- Email: [email protected]
Fast and simple fabrication procedure of whole-glass microfluidic devices with metal electrodes
Microelectronic Engineering 110 (2013) 441–445
Contents lists available at SciVerse ScienceDirect
Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee
Fast and simple fabrication procedure of whole-glass microfluidic devices with metal electrodes Jaroslav Kotowski a,⇑, Vít Navrátil a, Zdeneˇk Slouka b, Dalimil Šnita c a
Institute of Chemical Technology Prague, Department of Chemical Engineering, Technická 5, 166 28 Prague, Czech Republic University of Notre Dame, Notre Dame, IN 46556, United States c University of West Bohemia, Research Centre New Technologies, Univerzitní 8, 306 14 Plzenˇ, Czech Republic b
a r t i c l e
i n f o
Article history: Available online 29 March 2013 Keywords: Microfluidic system Glass chip Integrated electrodes Micro-device Fabrication
a b s t r a c t We developed a new process for fabrication of whole-glass microfluidic chips with integrated metal electrode arrays. Our process is based on a novel technique that enables reliable bonding of two glass substrates, one with a microfluidic channel and the other with an electrode array. The technique uses sodium silicate as an intermediate layer between the two glass substrates; this layer provides very tight bonding while preserving full functionality of the metal electrodes. Functionality of the electrode array was confirmed by impedance spectroscopy measurements of KCl solutions with different concentrations. To confirm the quality of bonding, we used a solution of fluorescein as a tracer of any poorly bonded areas. Our results showed that there was no fluorescein leakage out of the microfluidic channel and that the electrodes were free of sodium silicate where desired. Our method can be applied to fast and costeffective prototyping of whole-glass microfluidic chips with integrated electrode arrays. Ó 2013 Elsevier B.V. All rights reserved.
1. Introduction Microfluidic systems (Lab-On-a-Chip) are widely used in biological/analytical applications such as blood cell sorting [1], DNA analysis [2], or immuno-diagnostics [3] because they combine complex laboratory functions with the low consumption of samples. One of the key components of microfluidic chips is a microelectrode array which can be used either as an active (e.g., source electrodes for AC or DC electro-osmosis [4]) or passive element (e.g., electrical potential sensing [5]). The first Lab-On-a-Chip devices were made of glass or silicon substrates [6] by fabrication techniques commonly used in the semiconductor industry. Nowadays, polymeric materials such as PDMS, PMMA, and PC are often used for microfluidic system fabrication since they enable the low-cost and mass production of these systems through commercially available processes such as casting, hot-embossing, imprinting etc. Polymeric microfluidic systems are easy to fabricate, but they lack thermal stability and resistance to organic solvents. Problems can be also caused by the gas permeability of some polymeric materials (e.g., PDMS [7]). Unlike to polymeric materials, glass possesses very high thermal stability and a broad range of available surface chemistries enables to tailor its surface properties to be tailored to specific needs. ⇑ Corresponding author. Address: ICT Prague, Technická 5, 166 28 Prague, Czech Republic. Tel.: +420 22044 3239; fax: +420 220 444 320. E-mail address: [email protected] (J. Kotowski). 0167-9317/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.mee.2013.03.123
These properties are advantageously used in many applications e.g., chemical synthesis, the production of nanoparticles, reaction of supercritical solvents etc. [8–11]. Glass microfluidic systems in which integrated electrodes are usually made as two-layer systems where one layer bears microelectrodes (made by the deposition of metal layers followed by lithography) and the other layer contains microfluidic channels (made by etching in HF, laser machining, mechanical machining or sandblasting). A comprehensive summary of the fabrication processes used for glass microfluidic systems was published by Iliescu et al. [12]. The aforementioned manufacturing of microfluidic channels and electrode arrays on glass substrates through processes mentioned above is routinely conducted in many microfluidic laboratories. But the reliable, defect-free and leakage-free bonding of those glass substrates that preserve full functionality of electrodes is still quite challenging. The reason for this is the presence of a thin layer of electrodes, that makes the glass substrate surface rather uneven and heterogeneous. Heterogeneity and surface unevenness cause serious issues during the bonding process during which proper contacting of bonded substrates plays a crucial role. Several techniques of glass-to-glass bonding have been published: high temperature bonding [13], low temperature bonding [14,15], thermal anodic bonding [16], and two-step activation bonding [17]. But they practically always fail to provide a tight bond and good channel sealing in the case of glass substrates with integrated electrodes. An interesting solution for this issue was published by Henry et al. [18]. The authors created small grooves
442
J. Kotowski et al. / Microelectronic Engineering 110 (2013) 441–445
on a glass substrate which were later filled with a metal by a number of sputtering processes. However, repeated metal deposition requiring precise control of metal layer thickness makes the process quite difficult, costly, time consuming and not suitable for not well equipped microfluidics labs. Due to fabrication difficulties in bonding glass substrates, several hybrid glass/PDMS microfluidic chips with integrated electrodes were developed [19–21]. In those microfluidic chips, glass and PDMS are used as a carrier of an electrode array and microfludic channels, respectively. Bonding of those materials relies on traditional surface plasma activation which along with the flexibility of PDMS provides a good way of sealing microfluidic channels on glass substrates with uneven surface caused by the presence of electrodes. The group of Li [19] used a layer of SU8 photoresist coated on a silicon substrate to facilitate the bonding of PDMS by activation in ozone. Group of Shu-Ming prepared a hybrid glass/ PDMS chip by using UV curable glue as an intermediate layer [20]. The use of these hybrid chips, however, is limited to low temperature and low pressure applications. Also different surface properties of materials, e.g., a wettable glass and a non-wettable PDMS might introduce unexpected issues in some applications. In this paper, we present a fast and inexpensive method for the fabrication of whole-glass microfluidic chips with integrated metal electrode arrays. Our fabrication method requires minimum number of expensive devices and can be performed with standard laboratory equipment.
2. Experimental 2.1. Fabrication procedure The entire fabrication procedure is illustrated in Fig. 1. Microscopic glass slides (75 by 25 mm) were used as glass substrates. Electrode arrays were made by the standard process based on the combination of metal sputtering and photolithography (photoresist P-ma 1275 by Microresist) (Fig. 1A). In the first stage, electrodes made of inexpensive nickel were used to develop and optimize the whole fabrication procedure. Nickel was later replaced by more expensive gold. The electrode array was designed as a four electrode array with the following parameters: width of the electrodes 200 lm, the gap between the electrodes 400 lm. In order to investigate the influence of the metal layer thickness on the quality of bonding, we prepared electrode arrays with different thicknesses ranging from 200 to 700 nm. Microfluidic channels were etched into a second glass slide (Fig. 1B) through a structured polyester tape (type M42 by Europack Chrudim) in a solution made up of hydrofluoric acid, sulfuric acid, and DI water mixed in the volumetric ratio 1:2:1 [22]. The polyester tape serves as a mask for etching of the required structures. The tape was cut manually using a scalpel. Alternatively, a cutting plotter can be used [23]. The etching was performed in a laboratory shaker (IKA KS125) set to 300 cycles/min at room temperature. The etching rate was 5 lm/min. Via-holes located at the beginning and at the end of the microchannel were drilled by diamond coated ball burrs (by Dremel Motoflex) under water to prevent the local overheating and possible breakage of the glass substrate. The glass substrate bearing the microelectrodes is coated with a non-diluted solution of sodium silicate (10% NaOH and 26% SiO2 – known as water glass, by Sigma Aldrich) (Fig. 1C), and immediately covered with the glass substrate containing the microfluidic channel (Fig. 1D). The excess of sodium silicate located along the perimeter of the contacted glass substrates is rinsed off by DI water. The microchannel filled with the water glass is then connected to a
A B C D
E
F
Fig. 1. Scheme of the whole fabrication process of the glass microfluidic chip with integrated electrode array: (A) glass slide with an electrodes array, (B) glass slide with a microfluidic channel, (C) coating of the glass slide with sodium silicate, (D) attachment of the mating glass slide, (E) removal of the excess of sodium silicate by vacuum suction and rinsing with DI water and (F) drying of sodium silicate at 95 °C for 60 min.
vacuum pump (KNF N816.3KT) that removes the water glass from the channel. To ensure that all water glass is removed from the channel and the surface of the electrodes is free of any sodium silicate solution, we flush the channel with DI water (approximately 2 ml) while the channel is still hooked up to the vacuum (Fig. 1E). The assembled chip is then put in an oven set at 95 °C for 60 min to dry the remaining water glass which bonds the two glass substrates together. Fig. 2A and B shows the fabricated glass chip. 2.2. Confirmation of glass bonding To prove that the bond in glass/metal/glass sandwich is defectfree, we filled the micro channel with a fluorescein solution to trace any possible defects. Inspection was done visually after the chip illumination by UV light (using microscope OLYMPUS BX51WI/U-RFL-T). 2.3. Electrochemical impedance spectroscopy measurements An important step in the fabrication procedure is flushing of the microfluidic channel. Flushing assures that the surface of the electrode array is free of any sodium silicate. To test that the electrodes are fully functional, we measured electrochemical impedance spectra of KCl solutions with different concentrations. Electrochemical impedance spectroscopy (EIS) measures the impedance of the electrochemical cell in the dependence on frequency of a low amplitude AC electrical signal. The results from EIS are traditionally plotted as a dependence of total impedance on the frequency (Bode plot) or real and imaginary parts of impedance are
J. Kotowski et al. / Microelectronic Engineering 110 (2013) 441–445
443
2.4. Determination of the critical dimensions of microchannel The technology for bonding two glass substrates presented in this paper is limited to fluidic channels from which the solution of sodium silicate can be removed by a vacuum pump. To determine the critical dimensions and cross-section of our fluidic channel which can be successfully bonded with the use of sodium silicate, we made a set of 5 cm long straight channels with different channel heights and widths and tested them in our bonding experiment. The width and the height of the channel were determined by an optical microscope. The critical dimensions obtained for 5 cm long straight channel have informative character only. In general, critical dimensions will depend on the channel length, its curvature and also viscosity/concentration of sodium silicate solution. Longer and curved channels bonded with sodium silicate solution of larger viscosity will have larger critical dimensions than those determined here. To see the effect of sodium silicate viscosity on critical dimensions of the channels as well as on the quality of the created bond we conducted experiments with diluted solutions of sodium silicate. The sodium silicate solutions were prepared by mixing the stock solution with DI water in the following sodium silicate to DI water volume ratios: 2:1 (sol a), 3:2 (sol b) and 1:1 (sol c). 3. Results and discussions 3.1. Confirmation of bonding
Fig. 2. (A) The glass mifrofluidic chip with 200 nm thick gold electrodes and (B) The glass microfludic chip with 200 nm thick nickel electrodes and the channel connected to the connection tubes compared to the size of two euro coin.
displayed (Nyquist plot). Equivalent electrical circuits are often used to extract data of interest, e.g., resistance (conductivity) of a solution, capacity of electrical double layer at electrode–electrolyte interfaces etc. Randles circuit is the simplest but frequently used equivalent electrical circuit (see Fig. 3). It consists of four elements: (i) resistance of the electrolyte between the electrodes Rs, (ii) resistance associated with electron transfer reaction (electrochemical reaction) Rct, (iii) capacitance of the electrical double layer at electrode–electrolyte interface Cdl, and (iv) constant phase element associated with diffusion transport of electroactive species (Warburg impedance) Zw. Here we show the EIS results in the form of the Bode plot without further analysis of the obtained data. The EIS measurements were carried out in a two-electrode set up. The channel was filled with a KCl solution of various concentrations (1 M, 0.1 M and 0.01 M). We measured the dependence of impedance on the frequency of the applied sinusoidal signal with amplitude of 0.01 V. The frequency ranged from 1 to 105 Hz. The measurement was performed with a potentiostat (AUTOLAB PGSTAT 302N) using the NOVA 1.8 software.
Fig. 4A shows a microfluidic chip with 200 nm thick electrodes made with the sodium silicate assisted bonding technique. The green/black interface identifies the channel walls, the four whitedotted strips across the channel are metal electrodes. The microfluidic channel is well-sealed in the area of the electrode array. No leaking fluorescein was observed along the electrodes or between the glass substrates. The same quality of the bonding was also observed for all thicker electrodes layers up to 700 nm. The rest of the experiments confirming the functionality of the electrode array and showing the effect of sodium silicate concentration/viscosity on the quality of the bonding were carried out with the 200 nm thick electrodes. The chips with 200 nm thick electrodes assembled by using solution sol a (DI water:sodium silicate = 1:2) proved the same quality of the sealing as non-diluted sodium silicate solution. The use of sol b solution (DI water:sodium silicate = 2:3) led to a sealed channel in quality comparable to that of undiluted sodium silicate. However, during the assembly of both parts of the chip (glass with the microchannel, glass with electrode array), a significant amount of air bubbles was trapped between the glass slides. The air bubbles prevented the proper bonding over the surface of the two glass pieces. The presence of the bubbles significantly complicated the fabrication process and the assembly process had to be repeated many times until no air bubbles were present in the sodium silicate solution layer. We were not able to fabricate a chip with sufficient sealing quality using the solutions (DI water:sodium silicate = 1:1) (Fig. 4B). The fluorescein tracer leaked along the electrodes and in between the glass slides. A spontaneous disintegration of these glass chips was observed after three hours. 3.2. Verification of electrode array functionality
Fig. 3. Randles circuit as a equivalent electrical circuit for electrochemical impedance spectroscopy.
We measured EIS spectra of KCl solutions as described in Section 2 for both nickel (Fig. 5A) and gold (Fig. 5B) electrode arrays. The shape of the EIS curves in the graphs is qualitatively the same as is described in [24]. For higher frequencies (104–105 Hz), there is
444
J. Kotowski et al. / Microelectronic Engineering 110 (2013) 441–445
Fig. 4. The detail of the microchannel (with electrodes) filled with the fluorescent tracer observed under the UV light, (A) channel fabricated using undiluted sodium silicate is properly sealed, width of the channel is 954 lm. (B) Channel fabricated using diluted sodium silicate in ratio 1:1 with DI water. Fluorescent tracer leaks out ofthe channel. Width of the channel is 417 lm.
Impedance (Ω)
10
10
10
10
7
7
6
Au electrode
6
10
5
5
10
4
4
10
3
3
2
10 −1 10
C
B 10
Ni electrode
Z(Ω)
A 10
10
c = 1.00 M KCl c = 0.10 M KCl c = 0.01 M KCl 0
10
10
1
2
3
10 10 Frequency (Hz)
10
4
10
10 −1 10
10
0
1
10
10
2
10
3
4
10
10
5
Frequency (Hz)
9
10
2
5
c = 1.00 M KCl c = 0.10 M KCl c = 0.01 M KCl
D
Au electrode - without rinsing
8
7
10
Au electrod - after rinsing
6
10
10
7
5
10
Z(Ω)
Z(Ω)
10
6
4
10
10
5
3
10
4
10 0 10
10 c = 1.00 M KCl c = 0.10 M KCl c = 0.01 M KCl 10
1
10
2
2
10
3
10
4
5
10
10
6
Frequency (Hz)
10 0 10
c = 1.00 M KCl c = 0.10 M KCl c = 0.01 M KCl 10
1
10
2
10
3
10
4
5
10
6
10
Frequency (Hz)
Fig. 5. The dependence of impedance on the frequency of the applied sinusoidal signal with the amplitude of 0.01 V for three various KCl solution concentrations 1.0, 0.1 and 0.01 M KCl. (A) Microfludic chip with nickel electrode array, (B) microfludic chip with gold electrode array, (C) microfludic chip with gold electrode array fabricated without the flushing step (step e in Fig. 1), (D) microfludic chip with gold electrode array as presented in the Fig. 4C subsequently flushed with DI water.
a significant difference between the impedance curves for various electrolyte concentrations. KCl solutions of higher concentrations gave lower impedance at the same frequency. This part if the impedance spectra can be used for extraction of information on conductivity of electrolyte in the channel. On the other hand, the impedance curves collapse onto each other for higher frequencies. The dependence on concentration is surprisingly completely lost
showing that there is another effect which controls the behavior of the system in the low frequency regime. We suspect that this effect might be related to the geometry of electrodes with a meshlike pattern. To show that the flushing step is necessary, we carried out an EIS measurement in a microfluidic chip fabricated without rinsing the channel with water to remove sodium silicate from the surface
J. Kotowski et al. / Microelectronic Engineering 110 (2013) 441–445 Table 1 Critical dimensions of microchannel. Dilution ratio (sodium silicate:water)
Width (lm)
Depth (lm)
Cross-section (lm2)
0:1 2:1 3:2
425 350 310
20 12 10
8500 4200 3100
of electrodes. There is no significant frequency dependence of the impedance for the various KCl solutions both for lower and higher frequency signals (Fig. 5C). The same chip that was fabricated without flushing of the channel with DI water was subsequently flushed with DI water several times and the EIS measurement was repeated Fig. 5D. The shape of impedance curves in Fig. 5D is analogous to the impedance curves in Fig. 5B. Comparison of Fig. 5C and D shows that the sodium silicate, which covered the electrodes, can be dissolved and flushed from the channel even after the chip is completely fabricated and dried. However, it is desirable to perform the flushing immediately after bonding the two glasses together otherwise the remaining sodium can glue the channel. This experiment led us to the assumption that it is possible to dissolve the sodium silicate in the microfluidic channel. Therefore we kept the whole microfluidic chip with electrode array immersed in DI water for one week. No defect in the sealing/bonding of the microfluidic chip was observed. 3.3. Determination of the critical cross-sections The critical cross-sections of 5 cm long, straight channels for different sodium silica solutions were determined according to Section 2.4. The dimensions of channels tested are summarized in Table 1. The values given in the table are averaged values from two/three measurements taken at different positions along the channel (beginning, middle, and end). We found that the increasing dilution ratio of sodium silicate allows the reduction of the critical dimensions (i.e., cross-section). However the dilution also increases difficulty (e.g., error rate) during fabrication procedure. 4. Conclusions We showed that bonding of glass substrates by means of nondiluted sodium silicate solution represents a simple yet reliable method for low temperature fabrication of glass microfluidic chips with integrated electrode arrays. This bonding technique does not require any expensive commercial apparatuses or expensive material. The cost as well as very fast preparation time (3–4 h) makes our technique a suitable candidate for prototyping. We optimized this fabrication process for electrodes with the thickness up to 700 nm. We verified full functionality of electrodes by running an electrochemical impedance spectroscopy measure-
445
ment. This measurement also revealed the importance of rinsing the channel with sufficient amount of DI water to remove all sodium silicate from the surface of electrodes. The critical microchannel cross-section of 8500 ± 10% lm2 was evaluated for a 5 cm long straight microchannel when using non-diluted sodium silicate solution. We successfully fabricated microfluidic glass chips sealed by a sodium silicate solution diluted with water in a volume ratio of 3:2 and obtained a defect-free bond between glass/metal/glass layers. The use of lower viscosity sodium silicate solution led to the decrease in the critical cross-section to 3100 ± 10% lm2 for the 5 cm long straight channel. Acknowledgement The authors thank for the financial support by the Grant of the GAAV CR IAA401280904 and by the specific university research (MSMT No. 20/2013). The result was developed within the CENTEM project, Reg. No. CZ.1.05/2.1.00/03.0088, co-funded by the ERDF as part of the Ministry of Education, Youth and Sports’ OP RDI programme. References [1] B. Qu, Z. Wu, F. Fang, Z. Bai, D. Yang, S. Xu, Anal. Bioanal. Chem 392 (2008) 1317–1324. [2] W. Kubicki, R. Walczak, J. Dziuban, Optica Appl. XLI (2) (2011). [3] K. Suk Yang, H.J. Kim, J.K. Ahn, D.H. Kim, Curr. Appl. Phys. 9 (2009) e60–e65. [4] M.S. Arefin, T.L. Porter, J. Appl. Phys. 111 (2012) 054919. [5] M. Svoboda, Z. Slouka, W. Schrott, D. Šnita, Microelectron. Eng. 86 (2009) 1371–1374. [6] P. Abgrall, A.-M. Gue, J. Micromech. Microeng. 17 (2007) R15–R49. [7] M. Sadrzadeh, K. Shahidi, T. Mohammadi, J. Appl. Poly. Sci. 117 (2010) 33–48. [8] A. Abou-Hassan, O. Sandre, V. Cabuil, Angew. Chem. Int. Ed. 49 (2010) 6268– 6286. [9] S. Marre, A. Adamo, S. Basak, C. Aymonier, K.F. Jensen, Ind. Eng. Chem. Res. 49 (2010) 11310–11320. [10] W. Verboom, Chem. Eng. Technol. 32 (11) (2009) 1695–1701. [11] F. Trachsel, C. Hutter, P.R. von Rohr, Chem. Eng. J. 135S (2008) S309–S316. [12] C. Iliescu, H. Taylor, M. Avram, J. Miao, S. Franssila, Biomicrofluidics 6 (2012) 016505. [13] S. Queste, R. Salut, S. Clatot, J.-Y. Rauch, C.G.K. Malek, Microsys. Technol. 16 (2010) 1485–1493. [14] L. Chen, G. Luo, K. Liu, J. Ma, B. Yao, Y. Yan, Y. Wang, Sens. Actuators, B 119 (2006) 335–344. [15] H.Y. Wang, R.S. Foote, S.C. Jacobson, J.H. Schneibel, J.M. Ramsey, Sens. Actuators B 45 (1997) 199–207. [16] P. Maoa, J. Han, Lab Chip 5 (2005) 837–844. [17] Y. Xu, C. Wang, Y. Dong, L. Li, K. Jang, K. Mawatari, T. Suga, T. Kitamori, Anal. Bioanal. Chem. 402 (2012) 1011–1018. [18] C.S. Henry, Microchip Capillary Electrophoresis, Methods and Protocols, vol. 339, 2006, pp. 13–26. [19] P. Li, N. Lei, D.A. Sheadel, J. Xu, W. Xue, Sens. Actuators B 166–167 (2012) 870– 877. [20] S. Kuoa, C. Yanga, J. Shieab, C. Lin, Sens. Actuators B 156 (2011) 156–161. [21] S.M. Shameli, T. Glawdel, Z. Liu, C.L. Ren, Anal. Chem. 84 (2012) 2968–2973. [22] O.N. Dyatlova, V.V. Bykov, Glass Ceram. 19 (2) (1962) 77–81. [23] Po Ki Yuen, Vasiliy N. Goral, J. Chem. Educ. 89 (10) (2012) 1288–1292. [24] J. Janouš, P. Beranek, M. Pribyl, D. Šnita, Microelectron. Eng. 97 (2012) 387– 390.
Contents lists available at SciVerse ScienceDirect
Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee
Fast and simple fabrication procedure of whole-glass microfluidic devices with metal electrodes Jaroslav Kotowski a,⇑, Vít Navrátil a, Zdeneˇk Slouka b, Dalimil Šnita c a
Institute of Chemical Technology Prague, Department of Chemical Engineering, Technická 5, 166 28 Prague, Czech Republic University of Notre Dame, Notre Dame, IN 46556, United States c University of West Bohemia, Research Centre New Technologies, Univerzitní 8, 306 14 Plzenˇ, Czech Republic b
a r t i c l e
i n f o
Article history: Available online 29 March 2013 Keywords: Microfluidic system Glass chip Integrated electrodes Micro-device Fabrication
a b s t r a c t We developed a new process for fabrication of whole-glass microfluidic chips with integrated metal electrode arrays. Our process is based on a novel technique that enables reliable bonding of two glass substrates, one with a microfluidic channel and the other with an electrode array. The technique uses sodium silicate as an intermediate layer between the two glass substrates; this layer provides very tight bonding while preserving full functionality of the metal electrodes. Functionality of the electrode array was confirmed by impedance spectroscopy measurements of KCl solutions with different concentrations. To confirm the quality of bonding, we used a solution of fluorescein as a tracer of any poorly bonded areas. Our results showed that there was no fluorescein leakage out of the microfluidic channel and that the electrodes were free of sodium silicate where desired. Our method can be applied to fast and costeffective prototyping of whole-glass microfluidic chips with integrated electrode arrays. Ó 2013 Elsevier B.V. All rights reserved.
1. Introduction Microfluidic systems (Lab-On-a-Chip) are widely used in biological/analytical applications such as blood cell sorting [1], DNA analysis [2], or immuno-diagnostics [3] because they combine complex laboratory functions with the low consumption of samples. One of the key components of microfluidic chips is a microelectrode array which can be used either as an active (e.g., source electrodes for AC or DC electro-osmosis [4]) or passive element (e.g., electrical potential sensing [5]). The first Lab-On-a-Chip devices were made of glass or silicon substrates [6] by fabrication techniques commonly used in the semiconductor industry. Nowadays, polymeric materials such as PDMS, PMMA, and PC are often used for microfluidic system fabrication since they enable the low-cost and mass production of these systems through commercially available processes such as casting, hot-embossing, imprinting etc. Polymeric microfluidic systems are easy to fabricate, but they lack thermal stability and resistance to organic solvents. Problems can be also caused by the gas permeability of some polymeric materials (e.g., PDMS [7]). Unlike to polymeric materials, glass possesses very high thermal stability and a broad range of available surface chemistries enables to tailor its surface properties to be tailored to specific needs. ⇑ Corresponding author. Address: ICT Prague, Technická 5, 166 28 Prague, Czech Republic. Tel.: +420 22044 3239; fax: +420 220 444 320. E-mail address: [email protected] (J. Kotowski). 0167-9317/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.mee.2013.03.123
These properties are advantageously used in many applications e.g., chemical synthesis, the production of nanoparticles, reaction of supercritical solvents etc. [8–11]. Glass microfluidic systems in which integrated electrodes are usually made as two-layer systems where one layer bears microelectrodes (made by the deposition of metal layers followed by lithography) and the other layer contains microfluidic channels (made by etching in HF, laser machining, mechanical machining or sandblasting). A comprehensive summary of the fabrication processes used for glass microfluidic systems was published by Iliescu et al. [12]. The aforementioned manufacturing of microfluidic channels and electrode arrays on glass substrates through processes mentioned above is routinely conducted in many microfluidic laboratories. But the reliable, defect-free and leakage-free bonding of those glass substrates that preserve full functionality of electrodes is still quite challenging. The reason for this is the presence of a thin layer of electrodes, that makes the glass substrate surface rather uneven and heterogeneous. Heterogeneity and surface unevenness cause serious issues during the bonding process during which proper contacting of bonded substrates plays a crucial role. Several techniques of glass-to-glass bonding have been published: high temperature bonding [13], low temperature bonding [14,15], thermal anodic bonding [16], and two-step activation bonding [17]. But they practically always fail to provide a tight bond and good channel sealing in the case of glass substrates with integrated electrodes. An interesting solution for this issue was published by Henry et al. [18]. The authors created small grooves
442
J. Kotowski et al. / Microelectronic Engineering 110 (2013) 441–445
on a glass substrate which were later filled with a metal by a number of sputtering processes. However, repeated metal deposition requiring precise control of metal layer thickness makes the process quite difficult, costly, time consuming and not suitable for not well equipped microfluidics labs. Due to fabrication difficulties in bonding glass substrates, several hybrid glass/PDMS microfluidic chips with integrated electrodes were developed [19–21]. In those microfluidic chips, glass and PDMS are used as a carrier of an electrode array and microfludic channels, respectively. Bonding of those materials relies on traditional surface plasma activation which along with the flexibility of PDMS provides a good way of sealing microfluidic channels on glass substrates with uneven surface caused by the presence of electrodes. The group of Li [19] used a layer of SU8 photoresist coated on a silicon substrate to facilitate the bonding of PDMS by activation in ozone. Group of Shu-Ming prepared a hybrid glass/ PDMS chip by using UV curable glue as an intermediate layer [20]. The use of these hybrid chips, however, is limited to low temperature and low pressure applications. Also different surface properties of materials, e.g., a wettable glass and a non-wettable PDMS might introduce unexpected issues in some applications. In this paper, we present a fast and inexpensive method for the fabrication of whole-glass microfluidic chips with integrated metal electrode arrays. Our fabrication method requires minimum number of expensive devices and can be performed with standard laboratory equipment.
2. Experimental 2.1. Fabrication procedure The entire fabrication procedure is illustrated in Fig. 1. Microscopic glass slides (75 by 25 mm) were used as glass substrates. Electrode arrays were made by the standard process based on the combination of metal sputtering and photolithography (photoresist P-ma 1275 by Microresist) (Fig. 1A). In the first stage, electrodes made of inexpensive nickel were used to develop and optimize the whole fabrication procedure. Nickel was later replaced by more expensive gold. The electrode array was designed as a four electrode array with the following parameters: width of the electrodes 200 lm, the gap between the electrodes 400 lm. In order to investigate the influence of the metal layer thickness on the quality of bonding, we prepared electrode arrays with different thicknesses ranging from 200 to 700 nm. Microfluidic channels were etched into a second glass slide (Fig. 1B) through a structured polyester tape (type M42 by Europack Chrudim) in a solution made up of hydrofluoric acid, sulfuric acid, and DI water mixed in the volumetric ratio 1:2:1 [22]. The polyester tape serves as a mask for etching of the required structures. The tape was cut manually using a scalpel. Alternatively, a cutting plotter can be used [23]. The etching was performed in a laboratory shaker (IKA KS125) set to 300 cycles/min at room temperature. The etching rate was 5 lm/min. Via-holes located at the beginning and at the end of the microchannel were drilled by diamond coated ball burrs (by Dremel Motoflex) under water to prevent the local overheating and possible breakage of the glass substrate. The glass substrate bearing the microelectrodes is coated with a non-diluted solution of sodium silicate (10% NaOH and 26% SiO2 – known as water glass, by Sigma Aldrich) (Fig. 1C), and immediately covered with the glass substrate containing the microfluidic channel (Fig. 1D). The excess of sodium silicate located along the perimeter of the contacted glass substrates is rinsed off by DI water. The microchannel filled with the water glass is then connected to a
A B C D
E
F
Fig. 1. Scheme of the whole fabrication process of the glass microfluidic chip with integrated electrode array: (A) glass slide with an electrodes array, (B) glass slide with a microfluidic channel, (C) coating of the glass slide with sodium silicate, (D) attachment of the mating glass slide, (E) removal of the excess of sodium silicate by vacuum suction and rinsing with DI water and (F) drying of sodium silicate at 95 °C for 60 min.
vacuum pump (KNF N816.3KT) that removes the water glass from the channel. To ensure that all water glass is removed from the channel and the surface of the electrodes is free of any sodium silicate solution, we flush the channel with DI water (approximately 2 ml) while the channel is still hooked up to the vacuum (Fig. 1E). The assembled chip is then put in an oven set at 95 °C for 60 min to dry the remaining water glass which bonds the two glass substrates together. Fig. 2A and B shows the fabricated glass chip. 2.2. Confirmation of glass bonding To prove that the bond in glass/metal/glass sandwich is defectfree, we filled the micro channel with a fluorescein solution to trace any possible defects. Inspection was done visually after the chip illumination by UV light (using microscope OLYMPUS BX51WI/U-RFL-T). 2.3. Electrochemical impedance spectroscopy measurements An important step in the fabrication procedure is flushing of the microfluidic channel. Flushing assures that the surface of the electrode array is free of any sodium silicate. To test that the electrodes are fully functional, we measured electrochemical impedance spectra of KCl solutions with different concentrations. Electrochemical impedance spectroscopy (EIS) measures the impedance of the electrochemical cell in the dependence on frequency of a low amplitude AC electrical signal. The results from EIS are traditionally plotted as a dependence of total impedance on the frequency (Bode plot) or real and imaginary parts of impedance are
J. Kotowski et al. / Microelectronic Engineering 110 (2013) 441–445
443
2.4. Determination of the critical dimensions of microchannel The technology for bonding two glass substrates presented in this paper is limited to fluidic channels from which the solution of sodium silicate can be removed by a vacuum pump. To determine the critical dimensions and cross-section of our fluidic channel which can be successfully bonded with the use of sodium silicate, we made a set of 5 cm long straight channels with different channel heights and widths and tested them in our bonding experiment. The width and the height of the channel were determined by an optical microscope. The critical dimensions obtained for 5 cm long straight channel have informative character only. In general, critical dimensions will depend on the channel length, its curvature and also viscosity/concentration of sodium silicate solution. Longer and curved channels bonded with sodium silicate solution of larger viscosity will have larger critical dimensions than those determined here. To see the effect of sodium silicate viscosity on critical dimensions of the channels as well as on the quality of the created bond we conducted experiments with diluted solutions of sodium silicate. The sodium silicate solutions were prepared by mixing the stock solution with DI water in the following sodium silicate to DI water volume ratios: 2:1 (sol a), 3:2 (sol b) and 1:1 (sol c). 3. Results and discussions 3.1. Confirmation of bonding
Fig. 2. (A) The glass mifrofluidic chip with 200 nm thick gold electrodes and (B) The glass microfludic chip with 200 nm thick nickel electrodes and the channel connected to the connection tubes compared to the size of two euro coin.
displayed (Nyquist plot). Equivalent electrical circuits are often used to extract data of interest, e.g., resistance (conductivity) of a solution, capacity of electrical double layer at electrode–electrolyte interfaces etc. Randles circuit is the simplest but frequently used equivalent electrical circuit (see Fig. 3). It consists of four elements: (i) resistance of the electrolyte between the electrodes Rs, (ii) resistance associated with electron transfer reaction (electrochemical reaction) Rct, (iii) capacitance of the electrical double layer at electrode–electrolyte interface Cdl, and (iv) constant phase element associated with diffusion transport of electroactive species (Warburg impedance) Zw. Here we show the EIS results in the form of the Bode plot without further analysis of the obtained data. The EIS measurements were carried out in a two-electrode set up. The channel was filled with a KCl solution of various concentrations (1 M, 0.1 M and 0.01 M). We measured the dependence of impedance on the frequency of the applied sinusoidal signal with amplitude of 0.01 V. The frequency ranged from 1 to 105 Hz. The measurement was performed with a potentiostat (AUTOLAB PGSTAT 302N) using the NOVA 1.8 software.
Fig. 4A shows a microfluidic chip with 200 nm thick electrodes made with the sodium silicate assisted bonding technique. The green/black interface identifies the channel walls, the four whitedotted strips across the channel are metal electrodes. The microfluidic channel is well-sealed in the area of the electrode array. No leaking fluorescein was observed along the electrodes or between the glass substrates. The same quality of the bonding was also observed for all thicker electrodes layers up to 700 nm. The rest of the experiments confirming the functionality of the electrode array and showing the effect of sodium silicate concentration/viscosity on the quality of the bonding were carried out with the 200 nm thick electrodes. The chips with 200 nm thick electrodes assembled by using solution sol a (DI water:sodium silicate = 1:2) proved the same quality of the sealing as non-diluted sodium silicate solution. The use of sol b solution (DI water:sodium silicate = 2:3) led to a sealed channel in quality comparable to that of undiluted sodium silicate. However, during the assembly of both parts of the chip (glass with the microchannel, glass with electrode array), a significant amount of air bubbles was trapped between the glass slides. The air bubbles prevented the proper bonding over the surface of the two glass pieces. The presence of the bubbles significantly complicated the fabrication process and the assembly process had to be repeated many times until no air bubbles were present in the sodium silicate solution layer. We were not able to fabricate a chip with sufficient sealing quality using the solutions (DI water:sodium silicate = 1:1) (Fig. 4B). The fluorescein tracer leaked along the electrodes and in between the glass slides. A spontaneous disintegration of these glass chips was observed after three hours. 3.2. Verification of electrode array functionality
Fig. 3. Randles circuit as a equivalent electrical circuit for electrochemical impedance spectroscopy.
We measured EIS spectra of KCl solutions as described in Section 2 for both nickel (Fig. 5A) and gold (Fig. 5B) electrode arrays. The shape of the EIS curves in the graphs is qualitatively the same as is described in [24]. For higher frequencies (104–105 Hz), there is
444
J. Kotowski et al. / Microelectronic Engineering 110 (2013) 441–445
Fig. 4. The detail of the microchannel (with electrodes) filled with the fluorescent tracer observed under the UV light, (A) channel fabricated using undiluted sodium silicate is properly sealed, width of the channel is 954 lm. (B) Channel fabricated using diluted sodium silicate in ratio 1:1 with DI water. Fluorescent tracer leaks out ofthe channel. Width of the channel is 417 lm.
Impedance (Ω)
10
10
10
10
7
7
6
Au electrode
6
10
5
5
10
4
4
10
3
3
2
10 −1 10
C
B 10
Ni electrode
Z(Ω)
A 10
10
c = 1.00 M KCl c = 0.10 M KCl c = 0.01 M KCl 0
10
10
1
2
3
10 10 Frequency (Hz)
10
4
10
10 −1 10
10
0
1
10
10
2
10
3
4
10
10
5
Frequency (Hz)
9
10
2
5
c = 1.00 M KCl c = 0.10 M KCl c = 0.01 M KCl
D
Au electrode - without rinsing
8
7
10
Au electrod - after rinsing
6
10
10
7
5
10
Z(Ω)
Z(Ω)
10
6
4
10
10
5
3
10
4
10 0 10
10 c = 1.00 M KCl c = 0.10 M KCl c = 0.01 M KCl 10
1
10
2
2
10
3
10
4
5
10
10
6
Frequency (Hz)
10 0 10
c = 1.00 M KCl c = 0.10 M KCl c = 0.01 M KCl 10
1
10
2
10
3
10
4
5
10
6
10
Frequency (Hz)
Fig. 5. The dependence of impedance on the frequency of the applied sinusoidal signal with the amplitude of 0.01 V for three various KCl solution concentrations 1.0, 0.1 and 0.01 M KCl. (A) Microfludic chip with nickel electrode array, (B) microfludic chip with gold electrode array, (C) microfludic chip with gold electrode array fabricated without the flushing step (step e in Fig. 1), (D) microfludic chip with gold electrode array as presented in the Fig. 4C subsequently flushed with DI water.
a significant difference between the impedance curves for various electrolyte concentrations. KCl solutions of higher concentrations gave lower impedance at the same frequency. This part if the impedance spectra can be used for extraction of information on conductivity of electrolyte in the channel. On the other hand, the impedance curves collapse onto each other for higher frequencies. The dependence on concentration is surprisingly completely lost
showing that there is another effect which controls the behavior of the system in the low frequency regime. We suspect that this effect might be related to the geometry of electrodes with a meshlike pattern. To show that the flushing step is necessary, we carried out an EIS measurement in a microfluidic chip fabricated without rinsing the channel with water to remove sodium silicate from the surface
J. Kotowski et al. / Microelectronic Engineering 110 (2013) 441–445 Table 1 Critical dimensions of microchannel. Dilution ratio (sodium silicate:water)
Width (lm)
Depth (lm)
Cross-section (lm2)
0:1 2:1 3:2
425 350 310
20 12 10
8500 4200 3100
of electrodes. There is no significant frequency dependence of the impedance for the various KCl solutions both for lower and higher frequency signals (Fig. 5C). The same chip that was fabricated without flushing of the channel with DI water was subsequently flushed with DI water several times and the EIS measurement was repeated Fig. 5D. The shape of impedance curves in Fig. 5D is analogous to the impedance curves in Fig. 5B. Comparison of Fig. 5C and D shows that the sodium silicate, which covered the electrodes, can be dissolved and flushed from the channel even after the chip is completely fabricated and dried. However, it is desirable to perform the flushing immediately after bonding the two glasses together otherwise the remaining sodium can glue the channel. This experiment led us to the assumption that it is possible to dissolve the sodium silicate in the microfluidic channel. Therefore we kept the whole microfluidic chip with electrode array immersed in DI water for one week. No defect in the sealing/bonding of the microfluidic chip was observed. 3.3. Determination of the critical cross-sections The critical cross-sections of 5 cm long, straight channels for different sodium silica solutions were determined according to Section 2.4. The dimensions of channels tested are summarized in Table 1. The values given in the table are averaged values from two/three measurements taken at different positions along the channel (beginning, middle, and end). We found that the increasing dilution ratio of sodium silicate allows the reduction of the critical dimensions (i.e., cross-section). However the dilution also increases difficulty (e.g., error rate) during fabrication procedure. 4. Conclusions We showed that bonding of glass substrates by means of nondiluted sodium silicate solution represents a simple yet reliable method for low temperature fabrication of glass microfluidic chips with integrated electrode arrays. This bonding technique does not require any expensive commercial apparatuses or expensive material. The cost as well as very fast preparation time (3–4 h) makes our technique a suitable candidate for prototyping. We optimized this fabrication process for electrodes with the thickness up to 700 nm. We verified full functionality of electrodes by running an electrochemical impedance spectroscopy measure-
445
ment. This measurement also revealed the importance of rinsing the channel with sufficient amount of DI water to remove all sodium silicate from the surface of electrodes. The critical microchannel cross-section of 8500 ± 10% lm2 was evaluated for a 5 cm long straight microchannel when using non-diluted sodium silicate solution. We successfully fabricated microfluidic glass chips sealed by a sodium silicate solution diluted with water in a volume ratio of 3:2 and obtained a defect-free bond between glass/metal/glass layers. The use of lower viscosity sodium silicate solution led to the decrease in the critical cross-section to 3100 ± 10% lm2 for the 5 cm long straight channel. Acknowledgement The authors thank for the financial support by the Grant of the GAAV CR IAA401280904 and by the specific university research (MSMT No. 20/2013). The result was developed within the CENTEM project, Reg. No. CZ.1.05/2.1.00/03.0088, co-funded by the ERDF as part of the Ministry of Education, Youth and Sports’ OP RDI programme. References [1] B. Qu, Z. Wu, F. Fang, Z. Bai, D. Yang, S. Xu, Anal. Bioanal. Chem 392 (2008) 1317–1324. [2] W. Kubicki, R. Walczak, J. Dziuban, Optica Appl. XLI (2) (2011). [3] K. Suk Yang, H.J. Kim, J.K. Ahn, D.H. Kim, Curr. Appl. Phys. 9 (2009) e60–e65. [4] M.S. Arefin, T.L. Porter, J. Appl. Phys. 111 (2012) 054919. [5] M. Svoboda, Z. Slouka, W. Schrott, D. Šnita, Microelectron. Eng. 86 (2009) 1371–1374. [6] P. Abgrall, A.-M. Gue, J. Micromech. Microeng. 17 (2007) R15–R49. [7] M. Sadrzadeh, K. Shahidi, T. Mohammadi, J. Appl. Poly. Sci. 117 (2010) 33–48. [8] A. Abou-Hassan, O. Sandre, V. Cabuil, Angew. Chem. Int. Ed. 49 (2010) 6268– 6286. [9] S. Marre, A. Adamo, S. Basak, C. Aymonier, K.F. Jensen, Ind. Eng. Chem. Res. 49 (2010) 11310–11320. [10] W. Verboom, Chem. Eng. Technol. 32 (11) (2009) 1695–1701. [11] F. Trachsel, C. Hutter, P.R. von Rohr, Chem. Eng. J. 135S (2008) S309–S316. [12] C. Iliescu, H. Taylor, M. Avram, J. Miao, S. Franssila, Biomicrofluidics 6 (2012) 016505. [13] S. Queste, R. Salut, S. Clatot, J.-Y. Rauch, C.G.K. Malek, Microsys. Technol. 16 (2010) 1485–1493. [14] L. Chen, G. Luo, K. Liu, J. Ma, B. Yao, Y. Yan, Y. Wang, Sens. Actuators, B 119 (2006) 335–344. [15] H.Y. Wang, R.S. Foote, S.C. Jacobson, J.H. Schneibel, J.M. Ramsey, Sens. Actuators B 45 (1997) 199–207. [16] P. Maoa, J. Han, Lab Chip 5 (2005) 837–844. [17] Y. Xu, C. Wang, Y. Dong, L. Li, K. Jang, K. Mawatari, T. Suga, T. Kitamori, Anal. Bioanal. Chem. 402 (2012) 1011–1018. [18] C.S. Henry, Microchip Capillary Electrophoresis, Methods and Protocols, vol. 339, 2006, pp. 13–26. [19] P. Li, N. Lei, D.A. Sheadel, J. Xu, W. Xue, Sens. Actuators B 166–167 (2012) 870– 877. [20] S. Kuoa, C. Yanga, J. Shieab, C. Lin, Sens. Actuators B 156 (2011) 156–161. [21] S.M. Shameli, T. Glawdel, Z. Liu, C.L. Ren, Anal. Chem. 84 (2012) 2968–2973. [22] O.N. Dyatlova, V.V. Bykov, Glass Ceram. 19 (2) (1962) 77–81. [23] Po Ki Yuen, Vasiliy N. Goral, J. Chem. Educ. 89 (10) (2012) 1288–1292. [24] J. Janouš, P. Beranek, M. Pribyl, D. Šnita, Microelectron. Eng. 97 (2012) 387– 390.