Defect printability and inspection studies were conducted on a programmed EUV defect mask. The mask was fabricated using Ta-based absorber stack on a Mo/Si multilayer coated 6025 plate. The defect pattern contains a variety of types of defects. The defect printing was performed on the Engineering Test Stand (ETS), which is the 0.1 NA EUV scanner at Sandia National Laboratories in Livermore, CA. The result showed that the printability of defects depended on the defect type and that either notches in or protrusions from absorber lines were the first to print. The minimum printable defect size was approximately 15 nm (1X). Defect inspection was performed on a 257-nm wavelength mask inspection system in die-to-die mode. Seventy-eight out of 120 programmed defects were detected when using 50% detection sensitivity. Maximum detection sensitivity was also tried. However, the number of defects is overwhelmed by the nuisance defects. The minimum defect detected was 52 nm in width. Simulations with a 2-D scalar model are used to verify the results.
We have exposed 10 wafers on the Engineering Test Stand (ETS), the 0.1 NA EUV scanner at Sandia National Laboratories in Livermore, CA. The EUV reflective mask was fabricated in-house using a Ta-based absorber stack on Mo/Si multilayers. The printed wafers contained different line sizes and pitches, line-end shortening measurement structures, contact holes, and patterns for estimating absorber defect printability. The depths of focus of each feature are typically 2 um due to the small NA of the scanner, and these should decrease by at least a factor of 6.25 as the NA's increase to 0.25. The data from measurements of line size through pitch and line-end shortening test structures indicate that both 1D and 2D optical proximity correction will be required. Defects that are either notches in or protrusions from absorber lines are the first to print, and they begin to print when they reach approximately 15~nm (1X) in size. This size threshold is in accordance with the 2003 ITRS specifications. We also report the first printing of SRAM bitcells with EUV lithography.
We demonstrate direct flare measurements of 4-mirror projection optics in the Engineering Test Stand (ETS) using a conventional resist clearing method (the Kirk method). Two extreme UV lithographic projection optics, one with higher flare than the other, have been characterized and the results compared. The measured results have also been compared to analytical calculations based on measured mirror roughness and the extended point spread function. Full-field flare across the 24 mm field width has been measured, and we have verified that flare is constant across the field for EUV lithography as predicted. Horizontal (H) and vertical (V) flare bias has been observed and the cause of the H-V flare bias has been investigated. The main cause has been identified to be anisotropic mirror polishing. Simulations with the 2D Power Spectral density function have confirmed the experimental results.
Static and scanned images of 100nm dense features for a developmental set of l/14 optics (projection optics box # 1, POB 1) in the Engineering Test Stand (ETS) were successfully obtained with various LPP source powers last year. The ETS with POB1 has been used to understand initial system performance and lithographic learning. Since then, numerous system upgrades have been made to improve ETS lithographic performance to meet or exceed the original design objectives. The most important upgrade is the replacement of POB 1 with an improved projection optics system, POB2, having lower figure error (l/20 rms wavefront error) and lower flare. Both projection optics boxes are a four-mirror design with a 0.1 numerical aperture. Scanned 70-nm dense features have been successfully printed using POB2. Aerial image contrast measurements have been made using the resist clearing method. The results are in good agreement to previous POB2 aerial image contrast measurements at the subfield exposure station (SES) at Lawrence Berkeley National Laboratory. For small features the results deviate from the modeling predictions due to the inherent resolution limit of the resist. The intrinsic flare of POB2 was also characterized. The experimental results were in excellent agreement with modeling predictions. As predicted, the flare in POB2 is less than 20% for 2μm features, which is two times lower than the flare in POB1. EUV flare is much easier to compensate for than its DUV counterpart due to its greater degree of uniformity and predictability. The lithographic learning obtained from the ETS will be used in the development of EUV High Volume Manufacturing tools. This paper describes the ETS tool ETS tool setup, both static and scanned, that was required after the installation of POB2. The paper will also describe the lithographic characterization of POB2 in the ETS and cmpare those results to the lithographic results obtained last year with POB1.
While interferometry is routinely used for the characterization and alignment of lithographic optics, the ultimate measure of performance for these optical systems is the transfer of an image or pattern into photoresist. Simple yet flexible exposure systems play an important role in this task because they allow complex system-dependent effects to be isolated from the printing results. This enables the most direct lithography evalaution of the optical system under investigation. To address tehse issues for commercial-class EUV optics, a synchrotron-based programmable illuminator exposrue station has been implemented at Lawrence Berkeley National Laboratory (the Advanced Light Source). As previously presented, this static microfield exposure system has been used to lithography characterize a 4-mirror optical system designed for the EUV Engineering Test Stand (ETS) prototype stepper. Based on the lithographic characterization, here we present a detailed performance analysis of the 0.1-NA ETS Set-2 optic. Operation of the static printing system with the Set-2 optic yielded approximately 330 exposed wafers, where each wafer contains one or more focus-exposure matrices. A wide variety of parameters were studied includign, among others, illumination conditions, resist thickness, and mask tone. Here we present a subset of this data in terms of process-window results. The resutls demonstrate a depth of focus (DOF) approximately 2μm for isolated 70-nm line features, 1 μm for nested 70-nm line features, and 0.5μm for 70-nm contacts on 270-nm pitch.
Full-field imaging with a developmental projection optic box (POB 1) was successfully demonstrated in the alpha tool Engineering Test Stand (ETS) last year. Since then, numerous improvements, including laser power for the laser-produced plasma (LPP) source, stages, sensors, and control system have been made. The LPP has been upgraded from the 40 W LPP cluster jet source used for initial demonstration of full-field imaging to a high-power (1500 W) LPP source with a liquid Xe spray jet. Scanned lithography at various laser drive powers of >500 W has been demonstrated with virtually identical lithographic performance.
In this work simulation parameters are developed for Shipley EUV-2D photoresist under exposure at 13.4nm. Baseline parameter values are determined from theory and experiment. The simulation parameters were tuned from these values using a commercial automatic parameter optimisation software to match simulation results to experimental lithographic data generated using the ETS Set-2 projection optics in the subfield exposure station (SES). In an attempt to maximise parameter accuracy the experimental data set used included 4 different feature sizes and known non-idealities of the exposure set-up were accounted for (mask errors, lens aberrations and metrology bias). The resulting model described the experimental data very well with only a low level of residual error.
The relationships between polymer molecular weight, surface roughness measured by Atomic Force Microscopy (AFM), and EUV line edge roughness (LER), were studied in four separate rounds of experiments. In Round 1, EUV-2D (XP98248B) was prepared with seven levels of added base. These seven resists were patterned using EUV lithography; the LER was determined using 100 nm dense lines. The LER of the seven resist dramatically decreases with increasing level of base. These LER results were compared with the surface roughness of these resists after development for unexposed and DUV (248 nm) exposed surfaces. In Rounds 2-4, we evaluated three sets of EUV-2D type resists prepared with polymers having Mw of 2.9, 4.9, 6.1, 9.1, 16.1, and 33.5 Kg/mole. EUV LER and surface roughness were determined for each resist. In Round 2, the polymers were substituted into the EUV-2D resist matrix with no other formulation changes. In Round 3, the PAG level was decreased with increasing polymer Mw, to obtain a constant unexposed fill thickness loss (UFTL) for all six resists. In Round 4, both PAG level and base level were modified to yield six resists with similar sensitivity and EFTL. These experiments have led to conclusion about the impact of polymer molecular weight on imaging LER and AFM surface roughness, as well as elucidating the relationship between all three.
If EUV lithography is to be inserted at the 65-nm node of the 2001 International Technology Roadmap for Semiconductors, beta-tool resists must be ready in 2004. These resists should print 35-65 nm lines on a 130-nm pitch with LER below 4 nm 3s. For throughput considerations, the sizing dose should be below 4 mJ/cm2. The VNL and EUV LLC resist development program has measured the resolution, LER, and sizing dose of approximately 60 ESCAP photoresists with the 10X exposure tools at Sandia National Laboratories. The NA of these tools is 0.088, and every resist measured would support the beta-tool resolution requirement if the resolution scales with NA as predicted by optics. 50-nm dense lines have been printed with monopole off-axis illumination, but 35-nm resolution on a 130-nm pitch remains to be demonstrated. Only one photoresist met the LER specification, but its sizing dose of 22 mJ/cm2 is over five times too large. The power spectral density of the roughness of every resist has a Lorentzian line shape, and most of the roughness comes from frequencies within the resolution of the exposure tools. This suggests a strong contribution from mask and optics, but more work needs to be done to determine the source of the roughness. Many resists have sizing doses below the 4 mJ/cm2 target, and neither resolution nor LER degrades with decreasing sizing dose, suggesting that shot noise is not yet affecting the results. The best overall resist resolved 80-nm dense lines with 5.3 nm 3s LER on 100-nm dense lines at a sizing dose of 3.2 mJ/cm2. Thus, it comes close to, but does not quite meet, the beta-tool resist targets.
Static and scanned images of 100 nm dense features were successfully obtained with a developmental set of projection optics and a 500W drive laser laser-produced-plasma (LPP) source in the Engineering Test Stand (ETS). The ETS, configured with POB1, has been used to understand system performance and acquire lithographic learning which will be used in the development of EUV high volume manufacturing tools. The printed static images for dense features below 100 nm with the improved LPP source are comparable to those obtained with the low power LPP source, while the exposure time was decreased by more than 30x. Image quality comparisons between the static and scanned images with the improved LPP source are also presented. Lithographic evaluation of the ETS includes flare and contrast measurements. By using a resist clearing method, the flare and aerial image contrast of POB1 have been measured, and the results have been compared to analytical calculations and computer simulations.
While interferometry is routinely used for the characterization and alignment of lithographic optics, the ultimate performance metric for these optics is printing in photoresist. The comparison of lithographic imaging with that predicted from wavefront performance is also useful for verifying and improving the predictive power of wavefront metrology. To address these issues, static, small-field printing capabilities have been added to the EUV phase- shifting point diffraction interferometry implemented at the Advanced Light Source at Lawrence Berkeley National Laboratory. The combined system remains extremely flexible in that switching between interferometry and imaging modes can be accomplished in approximately two weeks.
The Engineering Test Stand (ETS) is an EUV lithography tool designed to demonstrate full-field EUV imaging and provide data required to accelerate production-tool development. Early lithographic results and progress on continuing functional upgrades are presented and discussed. In the ETS a source of 13.4 nm radiation is provided by a laser plasma source in which a Nd:YAG laser beam is focused onto a xenon- cluster target. A condenser system, comprised of multilayer-coated and grazing incidence mirrors, collects the EUV radiation and directs it onto a reflecting reticle. The resulting EUV illumination at the reticle and pupil has been measured and meets requirements for acquisition of first images. Tool setup experiments have been completed using a developmental projection system with (lambda) /14 wavefront error (WFE), while the assembly and alignment of the final projection system with (lambda) /24 WFE progresses in parallel. These experiments included identification of best focus at the central field point and characterization of imaging performance in static imaging mode. A small amount of astigmatism was observed and corrected in situ, as is routinely done in advanced optical lithographic tools. Pitch and roll corrections were made to achieve focus throughout the arc-shaped field of view. Scan parameters were identified by printing dense features with varying amounts of magnification and skew correction. Through-focus scanned imaging results, showing 100 nm isolated and dense features, will be presented. Phase 2 implementation goals for the ETS will also be discussed.
Minimizing image placement errors due to thermal distortion of the mask is a key requirement for qualifying EUV Lithography as a Next Generation Lithography (NGL). Employing Low Thermal Expansion Materials (LTEMs) for mask substrates is a viable solution for controlling mask thermal distortion and is being investigated by a wide array of researchers, tool makers, photomask suppliers, and material manufacturers. Finite element modeling has shown that an EUVL mask with a Coefficient of Thermal Expansion (CTE) of less than 20 ppb/K will meet overlay error budgets for <EQ 70 nm lithography at a throughput of 80 wafers per hour. In this paper, we describe the functional differences between today's photomask and EUVL masks; some of these differences are EUVL specific, while others are natural consequences of the shrinking critical dimension. We demonstrate that a feasible manufacturing pathway exists for Low Thermal Expansion Material (LTEM) EUVL masks by fabricating a wafer-shaped LTEM mask substrate using the same manufacturing steps as for fabricating Si wafers. The LTEM substrate was then coated with Mo/Si multilayers, patterned, and printed using the 10X Microstepper. The images were essentially indistinguishable from those images acquired from masks fabricated from high quality silicon wafers as substrates. Our observations lend further evidence that an LTEM can be used as the EUVL mask substrate material.
We report on the development of an electric capillary discharge source that radiates with comparable efficiency at both 13.5 nm and 11.4 nm, two wavelengths of interest for EUV lithography. The discharge source is comprised of a low- pressure, xenon-filled, small diameter capillary tube with electrodes attached to both ends. A high-voltage electric pulse applied across the capillary tube generates an intense plasma that radiates in the EUV. This source is capable of producing 7 mJ/steradian per pulse in a 0.3 nm bandwidth centered at 13.4 nm. In this paper we will address three significant issues related to the successful development of this source: minimization of debris generation, thermal management, and imaging quality.
The capabilities of the EUV 10x microstepper have been substantially improved over the past year. The key enhancement was the development of a new projection optics system with reduced wavefront error, reduced flare, and increased numerical aperture. These optics and concomitant developments in EUV reticles and photoresists have enabled dramatic improvements in EUV imaging, illustrated by resolution of 70 nm dense lines and spaces (L/S). CD linearity has been demonstrated for dense L/S over the range 100 nm to 80 nm, both for the imaging layer and for subsequent pattern transfer. For a +/- 10 percent CD specification, we have demonstrated a process latitude of +/- micrometers depth of focus and 10 percent dose range for dense 100 nm L/S.
The strong attenuation of extreme UV (EUV) radiation by organic materials necessities the use of a thin layer imaging (TLI) process for EUV lithography. Several TLI processes have been identified for potential use for EUVL, and the common theme in these approaches is the transfer of the aerial image to a thin layer of refractory-containing material, which is then used as a dry O2 etch mask during a subsequent pattern transfer to the device layer. One TLI process that has been extensively examined for EUVL is the silylated top-surface imaging (TSI) technology, which is discussed in this paper. Using a new disilane silylation reagent, dimethylaminodimethyldisilane (DMDS) and 13.4 nm exposure, the TSI process has been sued to print 100 nm lines and spaces at equal pitch and 70 nm lines and spaces at a higher 1:2 pitch. The line edge roughness for the printed lines has been determined using a custom image analysis program and, as expected, varies with the particular EUV exposure system and numerical aperture. Exposures done with 193 nm lithography and the TSI process using DMDS are also shown for comparison to the EUV results.
The Sandia EUV 10x microstepper system is the result of an evolutionary development process, starting with a simple 20x system, progressing through an earlier 10x system, to the current system that has full microstepper capabilities. The 10x microstepper prints 400-micrometers -diameter fields at sub- 0.10-micrometers resolution. Upgrades include the replacement of the copper wire target with a pulsed xenon jet target, construction of an improved projection optics system, the addition of a dose monitor a d an aerial image monitor, and the addition of a graphical user interface to the system operation software. This paper provides an up-to-date report on the status of the microstepper.