- Email: [email protected]
Etch rate optimization in reactive ion etching of epoxy photoresists
Procedia Chemistry Procedia Chemistry 1 (2009) 796–799 www.elsevier.com/locate/procedia
Proceedings of the Eurosensors XXIII conference
Etch rate optimization in reactive ion etching of epoxy photoresists M. Driesen*, K. Wouters, R. Puers ESAT/MICAS, Katholieke Universiteit Leuven, Belgium
Abstract In this paper, the influence of several reactive ion etching (RIE) process parameters on epoxy photoresist etch rates is compared and optimized. Four types of epoxies were examined: SU-8 2050, EpoCore, hardbaked and non-hardbaked EpoClad. The parameters that were tested are: O2 flow, SF6 flow, chamber pressure and RF power. For the experiments, a Box-Behnken design of experiment (DOE) [1] was selected, allowing us to examine 4 different non-linear parameters with 25 experiments. The collected data allows to optimize an etching process, taking into account compatibility with other materials and required etch rates. Keywords: Reactive ion etching, epoxy, photoresist, SU-8, EpoClad, EpoCore
1. Introduction Photosensitive epoxies have become essential in many MEMS devices. The best-known is probably SU-8, developed by IBM, with applications in microfluidics [2] and electroplating [3]. Other, more recent materials are EpoClad and EpoCore, originally developed for use in plastic optics [4]. A well-known challenge in using these photoresists is the difficult removal, e.g. when used as an electroplating mold. Cured epoxies are strongly crosslinked and chemically very stable. They can be removed by oxidizing mixtures, such as piranha (3-7:1 H2SO4:H2O2), but these will also attack many other materials, like metals. Another option, chosen here, is RIE with oxygen (O2) and sulfur hexafluoride (SF6). 2. Experiments
2.1. Setup A non-linear DOE was chosen because previous RIE results on SU-8 have shown non-linear dependencies [5]. The three tested levels of each parameter are shown in table 1.
* Corresponding author. Tel.: +32-16-321093; fax: +32-16-321975 E-mail address: [email protected]
1876-6196/09/$– See front matter © 2009 Published by Elsevier B.V. doi:10.1016/j.proche.2009.07.198
M. Driesen et al. / Procedia Chemistry 1 (2009) 796– 799
797
Table 1. Tested levels of all process parameters. Level
O2 flow [sccm]
SF6 flow [sccm]
Pressure [mTorr]
RF Power [W]
-1
10
0
50
50
0
30
2
100
100
1
50
4
150
150
2.2. Sample preparation For testing, silicon wafers were prepared with a thin film of sputtered copper (Cu) on a titanium (Ti) adhesion layer. Cu does not etch in the type of plasma used [6], and protects the bare silicon during the experiments. On top of this film, the photoresist layers were spun and patterned with holes. The process used is the standard process as advised by the manufacturer (MicroChem Corp. for SU-8, Microresist GmbH for EpoClad and EpoCore). SU-8 and EpoCore have large internal stresses, resulting in microcracks after processing. Therefore, a hardbake is necessary. EpoClad has fewer tendencies to crack, so it can be used without hardbake. This should result in less crosslinking, and easier removal. The layers were then etched for 5 minutes in a Plasma Technology 80-series etcher, at 20 °C. A different etcher configuration might yield different results. The full process is shown in Figure 1.
Figure 1. Process steps: (1) bare silicon wafer, (2) Ti-Cu deposition, (3) photoresist spinning, (4) patterning square holes, (5) RIE etch.
Before and after etching, the thickness of the layer was measured in three places on the wafer with a Dektak profile meter. 3. Results Figure 2 and Figure 3 show the results of the experiments. Only first-order effects are shown. The following individual observations can be made: • SF6 flow: the largest influence can be seen with non-hardbaked EpoClad and EpoCore. SU-8 is insensitive to higher flows; the other epoxies showed an increase in etch rates for increased flows. • Oxygen flow: a highly non-linear response is seen, with the fastest average etch rates at medium flow rates. The exception is non-hardbaked EpoClad, with a constant decrease in etch rates for increasing flows. • RF power: all four epoxies react similar. An increase in RF power results in a higher etch rate. The influence is nearly linear. • Chamber pressure: the effect on hardbaked EpoClad is the largest. When increasing the pressure from 50 to 150 mTorr, the average etch rate is 8 times as high. For the other epoxies, the influence was very limited. As predicted, non-hardbaked EpoClad has on average a higher etch rate than hardbaked EpoClad (respectively 0.28 µm/min and 0.22 µm/min on average over all runs). This makes it an option for applications that do not need the strength of a highly crosslinked polymer, e.g. an electroplating mold.
798
M. Driesen et al. / Procedia Chemistry 1 (2009) 796– 799
Figure 2. (a) Influence of SF6 flow on etch rate. (b) Influence of O2 flow on etch rate.
Figure 3. (a) Influence of RF power on etch rate. (b) Influence of chamber pressure on etch rate.
Figure 4 shows two outspoken second-order effects. Firstly, at 50 mTorr pressure, the RF power dependence is very different, with etch speeds lowest at medium power levels (100 W). Secondly, at high SF6 flows (4 sccm), the etch rate becomes more dependent on the chamber pressure, especially for both EpoClad varieties.
Figure 4. (a) Influence of RF power with 50 mTorr chamber pressure. (b) Influence of chamber pressure with 4sccm SF6 flow.
799
M. Driesen et al. / Procedia Chemistry 1 (2009) 796– 799
Another effect was observed: an increase in internal stress in the etched SU-8 layers, see table 2. The parameters for that run were: 100 W RF power, 10 sccm O2 flow, 4 sccm SF6 flow and 50 mTorr chamber pressure. Etch rate was nearly zero with these parameters. A similar result was reported before [7], but with an initial reversal to compressive stress. This difference can be explained by different ways of measuring a mean stress (buckling beams vs. Vernier scale indicators [8]), while in reality a stress gradient is present because only the surface is altered. Table 2. Internal stress evolution of SU-8. All stresses are tensile.
Etching time [min] Stress [MPa]
0 8.4
1 13.4
3 18.9
6 29.6
4. Conclusions RIE etch rates of four epoxy photoresists were examined under varying conditions of chamber pressure, RF power and gas flow. Aside from the observation that higher RF power results in higher etch rates for all tested epoxies, the following general conclusions can be drawn for each of them: • Non-hardbaked EpoClad: a rise in SF6 flow will increase etch rates, rising O2 flow will decrease it. Chamber pressure has little influence. • Hardbaked EpoClad: etch rates increase with increasing chamber pressure and O2 flow rate. The influence of O2 flow rate levels off after 30 sccm. SF6 has little influence. • EpoCore: SF6 flow and etch rate are directly proportional. Chamber pressure has little influence. The effect of O2 flow is highly non-linear, showing high etch rates at medium flows (30 sccm). • SU-8 2050: except for RF power, the only clear effect is seen in the O2 flow rate, with again higher etch rates at medium flows. These results can now be used to optimize an etching process. Examples are: high etch rates for use as a mold for electroplating; low etch rates for use as an etch mask for e.g. silicon or glass etching; etching with low SF6 flow for compatibility with glass or silicon substrates.
Acknowledgements This work was carried out in the frame of SBO projects 060830, HyperCool-IT and 060046, Gemini, sponsored by the “Instituut voor de aanmoediging van Innovatie door Wetenschap en Technologie in Vlaanderen” (IWT), Belgium.
References 1.
M. Evans. Optimisation of Manufacturing Processes. London: Maney Publishing; 2003.
2.
A. Ezkerra, L.J. Fernández, K. Mayora, J.M. Ruano-López. Fabrication of SU-8 free-standing structures embedded in microchannels for
3.
H. Lorenz, M. Despont, N. Fahrni, N. LaBianca, P. Renaud, P. Vettiger. SU-8: a low-cost negative resist for MEMS. J Micromech.
4.
F. Ceyssens, M. Driesen, K. Wouters, R. Puers. A low-cost and highly integrated fiber optical pressure sensor system. Sensors and
microfluidic control. J Micromech Microeng (2007); 17: 2264-2271. Microeng (1997); 7: 121-124. Actuators A (2008); 145: 81-86. 5.
G. Hong, A.S. Holmes, M.E. Heaton. SU-8 resist plasma etching and its optimization. Microsystem Technologies (2004); 10: 357-359.
6.
K.R. Williams, K. Gupta, M. Wasilik. Etch rates for micromachining processing – part II. J Microelectromech Syst (2003); 12: 761-778.
7.
J.B. Bureau, B. Legrand, D. Collard, L. Buchaillot. 3D self-assembling of SU-8 microstructures on silicon by plasma induced
8.
L. Elbrecht, U. Storm, R. Catanescu, J. Binder. Comparison of stress measurement techniques in surface micromachining. J.
compressive stress. Transducers ’05, Digest of Technical Papers (2005); 1: 19-22. Micromech. Microeng (1997). 7: 151-154.
Proceedings of the Eurosensors XXIII conference
Etch rate optimization in reactive ion etching of epoxy photoresists M. Driesen*, K. Wouters, R. Puers ESAT/MICAS, Katholieke Universiteit Leuven, Belgium
Abstract In this paper, the influence of several reactive ion etching (RIE) process parameters on epoxy photoresist etch rates is compared and optimized. Four types of epoxies were examined: SU-8 2050, EpoCore, hardbaked and non-hardbaked EpoClad. The parameters that were tested are: O2 flow, SF6 flow, chamber pressure and RF power. For the experiments, a Box-Behnken design of experiment (DOE) [1] was selected, allowing us to examine 4 different non-linear parameters with 25 experiments. The collected data allows to optimize an etching process, taking into account compatibility with other materials and required etch rates. Keywords: Reactive ion etching, epoxy, photoresist, SU-8, EpoClad, EpoCore
1. Introduction Photosensitive epoxies have become essential in many MEMS devices. The best-known is probably SU-8, developed by IBM, with applications in microfluidics [2] and electroplating [3]. Other, more recent materials are EpoClad and EpoCore, originally developed for use in plastic optics [4]. A well-known challenge in using these photoresists is the difficult removal, e.g. when used as an electroplating mold. Cured epoxies are strongly crosslinked and chemically very stable. They can be removed by oxidizing mixtures, such as piranha (3-7:1 H2SO4:H2O2), but these will also attack many other materials, like metals. Another option, chosen here, is RIE with oxygen (O2) and sulfur hexafluoride (SF6). 2. Experiments
2.1. Setup A non-linear DOE was chosen because previous RIE results on SU-8 have shown non-linear dependencies [5]. The three tested levels of each parameter are shown in table 1.
* Corresponding author. Tel.: +32-16-321093; fax: +32-16-321975 E-mail address: [email protected]
1876-6196/09/$– See front matter © 2009 Published by Elsevier B.V. doi:10.1016/j.proche.2009.07.198
M. Driesen et al. / Procedia Chemistry 1 (2009) 796– 799
797
Table 1. Tested levels of all process parameters. Level
O2 flow [sccm]
SF6 flow [sccm]
Pressure [mTorr]
RF Power [W]
-1
10
0
50
50
0
30
2
100
100
1
50
4
150
150
2.2. Sample preparation For testing, silicon wafers were prepared with a thin film of sputtered copper (Cu) on a titanium (Ti) adhesion layer. Cu does not etch in the type of plasma used [6], and protects the bare silicon during the experiments. On top of this film, the photoresist layers were spun and patterned with holes. The process used is the standard process as advised by the manufacturer (MicroChem Corp. for SU-8, Microresist GmbH for EpoClad and EpoCore). SU-8 and EpoCore have large internal stresses, resulting in microcracks after processing. Therefore, a hardbake is necessary. EpoClad has fewer tendencies to crack, so it can be used without hardbake. This should result in less crosslinking, and easier removal. The layers were then etched for 5 minutes in a Plasma Technology 80-series etcher, at 20 °C. A different etcher configuration might yield different results. The full process is shown in Figure 1.
Figure 1. Process steps: (1) bare silicon wafer, (2) Ti-Cu deposition, (3) photoresist spinning, (4) patterning square holes, (5) RIE etch.
Before and after etching, the thickness of the layer was measured in three places on the wafer with a Dektak profile meter. 3. Results Figure 2 and Figure 3 show the results of the experiments. Only first-order effects are shown. The following individual observations can be made: • SF6 flow: the largest influence can be seen with non-hardbaked EpoClad and EpoCore. SU-8 is insensitive to higher flows; the other epoxies showed an increase in etch rates for increased flows. • Oxygen flow: a highly non-linear response is seen, with the fastest average etch rates at medium flow rates. The exception is non-hardbaked EpoClad, with a constant decrease in etch rates for increasing flows. • RF power: all four epoxies react similar. An increase in RF power results in a higher etch rate. The influence is nearly linear. • Chamber pressure: the effect on hardbaked EpoClad is the largest. When increasing the pressure from 50 to 150 mTorr, the average etch rate is 8 times as high. For the other epoxies, the influence was very limited. As predicted, non-hardbaked EpoClad has on average a higher etch rate than hardbaked EpoClad (respectively 0.28 µm/min and 0.22 µm/min on average over all runs). This makes it an option for applications that do not need the strength of a highly crosslinked polymer, e.g. an electroplating mold.
798
M. Driesen et al. / Procedia Chemistry 1 (2009) 796– 799
Figure 2. (a) Influence of SF6 flow on etch rate. (b) Influence of O2 flow on etch rate.
Figure 3. (a) Influence of RF power on etch rate. (b) Influence of chamber pressure on etch rate.
Figure 4 shows two outspoken second-order effects. Firstly, at 50 mTorr pressure, the RF power dependence is very different, with etch speeds lowest at medium power levels (100 W). Secondly, at high SF6 flows (4 sccm), the etch rate becomes more dependent on the chamber pressure, especially for both EpoClad varieties.
Figure 4. (a) Influence of RF power with 50 mTorr chamber pressure. (b) Influence of chamber pressure with 4sccm SF6 flow.
799
M. Driesen et al. / Procedia Chemistry 1 (2009) 796– 799
Another effect was observed: an increase in internal stress in the etched SU-8 layers, see table 2. The parameters for that run were: 100 W RF power, 10 sccm O2 flow, 4 sccm SF6 flow and 50 mTorr chamber pressure. Etch rate was nearly zero with these parameters. A similar result was reported before [7], but with an initial reversal to compressive stress. This difference can be explained by different ways of measuring a mean stress (buckling beams vs. Vernier scale indicators [8]), while in reality a stress gradient is present because only the surface is altered. Table 2. Internal stress evolution of SU-8. All stresses are tensile.
Etching time [min] Stress [MPa]
0 8.4
1 13.4
3 18.9
6 29.6
4. Conclusions RIE etch rates of four epoxy photoresists were examined under varying conditions of chamber pressure, RF power and gas flow. Aside from the observation that higher RF power results in higher etch rates for all tested epoxies, the following general conclusions can be drawn for each of them: • Non-hardbaked EpoClad: a rise in SF6 flow will increase etch rates, rising O2 flow will decrease it. Chamber pressure has little influence. • Hardbaked EpoClad: etch rates increase with increasing chamber pressure and O2 flow rate. The influence of O2 flow rate levels off after 30 sccm. SF6 has little influence. • EpoCore: SF6 flow and etch rate are directly proportional. Chamber pressure has little influence. The effect of O2 flow is highly non-linear, showing high etch rates at medium flows (30 sccm). • SU-8 2050: except for RF power, the only clear effect is seen in the O2 flow rate, with again higher etch rates at medium flows. These results can now be used to optimize an etching process. Examples are: high etch rates for use as a mold for electroplating; low etch rates for use as an etch mask for e.g. silicon or glass etching; etching with low SF6 flow for compatibility with glass or silicon substrates.
Acknowledgements This work was carried out in the frame of SBO projects 060830, HyperCool-IT and 060046, Gemini, sponsored by the “Instituut voor de aanmoediging van Innovatie door Wetenschap en Technologie in Vlaanderen” (IWT), Belgium.
References 1.
M. Evans. Optimisation of Manufacturing Processes. London: Maney Publishing; 2003.
2.
A. Ezkerra, L.J. Fernández, K. Mayora, J.M. Ruano-López. Fabrication of SU-8 free-standing structures embedded in microchannels for
3.
H. Lorenz, M. Despont, N. Fahrni, N. LaBianca, P. Renaud, P. Vettiger. SU-8: a low-cost negative resist for MEMS. J Micromech.
4.
F. Ceyssens, M. Driesen, K. Wouters, R. Puers. A low-cost and highly integrated fiber optical pressure sensor system. Sensors and
microfluidic control. J Micromech Microeng (2007); 17: 2264-2271. Microeng (1997); 7: 121-124. Actuators A (2008); 145: 81-86. 5.
G. Hong, A.S. Holmes, M.E. Heaton. SU-8 resist plasma etching and its optimization. Microsystem Technologies (2004); 10: 357-359.
6.
K.R. Williams, K. Gupta, M. Wasilik. Etch rates for micromachining processing – part II. J Microelectromech Syst (2003); 12: 761-778.
7.
J.B. Bureau, B. Legrand, D. Collard, L. Buchaillot. 3D self-assembling of SU-8 microstructures on silicon by plasma induced
8.
L. Elbrecht, U. Storm, R. Catanescu, J. Binder. Comparison of stress measurement techniques in surface micromachining. J.
compressive stress. Transducers ’05, Digest of Technical Papers (2005); 1: 19-22. Micromech. Microeng (1997). 7: 151-154.