Two-step resist-development process of hydrogen silsesquioxane for high-density electron-beam nanopatterning

Transcription

1 Two-step resist-development process of hydrogen silsesquioxane for high-density electron-beam nanopatterning Hyo-Sung Lee, Jung-Sub Wi, Sung-Wook Nam, Hyun-Mi Kim, and Ki-Bum Kim a Department of Materials Science and Engineering, Seoul National University, 599 Gwanangno, Gwanak-gu, Seoul , Korea Received 15 August 2008; accepted 24 November 2008; published 4 February 2009 It is identified that the development of hydrogen-silsesquioxane resist after electron-beam exposure, by using a 25% tetramethylammonium-hydroxide TMAH developer, almost stops after 1 min of development time and it severely limits the delineation of high-density nanometer-scale patterns. By using x-ray photoelectron spectroscopy, the authors identified that this development-stopping phenomenon is due to the formation of a siloxane-type bond structure which is insoluble to the TMAH developer. Here, the authors propose a two-step development method that involves the removal process of siloxane layer using a dilute hydrofluoric acid between development processes. This method successfully eliminates the insoluble layer, thus generating isolated high-density dot patterns with 25 nm pitch American Vacuum Society. DOI: / I. INTRODUCTION The formation of high-density nanometer-scale features has been one of the critical process requirements for fabricating highly functional nanometer-scale devices in electronic, photonic, magnetic, and biological applications. 1 4 While there have been many processing schemes proposed, including both top-down and bottom-up approaches, electron-beam lithography is still considered as one of the most reliable and reproducible techniques to generate such features. In particular, the hydrogen-silsesquioxane HSQ electron-beam e-beam resist has been widely employed to define extremely small size patterns. For instance, it was reported that one can pattern even a sub-10-nm-size line feature by using this e-beam resist. 5 However, it should be noted that many of these results were obtained in low-density patterns where there is little reason to be concerned about the proximity effect. In a high-density pattern with a 1:1 ratio of feature size and spacing, the minimum feature size is still limited to a few tens of nanometers. 6,7 Obviously, this difference originates from the proximity effect. Thus, in order to overcome the proximity effect and to define high-density nanometer-scale patterns successfully in the resist point of view, it is essential to propose a new resist or a resistdevelopment process to increase the contrast value. Indeed, there have been several attempts to increase the contrast value of the HSQ resist, such as employing a prebake at room temperature, 8 development at high temperature, 8 a more concentrated developer, 9 and a salty developer. 10 It has also been reported that one can simply get higher contrast value by increasing the development time. For instance, in the case of the HSQ resist, the contrast value increases from 4.6 to 6.9 by simply extending the development time from 30 to 60 s using a 25% tetramethylammoniumhydroxide TMAH developer 11 and, in the case of the 2.38% TMAH developer, the contrast value shows a tendency to a Electronic mail: kibum@snu.ac.kr increase during 240 s of development time. 6,12 While both of these results clearly indicate that one can somewhat increase the contrast value of the resist by simply extending the development time, it is not at all clear whether this trend can continue. In this paper, we carefully investigate the effect of development time on development characteristics of the HSQ and propose a new development method which allows patterning of high-density isolated features. We found that HSQ resist development almost stops after 1 min in a 25% TMAH developer and identified that this phenomenon results from the formation of a TMAH-insoluble, siloxanelike surface layer on HSQ during the development process. We further noted that this layer can be removed by using a dilute hydrofluoricacid HF solution and, after removing this layer, one can continue the development process by using a TMAH solution. II. EXPERIMENTS HSQ FOx-12 purchased from Dow Corning Co. was spin-coated on a Si wafer to a thickness of 100 nm and prebaked for 1 min at 90 C on a hot plate. The e-beam exposure for contrast-curve measurement was carried out using a Leica Lion LV1 system operating at an acceleration voltage of 19 kv. After e-beam exposure, the HSQ resist was developed at various times 10 s, 30 s, 1 min, 3 min, and 5 min at 21 C using a 25 wt % TMAH aqueous solution. Then the resist was rinsed in de-ionized water for 2 min and dried by N 2 gas. Basically the processing scheme follows the one proposed by Henschel et al. 9 known to obtain the highest HSQ contrast value. The remaining resist thickness was measured by an optical-measurement system Nanospec AFT/4150, Nanometrics and the contrast values were extracted from the slope by the linear fitting between 0 and 0.7 of the normalized remaining thickness. As was mentioned in the Introduction, during the course of resist development by using TMAH, we identified that the resist development stops after about 1 min. Thus, in order to 188 J. Vac. Sci. Technol. B 27 1, Jan/Feb /2009/27 1 /188/5/$ American Vacuum Society 188

2 189 Lee et al.: Two-step resist-development process of hydrogen 189 characterize the structure of this TMAH-insoluble layer, we measured the Si spectra using the x-ray photoelectron spectroscopy XPS on the HSQ surface in the following four cases: 1 after HSQ spin-coating and the prebake step, 2 after the e-beam exposure step, 3 after the TMAH development step for 1 min, and 4 after the dilute HF dipping step for 1 min. The measurements were performed using a Thermo VG Scientific Sigma Probe spectrometer with monochromatic Al K at 15 kv. For these samples, e-beam exposure was carried out at a dose of 280 C/cm 2 using a Leica EBMF 10.5 system with 30 kv exposure energy over a range of 1 3 mm 2 due to the large x-ray beam size of about 400 m diameter. We have chosen this exposure dose to develop the HSQ resist partially in the TMAH developer to find out the mechanism of the development-stopping phenomena. In that exposure dose, the remaining HSQ thickness was about 65 nm after the first 1 min development step. Based on the XPS measurement results, we proposed a two-step development method which was carried out as follows: 1 the first development step, 2 the removal step of the siloxanelike layer, and 3 the second development step. In the first and second development steps the HSQ resist was developed for 1 min in the 25% TMAH solution. In between the first and second development steps, which we call the siloxanelike layer removal step, the HSQ resist was dipped in dilute HF 4000:1 water/hf solution for 1 min. An etch rate of about 2 nm/min was achieved for the HSQ resist in this dilute HF dip. The effect of the proposed two-step development TMAH-HF-TMAH method was compared with that of the typical one-step development TMAH, based on the results of contrast-curve measurement and the e-beam patterning of the dot array. Dot-array patterns with a large pitch size of 120 nm were exposed on 100-nm-thick HSQ resist using a Leica LION LV1 system at 19 kv. For high-resolution dotarray nanopatterning with 25 nm pitch size, we used a JEOL JBX-9300FS e-beam lithography system with 100 kv exposure energy and the 30-nm-thick HSQ resist. The patterned samples were inspected using a JEOL JSM-7401F scanning electron microscope SEM. III. RESULTS AND DISCUSSION A. Development-time effect Figure 1 shows the results of the contrast curve by varying the development time from 10 s to 5 min in the 25% TMAH developer. This figure clearly shows that the development continues when increasing the development time from 10 s to 1 min and the contrast value also increases from 3.8 to 5.2, accordingly, as was reported by several researchers Interestingly enough, however, we identified that the development rate gradually decreases with the development time and, after about 1 min, the development process almost stops. We do not see any further development of the resist by extending the development time up to 5 min. This phenomenon of development stop can possibly be explained by two effects. The first effect is to assume that the FIG. 1. Contrast curves of HSQ resist at various development times 10 s, 30 s, 1 min, 3 min, and 5 min using a 25% TMAH developer in 19 kv e-beam exposure. excessive energy is deposited inside the resist during the e-beam exposure step bulk effect. The second effect is to assume that the TMAH-insoluble layer is formed on the resist surface during the development step surface effect. In order to check the first assumption, we obtained the deposited-energy distribution through the thickness of the resist layer by using the Monte Carlo simulation. 13 While our simulation result under the condition of 19 kv e-beam energy for a 100-nm-thick HSQ resist shows that the deposited energy at 90 nm depth is about 1.15 times higher than the energy at 10 nm depth, this energy difference is not large enough to explain the development-stopping phenomenon. Furthermore, since the development stop occurred in the entire range of HSQ thickness, it is not reasonable to assume that the development stop resulted from the increase in deposited energy inside the resist layer. In order to check the possibility of the formation of a TMAH-insoluble layer at the surface, we analyzed the chemical structure of the HSQ surface by using XPS. Figure 2 presents the high-resolution Si 2p XPS spectra for HSQ surfaces after the processes of a spin-coating and prebake, b e-beam exposure, c TMAH development, and d HF treatment. The spectra for all the HSQ surfaces show two peaks: the peak of high binding energy ev and the peak of low binding energy ev. These high and low binding energies correspond to the previously reported values for Si 4+ and Si 3+ states of SiO x film. 14,15 Namely, the Si 4+ signal comes from Si, which formed a network structure with strong siloxane Si O Si bonds, while the Si 3+ signal comes from Si, which formed a cage structure with weak Si H bonds. The Si 2p spectra obtained from the sample right after spin-coating and prebake show a strong Si 3+ peak intensity Fig. 2 a. This indicates that the HSQ resist after this step mainly consists of a cage structure with Si H bonds. After e-beam exposure, it is noted that the Si 4+ peak intensity increases even though the Si 3+ peak intensity still dominates, as is shown in Fig. 2 b. This certainly indi- JVST B-Microelectronics and Nanometer Structures

3 190 Lee et al.: Two-step resist-development process of hydrogen 190 FIG. 2. High-resolution Si 2p XPS spectra of HSQ surface after a spincoating and prebake, b e-beam exposure, c TMAH development for 1 min, and d dilute HF treatment for 1 min. FIG. 3. a Contrast curves of HSQ resist for TMAH, TMAH/HF, and TMAH/HF/TMAH development and b HSQ development rate for the first and second TMAH developments and HF etch rate. cates that the fraction of the Si network structure Si 4+ peak intensity is increased by e-beam exposure, which is not developed by TMAH. This is considered as one of the main mechanisms of how we can define e-beam patterns by using HSQ. 16 Interestingly enough, the spectra after 1 min of TMAH development show a higher Si 4+ and a lower Si 3+ peak intensity than those before the development, as is shown in Fig. 2 c. This either means that TMAH selectively dissolves Si 3+ type elements in the HSQ layer when there is a mixture of Si 3+ and Si 4+ elements or, as was reported by Schmid et al., 12 the TMAH solution converts the silicon hydride group into the Si O Si cross-linked structure via a condensation reaction. In any case, it appears that further development stops when there are enough Si 4+ elements built up at the surface. Certainly, the time for development to stop is determined by the development rate. In our experimental condition with a high concentration of 25% TMAH developer, the development stops after about 1 min, and this time will be longer in the case of using a lower TMAH concentration. B. Two-step development method It thus appears that removing this insoluble layer formed at the surface is essential for continuing the development process. Here, we tried a dilute HF solution treatment to remove this layer, since this layer is composed of a network structure with a strong siloxane Si O Si bond. As shown in Fig. 2 d, the Si 4+ peak intensity decreases after removing the 2-nm-thick siloxane layer and the overall peak appearance is similar to that before the development process. This result clearly indicates that the insoluble layer is effectively removed by the HF-dipping process. Figure 3 a shows the HSQ contrast curves obtained from the one-step TMAH, one-step and HF dip TMAH-HF, and two-step TMAH-HF-TMAH development cases, respectively. This result shows that the development process can be continued by using the proposed two-step development method. It also shows that HF dipping does not remove the resist layer appreciably. By using these contrast curves, the development rate of each case was extracted in Fig. 3 b. J. Vac. Sci. Technol. B, Vol. 27, No. 1, Jan/Feb 2009

4 191 Lee et al.: Two-step resist-development process of hydrogen 191 FIG. 5. Plan-view SEM images of e-beam patterning results for dot array with 25 nm pitch size for various e-beam exposure doses. a One- and b two-step developments after 13.75, 15, 16.25, and 17.5 fc/dot exposure. FIG. 4. Plan-view SEM images of e-beam patterning results for dot array with 120 nm pitch. a One- and b two-step developments after 28 fc/dot exposure and c one- and d two-step developments after 37 fc/dot exposure. The insets in c and d show the contrast-enhanced images at higher magnification. These data show that the development rates of the first and second TMAH treatments are almost the same, indicating that the removal of the insoluble layer effectively converts the surface layer to that before development. This means that one can now continue the development up to 2 min. Certainly, it is possible to extend the development time more than 2 min by simply alternating the dilute HF treatment and TMAH development several times when it is necessary. The effect of the two-step development method, as compared to that of the typical one-step method, is first demonstrated by patterning a dot-array pattern with a relatively large size of pitch 120 nm. Figure 4 illustrates the planview SEM images of the resist features obtained at different exposure doses and developed by applying the aforementioned two different methods. At the exposure dose of 28 fc/dot, the isolated features with 40 nm diameter were formed after the one-step development, whereas after the two-step development, the dot size was reduced to 20 nm, as shown in Figs. 4 a and 4 b. This result demonstrates that one can further reduce the feature size by simply extending the development time. The more drastic improvement effect of the two-step development process can be observed at a higher dose. At the exposure dose of 37 fc/dot, the dot-array features were not fully isolated after the one-step development by the formation of resist scum between the dots, as is shown in Fig. 4 c. Again, this resist scum was cleanly eliminated by the two-step development method, as is shown in Fig. 4 d. Apparently, as the dose level is increased, there builds up an appreciable electron dose between the dots by the proximity effect. This resist layer is now not cleanly developed by using a one-step development process due to the formation of a TMAH-insoluble layer at the surface. The effect of the proposed two-step development process is more drastic in patterning high-density features with small-size pitches. Figure 5 a shows SEM images of the resist features patterned with a 25 nm pitch for various exposure doses after one-step development. At the low dose of fc/dot, the fully isolated resist features were formed without any resist residue. However, it was observed that several dot features collapsed due to the insufficient dose. For successful formation of dot-array features, a higher e-beam dose is necessary. However, when the exposure dose was increased above 15 fc/dot, the dot-array patterns were not fully isolated due to the formation of a residual resist between dot features. Again, by using two-step development, these connected features were fully isolated up to a high exposure dose of 17.5 fc/dot, as is shown in Fig. 5 b. Asa result, we can fabricate the resist features with a 25 nm pitch in the wide range of exposure doses. All of these results show that the two-step method is effective in generating isolated patterns in high density. IV. CONCLUSION It is identified that the HSQ resist development, by using TMAH solution, stops after a certain period of time due to the formation of a TMAH-insoluble, siloxanelike layer at the surface and that this layer is successfully removed by using a dilute HF dip. Furthermore, once this layer is removed, one can continue the resist development up to a certain period of time. Based on these results, we are proposing a two-step development process which includes a brief diluted HFdipping process in between the normal TMAH development processes. In our dilute HF-dipping step, a 2-nm-thick siloxane layer was removed and the HSQ development was maintained up to 2 min. It is shown that the proposed two-step development method cleanly removes resist scum between the patterns and, thus, is effective in generating isolated high-density dot-array patterns. ACKNOWLEDGMENTS This work was supported through the Frontier Research Program of Tera-Level Nanodevices TND funded by the Korean Ministry of Science and Technology MOST. The JVST B-Microelectronics and Nanometer Structures

5 192 Lee et al.: Two-step resist-development process of hydrogen 192 authors thank Professor Kukjin Chun and Jinkwang Kim from Seoul National University for supporting the e-beam lithography simulator and Dr. Jehyuk Choi from Korea Advanced Nano Fab Center for e-beam patterning at 100 kv. 1 D. Wang, B. Sheriff, and J. R. Heath, Nano Lett. 6, W. Huang, X. Duan, and C. M. Lieber, Small 1, S. Sun, C. B. Murray, D. Weller, L. Folks, and A. Moser, Science 287, Z. Li, Y. Chen, T. I. Kamins, K. Nauka, and R. S. Williams, Nano Lett. 4, H. Namatsu, J. Vac. Sci. Technol. B 19, F. C. M. J. M. van Delft, J. P. Weterings, A. K. Van Langen-Suusling, and H. Romijin, J. Vac. Sci. Technol. B 18, M. J. Word, I. Adesida, and P. R. Berger, J. Vac. Sci. Technol. B 21, L M. Häffner, A. Haug, A. Heeren, M. Fleischer, H. Peisert, T. Chasse, and D. P. Kern, J. Vac. Sci. Technol. B 25, W. Henschel, Y. M. Georgiev, and H. Kurz, J. Vac. Sci. Technol. B 21, J. K. W. Yang and K. K. Berggren, J. Vac. Sci. Technol. B 25, F. Fruleux-Cornu, J. Penaud, E. Dubois, M. Francois, and M. Muller, Mater. Sci. Eng., C 26, G. M. Schmid, L. E. Carpenter II, and J. A. Liddle, J. Vac. Sci. Technol. B 22, Monte Carlo simulation was performed using an electron-beam lithography simulator, which was developed by Seoul National University. 14 F. G. Bell and L. Ley, Phys. Rev. B 37, S. Iwata and A. Ishizaka, J. Appl. Phys. 79, H. Namatsu, Y. Takahashi, K. Yamazaki, T. Yamaguchi, M. Nagase, and K. Kurihara, J. Vac. Sci. Technol. B 16, J. Vac. Sci. Technol. B, Vol. 27, No. 1, Jan/Feb 2009