Monolithic-integrated piezoresistive MEMS accelerometer pressure sensor with glass-silicon-glass sandwich structure

Figure 1 shows the monolithic composite sensor structure before anodic bonding process which consists of an accelerometer and a pressure sensor. The accelerometer has double-cantilever-mass structure and the pressure sensor has rectangular sensing diaphragm structure. After one-mask wet anisotropic etching, the accelerometer’s rectangular diaphragm with a mass and the pressure sensor’s rectangular diaphragm are shaped simultaneously. Then, deep reactive ion etching (DRIE) is used to release the accelerometer’s diaphragm and form the suspended double-cantilever-mass structure. The thickness of accelerometer’s diaphragm is designed to be 15 μm, the same as that of the pressure sensor’s diaphragm, and the mass of accelerometer is designed to be greatest after wet etching to increase the sensitivity of accelerometer. Boron-doped silicon areas on the substrate work as piezoresistors for both accelerometer and pressure sensor. A layer of thermal-growing SiO₂ and a layer of LPCVD low-stress SiN_x are deposited and patterned on the substrate as passivation layers. Around bonding areas, passivation layers are etched as trenches which expose silicon. A layer of 1 μm-thick LPCVD α-Si is deposited and patterned subsequently to connect the silicon substrate by the exposed silicon in trenches, which can help to protect piezoresistors from p–n junction break-down during the α-Si-glass anodic bonding process on the top surface. To ensure the bonding strength and airtightness, instead of Al wires, heavily boron-doped silicon areas are buried beneath the SiO2/SiNx passivation layers to electrically connect Al wires with Al pads. The glass-silicon-glass sandwich structure (Roylance and Angell 1979; Zhang et al. 2013) is formed with α-Si-glass anodic bonding on the top surface and Si-glass anodic bonding at the bottom, as is shown in Fig. 2.

Dimensional parameters of the pressure-sensing diaphragm and piezoresistors are shown in Fig. 3. The pressure-sensing diaphragm (Santosh Kumar and Pant 2014, 2016; Bae et al. 2004; Kanda and Yasukawa 1997) is designed to be 950 μm × 450 μm × 15 μm, which features high mechanical linearity under 450-kPa ranged pressures. Four boron-doped silicon piezoresistors (R1, R2, R3, R4) are parallelly arranged on edges and the center of the diaphragm along <110> direction to form a Wheatstone full-bridge. The piezoresistors are boron-doped to be about 1 × 1017 cm^-3 to reduce the temperature effects on sensitivity. The dimensional parameters of the piezoresistors are designed to be 260 μm (length) × 20 μm (width) × 3 μm (boron-doping depth).

Dimensional parameters of accelerometer’s double- cantilever-mass structure are shown in Fig. 4. Two folded boron-doped silicon piezoresistors (R′ 1, R′ 2) are parallelly arranged on the centers of two cantilevers along <110> direction, respectively, and another two (R′ 3, R′ 4) are parallelly arranged on substrate along another direction. Four piezoresistors form a Wheatstone half-bridge with reference resistors R′ 3 and R′ 4. The dimensional parameters of the piezoresistors are designed to be 100 μm (length) × 10 μm (width) × 3 μm (boron-doping depth). The thickness of accelerometer’s diaphragm and cantilevers is designed to be 15 μm, the same as that of the pressure sensor’s diaphragm. Two cantilevers are designed to be 80 μm × 60 μm with the distance of 60 μm. The dimensional parameters of the diaphragm are 785 μm × 260 μm, and the distance between the center of etched mass and the edge of the substrate is 770 μm.

In order to increase sensitivity, the mass of the accelerometer should be etched to be greatest. The mass is formed by one-mask anisotropic wet etching process and the convex corner compensation etching technique (Fan and Zhang 2006; Offereins 1990; Schroder et al. 2001) is used to design the greatest mass. Because the thickness of silicon substrate is 380 μm and the thickness of accelerometer’s diaphragm is 15 μm, the depth of wet etching is calculated to be 365 μm. The mass etching mask design is illustrated in Fig. 5. The dimensional parameters of mask frame are 1550 μm (length) × 960 μm (width) to ensure the diaphragm area of 865 μm (length) × 260 μm (width) after 365 μm depth wet etching process. The center of the big 440 μm side-length square is 1080 μm away from the edge, and eight small 200 μm side-length squares are around four corners of the big square for wet etching compansation. All parameters designed for squares can guarantee the greatest mass after 365 μm depth wet etching process.

Shown in Fig. 6a, the model of the pressure sensor, which includes a rectangular pressure diaphragm, piezoresistors and wires, is built and meshed by using 3D Builder and then exported to TEM. The material properties of silicon, aluminum and boron-doped silicon (piezoresistor) are listed in Table 1. Figure 6a results show the stress distribution on the pressure diaphragm under 450 kPa pressure applied on the top surface. After the stress distribution extracted to electromechanical coupling module, Fig. 6b results show the electric potential distribution of the piezoresistors and wires with a 5 V power supply on the Wheatstone bridge. The output of Wheatstone bridge is 2.5434–2.4622 V = 81.2 mV and the sensitivity of pressure sensor is calculated to be 81.2 mV/(5 V·450 kPa) = 36.1 μV/V·kPa.

In this design, the mass of the accelerometer is formed using silicon anisotropic wet etching process from the backside of the wafer. To maximize the mass size and get a deeper understanding of etching process, silicon anisotropic wet etching simulation is conducted by IntelliEtch, as is shown in Fig. 7. IntelliEtch module is based on an atomic representation of the anisotropic wet etching dynamic process of silicon substrate, allowing the use of a wide range of Kinetic Monte Carli (KMC) and Cellular Automata (CA) time-evolution (Zhu and Liu 2000; Yan et al. 2008). SiO₂ and SiN_x layers patterned as Fig. 5 are applied on the silicon wafer bottom surface (100) as the etching mask. The anisotropic etch rate of aqueous KOH (45 %, 85 °C) has been calibrated using wagon wheel method (Gosálvez et al. 2011) and is input to the database in software. Figure 8 shows that shapes of accelerometer’s mass are formed at different etching depths. Simulation verifies that the etching mask as Fig. 5 can form the greatest mass which is separated to diaphragm edges to be an island.

The accelerometer with double-cantilever-mass structure is modeled according to IntelliEtch simulation results and subsequently electromechanical coupling simulations are conducted using TEM, as is shown in Fig. 8. Under a 5 V power supply and a 125 g acceleration input, the output of Wheatstone bridge is 2.5042–2.4959 V = 8.3 mV. The sensitivity of accelerometer is calculated to be 8.3 mV/(5 V × 125 g) = 13.3 μV/V g.

Vibration modal analysis of accelerometer is also conducted using TEM, which can predict whether the chip is able to work well in a certain noise circumstance. The first vibration mode of the accelerometer is 8.54 kHz, as is shown in Fig. 9.

Shown in Fig. 12, the performances of the pressure sensor in the monolithic composite sensor are measured in a temperature-controlled environmental chamber where the temperature can be adjusted from room temperature to 300 °C and pressure can be adjusted from 0 to 1000 kPa.

Test results are demonstrated in Fig. 13. Figure 13 shows the output voltage versus the pressure of the pressure sensor at 25 °C under a 5 V DC supply. Within the range of 0–450 kPa, by using fitting-line method the sensitivity of the pressure sensor is calculated to be 33.0 μV/V kPa.

Shown in Fig. 14, a precise centrifuge turntable and a centrifuge control system are used for the measurement of the accelerometer. To ensure the dynamic balance, two accelerometers are fixed at two edges of the centrifuge turntable. The rotation-generated centrifugal accelerations are adjustable by the centrifuge control system which is controlled by a computer. A DC 5 V power is supplied to the Wheatstone bridge of the accelerometer. The output signal of the accelerometer is directly input a digital multimeter for voltage readout without any amplification.

The output voltage versus the acceleration under a 5 V supply is shown in Fig. 15. Within the measured range of 0–135 g, by using fitting-line method the sensitivity of accelerometer is calculated to be about 11.2 μV/V g. The zero point output of the accelerometer is measured to be 3.81 mV under a 5 V supply.