Journal of the Korean Physical Society, Vol. 51, No. 3, September 2007, pp. 984∼988

Deep Reactive Ion Etching of Polyimide for Microfluidic Applications T. N. T. Nguyen and N.-E. Lee∗ School of Advanced Materials Science & Engineering and Center for Advanced Plasma Surface Technology, Sungkyunkwan University, Suwon 440-746 (Received 13 November 2006) In this work, deep reactive ion etching (DRIE) of polyimide (PI) was carried out for the fabrication of a microfluidic channel and reservoir by using a CF4 /O2 inductively coupled plasma. The DRIE characteristics of PI were investigated by varying process parameters such as the total gas flow rate, the gas flow ratio, and the dc self-bias voltage (Vdc ). The results show that the etch rate increases with increasing total flow rate and with increasing Vdc due to the increased fluxes of reactive radicals and ions and to the increased ion-bombarding energy, respectively. The lateral etch (undercut) also increases with increasing of Vdc . When the CF4 /(CF4 + O2 ) flow ratio was varied, a maximum etch rate as high as 2.2 µm/min was obtained at a 20 % CF4 flow ratio. The 50-µm-thick PI sheet could be etched, though vertically, for application to a microfluidic channel or reservoir. PACS numbers: 52.75.Rx, 81.65.Cf Keywords: Deep reactive ion etching (DRIE), Microfluidic channel, Reservoir, Polyimide

PI due to the relative difficulty in processibility compared to other polymeric materials, such as PDMS and PC. In the case of PI, subtractive methods are more appropriate for the fabrication of microfluidic components. Subtractive technologies for pattern formation on PI mainly include laser ablation [14, 15] and etching [11, 16]. Because the laser ablation method is expensive, etching is preferable. Although the wet etching of PI has some advantages, high etch rate, low cost, and simple fabrication, it has inherent limits, highly sloped sidewall and large feature dimensions [16]. The slope angle could be up to 30 ∼ 50◦ , depending on the mask layer [16]. Therefore, wet etching is only useful in applications for patterns with large dimensions and a large allowable sidewall slope. To overcome this slope problem, an other choice could be dry etching by using a reactive ion plasma. Plasma etching of PI has been studied using O2 /CF4 [17], O2 /SF6 [17,18], O2 /F2 [19,20] and O2 [19, 21] chemistries. However, there have been no studies on deep reactive ion etching of PI with a depth of several tens of micrometers. In this work, the deep reactive ion etching characteristics of PI were investigated for CF4 /O2 inductively coupled plasmas by varying the process parameters, such as the gas flow ratio, the total flow rate, and the dc self-bias voltage, Vdc . When the total gas flow and the Vdc were increased, the etch rate of PI was increased. There was an optimum CF4 gas flow ratio for the highest etch rate. A thru-etch process for a 50-µm-thick PI substrate was successfully developed.

I. INTRODUCTION The micro-total analysis system (µ-TAS, also called a lab-on-a-chip or a microfluidic chip) [1–3] has been developed rapidly since last decade for applications in bio-chemical engineering [1–3] and environmental monitoring [3, 4]. Originally, µ-TAS devices fabricated on silicon incorporated pretreatment, separation, and detection sections as fundamental components [1–3]. µTAS devices employ many microfluidic components, including microfluidic channels, reservoirs, valves, mixers and pumps. Fabrication technologies for the microfluidic components in a µ-TAS have been developed based on silicon [5], glass [6–8], and quartz [9, 10] substrate materials combined with other polymer cover materials such as, for example, polydimethylsiloxane (PDMS) and polymethyl methacrylate (PMMA). An alternative approach is to use polymer substrate materials in microfluidic devices [11,12]. Depending on the fabrication methods and applications, a variety of polymer substrate materials, such as PDMS, polyimide, polyethylene (PE), and polycarbonate (PC), have been used. With outstanding mechanical, thermal, and electrical properties, PI is widely used in flexible microelectronic devices [12–14]. For applications of PI substrates to µ-TAS devices, fabrication technologies of PI are needed. However, replication methods, including casting, injection molding, or embossing, are rather limited for the fabrication of microfluidic components using ∗ E-mail:

[email protected]

-984-

Deep Reactive Ion Etching of Polyimide for Microfluidic· · · – T. N. T. Nguyen and N.-E. Lee

-985-

Fig. 1. Ni mask fabrication and DRIE procedure: (a) SiO2 deposition on the backside of PI as an etch-stop layer, (b) Cr/Cu adhesion/seed deposition, (c) lithography with a KMPR negative PR, (d) Ni hard-mask electroplating, (e) removal of the PR, (f) selective wet etching of Cr/Cu and (g) reactive ion etching of the PI layer (CF4 +O2 ).

II. EXPERIMENT Plasma etching of PI was carried out in a commercial 8-inch inductively- coupled- plasma (ICP) reactor (TCP 9100, Lam Research Corp.) equipped with a turbomolecular pump (2000 `/sec) backed by a dry pump. Plasma generation was controlled by a 13.56-MHz top electrode power. The bottom electrode power was supplied by a 4-MHz r.f. generator to induce the negative dc self-bias voltage (Vdc ). During etching, the chamber pressure was kept at 18 mTorr. Figure 1 shows the Ni hard-mask fabrication and DRIE processes. A 2 × 2 cm2 PI substrate (Kapton, DuPont) with a thickness of 50 µm was used for the etching experiments. When the thru-etch of PI was carried out, a SiO2 etch-stop layer with a thickness of 200 nm was deposited on the backside of the PI substrate by using e-beam evaporation. Next, the Cr (50 nm)/Cu (100 nm) layers as adhesion/seed layers were deposited on the front side of the PI substrate by d.c. sputtering. Sequentially, the photolithography process was carried out with a negative-tone KMPR-1050 photoresist (PR) (Kayaku MicroChem Corp.). The Ni electroplating process was used to form the hard-mask for etching. The Ni hard-mask formed using the KMPR resist showed a better edge profile compared to that using SU-8 PR. After Ni electroplating, the PR was removed using a remover (PG remover, Kayaku MicroChem Corp.), and a

Fig. 2. (a) Etch rate of PI as a function of the (CF4 + O2 ) total flow rate, (b) Cross-sectional SEM image of etched PI with the top electrode at 1000 V, a Vdc at −70 V, a CF4 /(CF4 +O2 ) flow ratio of 10 %, the total flow rate of 300 sccm.

microwave remote CF4 /O2 /N2 plasma was used to clean the remaining PR at the edge of the Ni hard-mask. A selective wet etching of the Cr adhesion and the Cu seed layers followed. The Cu seed layer was etched in few seconds by using nitric acid (50 %) mixed with deioninzed water. The Cr adhesion layer was dip-etched at 75 ◦ C in a solution containing 60 g/l of potassium permanganate and 200 g/l of tri-basic sodium phosphate. The etch rate and the profile of the PI were measured by using a scanning electron microscope (SEM, HITACHI S-3500 H).

III. RESULTS AND DISCUSSION Figure 2(a) shows the etch rate versus the total gas flow rate with the top electrode at 1000 V, Vdc at −70

-986-

Journal of the Korean Physical Society, Vol. 51, No. 3, September 2007

Fig. 3. (a) Etch rate of PI as a function of the CF4 /(CF4 + O2 ) flow ratio, (b) Cross-sectional SEM image of etched PI with the top electrode at 1000 V, a Vdc at −70 V, a CF4 /(CF4 +O2 ) flow ratio of 20 %, the total flow rate of 300 sccm.

V and a CF4 /(CF4 +O2 ) flow ratio of 10 %. The results show that the etch rate increases with increasing total gas flow, presumably due to increases in the reactive radical and ion fluxes. An etch rate of 2.6 µm/min was obtained at a total flow rate of 400 sccm, and the profile was nearly vertical without a significant undercut. As an example, the SEM image of the sample etched at a total flow rate of 300 sccm is shown in Figure 2(b). As shown in Figure 2(b), the profile angle was close to 90◦ . The etch rate results indicate that the Ni hard-mask fabrication process was successfully developed using a KMPR photoresist because the PI sheet was kept flat after etching, the adhesion of Ni hard-mask was good enough to endure the entire deep etching process, and the damage to the Ni hard-mask during etching was minimal. The etch rate results obtained as a function of the

CF4 flow ratio with the top electrode at 1000 V, the Vdc at −70 V, and a total flow rate of 300 sccm are shown in Figure 3(a). The highest etch rate obtained was 2.2 µm/min at a 20 % CF4 flow ratio, and the etch rate was lower at a CF4 flow ratio of 30 %. The cross-sectional SEM image of the PI etched at a 20 % CF4 flow ratio shows a vertical profile with a slightly enlarged pattern width, (Figure 3(b)). Addition of fluorine to the oxygen plasma combined with ion-assisted plasma etching with the dc self-bias voltage (Vdc ) applied was found to greatly enhance the etch rate [17, 19, 20, 22, 23]. Both fluorine and oxygen easily attack the PI backbone with the carboncarbon, carbon-oxygen, carbon-hydrogen, and carbonnitrogen bonds to form carbon-fluorine (CHF2 , CH2 F2 and CF4 ), carbon-oxygen (CO and CO2 ), and carbonfluorine-oxygen (COF2 ) compounds [19]. The nitrogen in the PI may form fluorine-containing compounds, such as NF3 . The presence of fluorine in the plasma also increases the formation of H2 O and HF compounds [19]. If there is no ion-bombardment on the surface, the fluorine compound stays on the surface, which inhibits the reaction of oxygen with the carbon and the hydrogen in the PI layer. In this situation, the etch rate is independent of the fluorine concentration in the plasma [20]. In an electron cyclotron resonance (ECR) plasma, addition of SF6 gas was shown to decreased the PI etch rate due to the formation of nonvolatile fluorinated compounds on the etched surface [18]. Therefore, it is necessary to supply the Vdc for enhancing the sputtering effect to increase the etch rate. Under the biased condition, the etch rate enhancement with the addition of F containing gas is attributed to enhanced removal of hydrogen, carbon, oxygen, and nitrogen atoms in the PI layer, together with the presence of oxygen radicals in the plasma. However, with a large fluorine concentration in plasma, there would be passivation on the surface, decreasing the etch rate. A comparison between O2 /CF4 and O2 /SF6 plasmas was made by Turban and Rapeaux [17]. By varying the CF4 to SF6 flow ratio from 0 ∼ 100 %, they showed that at lower, 30 %, CF4 to SF6 flow ratios, the O2 /CF4 plasma gives a higher etch rate than the O2 /SF6 plasma, but at a 60 % CF4 to SF6 flow ratio, the etch rate of the O2 /SF6 plasma is higher. However, that etch rate is much lower than that of a 30 % CF4 flow ratio. The reason for the lower etch rate at the higher CF4 flow rate is presumably the formation of a fluorocarbon passivation layer on the surface, which would impede the effective removal of etch by-product molecules. The etch rate and undercut as functions of Vdc are shown in Figure 4. The etch rate increases with increasing Vdc . When the −Vdc is larger than 110 V, the etch rate slightly decreases because the PI substrate is bent due to heating at an increased Vdc . When the Vdc increases, the ion bombardment energy increases. So the increase in the ion bombardment energy rate increases the etch rate, as well as the lateral etch (undercut), due

Deep Reactive Ion Etching of Polyimide for Microfluidic· · · – T. N. T. Nguyen and N.-E. Lee

-987-

Fig. 4. Etch rate and undercut of PI as function of the dc self-bias voltage, Vdc . Fig. 6. SEM image of the reservoir etched down to the SiO2 etch-stop layer.

Fig. 5. Etch rate of PI as a function of the channel width.

to effective removal of etch by-products by an increased physical sputtering effect. At a Vdc of −110 V, a top electrode power of 1000 V, a CF4 flow ratio of 10 %, and a total flow rate of 300 sccm, the etch rate could reach 2.6 µm/min. As previously mentioned, the sputtering effect has an important role in removing the by-product compounds from the etched surface due to difficulty in vaporizing the fluorocarbon compounds completely from the film surface. When the Vdc increases, the ion bombardment energy delivered to surface species increases and, as a result, easily breaks the carbon-carbon bonds. Fluorine and oxygen atoms also penetrate deeply into the polymer chain from both the vertical and the lateral sides for

etching, which leads to a larger undercut, as well as an increased vertical etch rate. Moreover, the sidewall was still vertical due to the sputtering effect. Figure 5 shows the dependency of the PI etch rate on the channel width. The experiment was carried out at a top electrode power of 1000 V, a Vdc of −70 V, a total flow rate of 300 sccm, and a CF4 /(CF4 +O2 ) flow ratio of 20 %. As Figure 5 shown, the etch rate decreases as the channel width decreases. The main reason for the pattern- width- dependent etch rate is related to the change in the etching mechanism due to a change in the transport of ion and radical species into the channel being etched. In fact, while the channel width decreases, the transport of radical and ion fluxes into the channel changes. Initially, the ion and the radical fluxes in plasma could easily reach the open area for etching, but as the etch depth increases, the ion and radical fluxes have more difficulty reaching the bottom of the channel. Also, the difficulty in removing the etch by-products can increase the redeposition of the by-product molecules to the sidewall. Therefore, the etch rate may decrease, and the etched profile may change. This phenomenon is often called aspect-ratio-dependent etching (ARDE), a type of micro- loading effect [7]. Figure 6 shows the SEM image of the 50-µm-thick PI sheet etched vertically at a top electrode power of 1000 V, a Vdc of −110 V, a total flow rate of 300 sccm, and a CF4 /(CF4 +O2 ) flow ratio of 10 % for the microfluidic channel and reservoir fabrication. The measured etch rate was 2.6 µm/min. The image indicates a vertical profile, but with a rounded corner at the bottom of the reservoir. On the other side of the reservoir, a sensing element can be integrated.

-988-

Journal of the Korean Physical Society, Vol. 51, No. 3, September 2007

IV. CONCLUSION The DRIE characteristics of PI in CF4 /O2 inductively coupled plasmas were investigated. The etch rate was observed to depend on the total flow rate of CF4 and O2 , the CF4 flow ratio, the dc self-bias voltage (Vdc ), and the pattern width. While varying the CF4 flow ratio, a 20 % CF4 flow ratio was shown to give the highest etch rate. For application in a microfluidic device, a microfluidic channel and reservoir can be successfully fabricated by etching vertically down to a depth of 50 µm depth at an etch rate of 2.6 µm/min for a reservoir with a width of 100 µm.

ACKNOWLEDGMENTS This work was supported by the Center of Excellence Program of the Korea Science and Engineering Foundation (Grant No. R-11-2000-086-0000-0).

REFERENCES [1] D. R. Reyes, D. Lossifidis, P. A. Auroux and A. Manz, Anal. Chem. 74, 2623 (2002). [2] P. A. Auroux, D. Lossifidis, D. R. Reyes and A. Manz, Anal. Chem. 74, 2637 (2002). [3] T. Vilkner, D. Janasek and A. Manz, Anal. Chem. 76, 3373 (2004). [4] D. M. Rawson, R. Bhatia, J. W. Dilleen and A. L. Atkinson, Biosensors and Bioelectronics 18, 667 (2003). [5] I. W. Rangelow, Vacuum 62, 279 (2001).

[6] T. Ichiki, Y. Sugiyama, T. Ujiie and Y. Horiike, J. Vac. Sci. Technol. B 21, 2188 (2003). [7] J. H. Park, N.-E. Lee, J. Lee, J. S. Park and H. D. Park, Microelectron. Eng. 82, 119 (2005). [8] J. H. Park, N.-E. Lee, J. Lee, J. S. Park and H. D. Park, J. Korean Phys. Soc. 47, S422 (2005). [9] D. A. Zeze, R. D. Rorrest, J. D. Carey, D. C. Cox, I. D. Robertson, B. L. Weiss and S. R. P. Silva, J. Appl. Phys. 92, 7 (2002). [10] P. W. Leech, Vacuum 55, 191 (1999). [11] J. S. Rossier, C. Vollet, A. Carnal, G. Lagger, V. Gobry, H. H. Girault, P. Michel and F. Reymond, Lab. Chip 2, 145 (2002). [12] S. Metz, R. Holzer and P. Renaud, Lab. Chip 1, 29 (2001). [13] J. T. Mabeck and G. G. Malliaras, Anal. Bioanal. Chem. 384, 343 (2006). [14] B. C. Giordano, J. Ferrance, S. Swedderg, A. F. R. Huhmer and J. P. Landers, Anal. Biochem. 291, 124 (2001). [15] M. Roberts, J. Rossier, P. Bercier and H. Girault, Anal. Chem. 69, 2035 (1997). [16] J.-S. Han, Z.-Y. Tan and M. Shikida, J. Micromech. Microeng. 14, 38 (2004). [17] G. Turban and M. Rapeaux, J. Electrochem. Soc. 130, 2231 (1983). [18] S. H. Kim, C. H. Lee and J. Ahn, J. Korean Phys. Soc. 40, 99 (2002). [19] I. Tepermeister and H. Sawin, J. Vac. Sci. Technol. A 9, 790 (1991). [20] A. Furuya, F. Shimokawa, T. Matsuura and R. Sawada, J. Micromech. Microeng. 4, 67 (1994). [21] N. Agarwal, S. Ponoth, J. Plawsky and P. D. Persans, J. Vac. Sci. Technol. A 20, 1587 (2002). [22] J. W. Bae, C. H. Jeong, J. T. Lim, H. C. Lee, G. Y. Yeom and I. Adesida, J. Korean Phys. Soc. 50, 1130 (2007). [23] B. Kim, J. Kim, S. H. Lee, J. Park and B. T. Lee, J. Korean Phys. Soc. 47, 712 (2005).