Effective Optimization Strategy for Electron Beam Lithography of Molecular Glass Negative Photoresist

As the crucial dimension (CD) of logic circuits continues to shrink, the photoresist metrics, including resolution, line edge roughness, and sensitivity, are faced with significant challenges. Photoresists are indispensable in the integrated circuit manufacturing industry, and specifically in achieving smaller critical dimensions. In this study, the effects of two categories of photosensitive compounds on lithography performance are explored, through a series of sulfonium salt‐based photoacid generators (PAGs) with diverse reactivity and photodegradable nucleophiles (PDNs) with varying nucleophilicity. The detailed characterization and exposure experiments suggest that the reactive alterations of different PAGs are mostly associated with the amount of phenyl composed of cations in PAGs. The “PDN first, PAG second” strategy, which employs a combination of low reactivity PAG and high reactivity PDN and involves PDN decomposition first and PAG decomposition second in the electron beam lithography process, achieves high sensitivity (100–270 µC cm−2), high resolution (25 nm 1:1 line/space, L/S), and low line edge roughness (LER ≤ 3.3 nm) stripes. This approach outperforms conventional formulations and may provide a potentially effective and useful strategy to improve electron beam photoresists.


Introduction
As predicted by Moore's Law, the crucial dimension (CD) representing the level of chip production has continuously decreased, leading to a need for improved photoresist metrics such as line edge roughness and sensitivity. The size of CD is typically determined by the Rayleigh formula, [1] with the common strategy for achieving a smaller CD being to reduce the wavelength of the exposed light source. The lithography exposure source eventually evolved from g-line (436 nm) to the current extreme ultraviolet (EUV, 13.5 nm) to accommodate the shrinking process. [2] Presently, EUV lithography has been able to reach lithography nodes of 5 nm and below, albeit with a high cost in equipage. For this reason, electron beam lithography (EBL) has been frequently utilized to evaluate the performance of EUV photoresists, and it provides a valuable contribution in fields like high-precision mask manufacturing, prototype device research and development, small-volume production, and fundamental scientific research. [3][4][5][6] In this study, we performed optimization experiments using EBL to develop high-performance e-beam photoresists, which could also serve as a foundation for EUV lithography.
While e-beam photoresist research has made significant progress, meeting all the demands for high resolution, high sensitivity, and low line edge roughness (LER) simultaneously remains a major challenge. [7] Two commercially available nonchemical amplification photoresists, Hydrogen silsesquioxane (HSQ) and polymethyl methacrylate (PMMA), can achieve sub-30 nm dense lines but at the cost of high exposure doses. [8][9][10] To improve sensitivity, hybrid photoresists containing metals have been proposed but are challenged by the potential introduction of metal contaminants to microcircuits and difficulties in stripping them in subsequent processes. [11][12][13] When smaller pitches are required for patterning in the next-generation lithography, molecular glass compounds with a single molecular weight distribution, small molecular mass, and limited entanglement compared to polymers are thought to be potential candidates for the next generation photoresists. [14] One of which is positive-tone photoresist (PTR) with acid leaving groups, which undergoes solubility changes during acid decomposition and eventually develops patterns under alkaline water development conditions. [14][15][16][17][18][19][20][21][22] The high surface tension of alkaline water presents a challenge for developing patterns, as it can cause the pattern to collapse or distort easily. The other one is negative-tone photoresist (NTR), which forms dense structures through ring-opening polymerization between epoxides or dehydration condensation among hydroxyl groups under acid catalysis. [23,24] NTRs use an organic developer with low surface tension, offering greater potential for forming high-resolution structures. For the primary study for EBL, a molecular glass NTR, BPA-6Ep, which contains bisphenol A core and epoxy group was selected. [25] This molecular core has the potential to create fine stripes and has previously been used for PTR with tert-butyloxycarbonyl (BPA-10) and an isolated line of sub-20 nm was obtained under EUV exposure conditions (Figure 1). [20] In recent years, there has been a growing demand for photolithography technology to achieve higher resolution. To address this need, Lawson et al. [26] introduced the concept of photodegradable nucleophiles (PDN) in the NTRs. [27] This technique aims to enhance resolution from sub-30 to sub-20 nm by inhibiting the migration and diffusion of the active end from the exposed to the non-exposed region through the use of nucleophilic reagents. Moreover, the PDN' photolysis capability enhances the photoresist's contrast, leading to further improvements in lithography performance. However, it should be noted that if the de-composition rate of the PAG is faster than or similar to that of the PDN, incomplete decomposition of the PDN can lead to the formation of low cross-linkage polymers. This results in insufficient mechanical properties and distortion of the photolithographic patterns. To address this issue, the "PDN first, PAG second" photoresist system was proposed, wherein PDN decomposes first and PAG decomposes later in the photoresist formulations. To achieve this, a series of PAGs with varying reactivity and PDNs with high reactivity (some of which are commercially available) were designed and synthesized ( Figure 1). In addition, the amount of phenyl group in the cations of PAGs was adjusted to control their reactivity, while PDNs were produced by incorporating anions with various nucleophilicity and cations with high reactivity. [28,29]

Low Reactive PAG Selected to Optimize Photoresist Formula
In preliminary lithography studies using BPA-6Ep, it was observed that even modest amounts of the traditional PAG TPS-SbF 6 (less than 5 wt%) caused significant bridging at low exposure doses of 30 μC cm −2 when forming stripes with a half pitch (HP) of 50 nm ( Figure S1, Supporting Information). This phenomenon can be attributed to the high sensitivity of the molecular glass formula. To address this issue, we endeavored to lower PAG's sensitivity by designing and synthesizing a series of PAGs with varying amounts of phenyl to prevent the uneven distribution that occurs due to the gradual decrease of PAG content. The goal was to identify a PAG with moderate acid production efficiency that would facilitate the optimization of photoresist formulations (Figure 1). The selection of PAGs was done based on the well-established acid production mechanism of bond cleavage-recombination of cations. [30][31][32][33] According to the reaction mechanism, protons are usually produced by the phenyl group contained in the sulfonium salt, therefore, we presume that the more the number of phenyl group contained in the cation of PAG, the stronger the acid production ability.
In EBL, the electrophilic properties of PAGs are of utmost importance, and the lowest unoccupied molecular orbitals (LUMO) typically exhibit a direct correlation with the electron affinity potential. [34,35] Therefore, we utilized Gaussian 09 software with density functional theory (DFT) methodology to optimize the conformation of PAGs, using the B3LYP together with the 6-31G(d,p) basis set. We also calculated the spatial distribution of LUMOs for the cations of PAGs, as shown in Figure 2a. Notably, in all cases, the charged atom (S+) was located at the center of the LUMO region and thus expected to be directly involved in the mechanism of PAG decomposition through bond cleavage. [36] To verify the accuracy of the computation results, we performed cyclic voltammogram (CV) of PAGs using a standard threeelectrode apparatus with Ag/Ag+ as a reference electrode. [37] To eliminate the effect of water and oxygen, we prepared highly concentrated PAG solutions (10 mm) and supporting electrolyte (tetrabutylammonium hexafluorophosphate, 0.1 m) in acetonitrile, which provided a suitable potential window for the PAG reduction assay. Then, we swept the applied potential at a scan rate (SR = 100 mV s −1 ) in the cathodic direction and swept it back in the opposite direction when the peak potential (Ep) was detected. The resulting reduction potential peaks of PAGs were depicted in Figure S2 (Supporting Information) and the LUMO energies were calculated from E LUMO,CV = -(E onset -E 1/2 Fc+/Fc + 4.8) eV, showing consistency with the DFT calculation (see Table 1 and Figure 2b). [38] The electron affinity of PAGs did not increase with phenyl. To confirm the difference in acid production efficiency among the PAGs, we proceeded with lithography experiments. PAGs containing SbF 6 − and CF 3 SO 3 − , along with NTR BPA-6Ep and PTR BPA-10, respectively, were used in EBL. DMS/DPS/TPS-SbF 6 in BPA-6Ep could not yield better contrast experimental results due to relatively high sensitivity. The lithography stripe bridging gradually intensifies as the number  of phenyl in cations increases, whereas the formulation containing TMS-SbF 6 does not form a lithographic pattern because of its weak acid production capacity ( Figure S3, Supporting Information). Contrast experiments of BPA-10 provided quantitative characterization of the difference in sensitivity of PAG ( Figure  S4, Supporting Information). Figure 2c shows the sensitivity expressed in terms of E 50 (the dose required to lose half of the film thickness, Table 1) of various PAGs. Experimental results confirm that the acidity of both groups of PAGs increases with the number of phenyl in the sulfonium salt, irrespective of the counter-ion being SbF 6 − and CF 3 SO 3 − . In the EBL experiments, both dimethyl sulfonium (DMS) and diphenyl sulfonium (DPS) showed similar exposure effects. We further investigated the hypothesized structure-property relationship of the acid generators by conducting X-ray photoelectron spectroscopy (XPS) experiments. To minimize errors, different PAGs were added to the matrix material BPA-6Ep at a 20 wt% mass ratio, and each formula was exposed at 30 μC cm −2 . As shown in Figures S5-S8 (Supporting Information), the S of S-Csp3 was seen at 167 eV while the S of S-Csp2 was observed at 164 eV. After exposure, new C-F (688.8 eV) was observed in all PAGs, [39] and a new S-Csp2 was seen in the TMS formula, where a chemical bond is formed between the benzene in the photoresist material, which generates protons, and the S after its cleavage. We compared the change in S element before and after the reaction, using O element as a reference, and found that the S content in the TMS formula was greatly reduced whereas it remained practically unchanged in the other three PAGs (Figure 2d). Therefore, we reasonably conclude that S-Csp3 is cleaved upon DMS exposure, indicating that PAGs containing methyl groups preferentially eliminate the methyl group rather than the phenyl group. Based on this characteristic, we compared the reaction activities of DMS and DPS by evaluating the change in the S-Csp3 content. The relevant information of the S element and the ratio of the reaction calculated based on the reduction of S-Csp3 were included in Table 2, where DPS demonstrated significantly higher reaction activity than DMS. Furthermore, we found that the proportion of S-Csp3 before exposure was 0.4 and 0.2, respectively, which were lower than the actual 0.67 and 0.33. This could be attributed to the environment of the sulfonium salts in the film.

High Reactive PDN Selected to Improve Lithography Performance
Although we have successfully reduced the sensitivity of photoacid generators (PAGs) to mitigate the issue of lithographic stripe bridging, further adjustments to the photoresist components are still necessary. Initially, we introduced TPS-OH as a photo-decomposable nucleophile (PDN) and combined it with the commercial PAG TPS-SbF 6 for electron beam lithography (EBL). However, we observed distortion and pattern collapse despite successful bridging inhibition (as shown in Figure S9, Supporting Information). This highlighted the insufficient degree of cross-linking caused by incomplete decomposition of the PDN, in turn impeding the mechanical performance of the photoresist due to PAG-initiated polymerization. To address this challenge, we replaced TPS-SbF 6 with a moderately sensitive PAG, DMS-SbF 6 , and conducted EBL with three PDNs (TPS-Tf/TPS-Ts/TPS-OH) featuring different nucleophilic anions in the same weight ratio. The optimal lithography patterns for each formulation are shown in Figure 3, indicating that inhibiting bridging is improved as the nucleophilicity of PDNs' anionic components increases while minimizing bridging. As previously discussed, introducing nucleophiles limits active-end migration from exposed to non-exposed areas, yielding enhanced lithographic outcomes. Our results show that TPS-OH, with stronger nucleophilicity as a PDN, generates superior photolithography patterns.
To further enhance the quality of the lithographic stripe, we performed EBL successively using different ratios of PAG and PDN mass addition for the DMS-SbF 6 formulation (PAG: PDN = 1:1/1:2/1:3/1:4). TPS-OH, which has the greatest nucleophilicity and exhibits a notable inhibitory effect on bridging, was chosen as the PDN. As depicted in Figure 4, when PAG: PDN ≥ 1:2, dense lines with high resolution (25 nm, L/S, pitch 50 nm) were successfully fabricated, and the line-edge roughness (LER) was gradually reduced (LER ≤ 3.3 nm) with increasing PDN ratio, although sensitivity was correspondingly decreased (100-270 μC cm −2 ). An empirical formula called the Z-parameter, given by the equation: resolution 3 × LER 2 × sensitivity = Z, was used to evaluate the performance of the photoresist material for these formulations. The Z-parameter values for formulations of PAG: PDN = 1:2/1:3/1:4 were 1.70 × 10 −7 , 2.72 × 10 −7 , and 3.55 × 10 −7 μC nm 3 respectively, while no data were obtained for PAG: PDN = 1:1 owing to bridging. Remarkably, the optimal formulation displayed a Z-parameter an order of magnitude greater than the commercial e-beam photoresist PMMA, with a value of 1.80 × 10 −6 μC nm 3 . [22] In order to assess the pattern improvement profile of BPA-6Ep, a formulation of PAG:PDN = 1:2, which exhibited the minimum Z-parameter, was selected to investigate the crosssectional SEM images of 40 nm L/S (pitch 80 nm) and 25 nm L/S (pitch 50 nm) (as shown in Figure 5). Notably, there were no indications of mild footing in the spatial region, and the line width (LW) provided by the cross-sectional SEM image was in good agreement with the default layout. Furthermore, comparison with the cross-sectional profile of TPS-SbF 6 without PDN ( Figure S10, Supporting Information) illustrated the synergistic effect of PDN and low-sensitivity PAG on the epoxy NTR, thereby highlighting the effectiveness of the "PDN www.advancedsciencenews.com www.advmatinterfaces.de  first, PAG second" approach in optimizing photoresist performance for the precise generation of patterns with the desired contrast.

Conclusions
In summary, we have developed a variety of PAGs with different reactivities and verified the rationality and accuracy of the design through characterization and exposure experiments, including XPS and contrast lithography. The strategy of "PDN first, PAG second," which involves PDN decomposition prior to PAG decomposition in the EBL process, has been successfully implemented in the EBL of BPA-6Ep molecular glass photoresist, leading to the generation of 25 nm L/S stripes at a dose of less than 300 μC cm −2 with LER ≤ 3.3 nm. This approach has demonstrated the effectiveness of negative photoresist optimization with the "PDN first, PAG second" system, and is presently undergoing further evaluation for its potential application in EUV lithography.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.