Suchit Bhattarai

Study of energy delivery and mean free path of low energy electrons in EUV resists

The relative importance of secondary electrons in delivering energy in photoresist films was assessed by performing large area exposures and by quantifying the inelastic mean free path of electrons in a leading chemically amplified positive tone EUV resist. A low energy electron microscope was used to directly pattern large (~15μm x 20μm) features with 15-80 eV electrons followed by analyzing the resulting dissolution rate contrast curve data. In the 40 to 80 eV regime the energy delivery was found to scale roughly proportionally with electron energy. In 15 to 30 eV regime however, this energy scaling did not explain the resist thickness loss data. The dose required to lower the resist thickness down to 20 nm was found to be 2-5X larger for 15 eV electrons than for 20, 25 and 30 eV electrons. Using scattering models from the literature including phonon scattering and optical data deduced electron energy loss spectroscopy and optical reflectometry, the inelastic mean free path values at energies between 10 eV and 92 eV range between about 2.8 and 0.6 nm respectively.

Limitations of resist-based characterization of EUV mask surface roughness

The feasibility of wafer-plane measurements of EUV mask surface roughness has been analyzed through stochastic resist simulations at various defocus conditions, for mask surface roughness values ranging between 50 pm and 500 pm rms. With partial coherence of 0.5, NA of 0.25, defocus of 100 nm and mask surface roughness of 50 pm rms, 1.3% of the total resist LER is contributed by the mask surface roughness induced aerial image phase roughness, while 39.1% of the total LER contribution comes from the absorbed photon image. 31.4% of the LER contribution is from the acid image and 27.9% is attributable to the quencher image at the end of the PEB reaction/diffusion processes. For surface roughness values of interest ranging between 50 pm and 150 pm rms, partial coherence of 0.5 and 100 nm defocus, the sensitivity of wafer plane aerial image LER to mask surface roughness is 9.5 nm/nm-rms, while the resist LER sensitivity is 2.9 nm/nm-rms. With hypothetical scaling of the resist parameters, the resist LER sensitivity to mask surface roughness increases to 6 nm/nm-rms.

Comparative analysis of shot noise in EUV and e-beam lithography

Gray-scale e-beam lithography has been performed to match the EUV and e-beam aerial image log slope for studying shot noise fundamentals in the two mechanisms through line-edge roughness (LER) measurements for 50 nm lines and spaces patterned on a leading chemically amplified EUV resist. The measured e-beam exposure latitude decreased from 0.4 with binary patterning to 0.28 with gray-scale e-beam exposure designed to match the EUV incident image profile, closely matching the EUV exposure latitude of 0.26. Calculations of absorption statistics with EUV and e-beam suggest that the shot noise with e-beam patterning is expected to be 10% larger than the shot noise with EUV patterning. However, despite the matched image gradients and close to identical absorbed quanta predictions, the e-beam patterned LER is 2.5× larger than the EUV patterned LER.

Analysis of shot noise limitations due to absorption count in EUV resists

Both fundamental measurements of resist exposure events and measurements of line-edge roughness for similar exposure latitude images for e-beam and EUV patterning tools have been used to assess the relative role of exposure shot-noise in lithographic performance. Electron energy loss spectroscopy (EELS) has been performed to quantify the probability of absorption of 100 keV electrons in two commercially available EUV resists. About 1/3 of the incident electrons lose at least 2 eV in the materials and this absorption probability is larger than that for EUV photons in the two modern EUV resists. Exposure event count densities between EUV and e-beam differ by 11-13%, which results in an expected difference in the variation in exposure shot noise of only 6%. With matched image exposure latitudes and accounting for EUV mask LER contribution the measured LER distributions indicate a high (76% and 94%) confidence that EUV resist performance is currently not dominated by exposure event counts for two leading chemically amplified EUV resists.

Simulation study of the influence of PEB reaction rates on resist LER

A stochastic resist simulator has first been calibrated to experimental results performed on a commercially available EUV resist, and subsequently has been used to study the influence of acid/base quenching rate and the polymer deprotection rate on resist LER for 22 nm half-pitch lines/spaces. Results indicate that larger quenching rates and smaller deprotection rates result in improved LER performance by causing an increase in the dose to size. With nominal quenching rate determined from literature, halving the deprotection rate relative to nominal value reduces the LER by 33%, while the dose to size increases by 2x. With nominal deprotection rate determined from literature, results indicate a low sensitivity of LER to quenching rate. Expected noise at the line edge calculated by using a shot noise model accounting for absorbed photons, acid, and base count, provides a good explanation for the LER trends calculated for several reaction rate scenarios.

Contrast curves for low energy electron exposures of an EUV resist in a scanning electron microscope

We present an experimental technique for determining the energy delivery efficiency of secondary electrons in an EUV resist, by directly exposing a positive tone chemically amplified resist with 29- 91 eV electrons created by utilizing the deceleration technology in a scanning electron microscope. Charging is an important problem associated with thin film exposure experiments. We assess the feasibility of using the SEM frame rate as a knob for controlling charging related artifacts. Preliminary measurements of secondary electron emission signal from an unexposed region in the resist provide clues about the time domain surface potentials that may form while the sample charges during exposures. These signals are found to change as a function of the SEM frame rate and landing energies. We provide contrast curve data for resist exposures with 29 eV, 49 eV and 91 eV electrons at three frame rates of 33 ms/frame, 8 s/frame and 30 s/frame. The energy delivery efficiency of electrons estimated for all three frame rates are also provided.

Study of shot noise in photoresists for extreme ultraviolet lithography through comparative analysis of line edge roughness in electron beam and extreme ultraviolet lithography

In this paper, the influence of the absorption shot noise on line edge roughness (LER) of photoresists for extreme ultraviolet (EUV) lithography is studied experimentally through the comparative analysis of LER obtained by EUV (92 eV photons) and 100 keV e-beam lithography. Techniques for performing EUV and e-beam lithography with a matched image log slope for a fair comparison of LER values are described. Measurements of absorption of 100 keV electrons estimated through a transmissive electron energy loss spectroscopy measurement with a 120 keV electron beam showed that despite having access to core levels in the material (e.g., 284 eV edge in carbon), these electrons mostly just excite the energy levels less than 100 eV in the resist, with a mean deposited energy of 35 eV. By combining the incident flux and the absorption probabilities, the absorbed quanta for patterning of 50 nm half-pitch line/space features was found to be similar between the two patterning technologies.

Simulation analysis of LER and dose tradeoffs for EUV resists with photo-decomposable quenchers

The SuMMIT stochastic simulator has been used to conduct a simulation study of photo-decomposable quencher (PDQ) based EUV resists and performance comparison between PDQ resists and conventional quencher (regQ) resists analyzed from the standpoint of dose and LER metrics. The dose and LER Tradeoffs have been analyzed as a function of base loading, base diffusion lengths and relative deprotection/quenching rates. About 3.5% LER improvements with PDQ has been predicted at a dose of 15 mJ/cm2 with base loaded at 20% of PAG loading, for 25 nm half-pitch line-space patterns. Dose savings of 2 mJ/cm2 and LER improvement of 0.1 nm between regQ and PDQ resists are predicted with a base diffusion length equal to the acid diffusion length of 10 nm, and base loaded at 30% of PAG loading. Dose improvements of 1 mJ/cm2 for equal regQ and PDQ LERs of 3.5 nm is possible at a deprotection rate that is half as fast as the acid/base quenching rate of 10 nm3/s. Improvement in the deprotection gradient is found to be the dominant factor behind lower PDQ LERs, while the difference in deprotection noise between conventional quenchers and PDQs is found to be marginal.

Simulation analysis of LER and dose tradeoffs for EUV resist with photo-decomposable quencher

The SuMMIT stochastic simulator has been used to conduct a simulation study of photo-decomposable quencher (PDQ) based EUV resists and performance comparison between PDQ resists and conventional quencher (regQ) resists analyzed from the standpoint of dose and LER metrics. The dose and LER Tradeoffs have been analyzed as a function of base loading, base diffusion lengths and relative deprotection/quenching rates. About 3.5% LER improvements with PDQ has been predicted at a dose of 15 mJ/cm2 with base loaded at 20% of PAG loading, for 25 nm half-pitch line-space patterns. Dose savings of 2 mJ/cm2 and LER improvement of 0.1 nm between regQ and PDQ resists are predicted with a base diffusion length equal to the acid diffusion length of 10 nm, and base loaded at 30% of PAG loading. Dose improvements of 1 mJ/cm2 for equal regQ and PDQ LERs of 3.5 nm is possible at a deprotection rate that is half as fast as the acid/base quenching rate of 10 nm3/s. Improvement in the deprotection gradient is found to be the dominant factor behind lower PDQ LERs, while the difference in deprotection noise between conventional quenchers and PDQs is found to be marginal.