Sie sind auf Seite 1von 12

Microelectronic Engineering 7879 (2005) 381392 www.elsevier.

com/locate/mee

Maskless lithography
R. Fabian Pease
Stanford University, Stanford, CA 94305-4075, USA Available online 22 January 2005

Abstract The high and rising cost of photomasks (largely driven by writing times exceeding 24 h) is driving the exploration of maskless lithography for applications requiring throughput about 1 cm2/s which is about one tenth that of an optical projection exposure system. Achieving this throughput with charged particle lithography requires currents 10,000 times larger than those presently used and hence sets up the need for charged particle optics radically dierent from those being used today. Achieving this throughput with optical maskless lithography at the required minimum features sizes of 65 nm and below is a serious engineering challenge for the spatial light modulator. Meeting 10% or even 1% of the throughput requirement might still result in mask writing and inspection technologies that would lead to signicantly less expensive masks. Furthermore, relaxing the requirements on control of individual edge positions (i.e., a xed-shape projector) would signicantly ease the above challenges. 2005 Published by Elsevier B.V.
Keywords: Maskless; Lithography; Electron beam; Optical lithography

1. Introduction To try and avoid the high and rising costs of photomasks, two forms of maskless lithography are being seriously pursued. One is optical (OML), whose proponents claim enjoys no fundamental limit to throughput and the other is charged particle maskless lithography (CPML2) that is claimed to enjoy no practical limit to resolution. Needless to say the above claims are oversimplications. OML has recently been reviewed by Sandstrom, Hintersteiner and their colleagues
0167-9317/$ - see front matter 2005 Published by Elsevier B.V. doi:10.1016/j.mee.2005.01.009

at Micronic Laser and ASML [1] and will be only briey covered here. A notional requirement is an exposure rate of 1 cm2/s and minimum feature size of 65 nm extendable to 45 nm for OML and to 25 nm for CPML2.

2. Denitions (Fig. 1) Minimum Feature size (MFS): the nominal size of the minimum feature to be exposed on the wafer.

382

R.F. Pease / Microelectronic Engineering 7879 (2005) 381392

Fig. 1. Denitions.

Minimum Address Unit (MAU): the smallest increment by which we want to adjust the position of the edge of a feature (also called the design grid). Ray: the trajectory of a single photon or charged particle. Pencil: Ideally a collection of rays converging to a single point in the image; here, we mean a collection of rays converging to the best focus. Bundle: A collection of pencils whose landing areas are contiguous. Beam: The total ux of photons or charged particles in the system. Column: A source and one or more lenses axially symmetric about an optical axis. Space-charge blurring includes stochastic (scattering) and continuum (lens) eects. For example a pattern generator employing a single pencil beam may have a pencil size (FWHM) the same as the MAU. But, as shown below, a more economical strategy is to have a pencil size much larger and adjust the current in the pencil to adjust the position of the feature edge (Fig. 1c). A more advanced pattern generator may employ a beam that is a bundle dening a MFS onto the wafer and adjust the positions of feature edges using a variable-shape technique. Some systems are now being developed feature a beam comprising an array of bundles.

3. Four limitations to throughput W As pointed out above, we should aim for W = 1 cm2/s. One well-known limitation is the dose required by the resist. For OML this, is usually expressed in mJ/cm2; the development of increasingly powerful lasers for optical projection lithography at 10 cm2/s suggests that this is not a serious problem for OML. For CPML2 this dose, usually expressed in Q lC/cm2, is that used to bring about the required chemical change in the resist. In most instances the value of Q increases with the energy of the particle to keep constant the energy dissipated per unit volume in the resist. Obviously W 6 I/Q and so to maintain W = 1 cm2/s for Q = 1 lC/cm2 (corresponding to a sensitive resist) we need I P 1 lA. This might just be practical for an MFS of 200 nm in a single-bundle system but not for 25 nm (Fig. 2). Hence a multi-bundle system seems to be needed. The speed at which the beam is scanned across the target can also limit throughput. For example, if we employ in a CPML2 system a stage speed of v cm/s and sweep width y cm then for a single bundle system W 6 vy cm2/s; for a n-bundle system the W 6 nvy cm2/s. So for n = 1 and y = 100 lm, the stage speed must be at least 100 cm/s. This is about ten times faster than todays stages and may cause unacceptable blurring for dwell times exceeding

R.F. Pease / Microelectronic Engineering 7879 (2005) 381392

383

Fig. 2. Space-charge blurring of 1-bundle (shaped-beam) systems.

10 ns but with n = 10 this need not be a problem. Alternatively, we can use a high-speed electrostatic deector to stop the beam travel over the sample during the dwell time. So mechanical stage speed does not seem to be a serious limiting factor for CPML2 especially as on grounds of required total current multiple bundles are needed. The case of OML is trickier for two reasons: high-speed (300 MHz) deection of the beam is more dicult to achieve, and the source, instead of being continuous as in the charge particle case, is pulsed at a repetition frequency (PRF) about 10 kHz. Thus, in the absence of such high-speed deection, a single bundle of beams should be v/10,000 long and the sweep width is achieved by having sucient pencils in the y-direction (Fig. 3). For example, if we have 5 5 pencils per MFS, MFS = 50 nm, and v = 10 cm/s, then we need 1000 pencils in the x-direction; and to achieve W = 1 cm2/s, we need y = 0.1 cm so the total number of pencils in the y-direction is 100,000 or 1e8 pencils altogether. Hence, engineering a sucient array of spatial light modulators is challenging. An additional drawback to OML is that we need to delineate not just the nominal pattern, but the much more complicated pattern demanded by the resolution enhancement technologies (Fig. 4). Thus, even more pencils might be required.

Fig. 3. For a pulsed illumination on a stage moving at velocity v in the x-direction the exposure can be accomplished as a sequence of ashes such that the bundle of pencils lls the distance traveled between pulses (v/PRF). The throughput is then given by W = vy and the number of pencils is W/PRF/p2, where p is the distance between pencil centers.

The max frequency f at which we can modulate the beam can also limit W. For the na system, ve where the beam only exposes 1 MAU at a time (1 bit/MAU) W 6 f(MAU)2; for a MAU of 1 nm we need f = 100 THz. Obviously, we need a system that exposes many MAUs simultaneously. There is at least one commercially available EBL tool

384

R.F. Pease / Microelectronic Engineering 7879 (2005) 381392

Fig. 4. Comparison of nominal pattern, as exposed in a charged particle system (left) and an equivalent optical pattern incorporating resolution enhancement features.

that features a value of f = 320 MHz, so we should aim for a system that exposes at least 3 105 MAUs simultaneously. In the case of OML, the PRF of the source sets the upper limit of 10 kHz which should be within the range of the MEMS-based spatial light modulators being developed. So f does not appear to be a serious limitation. As feature sizes shrink below 50 nm shot noise appears to be the most serious limit for CPML2 and for EUV and is described in some detail in Section 4.

4. Fundamental shot noise limits To read, write or display, a pattern requires a certain number of quanta that make up the image. The quanta can be grains in a photographic emulsion, electrons striking a cathode ray tube or a resist, inkdrops in jet printer, X-rays in radiograph. These limits have been studied for decades by the photographic and television industries. For example, Rose [2] determined that for the eye to distinguish two gray levels in two adjacent resolvable areas their individual populations of quanta must dier by at least ve standard deviations. Everhart [3] and Oatley [4] found that a similar rule holds for images built up in a scanning electron microscope (SEM). Thus, if we wish to have a 5% distinguishable change in gray level we need 10,000 quanta for each area, assuming, as in a Poisson distribution of quanta, that one standard deviation is the square root of the mean. How this applies to resist patterns in lithography has been the subject of (remarkably inconclusive) debate for more than 25 years [5]. Despite the inconclusiveness of the debate on the shot noise limits in resist images it is clear that

the higher resolution resists require more dose; this is analogous to the case of photographic lm. Below is a treatment of the problem that is derived from a paper delivered by Oldham[6]. Designers today require that 1 = MFS/50 and each minimum feature must receive sucient quanta so that the MAUs on the perimeters receive sucient quanta to maintain control of edge position to 1 MAU. Determining this number depends upon a number of assumptions which even include the denitions of edge position and line edge roughness that take into account the corresponding spatial frequency spectra of these quantities (seemingly ignored in ITRS roadmaps) [7]. In other words, we do not yet have a clear quantitative value of the number of quanta per MAU or per MFS. As a very rough estimate we assume that 1 quanta per MAU is too small and 100 probably too safe; so we will pick 10 as a geometric mean; this translates to 25,000 per MFS to satisfy todays designers. Then, for a MFS of 50 nm (1 MAU = 1 nm) the required dose is 10 quanta/nm2. One quanta from an ArF laser has about 5 eV energy thus the required minimum dose for the resist is about 10 J/m2 or 1 mJ/cm2 so is not a signicant limit. However, the limit for EUV quanta is about 20 mJ/cm2 so could be serious as current projections for the throughput of EUV are based on a required dose of 5 mJ/cm2. In the above example for singly-charged particles, we need at least 160 lC/cm2 or about the chemically required dose of PMMA to 25 keV electrons. For a MFS of 25 nm, we will need about 640 lC/cm2 or about the dose required by HSQ, an electron sensitive material that is nding increasing popularity in this size range. Note that these minimum required doses are independent of the chemistry of the resist or the species or energy of charged particle. We might be tempted therefore to increase the electron energy since we then can deliver more current into a beam of given diameter. But many of those (higher energy) electrons penetrate the resist without scattering and so do not contribute to the number of quanta interacting with the resist. So it is not clear that increasing the electron energy will help.

R.F. Pease / Microelectronic Engineering 7879 (2005) 381392

385

Fig. 5. Tilting Micromirror (courtesy of Karel van der Mast).

In general,we need m quanta/(MFS), so W = I(MFS)2/mq, where q is the electronic charge, i.e., the throughput decreases as the square of the MFS for a given I. For example if MFS = 25 nm and n = 25,000, then I > 0.6 mA! This is about two orders of magnitude higher than that achieved by any electron beam lithography system under development. For all systems in use today, I decreases (at least) as the square of the MFS yet it now appears

that we are shot noise limited so the required dose increases inversely as the square of the MFS. Thus, the throughput decreases at least as the fourth power of MFS! Thus, a radically dierent approach is now almost certainly required. The prospect for ions is even worse because of their greater vulnerability to space charge eects. Thus, it seems that the main challenge for OML is engineering the enormous array (1e8) of spatial light modulators (SLM). Sandstrom et al. [1] have

Fig. 6. Outline of ZPAL in which an array of micro-zone plates replaces the projection optics [8].

386

R.F. Pease / Microelectronic Engineering 7879 (2005) 381392

described the system under development jointly by Micronic Laser and ASML. Their spatial light modulator employs an array of tilting mirrors which can deliver a gray-scale image (Fig. 5); alternative approaches to the SLM are being developed by Oldham and by Solgaard et al. Gil et al. [8] have described a system in which the refractive projection optics (of the ASML system) is replaced by an array of zone plates which might well save cost (of the optics) and reduce the mechanical travel needed (Fig. 6); results have shown sub-wavelength resolution. For CPML2, the main challenge is achieving the combination of current and resolution to overcome the shot-noise limitation. The following section is devoted to this issue.

5. CPML2 architectures (Fig. 7) CPML2 can include electron lithography, ion beam patterning and writing with charged ink droplets. This last form has never exhibited sub micron resolution and will not be considered. Ion beam patterning is widely used for the repair of photomasks without the use of resists. In this case the throughput is so small that we shall also ignore this form of patterning as a contender. Nearly, all

such CPML2 has been done with electron beams although the use of ions is being explored [9]. CPML2 in the form of direct-write electron beam lithography is already used in manufacture where only very small areas are required and resolution (<100 nm minimum feature size) is paramount. Examples include the gates vias of mm-wave devices and circuits, and pole trimming for thin lm heads for hard disk drives [10,11]. The currents used are less than 1 nA. CPML2 is also used for the manufacture of masks [12]. The currents are about 110 nA. Both applications have been served with systems that employ a single beam and represent an extreme case of low throughput, in the range of 0.00010.01 cm2/s. A throughput, about 1 cm2/s requires currents 10,000 times larger and this requirement leads to the need for charged particle optics radically dierent from those being used today.

6. Single-axis systems (Fig. 7a and b) [13,14] The most common conguration features one column, one axis, one pencil but as pointed out above this looks hopeless in terms of current available for feature size of even 200 nm.

Fig. 7. Dierent arrangements for charged particle maskless lithography.

R.F. Pease / Microelectronic Engineering 7879 (2005) 381392

387

The most popular way to increase the number of MAUs being exposed simultaneously is to employ a single bundle, shaped so that one minimum feature can be exposed in one ash; i.e., one column, one axis, one bundle. But even here spacecharge eects set a limit to the sharpness of the edges (Fig. 2) which indicates that at the currents ten times less than those envisioned there is unacceptable blurring. An alternative approach is the dot matrix approach described by Newman, Winograd and by Pfeier (Fig. 8) [14]; this allows the beamlets to ll the entire eld of view of the lens. This leads to an electron optical arrangement similar to that employed in electron projection

lithography; indeed the switching element can be thought of as an active mask. Although the beams might be widely separated near the object plane and near the corresponding conjugate planes, the beams still co-mingle near the pupil planes. Winograd [14], Han[15a] and Golladay et al. [16] have described how these systems are limited in resolution by space-charge (Fig. 9). Indeed Han [15b] rigorously developed, and experimentally veried, an electron optical scaling model so that dierent congurations can be examined. To reach the maximum currents, it appears that the focus must be modulated according to the instantaneous current to correct for space-charge defocusing; this

Fig. 8. Many bundles, one axis (Newman, 1983).

388

R.F. Pease / Microelectronic Engineering 7879 (2005) 381392

Fig. 9. Space-charge blurring in 25-pencil, 1-axis system (similar to EPL column). At low current (e.g. 100 nA) each pencil would be focused onto a grid point. At 25 mA, 100 KV there is both continuum and stochastic space charge blurring. From Winograd, 1999.

might be practical for a mask exposure system but seems quite impractical for a maskless system because of the much more rapid changes in total current. The Nikon corporation has been developing EPL and their published experimental results have not yet shown controlled feature sizes below 70 nm

at currents exceeding 1 lA. Furthermore, mask projection systems have a signicant advantage over maskless systems because all the feature edge biasing is done at the mask making stage so that relatively low resolution projection optics can be employed. This is one reason why optical projection of mask images has been so eective.

Fig. 10. IMS Vienna column concept to minimize space-charge blurring (courtesy of T.R. Groves, from [28]).

R.F. Pease / Microelectronic Engineering 7879 (2005) 381392

389

Recently, a European alliance has been formed to pursue a multi-bundle, single-axis system that employs projection optics designed to minimize space charge blurring (Fig. 10) [17,28]. The modulation is performed by deecting the beamlets over individual apertures rather than over a common aperture as in Fig. 8. At this time no experimental results have been reported.

7. Multiple-axis systems Many multiple-axis systems have been proposed. The earliest (Fig. 11) [18] was not maskless and was based on night vision tubes in which a chromium-on-quartz wafer was coated with a photoelectron emitter such as gold or cesium iodide. Photoelectrons from the clear regions were accelerated and focused in parallel, uniform E- and B-elds, at unity magnication, onto the resistcoated wafer. Working circuits were successfully

fabricated with such a tool in the late 1970s. The smallest features obtained were submicron. Although this was signicantly ner than the design rules current at that time, prototype optical steppers were already approaching the same resolution and development was abandoned. Moreover, there were problems with overlay errors caused by wafer bowing and because of contamination from the electrons striking the resist a new photocathode lm had to be evaporated onto the mask with each new batch of wafers; this raised concern about defects being generated. As described below, a maskless version of this approach is now being researched. A very dierent approach is to have an array of conventional single-axis, single-beam columns [20]. This approach was researched at IBM and developed at ETEC (12). Single column resolution approached 10 nm at 1 kV and an array of 2 2 columns occupying a cubic volume of about (50 mm)3 was demonstrated. However, the diculty of

Fig. 11. Distributed axis congurations (a) original ELIPS approach of OKeefe et al. [18]. (b) two-stage version [19]. The focusing of the aperture objects onto the wafer is entirely due to the uniform magnetic eld.

390

R.F. Pease / Microelectronic Engineering 7879 (2005) 381392

Fig. 12. Cross-section and STEM view of 50 nm diameter aperture formed by ion beam milling through 800 nm metal (courtesy of Daniel Pickard).

scaling this up to the much larger arrays needed to give total currents exceeding 10 lA appears to have stalled further development. Schemes featuring multiple columns each featuring multiple beams or shaped beams have also been proposed and built [21]. But the diculty of engineering an array of columns each matched in terms of focus and beam position has so far proved too dicult to attract serious industrial development. This hardly surprising considering that it still takes several weeks to install, with adequate control of beam position and beam focus, a commercial high-resolution electron beam writer featuring a single, xed-shape, beam. To mitigate the challenge of matching beams focused along dierent axes, Groves and co-workers [19a,b] proposed a re-incarnation and modication of the original photocathode system and a simplied version is being researched at Stanford University (Fig. 12). The focusing of the sources onto the wafer is brought about solely by a uniform magnetic eld thus, facilitating matching of the focus conditions and hopefully, eliminating the need for individual correction of astigmatism in the dierent beamlets. Moreover, the electrical deection is brought about by deection electrodes that are common to each row of beamlets which should facilitate the stitching of the sub patterns. The unity magnication of the uniform-eld focusing leads to the need for sources no larger than the required nal beam diameter and, to keep the elec-

tron optics simple, can be externally modulated. A photocathode was picked as the most promising source to illuminate a mechanical aperture that was fashioned by drilling through a Pt membrane with a focused ion beam system. Fifty nanometer diameter sources can be routinely fabricated in this way (Fig. 13) and apertures as small as 30 nm diameter have been demonstrated. Experimentally a resolution better than 50 nm has been demonstrated [19b]. The main obstacle to realizing this

Fig. 13. Recording-Erasure Cycle of thermoplastic hologram. A similar process could be used for generating a recongurable mask with light or electrons (courtesy of James C. Wyant).

R.F. Pease / Microelectronic Engineering 7879 (2005) 381392

391

source is the lack of a suciently bright and rugged photocathode source. Recently, some encouraging results have been obtained using GaN-based structures that exhibited an energy spread of less than 300 meV and current densities exceeding 100 A/cm2.

9.2. Re-congurable masks Nguyen [24] has suggested that between one extreme of a $ x00,000 mask (x > 1) that takes weeks to build and qualify and a maskless system which can pattern the area of a mask in 100 s there might be a middle ground in the form a recongurable mask. Mask repair of course is one form of reconguring but is too slow for full eld reconguring. The key is a reconguring process that is assured of introducing a negligible number of defects and could be performed in about an hour. One possibility is to use a writing charged particle beam to generate a chromeless phase mask directly using a thermoplastic recording [25] (Fig. 14). No development or etching is needed, just transient heating to soften the plastic to allow the charge induced deformation to take place. Lest this seem far-fetched the Eidophor system was used commercially for many years for theater television and employed much the same principle although an oil lm was used in place of the thermoplastic lm. A further embellishment is to use an array of AFM tips to locally heat and deform a lm. It could be argued that these technologies are really for mass storage but in general a photomask is simply a read-only memory. 9.3. Possible changes in circuit architecture An increasing fraction of system functions can be brought about using programmable gate arrays. Mask programmable gate arrays are already available and even electron beam maskless programmable gate arrays are now become available [26]. These are programmed by patterning just an array of vias so the cumbersome full service patterning described earlier is replaced by no frills patterning featuring a xed shape and only 1% area coverage. A similar scheme (Gate Flasher) was proposed some years ago by Berglund [27] for patterning just the gates of MOS circuits with a xed shape. The requirements for the gates are more stringent than those for upper-level vias and the proposal was not adopted at that time. Both schemes would benet from a system capable of covering (if not exposing) the total area of a wafer at 1 cm2/s.

8. Hybrid systems There are at least two systems that feature a combination of single- and multiple-axis optics. One is Mapper [22] that originated as a masked system and then migrated to the conguration shown. The modulation is carried out by an array of deectors activated by laser beams. A full description of this approach is given in a companion paper. The Canon company has described a developmental system [23]. The essence is multiple-axis component comprising a lens array (CLA) which is use to correct dynamically for space charge effects including eld curvature. Few, if any, experimental results are available at this time.

9. Strategic alternatives 9.1. A self-limiting situation? Suppose as a milestone, we are successful in developing a system featuring high resolution (<25 nm) in an array of pencils to give a throughput of 1 mm2/s, i.e., 1% of our target. Such a system could write a mask in about one tenth the time taken by a single bundle system and as such would make masks signicantly less (may be one quarter as) expensive in terms of dollars and turnaround time, thus eliminating much of the impetus of the project. It could be argued that inspection would still make masks prohibitively expensive but here again such a technology could be used also to accelerate inspection. The possibility of this scenario should not deter us from proceeding. Indeed it makes the project that much more attractive since even if it does not meet its original goal it will bring signicant benets.

392

R.F. Pease / Microelectronic Engineering 7879 (2005) 381392

10. Summary and conclusions Charged Particle Maskless Lithography (CPML2) is already being used in very low volume (0.001 cm2/s) production for features that are very dicult to achieve optically (65 nm and below). To be of signicant help to the semiconductor industry, the throughput must be increased to about 1 cm2/s. By the time such a system is available, the feature sizes of interest will be at and below 65 nm. The primary challenge for optical maskless lithography (OML) is realizing the enormous array (about 1e8) of modulatable light pencils. For presently available charged particle lithography (and EUV) systems, the main challenge appears to achieving the required current and resolution because of shot noise. For a given electron-optical conguration the throughput decreases as the fourth power of minimum feature size (below 50 nm). The goal of 25 nm features at 1 cm2/s might be achieved using a multi-axis electron beam approach in which the number of axes can be indenitely increased to keep up with the above fourth-power law. The author can see no other way of accomplishing this goal. Such a system could also be used for greatly accelerated (at least 100) SEM inspection. Alternative developments that might use a less ambitious maskless tool include programming gate arrays and reconguring masks.

References
[1] T. Sandstrom, J. Hintersteiner et al., in: SPIE Microlithography Symposium, 2004. [2] A. RoseAdvances in Electronics and Electron Physics, vol. 1, Academic Press, 1948. [3] T.E. Everhart, Ph.D Dissertation, Cambridge University, 1958. [4] C.W Oatley, The Scanning Electron Microscope, Cambridge University Press, Cambridge, 1976. [5] (a)C.A. Mead, I. Sutherland, T.E. Everhart, Report on DARPA Working Group on Lithography, 1976; (b)P. Leunissen, Determining the impact of statistical uctuations on resist edge roughness, MNE, 2004. [6] W.G. Oldham, in: Paper Presented to SPIE Symposium on Microlithography, Santa Clara, CA, 2002. [7] See ITRS website www.itrs.org. [8] D. Gil, R. Menon, H.I. Smith, J. Vac. Sci. Technol. B. 21 (2003) 28102814. [9] See, for example K.-N. Leung, J. Vac. Sci. Tech. B 17 (1999) 2776. [10] Record of the 5th LETI Conference, 2003. [11] W. Lu, et al., J. Vac. Sci. Technol. B 18 (2000) 3488. [12] See for example, almost any record of the SPIE BACUS meeting held annually. [13] D.R. Herriott, et al., IEEE T. Electron. Dev. 22 (July) (1975). [14] G. Winograd, et al., J. Vac. Sci. Technol. B 18 (2000) 3052. [15] (a)L. Han, et al., in: SPIE Conference on Charged Particle Optics, Denver Colo., SPIE, vol. 3777, 1999, p. 192; (b) L. Han, et al., J. Vac. Sci. Technol. B 18 (2000) 2999. [16] S. Golladay, et al., J. Vac. Sci. Technol. B 18 (2000) 3072. [17] T.R. Groves, private communication. [18] T.W. OKeefe, et al., in: Paper Presented at IEEE IEDM, 1967. [19] (a) T.R. Groves, R.A. Kendall, J. Vac. Sci. Technol. B 16 (1998) 1368; (b) D.S. Pickard, T.R. Groves et al., J. Vac. Sci. Technol. B 20 (2002) 2662. [20] T.H.P. Chang, et al., J. Vac. Sci. Technol. B 17 (1999) 2814. [21] E. Yin, et al., J. Vac. Sci. Technol. B 18 (2000) 3126. [22] P. Kruit, et al., in: Papers Presented at EIPBTN, 2004 and MNE 2004. [23] M. Muraki, S. Gotoh, J. Vac. Sci. Technol. B 18 (2000) 3061. [24] C. Nguyen, private communication, July 2003. [25] W.E. Glenn, Recording of Information by Electron Beams (1962), NBS# 6204005. (Box 191, folder 12). [26] For example, see E-Asic web site http://www.easic.com/ technolgy/ebeam.html. [27] C.N. Berglund, private communication, 2000. [28] C. Brandstatter, H. Loeschner, G. Stengl, G. Lammer, H. Bushbeck, E. PLatzgummer, H. Doering, T. Elster, O. Fortagne, Projection Maskless Lithography, in: Proceedings of SPIE, vol. 5374, Emerging Lithographic Technologies, May 2004, pp. 601609.

Acknowledgements The preparation of this paper was supported primarily by the DARPA Advanced Lithography Program and the Semiconductor Research Corporation. The author acknowledges valuable discussions with Mark McCord (KLA Tencor), Clark Nguyen (DARPA MTO), Dan Pickard (Stanford University), Bill Oldham (U.C. Berkeley), T.R. Groves (Leica MicroSystems), Pieter Kruit (T.U. Delft), H.I. Smith (MIT) and many others.

Das könnte Ihnen auch gefallen