Microfluidics for High Pressure: Integration on GaAs Acoustic Biosensors with a Leakage-Free PDMS Based on Bonding Technology

Microfluidics integration of acoustic biosensors is an actively developing field. Despite significant progress in “passive” microfluidic technology, integration with microacoustic devices is still in its research state. The major challenge is bonding polymers with monocrystalline piezoelectrics to seal microfluidic biosensors. In this contribution, we specifically address the challenge of microfluidics integration on gallium arsenide (GaAs) acoustic biosensors. We have developed a robust plasma-assisted bonding technology, allowing strong connections between PDMS microfluidic chip and GaAs/SiO2 at low temperatures (70 °C). Mechanical and fluidic performances of fabricated device were studied. The bonding surfaces were characterized by water contact angle measurement and ATR-FTIR, AFM, and SEM analysis. The bonding strength was characterized using a tensile machine and pressure/leakage tests. The study showed that the sealed chips were able to achieve a limit of high bonding strength of 2.01 MPa. The adhesion of PDMS to GaAs was significantly improved by use of SiO2 intermediate layer, permitting the bonded chip to withstand at least 8.5 bar of burst pressure. The developed bonding approach can be a valuable solution for microfluidics integration in several types of MEMS devices.


Introduction
Microfluidics field has emerged as a solution for the precise control and manipulation of fluids at microliter scales [1][2][3]. On-chip microfluidics integration is one of the most promising development vectors, particularly in the field of biosensors [4,5]. Microfluidicsintegrated lab-on-a-chip solutions have been widely used in many applications, such as clinical diagnostics on human physiological fluids, cell biology [6], detection of tumor cells, biochemical detections, electrophoresis, biochemistry, PCR [7], DNA analysis, singlecell trapping, droplets microfluidics [8], biosensors, and more [9]. Recently introduced microfluidic biosensors have the advantages of portability, high precision, easy application, and high-throughput parallel processing [10]. Ma et al. [11] showed that electrochemical biosensing made in microfluidic channels could lower the detection limit of endotoxin with a confined space and enhance van der Waals forces. They used a confined microfluidic channel and continuous flow forced the target molecules to bind to the sensing surface for fast preconcentration to enhance sensitivity and shorten detection times. Zhang et al. [12] used microfluidic channels with biosensors for detection of Salmonella using Fe-nanocluster amplification and smart phone imaging.
Micromachining processes open up the possibility to combine microsensors and microfluidics onto a single chip. Various sensing technologies have been integrated in order to improve biosensors. To address the challenge, some authors have attempted to increase the adhesion between the PDMS and the substrate by using a gold layer. However, this method is not suitable for seal-patterned PDMS with micromachined GaAs. Others have proposed 3-aminopropyl)trimethoxysilane (APTES) and achieved a bonding strength of 406 kPa for PMMA/PDMS bonding [35,36]. A thermal bonding method is used to bond four-layer microfluidic chip [37]. In the other approach, PDMS is mixed with a small amount of polyethylenimine solution to prepare a sticky thin layer, which works similarly to a sticky tape to adhere on glass, PMMA, and metal by contact pressing [38]. Lastly, Anil et al. [39] developed microfluidics-integrated microscale that comprise an isoporous nanostructured membrane. The GaAs on Ge/Si substrate was first flipped with the GaAs nanopyramids side bonded to a Polydimethylsiloxane (PDMS) substrate, whereas a transparent flexible polymer film was weakly bonded by Van der Waals forces [40,41]. The SiO 2 layer was used to increase PMMA bonding capability to PDMS in fabricating gas micro valves. Ahmad et al. showed a strong and irreversible bond of PDMS on PMMA when it is covered with SiO 2 nano particles [42].
In this study, we developed the approach for irreversible and leakage-free plasma assisted bonding to integrate PDMS microfluidic channel on GaAs substrate. For this purpose, we combined the thin-film SiO 2 intermediate layer on GaAs substrate with plasma O 2 treatment and low-temperature annealing. In brief, this bonding technology is obtained in four main steps: (1) SiO 2 deposition, (2) plasma O 2 treatment, (3) chip alignment and bonding, and (4) annealing at low temperature 70 • C. This method is appealing for its compatibility with traditional replication methods using PDMS, and the surface structures can be retained. The characterization of PDMS and GaAs/SiO 2 surfaces before bonding was verified by contact angle, Attenuated Total Reflectance-Fourier Transformed Infrared Spectroscopy (ATR-FTIR), AFM, and SEM analysis. Bonded chips were characterized using a tensile machine strength bonding equipment on PDMS and a leakage bench test.

Materials
Polydimethylsiloxane (PDMS, Sylgard 184) was obtained from Dow Corning Toray Corp. SU-8 3025 photoresist was purchased from MicroChem (Newton, MA, USA). Undoped, 3-inch in diameter, and 625 ± 25 µm thick double-side polished GaAs (100) ± 0.5 • wafers (AXT, Inc., Fremont, CA, USA) were used to fabricate biointerface chips. Acetone (ACP Chemicals, Saint-Léonard, QC, Canada) was used to clean substrates. Red color dye was purchased from Shanghai Macklin Biochemical Co., Ltd. (Shanghai, China). Hydrostatic pressure was applied to generate constant flow rate in microfluidic chip. Sharp blade was used to cut PDMS for placing microfluidic chip. SEM microscope (Thermofisher APREO S Low-vacuum SEM and 30 mm 2 SDD EDX) was used to observe the microfluidic channel in detail. AFM scans of different dimensions were recorded in order to have a representative sampling of the surface roughness of GaAs/SiO 2 . The AFM cantilever had a nominal resonance frequency of 330 kHz, a force constant of 42 N/m, a length of 125 µm, and a mean width of 30 µm.

Methods and Equipment
The fabrication process of the PDMS microfluidic channel is schematically illustrated in Figure 1. The SU-8 mold of 70 µm thickness was fabricated with SU-8 3050 onto a 1 mm thick silicon wafer using standard photolithography processes including spin coating, pre-baking, exposure, post-baking, and development. The SU-8 mold was then used to replicate PDMS microfluidic channel. PDMS with thickness of 3 mm was prepared by pouring the mixture of low-modulus PDMS (component ratio A:B = 1:10, Sylgard 184, Dow Corning).
Air bubbles that appeared during the mixing were removed using a vacuum desiccator, followed by baking at 80 • C for 2 h, followed by pouring onto the SU-8 mold to have a sticky layer of about 300 µm. Finally, the prepared PDMS structures were peeled off from the mold and small inlet and outlet holes were punched. Images of the SU-8 mold and the fabricated PDMS microfluidic channel are shown in Figure 1. The resulting channels have a hight of 60 µm and a width of 300 µm. The GaAS surface was covered with the SiO 2 layer by an RF reactive magnetron sputtering MP450S machine (Plassys, France). During the deposition of the SiO 2 layer, a plasma activation process was performed in 6 mTorr, power 350 W, and with 27% oxygen flow.  Air bubbles that appeared during the mixing were removed using a vacuum desiccator, followed by baking at 80 °C for 2 h, followed by pouring onto the SU-8 mold to have a sticky layer of about 300 μm. Finally, the prepared PDMS structures were peeled off from the mold and small inlet and outlet holes were punched. Images of the SU-8 mold and the fabricated PDMS microfluidic channel are shown in Figure 1. The resulting channels have a hight of 60 μm and a width of 300 μm. The GaAS surface was covered with the SiO2 layer by an RF reactive magnetron sputtering MP450S machine (Plassys, France). During the deposition of the SiO2 layer, a plasma activation process was performed in 6 mTorr, power 350 W, and with 27% oxygen flow. Figure 2 presents an experimental setup to measure flow rate and pressure in the microfluidic channel. Once the devices were fabricated and assembled, we connected 25 mL syringes to Tygon (Sigma-Aldrich, Lyon, France) tubing to control the volume of air pumped in and out the control channels. A hydrostatic pressure is mounted to accurately displace the syringe plunger. Additionally, we connected the inlet of the microfluidic device with Tygon tubing to a flow sensor to control the flow rate of our medium and cell sample.  Figure 2 presents an experimental setup to measure flow rate and pressure in the microfluidic channel. Once the devices were fabricated and assembled, we connected 25 mL syringes to Tygon (Sigma-Aldrich, Lyon, France) tubing to control the volume of air pumped in and out the control channels. A hydrostatic pressure is mounted to accurately displace the syringe plunger. Additionally, we connected the inlet of the microfluidic device with Tygon tubing to a flow sensor to control the flow rate of our medium and cell sample. Hence, the pressure difference varies from 0 to 8.5 bar. With the syringe pump single-layer valves in this device can be accurately controlled without using more com cated electropneumatic systems. An LG16 (Sensirion) microfuidic control system Hence, the pressure difference varies from 0 to 8.5 bar. With the syringe pumps, the single-layer valves in this device can be accurately controlled without using more complicated electropneumatic systems. An LG16 (Sensirion) microfuidic control system was used to deliver fluid to the channel of the test device and to monitor the applied pressure. Images of the microfluidic channel were obtained with optical microscope (Mitutoyo FS70, Mitutoyo Corp., Kawasaki, Japan) and a camera (IDS µEye, IDS Imaging Development Systems, Obersulm, Germany) with a spatial resolution of 5.5 µm/pixel. During our experiments, a flow sensor connected to a PC is placed for the continuous recording of the flow rate and pressure flow in the microfluidic channel.

SEM Characterization
The tested PDMS microfluidic structure about 3 mm in thickness was cut along the channel length by a sharp blade. Resultant PDMS membranes with exposed microchannel (see Figure 3A) were coated with Cr thin layer to be analyzed in the SEM microscope. As shown in Figure 3B, the shape of a 400 µm wide microchannel pattern casted in PDMS in a single millimeter scale is well preserved.

Activation-Characterization of PDMS and GaAs/SiO2 Surfaces-Interfaces
The hydrophobicity of PDMS is associated with the organic methyl groups present in the chemical structure of PDMS. The microchannel was cut out of the mold, followed by oxygen plasma treatment to render the PDMS's surface hydrophilic. We prepared our bonding technology by a combination of surfaces treatment and annealing (see Figure 4). Oxygen plasma treatment was demonstrated to be the most rapid process to increase the hydrophilicity of PDMS's surface by removing hydrocarbon groups and introducing polar silanol (Si-OH) groups via oxidation. The activation process duration was 60 s. During To show that the presented bonding method can preserve the channel profile as pure PDMS does, we bonded a PDMS microfluidic structure to a 650 µm thick, (100) oriented GaAs substrate, previously covered with 100 nm SiO 2 layer. Figure 3C shows the cross-sectional profile of the 80 µm high microchannel after bonding, proving that our method of chip packaging with sandwich structure is safe. The two subtrates (GaAs and PDMS) are bonded after plasma oxygen treatment of SiO 2 intermediate thin layer and thermal annealing.

Activation-Characterization of PDMS and GaAs/SiO 2 Surfaces-Interfaces
The hydrophobicity of PDMS is associated with the organic methyl groups present in the chemical structure of PDMS. The microchannel was cut out of the mold, followed by oxygen plasma treatment to render the PDMS's surface hydrophilic. We prepared our bonding technology by a combination of surfaces treatment and annealing (see Figure 4). Oxygen plasma treatment was demonstrated to be the most rapid process to increase the hydrophilicity of PDMS's surface by removing hydrocarbon groups and introducing polar silanol (Si-OH) groups via oxidation. The activation process duration was 60 s. During various bonding tests, we tested bonding at 50 • C and 60 • C; it turned out that the quality of the bonding is not ensured, while for the same conditions, bonding is strong at 70 • C. According to the literature. Bonding strength increases with an increase in temperature [43]; the bonding is due to the interdiffusion of polymer chains. For this, two surfaces were bonded by bringing them into contact followed by a heat treatment at 70 • C for 1 h.

Activation-Characterization of PDMS and GaAs/SiO2 Surfaces-Interfaces
The hydrophobicity of PDMS is associated with the organic methyl groups present in the chemical structure of PDMS. The microchannel was cut out of the mold, followed by oxygen plasma treatment to render the PDMS's surface hydrophilic. We prepared our bonding technology by a combination of surfaces treatment and annealing (see Figure 4). Oxygen plasma treatment was demonstrated to be the most rapid process to increase the hydrophilicity of PDMS's surface by removing hydrocarbon groups and introducing polar silanol (Si-OH) groups via oxidation. The activation process duration was 60 s. During various bonding tests, we tested bonding at 50 °C and 60 °C; it turned out that the quality of the bonding is not ensured, while for the same conditions, bonding is strong at 70 °C. According to the literature. Bonding strength increases with an increase in temperature [43]; the bonding is due to the interdiffusion of polymer chains. For this, two surfaces were bonded by bringing them into contact followed by a heat treatment at 70 °C for 1 h. To characterize the surface modifications of PDMS replica and GaAs/SiO 2 substrate after each step of plasma treatment, contact angles (CA) measurements were performed. Water droplets (5 µL) were deposited on the surface of each studied surface. As shown in Figure 5, the contact angle dropped from 100 • to 53.8 • after the oxygen plasma treatment of PDMS, and from 41.3 • to 11.9 • after the oxygen plasma-treated GaAs/SiO 2 . The drastic decrease in contact angles indicated that the hydrophobic surface of PDMS became hydrophilic due to the hydroxyl terminals on the plasma-activated PDMS's surface. The surface of PDMS after plasma treatment has low surface energy due to the weak intermolecular forces between the methyl groups and the strong (Si-O) and flexible (Si-O-Si) siloxane chain [44].
The surface functional group of GaAs, silicone dioxide, and PDMS were analyzed by using ATR-FTIR in the MIR spectral region from 4000 to 500 cm −1 (λ = 2.5-20 µm) in order to study the effect of oxygen plasma on surface modification. The IR transmittance spectra are presented in Figure 6. The peaks between 2.950 cm −1 and 2.970 cm −1 correspond to the asymmetric Si-OH bonds of PDMS. The peaks at 1.257 cm −1 and 1.010 cm −1 are attributed to CH 3 asymmetric deformation and Si-O-Si asymmetric deformation of PDMS respectively. A high transmittance on the PDMS substrate in a visible light domain is attained. In comparison, the transmittance of the PDMS substrate is 96% and that of the GaAs/SiO 2 is 85%, which results in PDMS transmittance being 11% better than GaAs/SiO 2 .
Water droplets (5 μL) were deposited on the surface of each studied surface. As shown in Figure 5, the contact angle dropped from 100° to 53.8° after the oxygen plasma treatment of PDMS, and from 41.3° to 11.9° after the oxygen plasma-treated GaAs/SiO2. The drastic decrease in contact angles indicated that the hydrophobic surface of PDMS became hydrophilic due to the hydroxyl terminals on the plasma-activated PDMS's surface. The surface of PDMS after plasma treatment has low surface energy due to the weak intermolecular forces between the methyl groups and the strong (Si-O) and flexible (Si-O-Si) siloxane chain [44]. The surface functional group of GaAs, silicone dioxide, and PDMS were analyzed by using ATR-FTIR in the MIR spectral region from 4000 to 500 cm −1 (λ = 2.5-20 μm) in order to study the effect of oxygen plasma on surface modification. The IR transmittance spectra are presented in Figure 6. The peaks between 2.950 cm −1 and 2.970 cm −1 correspond to the asymmetric Si-OH bonds of PDMS. The peaks at 1.257 cm −1 and 1.010 cm −1 are attributed to CH3 asymmetric deformation and Si-O-Si asymmetric deformation of PDMS respectively. A high transmittance on the PDMS substrate in a visible light domain is attained. In comparison, the transmittance of the PDMS substrate is 96% and that of the GaAs/SiO2 is 85%, which results in PDMS transmittance being 11% better than GaAs/SiO2. To ensure the success of surface modification, ATR-FTIR measurements were conducted for six different substrates to verify reproducibility with a maximum error of 5%. Due to the wafer thickness of 650 μm, GaAs disks support only internal reflections. In particular, around 1200 cm −1 , silicon oxide possesses a vibrational mode and, thus, reduces transparency in the so-called molecular fingerprint region. The presence of a peak at 1116 cm −1 corresponds to a thin layer of Si-O. In addition to the transmittance peak at 1018 cm −1 , which relates to methyl groups, there is trace of chloroform in the silicon dioxide intermediate layer. After the treatment of the oxygen plasma, a large amount of hydroxyl groups is produced on the surfaces of both the silicon dioxide and the PDMS. The later conformal contact of the two surfaces will yield a large amount of Si-O-Si bonds, which will form strong adhesion between the two surfaces. A reorganization of the short polymeric chains supporting the creation of polar groups can be considered, resulting in an increase in temperature. One major advantage of the temperature increase is that the polymerization takes place while preserving the functionality of the monomer. Moreover, the increased cross-linking density in the PDMS directly influences the strength of covalent bonding. To ensure the success of surface modification, ATR-FTIR measurements were conducted for six different substrates to verify reproducibility with a maximum error of 5%. Due to the wafer thickness of 650 µm, GaAs disks support only internal reflections. In particular, around 1200 cm −1 , silicon oxide possesses a vibrational mode and, thus, reduces transparency in the so-called molecular fingerprint region. The presence of a peak at 1116 cm −1 corresponds to a thin layer of Si-O. In addition to the transmittance peak at 1018 cm −1 , which relates to methyl groups, there is trace of chloroform in the silicon dioxide intermediate layer. After the treatment of the oxygen plasma, a large amount of hydroxyl groups is produced on the surfaces of both the silicon dioxide and the PDMS. The later conformal contact of the two surfaces will yield a large amount of Si-O-Si bonds, which will form strong adhesion between the two surfaces. A reorganization of the short polymeric chains supporting the creation of polar groups can be considered, resulting in an increase in temperature. One major advantage of the temperature increase is that the polymerization takes place while preserving the functionality of the monomer. Moreover, the increased cross-linking density in the PDMS directly influences the strength of covalent bonding.

Test with Plasma O 2 and Annealing
The surface of GaAs at different fabrication steps was analyzed with AFM within a scanning area of 3 µm × 3 µm. The surface morphology of the silicon dioxide was examined using the AFM, in which the grain sizes can be clearly observed and it is noticed that the surface average roughness was 293 pm. As it can be seen from the AFM images ( Figure 7A,B) the grain size is in the range of 20-58 nm. For a better reading of this paper, surface morphological analyses using AFM are reported for the average of six different scans, with a maximum error of 5%. It was found that the average roughness Ra of SiO2 surface decreased from 552 pm to 441 pm (Table 1) after plasma treatment. It is seen that both the average roughness and root mean square roughness of the PDMS surface are a factor around 465% higher than those of GaAs/SiO2. The obtained results indicate that smoother surfaces of GaAs/SiO2 can be achieved by performing the above-mentioned modification. However, it was shown by Zahid et al. [45] that the plasma O2 treatment of PDMS leads to a significant increase in its surface roughness, from Rq = 1441 pm in the case of fresh PDMS to Rq = 40,031 pm after plasma treatment. The results further confirmed that the PDMS layer adhered on the silicon dioxide wafer during the cast molding process.

Bonding Strength Evaluation
The bonding strength for all substrate was investigated using destructive mechanical tensile test method on square 10 × 10 mm 2 bonded pairs. In addition to the PDMS-SiO2/GaAs bonding configuration, the bonding strength of two other interesting systems, PDMS-Glass and PDMS-LiNbO3, was evaluated. PDMS, glass, and LiNbO3 substrates were cut using diamond saw dicing, whereas GaAs/SiO2 were cut with a cleavage method. All substrates were cleaned with acetone and isopropanol to remove dust and organic contaminants. The bonding strength between the different substrates and PDMS was determined using a mechanical tester for micro-components (Nordson DAGE 4000Plus), equipped with a 250 kg cartridge (50 kg range used). The specific horizontal setups of the tensile test and the assembled sample are shown in Figure 8A. Two steel sample holders were used to sandwich the bonded sample in between. The upper part of PDMS replica and the backside of all substrates were glued on steel sample holders using a single component instant adhesive Loctite 480 (Henkel Adhesives). The sandwiched sample and its holders were fixed horizontally in the special stud-pull fixture of the tester that eliminates side forces due to gimbal construction. During the tensile test, the motorized stage moved It was found that the average roughness Ra of SiO 2 surface decreased from 552 pm to 441 pm (Table 1) after plasma treatment. It is seen that both the average roughness and root mean square roughness of the PDMS surface are a factor around 465% higher than those of GaAs/SiO 2 . The obtained results indicate that smoother surfaces of GaAs/SiO 2 can be achieved by performing the above-mentioned modification. However, it was shown by Zahid et al. [45] that the plasma O 2 treatment of PDMS leads to a significant increase in its surface roughness, from Rq = 1441 pm in the case of fresh PDMS to Rq = 40,031 pm after plasma treatment. The results further confirmed that the PDMS layer adhered on the silicon dioxide wafer during the cast molding process.

Bonding Strength Evaluation
The bonding strength for all substrate was investigated using destructive mechanical tensile test method on square 10 × 10 mm 2 bonded pairs. In addition to the PDMS-SiO 2 /GaAs bonding configuration, the bonding strength of two other interesting systems, PDMS-Glass and PDMS-LiNbO 3 , was evaluated. PDMS, glass, and LiNbO 3 substrates were cut using diamond saw dicing, whereas GaAs/SiO 2 were cut with a cleavage method. All substrates were cleaned with acetone and isopropanol to remove dust and organic contaminants. The bonding strength between the different substrates and PDMS was determined using a mechanical tester for micro-components (Nordson DAGE 4000Plus), equipped with a 250 kg cartridge (50 kg range used). The specific horizontal setups of the tensile test and the assembled sample are shown in Figure 8A. Two steel sample holders were used to sandwich the bonded sample in between. The upper part of PDMS replica and the backside of all substrates were glued on steel sample holders using a single component instant adhesive Loctite 480 (Henkel Adhesives). The sandwiched sample and its holders were fixed horizontally in the special stud-pull fixture of the tester that eliminates side forces due to gimbal construction. During the tensile test, the motorized stage moved horizontally at a uniform speed of 10 µm/s, while load force and displacement were recorded simultaneously until failure occurred. The failure mode was determined by optical inspection of fracture surfaces (interface). The tensile strength of the tested sample was calculated by dividing the measured maximal load force at the bond failure. The bonding strength was calculated as follows: where F is the load force, and S is the bonding area. horizontally at a uniform speed of 10 μm/s, while load force and displacement were recorded simultaneously until failure occurred. The failure mode was determined by optical inspection of fracture surfaces (interface). The tensile strength of the tested sample was calculated by dividing the measured maximal load force at the bond failure. The bonding strength was calculated as follows: (1) where F is the load force, and S is the bonding area. All tensile tests were conducted on the samples that experienced plasma treatment and thermal annealing at 70 °C for one hour. The load force curves for all tested configurations are presented in Figure 8B   All tensile tests were conducted on the samples that experienced plasma treatment and thermal annealing at 70 • C for one hour. The load force curves for all tested configurations are presented in Figure 8B-left. For all samples, the strength response exhibits conventional behavior: increases over time until breakage and then sharply decreasing. Breaking was observed at 12 s, 18 s, and 19 s for GaAs/SiO 2 , LiNbO 3 , and glass, respectively. The load force for PDMS-SiO 2 /GaAs is consistently higher when compared to other configurations with glass and LiNbO 3 substrates, i.e., 200.75 N, 128.02 N, and 77.79 N, respectively. The use of an intermediate SiO 2 bonding layer increased the bonding strength on GaAs, resulting in a significant increase in bonding strength: increase of 55% compared with glass substrate and of 156% compared with LiNbO 3 . The achieved bonding strength of GaAs/SiO 2 to PDMS is~2.01 MPa. Figure 8B-right shows the average load force for different samples bonded by the presented process. A good reproducibility (bonding tests were carried out on five similar samples) for our measurements was observed, with a maximum error of 6% caused by the dispersive imprecision on the dimensions of the bonding area. Figure 8C shows photos of broken structures after the tensile tests. In the case of PDMS-SiO 2 /GaAs chips, failure was generally observed at the SiO 2 -PDMS bonding interface. However, in the case of Glass and LiNbO 3 samples, failure was observed in the PDMS volume which is in agreement with the literature reports [27,36,46]. The results can be explained by the higher bonding strength between PDMS-Glass and PDMS-LiNbO 3 than the coherence force of cured PDMS materials. In fact, when PDMS's surface is brought into contact at elevated temperatures where the molecular chain mobility is high, adhesion occurs at the interface. The chain ends penetrate to the opposite substrate, leading to high bonding strength. Moreover, the increase in surface roughness enhances the adhesion property of substrates. The surface roughness of LiNbO 3 was bigger than glass no matter what the plasma activation time was according the findings of Xu et al. [47]. To proceed further in our interpretation, we would like to hypothesize that the large mismatch in the coefficient of thermal expansion (CTE) should be analyzed in [47,48] between LiNbO 3 and glass (14.4 (x, y-axis)-7.5 (z-axis) × 10 −6 K −1 for LiNbO 3 and 0.56 × 10 −6 K −1 for glass). According to this hypothesis, there is bigger stress at PDMS-LiNbO 3 interface due to higher mismatch in (CTE) when compared to PDMS-Glass. Therefore, PDMS-Glass resists higher load.
Since the SiO 2 layer was also locally detached from the GaAs surface, this failure mode is partially affected by limited adhesion of SiO 2 layer on GaAs that indicates that the bonding strength is higher than the coherence of PDMS material. The high stiffness of GaAs promotes interface failure, we would like to hypothesize that the high value of the Young's modulus of GaAs (118 GPa) implies that the fracture takes place at the interface. The observed failure of the PDMS-SiO 2 /GaAs bond was consistently at the interface between SiO 2 , indicating a very strong bond between SiO 2 and PDMS. SiO 2 has been reported to be a reliable adhesive layer to bond PMMA substrates on PDMS for microfuidic applications [42].

Leakage Tests
In order to validate a leakage-free performance of the bonded PDMS-SiO 2 /GaAs chips, the channel inlet/outlet holes were equipped with epoxy-sealed connectors and the leakage test was performed under defined flow conditions ( Figure 9A). The flow rate increased from 10 µL/min to 4500 µL/min, looking for the maximum working pressure with no leakage appeared. A red dye solution was used for easier optical microscope inspection of potential leakage at the border of the channel.
In the leakage test, the design chips based on PDMS-SiO 2 /GaAs were kept intact (i.e., without dissemination of red dye solution into the bonded interface) until 8.5 bar of working pressure, which is maximum available in our experimental setup. It is clearly seen in Figure 9A that the channel is well defined, and no leakage is observed at flow rate of 4500 µL/min and 8500 mbar (the limit of control pressure sensor). The maximum working pressure and the maximum flow rate are shown in Figure 9B. In the leakage test, the design chips based on PDMS-SiO2/GaAs were kept intact (i.e without dissemination of red dye solution into the bonded interface) until 8.5 bar of work ing pressure, which is maximum available in our experimental setup. It is clearly seen i Figure 9A that the channel is well defined, and no leakage is observed at flow rate of 450 μL/min and 8500 mbar (the limit of control pressure sensor). The maximum working pres sure and the maximum flow rate are shown in Figure 9B.

Discussion
Due to the challenges related to the use of PDMS as the structural material, mos studies have focused on finding alternative materials. Although other polymers, such a PS, PMMA, TPE, COP, and photoresists, have been used for the fabrication of microfluidi devices [46,[49][50][51]. Many different problems related to the bonding of the microfluidi channel on piezoelectric substrate such as GaAs (100) were identified. Bonding processe are required for assembly of microfluidic devices, made of two or more components. Th can be achieved by using of double-sided tape, glues, or solvent bonding [52]. The bond ing of PDMS on different substrates has been reported in the scientific literature [36], bu the solvents used in the bonding process can also strongly influence the growth of cel cultured in the microfluidic devices [53].  trimethox ysilane (MPTMS), which was a chemical coupling reagent to modify the surfaces of th noble metals, and PDMS to improve their adhesion [54]. Yong et al. proposed a nove approach to fabricate multi-layer glass microfluidics chips, which comprises laser cuttin and thermocompression bonding [29]. Among the methods presented above, special trea ments processes or the processes of adding additional chemical reagents are required t achieve PDMS adhesion. This work presents an approach to assess how to improv PDMS/GaAs bonding based on the combination of low temperature (<100 °C plasma/thermal treatments and the use of appropriate intermediate bonding layer. W proposed a new solution for bonding of PDMS microfluidic cell on the gallium arsenid substrate, covered with silicon dioxide thin intermediate layer. The aim of the character zation step was to optimize the bonding quality of multilayer PDMS-SiO2/GaAs. Th study consisted of experimental investigations. We have characterized the bonding inter

Discussion
Due to the challenges related to the use of PDMS as the structural material, most studies have focused on finding alternative materials. Although other polymers, such as PS, PMMA, TPE, COP, and photoresists, have been used for the fabrication of microfluidic devices [46,[49][50][51]. Many different problems related to the bonding of the microfluidic channel on piezoelectric substrate such as GaAs (100) were identified. Bonding processes are required for assembly of microfluidic devices, made of two or more components. This can be achieved by using of double-sided tape, glues, or solvent bonding [52]. The bonding of PDMS on different substrates has been reported in the scientific literature [36], but the solvents used in the bonding process can also strongly influence the growth of cells cultured in the microfluidic devices [53]. Wu et al. used (3-Mercaptopropyl) trimethoxysilane (MPTMS), which was a chemical coupling reagent to modify the surfaces of the noble metals, and PDMS to improve their adhesion [54]. Yong et al. proposed a novel approach to fabricate multi-layer glass microfluidics chips, which comprises laser cutting and thermocompression bonding [29]. Among the methods presented above, special treatments processes or the processes of adding additional chemical reagents are required to achieve PDMS adhesion. This work presents an approach to assess how to improve PDMS/GaAs bonding based on the combination of low temperature (<100 • C) plasma/thermal treatments and the use of appropriate intermediate bonding layer. We proposed a new solution for bonding of PDMS microfluidic cell on the gallium arsenide substrate, covered with silicon dioxide thin intermediate layer. The aim of the characterization step was to optimize the bonding quality of multilayer PDMS-SiO 2 /GaAs. The study consisted of experimental investigations. We have characterized the bonding interface by various measurement techniques (SEM, AFM, ATR-FTIR, and CA).
In order to prove the usefulness of the proposed solution, the elements and the chemical bonds on PDMS's surfaces have been determined by ATR-FTIR analysis. The roughness and topography of various treated and non-treated PDMS and GaAs/SiO 2 surfaces were also analyzed using AFM. The PDMS-SiO 2 /GaAs samples fabricated according the proposed method were able to withstand the load force until 200.7 N without failure, which corresponds to bonding strength of 2.01 MPa, which was the highest value we obtained. This is substantially higher than the bonding strength of other tested microfluidic systems, such as PDMS-Glass or PDMS-LiNbO 3 . An analysis system was created to measure the  [27]. From our experiment, the bonding strength obtained is higher than other PDMS bonding methods.
In the plasma treatment process, there is only a small amount of hydroxyl groups (the inherent hydroxyl groups) on the surfaces of PDMS and silicon dioxide layer. Xiangdong et al. mentioned that the contact of such two surfaces will produce a small amount of Si-O-Si bonds, which can only form a week adhesion force [55]. Thermal annealing is another method used to improve adhesion performance. Indeed, annealing contributes to the stabilization of the bonding layer, improves the cross-linking density in the PDMS and favors the orientation of molecular chains. Although a bonding temperature around 100 • C is low for silicon or glass substrates, it affects the bonding performance of polymer substrates greatly. Winnie et al. [56] used hot embossing techniques for bonding the PDMS-PMMA system at 90 • C for 3 h. An additional annealing process at 65 • C for 1 h on PDMS and PS surfaces improved bonding and allowed the stabilization of higher SEF for a longer period of time (>3 days). In this work, to enhance bonding, we have combined plasma treatment and annealing. In our experiment, after plasma treatment, both parts are aligned and pressed together while undergoing the curing process at low temperature of 70 • C for 1 h. Pre-stress uniformly applied during bonding significantly influences the orientation of the polymer chains. Hammami et al. [57] showed that the combination of temperature and stretch promotes the orientation of molecular chains in the dielectric elastomer. Subsequently the combination of temperature and small deformation promotes bonding of our system.
The chip holder ensures that the leakage could only occurs at the bounded interface between PDMS and GaAs/SiO 2 substrates, i.e., not at the tubing connector. No leak was observed in the tested PDMS-SiO 2 /GaAs samples until maximal available working pressure of 8.5 bar. To our knowledge, it is the highest leakage-free pressure reported in the literature for PDMS-based bonding systems. A comparison between previously reported results of leakage test is shown in Table 2.  [55] Overall, this study presents a method for evaluating whether PDMS can be used as a reliable material for microfluidic devices in order to enhance the performance of acousticfluidics-biosensors based on GaAs. Our contribution is discussed only in terms of basic technological challenge to couple PDMS with GaAs. Nevertheless, the presented results are likely to contribute to the improvement of the performance of the microfluidics systems combining PDMS and GaAs.

Conclusions
We have presented a novel bonding solution between GaAs and PDMS, which enables the development of an increasingly in-demand array in biosensors field, including those requiring high flow rates and high pressures. The combination of SiO 2 intermediate layer with plasma oxygen and low-temperature annealing (70 • C) significantly improves the bonding of PDMS to the GaAs substrate. In our acoustic biosensor application, one can assume a maximum pressure driving around 8.5 bar for the fluid. The bonding area of microfluidic devices can withstand a stress about 2.01 MPa. Additionally, this bonding method does not require wet chemical treatments on bonded surfaces, which may be prohibited in some applications. Bonding features were evaluated using different methods, bonding strength, and leakage tests. Compared to the previous studies, our included bonding method distinguished a robust and rapid fabrication technology as well as superior bonding strength and leakage-free pressure. The obtained results can be valuable for the research and development of integrated microfluidic devices based on PDMS material in general.