Achieving the ultimate resolution limit of EUV lithography is greatly impeded by the 3D photomask geometry, including an absorber whose thickness is comparable to the minimum lateral dimensions of the pattern, and a reflection plane a similar depth beneath the surface of the multilayer mirror. Rigorous simulations have shown that these effects can in theory be mitigated by adopting a thinner absorber and a multilayer with a reflection plane closer to the surface. But regardless of how rigorously the design is optimized, there is clearly a need to experimentally confirm that the as-built photomask conforms to the simulation’s predicted complex electric field. This experimental confirmation is difficult because only the field’s intensity is directly observable. One promising approach to unambiguously make this measurement is Zernike phase contrast imaging, which determines the complex electric field from intensity images acquired from a single illumination condition with different phase shifts on the 0 order. In this work we present an extension to a hyperspectral version of the technique. By varying the wavelength, we are able to empirically observe the complicated interaction between absorber, multilayer, pattern, and illumination. We performed an experimental demonstration of the technique on a patterned EUV mask with 60nm TaN absorber using specially fabricated zone plates on the SHARP EUV microscope at the Center for X-Ray Optics. Our results demonstrate the sensitivity of hZPC to both the Fresnel reflectance as well as more subtle 3D effects also observed in rigorous simulations.
With growing interest in EUV attenuated phase shift masks due to their superior image quality for applications such as dense contact and pillar arrays, it is becoming critical to model, measure, and monitor the relative intensity and phase of multilayer and absorber reflections. We present a solution based on physical modeling of reflectometry data, which is capable of achieving single picometer phase precision. During repeated reflectometry measurements we observed a systematic change in absorber reflectivity which we attribute to the growth of a carbon film from 44-156pm, causing a change in the relative phase of 0.3°. This represents sensitivity to changes in the average film thickness to well below one atomic monolayer. After separating out systematic drift from random noise, we estimate our precision to be 3σ = 0.1°, corresponding to 3-4pm.
In this paper, we present two methods for directly measuring the effective complex reflectance function of a patterned EUV mask. Obtaining this measurement can provide important insight into a number of key areas in EUV mask development, including obtaining a deeper understanding of mask 3-D effects, and characterizing and quantifying the amplitude and phase generated by attenuated and etched phase shift masks. The first method, Quantitative Zernike Phase Contrast Microscopy (QZPCM), works by modulating the imaging pupil function with several known phase shifts, and obtaining through-focus images of a target area for each pupil setting. The second method, Lateral Shearing Imaging (LSI), works by splitting and interfering two copies of the complex amplitude function in the image plane separated by a distance s. The resulting fringe pattern gives information about the derivative of the complex amplitude in the direction of the shift. We present results from two experiments: the first demonstrates QZPCM at EUV on the Berkeley SHARP microscope, and the second utilizes LSI in an optical prototype using a visible light laser source.
We present a comparison of experimental techniques for measuring the as-built phase shift of EUV photomasks to meet the unique requirements for EUV lithography at the resolution limit. Attenuated phase-shift masks provide superior image quality for certain applications such as low-k1 contact and pillar arrays, offering increased throughput and reduced stochastic printing failures. But whereas the traditional phase-shift is π, rigorous electromagnetic simulations suggest the optimal phase-shift for an EUV photomask must be adjusted to account for Mask 3D effects, which are themselves difficult to measure. In this work, we explore at-wavelength metrology including reflectometry, scatterometry, and phase imaging for measuring multilayer and absorber reflectance, as well as complex scattering amplitudes for a grating with pitch p = 420nm and duty-cycle D = 0.33. Using rigorous electromagnetic simulations, we find that relying only on reflectometery and the Fresnel thin-mask model predicts the complex scattering amplitudes with 22% accuracy due to 3D effects, whereas a combination of scatterometry and through-focus imaging can achieve a promising 0.6% accuracy, and a combination of scatterometry and Zernike Phase-Contrast imaging can achieve a superior 0.1% accuracy. Experimental results based on imaging and scatterometry clearly display 3D effects that differ substantially from idealized rigorous simulations, suggesting the difficulty of accurately predicting 3D effects and hence the need to accurately measure them.
We demonstrate complementary reflectometry and scatterometry methods to measure the phase and amplitude of a patterned EUV photomask at its operating wavelength (13.5nm) and angle range (2 − 10°). We carried out experimental measurements at ALS Reflectometry and Scattering Beamline 6.3.2 on an EUV photomask with a 40-bilayer MoSi multilayer mirror and 60nm TaN absorber. We took three types of measurements: reflectometry for blank multilayer, reflectometry for blank absorber-coated multilayer, and scatterometry for line- space gratings. We used the reflectometry data to fit the Fresnel reflectance amplitude by adjusting the thickness, atomic density, and interface roughness of multilayer and absorber layers. We then fit the scatterometry data using a thin-mask approximation. The advantage of reflectometry is the higher level of model fidelity (2 − 4% vs 6% − 15% error), whereas the advantage of scatterometry is its direct sensitivity to relative phase through spatial interference. Despite differences between the two approaches, both gave similar phase values, mutually-consistent to within π/51 RMS. We observe the phase to vary from 0.78μ at 2° to 0.88μ at 10°, suggesting that engineering phase effects to improve image contrast will need to consider phase values across a range of illumination angles rather than simply the chief ray.
In this work, we demonstrate a method to design the Mo-Si multilayer stack of an EUV photomask to increase the optical efficiency of shadowing-orientation equal lines and spaces imaged under dipole illumination. We achieve this using a computational framework written in the PyTorch machine learning library, which is capable of optimizing the multilayer for partially-coherent imaging rather than specular reflectivity. After computing optimal multilayer designs for both 0.33 and 0.55 NA EUV systems, we verify the improvements via RCWA simulation. We demonstrate optical efficiency gains of up to 22%=14% for the 0.33/0.55 NA systems, respectively.
With EUV Lithography rapidly approaching maturity, accurate metrology to thoroughly characterize EUV photomasks is needed. We present an actinic EUV reflection-based scatterometry technique to measure key parameters of EUV photomasks to characterize both the multilayer mirror substrate as well as periodic absorber targets fabricated on the multilayer. We show these measurements can be used both in determining the physical dimensions on the mask, and also in directly quantifying optical effects, which can provide invaluable feedback in the mask optimization and manufacturing processes. In this paper, we present four different methods of data analysis for EUV mask scatterometry: two for characterizing the multilayer mirror based on measurements of the reflected light intensity from a flat open area of the mask, and two more for characterizing absorber gratings fabricated on the multilayer substrate based on measurements of the diffraction efficiencies. Key findings include that a simple neural net architecture containing a single fullyconnected hidden layer that can characterize the multilayer’s angularly-varying complex reflection coefficient to 7 × 10-4 accuracy, and that dictionary-based scatterometry with 7 wavelengths from 13.2 − 13.8nm can measure the absorber thickness of a grating to 0.4nm even in the presence of random and systematic errors. With the presented methods and findings, we hope to demonstrate that actinic EUV scatterometry has the capabilities to accurately characterize EUV masks in terms of both multilayer and absorber.
There are many applications where fast, accurate light scattering from EUV photomasks must be computed, including inverse mask design, actinic die-to-database inspection, and actinic scatterometry. However, so-called mask 3D effects make this calculation much more challenging than traditional optical lithography. These 3D effects arise from the optically thicker absorber, the lack of illumination symmetry about normal incidence, the multilayer mirror reflection function, and multiple scattering off the absorber. In this paper, we explore using actinic scatterometry at the CXRO EUV reflectometer to characterize both the multilayer and absorber of an EUV photomask; we then introduce the Multilayer Multiple Scattering (MLMS) mathematical model that conveniently separates the effects of the multilayer and the absorber and explore the implications of this model on the origins of mask 3D effects.
Achieving high-throughput extreme ultraviolet (EUV) patterning remains a major challenge due to low source power; phase-shift masks can help solve this challenge for dense features near the resolution limit by creating brighter images than traditional absorber masks when illuminated with the same source power. We explore applications of etched multilayer phase-shift masks for EUV lithography, both in the current-generation 0.33 NA and next-generation 0.55 NA systems. We derive analytic formulas for the thin-mask throughput gains, which are 2.42× for lines and spaces and 5.86× for contacts compared with an absorber mask with dipole and quadrupole illumination, respectively. Using rigorous finite-difference time-domain simulations, we quantify variations in these gains by pitch and orientation, finding 87% to 113% of the thin-mask value for lines and spaces and a 91% to 99% for contacts. We introduce an edge placement error metric, which accounts for CD errors, relative feature motion, and telecentricity errors, and use this metric both to optimize mask designs for individual features and to explore which features can be printed on the same mask. Furthermore, we find that although partial coherence shrinks the process window, at an achievable sigma of 0.2 we obtain a depth of focus of 340 nm and an exposure latitude of 39.2%, suggesting that partial coherence will not limit the feasibility of this technology. Finally, we show that many problems such as sensitivity to etch uniformity can be greatly mitigated using a central obscuration in the imaging pupil.
In DUV lithography, scatterometry enables precise measurement of mask dimensions such as the pitch, linewidth, and sidewall-angle of periodic patterns. However, substantial differences in the optical properties of DUV and EUV masks, such as angular sensitivity and mask 3D effects, makes simply extending existing technologies difficult. Using the EUV reflectometer at Lawrence Berkeley National Labs Center for X-Ray Optics with tunable wavelength and illumination angle, we explore how to extend scatterometry to EUV masks, with particular emphasis on using rigorous simulations and experimental data to quantify the accuracy of sensitive measurements such as sidewall-angle.
Mask scatterometry at EUV wavelengths has benefits but also poses challenges that are not present at DUV wavelengths. The benefits come primarily from using the same wavelength as lithography; due to the severe sensitivity of the multilayer mirror to wavelength, the diffraction patterns obtained at DUV wavelengths from EUV masks would be both highly attenuated and substantially distorted. However, stronger mask 3D effects and the sensitivity of the multilayer to angle of illumination add extra levels of complexity to modeling the spectra of EUV masks that are not present in traditional DUV masks.
We use rigorous FDTD (Finite Difference Time Domain) imaging simulations of patterned EUV multilayer masks to generate a library of spectra including gratings with a range of orientations, pitches, line widths, absorber heights, and side-wall angles under a wide range of illumination wavelengths and angles. We then perform SVD-based dimensionality reduction to find an efficient representation, or dictionary, for the spectra. Using this low-dimensional dictionary, we determine the sampling requirements, i.e. which measurements (angles and wavelengths of illumination) are necessary to measure all parameters of interest to a specified accuracy. We finally acquire experimental spectra of known mask features on the EUV reflectometer using different illumination conditions, and use the dictionary to recover the underlying dimensions of the features.
Contact-hole layer patterning is expected to be one of the first applications for EUV lithography. Conventional absorber masks, however, are extremely inefficient for these layers, placing even more burden on the already challenging source power demands. To address this concern, a chromeless checker-board phase-shift mask for 25- nm dense contacts has been shown to provide a throughput gain of 8x based on characterization with the SHARP EUV microscope and 7x based on micro field patterning with the Berkeley MET. These promising experimental results warrant both assessment for implementation in practice and rigorous simulations for diagnosing 3D mask effects. In this paper we verify the theoretical benefits of phase-shift masks over traditional absorber masks in idealized Kirchhoff analysis, explore the sensitivity of patterning to deviations from the ideal scattered orders, model the etched multilayer using thin-film characteristic matrix analysis, and finally use rigorous 3D Finite-Time Time Domain (FTTD) simulations of etched multilayer masks to explore mitigation of 3D effects to achieve optimal mask designs for minimum-pitch line-space and contact array patterns.
Feedback control of overlay errors to the scanner is a well-established technique in semiconductor manufacturing . Typically, overlay errors are measured, and then modeled by least-squares fitting to an overlay model. Overlay models are typically Cartesian polynomial functions of position within the wafer (Xw, Yw), and of position within the field (Xf, Yf). The coefficients from the data fit can then be fed back to the scanner to reduce overlay errors in future wafer exposures, usually via a historically weighted moving average. In this study, rather than using the standard Cartesian formulation, we examine overlay models using Zernike polynomials to represent the wafer-level terms, and Legendre polynomials to represent the field-level terms. Zernike and Legendre polynomials can be selected to have the same fitting capability as standard polynomials (e.g., second order in X and Y, or third order in X and Y). However, Zernike polynomials have the additional property of being orthogonal over the unit disk, which makes them appropriate for the wafer-level model, and Legendre polynomials are orthogonal over the unit square, which makes them appropriate for the field-level model. We show several benefits of Zernike/Legendre-based models in this investigation in an Advanced Process Control (APC) simulation using highly-sampled fab data. First, the orthogonality property leads to less interaction between the terms, which makes the lot-to-lot variation in the fitted coefficients smaller than when standard polynomials are used. Second, the fitting process itself is less coupled – fitting to a lower-order model, and then fitting the residuals to a higher order model gives very similar results as fitting all of the terms at once. This property makes fitting techniques such as dual pass or cascading  unnecessary, and greatly simplifies the options available for the model recipe. The Zernike/Legendre basis gives overlay performance (mean plus 3 sigma of the residuals) that is the same as standard Cartesian polynomials, but with stability similar to the dual-pass recipe. Finally, we show that these properties are intimately tied to the sample plan on the wafer, and that the model type and sampling must be considered at the same time to demonstrate the benefits of an orthogonal set of functions.