0.7 NA DUV STEP & SCAN SYSTEM FOR 150nm IMAGING WITH IMPROVED OVERLAY

Transcription

1 .7 NA DUV STEP & SCAN SYSTEM FOR 15nm IMAGING WITH IMPROVED OVERLAY Jan van Schoot, Frank Bornebroek, Manfred Suddendorf, Melchior Mulder, Jeroen van der Spek, Jan Stoeten and Adolph Hunter ASML BV De Run 111, 553 LA, Veldhoven, The Netherlands Peter Rümmer Carl Zeiss D Oberkochen, Germany This paper was first presented at the SPIE Symposium on Optical Microlithography March 1999, Santa Clara, California, USA

2

3 .7 NA DUV STEP & SCAN SYSTEM FOR 15nm IMAGING WITH IMPROVED OVERLAY Jan van Schoot, Frank Bornebroek, Manfred Suddendorf, Melchior Mulder, Jeroen van der Spek, Jan Stoeten and Adolph Hunter ASML BV, De Run 111, 553 LA Veldhoven, The Netherlands Peter Rümmer Carl Zeiss, D Oberkochen, Germany ABSTRACT To extend KrF lithography below the 18nm SIA design rule node in manufacturing, an advanced DUV Step & Scan system utilizing a lens with an NA up to.7 will be required to provide sufficient process latitude [1]. Towards the SIA s 15nm design rule node, manufacturing challenges for 248nm lithography include contact hole printing, iso-dense bias control and adequate across the field CD uniformity. All will benefit from higher NA lenses. In this paper, results obtained on a PAS 55/7B DUV Step & Scan system are presented. The system design is based on the PAS 55/5 [2, 3] with a new.7na Starlith lens, AERIAL II illuminator and ATHENA advanced alignment system. Imaging of dense and isolated lines at 18nm, 15nm and below as well as 18nm and 16nm contact holes is shown. In addition to imaging performance, image plane deviation, system distortion fingerprints, single-machine overlay and multiple-machine matching results are shown. Using the ATHENA alignment system, alignment reproducibility as well as overlay results on CMP wafers will be shown. It is concluded that this exposure tool is capable of delivering imaging and overlay performance required for mass production at the 15nm design rule node, with potential for R&D applications beyond. 1. INTRODUCTION The announcement of the 1997 National Technology Road Map for Semiconductors brought into sharp focus many of the challenges which IC manufacturers and equipment manufacturers together will have to overcome. Compared with the 1994 version of the road-map, the timing for first production output of several nodes had been accelerated. See Figure 1. Year: NTRS'94 NTRS'97 1/2 pitch* Isol line* Figure 1 inserted * Dimensions for minimum half pitch and isolated line in nm The SIA Roadmap for Semiconductors. In particular the First Product Shipment Technology Generation date for 18nm devices had been advanced from 21 to 1999, 13nm Technology I ILL Generation from 24 to 23, and a new node for a 15nm Technology Generation first shipment in 21 was introduced for the first time. In October of 1998 a proposal for further acceleration was made when First Product Shipment was revised to Year of First Volume Shipment, with the request for development capability to be available 2-3 years earlier [4]. The aggressive acceleration of the road-map paves the way for a rapid transition from 18nm to 15nm production. The timing of which can only be met by the extension of KrF DUV lithography beyond it s previously recognized application limit of 18nm. The new timing essentially validated the already on-going development of a new generation of high NA DUV exposure tools. To meet the technological and timing requirements of the road-map, IC manufacturers require a step and scan tool which is capable of 15nm production and available for R&D and pilot production in the first half of Optical system design and dynamic stage control critically determine the resulting imaging performance of step and scan lithography systems. High NA lenses capture more of the higher orders diffracted by the reticle, thus allowing for a more precise reconstruction 1

4 of the pattern from the reticle. This results in a higher resolution, a larger exposure latitude, a smaller iso-dense bias, a smaller Mask Error Factor [12] etc. Regarding required lens NA, simulation studies using Solid C, full resist model show that for 15nm L/S, maximum Exposure Latitude (EL) is optimal at NA>.68, see Figure 2. In Figure 3 the Iso-Dense bias, calculated using the same model, is shown. It is also clear here that for 15nm production, DUV tools with higher NA capability are required. The PAS 55/7B utilizes the PAS 55 architecture and dynamic stage control systems proven in other ASML Step & Scan tools. Commonality in architecture between the PAS 55/7B and other ASML Step & Scan tools facilitates an economical fab mix for critical, non-critical and back-end layer exposures. Wafer stage scanning speeds of up to 25mm/s are combined with on-the-fly levelling, while the ATHENA advanced alignment system, together with the existing Through-The-Lens (TTL) Phase Grating Alignment (PGA) system enables the PAS 55/7B to deliver the required overlay performance. In this paper we describe the major system enhancements of the PAS 55/7B namely the Projection Lens, AERIAL II Illuminator and ATHENA module. Finally imaging and overlay results that demonstrate the 15nm production capability of the tool are presented. 2. GENERAL SYSTEM DESCRIPTION The major new components of the PAS 55/7B Step & Scan system are a 4X, NA =.7 projection lens, AERIAL II high transmission illuminator and ATHENA advanced alignment system. The new Starlith 7 lens is a variable.7 -.5NA lens. Printing at a lower K 1 factor, even at.7na, requires a projection lens with low aberrations. Due to enhanced manufacturing methods, the lens specifications are improved compared to earlier DUV lens designs. The new AERIAL II illuminator is based on the illuminator developed for the PAS 55/9 193nm Step & Scan system [5]. Imaging performance of a scanning system is strongly determined by the synchronization between reticle and wafer stage during exposure [6]. Stage synchronization is characterized by Moving Standard Deviation (MSD) and Moving Average (MA) of the relative position of both stages over the time window in which each image point travels across the illumination slit width. MA represents a real stage error and contributes to the image distortion and overlay budget. MSD describes the loss of contrast (fading) of the aerial image that results from poor scan synchronization. Figure 4shows the results of simulations for the relation between CD uniformity and MSD for various resolutions. Simulations were performed using Solid C. The aerial image is convoluted with the Gaussian distributed positioning error. Also z-fading is taken into account at a fixed value of MSD z = 4nm. For the development a Weiss model is used with a resist contrast of 9. Exposure variations of ±4% around Best Energy are introduced in order to incorporate exposure and reticle errors, and thickness variations of the photoresist layer. The CD-range at MSD= is taken as a reference. A lens/illuminator setting of NA =.7, σ =.85/.55 is assumed. Decreasing the NA will enlarge the influence of MSD, especially on the dense lines I ILL.85.8 I-127.ILL outer Figure NA Maximum Exposure Latitude for 15nm dense lines outer Figure NA Iso-Dense Bias for 15nm dense lines. 2

5 These simulations show that low MSD is a major prerequisite for good imaging at low resolutions. For the PAS 55/7B further improvements are achieved to bring the MA back to 5nm and MSD to 15nm so together with the higher NA lens, 15nm is within the range of the system. Figure 5 shows peak MA and MSD values measured on a PAS 55/7B system. Data is collected over 7 wafers at the maximum Wafer Stage scan speed of 25 mm/s. Table 1 shows the maximum, mean and mean + 3σ values. MA-X MA-Y MSD-X MSD-Y max: mean: m + 3σ: Table 1 Maximum, Mean and Mean + 3σ formaand MSD over 7 wafers per die at maximum speed. Contribution to CDU [nm] Dense lines Isolated lines 13nm 15nm 18nm MSD [nm] Figure 4 MSD contribution to CD-uniformity. For features below 18nm, MSD values greater than 2nm have a severe impact on the CD-uniformity. I-1271.ILL The requirement for 15nm resolution as defined by the SIA roadmap is 5nm for matched machine overlay. From this a single machine overlay of 35nm can be derived. The PAS 55/7B is equipped with ASML s (TTL) Phase Grating Alignment system and the ATHENA alignment system, ASML s advanced Phase Grating Alignment system [7]. ATHENA has been developed as an addition to the current TTL alignment system to meet the current and future overlay requirements especially on processed CMP layers. ATHENA is used to align the wafer with respect to the wafer stage. The TTL system aligns the reticle to the wafer stage. ATHENA is able to collect independently the signals of 7 diffraction orders at 2 different wavelengths (633nm & 532nm). PAS 55 /7: WR peak MA-X & MA-Y I-1272.ILL PAS 55 /7: WR peak MSD-X & MSD-Y I-1279.ILL Figure 5 Peak MA and MSD values in X and Y over 7 wafers at maximum speed. 3

6 Using 2 lasers eliminates destructive interference that may occur in deep marks or on marks covered with multiple process layers. Additionally, it enables a design of an optical system that detects each diffraction order on a separate detector. The combination of high power lasers and separate detection of orders also guarantees sufficient signal strength at shallow marks. Per wafer mark up to 14 aligned positions, each associated with an individual diffraction order, can be determined. Alignment recipes can select the optimum combination of these positions yielding accurate alignment. This flexibility makes ATHENA adaptable to any wafer process. A standard set of recipes is supplied. An established application procedure enables the user to further optimize a standard recipe. For both the ATHENA and TTL systems, improved electronics and software have been developed. Data acquisition and data processing are now performed in parallel enhancing throughput by up to 8%, enabling a throughput of 14 wafers/hour (46 exposures, 16x33mm field, 3mJ/cm 2 ). The ATHENA sensor is mounted just above the wafer plane. A single frame made out of very low expansion material holds the optical elements as depicted in Figure 6. Baseline stability is maintained by referencing to a fixed target on the wafer stage, that is also measured by the TTL alignment system. The lens system images all diffraction orders on individual references on a single plate to guarantee mutual stability. The beam splitter separates both wavelengths and the subsequent special optical components direct every pair of diffraction orders towards its own reference grating. Aberrations of the alignment optics can easily be corrected for by applying software off-sets to the individual signals. The detected light is transmitted towards the detectors which have been integrated with the electronics that process the electrical signals. A CCD camera provides a monitor image for visual inspection. Detectors I-1273.ILL Modulator Wavelength 1 Wafer mark CCD Wavelength 2 Modulator Detectors Figure 6 ATHENA hardware. On the right side, both lasers have been schematically depicted. Fibers guide the laser light towards the ATHENA optical module (left hand side). Diffraction orders as reflected by the wafer target are imaged upon separate references. Subsequently, the light is transferred to a series of detectors. 4

7 The two lasers are contained in a separate box with cooling fans and optical isolators to guarantee their stability. The laser light is modulated and polarized. Lasers, CCD camera and electronics are placed at remote locations where they are not able to influence the temperature stable environment of the ATHENA optical module. Alignemnt reproducibility [StDev, nm] Figure 7 Aligned Position [nm] Figure 8 Alignment Reproducebility [StDev, nm] Figure 9 All 1st 3rd 5th 7th Diffraction Order (Red and Green) ATHENA alignment reproducibility for each of the different diffraction orders Time [min] ATHENA long term stability Relative Alignment Signal ATHENA dynamic range. 1 st order 3 rd order 5 th order 7 th order I-1274.ILL I-1275.ILL I-1276.ILL 1 The basic ATHENA performance (as measured on a PAS 55/5) can be expressed in alignment reproducibility per order, long-term stability and dynamic range, see Figures 7 to 9. Single-machine overlay and overlay performance on processed (CMP) wafers will be presented in section THE PROJECTION SYSTEM 3.1 Design goals optical column Projection lens The Starlith 7 projection lens surpasses the Starlith 5 lens design [2] by an increased NA of.7 and improved imaging quality. In addition to a reduction of the distortion, focal plane deviation and telecentricity values the lens was re-engineered with respect to wavefront variance. The design part of the wavefront aberrations budget could be reduced by more than 5% compared to the Starlith 5 lens used in the PAS 55/5 KrF step and scan system. The excellent wavefront correction has two advantages: firstly it results in a very uniform imaging behavior for a wide range of NA and σ settings for partial coherent imaging. Secondly the contrast is constant within a few percent over the full field. Both small FPD values and high contrast over the field are fundamental to 15nm imaging capability. During the design process special care has been taken in order to achieve small incidence angles on the individual lens surfaces. This eases both lens assembly and lens element coating technology resulting in an improved lens performance Illumination system The design goals for the PAS 55/7B-illuminator were threefold: Maximize the light intensity on wafer level by using an improved optical design with fewer elements and shorter path length in optical materials as well as improved anti-reflective and high-reflective coatings. For forming the pupil shape and the illumination field, diffractive optical elements have been developed which guarantee high efficiencies. A high illuminator lifetime is obtained by using selected quartz and CaF 2 -material for some lens elements and for the light mixing rod. Additionally, the whole illumination system is purged with nitrogen gas in order to prevent chemical and physical contamination. 5

8 Facilitate illumination settings over a full range of coherence and annularity as well as quadrupole settings without appreciable loss of efficiency. The quadrupole illumination modes can be realized by using specific diffractive optical elements. The big advantage of this approach is that shaping of the poles is done with an efficiency of almost 9%. The general layout of the optical path is presented in Figure 1, showing the following illumination components: The KrF-laser, line-narrowed to.6 pm and able to be operated up to 1 khz and 1 mj/pulse (1 W). The beam delivery system consists of a beam expander, bending points and an automated beam measuring and beam steering unit. The system is able to transport the laser beam over a maximum distance of 2 m. The illuminator with pupil shaping optics, to create conventional, annular and quadrupole illumination modes. The energy sensor used as dose control unit and located behind a partly transmitting mirror. The integrator rod for creating a uniform illumination field. The internal REticle MAsking (REMA) unit, which allows selection of required reticle areas. The REMA-lens, which images the created slit and the internal masking blades onto the reticle. 3.2 Manufacturing issues To develop a DUV system for 15nm imaging, lens adjustment methods were also improved. The approach at Carl Zeiss is based on the quantitative characterization of all relevant lens aberrations across the image field by a set of model functions discharge chamber line narrowing optics beam expander optics rema lens entrance group rema lens intergrator rod energy sensor incouple group filter internal rema pupil shaping optics I ILL bending point(s) projection lens beam stearing beam delivery Figure 1 Optical layout of the PAS 55/7B system. 6

9 implemented on the system qualification equipment as an integral part of the adjustment process. This allows for a systematic separation between the various orders of rotational symmetric and non-symmetric aberrations. These aberrations can then be corrected for by appropriate adjustment actions optimizing the convergence of the adjustment quality. The rotational symmetric aberrations are adjusted close to the design values by varying the spacing between individual lens elements. Further significant improvement was achieved by minimizing the non-rotational symmetric contributions with respect to the scanner slit geometry using dedicated algorithms to predict the best orientation of the optical elements. This advanced adjustment approach not only helps to ensure the production of high-quality lenses at high output rates, but also assists in decreasing Ramp-Up time for new systems like the Starlith 7 lens. Implementation of computer-aided processes throughout the whole production process has been essential in achieving the volume production of 15nm DUV systems. 3.3 First system results The Starlith 7 lens system shows improved optical properties compared to the NA =.63 Starlith 5 and 55 lenses. Figure 11 shows the results of lens qualification measurements done at Carl Zeiss (scaling: Starlith 5 lens = 1%, Starlith 7 data averaged over 6 lenses). 1 % 8 % 6 % 4 % 2 % % Zernike coefficients Wavefront RMS Integrated FPD Integrated Astigmatism Integrated Distortion Starlith 5 Starlith 55 Starlith 7 Figure 11 Three generations of DUV projection lenses. (Scale: Starlith 5 lens = 1%). I-1277.ILL Maximum Zernike coefficients (lower coma, spherical aberrations and threewave aberrations) measured by TTL interferometry over the used field are shown, as well as the maximum of the wavefront RMS. Also shown are the maximum aberrations of the aerial image for focal plane deviation, astigmatism and distortion integrated in scanning direction. The progressive reduction of the wavefront aberrations has been achieved by extensive use of TTL interferometry during the adjustment process. A large number of individual Zernike coefficients are controlled. This leads to a.7 NA lens with improvements of more than 3% compared to a typical Starlith 5 lens at NA =.63. Similarly, the lens contribution to IPD, Astigmatism and Distortion has been reduced by up to 4% compared to the Starlith 5 lens. 4. IMAGING RESULTS 4.1 Distortion and Focal Plane The optical performance of the system has been measured in terms of image distortion, focal plane and astigmatism. Both static and dynamic (scanning) results are shown in this section, representing projection lens and system performance respectively. Test wafers were exposed and measured using the TTL alignment system of the PAS 55/7B. In static mode, measurements were taken at 13 x 3 locations in the illumination slit. As the intensity distribution in the illumination slit is trapezoidal in the scanning direction, static measurements are restricted to an area in the centre of the slit, where the intensity is constant. At small sigma settings, this area of constant intensity is increased and thus allows the largest number of sample points. The full imaging field (26mm x 33mm) is used for dynamic measurements where 13 x 19 locations are sampled. The usual corrections for residual adjustable parameters (static and dynamic) have been applied [2] Image plane The FOCAL technique [9] was used to determine focal plane and astigmatism. Static and dynamic results are shown in Figures 12 to 15. A visual comparison of the lens and system fingerprints clearly reveals the similarity between static and scanning results. This 7

10 indicates that scanning induced fading is minimal and that the machine contribution to the system fingerprints is small. A summary of results can be found in Table 2. The Image Plane Deviation (IPD), defined as the total range of any focal position for horizontal and vertical lines within the field, is given. 1.8 Y [mm ] X [m m] I-1271.ILL BF [nm] Y [mm ] X [mm] I ILL BF [nm] Figure 12 Static focal plane measured in the illumination slit showing the average for horizontal and vertical lines, 39nm peak-to-valley, FPD = 78nm (NA =.7, σ =.33). Figure 14 Image plane of the PAS 55/7B measured over the full scanned image field. The Image Plane Deviation (IPD) is 98nm (NA =.66, σ =.65). 1.8 Y [m m] Figure X [mm ] AST -3 [nm] -5-7 Static astigmatism in the illumination slit. The maximum astigmatism in the measured field is 65nm (NA =.7, σ =.33). I ILL Y [mm] Figure X [m m] AST [nm] -5-1 Astigmatism of the PAS 55/7B over the full, scanned image field. The maximum astigmatism is 69nm (NA =.66, σ =.65). 5 I ILL 8

11 4.1.2 Image distortion Figure 16 shows the static distortion in the illumination slit. The maximum Non-Correctable Error (NCE) is 7nm and 12nm in X and Y direction respectively. The dynamic distortion over the whole scanned image field is shown in Figure 17. This vector plot represents the system fingerprint for distortion. The maximum NCE is 13nm in both X and Y directions. In Table 2, we have summarized Image Plane and Distortion results obtained at a number of different illumination settings. The spread over the settings used is minimal. The change in image distortion due to the use of different NA/σ settings can be an important contributor to overlay in device applications. Figure 18 shows a plot where we have superimposed the image distortion I-1278.ILL Table 2 Summary of Image Plane and Distortion results. The table also gives the largest distortion vector difference between any of the settings measured. Figure nm NCEmax X = 7 Y = 12 Static distortion in the illumination slit. The maximum Non-Correctable Error (NCE) is 12nm (NA =.7, σ =.33). NA=.66 =.65 I ILL Illumination Setting IPD Ast. NCE Max. vector difference [nm] # NA σ (σ out /σ in ) [nm] [nm] [nm] #5 #4 #3 # / / I ILL nm NCEmax X=13 Y= Figure 17 Dynamic (system) distortion of the PAS 55/7B over the full, scanned image field, for a conventional illumination setting (NA =.66, σ =.65). The maximum NCE is 13nm nm Figure 18 Superimposition of vector plots of the image distortion for five different conventional and annular illumination settings (see also Table 2). 9

12 vector plots for the illumination settings as listed in Table 2. The maximum non correctable distortion difference between any of the settings measured is only 8nm Resolution The PAS 55/7B imaging performance with 1:1 lines and spaces has been evaluated using photoresist C and Clariant AZ-DX331P photoresist. For the evaluation of contact holes TOK DP15 photo resist with a thickness of.68µm has been used. Process conditions are summarized in Table 3. The photoresist was processed using FSI Polaris wafer tracks. The system is purged with activated charcoal filtered air to prevent adverse process effects from airborne base contamination. Although the process tracks and the exposure tools are not interfaced, the time between coating, exposing and developing was kept below 2 minutes. Measured CD [nm] +1% 2-1% nominal CD [nm] I ILL SEM analysis was done with AMAT 783SI and Hitachi S-8C4 CD-SEM s for top down automated CD measurements and a Hitachi S-78H SEM for tilted inspection and photographs of the resist profiles. Figure 19 Linearity on photoresist C on Organic BARC for dense (1:1) lines, NA =.7, σ =.85/.55. Description Conditions Photoresist DX-331P TOK DP15 photoresist C Thickness.4/.5µm.68µm.56/.4µm AR2 or BARC SiON Softbake temperature Softbake time PEB temperature Using photoresist C on Organic BARC, the linearity was measured in the center of the scanned field with the best focus and best energy selected. Results for both dense (1:1) and isolated lines are shown. The NA was.7 and the annular partial coherence was.7/.4. Figures 19 and 2 show linear behavior down to 13nm with the used resist process. - Organic BARC 9 C 8 C 1 C 6s 9s 9s 11 C 11 C 11 C PEB time 9s 9s 9s Developer OPD-262 OPD-262 OPD-262 Develop Time 6s 6s 6s Table 3 Photoresist process conditions. Measured CD [nm] 3 2 Figure 2 +1% -1% nominal CD [nm] Linearity on photoresist C on Organic BARC for isolated lines, NA =.62, σ =.85/.55. Figures 21 and 22 show top-down SEM photographs of lines/spaces and Depth of Focus (DoF) with resolutions ranging from 15nm to 11nm using binary masks without OPC in order to demonstrate the pure I ILL 1

13 tool capability. Also 15nm isolated lines are shown. The illumination mode was annular for all cases. At the selected apertures and ring widths, the depth of focus is.8µm for 15nm dense lines, reducing to.5µm for 13nm dense lines. For 15nm isolated lines, the observed DoF is.5µm. At the bottom of Figure 22, it is shown that even a 11nm dense line pattern is resolved. For the used lens and illuminator setting, less than 1% of the light in the first orders is captured by the lens, proving the nearly perfect image transfer by the optical and mechanical systems. In Figure 23, SEM pictures of cross sections of dense lines are shown, NA =.66, σ =.75/.45 for the 15nm exposure. All other pictures are obtained using NA =.7, σ =.85/.55. Figure 24 shows top-down SEM pictures of 18nm and 16nm dense contact holes. The setting used was NA =.7, σ =.85/.55, the observed DoF is.8µm for 18nm contacts, and.5µm for 16nm contacts. Note the roundness of the contact holes which illustrates good dynamic stage control during scanning. All pictures shown in Figures are made with binary reticles without OPC, in order to show the pure tool behavior. CD-uniformity is an important measure for characterizing step and scan systems. The CD-uniformity of 18nm and 15nm lines and spaces are measured, and the results are summarized in Table 4. In order to reduce SEM and process contributions, six fields across the wafer are averaged at each field position. 15 nm Dense Lines, NA =.66, σ=.75/ BF nm Isolated Lines, NA =.62, σ=.85/ BF Figure 21 Top-down SEM photographs of 15nm dense and isolated lines. 11

14 14 nm Dense Lines, NA =.7, σ=.85/ BF nm Dense Lines, NA =.7, σ=.85/ BF nm Dense Lines, NA =.7, σ=.85/ BF nm Dense Lines, NA =.7, σ=.85/.55 BF Figure 22 Top-down SEM photographs of dense lines ranging from 14nm down to 11nm. 12

15 15nm 14nm 13nm 12nm Figure 23 Cross section SEM photographs of dense lines ranging from 15nm to 12nm. Exposed using Clariant AZ DX-331P on a polysilicon - SiON film stack. 18nm Contacts, NA=.7, σ=.85/.55 BF -.4 BF -.3 BF BF +.3 BF nm Contacts, NA=.7, σ=.85/.55 BF -.3 BF -.2 BF BF +.1 BF +.2 Figure 24 Top-down SEM photographs of 18nm and 16nm dense contact holes. 12nm Dense Lines, NA=.7, σ=.315 Figure 25 BF -.3 BF BF nm Dense Line pattern using Levenson type PSM on Clariant, on top of a SiON BARC/poly Si. 13

16 Feature 18nm DL 18nm iso 15nm DL 15nm DL* 15nm iso Setting NA =.6, σ =.7/.4 NA =.56, σ =.6/.3 NA =.66, σ =.75/.45 NA =.7, σ =.85/.55 NA =.62, σ =.85/.55 CD-uniformity ±.1 µm ±.2 µm ±.3 µm *Quadrupole illumination has been used Table 4 CD-uniformity 3σ values for 18nm and 15nm dense and isolated features at best focus and over a focus range. Especially in the case of Dense Lines the reticle contribution dominates the total measured CD variation at best focus. The reticles used for these experiments are specially designed to minimize the reticle CD variation contribution using the Picked CD technique [1]. For the 15nm dense lines, the reticle CD distribution has a 3σ of 15nm. The Mask Error Factor (MEF) [11, 12] has been obtained experimentally and equals 3.2. As a result, the expected 3σ at wafer level is 12nm (15/4 *3.2). From this it can be concluded that the reticle is responsible for approximately 8% of the entire CD uniformity. 5.2 Single Machine Overlay Figure 26 shows the single machine overlay performance at full scan speed (25mm/s). On 8 wafers, a layer of 2 full 26mm x 33mm fields was exposed. After reloading the wafers in the machine and aligning them, an identical second layer was exposed. The maximum measured 99.7% single machine overlay error between both layers was 14 and 18nm for X and Y, respectively. The bell -shape of the Gaussian curve for the results in both the X and the Y direction is well preserved. Count Overlay [nm] Figure 26 Single machine overlay at full scan speed: X = 14nm, Y = 18nm (99.7% of the data, 8 wafers, 2 fields/wafer, 49 points/field). In Figure 27, a detailed picture is shown of the overlay errors of the first wafer. The errors are scaled with respect to the single machine overlay specification of 4nm (lower right corner). 1 X Y I ILL I ILL Finally, in Figure 25 we show 12nm dense lines with a focus range of.6µm. These were obtained using an attenuated PSM. The asymmetry of the spaces is a well known effect using PSM and is due to reticle effects [13]. 5. OVERLAY PERFORMANCE 5.1 Introduction Nominal Position Y [mm] 5-5 In this section, the single machine overlay performance of the PAS 55/7B and the matched machine overlay performance of the PAS 55/7B to the PAS 55/5 and to the PAS 55/4 (I-line scanner) are presented. Additionally, results are presented using the ATHENA system. 4 nm Nominal Position X [mm] Figure 27 Wafer plot of the first of 8 overlay wafers, showing 49 overlay vectors per field 14

17 5.3 Matched Machine Overlay The matched machine overlay performance to the PAS 55/4 I-line scanner has been investigated. To get the PAS 55/7B in a matched state to the PAS 55/4 DUV Step & Scan tool, the lens errors of the PAS 55/7B system were minimized with respect to the PAS 55/4 lens using reference wafers exposed on the PAS 55/4. After putting the system in the matched state, a verification was performed using 3 wafers. For each wafer, the first layer was exposed on the PAS 55/4 and the second layer was exposed on the PAS 55/7B, using a wafer layout consisting of 12 fields. The maximum measured 99.7% matched overlay error between both layers was 49nm in X and 58nm in Y. In Figure 28, the overlay performance measured at 25 points per field is shown. The overlay errors for this experiment were 38nm in X and 44nm in Y. Count X Y I-1273.ILL Count X Y Overlay [nm] Figure 29 ATHENA single machine overlay (max vectors X = 12mm, Y = 11mm). Experiments [7] have proven that high spatial frequencies of the ASML alignment target are less affected by CMP than the low ones. High spatial frequencies are directly associated with high diffraction orders. To improve the accuracy on marks that have been polished asymmetrically due to CMP, ATHENA uses (a combination of) high orders. Figure 3 shows a wafer plot of a W-CMP polished wafer illustrating the alignment vectors compared to cleared marks (which are not affected by the CMP process) used as a reference. The improvements when using higher order diffraction compared to first order are clear. I ILL Overlay [nm] Figure 28 Matched machine overlay of the PAS 55/7B to a PAS 55/4 system: maximum error in X = 38nm, in Y = 44nm (for 99.7% of the data, using 3 wafers, 12 fields/wafer, 25 points/field). 1 ST ORDER ALIGNMENT RAW DATA 1 ST ORDER ALIGNMENT WITH PROCESS CORRECTIONS 7 TH ORDER ALIGNMENT RAW DATA I ILL 5.4 ATHENA The ATHENA alignment system as described in section 2 has been designed to increase the alignment process latitude. First, single machine overlay has been measured using ATHENA, see Figure 29. Figure 3 1 nm 1 nm 1 nm Wafer plots of a W-CMP wafer using 1st and 7th order alignment. ATHENA overlay on W-CMP wafer has been verified on a test lot of 8 wafers. The first layer is exposed on a PAS 55/2 i-line stepper. The 2nd layer is exposed on a scanner after aligning with the 7th order of the red 15

18 laser. KLA511 measurements are presented in Figure 31. After subtraction of the correctable terms, residuals remained in the X and Y direction of 51nm and 44nm, respectively. More details can be found in [7]. Count X Y Overlay [nm] Figure 31 ATHENA overlay on CMP Residuals of 8 wafers, 25 points per wafer, max X: 51nm, max Y: 44nm. 6. CONCLUSIONS In this paper we have described the requirements for achieving 15nm lithography with KrF step and scan machines. These can be summarized as: a variable high NA lens (NA =.7) excellent stage synchronization (MSD < 15nm) overlay capability on advanced layers (e.g. W-CMP). The PAS 55/7B has shown.8µm DoF for 15 nm dense lines,.8µm DoF for 18 nm contact holes and.5µm DoF for 16 nm contact holes. The CD uniformity is 14nm 3σ for 15 nm dense lines. An overlay of 51nm has been obtained on W-CMP processed layers. The results at resolutions from 13nm down to 11 nm indicate that the described system has good potential for extending applications to resolutions of 13nm and below. ACKNOWLEDGEMENTS The authors would like to thank Tammo Uitterdijk, Louis Jorritsma, Dennis Faas, Henry Megens, Geert Simons, Walter Schuller, Paul Hinnen, Frank Commissaris, Wilfred Kauffeld and Wil Pijnenburg for data gathering and analysis, Koen van Ingen Schenau and Stephan Sinkwitz for process setup and Yin Fong Choi, Jenny Swinkels, Ingrid Janssen and Paul Luehrmann as well as Patrick Jaenen from IMEC, Belgium for SEM support. I ILL Many thanks also to The Zeiss Project Team as well as Philips Research, for their work on the ATHENA alignment system and the alternating Phase Shift Mask. We thank all members of the PAS 55 Project at ASML. Finally, we would like to thank Bob Simpson and Frank Harmsen for their help with the publication of this manuscript. REFERENCES [1] W. Maurer et al., Pattern transfer at k 1 =.5: get.25um lithography ready for manufacturing, Proc. SPIE, Vol. 2726, 1996, pp [2] G. de Zwart et al., Performance of a Step & Scan system for DUV lithography, Proc. SPIE, Vol. 351, 1997, pp [3] J. van Schoot et al., Advanced imaging and overlay performance of a DUV Step & Scan system, Proc. Semicon Korea, [4] National Technology Roadmap for Semiconductors: Technology Needs, Rev 1/7/98. [5] J. Mulkens et al., ArF step and scan exposure system for.15 µm and.13 µm production nodes, proc. SPIE, Vol. 3679, [6] A. Erdmann et al., Lithographic process simulation for scanners, Proc. SPIE 3334, 1998, pp [7] J.H.M. Neijzen et al., Improved overlay performance using an enhanced phase grating adjustment system, proc. SPIE, Vol. 3677, [8] G.Davies, J.Stoeldraijer, H.Glatzel et al., 193nm Step and Scan Lithography, SEMI Technology Symposium 98, Chiba, Japan. [9] P. Dirksen, Latent image metrology for production of wafer steppers, proc. SPIE, Vol. 244, 1995, pp [1] J. Waelpoel, J. van Schoot and A. Zanzal, Demonstrating next generation CD uniformity with today s tools and processes, proc. SPIE, Vol. 3236, [11] A. Wong et al., Lithographic effects of mask Critical Dimension error, proc. SPIE, Vol. 3334, 1998, pp [12] J. van Schoot, J. Finders, K. van Ingen Schenau, M. Klaassen and C. Buijk, The Mask Error Factor: Causes and Implications for Process Latitude, proc. SPIE, Vol. 3679, [13] A.K. Wong and A.R. Neureuther, Mask topography effects in projection printing of Phase Shifting Masks, IEEE trans. on Electron Devices, Vol 6, 1994, pp

19

20