We have proposed a new inspection method of in-line focus and dose control for high-volume manufacturing of
semiconductor. And we have referred to this method as "Focus and Dose Line Navigator (FDLN)". This method can
raise a performance of semiconductor exposure tool and therefore they can go up a yield ratio of semiconductor
device. The method leads the exposure condition (focus and dose) to the center of process window. FDLN calculates
correct exposure condition using the technology of solving the inverse problem. The sequence involves following
process. 1) Creating a focus exposure matrix (FEM) on a test wafer for building some models as supervised data.
The models mean the relational equation between the multi measurement results of resist patterns (e.g. Critical
dimension (CD), height and sidewall angle) and exposure conditions of FEM. 2) Measuring the resist patterns on
production wafers and feeding the measurement data into the library to predict focus and dose. In this time, we have
evaluated the accuracy of FDLN. We made some sample wafers by Canon's exposure tool "FPA-7000AS7". And
we used Veeco's advanced CD-AFM "InSight" as a topography measurement tool.
In addition to hardware performance enhancement of exposure tool, new functions are needed to be developed to meet
the required performance for realizing double patterning. New functions to improve overlay accuracy are advanced
distortion control and stage control. We have developed a real-time lens magnification control system to enhance
distortion control, which can make peel type, barrel type and trapezoid type of distortion shape, resulting in improving
intra-shot overlay accuracy. Wafer stage grid control function can compensate for shot shift, shot rotation and
magnification for each single shot, realizing drastic advancement in overlay accuracy. As for CD performance
improvement, dose optimization is effective to compenste for CD uniformity according to CD metrology data from
processed wafers. On the other hand, process window enhancement is performed by optimizing illumination mode with
Canon's solution software k1 TUNE. In this paper, we will introduce these new functions.
Canon has developed an immersion exposure tool, the FPA-7000AS7 (AS7), with the industry's highest NA of 1.35.
This paper reports on its performance. The AS7's projection lens achieves ultra-low aberration with total RMS of less
than 5 mλ and flare of less than 0.5%. The resolution capability is 37 nm with k1 = 0.259, and DOF of 0.8 μm was
obtained owing to the ultra-low aberration and low flare. Regarding focus performance, a 3σ value of 19.3 nm for Lstage
and 16.1nm for R-stage were attained in a whole area. The result of CD uniformity of 1.91nm (3σ) was obtained
across the wafer with a total of 4032 measurement points. Distortion was 3.9 nm at the worst value. On the other hand
the most critical issue of immersion is defects, so the nozzle, lens and stage must be cleaned to reduce defects. The result
of defect evaluation of the AS7 was an average of 0.042 defect/cm<sup>2</sup> from 25 wafers in a lot and average 0.046 defect
count/cm<sup>2</sup> from long-term defect evaluation for two months. From these results, we are confident that the AS7 is capable
of 45-nm node device production.
Canon has renewed its platform of exposure tools. The new platform, the FPA-7000, is designed to cover multiple
generations. The lens performance of the FPA-7000AS5 achieves less than 6m&lgr;, while that of the AS7 is estimated to
be less than 4m&lgr;. The illumination performance meets the target required for the 45nm node. The in-situ aberration
monitor, called iPMI, attains the measurement repeatability of 1.45m&lgr;. Focus and overlay units have improved process
robustness. A solution tool for optimization is introduced to be connected with the FPA-7000. Moreover, latest studies
of immersion, such as nozzle pressure, temperature control and defect inspection result are reported, and we also
discuss the possibility of high-refractive-index immersion.
High-NA and immersion projection systems require RETs (Resolution Enhancement Techniques) that utilize polarized illumination. Therefore measuring aberrations that are dependent on illumination polarization (polarization aberration) also becomes important. Generally, metrology for polarization aberration measurement consists of polarizer, resulting in a large-scale apparatus and rising cost. Therefore, a simple and accurate metrology method is desired, one that can be easily installed then removed after testing. We have investigated a simple and accurate metrology method for polarization aberration measurement using Canon SPIN. Through this work, we developed a new theory, entitled BLP (Birefringence measurement by Linear Polarization of light), to characterize birefringence of the lens by rotating linear polarization illumination. One of the merits of BLP is its applicability to most of the conventional metrologies for lens aberration measurement. In this paper, we have used the SPIN method for BLP evaluation. We confirmed the accuracy of BLP by achieving 1.0 correlation coefficient with Jones theory for Retardance and Fast-Axis of birefringence. We also evaluated the validity of Pseudo-Jones-Pupil (PJP), which was generated from SPIN-BLP analysis, for imaging performance simulation. This resulted in identical imaging performance with the original Jones pupil for resolution and LRCD. As a polarization aberration monitor, SPIN can be used for qualification, periodic monitoring and evaluation of image performance in the field. Another advantage of SPIN is its portability. Therefore we also consider usage of SPIN as a machine-to-machine calibration tool.
Achieving accurate low k1 imaging performance requires that the illumination intensity profile (effective light source profile) no longer be neglected. Simultaneously, simulation techniques have taken on an unprecedented level of importance because it is not practical for all low-k1 imaging applications to be performed experimentally. The impetus is now on the simulation to efficiently narrow down the numerous those options. Moreover, we are concerned that current metrology methods, such as the SEM, will be no longer be used with full confidence in terms of data reliability and accuracy because the specification may reach its measurement limit and the sample reproducibility may dominate the CD budget. We therefore anticipate that a simulation, which incorporates all factors potentially impacting performance, will predict experimental results accurately and repeatedly. We have been newly developing a reticle-based metrology tool, entitled REMT (Reticle Effective light source Measurement Tool), to precisely quantify the illumination shape. The illumination light, which first passes through a pinhole and traverses an optical path within REMT, is then detected by a CCD camera located over the reticle stage to form the illumination intensity profile. The measurement reproducibility of the σ size for REMT is less than ±0.0002. We have developed a lens metrology tool, entitled SPIN (Slant projection through the PIN-hole), to accurately quantify lens aberrations. SPIN is also a reticle-based metrology tool, with repeatability of less than 1mλ. In this paper, we will investigate Left-Right CD Difference (LR-CD), the well-known detection method for coma aberration, comparing experimental results with those from simulations that consider both lens aberrations and illumination shape as measured by SPIN and REMT, respectively. In this discussion, the factors causing LR-CD for dipole illumination will be also analyzed and quantified.
The images projected by the first vortex via masks, which comprised arrays of closely spaced dark spots that could pattern contacts with critical dimensions smaller than a third of an exposure wavelength in negative photoresist, showed several unexpected anomalies. Under certain conditions, the contact holes were elliptical (rather than round), displaced from their ideal locations, and had major axes oriented in directions that broke the expected symmetries. These effects have now been explained in terms of errors in the mask transmission and phase that give rise to unwanted Fourier components of the image combined with aberrations in the projection lens. Both effects must be present to break all the pattern symmetries. Distortions can be controlled by setting the numerical aperture of the projection lens to filter out four of the nine Fourier components and by proper design of the reticle.
Vortex masks composed of rectangles with nominal phases of 0°, 90°, 180° and 270° have been shown to print sub-100nm vias and via arrays when projected into negative resist using 248nm light. Arrays with pitches down to 210nm and CDs as small as 64nm have been reported. While promising, 248nm vortex via images showed some anomalies: The developed contacts were somewhat elliptical, with four different repeating via shapes. The common depth of focus for these four classes of via was limited by their different behaviors through focus. Phase edges in isolated vortex pair structures tended to print, also limiting the useful DOF. These issues can be ameliorated by employing 193nm illumination and a new negative-tone resist. Smaller NAs and higher coherence extend the common depth of focus and larger NAs can be used to print even more tightly spaced patterns. Advanced optical proximity correction techniques can also be applied to reduce the via ellipticity and placement error, and a more optimal choice of geometrical phase depth reduces pattern variability. Further developments and incremental improvements in vortex via design and processing may make it the method of choice for via patterning at the 45nm node.
In an optical vortex, the wavefront spirals like a corkscrew, rather than forming planes or spheres. Since any nonzero optical amplitude must have a well-defined phase, the axis of a vortex is always dark. Printed in negative resist at 248 nm and NA0.63, optical vortices and optical vortex arrays produce contact holes with 64 nm<CD<160 nm (0.2<k1<0.4), depending on exposure dose. Arrays of vortices with kpitch>0.6 can be patterned using a chromeless phase-edge mask composed of rectangles with nominal phases of 0, 90, 180, and 270 deg. Lithography simulation and resist exposures have demonstrated process windows with 10%Elat and ~400-nm depth of focus (DOF) for 85-nm CDs at 210-nm pitch with =0.15, but the developed contacts are somewhat elliptical. No significant surface development has appeared due to phase-edge printing. However, the spacewidth alternation phenomenon familiar from linear chromeless phase-edge lithography does cause small positional errors for vortex vias, and each of the four vortices in the repeating pattern may behave somewhat differently through focus, potentially limiting the common process window. Smaller CDs and pitches are possible with shorter wavelength and larger NA, while larger pitches give rise to larger CDs. At pitch >0.6 µm, the vortices begin to print independently for 0.3. Such "independent" vortices have a quasi-isofocal dose that gives rise to 110-nm contacts with Elat>14% and DOF >400 nm. In an actual chip design, unwanted vortices and phase step images would be erased from the resist pattern by exposing the wafer with a second, more conventional, bright-field trim mask. Compared to other ways of producing deep subwavelength contacts, the vortex via process reduces the lithography and process control challenges.
The extendibility of optical lithography using KrF and ArF exposure tools is still being investigated, even, being demanded strongly now, due to the unforeseen issues, high cost, and general difficulty of NGLs - including F<sub>2</sub> and immersion lithography. In spite of these challenges Moore's Law requires continued shrinks and the ITRS roadmap still keeps its aggressive timetable. In order to follow the ITRS roadmap, the resolution must keep improving by increasing the lens NA for optical exposure tools. However, the conventional limit of optical resolution (kpitch=0.5) is very close for the current technologies, perhaps limiting progress unless NGL becomes available quickly. Therefore we need to find a way to overcome this seemingly fundamental limit of optical resolution. In this paper, we propose two practical two-mask /double-exposure schemes for doubling resolution in future lithography. One method uses a Si-containing bi-layer resist, and the other method uses Applied Materials' APF (a removable hard mask). The basic ideas of both methods are similar: The first exposure forms 1:3 ratio L/S patterns in one resist/hard mask layer, then the second exposure images another 1:3 ratio L/S pattern in-between the two lines (or two spaces) formed by the first exposure. The combination of these two exposures can form, in theory, kpitch=0.25 patterns. In this paper, we will demonstrate 70nm L/S pattern (140nm pitch) or smaller by using a NA0.68 KrF Scanner and a strong-RET reticle, which corresponds to kpitch = 0.38 (k1=0.19). We will also investigate the critical alignment and CD control issues for these two-mask/dual-exposure schemes.
The first vortex masks composed of rectangles with phases of 0°, 90°, 180° and 270° - as proposed at Photomask 2002 - have been fabricated and shown to print sub-100nm contacts. The walls of the phase trenches are very nearly vertical, with all four phase regions meeting at sharp corners which define the phase singularities. Arrays with pitches down to 210nm have been printed in negative DUV resist using KrF illumination with NA=0.73 and sigma=0.15. The developed contacts are somewhat elliptical, but their shapes can be corrected (if necessary) by OPC techniques. The depth of focus for +/-10% CD variation is >400nm for 85nm CD vias at 210nm pitch and >700nm for 100nm vias at 250nm pitch. The exposure latitude is ~15% at best focus. At constant exposure dose, the via CDs vary with pitch as predicted by simulations. Increasing exposure dose makes the openings smaller, more uniform and more circular. No significant surface development has appeared due to phase-edge printing. However, the spacewidth alternation phenomenon familiar from linear chromeless phase-edge lithography does cause small positional errors for vortex vias, and each of the four vortices in the repeating pattern behaves somewhat differently through focus, potentially limiting the common process window.
In an optical vortex, the wavefront spirals like a corkscrew, rather than forming planes or spheres. Since any nonzero optical amplitude must have a well-defined phase, the axis of a vortex is always dark. Printed in negative resist at 248nm and NA=0.63, 250nm pitch vortex arrays would produce contact holes with 80nm<CD<160nm(0.2<<i>k</i><sub>1</sub><0.4), depending on exposure dose. Arrays of vortices with <i>k</i><sub>pitch</sub>>0.6 can be patterned using a chromeless phase-edge mask composed of rectangles with nominal phases of 0°, 90°, 180° and 270°. Analytic and numerical calculations have been performed to characterize the aerial images projected from such vortex masks using the Kirchhoff-approximation and rigorous EMF methods. Combined with resist simulations, these analyses predict process windows with ≈10%<i>E</i><sub>lat</sub> and >200nm DOF for 80nm CDs on pitches greater than or equal to 250nm at σ greater than or equal to 0.15. Smaller CDs and pitches are possible with shorter wavelength and larger NA while larger pitches give rise to larger CDs. At pitch >0.8μm, the vortices begin to print independently for σ greater than or equal to 0.3. Such “independent” vortices have a quasi-isofocal dose that gives rise to 100nm contacts with <i>E</i><sub>lat</sub>>9% and DOF>500nm at σ=0.3. The extra darkness of the nominal 270° phase step can be accommodated by fine-tuning the etch depth. A reticle fabrication process that achieves the required alignment and vertical wall profiles has been exercised and test masks analyzed. In an actual chip design, unwanted vortices and phase step images would be erased from the resist pattern by exposing the wafer with a second, more conventional trim mask. Vortex via placement is consistent with the coarse-gridded grating design paradigms which would - if widely exercised - lower the cost of the required reticles. Compared to other ways of producing deep sub-wavelength contacts, the vortex via process requires fewer masks and reduces the overlay and process control challenges. A high resolution negative-working resist process is essential, however.
The industry demand for an extension of optical lithography using KrF and ArF exposure tools remains strong, concerning process maturity of NGL and the higher capital cost for implementing new tools. Many solutions exist for printing to print fine lines and spaces with k1=0.3 or smaller, these include the use of alt-PSMs, dipole illumination and other RETs. Moreover, the application of these RETs using NA0.85 ArF scanner enables feature shrinkage down to the 65nm Node. However, contact/via holes are the most difficult features to successfully print according to the sizes dictated by ITRS road map. One of the primary reasons for the inability to maintain the same shrinkage pace is the resolution limitation due to two-dimensional diffracted light. Another factor is that, with the exception of negative-tone resist, the complicated strong PSM (alt-PSM) applications have some difficulties to deploy. In 2002, Canon demonstrated a new method, entitled IDEALSmile, which simultaneously resolves 100nm dense and isolated contact holes while providing a robust process window. The advantages of the IDEALSmile technique are the high-resolution capability and the large process window with a conventional method such as single exposure using binary mask. In order to apply the IDEALSmile technique on actual device patterns, it is necessary to evaluate its through-pitch performance in terms of its process window, MEEF and etc. Using the NA0.73 KrF scanner, greater than a 0.3um common process window was achieved for 120nm holes, ranging from 1:1 dense features through 1um pitch isolated features. Moreover, 0.1.um dense holes were resolved with sufficient process window under the same exposure conditions. These results lead us to conclude that, using the IDEALSmile technique, we can achieve a sufficiently large through-pitch process window for the 65nm node using a state-of-the-art NA0.85 ArF scanner.
A PECVD deposited carbon hardmask is combined with dielectric anti-reflective coating (DARC) for the patterning of sub-90nm lines with 248nm lithography. Using this CVD dual layer stack, <1% reflectivity control is demonstrated for both 248nm and 193nm lithography. The film stack is tested with an etch integration scheme to reduce polysilicon gate critical dimension (CD). The dual layer stack can be defined with less than 100nm thick photoresist. Because of the minimal resist required to open the stack, this film stack enables an integration scheme that extends conventional photoresist trim processes up to 70% of the starting line width. In addition to conventional trim process, a resistless carbon mask trim process is investigated to further shrink the gate critical dimension. The results show that the carbon hardmask has greater than 6:1 etch selectivity to polysilicon, enabling the extension of the resist trimming technique to generate sub-30nm structures using 248nm lithography.
An optical vortex has a phase that spirals like a corkscrew. Since any nonzero optical amplitude must have a well-defined phase, the axis of a vortex (where the phase is undefined) is always dark. Printed in negative resist, lowest order vortices would produce contact holes with 0.2<k<sub>1</sub><0.5, roughly 80-200nm diameter, with 248nm exposure and NA=0.63. Arrays of vortices with k<sub>pitch</sub>>0.6 can be produced using a chromeless phase-edge mask composed of rectangles with phases of 0°, 90°, 180° and 270°. EMF and Kirchhoff-approximation simulations reveal that the image quality of the dark spots is excellent, and predict a process window with 15% exposure latitude and 400nm DOF for 80nm diameter spots on pitches ≥250nm at σ=0.15. EMF simulations predict that the 0-270° phase step will not be excessively dark if the quartz wall is vertical. Chrome spots at the centers can control the diameters which otherwise are set by the parameters of the imaging system and exposure dose. Unwanted vortices can be erased from the image by exposing with a second, more conventional, trim mask. This method would be superior to the other ways of producing sub-wavelength vias, but successful implementation requires the development of appropriate negative-tone resist processes.
The massive amount of data necessary to qualify a new 100nm generation technology can be efficiency collected using a CD-SEM and analyzed using Klarity ProData. By comparing the linewidths, space widths, and pitches printed in resists with different focus, exposure does, and numerical aperture with the measured reticle parameters, one can be determine optimal processing conditions and the required biasing rules for the new technology. The Sidewall Chrome Alternating Aperture Mask, a next generation alternating phase shift mask structure, is especially suitable for this as all relevant mask features are visible from the top surface which, however, is not planar and thus can confuse optical mask inspection tools. Resist patterns with line-space pitches from 220nm to 800nm and isolated lines - as well as the reticle - were measured sing a KLA-Tencor 8250 CD-SEM and analyzed with ProData. At the isofocal dose, the 70nm line - 150nm space reticle pattern printed with equal 110 nm lines and spaces at NA equals 0.63 on a Canon FPA-5000 ES3 248 nm step and scan tool, with a process window that overlapped those of less dense approximately 100 nm features.
Multiple exposure alternating phase-shifting mask lithography can exploit economies of scale to lower total manufacturing cost if chip designers adhere to simple rules, some of which also would apply to dipole illumination and other sub-wavelength lithography systems. The most promising system employs the Sidewall Chrome Alternating Aperture Mask structure which fulfills the theoretical expectations of strong phase-shift optics and offers the potential for low cost sub-100nm imaging with 248nm light. It has already printed 73nm semi-dense lines with 248nm light at k1 equals 0.19 and demonstrated a useful common processing window for less dense and isolated structures.
This paper presents the results of a joint development effort between Canon USA, Inc. and Photronics, Inc. on 150nm contact hole application. A double exposure technique, Canon's IDEAL technique, is used to achieve the very small dense contact hole and isolated contact hole simultaneously. Canon's IDEAL exposure technique has shown, through numerous documented investigations to be beneficial for extending the current lithography tool life with regards to line patterns. However, it is also now equally important to evaluate IDEAL's advantages for contact holes. We look to apply the IDEAL technique to contact holes by using a Hole-shaped alternating Phase Shift Mask for the grid and a binary mask for trimming. This experiment was performed on a Canon FPA- 3000EX6 5X stepper with maximum NA0.65, using JSR TMX1260Y 300nm thick resist. All masks were made by Photronics. Since image intensity imbalances of Hole shaped alt-PSMs were too large to generate a perfect grid, we exposed twice with the same Hole-shaped alt-PSM reticle. The second exposure was shifted to combine 0 degree and 180-degree space, thereby creating a well-balanced grid. Subsequently, we used a binary mask for trimming. Through this method, 0.15 micrometers dense holes and 0.15 micrometers isolated holes with simple reticle bias were resolved simultaneously, and over 0.6 micrometers common DOF was obtained. Due to the high accuracy alignment between the PSM hole mask and binary mask from this experiment, double and triple exposure schemes can be used in actual production. Based on these experimental result, we also confirmed that the IDEAL technique allows fora 50nm combination error of stage stepping and reticle alignment without including significant CD error. A well- balanced grid can be generated using the vertical line PSM and horizontal line PSM, by minimizing image intensity imbalances due to PSM structures, however, the three-reticle application may prove prohibitive due to the increase in reticle cost.
The Sidewall Chrome Alternating Aperture (SCAA) mask has now successftilly printed resist images with k1 factors as low as 0.20, without significant focus-dependent spacewidth alternation or other anomalies that affect common alternating-PSM structures. The SCAA mask process (reported at BACUS 2000') etches the phase topography first and then forms the transparent openings that define the image in a conformal chrome layer deposited afterwards. This process minimizes the differences between the 3-dimensional environments ofthe phase shifted and unshifted mask features. With all chrome supported and all quartz walls covered, only the size of a chrome aperture determines its transmission and only the height difference ofthe quartz surface affects the phase shift. SCAA masks are more stable mechanically than alt-PSM structures in which the quartz walls are undercut beneath the chrome edges to minimize the french walls effects. The chrome covering the phase edges also buries entire classes ofunrepairable phase defects. Initial experiments on Canon ES2 and ES3 exposure tools confirm that KrF SCAA masks project acceptable images of isolated line and line-space patterns down to lOOnm in 280mn thick JSR M1O8Y resist. The process windows, however, were limited by resist collapse, and there were strong optical proximity effects. Comparing the resist results to SEM scans of the masks confirmed the insensitivity ofthe image to overlay errors, so long as the phase steps were covered by chrome. The Phase Phirst paradigm exploits the SCAA mask structure to enable low cost strong PSMs. The key is to mass produce SCAA mask substrates with generally useful phase topographies using wafer fab techniques. These Phase Phirst substrates would arrive at the mask houses ready to write and guaranteed to be free ofphase-defects. At design houses, Phase Phirstenabled EDA tools would lay out the chips in such a way that all fme dark features lie at the predetermined phase shift locations on the Phase Phirst substrate while the larger-dimension structures appear on a more conventional trim-mask. Once the GDS-II tapes arrive at the mask house, the chip design would be matched to the specified Phase Phirst substrate and printed in the chrome. Cost and turn around time should be similar to those of a COG mask pair. The wafer yield and resolution, however, would be enhanced by the well-known advantages of strong phase-shifting.