam onto a point 107aon the surface of a wafer 108. The diffraction-limited, wafer-plane focused amplitude distribution is illustrated as profile 109. The respective clear aperture diameters D₁and D₂, and nominal focal lengths F₁and F₂of lenses L₁and L₂are D₁=5 μm, D₂=1 μm, F₁=171 μm, and F₂=3.25 μm. (More detail on the lens design is provided in §2.)

[0111] The “focusing” action of the microlenses is more analogous to that of a condenser lens than of an imaging lens, in that each microlens functions to condense incident illuminating radiation onto a target illumination spot. Unlike conventional imaging systems, the microlenses do not necessarily have well-defined focal planes, or if they do, they need not operate at exactly their design focal lengths. For example, the wafer plane is positioned slightly above lens L₂'s design focal plane in the present design because the diffraction-limited focus is not actually optimum at the design focal plane (as defined by geometric optics). Furthermore, lens L₁is not operated at its best (diffraction-limited) focus position because this would (according to optical simulations) significantly increase sensitivity to lens aberrations. (Tolerance issues are addressed in §11.)

[0112] The shutter 103ais shown in its open position in FIG. 1. In an alternative position 103bit blocks the radiation, switching off the pixel. The shutter is preferably located proximate to lens L₂in order to minimize its positional travel range. The travel range could, in principle, be reduced by an order of magnitude by positioning the shutter very close to the wafer, but in this design it is positioned above L₂so that it does not affect the working distance between the printhead and the wafer. (Alternative modulator mechanisms are discussed in §4.)

[0113] Lens L₂is preferably provided with a micromechanical actuator for fine-adjusting its lateral position. As illustrated in FIG. 1, focus point 107aon the wafer can be moved to proximate point 107bby moving the lens. The lens actuation mechanism is used to make nanometer-scale adjustments to maintain overlay alignment, and can also be used to counterbalance critical tolerance factors such as thermally induced magnification errors and asymmetry in lens L₂'s surface form.

[0114] The printer pixels are assembled into an array forming a printhead. FIG. 2 schematically illustrates a portion of a printhead 201 in cross-section. A silicon plate 202 supports the first-stage microlenses (such as lens L₁) on its top side and the modulators (such as modulator 102) and second-stage microlenses (such as lens L₂) on its bottom side. Tapered or terraced holes, such as hole 203, are etched through plate 202 to accommodate the converging EUV beam paths. The spaces between second-stage lens apertures provide accommodation for structural support and for the modulators' and second-stage microlenses' associated positional actuators and data paths (not shown).

[0115] The printhead optics could potentially be degraded by hydrocarbon outgassing from the wafer (Ref 25). This is particularly a concern for the second-stage lenses due to their close proximity to the wafer. Corrosion-resistant optical coatings may be used to mitigate degradation, but a more practical alternative may be to cover the bottom of each printhead with an EUV-transparent pellicle 204, such as a diamond membrane or a beryllium layer with a corrosion-resistant coating, which can be periodically cleaned or replaced. (The printheads could also function as vapor barriers to prevent hydrocarbon deterioration of the EUV illumination mirrors.) In contrast to pellicles used for projection lithography masks, the printhead pellicles need not be self-supporting over a large aperture area, and can hence be very thin and highly transmitting. For the present design, a pellicle transmission efficiency η_pellicleof 0.95 is assumed,

[0116] η_pellicle=0.95 Eq. 1.1

[0117] For example, a diamond film of thickness 7 nm, or a beryllium film of thickness 45 nm, would have 95% transmittance at an operating wavelength of 11.3 nm. A practical pellicle material could be beryllium with a protective coating similar to coatings used for EUV mirrors (Ref 26).

[0118] FIG. 3 is a plan view of printhead 201. The first-stage microlenses (e.g., lens L₁) are distributed in a triangular centering pattern across a square printhead aperture 301. The lens apertures are preferably circular for two reasons: First, for a given lens aperture area and focal length, the circular shape maximizes lens transmittance efficiency. Second, the circular symmetry minimizes the number of degrees of freedom that need to be controlled in the manufacturing process. Given that the apertures are circular, the triangular array layout illustrated in FIG. 3 is preferred because this optimizes the lens fill factor.

[0119] The center spacing d_lensbetween adjacent lens centers (FIG. 3) is approximately 6 μm,

[0120] d_lens≅6 μm Eq. 1.2

[0121] (This value is chosen to provide a 1-μm clearance between lens apertures.) The lens fill factor η_FF(defined as the fraction of the area of aperture 301 covered by the first-stage lens apertures) is approximately

\begin{matrix} η_{FF} = (\frac{π}{4} D_{1}^{2}) / (\frac{\sqrt{3}}{2} d_{lens}^{2}) ≅ (\frac{π}{4} {(5 µ m)}^{2}) / (\frac{\sqrt{3}}{2} {(6 µ m)}^{2}) = 0.630 & Eq.  1.3 \end{matrix}

[0122] The dimensions W_xand W_yof aperture 301 in the respective X and Y directions are both 20 mm,

[0123] W_x=20 mm Eq. 1.4

[0124] W_y=20 mm Eq. 1.5

[0125] The number of pixels per printhead, N_pixel/head, is approximately

\begin{matrix} N_{pixel / head} ≅ W_{x} W_{y} / (\frac{\sqrt{3}}{2} d_{lens}^{2}) ≅ {(20 mm)}^{2} / (\frac{\sqrt{3}}{2} {(6 µ m)}^{2}) = 1.283 \times 10^{7} & Eq.  1.6 \end{matrix}

[0126] The total number of pixels per printer system, N_pixel, is equal the number per printhead (N_pixel/head) multiplied by the number of printheads per wafer print module (45), and the number of modules per system (8)

[0127] N_pixel=N_pixel/head×45×8≅4.62×10⁹ Eq. 1.7

[0128] The EUV illumination covers a square illumination field 302 having dimensions W′_xand W′_yof 22 mm in the respective X and Y directions and overfilling aperture 301,

[0129] W′_x=22 mm Eq. 1.8

[0130] W′_y=22 mm Eq. 1.9

[0131] Thus the printheads have an aperture collection efficiency η_{head ap}of

\begin{matrix} η_{head ap} = \frac{W_{x} W_{y}}{W_{x}^{'} W_{y}^{'}} = {(\frac{20 mm}{22 mm})}^{2} = 0.826 & Eq.  1.10 \end{matrix}

[0132] Two position sensor arrays 303aand 303bcomprising position sensors (such as sensor element 304) border illumination field 302. The sensor elements could, for example, be confocal-imaging microlens arrays (Ref. 13) operating at a visible wavelength (e.g., 633 nm), or capacitance sensors, or a combination of such elements. The sensor arrays would detect focus height (e.g., by means of the confocal elements' depth response). Also, they would detect lateral alignment (X-Y position, rotation, and magnification) by sensing the printhead's lateral position relative to an alignment pattern such as a periodic tracking patterns 305aand 305bformed in the wafer scribe lines. (Tracking system options are discussed in §5.)

[0133] Multiple printheads are assembled into an array forming a wafer print module, which covers a 300-mm wafer. FIG. 4 schematically illustrates a portion of a wafer print module 401 in cross-section, and FIG. 5 is a plan view of module 401. Multiple printheads are attached to a structural framework 402 comprising a network of joist elements. For example, printhead 201 is supported by joists 403aand 403b(FIGS. 4, 5). The center spacing between the printheads is 2W_xin the X direction and 2W_yin the Y direction. Micropositioning actuators 404aand 404bbetween printhead 201 and joists 403aand 403braster-scan the printhead across the wafer during exposure. The scan range is approximately 250 μm. (The optics may be sealed off from wafer-generated hydrocarbon contaminants by means of bellows couplings or flexible seals between the printheads and their supporting joists.) Actuators 404aand 404bare controlled by position sensor feedback from sensor arrays 303aand 303b. In addition to performing the scanning function, the actuators control alignment, fine focus and tilt. Also, the second-stage lens actuators are used in conjunction with printhead actuators 404aand 404bto make fine adjustments in the focus spot positions (e.g., to correct magnification errors due to thermal expansion differences between the wafer and printhead).

[0134] As the printhead scans the wafer, each printer pixel follows a serpentine raster scan path and is modulated in synchronization with the pulsed LPP source to expose a rectangular array of image points. The scan paths of proximate pixels are interleaved so that isolated pixel defects have minimal influence on the printed image. The wafer 108 is exposed with four sequential raster scans. It is first scanned at position 501a(FIG. 5). The wafer is then stepped by a distance W_xin the X direction to position 501bwhere it is scanned a second time; it is stepped by a distance W_yto position 501cand scanned a third time; and is then stepped by a distance W_xin the negative X direction to position 501dand is scanned a fourth time. At each scan position, the wafer remains stationary while the printheads raster-scan the wafer.

[0135] The scanning method outlined above is termed a “lens-scan” technique. In an alternative “wafer-scan” method, the wafer moves during exposure while the printheads remain substantially stationary, except for slight positional corrections to maintain focus and alignment. (Scan methods are discussed in detail in §8.)

[0136] As mentioned above, for efficiency it is preferred that the system use EUV illumination that is collimated and of uniform irradiance, and which is efficiently partitioned among the illumination fields. A system of terraced fold mirrors is used to partition a large-aperture, collimated EUV beam into square illumination fields covering individual printhead apertures. For example, FIG. 4 illustrates a fold mirror 407bthat directs EUV illumination 104 onto illumination field 302 covering printhead aperture 301. Adjacent fold mirrors 407aand 407care positionally displaced from mirror 407bso that shadow zones 406aand 406bare not illuminated. Two such sets of fold mirrors operate in conjunction to subdivide the beam into a two-dimensional array of square illumination fields.

[0137] As illustrated in FIG. 6, the collimated beam is generated from a xenon LPP source 601 by two mirrors M1 and M2. Mirror M1 is a deep, aspheric, condenser element, and mirror M2 is a much larger but shallower, spherical collimator element. Two beam-shaping mirrors are provided to simultaneously satisfy the collimation and uniformity requirements. (The mirrors' reflectivity variation with incidence angle could affect beam uniformity, but the mirror geometry is designed to counterbalance this effect. Also, the design compensates for nonuniformity of the source's angular emission profile.) The collimated output from mirror M2 is distributed among eight wafer print modules and the system exposes eight 300-mm wafers in parallel.

§2 Microlens Design

[0138] Each microlens surface is designed as a “refractive conjugate surface” (RCS), whose geometry is defined relative to cylindrical coordinates {r,z} as illustrated in FIG. 7. An RCS 701 has axial symmetry about a z axis 702, and is designed to image a first axial conjugate point 703 onto a second axial conjugate point 704. For example, an incident ray 705, initially directed toward point 703, intercepts surface point 706 and is refracted toward point 704. The respective axial displacements of points 703 and 704 from the surface vertex point 707 (at z=z_vertex) are denoted z_f1and z_f2((In general, z_f1and z_f2can be positive, negative, or infinity.) If the z origin is at the surface vertex (i.e., z_vertex=0), then the {r,z} coordinates at any surface point 706 are related by the equation

\begin{matrix} η_{1} \frac{2 z - c_{f1} (r^{2} + z^{2})}{1 + \sqrt{1 - 2 c_{f1} z + c_{f1}^{2} (r^{2} + z^{2})}} = n_{2} \frac{2 z - c_{f2} (r^{2} + z^{2})}{1 + \sqrt{1 - 2 c_{f2} z + c_{f2}^{2} (r^{2} + z^{2})}} & Eq.  2.1 \end{matrix}

[0139] wherein c_f1and c_f2are the respective incident and refracted field curvatures at vertex 707,

\begin{matrix} c_{f1} = \frac{1}{z_{f1}} & Eq.  2.2 \\ c_{f2} = \frac{1}{z_{f2}} & Eq.  2.3 \end{matrix}

[0140] and n₁and n₂are the refractive indices on the incidence and transmitted sides of surface 701. (For weakly-absorbing media such as molybdenum n₁and n₂are the real parts of the complex refractive indices.) An approximate paraxial representation of the surface form is

\begin{matrix} z ≅ \frac{1}{2} c_{S} r^{2} + \frac{1}{8} c_{S}^{3} (1 + k_{S}) r^{4} & Eq.  2.4 \end{matrix}

[0141] wherein c_Sis the surface curvature and k_Sis the surface's conic constant,

\begin{matrix} c_{S} = \frac{n_{2} c_{f2} - n_{1} c_{f1}}{n_{2} - n_{1}} & Eq.  2.5 \\ k_{S} = \frac{n_{1} {n_{2} (c_{f2} - c_{f1})}^{2} (n_{2} c_{f1} - n_{1} c_{f2})}{{(n_{2} c_{f2} - n_{1} c_{f1})}^{3}} & Eq.  2.6 \end{matrix}

[0142] (If z_vertex≠0, the substitution z←z−z_vertexshould be made in Eqs. 2.1 and 2.4.)

[0143] The microlens design is determined not only by imaging requirements, but also by surface tolerance considerations. FIG. 8 illustrates the relationship between positional error of a refracting surface 801aand the resultant optical phase error of a transmitted electromagnetic wave 802a. If surface 801ais translationally displaced by an incremental distance s to position 801b, the plane wave 802awill be positionally shifted by a corresponding distance d to position 802b. The corresponding phase shift φ in the transmitted field is

\begin{matrix} φ = 2 π \frac{s}{λ} (n_{2} \cos θ_{2} - n_{1} \cos θ_{1}) & Eq.  2.7 \end{matrix}

[0144] wherein λ is the wavelength (in vacuum), n₁and n₂are the refractive indices on the respective incidence and transmitted sides of surface 801a, θ₁is the incident angle, and θ₂is the refracted angle. The right-hand term in the above expression defines a “surface tolerance factor” (stf),

[0145] stf=n₂cos θ₂−n₁cos θ₁ Eq. 2.8

[0146] which should be minimized in the lens design. For a refractive conjugate surface, the stf at point {r,z} (i.e., point 706 in FIG. 7) can be represented by the following paraxial approximation,

\begin{matrix} stf ≅ (n_{2} - n_{1}) (1 + \frac{1}{2} θ_{1} θ_{2}) ≅ (n_{2} - n_{1}) (1 + \frac{1}{2} r^{2} (c_{S} - c_{f1}) (c_{S} - c_{f2})) & Eq.  2.9 \end{matrix}

[0147] Generally, the stf distribution across a lens aperture takes on its worst-case value at the aperture edge, where the incident and refracted angles θ₁and θ₂are maximized.

[0148] The design operating wavelength λ and complex refractive index ñ_Mofor the molybdenum lenses L₁and L₂are

[0149] λ=11.3 nm Eq. 2.10

[0150] ñ_Mo=n_Mo−ik_Mo;

[0151] n_Mo=0.9516, Eq 2.11

[0152] k_Mo=0.00464

[0153] FIG. 9 shows a cross-section of lens L₁. The graph 901 illustrates the transmitted electric field's amplitude profile versus aperture radius position r₁, assuming a unit-amplitude incident beam. (The transmitted field intensity is proportional to the amplitude squared.) The lens's clear aperture diameter D₁is 5 μm, and its center thickness t_c1is 20 nm. The top surface 902 is planar, and the bottom surface 903 is designed to focus a collimated incident beam to a point 171 μm (the nominal focal length, F₁) below the top surface. (Depending on the manufacturing method, it may be more practical to make the bottom surface planar and the top surface curved; or both surfaces could be curved.)

[0154] FIG. 10 shows a cross-section of lens L₂, and also illustrates its incident field amplitude profile 106 (cf. FIG. 1) and transmitted field amplitude profile 1001 versus aperture radius position r₂. Lens L₂'s top surface vertex 1002 is located at a distance of 171 μm (the nominal focal length F₁) below lens L₁, and profile 106 represents the diffraction-limited focus spot produced by lens L₁. Lens L₂'s clear aperture diameter D₂is 1 μm and its center thickness t_c2is 20 nm. The top surface 1003 is designed to image the center of lens L₁to an intermediate conjugate point (not shown) 6.31 μm below vertex 1002, and the bottom surface 1004 images this intermediate conjugate onto a focal plane 3.25 μm (the nominal focal length F₂) below vertex 1002. This design prescription balances the optical power between surfaces 1003 and 1004 so that the stf (Eq. 2.8) of geometric edge ray 1005 is the same at points 1006 and 1007 where the ray intercepts the surfaces (i.e., the refraction angles at the two surfaces are matched). Thus, lens L₂'s worst-case stf is minimized.

[0155] (Actually, the definition of “worst case” may be over-conservative in this design because the transmittance of edge ray 1005 is essentially zero, as evidenced by transmission profile 1001.)

[0156] FIG. 11 illustrates the focused beam's amplitude profile 109 at the wafer plane versus the beam radius position r₃(cf. FIG. 1). This represents the diffraction-limited focus spot produced by lens L₂. FIG. 12 shows the peak-normalized intensity profile I[r₃] (which is proportional to the amplitude squared). This profile can be closely approximated as a Gaussian,

[0157] I[r₃]≅exp[−(r₃/34.5 nm)²] Eq 2.12

[0158] For comparison, FIG. 12 also shows a classical Airy-disk profile I′[r₃] that would be produced by an ideal, unapodized, circular-aperture lens of numerical aperture NA=0.1. The functional form of this profile is

\begin{matrix} I^{'} [r_{3}] = {(\frac{J_{1} [2 π NA r_{3} / λ]}{π NA r_{3} / λ})}^{2} & Eq.  2.13 \end{matrix}

[0159] wherein J₁is a Bessel function.

[0160] The lens design parameters are selected to balance a three-way tradeoff between optical resolution, efficiency, and practicality. Circular lens apertures are preferred because the circular shape maximizes EUV transmittance for a given aperture area and focal length, and the circular symmetry minimizes the number of degrees of freedom that need to be controlled in the manufacturing process. The aperture diameters D₁=5 μm and D₂=1 μm are selected as an estimated best compromise between optical efficiency and manufacturability. (For a specified optical resolution, smaller lenses would have less absorption loss because the edge thickness would be reduced, but they would also be more difficult to manufacture.) Given the diameter specification, the lenses' remaining design parameters are constrained to maintain a specified focus spot resolution at the wafer plane, and are optimized (subject to the resolution constraint) to maximize optical throughput. The resolution requirement is specified in terms of an “effective numerical aperture” parameter, as described below.

[0161] A lithography system's printing resolution (RES) can be roughly estimated by the equation

[0162] RES=k₁λ/NA Eq. 2.14

[0163] wherein λ is the exposure wavelength, NA is the illumination numerical aperture at the wafer, and k₁is a process-dependent constant that is typically close to 0.6. The geometric numerical aperture NA_geoof lens L₂, as determined by a geometric raytrace of edge ray 1005 (FIG. 10), is NA_geo=0.156. This value actually has little relevance to the lens's resolution performance because the lens is essentially opaque near its periphery. (In practice, the lens need not even be provided with an aperture stop.) However, an “effective numerical aperture” NA_effcan be defined to provide a common basis of comparison with projection lithography systems. NA_effis defined as the numerical aperture of an ideal, unapodized lens that would provide imaging resolution comparable to that of lens L₂, as defined by a criterion such as the focus spot's full-width-at-half-maximum (FWHM). As illustrated in FIG. 12, lens L₂'s focused intensity profile I[r₃] closely matches a classical Airy-disk profile I′[r₃] corresponding to a 0.1-NA lens, so the system's effective numerical aperture is NA_eff=0.1.

[0164] The effective numerical aperture can be used in the above resolution equation (Eq. 2.14) to roughly compare the microlens system to projection lithography systems. But this comparison is limited because unlike projection systems, the microlens system exposes proximate image points sequentially—not simultaneously—and the points therefore do not interact coherently. The printed image would thus be devoid of coherent proximity effects that are characteristic of projection lithography systems.

[0165] The combined transmission efficiency of lenses L₁and L₂(i.e., ratio of the wafer-plane energy to the incident energy intercepted by lens L₁'s clear aperture, neglecting any pellicle or modulator transmission losses) is 17.5%. This value is based on a diffraction simulation (with the conventional approximations of Fresnel-Kirchoff diffraction theory), assuming that an ideal monochromatic, collimated beam is incident on lens L₁. However, the incident beam produced by the EUV illumination system actually has a finite spectral bandwidth of approximately 0.3 nm and an angular divergence full angle of approximately 1 mrad. The bandwidth is insignificant, but the source divergence causes some spreading of the diffraction-limited source image on lens L₂'s aperture, resulting in some energy loss through L₂. Assuming (conservatively) that the plasma source can be modeled as a spherical, Lambertian (i.e., uniform-radiance) emitter subtending 1 mrad at the microlens L₁aperture plane, an efficiency attenuation factor of 0.963 is required to account for the source size. Thus the actual lens efficiency t_lensis

[0166] η_lens=0.175×0.963=0.169 Eq. 2.15

[0167] If the source divergence were significantly greater than 1 mrad the attenuation factor would be much lower; for example, at 2 mrad it would be reduced from 0.963 to 0.880. (But although the source size affects efficiency, it does not limit imaging resolution because the source is not imaged onto the wafer.)

[0168] FIG. 13 is a tabular summary of the lens design parameters. The surfaces are numbered in order of traversal by the EUV beam, and the “surface” column also indicates the surface reference numbers in FIGS. 1, 9, and 10. The “media” column indicates the optical media above and below each surface (e.g., “Vac/Mo” means vacuum above and molybdenum below). The “diameter” column indicates the aperture stop diameters. (As noted above, lens L₂does not actually require an aperture stop.) The “z_vertex” column indicates each surface vertex's position along the optical axis, as measured from the first surface, and the “z_f1” and “z_f2” columns indicate the positions of each surface's conjugate points relative to the surface vertex (cf. FIG. 7).

[0169] Estimated surface form tolerances are 4 nm RMS for L₁and 2 nm RMS for L₂. In addition, if a calibrated lateral position offset is not applied to L₂, the odd-symmetric component of L₂'s surface error (i.e., tilt and coma) would need to be limited to approximately 0.75 nm RMS. (Design tolerance issues are discussed in §11.)

[0170] The above design is not very well optimized for focus because the diffraction-limited focus of a finite-conjugate lens is not actually optimum at the nominal focal plane (as defined by geometric optics). For example, system efficiency could (according to simulations) be improved somewhat by moving L₂slightly closer to L₁. This would improve optical efficiency, but it would also increase image sensitivity to lens aberrations, so L₁'s operating focal plane is kept at its design position in the current design. However, L₂should operate at optimum focus because its focus range must include positioning tolerances and the wafer flatness tolerance. Therefore the operating wafer plane would be shifted slightly (by −134 μm) from its design position, as indicated by the footnote in FIG. 13. (This shift is based on simulated printing performance, which is discussed in §10.) Other design improvements can also be made. For example, the RCS design geometry (as defined by FIG. 7) could be varied to optimize optical performance. The above design is based on optimization of the point spread function (I[r₃], FIG. 12), using approximations of Fresnel-Kirchoff diffraction theory; whereas a more refined design would be based on simulated printing performance, using more accurate electromagnetic simulations. Also, the design procedure should take into account the image sensitivity to lens aberrations and should properly balance any tradeoffs between optical efficiency and aberration sensitivity.

[0171] The above design configuration represents just one of a variety of possible design approaches. One alternative would be to design lens L₁as a phase Fresnel lens (Ref. 27) in order to minimize the lens thickness and absorption losses. (Fresnel zone plate lenses, such as those described in Refs. 16-18, could possibly also be used, but phase Fresnel lenses would have better optical efficiency.) FIGS. 14A and 14B illustrate cross-sections of two such designs, a two-facet design comprising facets 1401 and 1402 (FIG. 14A), and a three-facet design comprising facets 1403, 1404, and 1405 (FIG. 14B). The facets on each surface can be designed as refractive conjugate surfaces with common conjugate points, and the step height h between adjacent facets is selected so that the transmitted field's phase discontinuity across each step is 2π at the design wavelength. (For a design wavelength of 11.3 nm, the step height on a molybdenum lens would be approximately 233 nm.) These designs could make it possible to use large (e.g., >10 μm diameter) or higher power (NA>0.02) first-stage lenses without incurring unacceptable absorption losses. (However, as the lens size or power are increased the resolution-limiting effects of the source size and spectral bandwidth become more significant, so even without absorption losses there would still be a practical limit on the maximum lens size or optical power.)

[0172] Phase Fresnel lenses would be less practical for the second-stage lens L₂due to its small size, but L₂could possibly be constructed by combining a number of comparatively low-power elements to form a composite high-power lens. An L₂design of this type is illustrated in FIG. 15. The lens is a triplet comprising a molybdenum top element 1501, a beryllium middle element 1502, and a molybdenum bottom element 1503. (A method for manufacturing this type of lens is discussed in §3.) In contrast to molybdenum, the real part of beryllium's complex refractive index is greater than 1 at 11.3 nm, so a convex beryllium lens will exhibit positive optical power. (The middle element could alternatively be a vacuum space.) The triplet's optical power should preferably be balanced between the three elements so that the stf (Eq. 2.8) is matched at the four points 1504, 1505, 1506, and 1507 where geometric edge ray 1508 intercepts the four surfaces.

[0173] Higher optical resolution could be achieved by using more steeply curved optical surfaces or more elements in the second-stage lens, although the resulting resolution improvement would come at the expense of reduced optical efficiency. Acceptable efficiency could be regained by using smaller optical components, but extreme miniaturization of the optics could result in an unmanufacturable design. On the other hand, it may be possible to circumvent this difficulty by adding additional microlens stages. For example, FIG. 16 illustrates a three-stage pixel configuration comprising a first-stage lens L₁, a second-stage lens L′₂, a third-stage lens L′₃, and a modulator 102. (Lenses L′₁and L′₂in this design are similar to lens L₁in FIG. 1, and lens L′₃is similar to the triplet L₂design in FIG. 15.) Lens L′₁focuses incident EUV illumination 104 toward a point 1601 centered in the lens L′₂aperture; lens L′₂focuses the illumination toward point 105 at the center of the L′₃aperture; and lens L′₃then focuses the illumination onto point 107aon the wafer 108 (cf. FIG. 1). (The L′₂and L′₃apertures may be overfilled due to diffraction spreading of the focus beams, or due to the illumination source's geometric image size.) Lens L′₃could be a very small, high-power element (e.g., NA_eff>0.2), while lens L′₁could still be of practical size (e.g., aperture diameter>5 μm).

[0174] Molybdenum is just one of several materials that could be used for the microlenses. Several alternative materials that have good EUV optical characteristics are niobium (Nb), ruthenium (Ru), rhodium (Rh), and yttrium (Y). Material selection criteria that apply to multilayer EUV mirrors (e.g., refractive power and absorption—Ref. 28) are also relevant to EUV microlenses. Additionally, the microlens surfaces may be coated with a thin (e.g., several nanometers thick) corrosion-resistant layer similar to passivating coatings used for EUV mirrors (Ref. 26). These material selection considerations apply primarily to EWV lithography. For operation at other wavelengths, such as DUV or visible light, more conventional optical materials such as silicon dioxide or silicon nitride could be used.

[0175] Each second-stage microlens L₂is preferably provided with a micromechanical actuator for making nanometer-scale adjustments in its lateral translational position. A possible design for the actuator is illustrated conceptually in FIG. 17. Lens L₂is housed in an inner silicon frame 1701 that is attached, via contiguous leaf springs 1702aand 1702b, to an outer silicon frame 1703. The outer frame is attached, via contiguous leaf springs 1702cand 1702d, to anchor points 1704aand 1704b. The frames and leaf springs are made from a thin silicon layer that has been etched through over the hatched areas in FIG. 17. The lens position is actuated by means of electrostatic attraction or repulsion between proximate capacitive conductor pads 1705, 1706 and 1707, which are charged by three voltage controls (indicated as V₁, V₂, and V₃). The actuator's positional range would typically be on the order of 10 nm.

[0176] It could also be possible to equip each second-stage microlens with an individually controllable focus mechanism, making possible conformal printing on complex surface topographies. Due to the system's large depth of focus, a focus mechanism would probably not provide much benefit unless its positioning range is over an order of magnitude greater than the lateral positioning range. However, focus control could be useful for ultra-high-resolution systems operating at high numerical aperture (e.g., NA_eff>0.2).

§3 Microlens Manufacture

[0177] EUV microlenses can be manufactured with nanometer-scale profile control by adapting shadow-mask deposition processes that are used, for example, in shadow-mask molecular beam epitaxy (MBE). Large-scale shadow masks have been used to deposit graded-thickness, multilayer EUV mirror coatings with subnanometer thickness accuracy (Ref. 29); and shadow-mask deposition processes have also been applied to microlens manufacture (Ref. 30). Microlens-type structures with submicron dimensions can be fabricated using shadow-mask MBE techniques (Refs. 21, 22). Some very excellent recent examples of shadow-mask process capabilities for submicron fabrication are exhibited in Refs. 23, 24.

[0178] The above methods are all deposition processes, but shadow masks can also be used with etching processes. For example, Ref. 31 describes a reactive ion etching (RIE) process that achieves nanometer-scale thickness accuracy on three-dimensional topographies by using a movable shadow mask. Molybdenum can be reactive-ion etched (Refs. 32, 33), and accurately profiled molybdenum microlenses could be fabricated by a combination of shadow-mask etching and deposition processes.

[0179] FIGS. 18A-18F illustrate a method for fabricating three-element second-stage lenses L₂of the type illustrated in FIG. 15. (The single-element L₂design illustrated in FIG. 10 could be similarly formed by leaving out the process steps shown in FIGS. 18C-18E.) In the first process step (FIG. 18A), a substrate 1801 is etched by a particle beam 1802 which is masked by a shadow mask 1803 comprising an array of circular apertures such as aperture 1804 (shown cross-sectionally). Either the substrate or the mask is scanned in a circular pattern during etching to form an array of rotationally symmetric convex mold surfaces such as surface 1805. (The substrate should preferably be stepped during the etch process so that each mold surface is etched through multiple mask apertures, in order to mitigate the effects of nonuniformities in either the apertures or the particle beam. A similar stepping technique could be used in subsequent processing steps, FIGS. 18B-18E.) Next (FIG. 18B), vapor-phase molybdenum 1806 is shadow-deposited onto the substrate, again using a circular scan so that rotationally symmetric, meniscus molybdenum structures (such as structure 1807) are formed over the molds. A beryllium layer 1808 is then deposited through the shadow mask (FIG. 18C); and the beryllium and molybdenum are subsequently shadow-etched to form convex beryllium lens elements (such as element 1809) over the meniscus structures (FIG. 18D). Then another molybdenum layer is shadow-deposited to form a second array of meniscus structures (such as structure 1810) over the beryllium elements (FIG. 18E). In subsequent process steps, the lens elements are lithographically isolated to form separate lens structures, such as lens L₂(FIG. 18F; cf. FIG. 15); and the substrate 1801 is back-etched to expose the lenses and to form a positional actuator mechanism (such as that illustrated in FIG. 17).

[0180] An advantage of the above process is that all of the critical lens deposition steps are performed with a single shadow mask, so accurate alignment is maintained between process steps. Moreover, the shadow mask aperture diameter can be significantly larger than the microlens clear aperture diameter. Thus, the shadow mask apertures could easily be formed by using an interference lithography technique, which is very accurate and more economical than alternative processes such as electron-beam or focused-ion-beam lithography.

[0181] The mask aperture array would be designed to match the triangular microlens centering layout (FIG. 3). A lens centering pattern comprising equilateral triangles can be produced by interference lithography, using a three-beam exposure technique in which three collimated laser beams are directed onto a photoresist-coated substrate at equal incidence angles and symmetrically distributed azimuthal angles. (A similar three-beam interference method has been used to lithographically pattern microlens arrays, Ref. 34. The three-beam interference technique is much simpler than alternative four-beam interference methods or image-reversal processes that are used for patterning rectangular hole arrays, Ref 35.) Each beam should preferably be polarized in its plane of incidence. Provided that the incidence angles are sufficiently large, the resulting interference pattern will contain localized intensity maxima that can be developed into etched holes with a high degree of circular symmetry and accurate center placement. (Symmetric intensity minima can similarly be generated if the beams are polarized parallel to the substrate.)

[0182] The shadow-mask lithography technique described above is not the only option for microlens fabrication. Refs. 36 and 37 review a variety of alternative methods such as gray-scale lithography. For non-EUV applications, any of a variety of lens forms such as molded microlenses, gradient-refractive-index planar microlenses, melted-resin microlens arrays, etc., may be used. One technique that might be suitable for EUV application would be laser-assisted deposition and etching (Ref. 36, Sect. 4.4.3) in combination with interference lithography (i.e., a stationary laser interference pattern would be used instead of a scanning laser beam).

[0183] In a variation of the interference lithography method, NIST has developed an “atom lithography” process (Ref. 38) that forms extremely accurate and uniform periodic structures by using intersecting laser beams to focus an atomic beam onto interference minima or maxima on a deposition substrate. The method could be used with the three-beam geometry described above to form EUV microlens arrays with triangular centering patterns. A lens material such as molybdenum could perhaps be deposited directly. Alternatively, deposited layers of a material such as chromium could be used as molds or as sacrificial, pattern-transfer layers to form the microlenses.

§4 Modulator Options

[0184] The modulator 102 (FIG. 1) is preferably a micromechanical shutter 103a, which is located proximate lens L₂in order to minimize its positional travel range. It is preferably located between L₁and L₂so that it does not affect the working distance between the printhead and the wafer. The shutter could, for example, comprise a comb-drive actuator similar to a mechanism that has been under development for printing applications (Ref 39). Another type of shutter mechanism which has been proposed for use in maskless x-ray lithography is a stress-biased, electrostatically activated coil or flap (Refs. 40, 18).

[0185] Other types of shutters could also be used. For example, FIG. 19 shows a sectional view of one such mechanism. The second stage microlenses, such as lens L₂, are supported by a planar structure 1901 which is attached by support posts (such as post 1902) to a silicon plate 202. (Structure 1901 may comprise a lens position actuator such as that illustrated in FIG. 17.) EUV radiation is focused through etched holes in plate 202, such as hole 203 (cf. FIG. 2). The radiation illuminating each lens L₂is shuttered by means of a movable aperture plate, such as plate 1903, which is attached to plate 202 by means of vertical leaf springs such as leaf spring 1904. (The aperture plate and leaf springs could be formed as a film deposited onto a sacrificial mold.) The aperture plates and leaf springs are conductive (or are coated with a conductive film) and are positionally actuated by means of electrostatic interaction with capacitive conductor pads such as pad 1905 deposited on plate 202.

[0186] Shutter mechanisms are generally binary-state (on/off) devices, which would not provide any gray scale modulation capability. Although a shutter proximate to L₂could be positioned to only partially obstruct the illumination on L₂, this would significantly degrade the focused beam quality and spot size on the wafer. However, the focused beam intensity could be effectively modulated by inserting a micromechanically-actuated obscuration in the beam path close to the first-stage lens L₁. Due to the spatial filtering action of L₂, the optical diffraction and scatter caused by the obscuration would not significantly affect the wafer-plane focus spot other than to modulate its intensity level.

[0187] Modulator mechanisms other than shutters could also be used. For example, FIGS. 20A and 20B illustrate a pixel 101 comprising microlenses L₁and L₂and modulator 102, wherein the modulator comprises two proximate phase diffraction gratings G₁and G₂. In the modulator's ON state (FIG. 20A), the gratings have minimal effect on the transmitted beam. The incident beam 104 is focused by L₁toward a point 105 centered on L₂; and L₂thence focuses the radiation toward focal point 107aon the wafer 108. (The radiation field amplitudes on L₂and on the wafer are illustrated as profiles 106 and 109, respectively; cf. FIG. 1.) In the modulator's OFF state (FIG. 20B), grating G₁is translationally displaced, causing the transmitted energy to be diverted into diffracted orders such as +1st and −1st orders directed toward points 2001aand 2001b. (The diffracted field amplitudes at these points are illustrated as profiles 2002aand 2002b.) The zero order at point 105 is extinguished in the OFF state. The grating modulator's operational principle is similar to that of a laser modulator described in Ref 41, and several design variations for EUV application (including a design with gray-scale modulation capability) are described in the above-referenced provisional application No. 60/162,684.

[0188] A significant drawback of grating modulators for EUV application is their high efficiency loss. For an operating wavelength of 11.3 nm, and assuming molybdenum grating lines, a grating modulator's maximum theoretically attainable ON-state transmittance is 75.7%. In practice, the efficiency would need to be much lower (e.g., 50%) to accommodate tolerance factors. Due to this limitation other micromechanical modulator mechanisms such as shutters may be more practical, although grating modulators could be useful for non-EUV applications.

[0189] For some specialized applications, multiple modulators would not be required; instead the system could simply use a single modulator at the EUV source. (For example, the LPP source's laser could be modulated by means of a grating modulator.) The printer pixels would all print identical exposure patterns, resulting in a periodic exposure distribution matching the microlens periodicity. This type of system could be used, for example, in the manufacture of porous membrane filters or periodic structures such as mass-replicated MEMS devices. For applications where individually addressable pixel modulators are required, the printer modulators could themselves be fabricated by using a microlens lithography system that is only source-modulated. A “mix-and-match” strategy could also be used in which high-resolution, periodic structures are fabricated using a source-modulated microlens lithography system, while e-beam or DUV lithography is used for other non-periodic structures.

§5 Focus and Alignment Tracking System

[0190] Positional alignment tolerances for EUV lithography are on the order of a few nanometers; thus the printheads must be equipped with very precise and accurate alignment sensors and actuators. (Focus tolerances would typically be over an order of magnitude looser than alignment tolerances, but would be more critical for very high-resolution systems.) The preferred alignment method uses an array of micro-sensors, such as electromagnetic proximity probes or optical sensors, which are formed integrally with the printheads and which sense each printhead's position relative to periodic tracking patterns on the wafer surface. For example, FIG. 3 illustrates a printhead 201 equipped with two position sensor arrays 303aand 303bbordering illumination field 302 and comprising position sensor elements such as element 304. A physical (e.g., electrical or optical) interaction between the sensor elements and periodic tracking patterns 305aand 305bon the wafer creates a positional feedback signal such as a Moire-type pattern in the distribution of signal levels over the sensor arrays. This pattern is digitally analyzed to determine positional information which is relayed to positioning actuators (such as printhead actuators 404aand 404bin FIG. 4) to maintain accurate lateral alignment and focus while the system is printing. Also, each second-stage microlens L₂may be individually actuated to provide further control of alignment and magnification (FIG. 1).

[0191] Ref. 13 describes a confocal, microlens-imaging system that could be used for alignment tracking, and provisional application No. 60/162,684 describes an enhancement of this method that provides phase-contrast imaging capability. Ref. 13 also describes an actuator mechanism that provides warp compensation capability, as well as alignment, fine focus and tilt control. (Warp distortion might also be controlled by means of an electrically activated film, such as a piezoelectric layer, which is deposited on the printhead surface and is activated to induce a compensating shear stress across the surface.)

[0192] Focus and alignment could alternatively be tracked by means of capacitance sensors similar to fringe capacitance sensors described in Ref. 42. Such sensors are generally designed only to provide focus sensing capability, but a sensor similar to the interdigitated comb capacitor illustrated in FIG. 7 of Ref. 42 could also be sensitive to lateral alignment relative to a periodic tracking pattern on the wafer. Other sensor technologies that could be applicable to focus and alignment include proximal probe arrays and near-field optical sensors.

6 EUV Illumination System

[0193] FIGS. 21-26 show several views of the EUV illumination system (cf. FIGS. 4 and 6). The system's main elements are a xenon LPP source (Refs. 7, 8), collection optics comprising condenser mirror MI and collimator mirror M2, and distribution optics comprising two sets of terraced fold mirrors, M3 and M4. The system's primary design objectives are to provide collimated EUV illumination of uniform irradiance, optimize optical collection efficiency, and provide sufficient clearance space around the plasma source.

[0194] Referring to FIG. 21, the plasma source is centered at the focal point 2101 of condenser mirror M1. (The plasma source is indicated as 601 in FIG. 6.) Mirrors MI and M2 are axially symmetric with respect to an axis 2102 containing point 2101. (M1 is aspheric, whereas M2 may be spherical.) The mirrors are designed to produce illumination that is collimated and of uniform irradiance, taking into account the mirrors' reflectivity variation with incidence angle and the plasma's directional emission profile. The plasma is driven by a pulsed laser beam (or clustered collection of laser beams) 2103 which enters the system along axis 2102 and is focused onto point 2101. (The EUV radiation may optionally be modulated by means of a modulator mechanism 2107 in the laser beam path.) The beam's axial alignment results in an ETV plasma emission profile that is axially symmetric around axis 2102, making it possible to compensate for the emission's directional nonuniformity with an axially symmetric mirror design.

[0195] The collimated EUV beam 2104 from M2 is partitioned into 56 smaller, rectangular-aperture beams by a group of fold mirrors, which are collectively designated M3. The M3 group comprises eight subgroups M3.1, . . . M3.8. FIG. 21 shows one such subgroup, M3.3, and FIG. 22 shows the aperture geometry of M2 and of all eight M3 subgroups (M3.1, . . . M3.8) from a perspective looking into collimated beam 2104 (View A—A in FIG. 21). (Mirror M2 is illustrated as having an annular clear aperture in FIG. 22, but in practice M2 could comprise eight separate mirrors, each covering one of the eight M3 subgroups. M1 could possibly also be partitioned into separate mirrors.) The azimuthal angular displacement between successive M3 subgroups is 41.25° (FIG. 22), and clearance space is provided between M3.8 and M3.1 to accommodate a xenon injection nozzle and associated mechanics (not shown).

[0196] Each M3 subgroup comprises seven individual fold mirrors. For example, FIG. 21 illustrates subgroup M3.3 comprising fold mirrors M3.3.1 . . . M3.3.7, and FIG. 23 shows dimensional details of these mirrors in the projected view of FIG. 22. Collimated beam 2104 is incident on the M3.3 mirrors at a 75° incidence angle (FIG. 21), and is reflected into seven rectangular-aperture collimated beams 2105. FIG. 24 illustrates the aperture geometry of mirrors M3.3 and beams 2105 from a perspective looking into beams 2105 (View B—B in FIG. 21). The M3.3 mirror dimensions and positions are defined so that the beam width and spacing dimensions are 22 mm and 18 mm, respectively, in the projected view of FIG. 24.

[0197] The beams reflected from the M3 mirrors are further partitioned by a second group of fold mirrors, which are collectively designated as M4. The M4 group comprises eight subgroups M4 . . . M4.8 which are associated with corresponding M3 subgroups M3.1, . . . M3.8. For example, FIG. 21 illustrates the M4 subgroup M4.3 corresponding to M3.3. Subgroup M4.3 is itself a hierarchy comprising seven lower-level subgroups M4.3 . . . M4.3.7, each of which receives illumination from a respective M3 mirror (M3.3.1 . . . or M3.3.7) and further partitions it into individual square illumination fields. FIG. 25 shows an edge view of the M4.3 mirrors (View C—C in FIG. 21). Each subgroup M4.i.j(i=1, . . . 8; j=1, . . . 7) comprises seven individual fold mirrors M4.i.j.1, . . . M4.i.j.7 (FIG. 21); except that M4.i.1 and M4.i.7 each comprises only five mirrors (designated M4.i.j.2, . . . M4.i.j.6; j=1 or 7).

[0198] The seven rectangular-aperture beams 2105 are incident on the M4.3 mirrors at a 75° incidence angle (FIG. 25), and are reflected into 45 square-aperture beams 2501. FIG. 26 illustrates the aperture geometry of mirrors M4.3 and beams 2501 from a perspective looking into beams 2501 (View D—D in FIG. 25). Each of the 45 beams 2501 covers an individual square illumination field such as field 302 in FIGS. 5, 26. (FIG. 5 is a top view of the printhead and wafer, whereas FIG. 26 shows the illumination fields from a perspective looking up through the bottom of the wafer. In FIG. 4, the incident beam 104 illuminating printhead aperture 301 corresponds to one of the beams 2501.) Each illumination field in FIG. 26 is indexed with row and column indices j,k=1, . . . 7 identifying the corresponding fold mirrors M3.3.jand M4.3.j.kthat define the field. For example, field 302 is positioned at row j=2 and column k=4. (In FIG. 24 the mirrors M3.3.jare labeled with corresponding row indices j; and in FIG. 4 mirrors 407a, 407b, and 407ccorrespond respectively to M4.3.2.3, M4.3.2.4, and M4.3.2.5.) The M4.3 mirror dimensions and positions are defined so that the illumination fields are column-aligned with column width and spacing dimensions of 22 mm and 18 mm, respectively, in the projected view of FIG. 26.

[0199] Each M4.imirror subgroup (i=1, . . . 8) comprises 45 fold mirrors, so the total number of M4 mirrors is 8×45=360. More specifically, this is the number of M4 mirror clear apertures corresponding to square illumination fields. In practice, the actual number of M4 mirrors could be less because multiple clear apertures can share a common substrate. For example, in the design example illustrated in FIGS. 21-26, the elements of each set of three mirrors M4.i.3.k, M4.i.4.k, and M4.i.5.k(i=1, . . . or 8; k=1, . . . or 7) are coplaner and can hence be formed on a single, planar substrate.

[0200] The partitioned beams 2501 illuminate a wafer print module 401 which prints onto a 300-mm wafer 108 (FIGS. 21, 25; cf. FIGS. 4, 5). The illumination system supplies EUV illumination to eight such modules, and eight wafers are exposed simultaneously. The collection optics' axis 2102 (FIG. 21) is vertical, and the wafer planes are inclined at a 41.4° angle to horizontal. (The system could alternatively be designed so that the wafer planes are all horizontal, although such a design would incur practical and performance-related compromises.)

[0201] The fold mirror design illustrated above exemplifies a general design methodology that can be applied with different design assumptions and requirements. For example, the size and shape of the illumination fields, or the number of fields, may differ from the configuration illustrated in FIG. 26. The general design methodology comprises the following steps:

[0202] (1) The illumination aperture layout is defined, as in FIG. 26.

[0203] (2) The aperture shapes are translated along the direction parallel to the M4 incidence planes (the “k” direction in FIG. 26) so that the apertures in each row are consolidated into a single beam aperture, as illustrated in FIG. 24. (The translational position of each consolidated aperture provides a degree of freedom that is used to optimize optical efficiency.)

[0204] (3) The consolidated row apertures are then translated in the direction parallel to the M3 incidence planes (the “j” direction in FIG. 24), consolidating them into a single, contiguous beam aperture, as illustrated in FIG. 23.

[0205] (4) Multiple copies of the aperture shapes defined in step 3 are arranged to efficiently fill the M2 clear aperture space, as illustrated in FIG. 22. The translational positions of the consolidated apertures in step 2, the number of wafer print modules, and other design parameters are adjusted to fill the M2 aperture as efficiently as possible, and to also satisfy other design requirements such as clearance constraints.

[0206] The illumination optics must be designed to provide adequate clearance space around the plasma source. FIG. 27 illustrates a projection of the EUV illumination system's clear aperture limits onto mirror M1's focal plane 2106 (FIG. 21). The M2 mirror's inner and outer annular clear aperture limits are indicated as 2701 and 2702, respectively; and the square illumination fields' projections back onto plane 2106 are also illustrated (2703). The space between mirror groups M3.8 and M3.1 (cf. FIG. 22) provides accommodation for a xenon injection nozzle 2704, which feeds the plasma 601. The inner clearance circle 2701 defined by mirror M2's inner clear aperture limit has a diameter of 100 mm.

[0207] Although M1 and M2 are designed to produce a collimated beam of uniform irradiance, the beam's polarization state will not be uniform; hence any polarization sensitivity in M3 and M4 could significantly degrade the intensity uniformity. An advantage of the fold mirror design described above is that the combined polarization sensitivity of the M3 and M4 mirrors can be entirely neutralized by simply relying on symmetry. The incidence planes on corresponding M3 and M4 mirrors are orthogonal, so if M3 and M4 have identical incidence angles and identical optical coatings, their polarization effects will automatically be self-canceling.

[0208] In practice, it may actually be advantageous to specify slightly different reflective coatings for different fold mirrors in order to induce reflectivity differences between fold mirrors. This would be done to counterbalance the effect of absorption by ambient xenon gas. Assuming a xenon pressure of 1 mTorr (Ref. 8), the xenon absorption would be approximately 8% loss per meter of optical path length at the 11.3-nm operating wavelength. The source-to-wafer path length differs by up to 200 mm between different illumination fields, so the absorption would induce a 1.6% exposure variation across the wafer. However, this variation can be eliminated by designing compensating reflectivity losses into the M3 and M4 mirrors coatings. (This modification does not require graded-thickness coatings; it just requires separate coating specifications for different fold mirrors.)

[0209] The maximum source-to-wafer path length is approximately 1.6 meter. Again assuming 1 mTorr pressure, the xenon transmittance η_Xeover this distance is

[0210] η_Xe=0.875 Eq. 6.1

[0211] Since the fold mirrors operate at a high incidence angle (75°), they can use a very simple reflective coating. For example, a bare ruthenium coating has good EUV reflection efficiency at this incidence angle. FIG. 28 illustrates the series reflectivity of two bare ruthenium mirrors with crossed incidence planes and 75° incidence angles, as a function of wavelength λ (curve 2801). At the design wavelength of 11.3 nm the mirrors' combined reflection efficiency η_M3,M4is

[0212] η_M3,M4=0.684 Eq. 6.2

[0213] In practice, the M3 and M4 mirrors could comprise ruthenium films with a thin (e.g., several nanometer thick) overcoat similar to passivating coatings used for EUV mirrors (Ref. 26). The overcoat would provide improved corrosion resistance, and its thickness would provide a degree of freedom that could be used to.adjust the mirror reflectivity to counterbalance other efficiency factors such as the ambient xenon absorption.

[0214] Multilayer Mo/Be reflective coatings are required for mirrors M1 and M2, which operate close to normal incidence. Moreover, graded-thickness coatings are required to accommodate the variation in incidence angle across the mirror apertures. At incidence angles exceeding 25° the maximum attainable reflection efficiency is seriously degraded, so the maximum incidence angles on M1 and M2 are constrained to 250. FIG. 28 illustrates the series reflectivity of two multilayer Mo/Be mirrors with parallel incidence planes (and unpolarized incident radiation), as a function of wavelength λ, for normal incidence (curve 2802), and for 25° incidence on both mirrors (curve 2803). Compared to normal incidence, the spectrally integrated series reflectivity at 25° incidence is reduced by a factor of 0.82.

[0215] The design of mirrors M1 and M2 is based on 70-bilayer Mo/Be coatings of the type described in Refs. 2 and 3. The molybdenum and beryllium film thicknesses t_Moand t_Beare selected so that each Mo/Be bilayer has a half-wave optical thickness at the design wavelength and incidence angle, with the bilayer's optical thickness evenly divided between the Mo and Be films,

\begin{matrix} t_{Mo} = (1 - f_{Be}) \frac{λ_{D}}{2 Re [\sqrt{{\tilde{n}}_{Mo}^{2} - \sin^{2} θ}]} & Eq.  6.3 \\ t_{Be} = f_{Be} \frac{λ_{D}}{2 Re [\sqrt{{\tilde{n}}_{Be}^{2} - \sin^{2} θ}]} & Eq.  6.4 \end{matrix}

[0216] wherein θ is the incidence angle (in vacuum), λ_Dis the design wavelength (11.3 nm), ñ_Moand ñ_Beare the respective complex refractive indices for molybdenum and beryllium at wavelength λ_D, and f_Beis the beryllium optical thickness fraction,

[0217] f_Be=0.5 Eq. 6.5

[0218] ñ_Mohas the value given in Eq. 2.11, and ñ_Beis

[0219] ñ_Be=n_Be−ik_Be;

[0220] n_Be=1.00818,

[0221] k_Be=0.000966 Eq. 6.6

[0222] The reflectivity curves 2802 and 2803 in FIG. 28 are based on the above film specification. (In practice the reflectivity model should take into account factors such as film scattering, interlayer diffusion, and surface corrosion, but these factors are neglected here.)

[0223] The M1, M2 mirror design is also based on assumed emission characteristics of the LPP source, which will be defined in relation to the optical geometry illustrated in FIG. 29 (cf FIG. 21). For the purpose of mirror design the LPP is considered to be an infinitesimally small source at M1's focal point 2101. Rotational symmetry with respect to the optical axis 2102 is assumed, so a typical ray segment 2901aoriginating from point 2101 can be characterized by a spectral radiant intensity, I_λ=dI/dλ (Ref. 43), which is a function only of wavelength λ and the emission angle Θ₁relative to axis 2102,

[0224] I_λ=I_λ[λ,Θ₁] Eq. 6.7

[0225] (A comment on notation: Square braces “[ . . . ]” are used throughout this discussion to delimit function arguments; parentheses “( . . . )” are used for subexpression grouping; and curly braces “{ . . . }” are used to delimit lists or vectors.) I_λ is assumed to be separable in λ and Θ₁,

[0226] I_λ[λ,Θ₁]=η_src[Θ₁]I_0λ[λ] Eq. 6.8

[0227] wherein η_src[Θ₁] is a dimensionless factor representing the plasma emission's angular distribution, and I_0λ[λ] is the source's spectral radiant intensity in the downward (Θ₁=0) direction. (I_0λ[λ] has units of watts per nanometer per steradian.) The functional form of η_src[Θ₁] is

[0228] η_src[Θ₁]=(0.88+0.13 cos[Θ₁])/1.01 Eq. 6.9

[0229] A polar plot of this function is illustrated in FIG. 30, and the function I_0λ[λ] is illustrated in FIG. 31. (Eq. 6.9 and FIGS. 30 and 31 are based on data in Refs. 7 and 8.)

[0230] Referring again to FIG. 29, the meridional optical geometry is defined relative to {y,z} Cartesian coordinates, wherein the z axis coincides with optical axis 2102 and the origin {y,z}={0,0) is at focal point 2101. Ray segment 2901 a originates from point {0,0), intercepts M1 at a point {y₁,z₁}, and is reflected into a ray segment 2902awith an axial angle Θ₂relative to axis 2102. Segment 2902aintercepts M2 at a point {y₂,z₂} and is reflected into a ray 2903a, which is directed parallel to axis 2102. The incidence angle of segment 2901aon M1 is θ₁, and the incidence angle of segment 2902aon M2 is θ₂. Similarly, a ray segment 2901bproximate to 2901ais reflected into ray segment 2902band ray 2903bproximate to 2902aand 2903a, respectively. The geometric parameters of this second ray sequence differ from those of the first by corresponding differential quantities; for example, the emission angle of ray segment 2901bis Θ₁+dΘ₁, etc.

[0231] The spectral radiant intensity of ray segment 2901ais defined (Ref. 43) as the radiant power per unit solid angle per unit wavelength interval, i.e.

\begin{matrix} I_{λ} = \frac{ⅆ Φ}{ⅆ ω ⅆ λ} & Eq.  6.10 \end{matrix}

[0232] wherein dΦ is the differential radiant power transmitted within a differential solid angle dω and differential wavelength interval dλ. FIG. 32 illustrates a solid angular range dω defined by the set of all rays emitted from source point 2101 between the emission angle limits Θ₁and Θ₁+dΘ₁). Proximate ray segments 2901aand 2901bin FIG. 29 define the limits of this angular range in the meridional plane of FIG. 29, and the differential solid angle covered by this range is

[0233] dω=2πsin Θ₁dΘ₁ Eq. 6.11

[0234] The rays emitted from the source within this range define a source beam that is mapped by M1 and M2 onto an annular-aperture collimated beam with a differential cross-sectional area dA, as illustrated in FIG. 33. A meridional section of this beam in the plane of FIG. 29 is delimited by proximate rays 2903aand 2903b, and the differential beam area is

[0235] dA=2πy₂dy₂ Eq. 6.12

[0236] The differential radiant power dΦ′ transmitted in the collimated beam is equal to the 10 differential power dΦ in the source beam, attenuated by a mirror efficiency factor η_M1,M2,

[0237] dΦ′=η_M1,M2dΦ Eq. 6.13

[0238] The spectral irradiance E_λ in the collimated beam is the radiant power per unit area per unit wavelength interval, i.e.,

\begin{matrix} E_{λ} = \frac{ⅆ Φ^{'}}{ⅆ λ ⅆ A} = (\frac{ⅆ Φ^{'}}{ⅆ Φ}) (\frac{ⅆ ω}{ⅆ A}) (\frac{ⅆ Φ}{ⅆ λ ⅆ ω}) = η_{M1, M2} \frac{\sin Θ_{1} ⅆ Θ_{1}}{y_{2} ⅆ y_{2}} I_{λ} & Eq.  6.14 \end{matrix}

[0239] Eqs. 6.10-13). The mirror efficiency factor η_M1,M2is a function of wavelength λ and of both incidence angles θ₁and θ₂,

[0240] η_M1,M2=η_M1,M2[λ,θ₁,θ₂] Eq. 6.15

[0241] Substituting Eqs. 6.8 and 6.15 in Eq. 6.14 and integrating over wavelength, an expression for the collimated beam's spectrally integrated irradiance E is obtained,

\begin{matrix} E = \int E_{λ} ⅆ λ = \frac{\sin Θ_{1} ⅆ Θ_{1}}{y_{2} ⅆ y_{2}} η_{src} [Θ_{1}] \int η_{M1, M2} [λ, θ_{1}, θ_{2}] I_{0 λ} [λ] ⅆ λ & Eq.  6.16 \end{matrix}

[0242] It is evident from the geometry of FIG. 29 that

[0243] Θ₂=2θ₂ Eq. 6.17

[0244] Θ₁=Θ₂+2θ₁=2(θ₁+θ₂) Eq. 6.18

[0245] thus the efficiency factors on the right side of Eq. 6.16 can be expressed as a function only of θ₁and θ₂,

[0246] η_src[Θ₁]∫η_M1,M2[λ,θ₁,θ₂]I_0λ[λ]dλ=I₀η[θ₁,θ₂] Eq. 6.19

[0247] wherein

\begin{matrix} I_{0} = η_{src} [0] \int η_{M1, M2} [λ, 0, 0] I_{0 λ} [λ] ⅆ λ & Eq.  6.20 \\ η [θ_{1}, θ_{2}] = \frac{1}{I_{0}} η_{src} [2 (θ_{1} + θ_{2})] \int η_{M1, M2} [λ, θ_{1}, θ_{2}] I_{0 λ} [λ] ⅆ λ & Eq.  6.21 \end{matrix}

[0248] Based on the assumed Mo/Be mirror coating design and source emission characteristics, Eqs. 6.20 and 6.21 can be numerically evaluated to obtain

[0249] I₀=6.0 W/sr Eq. 6.22

[0250] η[θ₁,θ₂]≅1−0.533(θ₁²+θ₂²)−0.804(θ₁⁴+θ₂⁴)+2.123θ₁²θ₂² Eq. 6.23

[0251] (θ₁and θ₂have units of radians in Eq. 6.23. This approximation has an accuracy of 0.003 for θ₁and θ₂in the range 0 to or π/6.) Substituting Eq. 6.19 in Eq. 6.16, the irradiance in the collimated beam is

\begin{matrix} E = \frac{\sin Θ_{1} ⅆ Θ_{1}}{y_{2} ⅆ y_{2}} I_{0} η [θ_{1}, θ_{2}] & Eq.  6.24 \end{matrix}

[0252] A highly desirable aspect of the mirror design is to make E constant over the beam aperture. Thus, E is assumed to be constant, and the two constants E and I₀in Eq. 6.24 are combined into a single constant F representing the collection system's effective paraxial focal length,

[0253] F={square root over (I₀/E)} Eq. 6.25

[0254] Eq. 6.24 is restated as

\begin{matrix} ⅆ y_{2} = F^{2} η [θ_{1}, θ_{2}] \frac{\sin Θ_{1}}{y_{2}} ⅆ Θ_{1} & Eq.  6.26 \end{matrix}

[0255] The design geometry of mirrors M1 and M2 is determined by numerically solving two coupled differential equations for the mirror surface profiles. Eq. 6.26 represents one of these equations, and the other is obtained from the law of reflection, which can be expressed as a differential equation. Applying this law at point {y₁,z₁} (FIG. 29) yields

\begin{matrix} (\frac{{y_{1}, z_{1}}}{\sqrt{y_{1}^{2} + z_{1}^{2}}} - \frac{{y_{2} - y_{1}, z_{2} - z_{1}}}{\sqrt{{(y_{2} - y_{1})}^{2} + {(z_{2} - z_{1})}^{2}}}) \cdot {ⅆ y_{1}, ⅆ z_{1}} = 0 & Eq.  6.27 \end{matrix}

[0256] Similarly, the reflection equation at point {y₂,z₂} is

\begin{matrix} (\frac{{y_{2} - y_{1}, z_{2} - z_{1}}}{\sqrt{{(y_{2} - y_{1})}^{2} + {(z_{2} - z_{1})}^{2}}} + {0, 1}) \cdot {ⅆ y_{2}, ⅆ z_{2}} = 0 & Eq.  6.28 \end{matrix}

[0257] The sum of Eqs. 6.27 and 6.28 yields an expression that can be directly integrated,

[0258] d({square root over (y₁²+z₁²)}+{square root over ((y₂−y₁)²+(z₂−z₁)²)}+z₂)=0 Eq. 6.29

[0259] Using this equation in lieu of Eq. 6.27, there remain only two differential equations (Eqs. 6.26 and 6.28) that need to be solved numerically.

[0260] Before Eqs. 6.26, 6.28, and 6.29 can be applied, they must be augmented with a set of algebraic equations defining the relationships between the Cartesian coordinate parameters (y₁, z₁, y₂, and z₂) and the angle parameters (θ₁, θ₂, Θ₁, and Θ₂). These relationships also involve the respective lengths l₁and l₂of ray segments 2901aand 2902a(FIG. 29). All of these parameters will be defined as algebraic functions of three independent parameters, θ₁, θ₂, and l₁; and then the two differential equations, Eqs. 6.26 and 6.28, will be used to solve for θ₂and l₁as functions of a single independent parameter, θ₁.

[0261] Based on the FIG. 29 geometry, the Cartesian coordinate parameters can be represented as

[0262] y₁−l₁sin Φ₁ Eq. 6.30

[0263] z₁=−l₁Cos Φ₁ Eq. 6.31

[0264] y₂=y₁+l₂sin Φ₂ Eq. 6.32

[0265] z₂=z₁+l₂cos Φ₂ Eq. 6.33

[0266] Eqs. 6.17 and 6.18 define Θ₁and Θ₂as functions of θ₁and θ₂, so the only remaining undefined parameters are θ₁, θ₂, l₁, and l₂. A defining relationship for I₂can be obtained by integrating Eq. 6.29,

[0267] l₁+l₂+z₂=2l₂⁽⁰⁾ Eq. 6.34

[0268] wherein l₂⁽⁰⁾is an integration constant, the axial value of l₂(FIG. 29). Eliminating z₂between Eqs. 6.33 and 6.34, a defining equation for l₂is obtained,

\begin{matrix} l_{2} = \frac{2 l_{2}^{(0)} - l_{1} - z_{1}}{1 + \cos Φ_{2}} & Eq.  6.35 \end{matrix}

[0269] Eqs. 6.17, 18, 30, 31, 35, 32, and 33 are applied, in that order, to define the respective parameters Θ₂, Θ₁, y₁, z₁, l₂, y₂, and z₂as functions of the three independent parameters θ₁, θ₂, and l₁. Then θ₂and l₁are represented as undetermined functions of θ₁,

[0270] θ₂=θ₂[θ₁] Eq. 6.36

[0271] l₁=l₁[θ] Eq. 6.37

[0272] and Eqs. 6.26 and 6.28 define a pair of coupled differential equations that can be numerically solved for θ₂[θ₁] and l₁[θ₁],

\begin{matrix} \frac{ⅆ y_{2}}{ⅆ θ_{1}} = F^{2} η \frac{\sin Θ_{1}}{y_{2}} \frac{ⅆ Θ_{1}}{ⅆ θ_{1}} & Eq.  6.38 \\ (y_{2} - y_{1}) \frac{ⅆ y_{2}}{ⅆ θ_{1}} + (z_{2} - z_{1} + l_{2}) \frac{ⅆ z_{2}}{ⅆ θ_{1}} = 0 & Eq.  6.39 \end{matrix}

[0273] The boundary conditions on θ₂and l₁are

[0274] θ₂[0]=0 Eq. 6.40

[0275] l₁[0]=l₁⁽⁰⁾ Eq. 6.41

[0276] wherein l₁⁽⁰⁾is the axial value of l₁(FIG. 29).

[0277] The right side of Eq. 6.38 is indeterminate at θ₁=0 because Θ₁and y₂are both zero. However, the indeterminacy can be resolved by using the following limiting condition,

\begin{matrix} \lim_{θ_{1} -> 0} \frac{\sin Θ_{1}}{y_{2}} = \frac{1}{F} & Eq.  6.42 \end{matrix}

[0278] (In principle, the integration of Eqs. 6.38 and 6.39 need not start at θ₁=0, because M1 and M2 have annular clear apertures and the axial portions of the mirrors are not optically functional. However, the advantage of integrating from θ₁=0, and using Eq. 6.40 as a boundary condition, is that the resulting surface shape for M2 is nearly spherical, and a slight design modification described below can make it exactly spherical.)

[0279] The above design prescription for M1 and M2 has three undetermined design constants: F, l₁⁽⁰⁾, and l₂⁽⁰⁾. Also, the annular aperture radius limits of M2, R_2minand R_2max(FIG. 22) are undefined. These constants can be selected to satisfy a variety of practical constraints or performance objectives. The values used in the above design example are

[0280] R_2min=154.0 mm Eq 6.43

[0281] R_2max=315.0 mm Eq 6.44

[0282] F=223.5 mm Eq 6.45

[0283] l₁⁽⁰⁾=133.1 mm Eq 6.46

[0284] l₁⁽⁰⁾=657.2 mm Eq 6.47

[0285] The R_2minand R_2maxvalues are based on the M3 aperture layout illustrated in FIG. 23. The other three constants, F, l₁⁽⁰⁾, and l₂⁽⁰⁾, are determined to satisfy three design constraints: The maximum incident angle θ₁on M1 and the maximum incident angle θ₂on M2 are both set to 25°, and the optical clearance diameter around the plasma source 601 is set to 100 mm (FIG. 27).

[0286] Some practical improvements in the above design process can be made. First, although the design produces illumination of uniform irradiance across the wafer print modules, the angular divergence and polarization state of the illumination may not be uniform. These factors could have a slight effect on the microlens efficiencies, resulting in exposure nonuniformity over the wafers. However, the efficiency factor η in Eq. 6.38 can be defined to include the micro-optics' efficiency factors so that these are also compensated for in the design. Also, the efficiency factor can be defined to at least partially compensate for the ambient xenon absorption, to the extent possible with axially symmetric optics. (Any residual uncorrected errors in these factors can be compensated for in the M3, M4 coating design.)

[0287] The M1 and M2 surface shapes obtained by the above process are both aspheric. However, mirror M2 is nearly spherical (with less than 1-mm asphericity), and the optical design can be slightly modified to make it exactly spherical. The basic procedure is to first obtain a preliminary aspheric design by the above process and replace M2 with a closely fitting spherical surface. M1's surface shape is then reoptimized to satisfy the collimation condition, but ignoring the uniformity constraint. Eq. 6.39 represents the collimation constraint, but in lieu of Eq. 6.38 (the uniformity constraint) the following sphericity constraint is imposed on M2,

[0288] y₂²+(z₂−z_C)²=R² Eq. 6.48

[0289] wherein R is the spherical radius of M2, and z_Cis the z coordinate of the sphere's center point. This results in a mirror design that satisfies the collimation condition, but which exhibits some nonunifornity in the illumination irradiance. The coating specification for M1 or M2 can then be adjusted to eliminate the nonuniformity (e.g., by varying the f_Befactor in Eqs. 6.3 and 6.4 across the mirror apertures to fine-adjust the mirrors' reflectivity distribution).

§7 System Throughput

[0290] The system's throughput performance can be estimated from the design data provided in the preceding sections. The irradiance E in the collimated output from M2 is obtained from Eqs. 6.25, 6.22, and 6.45,

[0291] E=I₀/F²=0.120 mW/mm² Eq. 7.1

[0292] This value includes the reflectivity losses of M1 and M2, but not of M3 and M4; and it also neglects the ambient xenon absorption loss. Factoring in the combined reflectivity for M3 and M4 (η_M3,M4=0.684, Eq. 6.2) and the xenon loss (η_Xe=0.875, Eq. 6.1), the illumination irradiance on the printheads, E_illum, is

[0293] E_illum=E η_M3,M4η_Xe=0.0719 mW/mm² Eq. 7.2

[0294] The total illumination area A_illumis the area per illumination field (W′_xW′_y; wherein W′_x=W′_y=22 mm, Eqs. 1.8, 1.9), multiplied by the number of fields per wafer print module (45) and the number of modules (8),

[0295] A_illum=W′_xW′_y×45×8=174,240 mm² Eq. 7.3

[0296] The total radiant power Φ_illumdelivered to the modules by the illumination system is

[0297] Φ_illum=E_illumA_illum=12.5 W Eq. 7.4

[0298] The maximum power delivered to the eight wafer printing planes, Φ_print, is obtained by multiplying the above value by the following efficiency factors: the printheads' aperture collection efficiency, η_{head ap}=0.826 (Eq. 1.10); the lens fill factor η_FF=0.630 (Eq. 1.3); the microlens transmittance η_lens=0.169 (Eq. 2.15); and the pellicle transmittance, η_pellicle=0.95 (Eq. 1.1).

[0299] Φ_print=Φ_illumη_{head ap}η_FFη_lensη_pellicle=1.04 W Eq. 7.5

[0300] The resist exposure threshold is assumed to be 20 mJ/cm². To achieve good-contrast imaging for mixed positive- and negative-tone patterns this threshold level should be at approximately 25% of the maximum exposure level. Thus the maximum level, Q_A, is specified as

[0301] Q_A=(20 mJ/cm²)/0.25=80 mJ/cm² Eq. 7.6

[0302] The exposure area coverage rate A_tis

[0303] A_t=Φ_print/Q_A=13.1 cm²/sec Eq. 7.7

[0304] The total wafer-plane printing area A_printscanned by the printer during a full wafer print cycle is equal to the area per scan per printhead (i.e., W_xW_y; wherein W_x=W_y=20 mm, Eqs. 1.4, 1.5) multiplied by the number of scans per cycle (4), the number of printheads per wafer print module (45) and the number of modules (8),

[0305] A_print=W_xW_y×4×45×8≅5760 cm² Eq. 7.8

[0306] (Note: This area slightly overfills the wafer apertures—see FIG. 5—so the wafer-plane power Φ print does not actually all intercept the wafers. However, the printing throughput is determined by the wafer-plane power per unit area, which is independent of the actual wafer size and area coverage.) The print time per cycle, t_cycle, is

[0307] t_cycle=A_print/A_t≅441 sec Eq. 7.9

[0308] Each cycle exposes 8 wafers, so the print time per wafer, t_wafer, is

[0309] t_wafer=t_cycle/8≅55.2 sec Eq. 7.10

[0310] This translates to 65 wafers per hour, not including the wafer load time and other throughput overhead factors. (A revised throughput estimate of 62 wafers per hour is obtained in the next section.)

§8 Scanning Methods

[0311] As each printhead scans the wafer, its pixels are modulated in synchronization with the LPP source to create a latent image in the resist in the form of a matrix dot pattern comprising overlapping image dots. (The “dots” could alternatively be termed “image pixels”, but in the present discussion the term “pixel” is reserved for use in connection with printer pixels.) Each dot is exposed through a single printer pixel during an LPP pulse of duration <10 nsec (Ref 8). The LPP source's pulse repetition rate ν_srcis

[0312] ν_src=6000 Hz Eq. 8.1

[0313] and the exposure energy Q_dotdelivered to each dot (with the pixel in its ON state) is approximately

[0314] Q_dot=Φ_print/(ν_srcN_pixel)=3.77×10⁻¹⁴J Eq. 8.2

[0315] which is approximately 2000 photon units (from Eqs. 1.7, 7.5, 8.1). (The exposure energy in the OFF state is assumed to be zero.) The dots are centered at the points of a rectangular printing grid on the wafer. Assuming that the grid comprises square unit cells with a step dimension G between grid points, the maximum area-averaged exposure level Q_A(i.e., energy per unit area) on the wafer is

[0316] Q_A=Q_dot/G² Eq. 8.3

[0317] Hence the grid step is

[0318] G={square root over (Q_dot/Q_A)}=6.86 nm Eq 8.4

[0319] (from Eqs. 7.6, 8.2).

[0320] Each dot's radial exposure profile has the form illustrated by the intensity profile I[r₃] in FIG. 12. The profile's FWHM is 58 nm, an order of magnitude larger than the grid step G. Thus, a typical exposure point will receive energy from multiple overlapping dots and will hence not be strongly influenced by statistical variations in the exposure level. Also, random or isolated defects in the printer pixels will not have too significant an effect on the exposure pattern, provided that the dots exposed by any particular pixel are sparsely distributed so that there is not much overlap between them. The intensity profile I[r₃] is at the 1% level at 75-nm radius (FIG. 12; Eq. 2.12), so there will be no significant overlap if the dots exposed by each pixel are separated by a minimum “proximity clearance” distance d_clearof 150 nm,

[0321] d_clear=150 nm Eq 8.5

[0322] The 6.86-nm grid step can be achieved, while maintaining the 150-nm dot separation for each pixel, by using an interleaved raster scan technique in which individual pixels print coarsely-spaced rectangular arrays of dots which are spatially interleaved to form much more finely-spaced dot patterns. Two versions of this method are described below. In one method, termed “lens-scan”, each printhead and its attached microlenses scan across a stationary wafer during wafer exposure. Four scans are performed, each covering approximately 25% of the wafer area, and the wafer is stepped between scans to provide nearly full wafer coverage. (FIG. 5 illustrates the wafer's four scan positions as 501a, 501b, 501c, and 501d.) In the alternative method, termed “wafer-scan”, the wafer moves in relation to substantially stationary printheads. (The printheads would actually have to be positionally actuated with either method to maintain accurate alignment and focus, but the required positional range would be at least an order of magnitude less with the wafer-scan method. In a variant of this approach the printheads could be actuated to only control focus, while the second-stage microlenses are individually actuated to maintain lateral alignment.)

[0323] The basic design principles underlying the scan methods can be conceptually illustrated by first considering several one-dimensional scan methods in which a single row of pixels exposes a single row of image dots. For simplicity, it is assumed here that the pixels are all kept in their ON states during exposure, so the resulting exposure pattern will comprise a row of exposed dots at the specified center spacing G (the printing grid step) without any gaps or duplicate exposures. Considering a particular dot row, its dots are grouped into a multilevel subset hierarchy as follows. The top level of the hierarchy comprises subgroups Dots[i]; wherein i is an integer index that labels the printheads, and Dots[i] is the set of dots that are exposed by printhead i. (This printhead is designated as Head[i].) At the next level, each subgroup Dots[i] itself comprises subgroups Dots[i,j]; wherein j is an integer index that labels separate scans, and Dots[i,j] is the set of dots in Dots[i] that are exposed during scan j. (This scan is designated as Scan[j]. The minimal rectangular area covering Dots[i,j] is termed a “scan field”, which is designated as Field[i,j].) Dots[i,j] further comprises subgroups Dots[i,j,k]; wherein k is an integer index that labels the specific pixels in Head[i] that cover the dot row, and Dots[i,j,k] is the set of dots in Dots[i,j] that are exposed during Scan[j] through pixel k in Head[i]. (This pixel is designated as Pixel[i,k].) Finally, Dots[i,j,k] comprises individual dots Dot[i,j,k,l]; wherein l is an integer index that identifies the exposure pulses during Scan[j], and Dot[i,j,k,l] is the specific dot in Dots[i,j,k] that is exposed by pulse l in Scan[j]. (This pulse is designated as Pulse[j,l].)

[0324] FIGS. 34A and 34B illustrate a simple sequential dot row exposure pattern in which each pixel exposes a contiguous sequence of dots. In this example, all the dots are exposed with a single printhead, in a single scan. For notational consistency, the printhead and scan will be identified respectively as Head[0] and Scan[0] (i.e., i=0 and j=0 in the above notation). The pixels are designated as Pixel[0,k] (k=. . . , −1, 0, 1, . . . ), and the dots printed by pixel Pixel[0,k] are designated as Dot[0,0,k,l] (l=0, 1, 2, . . . ). The large circles in FIGS. 34A and 34B and subsequent figures represent the pixels' first-stage lens apertures; the small circles indicate center points of image dots that have been exposed; and the “+” marks indicate center points of unexposed dots. One particular pixel (Pixel[0,0]) is highlighted, and the dots printed by this pixel (Dot[0,0,0,0], Dot[0,0,0,1], etc.) are represented by filled-in circles. FIG. 34A illustrates the dot exposure pattern when the scan is halfway to completion, and FIG. 34B illustrates the pattern after completion. The scan direction is indicated as X, and the dot center spacing in the X direction (i.e., the scan distance per exposure pulse) is G_x. (G_xis nominally equal to the printing grid step G defined in Eq. 8.4. For two-dimensional printing, the grid cells may not necessarily be perfectly square, so the cell dimensions will be denoted as G_xand G_y.) The pixel center spacing is ΔX_pixelG_x, wherein ΔX_pixelis an integer (e.g., ΔX_pixel=16 in FIGS. 34A and 34B), and the number of exposure pulses per scan line, N_pulse/line, is equal to ΔX_pixel,

[0325] N_pulse/line=ΔX_pixel Eq. 8.6

[0326] Thus, after the scan is completed, (FIG. 34B) the contiguous dot groups Dots[0,0,k] (k=. . . , −1, 0, 1, . . . ) form an uninterrupted, uniformly spaced row of dots without any gaps, duplicate exposures, or spacing discontinuities between groups.

[0327] With the above scan method the dots printed by each pixel are closely clustered in a small region, so pixel defects or non-functional pixels could result in unacceptable printing errors. This limitation is overcome by using the interleaved exposure pattern illustrated in FIGS. 35A and 35B. (FIG. 35A shows the scan halfway to completion, and FIG. 35B shows the completed scan.) With this method, the scan distance per exposure pulse is ΔX_pulseG_x, wherein ΔX_pulseis an integer; and each pixel prints a regularly-spaced dot sequence with a dot centering interval of ΔX_pulseG_x. The pixel center spacing is ΔX_pixelG_x, as in FIGS. 34A and 34B. ΔX_pixeland ΔX_pulseare relatively prime integers (e.g., ΔX_pixel=16 and ΔX_pulse=3 in FIGS. 35A and 35B),

[0328] GCD[ΔX_pixel,ΔX_pulse]=1 Eq. 8.7

[0329] wherein “GCD” denotes the greatest common divisor. The relative primality condition implies that an infinite row of pixels would fully expose an infinite, uniformly spaced row of dots at the grid step interval G_x, without gaps or duplicate exposures, if the number of exposure pulses N_pulse/lineis equal to ΔX_pixel(Eq. 8.6). Also, if ΔX_pulseis sufficiently large the dots printed by a particular pixel will not overlap significantly, although there will be significant overlap between dots printed by different pixels. Hence, printing errors due to isolated pixel defects or non-functional pixels will be significantly mitigated.

[0330] FIGS. 36A and 36B illustrate a variant of the interleaved scan technique in which a printhead comprising a pixel row with a finite number of pixels is iteratively stepped and scanned to cover a dot row of indefinite extent, with each scan comprising N_pulse/lineexposure pulses on the dot row. As in FIGS. 35A and 35B, the scan distance per pulse is ΔX_pulseG_x, the pixel center spacing is ΔX_pixelG_x, and N_pulse/lineis equal to ΔX_pixel. (The restriction on the number of pulses per scan line can be effectively removed by using a “transitionless scan” method, which is discussed below.) The number of pixels per printhead pixel row is denoted as N_pixel/row(e.g., N_pixel/row=5 in FIGS. 36A and 36B). FIG. 36A illustrates a pixel row comprising pixels Pixel[0,0], Pixel[0,1], . . . , Pixel[0, N_pixel/row−1] at two positions: at the start of Scan[0] (dashed circles), and at the end of Scan[0] (solid circles); and FIG. 36B similarly illustrates the pixels at the start and end of Scan[1]. The printhead (Head[0]) covers a scan field indicated as Field[0,0] during Scan[0], and covers Field[0,1] during Scan[1]. The wafer is stepped between successive scans, and the distance between the start positions of consecutive scans (e.g., the center distance between Dot[0,0,0,0] and Dot[0,1,0,0]) is ΔX_fieldG_x, wherein

[0331] ΔX_field=N_pixel/rowΔX_pixel Eq. 8.8

[0332] This distance is selected so that the position occupied by each pixel, Pixel[0,k] (k=0 . . . N_pixel/row−1) during Scan[j] (j=1, 2, . . . ) coincides with the position that would be occupied by Pixel[0, j N_pixel/row+k] during Scan[0] if the pixel row were extended to include more than N_pixel/rowpixels. In other words, the multi-scan exposure pattern is identical to the pattern that would be produced with a single scan, using an extended pixel row.

[0333] The dot pattern generated by an individual scan and printhead, such as Dots[0,0] or Dots[0,1], will have gaps in the exposure pattern near the ends of the corresponding scan field. For example, the end region 3601 of Field[0,0] in FIG. 36A is illustrated at a larger scale in FIG. 37, showing details of the completed exposure pattern with gaps. (FIG. 37 is the same as FIG. 35B, except that there are no negative-numbered pixels Pixel[0,−1], Pixel[0,−2], etc., and no negative-numbered scans.) The incompletely exposed regions bordering each scan field are termed “stitch zones”. The stitch zones of adjacent scan fields (e.g., zones 3602 in FIG. 36A and 3603 in FIG. 36B) are overlapped (i.e., “stitched”), with each seamlessly filling in the exposure gaps in the other.

[0334] When multiple printheads are employed, their scan patterns can be designed so adjacent scan fields printed by different printheads are similarly stitched. The design approach is illustrated as follows. Each dot row is exposed through multiple printheads (Head[0], Head[1], etc.) and multiple scans, Scan[0], Scan[1], . . . , Scan[N_{x scan}−1]; wherein N_{x scan}is the number of scans and Scan[j] comprises pulse exposures Pulse[j,0], Pulse[j,1], . . . , Pulse[j, N_pulse/line−1]. For example, FIG. 38 illustrates pixel rows on two printheads, Head[0] comprising Pixel[0,0], Pixel[0,1], etc., and Head[1] comprising Pixel[1,0], Pixel[1,1], etc. The pixels are illustrated at two positions: at Pulse[0,0] (i.e., the first pulse of Scan[0], indicated by dashed circles), and at Pulse[N_{x scan}−1, N_pulse/line−1] (the last pulse of Scan[N_{x scan}−1], solid circles). The positional displacement between adjacent printheads is ΔX_head·G_x, wherein

[0335] ΔX_head=N_{x scan}ΔX_field Eq. 8.9

[0336] This value is chosen so that if the exposure sequence were to be continued past the last scan, the position of Head[0] at Pulse[N_{x scan}, 0] would coincide with the position of Head[1] on Pulse[0,0]. For example, in FIG. 38 Pixel[0,0] (solid circle) would step to the first position occupied by Pixel[1,0] (dashed circle). Hence the exposure pattern obtained with multiple printheads is equivalent to the pattern that would be obtained with a single printhead by extending its printing range (i.e., increasing the number of scans). Consequently, adjacent printheads' stitch zones are exactly complementary, with each seamlessly filling in the other's exposure gaps.

[0337] It is presumed above that the number of scans N_{x scan}is integer-valued, but in some circumstances it need not be. The last scan could be terminated prior to completion, and ΔX_headcould be selected so that, for example, the exposure pattern printed by Head[0] is continued from its termination point by Head[1]. In this case, N_{x scan}could have a fractional value. A fundamental problem with this “fractional scan” printing method is that if, for example, Head[0] completely scans all but the last of its scan fields, with the remaining portion of the last field being covered by Head [1] on its first scan, then Head[1] will complete its first (fractional) scan before Head[0] completes its first (full) scan. Thus the field stepping between scans would not be synchronized between printheads. Of course, this limitation would not be relevant if only one printhead is employed (or if the system is not intended to provide seamless image stitching between printheads). Furthermore, there is one particular circumstance, described below, in which asynchronous stepping would be practical and the fractional scan method could be used to advantage.

[0338] Scan[j] comprises pulse exposures Pulse[j,0], Pulse[j,1], . . . , Pulse[j, N_pulse/line−1], wherein the number of pulses (N_pulse/line) is equal to ΔX_pixel(Eq. 8.6). The scanning element (i.e., the printhead or the wafer) moves a distance of ΔX_pulseG_xbetween pulses, so the scan distance traversed between the start and end positions of each scan is (ΔX_pixel−1) ΔX_pulseG_x(FIG. 36A). Also, the distance between the start positions of consecutive scans is ΔX_fieldG_x=N_pixel/rowΔX_pixelG_x(FIG. 36B, Eq. 8.8); thus the stepping distance from a scan's end position to the next scan's start position is ΔX_stepG_x, wherein

[0339] ΔX_step=N_pixel/rowΔX_pixel−(ΔX_pixel−1)ΔX_pulse Eq. 8.10

[0340] For some scan strategies (in particular, the wafer-scan method), it may be convenient to set N_pixel/row=ΔX_pulseso that ΔX_step=ΔX_pulseand the stepping distance (ΔX_stepG_x) is equal to the scan distance per pulse (ΔX_pulseG_x). In this case, the process of “stepping” between two consecutive scans is indistinguishable from simply continuing the first scan without interruption. Since there are no stepping transitions between scans, the printheads can step asynchronously and the fractional scan method can be employed. The separate “scans” (as defined above) can be considered to constitute portions of a single, extended scan, which will be termed a “transitionless scan”, and this scanning mode will be termed “transitionless scanning”.

[0341] The defining condition for transitionless scanning is

[0342] N_pixel/row=ΔX_pulse(transitionless scan) Eq. 8.11

[0343] Under this condition the N_{x scan}factor in Eq. 8.9 need not be integer-valued, and the only fundamental restriction on ΔX_headis that it must be an integer multiple of ΔX_pulsein order to maintain pulse synchronization between printheads;

[0344] ΔX_head=N_{x pulse}ΔX_pulse Eq. 8.12

[0345] wherein N_{x pulse}is the total number of exposure pulses on the dot row. (As a practical consideration, N_{x pulse}must be sufficiently large to ensure mechanical clearance between adjacent printheads.) The scan-related parameters defined above (N_pulse/line, ΔX_field, N_{x scan}) and related equations (Eqs. 8.6, 8.8, 8.9) are not very relevant to a transitionless scan. Instead, transitionless scanning can be characterized by corresponding primed quantities (N_pulse/line, ΔX_field, N_{x scan}), which are defined as follows:

[0346] N_pulse/line=N_{x pulse} Eq. 8.13

[0347] ΔX′_field=ΔX_head Eq. 8.14

[0348] N_{x scan}=1 Eq. 8.15

[0349] (i.e., each printhead's scan fields are effectively merged into a single transitionless scan field). If there is only one printhead, Eqs. 8.14 and 8.15 are not relevant and the printhead could simply traverse a scan field of unlimited extent without stepping.

[0350] The design principles outlined above can be readily generalized from one-dimensional to two-dimensional scanning. In the same way that image dots are combined in line scan patterns to form image lines, the image lines can be combined in raster scan patterns to form two-dimensional images; and techniques such as scan interleaving and field stitching can be similarly applied in the second dimension. The notational designations used above for one-dimensional printing (e.g., Head[i], Scan[j], Pixel[i,k], Pulse[j,l], Dot[i,j,k,l], etc.) can be adapted to two-dimensional printing. But in this context the printheads, scan fields, pixels, and exposure positions within each scan all generalize to two-dimensional arrays, so the indexing labels i,j,k, and l are reinterpreted as two-component indices, as outlined below:

[0351] i={i_x,i_y}: printhead indices

[0352] i={j_x,j_y}: scan indices

[0353] k={k_y,k_y}: pixel indices

[0354] l={l_x,l_y}: exposure pulse indices Eq. 8.16

[0355] For brevity, the following notation will be used for zero indices,

[0356] {right arrow over (0)}={0,0} Eq. 8.17

[0357] FIG. 39 illustrates a printhead (Head[{right arrow over (0)}]) comprising pixels designated as Pixel[{right arrow over (0)}, {k_x,k_y}]; wherein k_xand k_yare integers. The k_yindex labels the printer rows (i.e., all the pixels in any particular row, such as row 3901, have the same k_yindex), and k_xidentifies the pixels in each row. The figure also illustrates a portion of an exposure pattern generated by Head[{right arrow over (0)}] and Scan[{right arrow over (0)}]. The pattern comprises image dots designated as Dot[{right arrow over (0)},{right arrow over (0)}, {k_x,k_y}, {l_x,l_y}], each of which is printed by a corresponding pixel, Pixel[{right arrow over (0)}, {k_x,k_y}]; wherein l_x, and l_yare integers associated with Pulse[{right arrow over (0)}, {l_x,l_y}]. The dots exposed by any particular pixel are centered at the points of a two-dimensional rectangular grid, and the l_xand l_yindices label the dot column and row positions within this grid. (The “x” subscripts in the indices are associated with the scan direction, which is indicated as X in FIG. 39; and the “y” subscripts are associated with the cross-scan direction, indicated as Y. The pulse indices l_xand l_yare not necessarily related to the time-sequential order of the exposure pulses.)

[0358] Following is a reference list of terms, notation, and parameters that relate to two-dimensional scanning. (Some notational modifications would need to be made for transitionless scanning.)

[0359] Head[i] printhead

[0360] Scan[j] raster scan

[0361] Field [i,j] scan field corresponding to Head[i], Scan[j]

[0362] Pixel[i,k] individual pixel in Head[i]

[0363] Pulse[j,l] individual exposure pulse in Scan[j]

[0364] Dot[i,j,k,l] image dot printed by Pixel[i,k], Pulse[j,l]

[0365] Dots[i,j,k] union over l of Dot[i,j,k]

[0366] Dots[i,j] union over k of Dots[i,j,k]

[0367] Dots[i] union over j of Dots[i,j]

[0368] W_x, W_ynominal printhead aperture dimensions

[0369] G_x, G_yprinting grid cell dimensions

[0370] N_pulse/linenumber of exposure pulses per scan line

[0371] N_line/scannumber of scan lines per raster scan

[0372] N_pulse/scannumber of exposure pulses per raster scan

[0373] N_pixel/rownumber of pixels per printhead pixel row

[0374] N_row/headnumber of pixel rows per printhead

[0375] N_pixel/headnumber of pixels per printhead

[0376] N_row/periodnumber of pixel rows per period in the Y direction

[0377] N_period/headnumber of Y periods per printhead

[0378] N_{x scan}number of scans in the X direction

[0379] N_{y scan}number of scans in the Y direction

[0380] N_scannumber of scans, total

[0381] ΔX_pulseG_xscan distance per pulse

[0382] ΔY_lineG_yscan line spacing

[0383] ΔX_scanG_xraster scan X range

[0384] ΔY_scanG_yraster scan Y range

[0385] ΔX_pixelG_xpixel center spacing in the X direction

[0386] ΔX_offsetG_xX offset between pixel rows

[0387] ΔY_pixelG_ypixel row spacing in the Y direction

[0388] ΔY_periodG_ypixel periodicity dimension in the Y dimension

[0389] ΔX_fieldG_xX displacement between scan fields (=W_x)

[0390] ΔY_fieldG_yY displacement between scan fields (=W_y)

[0391] ΔX_stitchG_xstitch zone X width

[0392] ΔY_stitchG_ystitch zone Y width

[0393] ΔX_headG_xX displacement between printheads

[0394] ΔY_headG_yY displacement between printheads

[0395] X_dot[i,j,k,l] G_xX coordinate of Dot[i,j,k,l]

[0396] Y_dot[i,j,k,l] G_yY coordinate of Dot[i,j,k,l]

[0397] X_pixel[i,j,k] G_xX coordinate of Pixel[i,k] at Pulse[j,{right arrow over (0)}]

[0398] Y_pixel[i,j,k] G_yY coordinate of Pixel[i,k] at Pulse[j,{right arrow over (0)}]

[0399] FIG. 39 is a two-dimensional generalization of FIG. 35A, showing a typical raster scan halfway to completion. Each pixel traverses a serpentine, raster-scan path over the wafer, exposing a rectangular array of image dots in synchronization with the pulsed LPP source. For example, the highlighted pixel (Pixel[{right arrow over (0)},{right arrow over (0)}]) follows the scan path 3902 (with the time sequence indicated by the arrows), exposing dots designated as Dot[{right arrow over (0)},{right arrow over (0)},{right arrow over (0)},{l_x,l_y}]; l_x=0, 1, . . . 15, l_y=0, 1, . . . , 4. (The dots that have been exposed by this pixel at the scan halfway point are indicated as filled-in circles.)

[0400] The dot center spacing in the X direction is G_x, and the spacing in the Y direction is G_y. The scan distance covered between exposure pulses is ΔX_pulseG_x, and the scan line spacing is ΔY_lineG_y; wherein ΔX_pulseand ΔY_lineare integers (e.g., ΔX_pulse=3 and ΔY_line=6 in FIG. 39). The number of exposure pulses per scan line is designated as N_pulse/line, the number of scan lines per raster scan is designated as N_line/scan, and the total number of pulses per raster scan, N_pulse/scan, is

[0401] N_pulse/scan=N_pulse/lineN_line/scan Eq. 8.18

[0402] The region scanned by each pixel has nominal X and Y dimensions of ΔX_scanG_xand ΔY_scanG_y, wherein

[0403] ΔX_scan=N_pulse/lineΔX_pulse Eq. 8.19

[0404] ΔY_scan=N_line/scanΔY_line Eq. 8.20

[0405] The pixel center spacing in the X direction is ΔX_pixelG_x, the spacing between pixel rows in the Y direction is ΔY_pixelG_y, and there is an X offset of ΔX_offsetG_xbetween adjacent pixel rows; wherein ΔX_pixel, ΔY_pixel, and ΔX_offsetare integers (e.g., ΔX_pixel=16, ΔY_pixel=5, and ΔX_offset=8 in FIG. 39). Neglecting the printhead size limits, the pixel array is periodic in the X direction with a spatial period of ΔX_pixelG_x, and is periodic in the Y direction with a period of ΔY_periodG_y, wherein

[0406] ΔY_period=N_row/periodΔY_pixel Eq. 8.21

[0407] N_row/periodis the number of pixel rows per period,

\begin{matrix} N_{row / period} = {\begin{matrix} 1 & if Δ X_{offset} = 0 \\ Δ X_{pixel} / GCD [Δ X_{pixel}, Δ X_{offset}] & if Δ X_{offset} > 0 \end{matrix} & Eq . 8.22 \end{matrix}

[0408] (e.g., N_row/period=2 and ΔY_period=10 in FIG. 39).

[0409] N_pulse/lineis equal to ΔX_pixel, and N_line/scanis equal to ΔY_pixel,

[0410] N_pulse/line=ΔX_pixel Eq. 8.23

[0411] N_line/scan=ΔY_pixel Eq. 8.24

[0412] ΔX_pixeland ΔX_pulseare relatively prime, and ΔY_pixeland ΔY_lineare relatively prime,

[0413] GCD[ΔX_pixel,ΔX_pulse]=1 Eq. 8.25

[0414] GCD[ΔY_pixel,ΔY_line]=1 Eq. 8.26

[0415] It follows from Eqs. 8.23-26 that an infinite array of pixels would fully expose an infinite dot array at the specified grid cell dimensions G_xand G_y, without gaps or duplicate exposures, in a single raster scan consisting of N_line/scanline scans and N_pulse/lineexposure pulses per line.

[0416] The positional coordinates of a particular dot, Dot[i,j,k,l], relative to an {X, Y} coordinate system with the origin at Dot[{right arrow over (0)},{right arrow over (0)},{right arrow over (0)},{right arrow over (0)}], are denoted as {X_dot[i,j,k,l]G_x, Y_dot[i,j,k,l]G_y} (FIG. 40); wherein X_dotand Y_dotare integer-valued functions. These functions have the form

[0417] X_dot[i,j,k,l]=X_pixel[i,j,k]+l_xΔX_pulse Eq. 8.27

[0418] Y_dot[i,j,k,l]=Y_pixel[i,j,k]+l_yΔY_line Eq. 8.28

[0419] wherein the X_pixeland Y_pixelfunctions are related to the pixels' positional distribution. The range limits on the pulse indices l_xand l_yin Eqs. 8.27, 28 are

[0420] 0<l_x<N_pulse/line Eq. 8.29

[0421] 0<l_y<N_line/scan Eq. 8.30

[0422] The positional coordinate offset from Pixel[{right arrow over (0)},{right arrow over (0)}] at the time of Pulse[{right arrow over (0)},{right arrow over (0)}] to Pixel[i,k] at Pulse[j,{right arrow over (0)}] is {X_pixel[i,j,k]G_x, Y_pixel[i,j,k]G_y} (FIG. 41). For the configuration shown in FIG. 39 (i=j={right arrow over (0)}), the X_pixeland Y_pixelfunctions are specified as

[0423] X_pixel[{right arrow over (0)},{right arrow over (0)},k]=k_xΔX_pixel+k_yΔX_offset Eq. 8.31

[0424] Y_pixel[{right arrow over (0)},{right arrow over (0)},k]=k_yΔY_pixel Eq. 8.32

[0425] A printhead's pixels are labeled by {k_x,k_y} indices that are within the following range limits,

[0426] 0≦k_y<N_row/head Eq. 8.33

[0427] 0≦k_xΔX_pixel+k_yΔX_offset<N_pixel/rowΔX_pixel Eq. 8.34

[0428] wherein N_pixel/rowis the number of pixels per printhead pixel row and N_row/headis the number of pixel rows per printhead (e.g., N_pixel/row=5 and N_row/head=8 in FIG. 39). The total number of pixels per printhead, N_pixel/head, is

[0429] N_pixel/head=N_pixel/rowN_row/head Eq. 8.35

[0430] It may be convenient, from the perspective of design simplicity, to choose the number of pixel rows to be a whole number of periods, i.e., N_row/headis defined as

[0431] N_row/head=N_period/headN_row/period Eq. 8.36

[0432] wherein N_period/headis an integer (the number of Y periods per printhead; e.g., N_period/head=4 in FIG. 39). However, this is not a necessary design requirement, and N_period/headcan optionally be fractional.

[0433] The nominal printhead aperture dimensions W_xand W_y(FIGS. 3, 39) are

[0434] W_x=ΔX_fieldG_x Eq. 8.37

[0435] W_y=ΔY_fieldG_y Eq. 8.38

[0436] wherein

[0437] ΔX_field=N_pixel/rowΔX_pixel Eq. 8.39

[0438] ΔY_field=N_row/headΔY_pixel Eq. 8.40

[0439] Each printhead exposes a scan field that is slightly larger than the printhead aperture. For example, FIG. 42 illustrates a particular printhead (Head[{right arrow over (0)}], represented by its aperture boundary 4201), which scans a particular field (Field[{right arrow over (0)},{right arrow over (0)}]). As the printhead raster-scans the field, its boundary 4201 sweeps across a “stitch zone” (indicated by the hatched area 4202) which borders the field. The stitch zone's width dimensions in the X and Y directions are ΔX_stitchG_xand ΔY_stitchG_ywherein

[0440] ΔX_stitch=ΔX_scan Eq. 8.41

[0441] ΔY_stitch=ΔY_scan Eq. 8.42

[0442] The scan field's respective X and Y dimensions (including the stitch zone) are W_x+ΔX_stitchG_xand W_y+ΔY_stitchG_y.

[0443] In a multi-scan printing process, W_xand W_yare also the X and Y displacements, respectively, between scan fields. For example, FIG. 43 illustrates a printhead (Head[{right arrow over (0)}]), which prints four scan fields. It first raster-scans Field[{right arrow over (0)},{0,0}], steps and scans Field[{right arrow over (0)},{1,0}], steps and scans Field[{right arrow over (0)},{1,1}], and then steps and scans Field[{right arrow over (0)},{0,1}]. Each particular scan provides only partial exposure coverage over the corresponding field's stitch zone, cf. FIG. 37. (The stitch zones are indicated as hatched areas in FIG. 43.) But over the areas where adjacent scan fields overlap (indicated by intersecting hatched areas) the separate scans seamlessly fill in each other's gaps.

[0444] The positional coordinates of Pixel[{right arrow over (0)},k] in Head[{right arrow over (0)}] at the time of Pulse[j,{right arrow over (0)}] (i.e., at the start of Scan[j]) are {X_pixel[{right arrow over (0)},j,k]G_x, Y_pixel[{right arrow over (0)},j,k]G_y}, wherein

[0445] X_pixel[{right arrow over (0)},j,k]=X_pixel[{right arrow over (0)},{right arrow over (0)},k]+j_xΔX_field Eq. 8.43

[0446] Y_pixel[{right arrow over (0)},j,k]=Y_pixel[{right arrow over (0)},{right arrow over (0)},k]+j_yΔY_field Eq. 8.44

[0447] In Eqs. 8.43, 44 the range limits on the scan indices j_xand j_yare

[0448] 0≦j_x<N_{x scan} Eq. 8.45

[0449] 0≦j_y<N_{y scan} Eq. 8.46

[0450] wherein N_{x scan}and N_{y scan}are the number of scans in the X and Y directions, respectively. The total number of scans, N_scan, is

[0451] N_scan=N_{x scan}N_{y scan} Eq. 8.47

[0452] FIG. 44 illustrates a multi-printhead printing process. The figure shows four printheads, Head[{0,0}], Head[{1,0}], Head[{0,1}], and Head[1,1}], at a particular time during Scan[{right arrow over (0)}]. During this scan the printheads raster-scan respective scan fields, Field[{0,0},{right arrow over (0)}], Field[{1,0},{right arrow over (0)}], Field[{0,1},{right arrow over (0)}], and Field[{1,1},{right arrow over (o)}]. Subsequent scans would seamlessly fill in the spaces between these fields (cf. FIG. 43). The positional displacements between printheads in the X and Y directions, respectively, are ΔX_headG_xand ΔY_headG_y, wherein

[0453] ΔX_head=N_{x scan}ΔX_field Eq. 8.48

[0454] ΔY_head=N_{y scan}ΔY_field Eq. 8.49

[0455] The positional coordinates of Pixel[i,k] in Head [i] at the time of Pulse[j,{right arrow over (0)}] are {X_pixel[i,j,k]G_x, Y_pixel[i,j,k]G_y}, wherein

[0456] X_pixel[i,j,k]=X_pixel[{right arrow over (0)},j,k]+i_xΔX_head Eq. 8.50

[0457] Y_pixel[i,j,k]=Y_pixel[{right arrow over (0)},j,k]+i_yΔY_head Eq. 8.51

[0458] Substituting from Eqs. 8.48, 49, 43, 44, 39, 40, 31, 32; Eqs. 8.50, 51 translate to

[0459] X_pixel[i,j,k]=(k_x+N_pixel/row(j_x+N_{x scan}i_x))ΔX_pixel+k_yΔX_offset Eq. 8.52

[0460] Y_pixel[i,j,k]=(k_y+N_row/head(j_y+N_{y scan}i_y))ΔY_pixel Eq. 8.53

[0461] These equations define the pixels', printheads', and scan fields' positional geometry; and with these substitutions Eqs. 8.27 and 8.28 define the complete printed exposure pattern.

[0462] A procedure for determining scan parameters from the above relationships is illustrated by the following example. (System design parameters from the preceding sections are assumed in this example, but the procedure can be applied equally well with different design assumptions.)

[0463] The printing grid step dimensions, G_xand G_y, should be approximately equal to the value G from Eq. 8.4,

[0464] G_x≅G_y≅{square root over (Q_dot/Q_A)}=6.86 nm Eq. 8.54

[0465] ΔX_pulseand ΔY_lineare selected to approximately satisfy the 150-nm proximity clearance specification (Eq. 8.5),

[0466] ΔX_pulse≅d_clear/G_x≅21.9 Eq. 8.55

[0467] ΔY_line≅d_clear/G_y≅21.9 Eq. 8.56

[0468] (The proximity clearance requirement actually only determines lower limits on ΔX_pulseand ΔY_line, but these parameters should preferably be close to the lower limits in order to minimize the scan velocity and the stitch zone area.) The following design condition is stipulated so that the pixel distribution forms a triangular centering pattern, as illustrated in FIGS. 3 and 39,

[0469] ΔX_pixel=2 ΔX_offset Eq. 8.57

[0470] It follows from Eqs. 8.57, 22, and 21 that

[0471] N_row/period=2 Eq. 8.58

[0472] ΔY_period=2 ΔY_pixel Eq. 8.59

[0473] The centering triangles, such as triangle 3903 in FIG. 39, are preferably approximately equilateral and oriented with one side of each triangle parallel to the Y axis, implying the condition

[0474] ΔX_pixelG_x≅{square root over (3)}ΔY_periodG_y Eq. 8.60

[0475] (The triangles could alternatively be oriented so that each has a side parallel to the X axis, but this would result in a design in which the number of lines per raster scan, N_line/scan, is much greater, and the frequency of scan reversals would consequently be much higher.) ΔY_periodis selected to approximately match the 6-μm lens centering dimension (Eq. 1.2),

[0476] ΔY_period≅d_lens/G_y≅874.6 Eq. 8.61

[0477] and this estimate, in combination with Eqs. 8.60, 54, 59, yields

[0478] ΔX_pixel≅{square root over (3)}ΔY_period≅1514.9 Eq. 8.62

[0479] ΔY_pixel=ΔY_period/2≅437.3 Eq. 8.63

[0480] Eqs. 1.4, 1.5, 8.54, 8.37-40, 8.62, and 8.63 yield

\begin{matrix} N_{pixel / row} = \frac{W_{x}}{G_{x} Δ X_{pixel}} ≅ 1924.5 & Eq.  8.64 \\ N_{row / head} = \frac{W_{y}}{G_{y} Δ Y_{pixel}} ≅ 6666.7 & Eq.  8.65 \end{matrix}

[0481] The above integer parameter estimates (Eqs. 8.55, 56, 61-65) need to be adjusted to make them exact integers. Also, the relative primality conditions, Eqs. 8.25, 26, need to be satisfied. These conditions hold if (and only if) there exist integers U_x, V_x, U_y, and V_yfor which

[0482] U_xΔX_pixel+V_xΔX_pulse=1 Eq. 8.66

[0483] U_yΔY_pixel+V_yΔY_line=1 Eq. 8.67

[0484] Since most of the design conditions need only be satisfied approximately, there is a multiplicity of acceptable parameter choices, and additional design conditions can be imposed to narrow the choices. The printed image's sensitivity to pixel defects can be further mitigated by constraining U_x, V_x, U_yand V_yto have the form

[0485] U_x=−1 Eq. 8.68

[0486] V_x=2P_x+1 Eq. 8.69

[0487] U_y=−4 Eq. 8.70

[0488] V_y=2P_y+1 Eq. 8.71

[0489] wherein P_xand P_yare integers. Since the terms on the left sides of Eqs. 8.66, 67 would typically be much greater than 1, it follows from Eqs. 8.57, 59, 66-71 that

\begin{matrix} \frac{Δ X_{offset}}{Δ X_{pulse}} = \frac{Δ X_{pixel}}{2 Δ X_{pulse}} ≅ - \frac{V_{x}}{2 U_{x}} = P_{x} + \frac{1}{2} & Eq.  8.72 \\ \frac{Δ Y_{period}}{Δ Y_{line}} = \frac{2 Δ Y_{pixel}}{Δ Y_{line}} ≅ - \frac{2 V_{y}}{U_{y}} = P_{y} + \frac{1}{2} & Eq.  8.73 \end{matrix}

[0490] The above conditions are stipulated to avoid printing proximate image dots with adjacent pixels, in order to minimize the effect of pixel defects that straddle adjacent pixels. Eq. 8.72 implies that the X displacement (ΔX_offsetG_x) between diagonally adjacent pixels, such as Pixel[{right arrow over (0)},{0,0}] and Pixel[{right arrow over (0)},{0,1}] in FIG. 39, is approximately equal to the scan distance per pulse (ΔX_pulseG_x) times an odd multiple of ½. Thus the dot arrays printed by these pixels are approximately symmetrically interleaved in the X direction, and the minimum separation distance between the image dots in these two dot arrays is approximately ½ΔX_pulseG_x(i.e., 75 nm; Eqs. 8.54, 55). Similarly, it follows from Eq. 8.73 that pixels such as Pixel[{right arrow over (0)},{0,0}] and Pixel[{right arrow over (0)},{−1,2}], which are adjacent along the Y direction (FIG. 39), have a separation distance (ΔY_periodG_y) that is approximately equal to the scan line spacing (ΔY_lineG_y) times an odd multiple of ½. Thus the dot arrays printed by these pixels are approximately symmetrically interleaved in the Y direction, and hence the dots in these arrays have a minimum separation distance of approximately ½ΔY_lineG_y(i.e., also 75 nm; Eqs. 8.54, 56).

[0491] Given any selected values for ΔX_pulseand P_x, Eqs. 8.68, 69, and 66 yield a solution for ΔX_pixelthat is integer-valued; however it will not necessarily be even (as required by Eq. 8.57). ΔX_pulsemust be odd to ensure that ΔX_pixelis even, i.e.,

[0492] ΔX_pulse=2Q_x+1 Eq. 8.74

[0493] for some integer Q_x. The ΔY_pixelsolution obtained from Eqs. 8.70, 71, and 67 will not be integer-valued unless ΔY_lineis odd; and furthermore, the quantity ½(ΔY_line−1)+P_ymust be even. Thus, ΔY_linemust be of the form

[0494] ΔY_line=2(2Q_y−P_y)+1 Eq. 8.75

[0495] for some integer Q_y.

[0496] Substituting Eqs. 8.68-71 into Eqs. 8.66, 67, the relative primality conditions become

[0497] −ΔX_pixel+(2P_x+1)ΔX_pulse=1 Eq. 8.76

[0498] −4ΔY_pixel+(2P_y+1)ΔY_line=1 Eq. 8.77

[0499] Estimated values for P_x, P_y, Q_x, and Q_yare obtained from Eqs. 8.74-77, 55, 56, 62, and 63,

\begin{matrix} P_{x} = \frac{1}{2} (\frac{1 + Δ X_{pixel}}{Δ X_{pulse}} - 1) ≅ 34.2 & Eq.  8.78 \\ Q_{x} = \frac{1}{2} (Δ X_{pulse} - 1) ≅ 10.4 & Eq.  8.79 \\ P_{y} = \frac{1}{2} (\frac{1 + 4 Δ Y_{pixel}}{Δ Y_{line}} - 1) ≅ 39.5 & Eq.  8.80 \\ Q_{y} = \frac{1}{2} (\frac{1}{2} (Δ Y_{line} - 1) + P_{y}) ≅ 25.0 & Eq.  8.81 \end{matrix}

[0500] Based on the above estimates, the following design choices are made,

[0501] P_x=34 Eq. 8.82

[0502] Q_x=11 Eq. 8.83

[0503] P_y=40 Eq. 8.84

[0504] Q_y=25 Eq. 8.85

[0505] and from these the following parameter values are obtained,

[0506] ΔX_pulse=2Q_x+1=23 Eq. 8.86

[0507] ΔY_line=2(2Q_y−P_y)+1=21 Eq. 8.87

[0508] ΔX_pixel=(2P_x+1)ΔX_pulse−1=1586 Eq. 8.88

[0509] Δ_Y_pixel=((2P_y+1)ΔY_line−1)/4=425 Eq. 8.89

[0510] The estimated value of N_pixel/row(Eq. 8.64) is refined using Eqs. 8.88, 8.54, and 1.4,

\begin{matrix} N_{pixel / row} = \frac{W_{x}}{G_{x} Δ X_{pixel}} ≅ 1838.2 & Eq.  8.90 \end{matrix}

[0511] Based on this estimate, the following design choice is made,

[0512] N_pixel/row=1838 Eq. 8.91

[0513] and the following exact value for G_xis obtained (Eqs. 8.64, 1.4, 8.88, 8.91),

\begin{matrix} G_{x} = \frac{W_{x}}{N_{pixel / row} Δ X_{pixel}} = 6.861 nm & Eq.  8.92 \end{matrix}

[0514] The wafer-plane exposure dose is proportional to the product G_xG_y; i.e., if the printing grid cells are not square Eq. 8.3 generalizes to

\begin{matrix} Q_{A} = \frac{Q_{dot}}{G_{x} G_{y}} & Eq.  8.93 \end{matrix}

[0515] Thus, to maintain the design exposure dose the following revised estimate of G_yis obtained, using the exact G_xvalue from Eq. 8.92,

\begin{matrix} G_{y} ≅ \frac{Q_{dot}}{G_{x} Q_{A}} = {(6.86 nm)}^{2} / G_{x} = 6.859 nm & Eq.  8.94 \end{matrix}

[0516] The estimated value of N_row/head(Eq. 8.65) is revised using Eqs. 8.94, 8.89, and 1.5,

\begin{matrix} N_{row / head} = \frac{W_{y}}{G_{y} Δ Y_{pixel}} ≅ 6860.8 & Eq.  8.95 \end{matrix}

[0517] Based on this estimate, the following design choice is made,

[0518] N_row/head=6860 Eq. 8.96

[0519] (By choosing N_row/headto be an even integer, N_period/headis integer-valued, Eqs. 8.36, 8.58.) The following exact value for G_yis obtained (Eqs. 8.65, 1.5, 8.89, 8.96),

\begin{matrix} G_{y} = \frac{W_{y}}{N_{row / head} Δ Y_{pixel}} = 6.860 nm & Eq.  8.97 \end{matrix}

[0520] In the scan sequence illustrated in FIG. 5 (i.e., scan positions 501a, 501b, 501c, and 501d), the number of scans in the X direction and the number in the Y direction are both 2,

[0521] N_{x scan}=2 Eq. 8.98

[0522] N_{y scan}=2 Eq. 8.99

[0523] Combining these values and the preceding results, the lens-scan design parameters are summarized in FIG. 45.

[0524] The 65 wafer per hour throughput estimate in §7 can be refined, based on the above design parameters. The lens center spacing with these parameters does not exactly match the 6-μm nominal design value (Eq. 1.2), reducing the lens fill factor η_FFfrom 0.630 (Eq. 1.3) to 0.619. This efficiency loss results in a 1.8-% throughput reduction; also the throughput estimate should take into account the throughput overhead associated with scan reversal and wafer loading.

[0525] The scan time t_linefor each raster line is

[0526] t_line=N_pulse/line/ν_src=264. msec Eq. 8.100

[0527] and the scan time for a complete raster scan, t_scan, is

[0528] t_scan=N_line/scant_line=112.3 sec Eq. 8.101

[0529] This estimate does not include the time associated with scan reversal between raster lines. The scan velocity v_scanis

[0530] ν_scan=ΔX_pulseG_xν_src=0.95 mm/sec Eq 8.102

[0531] Assuming a 1-g reversal acceleration (i.e., 10 m/sec²), the velocity reversal time would be 0.19 msec. Also, the printhead must step in the Y direction by a distance of ΔY_lineG_y=144 nm between line scans. Again assuming 1-g acceleration, the step time would be approximately 0.24 msec. Thus, the expected reversal time overhead would be of order 1 msec or less per scan line. This would only add approximately 0.4 sec to the raster scan time, t_scan.

[0532] The wafer stepping time between scans should also be accounted for in the throughput calculation, but at 1-g acceleration the wafer could be stepped by 20 mm in less than 100 msec, so this is not a significant throughput factor. Thus the wafer cycle time (at 4 scans per cycle) can be approximately estimated at 452 sec (i.e., 113 sec per scan, times 4 scans), not including the wafer load time. Allowing 10 sec for swapping in each new wafer, the total wafer cycle time is estimated at 462 sec. With 8 wafers being printed per cycle, this translates to 62 wafers per hour.

[0533] The scanning system design illustrated above exemplifies a general design methodology that can be applied with different design assumptions and requirements. For example, the number of pixel rows per printhead (N_row/head) and the number of pixels per row (N_pixel/row) could be constrained to be integer powers of 2, in order to simplify the digital electronics design. This could be accomplished by increasing the printhead size in the above design to accommodate 8192 pixel rows with 2048 pixels per row. This would increase the printhead aperture dimensions to W_x=22.3 mm and W_y=23.9 mm. However, the first-stage lens diameter (D₁) and lens center spacing (d_lens) could be then adjusted, with N_row/headand N_pixel/rowheld constant, to scale either W_xor W_yto a desired design value. Also, although the pixel centering pattern is preferably based on approximately equilateral triangles, this is not an absolute design requirement and the pixel centering grid could be modified to control the printhead aperture aspect ratio (W_y/W_x) without changing N_row/heador N_pixel/row.

[0534] The above design outline applies to the lens-scan method in which the wafer remains stationary while the printheads raster scan over the scan fields. In the alternative wafer-scan method, the wafer moves while the printheads remain substantially stationary (except for slight focus and alignment adjustments). Due to the high inertia of the wafer and scanning mechanism, the wafer-scan method would preferably comprise a transitionless scan that does not require frequent scan reversals.

[0535] The transitionless scan method, which was described previously in the context of one-dimensional scanning, is also applicable in two dimensions. In this context, the method can be applied in either of two modes termed “transitionless X scanning” or “transitionless Y scanning”. (Both modes can be used in combination.) The defining conditions for these scanning modes are

[0536] N_pixel/row=ΔX_pulse(transitionless X scan) Eq. 8.103

[0537] N_row/head=ΔY_line(transitionless Y scan) Eq. 8.104

[0538] (cf. Eq. 8.11).

[0539] For the transitionless X scanning mode, N_{x scan}can be fractional, and ΔX_headis defined by

[0540] ΔX_head=N_{x pulse}ΔX_pulse Eq. 8.105

[0541] wherein N_{x pulse}is the total number of exposure pulses per dot row. The individual scans, Scan[{1,j_y}], Scan[{1,j_y}], etc. can (for each particular index j_y) be considered to constitute portions of a single “transitionless X scan”, which is designated as Scan′[{0,j_y)}] and is characterized by scan parameters N′_pulse/line, N′_pulse/scan, ΔX′_scan, ΔX′_field, and N′_{x scan}defined by

[0542] N′_pulse/line=N_{x pulse} Eq. 8.106

[0543] N′_pulse/scan=N′_pulse/lineN_line/scan Eq. 8.107

[0544] ΔX′_scan=ΔX′_field=ΔX_head Eq. 8.108

[0545] N_{x scan}=1 Eq. 8.109

[0546] (Other entities that are associated with a transitionless X scan can also be distinguished by using primed symbols. For example, Dots′[i,j,k] represents the set of image dots printed by Pixel[i,k] during Scan′[j].)

[0547] Similarly, for transitionless Y scanning, N_{y scan}can be fractional, and ΔY_headis defined by

[0548] ΔY_head=N_{y line}ΔY_line Eq. 8.110

[0549] wherein N_{y line}is the total number of scan lines covered by all the scans, Scan[{j_x,0}], Scan[{j_x,1}], etc. (for any particular index j_x). These scans constitute a “transitionless Y scan”, Scan′[{j_x,0}] which is characterized by scan parameters N′_line/scan, N′_pulse/scan, ΔY′_scan, ΔY′_field, and N′_{y scan}defined by

[0550] N′_line/scan=N_{y line} Eq. 8.111

[0551] N′_pulse/scan=N_pulse/lineN′_line/scan Eq. 8.112

[0552] ΔY′_scan=ΔY′_field=ΔY_head Eq. 8.113

[0553] N′_{y scan}=1 Eq. 8.114

[0554] If Eqs. 8.103 and 8.104 both hold, all of the scans can be merged into a single transitionless scan, Scan′[{0,0}], and Eqs. 8.105, 106, 108-111, 113, 114 all apply. For this case, the total number of pulses per scan, N′_pulse/scan, is

[0555] N′_pulse/scan=N_pulse/lineN_line/scan Eq. 8.115

[0556] The distinction between a transitionless X scan, Scan′_X[{0,j_y}], and its constituent scans Scan[{0,j_y}], Scan[{1,j_y}], . . . is not merely one of definitions and notation. The constituent scans' raster lines are merged in the composite, transitionless scan, thus reducing the number of scan reversals and associated throughput overhead. This is illustrated in FIGS. 46A and 46B. FIG. 46A shows several sets of image dots, Dots[{right arrow over (0)},{0,0},{right arrow over (0)}], Dots[{right arrow over (0)},{1,0},{right arrow over (0)}], etc., that are printed by Pixel[{right arrow over (0)},{right arrow over (0)}] during respective scans, Scan[{0,0}], Scan[{1,0}], etc. Each set is raster-scanned, and the arrows indicate the exposure time sequence in each scan. Assuming that Eq. 8.103 holds, the X-displacement between the each scan's end point (e.g., point 4601) and the succeeding scan's start point (e.g., point 4602) is equal to ΔX_pulseG_x, the scan distance per pulse. Hence, the multiple raster scans can be merged into a single, transitionless scan in which Pixel[{right arrow over (0)},{right arrow over (0)}] traverses all the sets, as illustrated in FIG. 46B. (The union of Dots[{right arrow over (0)},{0,0},{right arrow over (0)}], Dots[{right arrow over (0)},{1,0},{right arrow over (0)}], etc. is indicated as Dots′[{right arrow over (0)},{0,0},{right arrow over (0)}] in FIG. 46B, in conformance with the notational conventions outlined above.)

[0557] A fundamental limitation of the transitionless scan method is that the scan fields' stitch zones are very large—in fact, they are as large as the printhead apertures themselves. For a transitionless X scan

[0558] ΔX_stitch=ΔX_pixelΔX_pulseW_x×G_x Eq. 8.116

[0559] and for a transitionless Y scan

[0560] ΔY_stitch=ΔY_pixelΔY_line=W_y/G_y Eq. 8.117

[0561] (from Eqs. 8.41, 42, 19, 20, 23, 24, 103, 104, 39, 40, 37, 38). This results in a significant efficiency loss, either in terms of throughput or wafer coverage, due to incompletely exposed stitch zones bordering the wafer print module. The efficiency loss can be mitigated by minimizing the printhead aperture dimension in the direction of the transitionless scan. For example, in the following design illustration a transitionless X scan is used and the printhead aperture dimensions are specified as

[0562] W_x=10 mm Eq. 8.118

[0563] W_y=20 mm Eq. 8.119

[0564] wherein the smaller X dimension (W_x) is chosen to limit the efficiency loss.

[0565] FIG. 47 illustrates the wafer-scan method, which is based on the transitionless X scan described above. (For comparison, the lens-scan method is illustrated in FIG. 5.) A wafer print module 4701 comprises a set of printhead apertures such as aperture 4702 whose dimensions in the respective X and Y directions are W_xand W_y. In this example, the apertures are arranged in a rectangular array with X and Y periodicities of 2W_xand 2W_y, respectively. (In general, the X periodicity could be any integer multiple of ΔX_pulseG_x, and the Y periodicity could be any integer multiple of W_y. The illumination system would be designed to accommodate the aperture geometry, based on the design methods described in §6.) The wafer 108 is exposed in two scans. It is first exposed as it raster-scans between positions 4703aand 4703b, covering an X travel range of 2W_x. (The Y travel range is very small, e.g., 1.5 μm in this example.) It then steps by a distance of W_yin the Y direction and is exposed while raster-scanning between positions 4703cand 4703d. (This contrasts with the lens-scan method in which the wafer remains stationary during exposure.)

[0566] The pixel layout design used in the wafer-scan method is similar to a design configuration described in Ref. 44, except that the present method uses scan interleaving to minimize the effect of pixel defects, and it also better optimizes the pixel fill factor (Eq. 1.3). FIG. 48 conceptually illustrates the pixel distribution over a printhead aperture 4801. (The dimensions and proportions shown in the figure do not correspond to the actual design values derived below.) The pixel array comprises a set of closely spaced pixel rows, such as row 4802 comprising Pixel[{0,0}], Pixel[{1,0}], Pixel[{2,0}] and row 4803 comprising Pixel[{0,1}], Pixel[{1,1}], Pixel[{2,1}]. The pixels in each row are very sparsely spaced, with a pixel centering separation of ΔX_pixelG_x. The spacing is necessarily sparse because the number of pixels per row, N_pixel/row, is equal to ΔX_pulse(Eq. 8.103), which must have a small value because the scan velocity is proportional to ΔX_pulse(Eq. 8.102). Adjacent pixel rows are separated by a Y displacement distance of ΔY_pixelG_yand an X offset distance of ΔX_offsetG_x. The pixel array is periodic in the Y direction with a periodicity dimension of ΔY_periodG_y. (The above parameter definitions are consistent with the corresponding lens-scan parameter definitions, as illustrated in FIG. 39.)

[0567] The printing grid dimensions G_xand G_y, and the scan parameters ΔX_pulseand ΔY_line, are defined by the same conditions as in the lens-scan design (Eqs. 8.54-56; cf. FIG. 39), and they have the approximate values

[0568] G_x≅G_y≅−6.86 nm Eq. 8.120

[0569] ΔX_pulse≅ΔY_line≅−21.9 Eq. 8.121

[0570] ΔX_pixelcan be estimated from Eqs. 8.103, 39, 37, 118, and 120,

[0571] ΔX_pixel=W_x/(G_xΔX_pulse)≅66666.7 Eq. 8.122

[0572] The array's X periodicity dimension (ΔX_pixelG_x) is two orders of magnitude larger than the pixel dimensions. However, the pixels are centered at the points of a periodic tiling pattern comprising unit cells, such as cell 4804 in FIG. 48, which have dimensions comparable to the pixel size. FIG. 49 shows a magnified view of four pixels 4901, 4902, 4903, and 4904, which are centered at the points of unit cell 4804. Pixels 4901 and 4902 have the same X coordinate, and their centering distance in the Y direction is ΔY_periodG_y. ΔY_periodhas the same estimated value as in the lens-scan design (Eq. 8.61),

[0573] ΔY_period≅d_lens/G_y≅874.6 Eq. 8.123

[0574] The X displacement distance from pixels 4901 and 4902 to both pixel 4903 and pixel 4904 is C G_x, wherein

[0575] C=GCD[ΔX_pixel, ΔX_offset] Eq. 8.124

[0576] (It follows from Eq. 8.31 that all pixels' X coordinates, X_pixel[{right arrow over (0)},{right arrow over (0)},k] are integer multiples of C G_x. Pixels 4903 and 4904 are the specific pixels that have Y coordinates between the Y coordinates of pixels 4901 and 4902, and that are closest in X to pixels 4901 and 4902.) The unit cell 4804 forms a parallelogram whose sides are preferably approximately equal to the centering distance between pixels 4901 and 4902 (i.e., cell 4804 comprises two approximately equilateral triangles). Thus C has the approximate value

\begin{matrix} C ≅ \frac{\sqrt{3}}{2} Δ Y_{period} G_{y} / G_{x} ≅ 757.5 & Eq.  8.125 \end{matrix}

[0577] The number of pixel rows per period in the Y direction, N_row/period, is

[0578] N_row/period=ΔX_pixel/C Eq. 8.126

[0579] (cf Eq. 8.22). A related parameter, M, is defined by

[0580] M=ΔX_offset/C Eq. 8.127

[0581] It follows from Eqs. 8.124, 126, 127 that N_row/periodand M are relatively prime,

[0582] GCD[N_row/period, M]=1 Eq. 8.128

[0583] Hence there exist unique integers K_xand K_ysatisfying the conditions

[0584] K_xN_row/period+K_yM=1; 0<K_y<N_row/period Eq. 8.129

[0585] The Y offset between pixels 4903 and pixel 4904 is ΔY_slopeG_y(FIG. 49), wherein

[0586] ΔY_slope=(N_row/period−2K_y)ΔY_pixel Eq. 8.130

[0587] (This follows from Eq. 8.32. For example, if pixel 4901 is designated as Pixel[{right arrow over (0)},{right arrow over (0)}], then pixel 4903 has position indices {k_x,k_y}={K_x,K_y} in Eqs. 8.31, 32, and pixel 4904 has position indices {k_x,k_y}={−M −K_x, N_row/period−K_y}, from which Eq. 8.130 follows.) The scanning system design should preferably minimize ΔY_slopeso that the sides of unit cell 4804 are approximately equal.

[0588] N_row/periodis estimated from Eqs. 8.126, 122, and 125,

[0589] N_row/period=ΔX_pixel/C≅88.0 Eq. 8.131

[0590] Based on this estimate and Eqs. 8.21 and 8.123, ΔY_pixelis estimated as

[0591] ΔY_pixel=ΔY_period/N_row/period≅9.9 Eq. 8.132

[0592] The following design values are selected for ΔY_pixeland ΔY_lineto fit Eqs. 8.132 and 8.121 and satisfy the relative primality condition (Eq. 8.26),

[0593] ΔY_pixel=10 Eq. 8.133

[0594] ΔY_line=23 Eq. 8.134

[0595] Working backward through the above estimates (Eqs. 8.132, 126, and 122), the selected ΔY_pixelvalue results in a revised estimate for ΔX_pulsethat is almost exactly integer-valued,

[0596] N_row/period=ΔY_period/ΔY_pixel≅87.5 Eq. 8.135

[0597] ΔX_pixel=N_row/periodC≅66249.9 Eq. 8.136

[0598] ΔX_pulse=W_x/(G_xΔX_pixel)=22.0 Eq. 8.137

[0599] Thus, the following design choice is made,

[0600] ΔX_pulse=22 Eq. 8.138

[0601] With ΔX_pulsedetermined, the ΔX_pixeland N_row/periodestimates (Eqs. 8.122, 126) are again revised (although in this example the selected value for ΔX_pulseis so close to Eq. 8.137 that the revised values are practically unchanged),

[0602] ΔX_pixel=W_x/(G_xΔX_pulse)=66260.3 Eq. 8.139

[0603] N_row/period=ΔX_pixel/C≅87.5 Eq. 8.140

[0604] Based on Eq. 8.140 the following design choice is made,

[0605] N_row/period=87 Eq. 8.141

[0606] and ΔY_periodis obtained from Eq. 8.21, 133, and 141

[0607] ΔY_period=N_row/periodΔY_pixel=870 Eq. 8.142

[0608] The estimated value of C (Eq. 8.125) is adjusted to fit Eq. 8.126, using the above selection for N_row/periodand the revised estimate for ΔX_pixelfrom Eq. 8.139

[0609] C=ΔX_pixel/N_row/period≅761.6 Eq. 8.143

[0610] Thus the following design selection is made

[0611] C=761 Eq. 8.144

[0612] and an exact value for ΔX_pixelis obtained from Eq. 8.126

[0613] ΔX_pixel=N_row/periodC=66207 Eq. 8.145

[0614] The selected values for N_row/periodand C (Eqs. 8.141, 144) are both relatively prime with ΔX_pulse(Eq. 8.138); hence so is ΔX_pixel, as required by Eq. 8.25.

[0615] ΔX_offsetis in the range

[0616] 0<ΔX_offset<ΔX_pixel Eq. 8.146

[0617] Hence the M parameter (Eq. 8.127) is in the range

[0618] 0<M<ΔX_pixel/C=87 Eq. 8.147

[0619] The selection criteria for choosing M are that it should be within the above range; it must be relatively prime with N_row/period(Eq. 8.128); and subject to these constraints M is selected to minimize |ΔY_slope| (Eq. 8.130). The purpose of the minimization condition is to minimize the Y offset between pixels 4903 and 4904 in FIG. 49, so that the sides of parallelogram 4804 are as nearly equal as possible. Based on the above selection criteria, the following value for M is chosen

[0620] M=2 Eq. 8.148

[0621] and ΔX_offsetis determined by Eqs. 8.127, 144, and 148

[0622] ΔX_offset=M C=1522 Eq. 8.149

[0623] The K_xand K_yvalues determined from Eq. 8.129 are

[0624] K_x=1 Eq. 8.150

[0625] K_y=44 Eq. 8.151

[0626] and ΔY_slopeis obtained from Eqs. 8.130, 133, 141, and 150,

[0627] ΔY_slope=(N_row/period−2K_y)ΔY_pixel=−10 Eq. 8.152

[0628] (Actually, if ΔX_pulseis even the selection M=2 will always minimize |ΔY_slope|, and ΔY_slopewill be equal to −ΔY_pixel.)

[0629] N_pixel/rowis obtained from Eqs. 8.103, 138; and an estimate for N_row/headis obtained from Eqs. 8.40, 38, 119, 120, and 133,

\begin{matrix} N_{pixel / row} = Δ X_{pulse} = 22 & Eq.  8.153 \\ N_{row / head} = \frac{W_{y}}{G_{y} Δ Y_{pixel}} ≅ 291545.2 & Eq.  8.154 \end{matrix}

[0630] Thus N_row/headis defined as

[0631] N_row/head=291545 Eq. 8.155

[0632] and exact design values for G_xand G_yare obtained (from Eqs. 8.37-40),

\begin{matrix} G_{x} = \frac{W_{x}}{N_{pixel / row} Δ X_{pixel}} = 6.866 nm & Eq.  8.156 \\ G_{y} = \frac{W_{y}}{N_{row / head} Δ Y_{pixel}} = 6.860 nm & Eq.  8.157 \end{matrix}

[0633] For this design example the N_{x pulse}factor in Eqs. 8.105, 106 (the total number of pulses per scan line) is defined as

[0634] N_{x pulse}=N_pulse/line=2ΔX_pixel=132414 Eq. 8.158

[0635] (In practice N_{x pulse}could have any integer value that is sufficiently large to satisfy mechanical clearance constraints between printheads.) From Eqs. 8.105 and 8.108,

[0636] ΔX′_scan=ΔX′_field=ΔX_head=N_{x pulse}ΔX_pulse=2913108 Eq. 8.159

[0637] The number of scans in the X direction is 1 (Eq. 8.109), and the number of scans in the Y direction is

[0638] N_{y scan}=2 Eq. 8.160

[0639] FIG. 50 summarizes the design parameters for the wafer-scan method.

[0640] The wafer-scan method is inherently insensitive to pixel defects that straddle adjacent pixels. For example, the dot rows printed by pixel 4901 in FIG. 49 are not proximate to dot rows printed by any adjacent pixels such as pixels 4902, 4903, or 4904 because the Y displacement between adjacent pixels is much larger than the entire Y scan range (ΔY_scanG_y). Thus, defects that straddle adjacent pixels will not significantly affect the image. The scan pattern could also be designed to minimize sensitivity to defects that straddle pixels such as pixels 4903 and 4904 that are not directly adjacent, but which print proximate dot rows. For example, if the selected C value (Eq. 8.144) were increased slightly from 761 to 765, the dot columns printed by pixels 4903 and 4904 would be approximately symmetrically interleaved in the X direction with a minimum separation distance of approximately 75 nm, thereby minimizing the effect of defects that straddle these pixels.

[0641] Several improvements and variations of the above-outlined scanning methods are possible. For example, the system would not necessarily need to use the same raster scan pattern for all printing processes. Scan parameters such as N_pulse/line, ΔX_pulse, N_line/scan, ΔY_line, G_x, and G_ycould be adjusted for a particular process to optimize print resolution, exposure level, and throughput. The relative primality conditions (Eqs. 8.25, 26) could perhaps be relaxed somewhat, although the scan parameters should, at a minimum, be defined so that the exposure pattern produced with all the pixels ON does not have any structure within the system's optical resolution limit. This criterion implies the following conditions,

[0642] GCD[ΔX_pixel,ΔX_pulse]≦λ/(2NA_geoG_x) Eq. 8.161

[0643] GCD[ΔY_pixel,ΔY_line]≦λ/(2NA_geoG_y) Eq. 8.162

[0644] wherein λ is the exposure wavelength (11.3 nm) and NA_geois the geometric numerical aperture defined in §2 (NA_geo=0.156). (The optical resolution limit is discussed in §10.)

[0645] Although the above methods are based on a pulsed EUV source such as a xenon LPP source, the methods can also be adapted with very little change for use with a continuous source such as a synchrotron. (In this case, the pixels would expose continuous scan lines rather than discrete image dots, and proximate pixels' scan lines would be overlapped rather than interleaved in the X direction.) The raster scanning technique is suitable for general-purpose digital printing applications, but for some specialized applications the system could use an alternative vector-scanning method in which the scan path follows the lines and contours of the printed image. Vector scanning could be used, for example, when the print image is a periodic pattern matching the pixel layout. (Such applications would not typically require individual pixel modulators. Instead the system could simply use a single modulator at the EUV source, as outlined at the end of §4.)

§9 Data Compression

[0646] A maskless EUV lithography system would have enormous data storage and data flow requirements. For example, a fully buffered bit map image at a 6.86-nm sampling resolution (i.e., the grid step dimension G, Eq. 8.4) would require 2×10¹²data bits per cm²of wafer area. 2×10¹²bits corresponds to approximately $1 million of DRAM storage at current prices; thus the data would have to be converted to a compressed format that can be efficiently decompressed at high data rates. The system's total data rate for pattern generation is 2.8×10¹³bits per sec (i.e., N_pixelν_src; Eqs. 1.7, 8.1). However, if all eight wafer print modules are supplied data from a common data stream the data rate would be (only) 3.5×10¹²bits per sec.

[0647] Perhaps the most practical and straightforward approach to data compression would be to convert the image data into a sequence of machine instructions that can be processed by digital hardware to regenerate the image bit stream at a high data rate. Typical lithography design patterns comprise simple rectangular shapes and periodic structures that can be very compactly represented in terms of repeat loops and subroutines, so high compression rates could be achievable with this approach. Assuming, for example, that the system uses multiple processing units, each of which generates a 32-bit data stream at a 100 MHz data rate, the total output from each unit would be 3.2×10⁹bits per sec. (The instruction feed rate could be orders of magnitude lower, depending on the level of data compression that is achieved.) Approximately 1000 processors could satisfy the system's data flow requirement. This assumes data sharing between modules; otherwise the required number of processors would be of order 10,000.

[0648] Although the modules may receive data from a common data stream, the system could nevertheless provide a limited degree of module-specific customization of the print image. Each module could have an associated database of “patch” image patterns that are spliced into the bit stream as it is received. This facility would be used, for example, to compensate for defective pixels or to insert wafer die serialization characters into the image.

[0649] With the lens-scan technique described in §8, each printhead pixel row could be provided a dedicated line buffer that is filled prior to each line scan. Each raster scan comprises 425 line scans (i.e., N_line/scan=425, FIG. 45); thus less than 0.3% (i.e., {fraction (1/425)}) of the entire wafer bit image would be buffered. With the wafer-scan technique the line buffering approach is not very practical because N_line/scanis much lower (10) and the number of pixel rows per printhead (N_row/head) is much higher (291545 for the wafer scan, compared to 6860 for the lens scan; FIGS. 45, 50). But the number of pixels per row (N_pixel/row) is much lower (22, versus 1838 for the lens scan), so it could be feasible to provide each pixel a small data buffer and an associated instruction pointer and return stack for instruction processing. Each processing unit would be time-shared between multiple pixels, keeping their data buffers filled as the scan progresses. Since the number of grid steps per pulse (ΔX_pulse) is 22, the processors need only transmit every 22nd image bit to the bit stream; however the instruction feed rate and processing rate would be 22 times higher than with the lens-scan method. Also, the processor instruction set should preferably be designed so that it can be processed in either forward or reverse sequential order to accommodate bi-directional raster scanning. For example, rather than terminating subroutines with return instructions, branch instructions could include both the starting and ending addresses of subroutine instruction blocks, which may be processed in either forward or reverse order. (An advantage of this branching method is that subroutine blocks could be overlapped, further improving data compression efficiency.)

§10 Printing Performance

[0650] FIG. 51 illustrates a radial point exposure intensity profile 5101 for a single, isolated image dot, based on the imaging performance of the lens design outlined in FIG. 13 (with focus optimized as indicated by the footnote in FIG. 13). This “point exposure profile” is identical to the intensity profile I[r₃] illustrated in FIG. 12, except that the vertical scale is normalized to units of the maximum attainable exposure level, Q_A=80 mJ/cm²(Eq. 7.6). On this scale, the profile peaks at 0.0134 (corresponding to 1.075 mJ/cm₂in absolute units). Also, FIG. 51 illustrates a cross-sectional “line exposure profile” 5102 of an isolated dot row comprising dot exposures distributed along a line at the grid step interval of G=6.86 nm (Eq. 8.4). This profile peaks at 0.116 in Q_A-relative units (or 9.32 mJ/cm²in absolute units).

[0651] On the scale of FIG. 51 the assumed resist exposure threshold (20 mJ/cm²) is at the 0.25 level. The point and line exposure profiles 5101 and 5102 are both well below this level, so the requisite exposure level can only be attained by combining multiple overlapping dot exposures. FIGS. 52-55 illustrate several such dot patterns that could be used to form basic semiconductor elements such as conductor lines and contact holes, and corresponding exposure profiles are shown in FIGS. 56 and 57. The resolution parameter k₁for all of these images is 0.6, based on an operating wavelength λ=11.3 nm, an effective numerical aperture NA_eff=0.1, and imaging resolution RES=70 nm (Eq 2.14).

[0652] FIG. 52 illustrates a dot pattern exposing a positive-tone disk image, which would develop into an isolated hole in a positive resist. The illustrated dots such as dot 5201 in FIG. 52 (and FIGS. 53-55) represent the center points of image dots that are exposed with pixels in their ON state, while the “+” marks such as mark 5202 represent unexposed image dots. The dots are exposed on a square grid with unit cell dimension G=6.86 nm (Eq. 8.4). The 25% threshold contour (representing the exposure distribution's aerial image) is illustrated as curve 5203, which has a diameter of 70.9 nm.

[0653] FIG. 53 illustrates a complementary pattern comprising a negative-tone disk image, which would develop into an isolated mesa or post in a positive resist. The 25% contour 5301ahas a diameter of 69.5 mn.

[0654] FIG. 54 illustrates a positive-tone line image that would develop to an isolated space (i.e., trough) with a positive resist. The space dimension between the 25% contours 5401 is 69.8 nm. A complementary negative-tone line pattern shown in FIG. 55 has a line dimension of 71.4 nm between the 25% contours 5501. This pattern would develop to an isolated raised line (or ridge) in a positive resist. (Features of these dimensions would exhibit minimal proximity effects. For example, a simulated dense line/space pattern comprising features of the type illustrated in FIGS. 54 and 55, with a 137-nm pitch, exhibits an uncompensated iso-dense bias of less than 3 nm.)

[0655] FIGS. 56 and 57 are exposure profiles (in Q_A-relative units) through the X axes of FIGS. 52-55, showing the high contrast in the exposure distributions. Profiles 5601 and 5602 correspond to the FIG. 52 and FIG. 53 exposure patterns, respectively; and profiles 5701 and 5702 correspond respectively to FIG. 54 and FIG. 55.

[0656] The microlenses exhibit a useful frequency-filtering function that mitigates the effects of digitization and aliasing in the printed image. Each diffraction-limited image dot comprises spatial frequencies within a maximum frequency limit corresponding to a spatial period of λ/(2NA_geo)=36.3 nm (wherein λ=11.3 nm and NA_geois the geometric numerical aperture defined in §2, NA_geo=0.156), and a composite exposure comprising a linear superposition of dot exposures will similarly be frequency-limited. This dimension defines the imaging system's optical resolution limit (which is significantly lower than the practical printing resolution). The printed image cannot contain any frequency components comprising spatial periods below the 36.3-nm limit. For example, one edge of the dot exposure patterns in FIG. 54 has a dithering pattern 5402 with a periodicity of 4 G=27.4 nm, but the exposure contours 5401 are nevertheless (theoretically) perfectly straight because the 27.4-nm dimension is below the 36.3-nm resolution limit. Generally, any one-dimensionally-periodic exposure pattern with a periodicity below 36.3 nm would print with perfectly straight exposure contours; and a two-dimensionally periodic pattern (such as a checkerboard pattern) within this periodicity limit would produce a perfectly uniform exposure distribution. (In reality, there may be some line edge roughness and exposure nonuniformity due to shot noise, Ref. 20. But at the assumed 20-mJ/cm²exposure threshold the shot noise effect would probably not be too significant.)

[0657] A variety of periodic dot patterns can be formed with periodicities below the optical resolution limit, and such patterns can be used to effectively provide grayscale control, to form straight-edged diagonal features, and to provide very precise control over dimensions and placement of print features. “Digital dithering” techniques based on sub-resolution patterning can give the lithography designer a great deal of flexibility in controlling image feature size, shape, orientation, and placement.

§11 Tolerance Factors

[0658] The print quality is affected by a variety of tolerance factors, including defective or non-functional pixels, statistical fluctuations in the LPP pulses, thermally-induced image magnification errors, focus errors, and lens surface form errors. The cumulative effect of such factors on the positions and dimensions of print features must generally be an order of magnitude smaller than the printing resolution, and the effect of individual tolerance factors must typically be two orders of magnitude smaller. Thus, for a system with 70-nm print resolution capability, the allowable printing errors from individual factors would be of order 1 nm. However, the tolerance requirements can be relaxed somewhat by employing tolerance compensation mechanisms such as micromechanical centration actuators on the second-stage microlenses (FIG. 1).

[0659] The printing system is very tolerant of isolated pixel defects. For example, FIG. 58A illustrates how a printer pixel stuck in its ON state would affect the negative-tone disk image of FIG. 53. FIG. 58A is identical to FIG. 53, except that an extra image dot 5801 proximate to the 25% threshold contour 5301ahas been exposed by the defective pixel. This distorts the contour (as illustrated by the distorted contour 5301b), reducing its X dimension from a calculated value of 69.55 nm to 68.86 nm. This is only a 1% dimensional change. Moreover, if certain pixels are known to be defective, the dot pattern can be modified to compensate for the defects. For example, FIG. 58B illustrates the same dot pattern as FIG. 58A, except that two dots 5802 and 5803 are left unexposed to compensate for the extra dot 5801. The resulting threshold contour 5301chas an X dimension of 69.43 nm, and the dimensional error is an insignificant 0.11 nm.

[0660] Of course, if the defective pixel printed multiple dots clustered in a very small area its effect would be much greater. But the interleaved raster scan methods described in §8 ensure that the image dots that significantly overlap any particular wafer point will have all been exposed through different pixels, thereby avoiding any such clustering effect.

[0661] The image dots overlapping any particular wafer point will have also all been exposed by different LPP pulses. (Dots exposed during a single pulse are separated by at least 6 μm, the pixel center spacing; and hence they cannot overlap.) Thus the effect of statistical variations in pulse intensity will tend to cancel out in the printed image. For example, if the pulse variations are uncorrelated and normally distributed with a 5% standard deviation, the diameter of the 25% threshold contour 5203 (FIG. 52) would have an estimated corresponding standard deviation of 0.38 nm (or 0.54%). In contrast, a constant (time-independent) pulse intensity error of 5% would induce a threshold diameter error of 2.3 nm (3.2%). The negative-tone image in FIG. 53 would exhibit a corresponding 0.24-nm (0.34%) standard deviation for uncorrelated pulse variations and 2.0-nm (2.9%) for constant errors.

[0662] The dot centering accuracy and effective image magnification would be affected by thermal expansion in the printheads. Assuming that the printheads comprise silicon plates, a 0.1° C. temperature change would cause the plates to expand by a factor of 1+5×10⁻⁷, resulting in a 5-nm positional shift at the periphery of a 20-mm printhead aperture. The second-stage microlenses' centration controls can be used to counterbalance this shift.

[0663] The image sensitivity to focus error is illustrated in FIGS. 59 and 60. FIG. 59 shows plots of the disk image diameter (i.e., the diameter of the 20 mJ/cm²exposure contour, such as contour 5203 or 5301ain FIGS. 52, 53) as a function of lens L₁'s defocus ΔZ₁. Curve 5901 is for the positive-tone disk image of FIG. 52, and curve 5902 is for the negative-tone disk image of FIG. 53. Similarly, FIG. 60 plots the image diameter versus lens L₂'s defocus ΔZ₂for the positive- and negative-tone disk images (curves 6001 and 6002, respectively). The optical design has a slight focus correction (footnote in FIG. 13) to minimize the negative-tone image's sensitivity to L₂defocus (curve 6002). (All of the image and tolerance simulations in this section and in §10 assume this same focus correction.) However, a similar correction is not made for L₁because this would (according to optical simulations) make the image overly sensitive to lens aberrations. The focus tolerance range corresponding to a ±1-nm diameter shift of the negative-tone image (FIG. 53) is ±2.9/-3.6 μm for L₁and ±1 nm for L₂. For the positive-tone image (FIG. 52) the corresponding ranges are ±3.4/-5.2 μm for L₁and +222/-164 nm for L₂.

[0664] The lens surface form tolerance is the most critical tolerance factor. The system would be most sensitive to errors that are correlated between proximate pixels (and therefore have a cumulative effect on print quality), but which are uncorrelated on the scale of a printhead or wafer print module (and thus cannot be compensated for by adjusting focus, alignment, illumination level, or wafer processing conditions). Considering the disk images in FIGS. 52 and 53, the primary effect of pixel-correlated, uncompensated surface errors would be to change the size and centering position of the threshold contours 5203 and 5301a. (The contour shape distortion would typically be insignificant.) Image centering errors result from tilt and coma aberrations in L₂; whereas size errors are most strongly affected by lens thickness errors (which affect pixel transmittance) and L₂aberrations of low azimuthal order, such as spherical aberration.

[0665] The negative-tone disk image of FIG. 53 is especially sensitive to lens aberrations in L₂. The image center coordinates have an approximately linear dependence on tilt and coma in L₂, and an RMS surface deviation of approximately 0.75 nm can result in a positional shift of up to 1 nm. The disk diameter has an approximate linear dependence on lens thickness, with a uniform surface deviation of 3 nm corresponding to a 1-nm diameter shift. The diameter sensitivity to spherical aberration is approximately quadratic, with an RMS surface deviation of 2 nm corresponding to approximately a 1-nm diameter error. (These estimates are based on simulated surface deviations having the form of Zernike polynomials up to 36th order.)

[0666] The image is much less sensitive to L₁surface errors. The image centering is essentially unaffected by L₁(unless the image is significantly out of focus). The image size has an approximately linear dependence on L₁thickness, with a 4-nm thickness error corresponding to a 1-nm diameter shift. With the exception of the thickness factor, the diameter sensitivity to all L₁aberrations is approximately quadratic, with an RMS surface error of approximately 4 nm corresponding to a 1-nm diameter error.

[0667] Lens surface tolerance requirements can be estimated from the above sensitivity factors. The estimation criterion is that the image diameter and centration errors associated with each surface should not exceed 1 nm. Based on this criterion, the total RMS surface deviation for each surface (including the lens thickness error) should be within 2 nm for L₂and within 4 nm for L₁. Furthermore, the tilt and coma-related component of the L₂surface deviation should be held to 0.75 nm RMS; although this latter requirement can be eliminated if a calibrated centration offset is applied to each L₂lens. (These surface form tolerances could be relaxed somewhat by designing the system to optimize a design performance metric that includes tolerance sensitivity factors.)

[0668] Centration controls can only be effective if the lens centration errors can be accurately characterized. One possible characterization method is the following (see FIG. 61). First, a periodic test pattern of submicron holes (such as hole 6101, shown cross-sectionally in FIG. 61) is patterned in a mask material on a UV-transparent substrate 6102 such as fused silica. The test pattern is formed using interference lithography to ensure accurate placement and size of the holes. A UV-transparent layer 6103 is deposited over the hole array, and then a second mask layer 6104 is deposited and a complementary array of isolated mesa patterns (such as mesa 6105) is lithographically patterned, using the EUV printing system. The pattern is back-illuminated with a beam of UV radiation 6106, and the transmitted radiation is analyzed with a UV micro-imaging and metrology system (not shown) which measures the energy transmitted through the area around each mesa. The transmitted energy distribution over the test pattern is measured as a function of illumination direction and is analyzed to determine the mesa size and position errors.

[0669] Variations of the test method outlined above can also be used to evaluate system performance factors other than lens aberrations. For example, the method can identify defective pixels. Also, printhead alignment errors and stitching alignment errors between printheads can be characterized. (For alignment testing, the same interference lithography system that is used to form the test pattern could also be used to form the alignment tracking patterns, such as patterns 305aand 305bin FIG. 2.)

§12 Control System Architecture

[0670] FIG. 62 is a block diagram illustrating the control system's main architectural components and interactions, in a preferred embodiment. The various elements were described in detail above, and will only be summarized here.

[0671] The printer comprises major subsystems for EUV illumination, position control, and image generation, which are synchronized to a common system clock. The illumination system includes the LPP pulse generator, the EUV source, and the EUV illumination mirrors. Illumination is supplied to eight wafer print modules (“Module 1” . . . “Module 8”), each of which prints onto a separate corresponding wafer. (Bold lines in FIG. 62 represent EUV radiation paths, whereas light lines represent data or informational paths.) Each module comprises 45 printheads, and each printhead comprises an array of printer pixels, a printhead actuator, and a focus/alignment system. The control elements within each pixel include a modulator and a lens actuator (on the second-stage lens).

[0672] The modulators are controlled by the image generation subsystem, which typically includes real-time image decompression hardware, high speed data lines (e.g., fiber optics) feeding the printheads, and integrated data demultiplexers and line buffers on the printheads. The system would typically supply a common data stream to all eight modules in order to minimize the cost and complexity of the image generation hardware.

[0673] The position control subsystem receives positional information from focus/alignment sensors (such as confocal-imaging microlens arrays) formed integrally on the printheads, which sense each printhead's position relative to wafer tracking patterns (the “Tracks” in FIG. 62). This information is processed by the position control system and is used to control the printhead actuators; also the lens actuators are controlled to compensate for thermal expansion and possibly other tolerance factors.

§13 References

[0674] 1. D. W. Sweeney et al., “EUV Optical Design for a 100 nm CD Imaging System”, SPIE Proc., Vol. 3331 (1998), pp. 2-10.

[0675] 2. C. Montcalm et al., “Multilayer reflective coatings for extreme-ultraviolet lithography”, SPIE Proc., Vol. 3331 (1998), pp. 42-51.

[0676] 3. J. A. Folta et al., “Advances in multilayer reflective coatings for extreme-ultraviolet lithography”, SPIE Proc., Vol. 3676 (1999), pp. 702-709.

[0677] 4. Ruth DeJule, Ed., “Masks: What's Behind Rising Costs?”, Semiconductor International, September, 1999, pp. 58-64.

[0678] 5. Yashesh Shroff and William G. Oldham, U. C. Berkeley Microfabrication Laboratory, “EUV Nanomirror Light Modulator Array for Maskless Lithography”, http://ilpsoft.eecs.berkeley.edu:9636/˜ilpsoft/99abstracts/yashesh.2.html (DARPA) MDA972-97-1-0010 and (SRC) 96-LC460.

[0679] 6. Neha Choksi, Mark McCord, David Markle and R. F. W. Pease, Solid State and Photonics Laboratory, Stanford University, “Maskless EUV Lithography Based on Elastic Micromirrors”, http://chomsky.stanford.edu/˜kevbert/neha poster/sld001.htm.

[0680] 7. G. D. Kubiak et al., “High-power extreme ultraviolet source based on gas jets”, SPIE Proc., Vol. 3331 (1998), pp. 81-89.

[0681] 8. G. D. Kubiak et al., “Scale-up of a cluster jet laser plasma source for Extreme Ultraviolet Lithography”, SPIE Proc., Vol. 3676 (1999), pp. 669-678.

[0682] 9. J. A. Folta et al, “High Density Arrays of Micromirrors (Final Report)”, California Univ., Livermore. Livermore Radiation Lab. Report Number: UCRL-ID-133164, February 1999.

[0683] 10. William C. Sweatt and Richard H. Stulen, “Maskless Lithography”, U.S. Pat. No. 5,870,176, issued Feb. 9, 1999.

[0684] 11. Gregory D. Cooper and Richard M. Mohring, “Transferring a programmable pattern by photon lithography”, pending patent application, publication WO9900706, filed Jun. 25, 1998.

[0685] 12. Natale M. Ceglio and David A. Markle, “Maskless, Reticle-Free Lithography”, U.S. Pat. No. 5,691,541, issued Nov. 25, 1997.

[0686] 13. Kenneth C. Johnson, “Microlens Scanner for Microlithography and Wide-Field Confocal Microscopy”, pending patent application, publication WO9734171, filed Feb. 20, 1997.

[0687] 14. G. R. Padmanabhan, “Lithography Systems Employing Programmable Reticles”, pending patent application, publication WO9705526, filed Jul. 25, 1996.

[0688] 15. Kanti Jain and Jeffrey M. Hoffnan, “Seamless, Maskless Lithography System Using Spatial Light Modulator”, pending patent application, publication WO9804950, filed Jul. 25, 1996.

[0689] 16. Henry I. Smith, “A proposal for maskless, zone-plate-array nanolithography,” J. Vac Sci. Technol. B 14(6), November/December 1996, pp. 4318-4322.

[0690] 17. Henry I. Smith, “Maskless lithography using a multiplexed array of Fresnel zone plates”, U.S. Pat. No. 5,900,637, issued May 4, 1999.

[0691] 18. D. J. D. Carter et al., “Zone-Plate Array Lithography (ZPAL): A new maskless approach”, SPIE Proc., Vol. 3676 (1999), pp. 324-332.

[0692] 19. J. R. Sheats and B. W. Smith (Ed.), “Microlithography Science and Technology”, Marcel Dekker, Inc. (1998).

[0693] 20. John M. Hutchinson, “The Shot Noise Impact on Resist Roughness in EUV Lithography”, SPIE Proc., Vol. 3331 (1998), pp. 531-536.

[0694] 21. S. Malzer et al, “Properties and applications of the ‘epitaxial shadow mask molecular beam epitaxy technique’”, J. Vac. Sci. Technol. B, Vol. 14(3) (1996), pp. 2175-2179.

[0695] 22. M. Kim et al., “The growth of nanometer Si/SiGe/Si quantum well wires with local molecular beam epitaxy in dependence on the shadow mask geometry”, Journal of Crystal Growth, Vol. 167 (1996), pp. 508-515.

[0696] 23. J. Köhler et al, “Direct growth of nanostructures by deposition through an Si₃N₄shadow mask”, Physica E, Vol. 4 (1999), pp. 196-200.

[0697] 24. Roli Lüithi et al, “Parallel nanodevice fabrication using a combination of shadow mask and scanning probe methods”, Appl. Phys. Lett., Vol. 75(9) (1999), pp. 1314-1316.

[0698] 25. M. P. Kanauff et al., “A Gas Curtain for Mitigating Hydrocarbon Contamination of EUV Lithographic Optical Components”, SPIE Proc., Vol. 3676 (1999), pp. 735-742.

[0699] 26. Claude Montcalm et al, “Passivating Overcoat Bilayer for Multilayer Reflective Coatings for Extreme Ultraviolet Lithography”, U.S. Pat. No. 5,958,605, issued Sep. 28, 1999.

[0700] 27. K. Miyamoto, “The Phase Fresnel Lens”, J. Opt. Soc. Am., Vol. 51, No. 1 (1961), pp. 17-20.

[0701] 28. B. W. Smith et al., “Materials for reflective multilayer coatings for EUV wavelengths”, SPIE Proc., Vol. 3331 (1998), pp. 544-554.

[0702] 29. C. Montcalm et al., “Multilayer coatings of 10× projection optics for extreme-ultraviolet lithography”, SPIE Proc., Vol. 3676 (1999), pp. 710-716.

[0703] 30. R. Grunwald et al., “Microlens formation by thin-film deposition with mesh-shaped masks”, Applied Optics, Vol. 38(19) (1999), pp. 4117-4124.

[0704] 31. J. G. Crockett et al., “Thickness Control of Semiconductor Device Layers in Reactive Ion Etch Processes”, U.S. Pat. No. 5,882,468, issued Mar. 16, 1999.

[0705] 32. M. L. Schattenburg et al., “Reactive-ion etching of 0.2 μm period gratings in tungsten and molybdenum using CBrF₃”, J. Vac. Sci. Technol. B 3(1) (1985), pp. 272-275.

[0706] 33. C. K. Malek et al., “Fabrication of high-resolution multilayer reflection zone plate microlenses for the soft x-ray range”, Optical Engineering Vol. 30(8) (1991), pp. 1081-1085.

[0707] 34. M. C. Hutley, “Optical techniques for the generation of microlens arrays”, Journal of Modern Optics, Vol. 37(2) (1990), pp. 253-265.

[0708] 35. A. Fernandez et al, “Methods for fabricating arrays of holes using interference lithography”, USDOE, Washington, D.C. Report Number: UCRL-JC-125997; CONF-9705218 (1997).

[0709] 36. H. P. Herzig (Ed.), “Micro-optics: Elements, systems and applications”, Taylor & Francis (1997).

[0710] 37. S. Sinzinger and J. Jahns, “Microoptics”, WILEY-VCH (1999).

[0711] 38. J. J. McClelland et al., “Nanofabrication via Atom Optics with Chromium”, SPEE Proc., Vol. 2995 (1997), pp. 90-96.

[0712] 39. John R. Karpinsky et al, “MEMS Micro-Shutter SLM for Intensity Modulation”, SPIE Proc., Vol. 3633 (1999), pp. 254-259.

[0713] 40. Carl O. Bozler and Steven Rabe, “Spatial Light Modulator”, U.S. Pat. No. 5,959,763, issued Sept. 28, 1999.

[0714] 41. D. M. Bloom, “Variable Attenuator for Laser Radiation”, U.S. Pat. No. 4,327,966, issued May 4, 1982.

[0715] 42. Zhenhai Chen and Ren C. Luo, “Design and Implementation of Capacitive Proximity Sensor Using Microelectromechanical Systems Technology”, IEEE Transactions on Industrial Electronics, Vol. 45(6) (1998), pp. 886-894.

[0716] 43. “Nomenclature and Definitions for Illuminating Engineering”, ANSI/IESNA RP-16-96, Illuminating Engineering Society of North America (1996).

[0717] 44. M. Feldman, “Use of Zone Plate Arrays in Projection X-Ray Lithography”, OSA Proceedings on Soft-X-Ray Projection Lithography, Vol. 18 (1993), pp. 207-209.

§14 Conclusion

[0718] The EUV printer system design concept described in the preceding specification illustrates the potential benefits and performance characteristics of maskless EUV lithography. The system's most important benefits would be eliminating the need for EUV masks, and eliminating the requirement for optical elements meeting Angstrom-level figure tolerances over large apertures. The design presented herein may not be directly applicable in its specifics to practical applications since it is based on particular assumptions about the illumination source characteristics, manufacturing and material constraints, design objectives and optimization criteria, etc., which may not accurately represent practical design conditions. Nevertheless, the specification provides, by way of example, a general design concept and methodology that can be applied to a diversity of micro-printing applications ranging, for example, from high-speed, digital document printers to x-ray lithography systems employing synchrotron radiation.

[0719] While the above is a complete description of specific embodiments of the invention, various modifications, alternative constructions, and equivalents may be used. Therefore, the above description should not be taken as limiting the scope of the invention as defined by the claims.

Claims

Certificate of Correction present

The claims shown below may not include correction changes. Use the corrected PDF for authoritative text.

What is claimed is:

1. A printing exposure apparatus comprising a radiation source, an illumination system, a modulator mechanism, a printhead, and a scanning mechanism, wherein

the printhead comprises a printhead aperture and an array of printer pixels distributed across the printhead aperture,

each pixel comprises a sequence of corresponding microlenses, which are symbolically designated as L₁, L₂, . . . , L_N, wherein N is an integer greater than 1,

the illumination system conveys radiation from the radiation source to the printhead, whereupon the first microlens L₁in each sequence receives illuminating radiation,

each microlens L_mother than the last microlens L_Nin the sequence focuses the illuminating radiation onto the next corresponding microlens L_m+1,

the last microlens L_Nin each sequence focuses the illuminating radiation onto a corresponding focal point on a printing surface,

the modulator mechanism modulates the radiation, whereby the focal points' exposure intensity levels are controllably varied,

the modulator mechanism is a spatial light modulator comprising an array of modulator elements,

each pixel further comprises one such modulator element disposed between that pixel's microlens L₁and the printing surface,

the exposure intensity at each pixel's corresponding focal point is modulated independently of other pixels, and

the scanning mechanism establishes relative motion between the printing surface and the printhead in synchronization with the modulator mechanism as the printing surface is exposed, whereby a synthesized, high-resolution exposure image is formed on the printing surface.

2. A printing exposure apparatus comprising a radiation source, an illumination system, a modulator mechanism, a printhead, and a scanning mechanism, wherein

the printhead comprises a printhead aperture and an array of printer pixels distributed across the printhead aperture,

each pixel comprises a sequence of corresponding microlenses, which are symbolically designated as L₁, L₂, . . . , L_N, wherein N is an integer greater than 1,

the illumination system conveys radiation from the radiation source to the printhead, whereupon the first microlens L₁in each sequence receives illuminating radiation,

each microlens L_mother than the last microlens L_Nin the sequence focuses the illuminating radiation onto the next corresponding microlens L_m+1,

the last microlens L_Nin each sequence focuses the illuminating radiation onto a corresponding focal point on a printing surface,

the modulator mechanism modulates the radiation before it is conveyed to the printhead and all the focal points' exposure intensity levels are subjected to the same modulation, and

3. A printing exposure apparatus comprising:

a radiation source;

a printhead comprising a printhead aperture and an array of printer pixels distributed across the printhead aperture, wherein each pixel comprises

a sequence of corresponding microlenses including at least first and last microlenses, and

a modulator element disposed between the first microlens and the printing surface;

an illumination system that conveys radiation from the radiation source to the printhead, whereupon for each pixel in the printhead

the first microlens in the sequence receives illuminating radiation,

each microlens other than the last microlens in the sequence focuses received illuminating radiation onto the next corresponding microlens, and

the last microlens in the sequence focuses received illuminating radiation onto a corresponding focal point on a printing surface;

a modulator control mechanism that controls the modulators in the printhead independently of each other so that the exposure intensity at each pixel's corresponding focal point is modulated independently of other pixels; and

a scanning mechanism that establishes relative motion between the printing surface and the printhead in synchronization with the modulator control mechanism as the printing surface is exposed.

4. A printing exposure apparatus comprising:

a radiation source;

a printhead comprising a printhead aperture and an array of printer pixels distributed across the printhead aperture, wherein each pixel comprises

a sequence of corresponding microlenses including at least first and last microlenses, and

an illumination system that conveys radiation from the radiation source to the printhead, whereupon for each pixel in the printhead

the first microlens in the sequence receives illuminating radiation,

each microlens other than the last microlens in the sequence focuses received illuminating radiation onto the next corresponding microlens, and

the last microlens in the sequence focuses received illuminating radiation onto a corresponding focal point on a printing surface;

a modulator mechanism that modulates the radiation before it is conveyed to the printhead so that the exposure intensity at each pixel's corresponding focal point is subjected to the same modulation; and

a scanning mechanism that establishes relative motion between the printing surface and the printhead in synchronization with the modulator control mechanism as the printing surface is exposed.

5. The exposure apparatus of claim 1 or 2 or 3 or 4, wherein the radiation comprises EUV radiation in the 10 to 15 nanometer wavelength range.

6. The exposure apparatus of claim 5, wherein the radiation source is a xenon laser-produced plasma source.

7. The exposure apparatus of claim 1 or 2 or 3 or 4, wherein the radiation source is a synchrotron and the radiation is EUV or x-ray radiation.

8. The exposure apparatus of claim 1 or 2 or 3 or 4, wherein at least one of each pixel's corresponding microlenses is a singlet, continuous-profile lens element.

9. The exposure apparatus of claim 5, wherein at least one of each pixel's corresponding microlenses is a singlet, continuous-profile meniscus lens element composed of an optical material whose complex refractive index has a real part less than 1 over the operating EUV wavelength range.

10. The exposure apparatus of claim 9, wherein the singlet element is composed of one of the materials molybdenum, niobium, ruthenium, rhodium, or yttrium.

11. The exposure apparatus of claim 1 or 2 or 3 or 4, wherein each pixel's last microlens is a multiple-element lens.

12. The exposure apparatus of claim 5, wherein each pixel's last microlens is a triplet lens comprising two continuous-profile, meniscus elements with a convex spacer element between them.

13. The exposure apparatus of claim 1 or 2 or 3 or 4, wherein at least one of each pixel's corresponding microlenses is a phase Fresnel lens.

14. The exposure apparatus of claim 1 or 2 or 3 or 4, wherein at least one of each pixel's corresponding microlenses is a Fresnel zone plate.

15. The exposure apparatus of claim 1, wherein each pixel further comprises a micromechanical actuation mechanism for fine-adjusting the lateral position of the pixel's corresponding last microlens.

16. The apparatus of claim 1 or 2 or 3 or 4 wherein the number of microlenses per pixel is two.

17. The apparatus of claim 1 or 2 or 3 or 4, further comprising at least one additional printhead to define a plurality of printheads that define a wafer print module, wherein the printing surface is on a semiconductor wafer, and the printheads simultaneously expose a common printing surface on the wafer.

18. The exposure apparatus of claim 1 or 2 or 3 or 4, wherein each modulator element comprises a micromechanically-actuated shutter.

19. The exposure apparatus of claim 1 or 2 or 3 or 4, wherein each modulator element is a grating modulator comprising two proximate, lamellar transmission gratings, one of which is micromechanically actuated to modulate the gratings' combined transmittance in the zero order.

20. The exposure apparatus of claim 1 or 2 or 3 or 4, wherein the printhead further comprises a position sensor system that detects the printhead's position relative to the printing surface, and wherein the detected positional information is used to accurately control the relative alignment between the printing surface and the printhead while scanning.

21. The exposure apparatus of claim 3, wherein the illumination system comprises collection optics that produce a collimated radiation beam of substantially uniform irradiance, and distribution optics that efficiently partition the collimated beam into individual illumination zones covering individual printhead apertures.

22. The exposure apparatus of claim 17, wherein the collection optics comprises a condenser mirror and collimator mirror, and wherein the mirrors' surface shapes are configured to simultaneously satisfy the collimation and uniformity conditions, taking into account the radiation source's angular emission distribution and the mirror's reflectivity dependence on incidence angle.

23. The exposure apparatus of claim 21, wherein the collection optics comprises a condenser mirror and collimator mirror, the mirrors comprise reflective coatings, the collimator is spherical, and the mirrors' surface shapes and reflective coatings are configured to simultaneously satisfy the collimation and uniformity conditions, taking into account the radiation source's angular emission distribution and the mirror's reflectivity dependence on incidence angle.

24. The exposure apparatus of claim 21, wherein the distribution optics comprise two sets of flat, terraced fold mirrors, including a first set of fold mirrors that partition the collimated beam from the collection optics into a set of first reduced-aperture collimated beams, and a second set of fold mirrors that partition each first reduced-aperture beam into a set of second reduced-aperture collimated beams, wherein each second reduced-aperture collimated beam illuminates a particular corresponding printhead aperture.

25. The exposure apparatus of claim 24, wherein the radiation comprises EUV radiation in the 10 to 15 nanometer wavelength range, the radiation source is a xenon laser-produced plasma source, and the fold mirrors comprise reflective coatings whose reflectivities counterbalance and compensate for exposure illumination variations between printheads due to EUV absorption by ambient xenon gas.

26. The exposure apparatus of claim 1 or 2 or 3 or 4, wherein the scanning mechanism translates the printhead relative to a stationary printing surface during exposure.

27. The exposure apparatus of claim 26, wherein the printing surface is exposed through multiple exposure scans and is stepped between scans so that the printhead raster-scans multiple exposure fields.

28. The exposure apparatus of claim 1 or 2 or 3 or 4, wherein the scanning mechanism translates the printing surface during exposure, while the printhead remains substantially stationary except for small positional corrections to maintain accurate positional alignment between the printhead and printing surface.

29. The exposure apparatus of claim 1 or 2 or 3 or 4, wherein the scanning motion interleaves or overlaps exposure patterns from different pixels so that any typical exposure point on the printing surface receives exposure energy from multiple pixels, whereby the resulting exposure image is not significantly affected by isolated pixel defects or nonfunctional pixels.

30. The apparatus of claim 3, further comprising at least one additional wafer print module to define a plurality of wafer print modules, wherein the modules simultaneously expose different print surfaces on separate wafers.

31. The exposure apparatus of claim 12, wherein the meniscus elements are made of molybdenum.

32. The exposure apparatus of claim 12, wherein the spacer is made of beryllium.

33. The exposure apparatus of claim 12, wherein the spacer is a vacuum space.

Patent Citations (20)

Patent	Date	Inventor	Cited By
US4327966(A)	1982-05-01	Bloom	Applicant
US4638334(A)	1987-01-01	Burnham et al.	Applicant
US5263073(A)	1993-11-01	Feldman	Applicant
US5517279(A)	1996-05-01	Hugle et al.	Applicant
US5619245(A)	1997-04-01	Kessler et al.	Applicant
US5625471(A)	1997-04-01	Smith	Applicant
US5691541(A)	1997-11-01	Ceglio et al.	Applicant
US5870176(A)	1999-02-01	Sweatt et al.	Applicant
US5882468(A)	1999-03-01	Crockett et al.	Applicant
US5900637(A)	1999-05-01	Smith	Applicant
US5958605(A)	1999-09-01	Montcalm et al.	Applicant
US5959763(A)	1999-09-01	Bozler et al.	Applicant
US6016185(A)	2000-01-01	Cullman et al.	Applicant
JP5224396	1992-10-01		Applicant
WO97/05526	1997-02-01		Applicant
WO97/34171	1997-09-01		Applicant
WO98/04950	1998-02-01		Applicant
WO98/12603	1998-03-01		Applicant
WO99/00706	1999-01-01		Applicant
WO-00/42618(A1)	2000-07-01		Examiner

Non-Patent Literature (40)

Davidson, Mark, “A Microlens Direct-Write Concept for Lithography,” SPIE vol. 3048, pp. 346-355 (1977).Applicant
Carter, et al., “Zone-plate array lithography (ZPAL): A new maskless approach,” SPIE, vol. 3676, Mar. 1999, pp. 324-332 (Santa Clara, CA).Applicant
Carter et al., “Maskless, parallel patterning with zone-plate array lithography,” J. Vac. Sci. Technol. B 17(6), Nov./Dec. 1999, pp. 3449-3452.Applicant
Chen et al., “Design and implementation of capacitive proximity sensor using microelectromechanical systems technology,” IEEE Transactions on Industrial Electronics, vol. 45(6) Dec. 1998, pp. 886-894.Applicant
Choksi et al., “Maskless EUV lithography based on elastic micromirrors,” Slides 1 through 10, Solid State and Photonics Laboratory, Stanford University, Aug. 7, 1999.Applicant
Choksi et al., “Maskless extreme ultraviolet lithography,” J. Vac. Sci. Technol. B 17(6), Nov./Dec. 1999, pp. 3047-3051.Applicant
DeJule, R., “Masks: What's behind rising costs?” Semiconductor International, Sep. 1999, pp. 58-64.Applicant
Feldman, M., “Use of zone plate arrays in projection x-ray lithography,” OSA Proceedings on Soft X-Ray Projection Lithography, 1993, vol. 18 (Andrew M. Hawryluk and Richard H. Stulen, eds.), pp. 207-209.Applicant
Fernandez, A. et al., “Methods for fabricating arrays of holes using interference lithography,” J. Vac. Sci. Technol. B 15(6), Nov./Dec. 1997, pp. 2439-2443.Applicant
Folta, James A. et al., “Advances in multilayer reflective coatings for extreme-ultraviolet lithography,” SPIE Proc. vol. 3676 (Mar. 1999), pp. 702-709.Applicant
Folta, J.A. et al., “High density arrays of micromirrors,” Feb. 1999, pp. 1-9 (UCRL-ID-133164/National Technical Information Service, US Dept. of Commerce, Springfield, VA 22161).Applicant
Gale, M.T., “Direct writing of continuous-relief micro-optics,” Micro-optics, Elements, systems and applications, (Chapter 4) (Hans Peter Herzig, ed.) (Taylor & Francis, 1997), pp. 87-152.Applicant
Grunwald, Rüediger et al., “Microlens formation by thin-film deposition with mesh-shaped masks” Applied Optics, vol. 38(19), (Jul. 1, 1999), pp. 4117-4124.Applicant
Hutchinson, John M., “The shot noise impact on resist roughness in EUV lithography,” SPIE Proc., vol. 3331 (1998), pp. 531-533.Applicant
Hutley, M.C., “Optical techniques for the generation of microlens arrays,” Journal of Modern Optics, vol. 37(2) (1990), pp. 253-265.Applicant
“Nomenclature and Definitions for Illuminating Engineering,” ANSI/IESNA RP-16-1996, Illuminating Engineering Society of North America (ANSI Approval Date: Nov. 29, 1996), pp. 1-2.Applicant
Kanouff, Michael P. et al., “A gas curtain of mitigating hydrocarbon contamination of EUV lithographic optical components,” SPIE Proc. vol. 3676 (Mar. 1999), pp. 735-742.Applicant
Kim, Myeongcheol et al., “The growth of nanometer Si/SiGe/Si quantum well with wires local molecular beam epitaxy in dependence on the shadow mask geometry,” Journal of Crystal Growth 167 (1966), pp. 508-515.Applicant
Köhler, J. et al., “Direct growth of nanostructures by deposition through an Si3N4 shadow mask,” Physica E 4 (1999) (Elsevier Science B.V.), pp. 196-200.Applicant
Kubiak, Glenn D. et al., “Scale-up of a cluster jet laser plasma source for extreme ultraviolet lithography,” Spie Proc. vol. 3676 (1999), pp. 669-678.Applicant
Kubiak, Glenn D. et al., “High-power extreme ultraviolet source based on gas jets,” Spie Proc. vol. 3331 (1998), pp. 81-89.Applicant
Lüthi, Roli et al., “Parallel nanodevice fabrication using a combination of shadow mask and scanning probate methods,” Appl. Phys. Lett., vol. 75(9), Aug. 1999, pp. 1314-1316.Applicant
McClelland, J.J. et al., “Nanofabrication via atom optics with chromium,” SPIE Proc. vol. 2995 (1997), pp. 90-96.Applicant
Karpinsky, John R. et al., “MEMS Micro-shutter SLM for Intensity modulation,” SPIE Proc. vol. 3633 (1999), pp. 254-259.Applicant
Malek, C. Khan et al., “Fabrication of high-resolution multilayer reflection zone plate microlenses for the soft x-ray range,” Optical Engineering, vol. 30(8) (Aug. 1991), pp. 1081-1085.Applicant
Malzer, S. et al., “Properties and applications of the ‘epitaxial shadow mask molecular beam epitaxy technique,’” J. Vac. Sci. Technol. B (14(3), May/Jun. 1996; 2175-2179.Applicant
Miyamoto, Kenro, The phase fresnel lens, J. of the Optical Society of America, 51(1), Jan. 1961, pp. 17-20.Applicant
Montcalm, Claude et al., “Multilayer coatings of 10X projection optics for extreme-ultraviolet lithography,” SPIE Proc. vol. 3676 (1999), pp. 710-716.Applicant
Montcalm, Claude et al., “Multilayer reflective coatings for extreme-ultraviolet lithography,” SPIE Proc. vol. 3331(1998), pp. 42-51.Applicant
Schattenburg, M.L., Reactive-ion etching of 0.2 μm period gratings in tungsten and molybdenum using CBrF3, J. Vac. Sci. Technol. B 3(1), Jan./Feb. 1985, pp. 272-275.Applicant
Microlithography, Science and Technology, Alignment and Overlay, (James R. Sheats and Bruce W. Smith, eds., Marcel Dekker, Inc., NY) (1998) pp. 319-321.Applicant
Shroff, Yashesh et al., “EUV nanomirror light modulator array for maskless lithography,” U.C. Berkeley Microfabrication Laboratory, http://ilpsoft.eecs.berkeley.edu:9636/˜ilpsoft/99abstracts/yashesh.2.html (DARPA) MDA972-97-1-0010 and (SRC) 96-LC-460.Applicant
Sinzinger, Stefan et al., “4. Refractive microoptics,” Microoptics (Wiley-VCH, NY, 1999); pp. 86-103.Applicant
Smith, Bruce W. et al., “Materials for reflective multilayer coatings for EUV wavelengths,” SPIE Proc. vol. 3331 (1998), pp. 544-549.Applicant
Smith, Henry I., “A proposal for maskless, zone-plate-array nanolithography,” J. Vac. Sci. Technol. B 14(6), Nov./Dec. 1996, pp. 4318-4322.Applicant
Sweeney, Donald W. et al., “EUV optical design for a 100 nm CD imaging system,” SPIE Proc. vol. 3331 (1998), pp. 2-10.Applicant
Völkel, R. et al, “Microlens lithography,” Conference: 1996 Display Manufacturing Technology Conference, Digest of Technical Papers, First Edition, pp. 95-96 (Publisher: Soc. Inf. Display, Santa Ana, Ca).Applicant
Völkel, R. et al., “Microlens array imaging system for photolithography,” Opt. Eng. 35(11) (Nov. 1996); pp. 3323-3330.Applicant
Völkel, R. et al., “Microlens lithography: A new approach for large display fabrication,” Microelectronic Engineering 30 (1996); pp. 107-110.Applicant
Völkel, R. et al., “Microlens lithography and smart masks,” Microelectronic Engineering 35 (1997); pp. 513-516.Applicant

Abstract