Optimizing SU-8 resist to fabricate micro-metallic structures
04/01/2006
Conventional precision engineering technologies are reaching their limits concerning minimum feature size and precision demands, so alternative technologies, such as LIGA (from the German acronym: lithographie, galvanik, abformung), gain importance [1]. One example is the fabrication of metallic micro-gear wheels, which are used in precision, zero backlash micro-harmonic drives [2]. In order to reach their performance, metallic components of these complex and small gear systems require the highest level of geometrical and dimensional precision, which can only be reached by means of LIGA technology (Fig. 1).
J. Kouba, M. Bednarzik, BESSY GmbH, Berlin, Germany; R. Engelke, G. Ahrens, micro resist technology GmbH, Berlin, Germany; H. Miller, MicroChem Corp., Newton, Massachusetts; D. Haase, Jenoptik Mikrotechnik GmbH, Jena, Germany
The customer-given specifications including precision requirements are given in Table 1 and clearly illustrate the need for the use of LIGA. Fabrication of the most critical parts of the gear assembly-the flex spline-poses the largest challenge, especially regarding the sidewall tolerances. The fabrication of all the components occurs by means of direct LIGA [3], which eliminates the need for x-ray lithography and improves the process stability, repeatability, and cost effectiveness.
Until recently, gear wheels, such as flex splines that are ~1mm tall, were fabricated using a PMMA (polymethyl methacrylate) process with sufficient quality and yield. However, due to the extremely long exposure times and the cost constraints of such exposures, switching to a more sensitive photopolymer, such as SU-8, would be preferred.
 |
The costs for x-ray exposure are the bottleneck of the process, and cutting them along with the associated pre- and post-processing steps would reduce single-part costs significantly. However, the conventional SU-8 resist process failed to meet the precision requirements, thus giving impetus for this study. The aim of the work was to optimize the SU-8 formulation and process to reach the quality of final micro-gear wheels similar to those made using the PMMA process.
Optimizing SU-8
SU-8, a well-known x-ray resist, has some advantages and disadvantages compared to PMMA (Table 2). The photospeed advantage of SU-8 vs. PMMA was applied to this x-ray LIGA process as shown in Fig. 2. The threshold dose of SU-8 is on the order of 300× smaller than that of PMMA, making the resist much more sensitive to x-ray radiation, while at the same time making the process window (for SU-8) much narrower than that of PMMA. Furthermore, by comparing the slope of the curves, one can see that both resists show about the same contrast (Table 3). The exposure time is ~20× shorter in the case of SU-8.
 |
null
 Figure 2. Gradation curves for SU-8 and PMMA with 1mm resist thickness and standard processing conditions. |
For the SU-8 process, 4-in. Si wafers with a Ti/Au seed layer were used as a substrate. SU-8 resists were custom prepared by MicroChem Corp. with photoacid generator (PAG) at concentrations of 2% and 6%.
The SU-8 resist layers were prepared by casting a constant volume of liquid resist on the substrate and baking it on a leveled hot plate. By precise control of the process, resist layers 1100µm in thickness and with a variation of <0.25% could be repeatedly prepared [4].
 |
By varying the softbake temperature and time between 105-130°C and 8-40 hrs, respectively, residual solvent content (the amount of solvent remaining in the resist film after softbake) was varied between 2-10%. By adjusting the softbake regimes, some variation in the solvent content distribution was also achieved [5].
X-ray exposures were generated on the BESSY dipole beamline, featuring a Jenoptik DEX02 scanner. The BESSY storage ring was operated at 1.7GeV; the radius of the dipole magnet was 4.359m and the magnetic field was 1.35T. The result was a critical energy of 2.5keV, or a critical wavelength of 0.5nm, respectively. A 200µm-thick Be window was used to separate the high vacuum region from the scanner. Typically, the ring is operated at 150-250mA.
Bottom dose was modulated between 10-40J/cm3 using various spectra filtration. A special testing mask for the process optimization was fabricated at BESSY. The mask was made using a 180µm-thick graphite substrate and 60µm gold, x-ray attenuating structures. The mask carried 10 identical fields allowing multiple exposures with different conditions on a single wafer.
For the fabrication of metallic gears, x-ray masks fabricated at the Center for Advanced Microstructures and Devices (CAMD) [6] and consisting of a 550µm-thick Be substrate and a 50µm-thick gold attenuating structure were used. Post exposure bake (PEB) was accomplished on the hot plate using temperatures from 60-95°C and times from 20-60 min.
For the PMMA process, 4-in. Si substrates with oxidized Ti as a plating base were used. High molecular weight, 2mm-thick PMMA sheets obtained from Goodfellow were glued to the substrate, using a method described in [7] and fly-cut to a final thickness of 1050µm. X-ray exposures were produced using a wavelength beamline shifter with a Jenoptik DEX01 scanner. The critical energy of the spectra was 7.69keV, which corresponded to a wavelength of 0.16nm. A 200µm-thick Be window was used to separate the high-vacuum region from the scanner. X-ray masks fabricated at CAMD consisting of a 550µm-thick Be substrate and a 50µm-thick gold attenuator mask were used. A bottom dose of 4.0kJ/cm3 was used together with a 160µm-thick graphite filter to adjust the spectra. In both cases, electroforming was performed by Micromotion GmbH using a commercially-available Kissler plating bench and a sulphite based Ni-Fe plating solution.
Sidewall bow was evaluated using a Wyko NT 1000 white light interferometer at the U. of Dortmund [8]. To evaluate the tilt of the gear wheel sidewalls, a video-based contact angle measurement method [9] was used. Top scum, resist residue, and structure collapse were visually assessed using a LEO 1560 scanning electron microscope.
Results
Figure 3 shows the typical sidewall bow of a gear wheel fabricated using the conventional SU-8 and PMMA fabrication processes. As can be seen, the amplitude of the bow when using SU-8 approaches 5µm, which is an unacceptably large dimensional deviation that would eventually disable the assembly of the gears.
 Figure 3. Sidewall bow on flex splines fabricated using SU-8 and PMMA fabrication processes. |
The unacceptably large deviation as shown in Fig. 3 was the starting point of the SU-8 process optimization. Because of the large number of processing parameters in the study of SU-8, a half-factorial, Taguchi-style design of experiments (DOE) was used to identify the significant parameters.
The amount of PAG, cast film solvent content, PEB temperature, PEB time, and bottom dose were the independent variables. The responses were top scum (also called skin effect), amount of observable residue after development of the resist, collapse of the structures, and sidewall bow. The sample evaluation and measurements were conducted only on the resist forms, and the electroplating step was eliminated. Table 4 shows a brief summary of the results.
 |
Two independent variables-solvent content and PEB temperature-were found to be significant in controlling structure collapse and sidewall bow. The interaction of these two variables was also found to be significant with the best results at a solvent content of 6% and a PEB temperature of 60°C. A third independent variable-bottom dose-was also found to be significant in affecting top scum, sidewall profile and residue, but not structure collapse.
 Figure 4. Progress of sidewall bow along the process of optimization (right) and PMMA reference measurements (left). The working range is highlighted. |
Using the results of the previous study, by further varying the significant parameters, the SU-8 process was optimized with respect to the sidewall bow. Figure 4 shows the evolution of the sidewall profile along the time scale of the SU-8 process optimization. Each point on the graph represents the average value of the amplitude of all the measurements of metallic flex splines produced under the same conditions. The error bars represent the maximum and minimum of all measured values. The left part of the chart in Fig. 4 shows the best and the worst corresponding values found on metallic flex splines fabricated using the PMMA process as described above. In these two cases, the assembled gear wheels were fully functional and exceeded their required lifetime, so that for this study, these two cases represented the working range for the sidewall bow.
The nonoptimized, conventional SU-8 process was not satisfactory, as amplitudes of sidewall bow approached 12µm. After optimizing the SU-8 process, both the average values and the variation decreased, so that at the end, the amplitudes of the bowing of the SU-8 based fabricated flex splines reached those of flex splines fabricated using the PMMA process.
The most significant parameters were found to be the solvent content, the PEB bake temperature, and the bottom dose. It was also found that the distribution of solvent content in the resist film plays an important role and that some specific softbake procedures yielded better results.
Figure 5 depicts the typical sidewall bow of a metallic flex spline that was fabricated using an optimized SU-8 process compared to the corresponding sidewall bow in the case of the non-optimized SU-8 process and the PMMA process, respectively. The significant improvement for the optimized SU-8 process can be clearly observed.
The overall sidewall profile consists both of the sidewall bow and the sidewall tilt. Using white light interferometry, only the bow can be measured. Using the contact angle measurement method, the sidewall tilt was estimated using several flex splines that were fabricated with the optimized SU-8 process and then compared to the sidewall tilt of flex splines fabricated using the PMMA process. The results of this measurement are shown in Fig. 6. The standard deviation of the sidewall tilt was estimated to be 3.5° in the case of the optimized SU-8 process and 3.9° in the case of the PMMA process.
 Figure 6. Sidewall tilt measured on metallic flex splines fabricated using a) optimized SU-8 process; and b) standard PMMA process. |
Comparing the standard deviation values, one can see that the sidewall tilt is comparable. The overall large value of the tilt of ~4° is believed to result from the electroforming process and is currently under investigation. As in the case of sidewall bow, the flex splines that were investigated and that were fabricated using the PMMA process were all well-working and acceptable parts, so that these high values of tilt obviously do not disturb the application.
Conclusion
An SU-8 process for the fabrication of precise metallic gear wheels with heights of 1mm and a smallest feature size of 25µm, along with aspect ratios >40 and overall very tight tolerances, was established. Starting from a nonoptimized SU-8 process, and carefully using design of experiments (DOE) methodology, significant process parameters for process optimization were identified. The remaining solvent concentration, PEB temperature, and bottom dose were identified to be among the most significant parameters. By varying significant parameters, the SU-8 process was optimized with respect to the sidewall bow. Among optimal conditions, 6% of the remaining solvent content and PEB temperature of 60°C were identified.
Using the optimized SU-8 process, the sidewall bow of the electroplated gear wheels was found to be comparable to that of the electroplated gear wheels that were fabricated using the PMMA process. Furthermore, the sidewall tilt of metallic gear wheels fabricated using the optimized SU-8 process was found to be comparable to that of the gear wheels fabricated using the PMMA process (Table 5). SU-8 was proven to be well suited for the particular application.
 |
The authors of this study see further potential in the optimization of the whole fabrication process, both on the resist processing side, as well as on the electroplating side. Further studies and optimization processes are being conducted.
Acknowledgments
This work makes broad use of the contents of the presentation, “Comparative Study of the Sidewall Profile of PMMA and SU-8 Moulds Made by UDXRL and of Electroformed Metallic Counterparts,” from the High Aspect Ratio Micro Structure Technology (HARMST) Workshop 2005, Gyeongyu Republic of Korea, June 10-13, 2005.
Heinz-Ulrich Scheunemann and B. Loechel, BESSY GmbH, Berlin, and G. Gruetzner, micro resist technology GmbH, Berlin, were co-authors of this article. The authors would like to thank all participants and contributors to this work, especially co-workers from BESSY GmbH, micro resist technology GmbH, Micromotion GmbH, CAMD, and the U. of Dortmund. This research was partially supported by the initiative of the Federal Ministry of Education (BMBF), Berlin.
References
- W. Menz, J. Mohr, Mikrosystemtechnik fuer Ingenieure, VCH Verlaggesselschaft mbH, Weinheim, 1997.
- R. Degen, R. Slatter, “Zero Backlash Micro-Gears and Actuators for Microassembly Applications,” 2004, www.harmonicdrive.de/de/pdf/fachauf_18.pdf.
- B. Loechel, L. Jian, H.-U. Scheunemann, D. Schondelmaier, J. Goettert, Y.M. Desta, et al., “Providing a Direct-LIGA Service: A Status Report,” COMS 2002, Ypsilanti, MI, 2002.
- G. Gruetzner, R. Ruhmann, G. Ahrens, M. Bednarzik, B. Loechel, P. Limbecker, et al., “Improved Process Parameters and Impact on Structure Quality of High-aspect-ratio SU-8 Microstrustructures,” HARMST 2003.
- R. Ruhmann, G. Ahrens, G. Gruetzner, N.A.-Staufenbiel, H. Schroeder, “Residual Solvent Content Distribution in Ultra-thick SU-8 Films and its Influence on the Imagin Quality,” HARMST 2005.
- http//www.camd.lsu.edu.
- S. Achenbach, “Optimierung zur Herstellung von Mikrostrukturen durch Ultratiefe Röntgenlithographie (UDXRL),” PhD Thesis, U. Karlsruhe, 2000.
- http://www-mst.e-technik.uni-dortmund.de/.
- R. Engelke, G. Ahrens, R. Ruhmann, S. Kopetz, J. Kastner, K. Wiesauer, et al., “Possibilities of Inline Process Inspection of High Aspect Ratio LIGA Micro Structures,” HARMST 2005.
For more information, contact Josef Kouba at the Application Center for Miroengineering at BESSY GmbH, Albert-Einstein-Str. 15, 12489 Berlin, Germany; ph 49/30 6392 3125, fax 49/30 6392 4709, email [email protected].