Proceedings of the 2007 International Conference on Information Acquisition July 9-11, 2007, Jeju City, Korea

Design and Fabrication MEMS-Based Micro Solid State cantilever Wind Speed Sensor Lidong Du , Zhan Zhao and Cheng Pang State Key Laboratory of Transducer Technology Institute of Electronics, China Academy of Sciences No.2 North Second Street, Zhongguancun, Haidian, Beijing, 100080, China [email protected] Abstract - The MEMS-based micro solid state cantilever wind speed sensor has advantages of small size and high sensitivity of low wind speed. The designed sensor was simulated with ANSYS software and was analyzed with fluid mechanics principle. The fabrication of the sensor mainly consists of wet anisotropic etching and DRIE (deep reactive ion etching) on a 3″ silicon wafer. The designed sensor comprises two silicon cantilevers which thicknesses are 12µm, widths 500µm, and lengths 1530µm. Each sensor detects sensitive signal with Wheatstone bridge which comprises two platinum resistances on the cantilevers and two platinum resistances on non cantilever area. The fabricated sensor was tested within small wind tunnel. The result suggests that the designed sensor can detect low wind speed as low as 0.08m/s and has high sensitivity of low wind speed. Index Terms - MEMS technology; ANSYS analysis; Wind Speed; silicon cantilever; Wheatstone bridge

I. INTRODUCTION Wind speed measurement is a very important factor in various fields such as aeronautics, meteorology, sealing and farming. The conventional wind sensor comprises small mechanical apparatus such as propeller and cup anemometers, or consists of thermal element such as hot wire anemometers, or has acoustic part such as acoustic radar. Propeller and cup anemometers are most commonly employed in the meteorology stations and airports. They are equipped with a dc generator. Such sensors are robust but their dc generator must be regularly maintained. Hot wire anemometers use fine heating elements that double as temperature sensors. The flow rate is inferred from the extent of forced convective heat transfer from the hot wire [1]. The acoustic anemometers based on Doppler frequency shifts consist of an acoustic launcher and receiver [2]. All of these wind sensors are big size while MEMS-based wind sensor has the advantages of small size, light weight, and low cost. Therefore, many schools have devoted into this field. Recent years some kinds of wind sensor have been designed based on mechanical principle or thermal principle. This paper concentrates on fabrication a double cantilever wind speed sensor which is sensitive to low wind speed. The sensor that is based on the mechanical anemometer principle was designed and fabricated using MEMS technology. In the designed sensor platinum was used as resistive material and was patterned by a lift-off process. Two cantilevers were formed on the front and back by front and back wet anisotropic etching and DRIE (deep reactive ion etching) process.

1-4244-1220-X/07/$25.00 ©2007 IEEE

336

Through those MEMS processes we simplify the fabrication processes of silicon cantilever sensor. The electric resistance of metal or nonmetal depended on the temperature of environments and the unexpected effects of material caused by the temperature variation of environments need to be compensated. So the electric resistances on cantilevers of the sensor were designed equal to the electric resistances on non cantilever area. All of them were formed into Wheatstone bridge [3]. When constant voltage was selected as the input of Wheatstone bridge, the static outputs of the sensor is zero. Thus, it lowers the difficulty of temperature compensation. At the same time the resistances change of the double cantilevers sensor is as twice as the single cantilever sensor. This paper will analyze the principle in section II. Then it will present the result of ANSYS CFD (Computational Fluid Dynamics) software analysis in section III. In the IV section, the developing processes based on MEMS technology will be given. In section V, the test result will be discussed. II.

OPERATION PRINCIPLE OF SENSOR

The front pattern and cross-sectional view of MEMS-based cantilever wind sensor are shown in Fig. 1. A comprehensive understanding of the relation between the output voltages and the wind speed must take into consideration many wind conditions. Sometimes the finite element method must be used to analyze the conditions of wind flow model. But the wind flow model is beyond the scope of this paper, this paper will concentrate on the discussion of the relation between output voltage and the wind speed. The mean wind speed was denoted as u0, the diameter of the wind tunnel was denoted as d. So the formulation for

Fig. 1 The front pattern and cross-sectional view of MEMS-based cantilever wind sensor. Each cantilever covers with platinum resistance.The four platinum resistances were formed into Wheatstone bridge.

Reynolds Number [4] is as shown in (1). Re = u0d/ν

(1)

The parameter ν is the air viscous friction coefficient. The wind is classified into two kinds by the Reynolds Number. If Reynolds Number Re is less than 2300 the wind flow is laminar flow; if Reynolds Number Re is more than 2300 the wind flow is turbulent flow [4]. In this paper the diameter d of the experiment instrument is equal to 4.52×10-2m. The Reynolds Number of experiment instrument calculated using (1) is more than 2300 when the mean speed of the wind flow is more than 0.799m/s. So the wind flow in experiment instrument is mainly turbulent flow. If the designed sensor was placed into the center of the turbulent flow the experiential function [4] of wind velocity distribution can be denoted by (2). u(y) = 0.81u0(y/r)n

(2)

The terms y, r, and n are the distance of cantilever bottom, the semidiameter of the experiment instrument, and the index related to the Reynolds Number Re. The following table is the relation of n and the Reynolds Number Re [4]. TABLE I.

INDEX OF VELOCITY DISTRUBUTION OF TURBULENT FLOW TABLE

Re

4.0×103

2.3×104

1.1×105

1.1×106

2.0×106

n

1/6

1/6.6

1/7

1/8.8

1/10

(3)

In order to calculating the drag force the cantilever was divided into slices, each with height dy. So the drag force [5] acting on each slice is denoted by (4). dF(y) = 0.5CDρu2(y) wdy

(4)

The term CD is the local drag coefficient, w is the width of the cantilever width in the direction facing the wind flow, and ρ is the density of air. Among those parameters, the parameter CD can be calculated by empirical formulae and is depend on the structure of material. The moment applied to cantilever is calculated by performing finite integration through the length of the cantilever, yielding h M = ∫0 ydF (y ) .

(5)

The term h is the height of the cantilever. The magnitude of induced strain of the cantilever beam is calculated as ε = Mt/ (2EI).

dR/R = dl/l – dS/S + dρ/ρ≈(1+2µ) ε .

ΔR/ R = (1+2µ)Mt / (2EI).

ΔR/R = 0.8547(1+2µ) CD u0 ρr2 (h/r)7.6/3.3/(Et2). 2

(11)

Because the two cantilevers were designed as equal to each other the relation between the output voltage and the wind speed can be obtained by Uo = 0.42735 Ui(1+2µ) CD u0 ρr2 (h/r)7.6/3.3/(Et2) . 2

(12)

Equation (11) is suggested that the output of the double cantilever sensor is as twice as the output of the single cantilever sensor and (12) is suggested that the magnitude of 2 the output signal is proportional to u0 . Now the Wheatstone bridge with constant voltage input was discussed. The diagram of Wheatstone bridge is shown in Fig. 2. When the environment temperature is T0 the static output of the Wheatstone bridge based on ohm law can be obtained by UT0 = UR2 – UR3 = U0(R2/(R1+R2) – R3/(R3+R4)).

(13)

When the environment temperature becomes T the real magnitude of the four resistances are as shown in (14), (15), (16), and (17). R1′ = R1(1+α(T–T0))

(14)

R2′ =

(15)

R2(1+α(T–T0))

(7)

337

(10)

When the constant DC voltage Ui was taken as the input of Wheatstone bridge the output of the Wheatstone bridge Uo can be obtained by

(6)

The term t is the thickness of the cantilever and w is the width of the cantilever.

(9)

By assuming that the velocity distribution function is given by (3), the formula for single cantilever resistance change can be obtained by

The term E is the Young’s modulus and the term I is the momentum of the inertia of the cantilever. The momentum I can be approximated by I = t3w/12.

(8)

The term µ, R, S, and ρ are the Passion ratio of platinum metal, metal resistance, cross-section area of metal, and resistance coefficient. Thus, from (6) and (8) the equation of the relative change of resistance in wind flow can be obtained by

Uo = Ui (ΔR1/ R1 +ΔR2/ R2 )/4.

The maximum Reynolds Number of our experiment instrument calculated using (1) is about 55000. So the index of wind velocity distribution function n was selected as 1/6.6. The function of wind velocity distribution is as shown in (3). u(y) = 0.81u0(y/r)1/6.6

Because the platinum resistance is adhered on the cantilever the strain of platinum resistance can be taken as equal to the cantilever induced strain. So the relative change of the resistance of the cantilever can be obtained by calculating the resistance change of the platinum resistance. The relative change of platinum resistance can be approximated easily by

Fig. 2 Diagram of Wheatstone bridge

R3′ = R3(1+α(T–T0))

(16)

R4′ =

(17)

R4(1+α(T–T0))

So the static output of the Wheatstone bridge is as shown in (18) UT = U′R2 – U′R3 =U0(R2(1+α(T–T0))/(R1(1+α(T–T0))+R2(1+α(T–T0))) – R3(1+α(T–T0))/(R3(1+α(T–T0))+R4(1+α(T–T0)))) = U0(R2/(R1+R2) – R3/(R3+R4)).

(18)

Fig. 3 Diagram of sensor model (Three demension)

When (13) and (18) were compared the conclusion can be drawn that when constant voltage was taken as the input of Wheatstone bridge the static output of Wheatstone bridge caused by the environment temperature variation is zero. III.

ANSYS CFD ANLYSIS OF SENSOR

The schematic diagram of cantilever sensor model is shown in Fig. 3. In the model the cantilever thickness is 10µm, width 500µm, and length 1500µm.

When the data of Fig. 5 and Fig. 6 were fitted with data analysis software it will be found that the maximum deflection and maximum stress of cantilever are proporational to the square of the mean flow speed. This result is accorded with the result of analysis based on fluid mechanics principle. IV.

FBRICATION OF SENSOR

The cantilever wind speed sensor was designed with two cantilevers. Its dimension is 6mm×5mm. The thickness of the cantilever is 12µm, width 500µm, and length 1530µm. The silicon wafer used in fabrication is double-side polished, n-type, and single-crystal silicon wafer (thickness = 350µm, diameter = 3 inch). The fabrication process of the cantilever is discussed in the following. The fabrication process was began with a thermal oxidation process forming a silicon dioxide layer which thickness is about 5000 Å after cleaning the silicon wafer. Then a silicon nitride layer was deposited about 2400 Å by the LPCVD (Low Pressure Chemical Vapor Deposition) process. The target of the silicon nitride layer is protected the silicon dioxide layer and silicon layer in the front and back wet anisotropic etching process. The nitride silicon grown by LPCVD process has tensile stress about 1000MPa; the dioxide silicon grown by thermal oxidation process has pressure stress about 300MPa [5]. When the sensor was fabricated the stress

338

1200 Maximum deflection (nm)

Through the simulation the data of maximum stress and maximum deflection of cantilever under the given wind flow were obtained. The curve of maximum deflection is displayed in Fig. 5. The curve of maximum stress is displayed in Fig. 6.

Fig. 4 Diagram of deflection of cantilever

1000 800 600 400 200 0 0

5

10 15 20 wind speed(m/s)

25

30

Fig. 5 Diagram of relation between maximum deflection and flow speed 14000 12000 Maximum stress (KPa)

The behavior of the cantilever sensor is simulated with ANSYS CFD software. For simplification of the sensor model, the 3D computing model includes only one cantilever. In the model the FLUID142 was used as the model of flow and Solid45 as the model of cantilever in ANSYS CFD software. The maximum stress and maximum deflection of the cantilever in the flow are simulated using ANSYS CFD software at the room temperature at the flow speed from 0.4m/s to 30m/s. The deflection result is shown in Fig. 4.

10000 8000 6000 4000 2000 0 0

5

10 15 20 wind speed(m/s)

25

30

Fig. 6 Diagram of relation between maximum stress and flow speed

balance should be taken into account. In our study the ratio about 1:2 was selected. After thermal oxidation process and LPCVD process, platinum thin membrane was sputtered on the silicon wafer. Chromium was used as the interlayer and lift-off process was used to remove the unwanted platinum. Then the negative photoresist was used to pattern the figure of cantilever. Then the SF6 plasma etching was used to remove the non masked nitride silicon. Because platinum is difficult to etch by HF acid the dioxide silicon layer was directly etched in HF acid about 20 seconds till that the non masked dioxide silicon was disappeared. After the front dioxide silicon and nitride silicon were removed the KOH etching process was used about 30 minutes at 70℃. The depth of front KOH etching is about 15µm. The front MEMS technology processes were completed after front KOH etching step. The micrograph of the designed sensor before the back MEMS technology processes is illustrated in Fig. 7. Before the back MEMS technology processes the front figure should be protected with black glue. So we protected the front figure with black glue. After this step the same processes as removing the front dioxide silicon and nitride silicon were repeated. The KOH etching process was used again about 5 hours at 70℃ when those steps were completed and the depth of back KOH etching is about 200µm. The last step was DRIE (Deep Reactive Ion Etching) process. This process was used to etch the silicon wafer from the back till obtaining the silicon cantilever. The whole fabrication processes of the cantilever are exposed in Fig. 8. Finally the wafer was diced into pieces and the wires were bonded to pads which connect the outer electric circuit. A packaged wind speed sensor is illustrated in Fig. 9. The resistance of the cantilever is about 1.1KΩ. Through those processes we simplify the fabrication processes of silicon cantilever wind speed sensor. V.

(a)

(f)

(b)

(g)

(c)

(h)

(d)

(i)

(e) Fig. 8 Diagram of technology process (a) Step of thermal oxidation; (b) Step of Low Pressure Chemical Vapor Deposition (LPCVD); (c) Step of lift-off Platinum; (d) Step of front etching SiO2; (e) Step of front SF6 plasma etching; (f) Step of front KOH etching; (g) Step of back etching SiO2; (h) Step of back SF6 plasma etching; (i) Step of Deep Reactive Ion Etching (DRIE);

Fig. 9 Photograph of packaged wind speed sensor

The blue lines are the sensor output curve at 21.4℃ and the green dot curve is the best-fit curve of the measurement value. The best-fit curve of the output follows (19) Y = a × Xb.

TEST RESULTS OF SENSOR

The fabricated sensor was installed in a small wind tunnel. Wind speed was set from 0m/s to 8m/s. The input of the Wheatstone bridge was set to DC voltage 5V which came from the DH1718 Dual Tracing Powers Supplies; the output of the Wheatstone bridge was measured by KEITHLEY 2001 MULTIMETER. The experiments were conducted at 21.4℃. The output curve and the best-fit curve of the sensor are shown in Fig.10.

(19)

Equation (19) is different with the earlier principle analysis. The reason is that the four resistances of the Wheatstone bridge are not completely equal to each other. When the input of the Wheatstone bridge was set to DC voltage 5V the static output of the Wheatstone bridge is not equal to zero. The resistances of the cantilevers simultaneously increased with wind speed increasing. Thus, the output of the sensor decreased. When the static output value was used minus the measurement values another curve of the sensor was gotten in Fig.11. Fig.11 suggested that the magnitude of the output signal is not proportional to the square of the wind speed. The first reason is that the thickness of the cantilevers is thicker. The second is that the wind flow brings about the resistance variation of the platinum resistance. When increased the wind speed the cantilevers gradually lose their elastic property and the platinum resistance lose more quantity of heat by convection. But the principles earlier analysis are still work. The designed sensor can be used to measure the low wind speed. The precision of the sensor is 5% and the sensitivity of

Fig. 7 The micrograph of designed sensor before the back MEMS technology process.

339

26.0

4.0 3.5

Best-fit curve of sensor Measurement curve of sensor

25.0

Y Caiculation Value (mv)

Y Difference Voltage (mv)

25.5

24.5 24.0 23.5 23.0 22.5 22.0

3.0 2.5 2.0 1.5 1.0 0.5 0.0

21.5 -1

0

1

2

3

4

5

6

7

-0.5

8

-1

X Wind speed (m/s)

0

1

2

3

4

5

6

7

8

X Wind speed (m/s)

Fig. 10 The measure value curve and best-fit curve of sensor at 21.4℃

Fig.11 The measure value curve and best-fit curve of sensor at 21.4℃

the sensor is 0.49mv/ (m/s). The lowest wind speed which the sensor can detect is 0.08m/s. CONCLUSION A cantilever wind speed sensor was designed and fabricated. The operating principle for detecting the wind speed was illustrated; the cantilever sensor was simulated in ANSYS CFD software; the silicon cantilever sensor was fabricated using MEMS technology. Experiments were conducted to observe the characteristic of the designed sensor. The figures shown in earlier section suggested that the sensor has high sensitivity to low wind speed. Through theoretic calculation we can get that the double cantilevers structure has augmented the output of designed sensor, and through those MEMS processes we simplify the fabrication processes of the silicon cantilever wind speed sensor. Further studies are necessary to extend the measurement range of the sensor, necessary to find out the reason, which

340

causes the difference between the result and the principle, and necessary to measure the wind direction and the wind speed at the same time. REFERENCES [1]

[2]

[3]

[4] [5]

M. Laghrouche, A. Adane, J. Boussey, S. Ameur, D. Meunier, S. Tardu “A miniature silidon hot wire sensor for automatic wind speed measurements,” Renewable Energy, vol. 30, pp. 1881–1896, February 2005. Zhifang Fan, Jack Chen, Jun Zou, Davis Bullen, ChangLiu, and Fres Delcomyn, “Design and fabrication of artificial lateral line flow sensors,” J. micromech. Microeng, vol. 12, pp. 655–661, June 2002. Wang Qi, Zhao zhan, Zeng Huanhuan, Fang Zhen, Zhang Bojun, “Pressure sensor based on Si3N4 membrane and the compensation,” Electronic Device, vol 29(3), pp.726–729 September 2006. Xia Taichun, “ENGINEERING FLUID MECHANICS,” Shang Hai Jiao Tong university press 2006. Stephen D. Senturia, “Microsystem design,” Publising House of Electronics Industry 2004.