No recent searches

  • to select
  • to navigate
  • to close

Design, fabrication and testing of MEMS pressure sensor with silica-Al heterogeneous structure

Since Smith proposed the piezoresistive effect of bulk silicon in 1954, silicon-based MEMS piezoresistive pressure sensors have been widely used in consumer electronics, the automotive industry, biomedical, meteorological and aerospace fields. However, due to the low sensitivity of bulk silicon material (silicon piezoresistive strain coefficient is about 100), serious temperature drift and other problems, the performance of silicon-based piezoresistive pressure sensor is affected. Many related studies have been carried out on the above issues [6@5]. For example, in order to improve the sensitivity of piezoresistive sensors, Yang Peidong et al. proposed the use of surface-modified silicon nanowires to achieve the giant piezoresistive effect. The piezoresistive (strain) coefficient of the nanowires they made can be two orders of magnitude higher than that of conventional bulk silicon. Obviously, the giant piezoresistive effect can greatly improve the sensitivity of pressure sensors. However, the generation mechanism of piezoresistive effect is very complex, which is closely related to the surface state of nanowire, so the stability is poor. The process requirements of the structure of the giant piezoresistive sensor are very high, which leads to the huge piezoresistive effect being widely used in pressure sensors. Recently, R〇we et al. studied the realization of micron-scale devices with huge piezoresistive effect. By using conventional silicon piezoresistive strips and aluminum strips to form a gold conductor hybrid piezoresistive device, geometric amplification of piezoresistive effect can be obtained, and its strain coefficient can reach 843, which can significantly improve the sensitivity of the sensor. Compared with the silicon nanowire sensor, the piezoresistive sensor with the metal silicon mixed structure (Si/Al isotherm) of the base micron scale is easy to be prepared and realized by constant etching. It is worth noting that the resistance, which is isomerized by silicon aluminum, is easily affected by ambient temperature and therefore requires temperature drift. For the temperature drift of the sensor, there are mainly hardware and software two methods, the hardware is often used in parallel temperature: resistance way to achieve, but because of the calculation, resistance itself temperature drift and other problems, the guiding effect is too ideal. Commonly used methods such as linear regression analysis, two-dimensional interpolation, vector holding machine and artificial neural network learning, etc., these methods effectively improve the effect of temperature drift and nonlinear error, and are more widely used at present.

In order to improve the sensitivity of MEMS piezoresistive pressure sensor and the influence of effective temperature drift and other factors, this redesign and research model is based on the high sensitivity of Si-Al isomerism pressure sensor. Firstly, the sensor structure and performance design are analyzed by using parts and theoretical calculation. The Si-Al isomerism pressure sensor with hardware temperature structure is manufactured by traditional MEMS. At the same time of hardware realization, the improved wavelet neural network based on genetic algorithm is used to carry on the sensor parts, so that the performance is greater. The research is a good reference price for the development of highly sensitive piezoresistive sensors.

1 Principle of piezoresistive effect of silica-Al heterostructures

The structure diagram of the traditional piezoresistive pressure sensor is shown below. It consists of a Wheatstone bridge formed by 4 piezoresistive strips distributed on the edge of the strain film. Under the action of stress, the resistance or resistivity of the piezoresistive bar changes due to piezoresistive effect, which can be expressed as:

AA-1

Where, R is the initial value, P is the initial resistivity, π is the piezoresistive coefficient, E is Young's modulus, g is the strain, and K is the strain sensitivity coefficient.

AA-2

Different from the working principle of traditional piezoresistive pressure sensors, Si-Al isomerism is realized by using the anisotropy of Si-doped silicon caused by stress to make the current leave the high conductive metal for pressure measurement [9]. The Si-Al heterogeneous structure pressure sensor structures 1 (b) and 1 (c) are shown. The Si Al isomerism pressure sensor chip of the meter comprises a pair of Si Al isomerism pressure sensing structure and a pair of Si Al isomerism temperature parameter, both of which have the same material and structural form. Each single Si Al isomerism comprises a Si piezoresistive strip and a metal Al strip. The silicon piezoresistive strips are connected with four lead-out pads, the outer two for constant current source power supply, and the inner side for output voltage measurement. The equivalent resistance of silica-al heterostructure can be expressed by the formula.

AA-3

In the formula, L is half of the distance between the two inner voltage measurements; L is half of the length of the piezoresistive bar; h is the thickness of the piezoresistive bar; b is the width of the piezoresistive bar; p1 and p2 are the longitudinal and transverse resistivity respectively. When external pressure is applied, the stress image.png generated in the silicon heterogeneous structure will change the longitudinal and transverse resistivity, thus affecting the equivalent resistance of the heterogeneous structure, which is the piezoresistive effect of the silica-aluminum heterogeneous structure. The expression of p1 and p2 bodies is as follows:

AA-4

Where, π1 and π2 are the longitudinal and transverse piezoresistive coefficients respectively, and the amplification factor SG related to sensor sensitivity can be expressed as:png. It is not difficult to find that when the ratio of L/b is large, image.png. Therefore, it can be seen that the piezoresistive bar with an appropriate aspect ratio can obtain a larger amplification factor and sensitivity. Considering the actual processing technology level, the preliminary size design of Si/Al heterostructure is as follows: b = 55um, l = 60um, L = 150um. The initial resistivity P0 is approximately 0.043 for 1.5um ion injection depth and 1X108 Dose /cm3 boron doping. Due to the removal of the top layer of silicon of 0.5um by the thermal oxidation process in the production process, the actual thickness of the piezoresist bar h= 1um. At this time, the initial theoretical resistance value of Si-Al heterostructure can be calculated as 31.8 according to the formula. It is worth noting that at normal temperature or normal working temperature, the resistivity of metal materials and semiconductor materials increases with the increase of temperature, and the resistance calculated according to the conventional resistance formula increases with the increase of temperature. However, the working principle and resistance formula of the sensor with Si-Al heterogeneous structure have changed. According to the calculation formula (2) ~ (4) of Si-Al heterogeneous structure resistance, considering the influence of temperature on resistivity and piezoresistive coefficient, it can be calculated that under different pressure conditions, the resistance of Si-Al heterogeneous structure decreases with the increase of temperature. This is consistent with the experimental results later in this paper.

As mentioned earlier, the sensing structure located inside the pressure sensitive film of the silicon aluminum heterostructure chip is used to measure the external pressure changes, while the reference structure located outside the pressure sensitive film is basically unaffected by the external pressure. They are mainly used to eliminate the temperature drift error. Ideally, since the four Si-Al heterostructures have the same shape and size and are located in the same temperature environment, the initial output without pressure should be U0+△Ut under the same constant current supply and temperature condition, where △Ut is the output change caused by temperature. When external pressure P is applied, the output voltage of the sensor structure is Uref=Uo+△Ut. Since the reference structure is not affected by stress, its output voltage is Uout and Uref differential treatment can obtain the final output voltage of the Si Al allostructure pressure sensor:

AA-5

It can be seen from the formula that the final output result of the pressure sensor is the voltage output change caused by the external pressure change, which has nothing to do with the temperature change. Theoretically, the temperature drift error can be suppressed by the hardware temperature compensation method.

2. Finite element stress simulation

In order to obtain good linearity and sensitivity of Si/Al heterogeneous sensor, it is necessary to select reasonable diaphragm parameters. Too thin film will lead to relatively large nonlinear error, and too thick film will lead to decreased sensitivity. Considering the level of processing technology, the thickness of the diaphragm is set as 20um in this paper. Under the condition of 0 ~ 1 000 kPa full scale, the length and thickness of the sensor diaphragm shall meet the formula.

AA-6

Where, is the Poisson's ratio of silicon, G is the Young's modulus of silicon, and B is the full-scale pressure. According to the formula, it can be calculated that the strain film side length a≤1 182 um. In this paper, 900 um was selected as the edge length of the membrane. As the thickness of the silicon substrate of the SOI wafer used for making the sensor is 650um and the wet corrosion Angle is 57.74°, the size of the C-type silicon cup window is 1 792 um.

In order to verify the rationality of the theoretical analysis, we use ANSYS finite element analysis software to conduct modeling and stress analysis for the pressure sensor of silica-aluminum heterogeneous structure designed in this paper, so as to finally determine the size parameters of the sensor. The physical properties of materials used in finite element simulation are shown in Table 1.

AA-7

Table 1 Material parameters of finite element simulation

FIG. 2(a) and FIG. 2(b) respectively show the diaphragm displacement nephogram and Von Mises stress distribution nephogram of the heterogeneous sensor under external pressure of 100kPa. As can be seen from Figure 2 (a), under the action of stress, the maximum displacement of the sensor occurs in the center of the strain film, which ensures the symmetrical distribution of stress. In FIG. 2 (b), the maximum stress is distributed in the middle of the four edges of the strain film. It is not difficult to find that the two silica-aluminum allostructured sensing structures are located at the maximum stress of the strain film, while the reference structures are located outside the strain film, basically unaffected by the stress.

Then, the applied pressure is gradually changed within a large range of 0~1000 kPa. According to the results of ANSYS finite element simulation, the average internal stress of the sensing structure and the reference structure can be extracted. The relationship between the internal stress and the external applied pressure is shown in Figure 3 (a). As can be seen from FIG. 3, with the increasing external pressure, the partial stress of the Si-Al heterogeneous sensing structure shows great linearity, while the internal stress of the parametric structure basically does not change, which is consistent with the expected results. Under the condition of 1mA, combining equations (2) ~ (4), we can calculate the theoretical value of the equivalent resistance of the Si-Al isomerism sensor changing with the external pressure, and 3 (b). It can be calculated from Figure 3 (b) that the sensitivity of the sensing structure at normal temperature is 0.098 5 m V/(V • kPa). When the two sensing structures are measured simultaneously, the sensitivity will be doubled.

AA-8

AA-9

Figure 2. Finite element simulation results of the Si-Al heterogeneous structure pressure sensor

AA-10

(a) The relationship between internal stresses and external pressures in the sensing structure and reference structure

AA-11

(b) The relationship between the equivalent resistance of the sensing structure and the reference structure and the external pressure


FIG. 3 Finite element simulation results of the Si-Al heterogeneous pressure sensor


3. Sensor manufacturing process

In this paper, the standard MEMS process is used to produce aluminum heterogeneous structure pressure sensor, special process and materials. The selected SOI wafer has a monocrystalline silicon device layer thickness of 1.5 um, a silicon dioxide layer thickness of 1 um and a substrate silicon thickness of 650 um. The specific production process is shown as follows:

AA-12

FIG. 4 Schematic diagram of the fabrication process of the pressure sensor with Si-Al heterogeneous structure

Step 1 Ion implantation 1x1018 Dose/cm3 boron ions were injected into the SOI device layer at 7° Angle and 20 keV energy. The SOI silicon wafers were then placed in a 1000 annealing furnace for 30 min to rapidly anneal the boron ions.

Step 2 Thermal oxidation The SOI silicon wafers are thermally oxidized to form a blunt silica layer with a thickness of about 1 um.

Step 3 Photoetch adhesive, using silicon piezoresist bar structure and the mask plate at the outlet end to advance lithography development. However, SFb, * mixed ICP method was used to etch silica passivated layer and top layer silicon, and the structure of silicon piezostopper strip and its extraction end with silica-Al heterostructure. The structure of the silicon piezoresistive strip and the induced silica passivation layer are photoetched to form a contact hole between the silicon piezoresistive strip and the metal aluminum strip.

Step 4 Aluminum sputtering adopts O R IO N-8-U H V thin film deposition equipment layer thickness of aluminum, and then uses mask plate photolithography to etch metal aluminum strips, metal aluminum leads and aluminum.

Step 5 Etching silicon cup First, the silicon dioxide passivity layer on silicon substrate is photoetched, then the underlying silicon is etched by TMAH method at a temperature of 80 _ to form C-type silicon cup. The top of the silicon cup is the strain film of the sensor, the size is 900 um x 900 um x 20 um, and the silicon oxide is connected.

Step 6 Bonding at an environment of 300 _, the SOI wafer is bonded to the silicon base with the bonding technique and sliced in parallel.

FIG. (a) shows the microscope photo of the silicon aluminum heterostructure pressure sensor chip, and FIG. (b) the actual sensor mounted. The entire chip has an area of 3 000 um x 3 000 um. As shown in the figure below, the inner side of the strain film of the two silica-al isomerism (sensing structure) sensors and the outer side of the left and right (reference structure) are in line with the design requirements.

AA-13

AA-14

Figure 5. The physical picture of the silicon Al heterogeneous pressure sensor chip

4. Experiment and result analysis

The schematic diagram of the calibration test platform is shown in Figure 6. The German G E-D ruck pressure controller PACE5000 is used as the standard pressure generator in the experiment, and the measurement accuracy is better than ± 0.03%FS. Firstly, the Si-Al isomerization pressure sensor is put into the temperature box, and the guide pipe is connected to the sensor and pressure controller. The measuring lead of the sensor is then drawn and connected with a multimeter for measurement. The temperature of the regulating chamber changes from -2 0℃ to 60℃, and the pressure changes from 0 kPa to 1000 kPa at the same temperature point, and 100 kPa is used as a pressure test point. Under the condition of 1 mA, the voltage output of the pin is measured by the Si/Al heterogeneous sensing structure and parameters at the same temperature and different pressure.

FIG. 7 shows the output curves of the Sial-Al heterostructure pressure sensor at different temperatures. Corresponding to the theoretical analysis, considering the constant current source power supply, the change of output voltage translates into the change of resistance. According to 7 (a), the external pressure applied to the Si/Al heterogeneous sensing structure increases at different temperatures, and the resistance is 32.32Ω at room temperature, which is basically consistent with the theoretical calculated value. When the temperature increases from -20℃ to 60℃, the resistance of the sensor structure becomes smaller, and the sensitivity of the Si-Al isomeric pressure sensor changes from 0.107 2 m V /(V. kP a) to 0.123 4 m V /(v. kP a). The sensor temperature drift. As described above, the Si-Al heteroparametric structure outside the strain film on the same chip is discussed as shown in Measurement 7 (b) of the pressure-sensitive Si-Al heteroparametric structure with temperature-varying resistance. As shown in Figure 7, the temperature increases while the temperature decreases with the increase of the base pressure, and the resistance is 30.35Ω at room temperature. The difference between the equivalent resistance of the sensor structure and the reference structure is mainly caused by the deviation in the fabrication process, such as the blending degree, the defects of the silicon and aluminum material itself and the deviation of the phottography, which will affect the effect of the hardware temperature compensation to some extent.

AA-15

AA-16

FIG. 7 Output curves of Si/Al heterogeneous pressure sensors at different temperatures

The difference between the Si/Al heterogeneous sensor structure and the parameter structure can improve the temperature drift of the sensor results in hardware temperature, and the parameter structure is effective. The sensitivity of the resistance difference △R at the same temperature is about 0.116&m V /(V·kPa) according to the average curve calculation. Because the silicon aluminum isomerism pressure sensor chip contains two identical sensing structures, the sensitivity is twice as high. This is greater than the sensitivity of the pressure sensor based on the piezoresistive effect of traditional bulk silicon (0.061G V /(V • k P K)

AA-17

Figure 8 Results after hardware temperature compensation

5. Data fusion processing based on genetic algorithm wavelet neural network

The basic principle of the improved wavelet neural network based on genetic algorithm is as follows: the genetic algorithm is used to form a continuously evolving sequence, and the global basic solution is obtained according to the special way. However, the calculated basic solution is used as the neural network state to carry on the neural network, and the random network is saved to make the basic solution of the problem easier and faster. The neural network process of genetic algorithm 9, the first neural network connection factor and transfer factor of random, basic genetic calculation of the first generation, repeated operation of the submode race. However, the output of each network corresponding to the group network is obtained by the forward propagation algorithm. The fitness function used is as follows:

AA-18

Where :M is the number of learning samples, and e is the error between the expected output value Yk and the actual output value yk of the wavelet network. It can be seen from equation (9) that the larger the error, the smaller the fitness. Calculate the fitness value of each individual, eliminate the less fitness value, and then carry out cross mutation and other operations, repeat training until the relationship is satisfied or the number of iterations is reached. See the literature for a detailed introduction of the genetic algorithm wavelet neural network.

Figure D Flowchart of the improved wavelet neural network based on genetic algorithm

The hardware temperature data is taken as the sample data of G, A, W, NN algorithm, and the sample data is normalized at first. Then, an improved mathematical model of wavelet neural network based on genetic algorithm is established by using M ATLA B to fuse the data. The input layer node of the meridians has 2 equivalent resistance values and temperature values, the hidden layer has 11 points, and the output layer is the pressure of 1 node. The momentum factor is 0.01, the learning speed is 0.001, and the maximum number of times is 2 000. The parameters of genetic algorithm are as follows: intersection 0.75, variation 0.08, and initial population size of 200. In the full scale range of 0 ~ 1000kPa, the predicted output and prediction errors of the Si-Al isomer pressure sensor after GA-WNN algorithm processing are shown in 10(a) and 10(b).

AA-19

AA-20

GA-WNN algorithm compensation results

As shown in Figure 10 (a), the output is consistent with the predicted output, the influence of temperature drift is basically eliminated, and the performance of the sensor is 0(F). It is found that the maximum predicted absolute error is more than 15kPa, and the measurement error is ±1.5% FS. It can be seen that the measurement error of Si-Al isomerism pressure sensor is significant after data fusion of GA-W N N N algorithm, which is mainly attributed to the fact that GA-W algorithm significantly reduces the nonlinear error of Si-Al isomerism sensor. According to the data in 10(a), the overall nonlinear error at the same temperature is 1.36%, which is significantly less than the overall nonlinear error before compensation of 3.56%.

6 Conclusion

This paper focuses on the design and research of a new type of large range pressure sensor based on the silicon aluminum heterogeneous sensing structure and parameters. Firstly, ANSYS carries out stress and sensitivity simulation of the sensor to verify the high sensitivity of the silicon aluminum heterogeneous sensor. The second design of the sensor manufacturing engineering, and according to the process flow, process and design model of the silicon - aluminum heterogeneous sensor chip. However, the temperature chamber and high precision pressure generator are used to test and calibrate the Si - Al isomerization pressure sensor. The hardware temperature method and the improved neural network algorithm based on genetic algorithm are mainly used to carry out the sensor. The GA-WNN algorithm is studied for significant sensor temperature drift and nonlinear error and sensor performance. The research of this paper is of reference value to the development of high sensitivity sensor.

References:

1, Smith C s. Piezoresistance Effect in Germanium and Silicon * J +. Physical Review, 1954, 94 (1) : 4 2-4 9.

2. Barlian A, Parle W T, Mallon J R, et al. Review: Semiconductor Piezoresistance for Microsystems * J + Proc IEEE 2009, 9 "3-552.

3. Khoshnoud F.,  Silva C W de. Recent Advances in MEMS Sensor Technology-Biomedical Applications * J ] . IEEE Instrumentation  and  Measurement Magazine, 2 01, 1 5 (1) : 8-1 4.

4. Zhang J H, Wu Y S, Liu b Q, et al. Research on High-Precision,  Low Cost Piezoresistive ME MS-Array Pressure Transmitters Based  on Genetic Wavelet Neural Networks for Meteorological  Measure  ments. Micromachines, 2015, 6:5 5 4-573.

5. Nie Meng, Yang Hengshan. • Chinese Journal of Sensors and Actuators, 2 0 1 7,3 0 (1 2):1 8 3 4-1 8 3 8 8.

6 and He R, Yang p. Giant Piezoresistance Effect in Silicon Nanowires * J] Nat Nanotechnol, 2 00 6, 1 (1) : 4 2-4 6.

7. Milne J S, Rowe A C, Arscott S, Et al. The Giant Piezoresistance Effects in Silicon Nanowires and M icrowires * J]. Phys Rev Lett, 2 0 0, 1, 105 (2, 2) :  226802.

8. Winkler K, Bertagnolli E, Lugstein a. Origin of Anomalous Pie  zoresistive Effects in VLS Grown Si Nanowires * J] the Nano t L t, 2 0 1 5,1 5:1 7 8 0 -1 7 8 5.

9. Rowe A C, Donoso-Barrera A, Renner C h, Et al. The Giant Room - Tem  perature Piezo to hold a Metal - Silicon in Hybrid Structure * J]. Phys Rev Lett, 2 00 8,100:145501.

10, Hansen O, Reck K, Thomsen E V Giant Geometrically Amplified Piezoresistance in Metal - ^ Semiconductor H b U y Resistors * J] Jour  nal Journal of Applied Physics, 2 00 8,1 0 4 (11) : 114510.

11, Ngo H D, Tekin T, Vu T C, Et a l MEMS Sensor with Giant Pie  zoresistive Effect Using Me tall - Semiconductor Hybrid Structure * C] / / IE 16 th E E International Solid-State Sensors, Actuators and Microsystems Conference(Transducers 2011), 2011:1018-1021.

12, Long Jun, Guan Wei, Wang Xudong, et al. • High Precision Measurement Model of Pressure Sensor Based on Ridge Regression * J]. Chinese Journal of Sensors and Actuators, 2 0 1 7,3 0 (3):3 9 1-3 9 6.

13, Long, Chen Jun • Pressure sensor temperature self-compensation strategy based on L S-S V R *J] • Chinese Journal of Sensors and Actuators, 2 0 1 7,3 0 (7):1 0 5 7-1 0 6 1.

14, Xu, Ke, Zuo • Research on Temperature of Optical Fiber Pressure Sensor based on P, S, O-B, P neural Network * J]. Instrument Technology and Sensors, 2 0 1 6 (1 0): 1-5, 9.

15, Chen, Feng, et al. • Chinese Journal of Sensors and Actuators,2 0 1 6,2 9 (7) : 1109-1114.

16, Sun Yicai, Liu Yuling, Meng Qinghao. Design, Manufacture and Application of Pressure Sensor *M] • Metallurgical Industry Press, 2000.

17, Yang. Research on a Silicon micropiezoresistive pressure sensor * J]. Machinery Design & Manufacture, 2 0 1 2, (1) : 103-1 0 5.

18, Dong Yang. Fabrication and Characteristic Research of Nanometer Polysilicon thin film Pressure Sensor Based on M EM S Technology [D]. Heilongjiang: Heilongjiang University, 2014.

19, Zhang Jiahong, Yang Min, Gu, et al. Development of high precision silicon piezoresistive pressure transmitter based on measurement and data fusion * J] • Chinese Journal of Sensors and Actuators, 2 01 6, 29 (6): 8 26-8 3 3.

  
Copyright @banning 2024. All rights reserved.