Thermal assisted direct bonding between structured

0 downloads 0 Views 253KB Size Report
glass chips sealing. For example, by anodic bonding, all fabrication steps must be carried out under clean room conditions, which, together with the very strict ...
Microsyst Technol DOI 10.1007/s00542-009-0911-5

TECHNICAL PAPER

Thermal assisted direct bonding between structured glasses for lab-on-chip technology Qiuping Chen Æ Qiuling Chen Æ Daniel Milanese Æ Monica Ferraris

Received: 29 June 2009 / Accepted: 24 July 2009 Ó Springer-Verlag 2009

Abstract A thermal assisted direct bonding (TADB) technique between structured glasses is proposed for microfluidic device fabrication. The bonded glass pairs were characterized by optical microscopy, scanning electron microscopy (SEM), apparent shear strength tests and Vickers hardness measurements across the bonded interface. The optimisation of TADB parameters on flat glasses and on structured glasses was analysed. This technique is user-friendly and low cost, and can be considered for mass production of glass-based micro-fluidic devices.

1 Introduction Microfluidics, as a technology for lab-on-chip devices, is an emerging science which deals with volumes of fluid of the order of magnitude of nanoliters or picoliters, which have to flow through micro channelled chips (Werdich et al. 2004). Among chips of different materials, due to its unique properties such as good chemical and thermal stability and easy surface modification, glass chips have great advantages over the chips of other substrates (Mijatovic et al. 2005; Czaplewski et al. 2003). For example polymers have poor temperature and chemical resistance; cost and incompatibility with certain biological fluids limit the silicon chips to few applications; metal-based devices have poor corrosion resistance and they are typically not compatible with some biological fluids. Q. Chen (&)  Q. Chen  D. Milanese  M. Ferraris Dipartimento di Scienza dei Materiali ed Ingegneria Chimica, Politecnico di Torino, C.so Duca degli Abruzzi 24, Turin, Italy e-mail: [email protected] Q. Chen e-mail: [email protected]

Even though success has been achieved with microfluidic devices based on glass chips, there are several limitations involved with the fabrication techniques: i.e. anodic bonding or fusion bonding related limitations for glass chips sealing. For example, by anodic bonding, all fabrication steps must be carried out under clean room conditions, which, together with the very strict requirements on surface roughness add considerable overhead cost to the production (Gui et al. 1997). A thermal assisted direct-bonding technique (TADB) in place of anodic bonding or fusion bonding was proposed and studied by the authors on two flat rare-earth doped multi-component silicate glasses (Chen et al. 2007, 2008) for waveguide lossless splitter fabrication. The same technique has been optimised and results are reported in this paper, on glass-based microfluidic bonding: the objective of microfluidic bonding is to seal the microchannels on one glass chip with another glass chip in order to keep the nanoliters or picoliters fluids inside the device or to make them flow through the micro-channels, i.e. for chemical analysis. Through the combination of surface treatment (cleaning and activation) and low temperature thermal annealing under slight pressure (kPa), typical of the TADB technique strict requirements that are needed by anodic bonding (e.g. clean room conditions and high glass surface quality) are not required. A good interface quality between the bonded glass chips is easily obtained and at low cost. Compared with TADB between flat (non-structured) glasses, the bonding technique between structured glasses is more difficult to obtain, because of the structures existing on glass surface: first, the structures (micro-channels) weaken the surface hydrophilic/activation effect; second, they decrease the contact area for bonding. Together they result in the decrease of TADB bonding efficiency

123

Microsyst Technol

e.g. decreasing the bonding strength. Moreover, besides the requirement of good interface and bonding strength, the TADB of glasses for microfluidics must guarantee the integrity and uniformity of the micro-channels. The microchannels should be kept open and not to be blocked or deformed after the TADB. In this work, based on our previous studies (Chen et al. 2007, 2008, 2009), the TADB was employed on flat and structured (micro-channelled) soda-lime glass chips for micro-fluidic devices fabrication. The optimum parameters of TADB on structured glasses were determined through the study of bonding temperature and pressure. The bonded glass pairs were characterized by optical microscopy, SEM, apparent shear strength tests and Vickers hardness measurements across the bonded interface.

2 Experiments Commercial soda-lime glasses were employed in this study: the thermal properties such as glass transition (Tg) and softening temperature (Tsoft) were measured using Perkin-Elmer DSC7 and TMA7. The glass samples were cut in 20 mm 9 20 mm slices and optically polished by Logitech PM5. The average roughness of all the glass samples was measured to be around 50 nm by profilometer P15 KLA-Tencor. The structured glasses used for TADB for microfluidic devices were obtained by hot-imprinting technique (Chen et al. 2009). The channels’ width was 100 lm, and the depth was 50 lm. After pre-cleaning with suitable wax removing solvent (acetone, isopropanol), all polished samples were immersed into a solution of NH4OH:H2O2:H2O (0.5:1:5) at 75°C for 10 min and finally rinsed with deionized water for 2 min to remove the residual chemicals. After cleaning, samples were dipped into solution H2SO4:H2O2 (3:1) (Chen 2008) at 75°C for 30 min, and then transferred into 75°C solution of (65%) HNO3 for 30 min after DI-water rinsing for surface activation, finally samples were rinsed and dried with N2 gas for 2 min. All the operations were carried out in a ventilated clean room. After surface activation, the samples to be bonded were brought into contact at room temperature and finger pressure applied in order to promote attractive forces to draw the surfaces together, forcing out any excess air or liquid. Subsequently, a thermal treatment was performed under different bonding temperatures (Tb) with different pressures in order to get an optimum bonding quality. Based on previously study, the bonding time and cooling rate were

123

set as 240 min and 5°C/min, respectively and the bonding temperature between 550 and 610°C (Chen et al. 2008). The bonding temperature and pressure employed for TADB of flat and structured glasses are shown in Tables 1 and 2, respectively. The bonded glass pairs were characterized by optical microscope (Reichert-Jung MeF3), SEM (FEI, QUANTA INSPECT 200), apparent shear strength tests (single lap in compression, adapted from ASTM D1002-05, sample size is 12 mm 9 6 mm) (Sintech 10/D) and Vickers microhardness across the bonded interface (200 g load for 15 s) (Leica, Leitz).

3 Results An almost joint-less interface was obtained after optimising TADB parameters of two flat soda-lime glasses as shown in Fig. 1, where the bonded area is underlined: the interface is uniform and no cracks or bubbles were observed along it. Micro-Vickers cracks (Fig. 2) propagate across and not along the joint of a flat bonded glass pair under optimum bonding conditions, suggesting a good bonding strength. The apparent shear strength of these bonded glasses measured by single lap in compression on four samples gave an average value of 33 MPa. Bubble/cracks free interface and channels in the glasses were kept open with good integrity and uniformity for the TADB of structures soda-lime glasses. Figure 3 (SEM) shows the cross section of the bonded interface under different magnifications: micro-channels were not deformed or filled after bonding. The apparent strength for the optimized TADB structured soda-lime glasses gave an average value of 32 MPa, measured on six samples.

4 Discussion To optimize the bonding process, reactions at the interface between the two glasses during TADB have been investigated. The activated glass surface is covered by OH groups (Lee et al. 1999), and therefore hydrogen bonds will form when two glasses are put in contact at room temperature, with finger pressure. These hydrogen bonds will link the two surfaces across the interface, and increases the bond strength (Lai et al. 2004) as in Eq. 1. Si - OH þ HO - Si $ Si - O - Si þ HOH

ð1Þ

This reaction in Eq. 1 is reversible at temperature lower than 425°C if water is present. To obtain strong siloxane bonds (Si–O–Si), water must be removed. Thermal treatment can remove the water molecules and

Microsyst Technol Table 1 Study of conditions (temperature and pressure) for TADB of flat sodalime glass

Table 2 Study of conditions (temperature and pressure) for TADB of structured sodalime glass

TADB pressure P(kPa) Tb = 550°C

TADB temperature Tb(°C)P = 28 kPa

Samples

P

Samples

T

SD1, SD2

20

SD11, SD12

550

SD3, SD4

30

SD13, SD14

560

SD5, SD6

24

SD15 SD16

570

SD7, SD8

26

SD17, SD18

580

SD9, SD10

28

SD19, SD20

590

TADB pressure P(kPa) Tb = 580°C

TADB temperature Tb(°C)P = 18 kPa

Samples

P

Samples

T

SB1

28

SB3–SB4

580

SB2

23

SB9–SB10

590

SB3–SB4

18

SB11–SB12

600

SB5–SB6

15

3B13–3B14

610

Fig. 1 Interface of a flat bonded glass pair under optimum bonding conditions

the bonding energy will increase with the annealing temperature (Tong and Gosele 1999). In this study, finger pressure was applied on two hydrophilic glass surfaces at room temperature and it forced the surfaces to form hydrogen bonds and Van der Waals bonds. During the subsequent thermal process, slight pressure (kPa, bonding pressure) also was applied on the samples to be bonded: it can enhance the removal of water molecules as mentioned above and promote the Si–O–Si bond formation (Tong and Gosele 1999). However, the bonding pressure must be carefully optimised: a too high bonding pressure can induce stresses on the two bonded surfaces and cause cracks when they are cooled down to room temperature.

Fig. 2 Vickers cracks propagate across and not along the joint of a flat bonded glass pair under optimum bonding conditions

The Tg of the employed glass was measured to be 548°C and the softening point is 716°C. In order to optimize TADB parameters for flat glasses, TADB was carried out under different pressures ranging from 20 to 30 kPa, by keeping constant the bonding temperature (550°C, around Tg), then with Tb between 550 and 590°C by keeping constant the bonding pressure (28 kPa) (Table 1). It was found that glasses bonded at temperature 550, 560 and 570°C showed poor bonding strength and they cannot withstand cutting and polishing processes. Glasses bonded at 580 and 590°C exhibited good bonding strength, but at temperature higher or equal to 590°C an unacceptable deformation of the sample due to

123

Microsyst Technol Fig. 3 Cross section of microchannelled soda-lime glasses after optimised TADB measured by SEM

Fig. 4 Comparison of bonded structured soda-lime glasses cross section after TADB under different pressure

the applied pressure (28 kPa) at T [ Tg was observed; it is well known that when a glass is heated above its glass transition temperature and under pressure, its viscosity quickly decreases and it becomes easy to deform. Glass samples bonded under pressure of 28 kPa at 580°C exhibited good bonding strength without any deformation: these parameters have been considered as the optimized ones for flat glass bonding. Based on study of TADB for flat glasses, the temperature for structured glass bonding was initially kept at 580°C. (Table 2) and the bonding pressure ranged between 15 and 28 kPa. For TADB on structured glasses the bonding under pressure can cause channels distortion. Therefore in order to keep the structures’ integrity, the pressure for TADB is critical and must be well controlled. When the TADB was carried out under higher pressure such as 28 and 23 kPa, the channels were deformed as can be seen in Fig. 4. Low pressure such as 13 and 15 kPa

123

resulted in a poor bonding strength and the bonded samples can not withstand the cutting process. Bonding pressures of 18 kPa at 580°C gave a good compromise. Hence, shear strength and Vickers test were performed on the samples bonded by these parameters: results show that cracks appeared along the interface when Vickers indentation of 200 g was applied as shown in Fig. 5, which suggested that the bonding temperature of 580°C and 18 kPa were still unsuitable for obtaining good bonding strength. Other three temperatures (590, 600 and 610°C, at 18 kPa, Table 2) were tested in order to find the optimised Tb for structured soda-lime glass bonding. Cracks occurred along the channel/interface for the some samples bonded at 590°C during cutting. Similar results were obtained by Vickers test on non-cracked samples. It is found that 18 kPa at 600°C were the most suitable parameters with which the channels on glass were kept open after the TADB. In addition, the interface was free of bubbles or cracks.

Microsyst Technol

micro-channels and without, were successfully bonded with sound interface, good bonding strength and open channels. The optimised conditions for TADB both on flat and on structured glasses were found and discussed. Strict requirements on surface roughness and clean room conditions were not necessary, thus reducing the cost of devices obtained by this technique. Acknowledgments This work has been partially funded by AsiaLink EU (ASIA-LINK-CN/ASIA-LINK/004 (81206)) project. M.Salvo is kindly acknowledged for mechanical tests.

References

Fig. 5 Vickers crack propagation on TADB micro-channelled sodalime glasses. (18 kPa, at 580°C)

However, when the bonding temperature was further increased to 610°C, it was found that some channels were deformed. The optimised bonding temperature for structured glasses were 600°C with a bonding pressure of 18 kPa, while the Tb for flat glasses of the same composition was 580°C with a bonding pressure of 28 kPa. A possible explanation for this difference can be that the micro-channels on the glass surface weaken the surface hydrophilic/activation effect and decrease the contact area for bonding. Both these features result in the decrease of TADB bonding efficiency, e.g. decreasing the bonding strength. In order to obtain the same good bonding as with the flat glasses, the bonding temperature Tb has been increased for the structures glasses and the bonding pressure decreased, to avoid deformation of the micro-channel. Moreover, a lower pressure induces lower stress on the two glass surfaces, which can be released. In addition, channels can dissipate more efficiently stresses caused by higher temperature (Lapides et al. 2004).

5 Conclusion Using the optimised TADB parameters, i.e. bonding temperature and bonding pressure, soda-lime glasses both with

Chen Q (2008) Study of thermal assisted direct bonding techniques for optical waveguide and micro-fluidic device fabrication. PhD thesis, Politecnico di Torino, Italy Chen Q, Chen Q, Milanese D, Ferraris M, Fokine M (2007) Direct bonding and imprinting techniques for micro-fluidic devices fabrication in commercial soda-lime glass, Patent N. TO2007A000345 Chen Q, Milanese D, Chen Q, Ferraris M et al (2008) Fabrication and direct bonding of photosensitive multi-component silicate glasses for lossless planar waveguide splitter. J Non-Cryst Solids 354:1230–1234. doi:10.1016/j.jnoncrysol.2006.11.045 Chen Q, Chen Q, Milanese D, Ferraris M (2009) Micro-structures fabrication on glasses for micro-fluidics by imprinting technique. J Microsyst technol 15:1067–1071 Czaplewski DA, Kameoka J, Mathers R, Coates et al (2003) Nanofluidic channels with elliptical cross sections formed using a nonlithographic process. Appl Phys Lett 83:4836–4838 Gui C, Albers H, Gardeniers JGE (1997) Fusion bonding of rough surfaces with polishing technique for silicon. J Microsyst Technol 3:122–128 Lai SI, Lin HY, Hu CT (2004) Effect of surface treatment on wafer direct bonding process. Mater Chem Phys 83:265–272 Lapides AA, Botvinkin MI, Zhabotinskii ME (2004) New method to determine the surface tension and viscosity of glass. J Chem Mater Sci 38:276–278 Lee DJ, Byeong Kwon Ju, Jang Jin (1999) Effects of a hydrophilic surface in anodic bonding. J Micromech Microeng 9:313–318 Mijatovic D et al (2005) The promise of nanotechnology for separation devices–from a top-down approach to nature-inspired separation devices. Lab Chip 5:492–500 Tong Q-Y, Gosele UM (1999) Wafer bonding and layer splitting for Microsystems. J Adv Mater 17:1409–1425 Werdich AA, Lima EA, Ivanov B, Ges I et al (2004) Microfluidic device to confine a single cardiac myocyte in a sub-nanoliter volume on a planar microelectrode array for cellular activity detection. Lab Chip 4:357–362

123