Thin mask model has been conventionally used in optical lithography simulation. In this model the diffracted waves from the mask are assumed to be Fourier transform of the mask pattern. In EUV (Extreme UltraViolet) lithography thin mask model is not valid because the absorber thickness is comparable to the mask pattern size. Fourier transformation is not suitable for calculating the diffracted waves from thick masks. Rigorous electromagnetic simulations such as finitedifference time-domain method, rigorous coupled wave analysis and 3D waveguide model are used to calculate the diffracted waves from EUV masks. However, these simulations are highly time consuming. We reduce the calculation time by adapting a CNN (Convolutional Neural Network). We calculate the far-field diffraction amplitudes from an EUV mask by using the 3D waveguide model. We divide the diffraction amplitudes into the thin mask amplitudes (Fourier transform of the mask pattern) and the residual mask 3D amplitudes. The incident angle dependence of the mask 3D amplitude for each diffraction order is fitted by using three parameters which represent the on-axis and the off-axis mask 3D effects. We train a CNN where the inputs are 2D mask patterns and the targets are the mask 3D parameters of all diffraction orders. After the training, the CNN successfully predict the mask 3D parameters. The CNN prediction is 5,000 times faster than the electromagnetic simulation. We extend the transmission cross coefficient formula to include the off-axis mask 3D effects. Our formula is applicable to arbitrary source shapes and defocus. We can use the eigen value decomposition method to accelerate the calculation.
Thin mask model has been conventionally used in optical lithography simulation. In this model the diffracted waves from the mask are assumed to be Fourier transform of the mask pattern. This assumption is the basis of Hopkins' method and sum of coherent system model. In EUV (Extreme UltraViolet) lithography thin mask model is not valid because the absorber thickness is comparable to the mask pattern size. Fourier transformation cannot be applied to calculate the diffracted waves from thick masks. Rigorous electromagnetic simulations such as finite-difference time-domain method, rigorous coupled wave analysis and 3D waveguide method are used to calculate the diffracted waves from EUV masks. However, these simulations are highly time consuming. We reduce the calculation time by adapting a convolutional neural network. We construct a convolutional network which can predict the diffracted waves from 1D EUV mask patterns. We extend the TCC method to include the off-axis mask 3D effects. Our model is applicable to arbitrary source shapes and defocus.
We directly extracted the phase-shift values of an EUV mask by measuring the reflectance of the mask. The mask had
gradient absorber thickness along vertical direction. We measured the reflectance of the open multilayer areas and the
absorber areas by using an EUV reflectometer at various absorber thicknesses. We also measured the diffracted 0th order
light intensities of grating patterns having several sizes of lines or holes. The phase-shift values were derived from these
data assuming a flat mask interference model of the diffracted lights. This model was corrected by including the
scattering amplitude from the pattern edges. We recalculated the phase-shift values which was free from the mask
topological effect. The extracted phase-shift value was close to 180 degrees at 67 nm and 71 nm absorber thicknesses.
The phase measurement error around 180 degree phase shift was 5 degrees (3σ).
Phase-shifting effect of EUV masks with various absorber thicknesses has been studied both by simulations and
experiments. In EUV lithography, masks with 180 phase shifting absorber work like embedded attenuated phase-shifting
masks. At 66nm thickness of TaN/TaON absorber, 180 degree phase shifting can be achieved in theory. Based on the
experiments, we observed that the true180 degree phase shifting can be achieved with absorber thickness between 66 and
76 nm. In this paper, phase shifting impact of the various thickness absorbers has been characterized. Imaging
performance of masks with 51 nm, 66 nm and 76 nm thick absorber has been experimentally compared. The process
window of various thickness absorber masks are rigorously studied.
When compared to a thick absorber mask, a thin absorber EUV mask is expected to have a comparable process
window, a reduced shadowing effect, and lower MEEF. However, regardless of the mask absorber thickness, the
dark-field in EUV lithography is never 100% dark. Using the same absorber stack composition, EUV masks with
thinner absorbers have inherently higher leakage due to the background transmission propagating through the absorber
stack. While this does act to improve resist sensitivity or throughput, the leakage reduces the image contrast and can
cause CD degradation in "double" exposed regions at the edge of adjacent fields. In this study, EUVL lithographic
benchmarking of both thin and thick absorber masks on the ASML Alpha Demo Tool (ADT) at IMEC is presented.
Herein, we experimentally quantify the process window, EL, LWR, MEEF, Esize, ultimate resolution, and impact of
dark-field background exposures on CDs for both thin and thick absorber masks. There are additional issues when
field edges overlap with adjacent fields, and mitigation strategies for EUV leakage emanating from dark-field regions
Electron backscattering from Extreme Ultraviolet (EUV) masks during Electron Beam (EB) exposure was studied by
simulations and experiments. The film structure of EUV masks is quite different from that of photomasks. The Mo/Si
multilayer on the EUV substrate is very thick (280 nm) and heavy metal material such as Ta is used for the absorber.
Monte Carlo simulations suggest that the absorbed energy inside the resist caused by the backscattered electrons from
these films is non-negligible, about 1/10 of the forward scattering electrons and 1/4 of the backscattered electrons from
the substrate. Also the simulations show that the influence range is very short because the backscattering happens near
the mask surface. These simulations were verified by conducting EB exposure experiments. Short-range proximity
effect was clearly observed by measuring the resist Critical Dimentions (CDs) of short bars laid beside the large exposed
area. The data were fitted by assuming a backscattering electron distribution which has an exponential form with 0.4 μm
range. The range is very short compared with the conventional proximity range of 10 μm. We conclude that the
conventional EB proximity effect correction method needs to be revisited for EUV masks.
Proc. SPIE. 6607, Photomask and Next-Generation Lithography Mask Technology XIV
KEYWORDS: Wafer-level optics, Light sources, Spatial frequencies, Fourier transforms, Scanning electron microscopy, Photomasks, Line width roughness, Line edge roughness, Semiconducting wafers, Decision support systems
Contribution of mask line edge roughness (LER) to resist LER on wafers was studied both by simulations and experiments. LER transfer function (LTF) introduced by Naulleau and Gallatin was generalized to include the effect of mask error enhancement factor (MEEF). Low spatial frequency part of LTF was enhanced by MEEF while high spatial frequency part was suppressed due to the numerical aperture limit of a stepper. Our model was experimentally verified as follows. First LER of a mask was measured by a scanning electron microscope. Then the mask LER was multiplied by LTF to simulate the aerial image LER on wafers. It was confirmed that the simulated LER agreed well with the LER measured by AIMSTM. Based on our model the contribution of the mask LER to the resist LER on wafers was estimated. According to our estimation the requirement of the mask LER should be as tight as that of the resist LER on wafers.
We discussed the following topics at the panel discussion of Photomask Japan 2002. 1) Pushing the limit of ArF lithography: How far ArF lithography will extend? 2) Current status and issues for F2 lithography, 3) Current status and issues for EPL, 4) Current status and issues for F2 and NGL masks, 5) Lithography tool selection from 90- to 65- nm nodes.
ArF lithography could extend to 65-nm node by using alternating phase shift masks. F2 lithography and EPL are not yet established and we need to solve many issues for practical applications. The choice of lithography tools in 65-nm node depends on devices and layers. Multiple lithography tools might be used in 65-nm node.
We discussed the following topics at the panel discussion of Photomask Japan 2001. 1) Lithography roadmap from 130-nm to 90-nm node. 2) Lithography tool selection, KrF, ArF, or F2 for each node. 3) Mask technology issues for ArF and F2 lithography. 4) CD control and MEEF reduction. 5) OPC and PSM applications. Panelists agreed that critical issues for 100-nm lithography are CD control, defect control and mask cost. Mask suppliers presented potential solutions for these issues: improvement of mask writing tools, refinement of resist process etc. By solving these issues 100-nm lithography can be realized by 2004.
We studied the mask error enhancement factor (MEF) for four 0.15-micrometers patterns, isolated lines, line and space (L and S), isolated holes, and dense holes, for various process conditions. The MEF for isolated lines was the smallest of all. The MEF for L and S was not as small as that of the isolated lines. We obtained a less than 15-nm wafer critical dimension (CD) variation, when we reduced the mask CD variation to 20 nm. For the isolated-hole patterns using an attenuated phase shift mask (PSM) with large mask biasing can reduce the wafer CD variation. On the other hand, it is very difficult to reduce the MEF and the wafer CD variation for the dense-hole patterns. The alternating PSM was the best of the evaluated process, but it was not good enough to reduce the mask CD variation.
The influence of spherical aberration on imaging performance was evaluated by resist simulation for various resist thicknesses and other resist parameters. The best focus variation in terms of pattern size in L and S was not appropriate as a lithographic criterion because it varied not only with pattern size but also with resist characteristics and thickness. General rules for the best focus of L and S based on CD-defocus characteristics are proposed. The lithographic performance represented by the CD uniformity and the best focus variation of isolated patterns were predicted approximately from the result of aerial image calculation for ideal resist performance. In conclusion, the reduction in spherical aberration that leads to improved CD accuracy is not always achieved by the decrease in best focus shift.
The lithographic performance and environmental stability of 193 nm chemically amplified resists based on norbornene t- butyl ester/maleic anhydride copolymer were investigated. Tert-butyl (t-Bu) cholate was used as a plasticizing additive to control the glass transition temperatures (Tg) of the resists. The resist softening temperature was decreased from > 180 degrees Celsius to 150 degrees Celsius by the addition of t-Bu cholate. The resist sensitivity and resolution were almost unchanged, when post-exposure bake (PEB) temperatures were below the resist softening temperature. High environmental stability of 1 hour was achieved by optimizing the resist softening temperature, because high temperature baking at around the resist softening temperature reduced the amount of residual solvent, and thus made the resist absorb basic airborne contaminants less readily. The environmental stability of the optimized resist was almost the same as that of 248 nm acetal type resist, and it was sufficient for practical use.
We compared the topography effect of two types of alternating PSMs; single-trench type with side etching and dual-trench type. The side etching value and dual-trench depth were adjust to give same linewidth in 0 degree and 180 degree regions for 0.2 micrometer L/S pattern. Several test patterns having different width and length were formed on these alternating PSMs. These two PSMs were evaluated by using an x4, 0.6 NA, KrF exposure tool. For longer patterns (similar to L/S pattern), pattern size differences were very small; the mask topography effect was negligible. However, pattern size differences of shorter patterns (similar to window pattern) were large with both Alt PSMs. Therefore, optimization of the side etching value or the trench depth is required for each mask pattern.
The practical OPC simulation system suitable for memory devices is developed with a simple threshold model considering acid diffusion in chemically amplified resists. And the resist parameter extraction method is also presented. The simulation performance is within 0.01 micrometer CD error, and within a few second computation time for 4 micrometer2 area memory cell on a EWS.
This paper describes the effects of the blocking level of polymer on both dissolution characteristics and lithographic performance in chemically amplified positive 193 nm resists consisting of a alicyclic methacrylate polymer and a photoacid generator. There were clear relationships between the blocking level and both the dissolution rate characteristics and the lithographic performance. We found that the dissolution contracts and developer selectivity improved as the blocking level of polymer increased, and then the resolution capability was improved. However, the dry-etch resistance and adhesion property of the resist film deteriorated as the blocking level increased, although these were at the same levels as those of conventional i-line novolak resist or polyhydroxystyrene base KrF resists. Therefore, these resist materials show potential for the next generation of LSI devices. Ideal dissolution parameters for improving the resolution capability were obtained by using a resist profile simulator. By using a high-contrast resist incorporating these parameters, and by using a higher numerical aperture lens and annular illumination, the mass production of next-generation devices with a 0.12 micrometers design rule can be achieved.
Optical proximity correction (OPC) is applied to the cell patterns of 0.15-micrometer-rule memory devices. Two kinds of memory cell patterns are studied. The first is a wire pattern which has small gaps between two wires. The small gaps can be clearly resolved by using OPC such as jogs or resizing. The second pattern is a storage node pattern which has a rectangular shape. The area of the storage node is enlarged by using OPC such as resizing, hammer heads or serifs. These OPC masks are successfully fabricated by using dry etching process.
Defect printability and sensitivity of reticle inspection systems were studied for 0.18 micrometer-rule devices. For our evaluation, an OPC test reticle was designed, and fabricated with E-beam and dry-etching. Base patterns are 0.18 micrometer-rule memory cells. Serif length is 0.3 micrometer and step is 0.1 micrometer (on reticle). The programmed defects have varieties of types, locations, and sizes. For the defect printability test, we used a 4 X KrF scanner (NA equals 0.6, (sigma) equals 0.75), and resist image was measured by CD-SEM. The defects which cause more than plus or minus 5% CD error were defined as 'printable' defects. It was cleared that very small defects can be printed on the wafer. For instance, 50 nm side placement defects were printed. Several inspection systems were evaluated and compared with our printability specification. From our result, there were no systems which have better performance than our specification. However, some latest systems were very close to our specification.
An assist-feature mask was fabricated for 0.2 micrometers window pattern formations using a dry etching process. Although the mask's assist-features were as small as 0.68 micrometers , mask inspection was successfully carried out using the cell-shift method. In addition, defects in assist-features were repaired by use of a laser mask repair system. The lithographic performance of this assist-feature mask was compared with that of a conventional mask, using a 4x KrF excimer laser exposure tool and a 0.7 micrometers thick positive resist. The numerical aperture (NA) of the exposure tool was 0.55 and annular illumination was used. The depth of focus of the 0.2 micrometers window was improved from 0.4 to 0.6 micrometers . Moreover, it was confirmed that defects in the assist- feature have little influence on the focus latitude of the main pattern. The DOF of patterns repaired with this technique recovered to nearly the same as that of the no- defect pattern.
The lithographic performance of high- and standard- transmittance attenuated phase shift masks (PSMs) was investigated in order to determine the suitability of applying attenuated PSMs to the fabrication of 0.15-micrometers hole patterns. Both PSMs had rim structures to eliminate side lobes, and they have two layers on the quartz substrate: a chromium-fluoride attenuated phase shifter layer and an opaque chromium layer. Both PSMs had similar lithographic performances that were high enough for 0.15 micrometers hole patterns.
Deblocking reaction mechanisms and lithographic performance in chemically amplified positive ArF resists were investigated by analyzing acid concentration and blocking level. The resists consisted of triphenylsulfonium triflate as a acid generator and either the copolymer, poly(carboxy- tetracyclododecyl methacrylate70-co- tetrahydropyranylcarboxy-tetracyclododecyl methacrylate30) or the terpolymer, poly(tricyclodecylacrylate60- co-tetrahydropyranylmethacrylate20-co-methacrylic acid20). The deblocking reaction mechanisms were evaluated from Arrhenius plots of the deblocking reaction rate constant. It was found that the deblocking reaction of both resists is ruled by two rate-determining steps, i.e., reaction-controlled in the low-temperature region and acid- diffusion-controlled in the high-temperature region. Furthermore, the copolymer resist had better post-exposure- delay (PED) stability. To clarify this result, acid loss caused by air-born contamination effect on deblocking reaction was investigated. The change of amount of blocking group by acid loss was small for the copolymer. Therefore the copolymer resist had better PED stability. Furthermore, the post-exposure bake (PEB) sensitivity of linewidth of the copolymer resist was smaller than that of the terpolymer resist. Both deblocking reaction rate constant and reverse reaction rate constant of the copolymer resist increased with PEB temperature. As a result, equilibrium constant of the copolymer was not valuable with temperature. This is the reason why the copolymer resist has low PEB sensitivity. It is concluded that small acid loss effect on deblocking reaction induces better PED stability. A resist with reverse reaction has an advantage for PEB temperature sensitivity.
The effects of acid structures and blocking groups in chemically amplified resists on compatibility with bottom anti-reflective coatings (BARCs), were investigated. The resists consisted of tert-butoxy carbonyl (t-BOC) or acetal blocked polyhydroxystyrene with three types of photoacid generators (PAGs) which generate trifluoromethanesulfonic acid (acid 1), 2,4-dimethyl benzensulfonic acid (acid 2) and cyclohexanesulfonic acid (acid 3). Three types of commercially available BARCs, Brewer Science CD9, DUV11 and DUV18 were used for this study. CD9 was decomposed by exposure and generated an acid substance, which induced the necking at the bottom of the resist films. In the case of DUV11, the generated acid from the PAG was neutralized, and footing was observed in t-BOC type resists. Acetal type resists had no footing on DUV11 because the deblocking reaction progressed without post-exposure baking. DUV18 had good compatibility with most of the resist materials because of its neutral acidity. From the viewpoint of resist materials, it was found that the acetal type resists tended to have necking, because the deblocking reaction occurred at lower acid concentration compared with t-BOC type resists. Moreover, the tendency to have a necking profile, in increasing order, was acid 3 > acid 2 > acid 1. This order corresponded with the reverse order of the efficiency of the deblocking reaction. A weak acid might be greatly affected by some substance diffused from a BARC. The acetal type resist with acid 1 had excellent compatibility with various BARCs. However, the resolution capability of the acetal type resist with acid 1 was lower than that of the acetal type resist with acid 3, because the acid diffusion length of acid 1 was larger than that of acid 3. It was concluded that good compatibility of the resist with the BARC is achieved by the high deblocking reaction efficiency and moderate diffusion length of acid in acetal type resists.
To improve the depth of focus of isolated windows, large assist feather technique has been proposed. This large assist method uses the assist features having almost the same size as main patterns, and the quartz substrate was vertically etched at the assist features. These large assist features were not printed on a wafer by mask topography effect; that is, the light intensity at large assist feature was decreased by the scattering effect of the vertical quartz edges. In this large assist feature masks, the phase shift angle of an assist feature has large effect on focus latitude. We chose two phase shift angles: 180 degrees for small sigma illumination and 360 degrees for annular illumination. The performances of two large assist feature masks were evaluated by using a 0.55 NA, valuable sigma, and KrF excimer laser stepper. Moreover, we applied surface insoluble layer to the assist feature method. Large assist features having the same size as main patterns were not printed on resist surface for 0.16 - 0.2 micrometer windows. Wide DOF (0.8 micrometer) of 0.16 micrometer window was obtained by using this method.
Optical proximity correction (OPC) was applied to alternating phase shift masks to improve printed resist pattern fidelity. Mask patterns were modified with jog type corrections. DRAM cell patterns were exposed by using a 0.55 NA, 0.36/0.55 (sigma) , KrF excimer laser stepper onto 0.5 micrometers thick chemically amplified negative resist. With 0.55 (sigma) , OPC was effective and printed resist pattern was very close to designed one. However, with 0.36 (sigma) , large pattern deformation was observed due to coma aberration.
We applied deep UV attenuated phase shift masks (PSMs) to the quarter micron level contact hole pattern layer of a DRAM. There were two different hole sizes: 0.26 micrometer hole in the memory cell region, and 0.35 micrometer in the peripheral circuit region. We examined two methods to reduce the side lobe effects in the peripheral circuit region. The first method was a chrome (Cr) shield method: the peripheral circuit region was covered by Cr. The second method was a mask bias method: large mask bias was added to contact hole patterns in memory cells. Both methods sufficiently reduced the side lobe effect in the peripheral circuit region.
A Xe2 excimer lamp (172 nm) has been used for submicrometer patterning of photoresist films. We resolved 0.35 micrometers lines and spaces using a contact printing system. We exposed a chemically amplified photoresist customized for ArF. The sensitivity of the photoresist was 130 mJ/cm$=2) or about ten times smaller than the sensitivity for ArF.
We delineated 0.088 micrometers line and space patterns by using an etched-in phase-shifting mask. The etched area of the mask had good morphology and high transmittance for deep UV light. The phase-shifting angle of the etched area was well controlled within 180 +/- 5 degrees.
Rigorous simulation has been shown to be useful for predicting complicated effects in the optical lithography process. Rigorous simulation models are necessary for accurately predicting behavior from non planar mask and substrate structures and for non-vertical light propagation. As previously shown[l], the benefits of 2D rigorous lithography simulation are many, however, these models are inherently limited. 2D models are not general enough to model contact openings, corners, 3D defect structures or off-axis illumina tion systems. These are important issues in lithography today and require true 3D simulation to help analyze them.
In automating optical proximity correction, calculation speed becomes important. In this paper we present a novel method for calculating proximity corrected features analytically. The calculation will take only several times the amount it takes to calculate intensity of one point on wafer. Therefore, the calculation will become extremely faster than conventional repetitive aerial image calculations. This method is applied to a simple periodic pattern. The simulated results show great improvement on linearity after correction and have proved the effectiveness of this analytical method.
A new model is presented to calculate the electromagnetic fields inside a resist according to vector potentials. The model can handle three-dimensional electromagnetic fields with fewer variables than Maxwell’s equation. The perturbation theory is applied to this model to describe the resist bleaching process.
In order to improve resolution and depth of focus (DOF) in reduction projection aligner, we investigate annular illumination method, in which the center portion of light source is screened.
This paper describes the relationship between resist pattern formation capability and the screened ratio of condenser aperture. And also, the pattern deformation induced by annular illumination is investigated.
First, on the basis of simulation analysis of this illumination method, the optimum optical parameters were selected to obtain high resolution and enough DOF. The effect of annular illumination was confirmed experimentally in i-line exposure. Secondly, to explore optical contrast dependence, we compared the resolution capabilities of both positive and negative resists, having different dissolution characteristics each other. It was found that annular illumination is more effective in low contrast region of light intensity. Thirdly, the dependence on pattern periodicity (L/S and isolated line), and pattern deformation were also examined. This method is effective in periodic pattern but not in isolated pattern, and induces a little pattern deformation in the edge region.
Annular illumination that is a simple method compared with phase shift mask, is a promising for expanding the process margin of 0.35 ?m resist patterning. Furthermore, this method becomes more effective, if adapting thin resist or new high resolution resist, which can be resolved even in low contrast light condition.