ADVANCED TECHNOLOGY FOR EXTENDING OPTICAL LITHOGRAPHY

ADVANCED TECHNOLOGY FOR EXTENDING OPTICAL LITHOGRAPHY Christian Wagner a, Winfried Kaiser a, Jan Mulkens b, Donis G. Flagello c a Carl Zeiss, D-73446 Oberkochen, Germany; b ASM Lithography, De Run 1110, LA Veldhoven, The Netherlands; c ASM Lithography, 8555 S. River Parkway, Tempe AZ, U.S.A. ABSTRACT This paper discusses the technological consequences of extending optical lithography down to 50 nm. A total systems approach is used to define the overall error budgets on CD and overlay. The feasibility of extremely high NA optics is studied. With extremely high-na optics, Depth of Focus (DoF) decreases and focus control becomes more critical. Using a Step & Scan system, the potential of system dynamics is evaluated. Mask issues are briefly reviewed. Experimental data on field extension with butting is included. As well as being technically feasible, the lithography step for the future technology nodes must be cost effective. It is concluded that optical extension should preferably be supported by reduced field (<22mm), large magnification (>5x) scanners. The transition to Next Generation Lithography (NGL) will be pushed to at least the 50 nm node. 1. INTRODUCTION This paper presents a vision of how optics-based exposure tools will look, when the transition to NGL techniques finally happens. The ongoing acceleration of the semiconductor roadmap forces the industry to extend optical lithography much further than ever expected. Including 157 nm wavelength radiation, it is commonly accepted today that optical lithography will enable manufacturing at the 70 nm node. Translating the semiconductor roadmap into an exposure tool roadmap, it becomes apparent that new wavelengths and extremely high-na optics will be needed. To assure sufficient process latitude, resolution enhancement methods will be implemented in volume manufacturing. As well as advanced mask technology, layer-tailored illumination schemes and processes will be used. With the use of phase shift masks, the wavefront aberrations become increasingly important in both the CD and overlay budgets. With optics exceeding 0.8 NA, polarization effects on both the reticle and the resist film must be critically reviewed. To control the polarization level, and to support the pupil illumination schemes, advanced illumination technology is needed. The paper is organized as follows: - Section 2 focuses on the basic requirements for wavelength reduction, high-na optics and low-k 1 imaging. - Section 3 assesses the technology needed for these requirements and options - Section 4 summarizes and discusses the findings of Sections 2 and 3 - Section 5 is the Conclusion Please note that the lens and machine performance requirements are discussed in detail in an accompanying paper. [1] k 1 -factor 0.9 0.8 0.7 0.6 0.5 0.4 0.3 NA (248) k 1 (248) 0.2 NA (193) k 1 (193) NA (157) k 0.1 1 (157) NA (126) k 1 (126) 0 250 230 210 190 170 150 130 110 resolution node (nm) I-14526.ILL 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 90 70 50 numerical aperture NA To summarize the choice is not whether to use extreme NA, low k 1 or lower wavelength, but rather to use all these features. This is summed up in Figure 1, showing the various applicable wavelengths, the exposure-tool NA and the k 1 that will be attached to the desired resolution node. Figure 1: Exposure tool NA and k 1 for resolution nodes from 250 nm down to 50 nm. The wavelength is indicated by the shape of the data points. 1

2. EXTENDING OPTICAL LITHOGRAPHY Figure 2 shows the major building blocks of the roadmap as discussed in the introduction. To reach the 50 nm node, wavelength, NA and k-factors have to be driven to the limits simultaneously. Wavelength Resolution = k1 Numerical Aperture /NA Low-k1 I-14527.ILL New lasers, optical design concepts, materials, coatings and resist will have to be developed to reduce the wavelength. The numerical aperture leads to a reduced DoF, thus even tighter focus budgets than today. Extremely high-na optics impose major challenges on aberration control for lens design and large lens elements must be introduced. The low-k 1 imaging needs layer-tailored illumination schemes, feature-specific mask layout (OPC), thin-film imaging and extremely tight aberration control. In the following sections, we discuss the requirements driven by the three arms of the diagram; wavelength, NA and low k 1. Laser Optics Design Materials/ Coatings Purging Resist Focus Budget Lens Element Size Design Illumination Mask Process Aberrations 2.1 Wavelength Reduction We have summarized in Table 1 the issues and requirements of extending lithography by wavelength. Down to 157 nm, lasers will be available as well as optics designs and materials for building the lenses. 248 nm resist has allowed the resolution to be initially scaled down from 0.25 µm to below 0.13 µm. 193 nm resists are rapidly improving at present. High-contrast resists for 0.13 µm resolution volume production are within reach, however, production-worthy 0.10 µm resists need to be proven. For 157 nm resist, absorption is a challenge. Regarding mask technology, the attenuated phase shift material for 193 nm and 157 nm especially needs to be developed. At 157 nm, the pellicle and contamination issues have to be solved. Figure 2: Rayleigh diagram shows the resolution limit and the issues involved. From 157 nm and below purging of the optical paths will be required. Extending optical lithography further to 126 nm does not seem very likely. With pure mirror designs (catoptric) as applied, for example, in EUVL it seems unlikely that the necessary NA level of e.g. > 0.7, as shown in Figure 1, will ever be reached. In order to make 126 nm an option, only catadioptric designs seem applicable. MgF 2 and LiF are the only candidates for a refractive material. MgF 2 is birefringent in the level of 0.01. For LiF there is no data on laser hardness and the feasibility to grow large crystals. In addition to the optics, mask and resist materials must be developed. Further, the feasibility of a 126 nm laser source Laser Optics Resist Mask Technology System Issues 248 nm KrF laser available high-na optics with fused silica high-na optics with fused silica and calcium fluoride supports evolution to <130 nm workable 193 nm ArF laser available contrast, production worthiness need material for attpsm 157 nm F 2 laser under development high-na optics with calcium fluoride and possibly alternative fluorides catoptric has limited NA, catadioptric needs lens material high absorption modified fused silica substrate, develop pellicle purging, contamination 126 nm no laser to be developed no material (?), reflective masks purging with He Table 1: The issues concerning wavelength reduction. 2

must be shown. Although it cannot be said that 126 nm can never be achieved, it does not appear to be possible either 2.2 High-NA Optics Design This section begins with an overview of the preferred lens design options and issues for the next steps for optical extension. Refractive designs have been greatly improved in terms of NA and residual aberrations over the last decade. 248 nm is used as a starting point. The bandwidth of line-narrowed KrF lasers allows refractive optics to be made of a single material, namely fused silica, only. 193 nm refractive designs need CaF 2 material in order to achromatise the lenses to today s state-of-the-art lasers which have a bandwidth of < 0.5 pm. Currently, the most likely solution for 157 nm is a catadioptric design (preferably without beam splitter and without obscuration as well). The ongoing use of refractive designs might again have a chance with a CaF 2 /BaF 2 or CaF 2 /NaF combination. This however depends not only on the availability of the crystals but also on the minimum laser bandwidths that might be achieved in the timeframe of 157 nm production tools. The challenge for catadioptric designs without beam splitter is either a central obscuration or an off-axis field. Off-axis fields would lead to changes in the exposure tool layout as well as new calibration and optimization strategies. Note that all design types have to be extendible to NA>0.8 in order to support at least the 70 nm node with optical lithography. With increasing NA, mirrors have the undesired tendency to obscure light that should reach the wafer. Besides the challenge to correct for higher aberrations, the geometrical boundary conditions imposed by mirror optics are an additional challenge for lens design. This is one of the reasons why refractive designs are a better choice than catadioptric solutions. At this point, it is worthwhile to discuss the NA and wavelength dependency of the laser bandwidth. To connect with today s scanners a slit height of 26 mm is assumed. For 248 nm we use an all-fused-silica refractive design and, for 193 nm, a typical achromatization with CaF 2. For 157 nm the bandwidth of state of the art single-line lasers is about 1 pm. This is consistent with the catadioptric design solutions as discussed above. For refractive solutions, much still depends on material dispersion (which is still not known exactly for all crystal types) and the design layout. For the existing 1 pm single-line 157 nm lasers, refractive designs with a CaF 2 /BaF 2 combination exist. However the large material volume needed in order to build these lenses makes this solution only feasible in case of a reduced (< 22 mm) slit height. In order to reduce material volume to a realistic level, a laser bandwidth < 0.5 pm will be needed. It should be noted that in any case the diameter of the lens image-field height (by track length and thus optical path) affects the laser bandwidth in a linear fashion. Thus, a reduction in field size will help to relax laser bandwidth requirements and thus laser cost. This argument is in favour of scanners since here the lens field is significantly smaller than the stepper field. Figure 3(b) charts the relative optics technology effort as a function of NA, independent of wavelength-specific technology. To a first-order approximation, lens element diameters are proportional to the Tangens of the aperture angle (again for the 26 mm slit height). Material volume follows the second order of this, when it is assumed that the Preferred High NA-Design Option Coatings Materials Laser System Integration 248 nm Refractive evolution 193 nm 157 nm Table 2: Refractive Beam splitter, Catadioptric w/o beam splitter Refractive development for higher NA additional HR coating development AR coatings under development Preferred design options for 248 nm, 193 nm and 157 nm. fused silica available, larger blank sizes needed fused silica available, larger blank sizes needed, CaF 2 volume and quality CaF 2 volume and quality 2nd material for color correction 0.5 pm to 0.3 pm bandwidth 0.5 pm to 0.3 pm under development >1 pm bandwidth available <1pm bandwidth under development standard standard solution with either off-axis field or obscuration standard 3

bandwidth (pm) 1.2 1.0 0.8 0.6 0.4 0.2 248nm 193nm 0.0 0.5 0.6 0.7 0.8 0.9 I-14528.ILL optics technology effort 3.0 2.5 2.0 1.5 1.0 0.5 0.0 0.5 0.6 0.7 0.8 0.9 I-14529.ILL numerical aperture (NA) numerical aperture (NA) (a) (b) Figure 3: (a) Laser bandwidth vs NA for 248 nm and 193 nm refractive designs and (b) Optics technology effort vs NA track length of the lens stays relatively constant summarizing design innovations, technology advances and continuous improvement processes. Note that the reduction to a 22 mm wide field will, by scaling, reduce lens material at least by the diagonal of the field squared. Theoretically, the amount of lens material follows the third power of the field diagonal. For practical designs, mechanical boundary conditions lead to a somewhat smaller sensitivity. Thus, one calculates that a 22 mm square field stepper needs about twice as much material as a scanner with 22 mm slit height. 2.3 Low-k 1 Imaging Low k 1 -imaging only works when mask layout, illumination condition and resist process are optimized simultaneously. With alternating phase shift masks entering the scene, chip design rules will have to be integrated into this global optimization strategy. Considering the overwhelming cost of NGL techniques, low k 1 imaging may be explored as a comparatively low-cost application (even if it is common knowledge that mask costs go up and double exposure will lead to reduced throughput). Table 3 below summarizes the commonly accepted set of low k 1 imaging optimization strategies. k 1 -factor >0.5 ~ 0.5 0.42 ~ 0.42 0.35 < 0.35 Mask Illum Mask Illum Mask Illum Mask Illum Memory binary Conventional half tone annular half tone QUASAR hard PSM coherent binary Annular binary annular half tone annular Logic binary Conventional Table 3: binary assist features binary annular conventional Binary assist features halftone assist features QUASAR halftone assist features QUASAR QUASAR hard PSM coherent Imaging optimization strategies for low-k 1 imaging. Note that the k 1 -boundaries are softer than the table might suggest. Where possible, two solutions are given. 4

Exposure latitude for dense and isolated lines depends significantly on the distribution of the diffracted light in the entrance pupil of the lens. In the range of k 1 factors between the theoretical limit of 0.25 and 0.35 the interfering rays are roughly located on opposite sides of the entrance pupil. This situation is created by either using alternating phase shifting masks or using chrome masks and dipole illumination. [2,3] We give a simple analytical analysis for dense lines realized by alternating phase shift masks. The lens contrast K is given by Equation (1) as written below: K = 100% for 1/4k 1 <1 σ (1a) K = (1-1/4k 1 )/(1/2+σ/2-1/8k 1 ) * 100% for 1 σ <1/4k 1 <1( (1b) K = 0% for 4k 1 <1 (1c) To arrive at a readable formula, we made a 1st order approximation with square illumination poles (surface 4σ 2 ) penetrating the NA boundary, which was assumed to be a straight line (it is easier to cut rectangles than to cut circles). For a PSM situation the minimum resolution CD min is given by k 1 =1/4. Figure 4 shows the light distribution in the pupil for a low σ that is close to minimum resolution. Figure 4: 1/4k 1 Pupil diagram showing the intensity distribution for phase-shift imaging at k 1 close to 0.25. The pupil radius (NA) is set to unity. As can be seen in Figure 5, if one approaches the 0.25 limit for the k 1 factor, small σ settings become extremely helpful. Although the theoretical resolution limit does not I-14530.ILL depend on σ, the minimum resolution that gives a certain level of contrast (exposure latitude) may be reduced by decreasing the coherence. exposure latitude 35 30 25 20 15 Figure 5: 10 s=0.1 5 s=0.2 s=0.3 0 0.40 0.35 0.30 0.25 k 1 -factor Exposure latitude for different s in the approximation of formula (1). In order to arrive at exposure latitude the contrast was scaled with 0.3, which is a rule of thumb supported by today s DUV experience. [2] Let us discuss the relation between two-beam imaging and polarization effects. The light reaches the wafer only from the outer regions of the pupil. The angle between the interfering rays is thus about 2arcsin(NA/n) with n being the refractive index of the resist. For numerical apertures exceeding 0.7 and k-factors around 0.3 polarization effects at image recombination in the aerial image will play a major role. At first dose effects must be controlled. Depending on whether the light is polarized in or out of the plane of the incoming rays, the Fresnel equations predict dose differences in the order of >20%. In order to limit HV differences the residual polarization should be below 10%. [1] Thus, in addition to intensity uniformity, polarization uniformity must be established. The second effect is that light being polarized in the plane of the incoming rays does not fully interfere anymore (the extreme case of a 90 angle is prevented in the resist for typical refractive indices). In order to optimize the contrast the polarization vectors should be orthogonal to the plane of the incoming rays. For dipole illumination using double exposure this may be easily realized by using linear polarization parallel to the mask features for each of the two masks. In the following we review illumination requirements. A summary is given in Table 4 below. Referring back to Figure 5, an insight into the additional requirements that come from the dipole and PSM exposure conditions can be seen. As shown the exposure latitude depends I-14531.ILL 5

strongly on the actual coherence (σ). This, at first, means that the absolute accuracy of s has to come down well below 0.01. This, in turn, leads to the requirement, that the across-field variation of σ has to be controlled extremely well. However, the across field variation of s will be more difficult to reach because of the small poles. In order to preserve high machine throughput, it is clear that the small poles cannot be realized by aperture stops in the illuminator, but by actively distributing the light, for example, by the use of customized diffractive optical elements Energy Lens rms Reticle CD CD Budget I-14533.ILL Focus Leveling Lens Focal Plane The pole-balance error goes into the telecentricity budget at wafer level. The intensity imbalance between the opposite sides of the pupil, lead to a telecentricity error. Imaging simulations show that for high-na, low k 1 imaging the pole-balance error should be limited to below 2%. Specification σ-stability across field < 0.01 pole balance < 2% residual polarization < 10% Table 4: Illuminator requirements for 2-beam interference. 3. TECHNOLOGY SUPPORTING THE OPTICAL EXTENSION 3.1 Introduction The section discusses the technical efforts to support the optical extension strategy. We will link the formal solutions of the extension path with CD and overlay performance requirements as given by the ITRS. [4] CD and overlay budgets are shown schematically by the simplified budget trees below. Note that the energy budget is an extended dose concept where all contributions to CD-error in best focus are summarized. Higher order lens aberrations are grouped by focus (even aberrations) and overlay (odd aberrations). For a first estimate one may assume that all sub-budgets will scale down with resolution. 3.2 Optics Technology The imaging requirements of the technology nodes must be broken down to the level of optical component specifications. As a first step, the dependence of CD-uniformity and overlay on lens aberrations must be determined. The absolute sensitivity depends on feature type, mask and illumination. For example, in a k 1 =0.3 process with 5% CD-uniformity due to lens aberrations, a wavefront rms of 0.020λ is needed. Regarding overlay the most stringent requirements come from Figure 6: Dose MSD Overlay Budget Aligner Reticle Lens (a) CD budget (b) overlay budget Even Aberrations Distortion Odd Aberrations printing isolated lines with alternating phase shift masks. The reason is that 2-beam imaging which uses a small σ is extremely sensitive to odd aberrations. To minimize pattern shift, both lens distortion and pattern-specific specification of single odd Zernike coefficients will be necessary. A detailed discussion on the imaging requirements of future exposure systems is given by Flagello et. al. in this volume. [1] For all types of lens aberrations an error budget must be broken down and component specifications must be calculated. The lens error budget consists of 6 major parts, as shown in Figure 7. One part comes from residual design aberrations. A major part is the component errors that arise from surface figure and homogeneity contributions. Here it must include only the residual errors after optimization in the budget, because these errors are optimized by adjusting and compensation. The same is true for mounting-induced lens deformations and lens element tilt and decenters. I-14534.ILL 6

Total Lens Error Budget I-14535.ILL Design Component Errors Assembly Lifetime Errors Lens Heating Environmental Errors Homogeneity of Material Lens Elements Surface Figure Mounting Induced Lens Deformation Decenters Tilts Material Absorption Coating Absorption p T ( ) Figure 7: Top level error budget The next part is made up of lifetime (compaction, induced absorption) and lens heating errors. Lens heating arises from the light absorbed either in the coating or in the bulk material. The last part is the so-called environmental errors. These errors include variations in ambient pressure and temperature. In the budget we only include the residual errors after optimizing the lens with a specific set of lens actuators. In the environmental errors we also include the residual errors after adjusting the lens magnification. The optics technology roadmaps are driven by (a) the requirements that come from imaging, overlay and throughput and (b) the increase in lens NA in order to support future resolution requirements. The core competencies in building high-na optical systems may be described along the process chain, with incoming materials, polishing and coating of the lens elements, mounting, assembly, adjustment and metrology capabilities. In order to take care of the requirements for 2-beam imaging (dipole and PSM), coma and all odd aberrations must be controlled extremely well. Odd aberrations come from homogeneity and surface errors and from tilt and decenter errors in the lens at assembly. The difficulty in reaching these specifications increase along with NA, because the sensitivity of the higher-order odd aberrations to lens element tilts and decenters increases. In order to limit these errors, all mounting and assembly processes have tolerances in the sub-µm regime and thus push mechanical fabrication to the limits. Figure 8 shows the improvement roadmap of surface figure on fused silica. The years given in the time scale refer to system delivery and thus mark the time from the first 248 nm volume production tools to the introduction Technology fused silica calcium fluoride Polishing Coating Mounting assembly and adjustment Metrology Table 5: Imaging, Overlay, Throughput homogeneity transmission induced effects homogeneity internal quality continuous quality improvement (surface figure) transmission absorption tilt and decenter requirements improved environmental control advanced optimization strategies improve accuracy for 100 nm node blank size metrology blank size metrology metrology of 193 nm production tools. In order to support future low k 1 imaging requirements the pace of the improvements must be maintained. Figure 9 shows how improvements of the optimization process lead to significant reduction of the wavefront rms as a function of the field position for a 0.7 NA KrF lens. The across field maximum of the rms was improved from 6.2 nm to 4.6 nm. NA high incidence angles (i max ) broad bandwidth ( i) polarization effects tilt and decenter requirements Table 5: Overview of high-na technical challenges 7

surface figure a.u. 1.2 1.0 0.8 0.6 0.4 0.2 0.0 1995 1996 1997 1998 1999 2000 2001 system delivery Figure 8: standard Figure 9: Improvement of critical surface figure on fused silica. Data scaled to fused silica performance in 1996. improved Interferometer data showing the reduction of across-field wavefront rms after advanced optimization. Wavefront levels < 0.020λ are necessary in order to support 0.30<k 1 <0.40 range. [1] In addition to rms, single Zernike coefficients need to be controlled extremely well. 3.3 Machine Technology rms (nm) 6 Extending optical lithography with high-na optics and decreasing wavelengths also sets challenging requirements on the Step & Scan system performance. Focus and machine dynamics are seen as especially critical. Other items are related to the implementation of 157 nm, which requires a completely N 2 purged optical path, and implementation of low k 1 imaging schemes like multiple exposures. 5 4 3 I-14537.ILL I-14536.ILL 3.3.1 Focus and MSD DoF is one of the most obvious issues related to high-na imaging. The traditional k 2 DoF scaling law, derived for dense lines and three beam imaging, leads to extremely low DoF for high-na optics. On the other hand, low k 1 imaging enhancements offer a possible increase in focus latitude. An example of this, using assist features, can be found in Ref. [2]. As a consequence, it is difficult to predict what the exact workable DoF will be at a certain technology level. Figure 10 shows a tentative DoF-roadmap (isolated line), based on assumptions of applicable wavelength, NA and enhancements per node. Both, the introduction of new wavelength technology and the use of imaging enhancements is assumed to be conservative, thus making the DoF numbers represent a worst-case scenario. As a reference the ITRS UDoF numbers are included. [4] DoF (nm) 800 600 400 200 0 Figure 10: KrF ArF 130 100 70 50 node (nm) DOF roadmap for isolated lines Roadmap NA-scaled Referring to the focus budget diagram in Figure 11, improvements are needed to lower the effects of the optics, the leveling and focus calibration components. Lens-induced focus errors include focal plane deviation (FPD) and high-order even-aberrations, leading to additional focal plane deviation varying with NA and σ settings. Additionally lens drift contributions caused by environmental changes (like ambient pressure) or radiation-induced effects (like lens heating), must be accounted for in the system metrology concept. The leveling concept interacts with user-related variations like wafer topology and reticle unflatness. Although both are expected to follow the focus roadmap as well, advancements in Step & Scan system technologies might alleviate the impact. As an example, reticle unflatness information can be partly corrected for by adjusting the reticle stage height and tilt position during a scan. Wafer topography not only directly consumes focus budget but also induces fading. In Figure 12 it can be seen that, during a scanning exposure, the wafer is leveled and focused depending on the actual wafer topography information in the exposed slit area. The focus loss due F2 I-14538.ILL 8

to topography is expressed as Moving Average (MA). Topography also reduces image contrast because of fading. Fading, quantified with Moving Standard Deviation (MSD) originates from local fluctuations of the wafer surface on a scale smaller than the slit width. If the wafer topography as a spatial frequency power spectrum is described, then MA(z) and MSD(z) can be calculated as a function of this spatial frequency. Focus Budget Aerial Image Levelling Calibration I-14539.ILL worst case focus (nm) 200 180 160 140 120 100 80 60 40 20 0 10 15 20 25 30 slit height (mm) MA MSD I-14541.ILL Focal Plane Deviation Even Higher Order Aberrations Lens Drift Wafer Topology Wafer Clamping Level Sensor Best Focus Process Offsets Figure 13: Worst case MA(z) and MSD(z) as a function of the slit dimension for a critical topography wafer. Image contrast is also degraded because of stage synchronization errors. These errors, quantified by MSD (xy), should scale with CD. To support sub-100 nm resolution a new vibration isolation concept is implemented in the new Step & Scan platform. 3.3.2 Multiple Exposure Reticle Surface Figure 11: Figure 12: Focus budget α slit area focus error Impact of wafer topography on MA and MSD focus behavior. Relative to a square stepper field, the topography-induced focus performance is strongly improved because of the Step & Scan on-the-fly leveling principle. Figure 13 makes clear that a further reduction of the field size in the non-scanning direction could improve the MA and MSD focus behavior even more. A reduction of the slit height of about 25% already decreases MSD by about the same relative number. The relative improvement in MA is somewhat less. Although the behavior on CMP wafers is better than the data given by this case study, the use of reduced slit sizes supports the high NA implementation. I-14540.ILL Double exposure technology, either using c:psm [3] or dual illumination, is expected for the extremely low k 1 applications. The Step & Scan machine must support this enhancement method, both in throughput as well as accuracy. Image placement errors of the sub-images are regarded as critical. They can be optics-induced as well as machine-induced. Another possible application for multiple exposure schemes can be found in so-called image field extensions. One might consider reduced field optics, especially for the hyper-na optics where costs and focus budget become critical. For large chip sizes however, fields then need to be enlarged, either using a two-mask field-stitching approach or using a raster scan approach. Some simple dual exposure experiments are performed to evaluate current sensitivity for placement errors. As shown in Figure 14, 0.30 µm lines were exposed as the first layer and, subsequently, 0.15 µm lines as the second layer. Both line sizes have been butted together resulting in a fork-shaped pattern extension. The 0.30 µm lines were printed with conventional illumination and the 0.15 µm lines were printed with annular illumination. Line placement errors in this experiment can come from the odd aberrations in combination with stage positioning errors. In Figure 14 the measured pattern shift as a function of the slit height is given. The maximum observed line shift is 6 nm, which is well below the allowed 9

value of 10 nm in a field stitching budget for 150 nm lines. A dedicated machine design for stitching should make seamless stitching possible. 3.4 Mask Performance Low k 1 imaging imposes stringent requirements on mask technology. While supporting the resolution roadmap, additional sub-resolution features (serifs, assist lines etc.) are used in order to preserve depth of focus and exposure latitude requirements. Given below are the requirements for the 100 nm and 70 nm node, with remarks on the feasibility. Compared to today s best performance data, it is believed that improving the mask parameters by more than 50% will be extremely difficult. CD uniformity is seen as especially critical. A change of reduction ratio to 5X or even 6X would significantly relieve the problem. We foresee that this change will be needed for the 70 nm node. pattern shift (nm) 9 6 3 0-3 -6-9 -15-10 -5 0 5 10 15 slit position (mm) Figure 14: Critical pattern extension experiment with double exposure. Pattern shift for 0.15 µm dense lines relative to 0.30 µm dense lines, obtained with a double exposure using a annular and conventional illumination setting respectively. I-14543.ILL Minimum feature size @ mask Image placement error @ wafer CD uniformity @ mask Best today 100 nm node 70 nm node 50 nm node ~ 250 nm -20% -45% -60% 4X okay -30% -50% 5X okay -15% -40% 6X ~ 25 nm -20% -45% -60% 4X ~ 10 nm okay -25% -45% 5X okay -15% -40% 6X -20% -50% -70% 4X okay -40% -60% 5X okay -30% -55% 6X critical MEF 2.0 2.3 2.6 Table 6: Necessary relative improvements of the critical mask performance parameter for different reduction ratio scenario. [5] 10

4. Discussion Combining the data of this paper the most likely lithography roadmap will be: 70 nm, the last fully optical node: - 193 nm and 157 nm as wavelength - Possible double exposure (c:psm for logic and dipole illumination for DRAM) - Lens numerical apertures of around 0.8 for 193 nm give a k 1 of 0.3 - Efforts are needed to make a production-worthy resist for 157 nm at the 70 nm node - CaF 2 must be ramped up to mass production - Considering both mask issues and necessary field sizes, a 5X reduction ratio is recommended, resulting in a 22 mm x 26 mm maximum exposure field. 50 nm, may be supported optically, however, there are competing NGL options (e.g. EUVL): - 157 nm as wavelength - Phase-shift two-beam technology (imaging at k 1 <0.3) - Low σ (<0.2) illumination with tight polarization control - Hyper NAs of close to 0.9 - Extreme sensitivity to all kinds of lens aberrations must be accounted for in lens design and manufacturing. At 50 nm the reduction in DoF and the increased complexity in lens technology may enforce a further reduction of field size, this, possibly, in combination with higher reduction ratio (6X). The latter is necessary to alleviate the pressure on the mask CD requirements, which become increasingly difficult because of the high MEF. Depending on the performance of the by-then-emerged EUV β-systems, EUVL might be inserted in that time frame as well. With the current issues, we do not see further extension of optical lithography below 50 nm. 126 nm lacks infrastructure on all the important sub-systems (laser, materials, resists, masks). In this case, EUVL will completely take over. We should not forget that the semiconductor industry is not only technology driven. It is also the most cost-effective technology that will be used for IC production. Manufacturing cost will increase per technology node. To minimize the cost increase, both in exposure tools and masks, one should consider a paradigm shift in chip size developments and reduction ratio. In accordance with the 1999 edition of the roadmap, 22 mm x 22 mm field size machines are suitable for both DRAM and Logic manufacturing. For larger chip sizes (e.g. SoC) one should consider stitching. It seems obvious that a 22 mm slit combined with 5X reduction is a better choice than today s standard of 26 mm slit and 4X reduction. 5. CONCLUSION We conclude that optical lithography is able to go well into the 70 nm node, and, for most applications, can be extended to 50 nm as well. The latter step requires 157 nm with extreme numerical apertures, phase shifting masks or dipole exposure. Using optics at the 50 nm node is technologically challenging. In order to be cost-effective, optimize yield and reduce the technical risks of very high NA lenses, field sizes of 22 mm or smaller and reduction ratios of 5X or larger are recommended. ACKNOWLEDGEMENTS The authors would like to thank Karl-Heinz Schuster, Daniel Krähmer, Willy Ulrich, Christian Hembd-Söllner, Reiner Garreis, Gerd Reisinger, Karlfrid Osterried, Uwe Mickan, Jan Baselmans, Pete Jenkins, Jo Finders, Marc Boonman, Denis Faas, Tammo Uitterdijk, Will Pijnenburg, and Judon Stoeldraijer for inputs and discussions on low-k 1 -imaging technology. We would also like to thank Vivian Kim, Jon Beckton, Marcel van Dijk and Lies Lisens for their help in preparing the manuscript. REFERENCES [1] D.G. Flagello et. al., Optical Lithography into the millennium: sensitivity to aberrations, vibration and polarization, SPIE Vol. 4000. [2] Jo Finders et. al, DUV Lithography (KrF) for 130 nm using off-axis illumination and assisting features, Semicon Japan 1999. [3] Michael E Kling et al., Practicing extension of 248 DUV optical lithography using trim-mask PSM, SPIE Vol. 3679. [4] International Technology Roadmap for Semiconductors, 1999 Edition, Lithography. [5] J. Mulkens et. al., Challenges and Opportunities for 157nm Mask Technology, SPIE Vol. 3873. 11

4022 502 94109