Energy Dissipation Contribution Modeling of Vibratory Behavior for Silicon Micromachined Gyroscope

Energy dissipation contribution of micro-machined Coriolis vibratory gyroscope (MCVG) is modeled, numerically simulated, and experimentally verified in this paper. First, the amount of independent damping dissipation consisting of thermoelastic loss, anchor loss, surface loss, Akhiezer loss, and air damping loss during vibration is obtained by simulation model, PML-based method, and numerical calculation, respectively. Then, temperature and pressure dependence characteristic of the corresponding quality factor (Q) for the MCVG are obtained. Meanwhile, dominant sources of damping dissipation are determined, which paves the way to improve Q. Finally, the temperature-dependent and pressure-dependent characteristics of the total Q are measured with errors of less than 10% and 18% compared with the simulated total Q, respectively, in which accuracy is acceptable for predicting the damping dissipation behavior of MCVG in design stage before high-cost fabrication.


Introduction
Vibration behavior is an important feature of MEMS twoorder mass-damping-stiffness system.Energy conversion is a fundamental characteristic of these vibrational system.For some type MEMS devices, vibration regarded as energy is collected as power of electrical components, such as vibration energy harvest [1][2][3].For another type devices, this vibration behavior can be localized to improve signal noise ratio (SNR) greatly [4].Nevertheless, more common but inevitable phenomenon for all of MEMS devices during vibration occurrence is vibration energy dissipation behavior, typically such as micromachined Coriolis vibratory gyroscope (MCVG) [5][6][7].The performance of this type of vibration system is significantly limited by the rate at which the vibration energy is dissipated, which is called time constant. is generally used as a measurement of this time constant of the vibration energy and is defined as ratio of the total strain energy to the dissipated energy per vibration cycle.Higher  means larger time constant and lower vibration energy dissipation, for the MCVG, which means achievement of excellent bias stability [8][9][10][11][12].
In order to achieve high performance of the MCVG, analysis of energy dissipation mechanisms of the Q value of the MCVG during vibration becomes an important issue [13][14][15][16].For the single-mode mass-damping-stiffness resonator, the  value has been widely regarded as a combination of five energy dissipation mechanisms and inverse proportion to the total energy dissipation [17][18][19].Energy dissipation usually consists of two categories: intrinsic dissipation such as thermoelastic damping loss 1/Q the , anchor damping loss 1/Q anc , and Akhiezer damping loss 1/Q akh and extrinsic dissipation such as surface damping loss 1/Q sur and air damping loss 1/Q air .
These damping loss mechanisms of the single-mode resonators have been extensively investigated for achieving the expected Q.In summary, influencing factors on damping losses can contribute to three aspects: vacuum packaging technology, materials properties, and mechanical structure topology.First, vacuum packaging scheme as the most common one is used to greatly decrease air damping, such as epi-seal encapsulation process [20][21][22][23][24][25], hermetic lid sealing process [13,[26][27][28], Through-Silicon Via (TSV) process [29,30], and SPIL MEMS WLP process [31].Second, some different kinds of materials are tried to yield lower the thermoelastic, surface deflect, and Akhiezer damping.For example, crystalline and doped silicon are chosen frequently to make high-Q resonators because of their excellent semiconductor properties [32][33][34][35][36][37].Researchers also take advantage of single crystal (SC) [38], microcrystalline (MC) [39][40][41][42], nanocrystalline (NC) [43,44], and ultra NC [45] diamonds to achieve high-Q resonators by trading off contributions of different damping loss items, respectively.As for the structure topologies which mainly affect anchor damping, anchor loss models of simple cantilever beam have been established [46][47][48].However, for some complex fully 3D resonators, contributions of another geometrical resonators including DETF [21], tether geometry [49][50][51], microsphere [52], cupped [52,53], wineglass [54], hemispherical [39], and disk shapes [55][56][57] to anchor loss are calculated hard by the analytical solution if no simplifying assumption is present.Damping loss model on accurately estimating the experiment results for these complex structure topology is very limited.Main reason is that it is difficult to obtain precise a priori knowledge of the stress profile at the anchor area of these design.Especially for the MCVG with both the drive and sense modes, the elastic beams connecting with anchors exist mechanical vibration coupling between the drive and sense modes.
Therefore, we will try to model and simulate the energy dissipation mechanisms for MCVG by numerical analysis technologies in this paper.Furthermore, the contribution of each damping dissipation to the total quality factor of the MCVG is predicted.The experiments are also implemented to verify the theoretical analyzes.
This paper is organized as follows: in Section 2, modeling and simulation of the independent damping dissipation for the MCVG are presented.In Section 3, temperature and pressure dependence characteristic measurements of the quality factor are implemented.Finally, the conclusions are provided in Section 4.

Numerical Simulation Analysis of Damping Dissipation
Here, we consider our single-mass z-axis MCVG as the example to [8,[58][59][60] to analyze the contribution of the damping loss, as shown in Figure 1.The symmetrical SOI-based MCVG structure consists of a set of the electrodes, anchors, elastic beams, and three masses including the drive-mode mass, the sense-mode mass, and the proof mass.The structure is fixed on the substrate through four anchors, which connect with eight elastic beams.The vibration between both modes is decoupled through eight other elastic beams located at the proof mass.In order to enhance the performance, Q value of the MCVG should be as high as possible.

Contribution Of Damping Dissipation. According to
Zener's theory [17,18], for the MCVG, the contribution of all the damping losses items to the total quality factor Q t can be further expressed as where Q the , Q anc , Q sur , Q akh , and Q air represent the quality factor related to thermoelastic damping loss, anchor damping loss, surface deflect damping loss, Akhiezer damping loss, and air damping loss.
2.1.1.Thermoelastic Damping Loss.Thermoelastic damping is part of mechanical vibration energy dissipated in the form of heat energy.This type of damping loss exists in the place of the mechanical structure deformation caused by elastic strain.It depends closely on geometric size and material property of the structure.For the simple mechanical structure such as cantilever beam, analysis solution of thermoelastic damping has been obtained fairly accurately.However, for the MCVG, the accurate analysis solution is deduced hardly because of the existing coupled elastic vibration between the drive and sense mode.So, finite element model of the thermoelastic damping is established.Then, the thermoelasticity modulus in Comsol software is used to extract Q the by solving the coupled mechanical and thermal differential equations.The detailed calculation process is demonstrated as follows.
The first step is the finite element model of the MCVG established in Comsol and then the elastic vibration is simulated to obtain the elastic stain tensor  xy at any point inside the structure and stored maximum vibration energy W.
The second step is to obtain transient temperature ∇Tof the MCVG by heat transfer simulation.∇ = / + / + / is Hamilton operator.The above elastic stain is first substituted into the expression  0 ζ  ( 11 +2 12 ) [19] regarded as internal heat source, where  0 is initial temperature of 20 ∘ C,  11 and  12 are the second-order elastic stiffness tensors, and  is tensor of the thermal expansion.Then, transient temperature ∇T of the MCVG internally is simulated by the thermoelastic modulus.
The last step is to derive Q the during the mechanical structure vibration.Considering heat transfer in the mechanical structure made from a thermally isotropic material of single crystal silicon, increment rate Ṡ  of the total entropy yielded from the thermal conduction of the internal heat source can be written as where  =  ⋅ ∇ is the heat flux and  is thermal conductivity coefficient.dV is small volume element during vibration.According to the thermodynamics theory, heat amount increment dQ produced during vibration can be written as Further, heat amount increment ΔQ during a period t s of vibration can be written as According to the above vibration energy , quality factor Q the caused by thermoelastic dissipation can be obtained as follows: According to Figure 1, geometrical parameters of the single crystal silicon (SCS) MCVG and the material parameters are first listed in Table 1.
Through setting the parameters listed in Table 1, energy loss distribution of the thermoelastic damping for the MCVG is obtained at the initial temperature of 20 ∘ C, as shown in Figure 2. It can been seen that the thermoelastic damping loss mainly exists at connection region between the mass and the elastic beams in both of the modes; meanwhile, the rest of the thermoelastic loss occurs at the anchors region connecting the elastic beams.This means that higher strain gradient exists in these regions, where temperature gradient is caused and further generates the thermal flow.Further, Q ted within the temperature of -40 ∘ C to 60 ∘ C is simulated.Due to the SCS properties variation with temperature such as E, K, C, and , the material properties as a function of temperature are set in the thermoelasticity modulus of Comsol.Figure 3 shows Q the of thermoelastic damping versus temperature.
It can be seen from Figure 3 that Q the caused by thermoelastic damping is five orders of magnitude, reaching about 0.3 million.Further, Q the decreases gradually with temperature rising.When temperature is 20 ∘ C, Q the is 241000.

Anchor Damping Loss.
Anchor damping is part of elastic-beam vibration energy radiated and dissipated into the substrate in the form of elastic wave.Perfectly matched layers (PML-based) method is induced to calculate Q anc of anchor damping in this work.Berenger first introduced the PML concept to analyze an issue on electromagnetic propagation in 1994 [61].Recently, Bindel applied the PML to estimate the anchor loss of 2D MEMS device [62].However, analytical solution of anchor loss Q anc for the fully 3D structure is still hard to achieve until now.So, we try to estimate the anchor damping Q anc by establishing the PML-based numerical model of the MCVG.
Figure 4 shows the finite element model (FEM) of the MCVG using a PML to model the substrate as a finite space by  Figure 5 shows the PML-based elastic wave simulation results of the anchor damping.It can be seen that the elastic waves propagate into the substrate through the connection between the elastic beams and anchors.The occurrence of the elastic waves is mainly caused by the elastic beams deformation during vibration, which gives rise to stress at the connection area.Finally, this vibration energy as the anchor loss is dissipated through the elastic wave propagation in the PML substrate.According to the above analysis, anchor dissipation can be reduced by the following ways.One is to locate the beams connecting with anchor at the center of mass to decrease deformation degree of the beams.The other is to reduce the total number of the beams connecting with anchors to decrease energy dissipation.Figure 6 plots anchor damping Q anc versus the sweeping layer number n at the initial temperature of 20 ∘ C. Q anc is an exponential function of the sweeping layer number by fitting the scatter spots.When the number n increases from 5 to 15, Q anc decreases significantly.This means that estimation accuracy of Q anc is a bit low.Nevertheless, with n further increasing, Q anc approximately keeps constant, reaching 593700.Its estimation accuracy is significantly improved.Nevertheless, computational cost of PC work station rises sharply.Further, the estimated anchor damping Q anc is a little larger than the thermoelastic damping Q the .

Surface Damping Loss.
The surface damping is related to the surface layer conditions of the structure such as the surface contamination and deformation induced during the fabrication process [47,63].For the MCVG, first of all, adsorbates on the surface layer can be fully cleaned up by careful operation in the clean room, as shown in Figure 7. Second, when the MCVG vibrates along the drive axis, the surface deformation of the proof mass is negligible because of its large rigid body characteristic, although lots of tiny release roles distribute in the proof mass.So, surface layer deformation only occurs on the elastic beams.It is further regarded as surface damping loss caused by the layer deformation which is given rise solely by the beams deformation.Nevertheless, because the ratio reaching 0.001 between the surface of the beams and the volume of the proof mass is that small, as shown in Figure 7, surface damping loss caused by the surface deformation of the beams can be also neglected for the MCVG.[21] happens when the period of the MCVG vibration is slower than inelastic phonon scattering process, which mainly restores the local thermal equilibrium disturbed by strain during the vibration of the single crystal.This loss is related to the strain caused by acoustic wave propagation process within the silicon material.Quality factor Q akh induced by the Akhiezer damping can be estimated as follows:

Akhiezer Damping Loss. Akhiezer damping loss
where , f, K, and T are density of single crystal silicon, resonance frequency of the MCVG, thermal conductivity, and temperature, respectively.c,  avg , and K are acoustic velocity within the silicon material and average Grüniesen's coefficient, which are equal to 1.04 × 10 5 m/s and 0.51, respectively.Q akh is inversely proportional to temperature.When the temperature is set as 20 ∘ C, Q akh is calculated to be 1.09 × 10 11 .Compared with the thermoelastic and anchor damping, Akhiezer damping is so small that its effect to quality factor can be neglected.

Air Damping Loss.
Air damping is part of extrinsic energy dissipation occurring on the moving surface of the device.The pressure determined by the vacuum level of the device cavity has significant influence on estimation of this damping.For the MCVG, the surface movement mainly occurs between the fixing comb fingers and moving comb fingers, bottom surface of the proof mass, and the substrate, respectively.Quality factor Q air of the MCVG can be obtained by calculating as follows.
First, mean free path of air molecules in the MCVG cavity can be written as where P is air pressure in the cavity and Boltzmann constant k B and effective diameter of air molecule d are equal to 1.38 × 10 -23 J/K and 3 × 10 -10 m, respectively.Due to slipping effect of air existing on the moving surface, air viscosity coefficient  should be modified as its effective value  eff [64] where  is equal to 1.8 × 10 -5 Pa⋅s at the temperature of 20 ∘ C, Knudsen number K n is equal to /D.and D is gap between the adjacent comb fingers and the bottom surface layer of the proof mass and the substrate layer, which are 2m and 5m, respectively.Then, damping coefficient c dmp is obtained as follows: where A are area between the adjacent comb fingers and the bottom surface layer of the proof mass and the substrate layer, which are 30m × 10m and 4000m × 4000m, respectively.Through the calculation, sum of the damping coefficients c dmp cmb between the comb fingers and c dmp mass between the bottom surface and the substrate is 4.51 × 10 -7 N/m/s.According to the equation   =   × 2 × (   +    ), where m d is sum of the drive-mode mass and proof mass, Q air induced by air damping is calculated to reach 56345 at the temperature of 20 ∘ C and pressure of 10Pa.Further, quality factor Q air as a function of pressure P at the temperature of 20 ∘ C is plotted in log scales in Figure 8.
It can be seen from Figure 8 that quality factor Q air is approximately linear with pressure P in log scales.Quality factor increases with pressure decreasing.Further, temperature dependence of the quality factor Q air is investigated.According to [65], viscosity coefficient  versus temperature T can be written as where viscosity coefficient  0 is 1.71 × 10 -5 Pa⋅s at the temperature of 0 ∘ C. Putting ( 8) and ( 10) to ( 9), through calculation, Q air as a function of temperature at the pressure of 10Pa is plotted in Figure 9.In Figure 9, with temperature increasing, Q air decrease gradually, which keeps in the order of ∼10 4 within the temperature of -40 ∘ C to 60 ∘ C.

Shock and Vibration
According to (1), contribution of the damping dissipations to the quality factor at the pressure 10Pa and temperature of 20 ∘ C can be listed in Table 2.
It can be seen that the total quality factor at the given condition of temperature and pressure is 42405.The damping dissipation is mainly determined by the thermoelastic damping, anchor damping, and air damping, which is a dominant source of dissipation, reaching 75.26%.

2.2.
Pressure-Dependent Characteristic of Quality Factor.Due to pressure-independent characteristic of the thermoelastic, anchor, and Akhiezer damping, these three quality factors are approximately constant over the pressure.Considering (1) and Figure 8, total quality factor Q tot versus pressure P can be calculated, as shown in Figure 10.
It can be seen that the total quality factor Q tot tends to be constant, reaching the order of 10 5 , when the pressure level is lower than 5Pa.This means that the thermoelastic damping and anchor damping are dominant sources of dissipation within this range of pressure level.When the pressure level is beyond 5Pa, the relatively low Q tot approaches Q air , which implies that air damping become dominant dissipation source.Further, the simulated Q tot is about 42405 at the pressure level of 10Pa.

Temperature-Dependent Characteristic of Quality Factor.
Due to weak temperature-independent characteristic of the anchor damping [21], Q anc is approximately regarded as be constant over the temperature.Considering (1) and Figures 3 and 8, total quality factor Q tot versus temperature T at the pressure of 10Pa can be calculated, as shown in Figure 11.
It can be seen that the total quality factor Q tot decreases slowly from 52978 to 37313 with temperature increasing, which approximately keeps in the order of ∼10 5 at the pressure level of 10Pa.Q tot varies 15665 within the temperature range of -40 ∘ C to 60 ∘ C. The simulated Q tot is 42405 at the temperature of 20 ∘ C. Further, Q tot is close to Q air .This means that the air damping is dominant source of dissipation at this pressure level.

Measurement and Comparison
The presented MCVG is packaged on the vacuum operation platform, as shown in Figure 12.The packaging pressure is 10Pa.First, Q tot as a function of temperature T is tested in the temperature controlling chamber, as shown in Figure 13.Q tot of three gyroscopes are calibrated by using the 3dB bandwidth method, respectively.
The measured Q tot is plotted by the scatter spots, as shown in Figure 14.It can be seen that the measured total Q tot of the three gyroscopes decreases with temperature increasing at the pressure level of 10Pa.Their change trend is almost consistent.The error bars represent accuracy of the measured values of Q tot .The average value of the Q tot of three gyroscopes reaches 49040 at the 40 ∘ C, 39116 at the 60 ∘ C, and 41087 at the 20 ∘ C. Compared with the simulated results shown in Figure 11, errors at these corresponding temperature spots are 7.4%, 4.8%, and 3.1%, respectively.In the whole temperature range, the error is all less than 10%, which means that the numerical simulation of the quality factor versus temperature for the MCVG is acceptable for predicting the quality factor characteristic.After the temperature dependence experiment of quality factor for the vacuum-packaged MCVG, the cover of the metal package is opened and the exposed MCVG is then put into the vacuum exhaust chamber to implement the pressuredependent experiment of the quality factor, as shown in Figure 15.
Through controlling the pressure in the exhaust chamber, pressure-dependent characteristic of Q tot is obtained when the temperature is 20 ∘ C. Test data is plotted by the scatter spots, as shown in Figure 16.
In the pressure range from 0.017Pa to 3000Pa, the measured Q tot increases slowly when the pressure is reduced from 5Pa to 0.017Pa, reaching the maximum value of 95872, which means that Q the and Q anc are dominant sources of damping dissipation.Nevertheless, Q tot drops inversely to 511 when the pressure is beyond 10Pa.Q tot is 34766 at the pressure level of 10Pa with an error of 18% compared with the simulated Q tot .Overall, the measured Q tot is relatively a little small, whereas both of their trends keep consistent.The error between the simulation data and measurement data may be mainly caused by uncertain material property and measurement accuracy.

Conclusions
In this paper, the energy dissipation contributions of the quality factor of the MCVG during vibration are established and verified experimentally.Some conclusions can be obtained.First, the thermoelastic damping and anchor damping dominate the quality factor of MCVG when the pressure level is less than 5Pa.Corresponding Q of these two damping reach the order of 10 5 , and Q the is a little less than Q anc .Nevertheless, air damping becomes dominant dissipation source with the pressure level beyond 5Pa.Second, temperature dependence characteristic of the total Q is dominated by air damping dissipation when the pressure keeps at the level of 10Pa.It can be also predicted that contribution of the thermoelastic damping and anchor damping to temperature dependence characteristic of total Q becomes larger with the pressure level decreasing.Last, the practical measurements prove that the error of the established damping dissipation model could predict the quality factor of the MCVG with an acceptable error.

Figure 1 :
Figure 1: The schematic of the simulated MCVG.

Figure 2 :
Figure 2: Energy distribution of thermoelastic damping of the MCVG at the temperature of 20 ∘ C.

Figure 3 :
Figure 3: Q the as a function of temperature.

Figure 4 :
Figure 4: FEM of the MCVG using PML substrate as finite space marked in yellow.

Figure 5 :Figure 6 :
Figure 5: PML-based energy propagation of the anchor damping at the connection between the beam and anchor.

Figure 8 :
Figure 8: Quality factor Q air versus pressure P at the temperature of 20 ∘ C.

Figure 9 :
Figure 9: Quality factor Q air versus temperature T at the pressure of 10Pa.

Figure 10 :
Figure 10: Total quality factor Q tot versus pressure P at the temperature of 20 ∘ C.

Figure 11 :
Figure 11: Quality factor Q tot versus temperature T at the pressure of 10Pa.

Figure 13 :Figure 14 :
Figure 13: Test system of Q tot versus temperature.

Figure 15 :Figure 16 :
Figure 15: Test system of Q tot versus pressure.

Table 1 :
GEOMETRICAL PARAMETER AND MATERIAL PROPERTIES.

Table 2 :
QUALITY FACTOR OF DAMPING DISSIPATION.