NANOHUB TOOL MANUAL Micro-Electro-Mechanical Simulation Lab (MEMSLab) User’s Guide Oluwatosin Adeosun Sambit Palit Ankit Jain Muhammad A. Alam Sponsors: Network for Computational Nanotechnology (NCN) Purdue University Center for Prediction of Reliability, Integrity and Survivability of Microsystems (PRISM) Summer Undergraduate Research Fellowship (SURF) 0 Introduction Electro-mechanical (EM) actuators have diverse applications both as an analog (tunable) and a digital (switch) element. Analog applications involve continuous position control of a movable electrode, e.g., external cavity tunable lasers, reflective diffraction grating, deformable mirrors for adaptive optics, RF-MEMS varactors, etc. Digital operation requires only a binary position control of the movable electrode e.g., RF-MEMS capacitive/ohmic switches, interferrometric Mirasol displays, etc. An electro-mechanical actuator can be modeled as a spring-mass system, with a point mass of value π, attached to a linear spring of spring constant π. The figure below illustrates the system. The dielectric on the bottom electrode is present to avoid electrode-electrode contact. The dynamic behavior of the spring mass system is as follows – π π2 π¦ ππ¦ = πΉππ − πΉπ − π . 2 ππ‘ ππ‘ Here πΉππ is the downward acting electrostatic force, πΉπ is the restoring spring froce, π¦ is the displacement of the mass from the equilibrium position and π is the damping coefficient. In this case, πΉπ = π(π¦0 − π¦) and πΉππ = π0 π΄π 2 2 π¦ 2( π +π¦) , where π΄ is the elecrode area, π is the applied voltage, π¦π is the dielectric thickness, ππ is the dielectric ππ constant and π¦0 is the air-gap. In general, the top electrode of electromechanical actuators have a specific support configuration, and the electrode does not move freely unlike the spring mass system. In order to determine the shape of the beam for a given voltage, the Kirchhoff-Love plate theory is used. The dynamic equation for a differential area element on the electrode is given by – ππ» π2π¦ π0 π 2 πΈπ»3 π4π¦ ππ¦ = − −π . 2 π¦π ππ‘ 2 12(1 − π 2 ) ππ₯ 4 ππ‘ ( + π¦) ππ This equation is solved using the Finite Difference method for boundary conditions imposed based on the electrode geometry and support configurations. Here π is the density of the top electrode material, π» is the thickness of the electrode, πΈ is the Young’s modulus, and π is the Poisson’s ratio. 1 The actuation behavior of a typical electromechanical actuator is hysteretic in nature, primarily defined by four performance metrics – the pull-in voltage (πππΌ ), the pull-out voltage (πππ ), the below pull-in capacitive response (πΆππ − π) and the post pull-in capacitive response (πΆππ − π). These are indicated in the figure below. In the compact model developed by Palit et. al., these four performance metrics can be represented by simple analytical scaling expressions, and can be estimated without the need of computationally expensive finite element simulations and/or experimental fabrication and subsequent characterization. The scaling relations for the pull-in voltage (πππΌ ), pull-out voltage (πππ ), below pull-in πΆππ − π response and post pull-in πΆππ − π response are given by – πΎ0 π§02 πππΌ = πΌππΌ √ (π − 1)πΆππΉπΉ πππ = πΌππ √ = πΌππΌ π0 , πΎ0 π§02 π§0 πΎ ( ) = πΌπ0 π1 , (π − 1)πΆππ ππ πΆππ π = πΜ ( ), πΆππΉπΉ πππΌ πΆππ − πΆππ π π = ( ) . πΆππ − πΆππ πππ Here, πΌππΌ , πΌππ , πΎ, πΜ and π are scaling variables/functions which depend purely on the top electrode geometry and support configuration, but are independent of the specific physical dimensions and material properties. Here πΎ0 = πΈπ» 3 6(1−π2 )πΏ3 , π§0 = π¦0 + π¦π ππ , πΆππΉπΉ is the off-state capacitance, πΆππ is the theoretically possible maximum on-state capacitance (when the entire top electrode is in contact with the dielectric), πΆππ is the capacitance at the point of pull-out instability. For exact values of these scaling parameters for cantilever, fixedfixed and circular configurations, please refer to Palit et. al. MEMS actuators have multiple design applications. Understanding their behavior as well as the ability to predict their actuation characteristics and voltage response is important when designing these actuators. In order to determine how these devices will behave, designers have to perform computationally expensive finite element 2 simulations or design experiments that can be very time consuming. This tool is created to allow users to enter basic information about a MEMS actuator and obtain a reasonably accurate estimation of the actuation response of the actuator. A compact model based on a scaling theory provides an almost instantaneous estimation of the actuation voltages (pull-in voltage and pull-out voltage) and the hysteretic CV characteristics (for both below pullin and post pull-in states). A more comprehensive numerical simulation option, which employs the Kirchoff-Love plate equation, is also available. Quasi-static and dynamic response of the cantilever, fixed-fixed beam and circular shapes can be obtained using the numerical simulation. A method-of-moments based 3D Poisson solver can also be invoked if necessary. For users who want to understand the basics of MEMS actuator operation, quasi-static and dynamic simulations can be performed for a simple spring-mass model. In this manual, we will explain how to use the tool, describing the input pages and providing illustrative examples of outputs. 3 Inputs and Settings Input page 1: Simulation settings The following input options are available in this input panel: a) Electrode geometry: This lets the user choose the geometrical shape (and support configuration) of the actuating top electrode. There are four available choices: a. Cantilever: Rectangular beam, fixed support at one end, free on the other end. b. Fixed-Fixed: Rectangular beam, fixed support at both ends. c. Circular: Circular beam, fixed support along the circumference. d. Spring-mass: A point mass attached to a linear spring. These options for the available electrode geometries have been explained pictorially on the tool. b) Simulation type: There are two available options – “Compact model” and “Numerical model”. Choosing the compact model will utilize the scaling theory developed by Palit et. al. for obtaining actuation voltages and CV characteristics for the cantilever, fixed-fixed beam and circular geometries. The numerical model on the other hand will solve the Kirchhoff-Love plate equation instead. Note that choosing this option will increase the computational time significantly. If the electrode geometry is chosen to be “Spring-mass”, this option is disabled because a spring-mass system is always solved numerically in this tool. c) 3D Poisson: This switch enables the use of a method-of-moments based Poisson solver for calculating the electrostatic force on the actuating membrane. This option is available only for numerical simulation. Note that choosing this option will increase the computational time significantly. When disabled, the electrostatic force is calculated by the parallel-plate assumption i.e., πΉππ = π0 π 2 2 π¦π 2( +π¦) ππ . 4 Input page 2: Voltage settings This page allows the user to select the type of analysis to be performed on the geometry configuration based on a simulation mode selected by the user on input page 1. Even though all the analysis modes are initially turned off by default, at least one analysis mode must be enabled. Out of the following, only the “DC analysis” mode is enabled if compact model simulation is invoked on input page 1. The following input options are available in this input panel: a) DC analysis: Perform quasi-static forward and reverse voltage sweeps on the actuator. There are three user input parameters for DC analysis. a. Voltage range: User can chose between a “Default” sweep, or specify a “Custom” sweep. The default sweep performs a forward and reverse sweep between 0 and an estimated value for 1.5πππΌ , using approximately 200 voltage points for calculation. b. Increment by: Voltage step (Δπ) for custom sweep. Units: V. Default: 0.1V. c. Stop at: Maximum voltage (ππ ) for custom sweep. Units: V. Default: 10V. Both forward and reverse sweeps between 0 and ππ are performed if custom sweep is selected. The latter two options are disabled if default sweep has been selected. b) Transient analysis: Performs the time-dependent dynamic simulation on the actuator. There are a total of five parameters available to the user for specifying the time-dependence of input voltage with time. a. Voltage type: User can choose between a “Step voltage” or “Ramp voltage”. b. Run Time: Duration for which the step voltage is applied (π‘π π‘ππ ). Units: s. Default: 0.01s. c. Voltage: Step voltage value (ππ π‘ππ ). Units: V. Default: 3V. 5 d. e. Ramp rate: Rate of change of applied voltage (π ππππ ). Units: V/s. Default: 100V/s. Final voltage: Maximum ramped voltage value (πππππ ). Units: V. Default: 3V. Inputs (b) and (c) are enabled if step voltage has been selected as the “Voltage type”. On the other hand, inputs (d) and (e) are enabled if ramp voltage has been selected. For step voltage analysis, the time dependence of applied voltage is given by: π(π‘) = {π 0 π π‘ππ π‘<0 π‘π π‘ππ ≥ π‘ ≥ 0 For ramp voltage analysis, the time dependence of applied voltage is given by: 0 π(π‘) = {π ππππ π‘ π‘<0 πππππ ≥π‘≥0 π ππππ Note that a nominal damping coefficient (100 N.s/m3 – value is per unit area) is used for transient analysis so that any resulting oscillations do not continue forever. The simulation is stopped as soon as a pull-in event is detected and no bouncing of the electrode at the dielectric surface has been accounted. 6 c) AC analysis: Performs time-dependent small-signal dynamic simulation on the actuator. There are a total of four input parameters available to the user for specifying the variance of input voltage with time. a. DC bias: DC biasing point for small-signal analysis (πππ ). Units: V. Default: 2V. b. AC amplitude: Small signal sinusoid amplitude (πππ ). Units: V. Default: 0.1V. c. Frequency: Small signal sinusoid frequency (π). Units: Hz. Default: 1000Hz. d. Time: Duration for which small signal voltage is applied (π‘ππ ). Units: s. Default: 0.01s. For this analysis mode, the time dependence of the applied voltage is given by: 0 π(π‘) = { πππ + πππ sin(2πππ‘) π‘<0 π‘ππ ≥ π‘ ≥ 0 Note that a nominal damping coefficient (100 N.s/m3 – value is per unit area) is used for AC analysis. AC analysis is performed only in the below pull-in state. The simulation is stopped as soon as a pull-in event is detected and no bouncing of the electrode at the dielectric surface has been accounted. 7 8 Input page 3: Dimensions and Material Properties The following input parameters are available in the two tabs of this input page: a) Length of Electrode: Length scale of the top actuating electrode (πΏ). Units: οm. Default: 400 πm. b) Width of Electrode: Width of the top actuating electrode (π). This option is disabled if the circular geometry is selected on input page 1. Units: οm. Default: 100 πm. c) Thickness of Electrode: Electrode thickness (π»). Units: οm. Default: 2 πm. d) Air-gap: Initial air-gap between actuating top electrode and top of the dielectric (π¦0 ). Units: οm. Default: 3 πm. e) Dielectric thickness: Thickness of the dielectric layer deposited on the bottom electrode (π¦π ). Units: nm. Default: 250 nm. f) Number of grids: Number of grids placed along πΏ (ππ₯ ). The same grid spacing is used along the width direction. This input is enabled only if Numerical model is selected in input page 1. Default: 40. 9 g) Choose electrode material: Select between a set of pre-defined materials, or define a “Custom” material for the top actuating electrode. Default: Nickel. h) Young’s modulus: Young’s modulus (πΈ) of the top actuating electrode. Units: GPa. Default: 200GPa. i) Poisson’s ratio: Poisson’s ratio (π) of the top actuating electrode. Units: none. Default: 0.31. j) Density: Density (π) of the top actuating electrode. Units: kg/m3. Default 8900 kg/m3. k) Choose dielectric material: Select between a set of pre-defined materials, or define a “Custom” material for the dielectric layer on the bottom electrode. Default: Silicon nitride. l) Dielectric constant: Dielectric constant (ππ ) of the dielectric layer. Units: none. Default: 7.9. Inputs (h), (i) and (j) are enabled only if “Custom” option is selected while specifying the top electrode material. Similarly, input (l) is enabled only if “Custom” option is selected while selecting the dielectric material. If the “Spring-Mass” geometry is selected from input page 1, the effective spring constant (π) and the mass (π) is estimated using the inputs on this page as follows: π= πΈπ»3 π (1 − π 2 )πΏ3 π = ππ»πΏπ 10 Illustrative examples DC analysis: Spring mass model This section summarizes the results obtained for DC analysis performed on the spring-mass model. The figure below shows the πΆ − π characteristics of the system. Default parameters are used. The actuation voltages πππΌ and πππ and indicated on the figure. The blue curve is the forward sweep, the red curve is the reverse sweep. VPO VPI The figure below plots the steady state position of the actuating electrode as a function of voltage for the forward sweep. Observe that at πππΌ , the electrode snaps down and comes into contact with the dielectric layer. The initial beam position is given by π¦0 . VPI 11 DC analysis: Cantilever with compact model This section summarizes the results obtained for DC analysis performed on the cantilever geometry, using the compact model based on scaling equations derived by Palit et. al. Similar qualitative results may be obtained for fixed-fixed beam and circular geometries as well. The figure below shows the πΆ − π characteristics of the system. Default parameters are used. The actuation voltages πππΌ and πππ are indicated on the figure. The blue curve is the forward sweep; the red curve is the reverse sweep. Observe that the πΆ − π behavior in the pulled-in state is markedly different from that obtained using the spring-mass model because the electrode area in contact with the dielectric varies with applied voltage. VPO VPI 12 DC analysis: Cantilever with numerical model This section summarizes the results obtained for DC analysis performed on the cantilever geometry using numerical simulation based on the Kirchhoff-Love plate theory. Similar qualitative results may be obtained for fixed-fixed beam and circular geometries as well. The figure below shows the πΆ − π characteristics of the system. Default parameters are used. The blue curve is the forward sweep; the red curve is the reverse sweep. VPO VPI The figure below plots the steady state position of the maximum deflection point on the actuating electrode as a function of voltage for the forward sweep and reverse sweep. VPO VPI 13 If we increase the number of grid points, the steps appearing on the πΆ − π characteristics disappear, and we get smoother curves. The figure below compares the results obtained from using the default parameters (ππ₯ = 40) and those obtained by using ππ₯ = 100. Also plotted are πΆ − π characteristics as predicted by the compact model. The steps, therefore are a result of coarse gridding. However, increasing the number of grid-points will increase the simulation time drastically. However, one observes that even by using a coarse grid, the actuation voltages are predicted reasonably well. Nx=100 Nx=40 Compact model In the figure below, we evaluate the effect of using the 3D Poisson solver over the simple parallel-plate assumption for computing electrostatic force. Again, we observe that even though using the 3D solver gives us a slightly different value for the actuation voltage; the post pull-in πΆ − π characteristics are identical to the results from the parallel plate assumption. One must therefore evaluate the tradeoff between getting a more accurate result using the 3-D Poisson solver and the massively high computational cost required for the simulation. Without 3D With 3D Compact model 14 In the DC analysis mode, it is possible to observe the actual deflections on the beam for all voltage points on the πΆ − π sweep. The figures below show the beam shapes at voltage just below πππΌ (voltage point = 66) during the forward sweep, and at voltage just above πππ (voltage point = 170) during the reverse sweep. The beam shapes can be viewed in different modes as shown. For these simulations, default parameters were used. Voltage point 66: Voltage point 170: 15 Transient analysis: Spring-mass with step voltage This section summarizes the results obtained for transient step voltage simulation using the spring-mass model. The figure below shows the πΆ − π‘ characteristics for the simulation. In this case, ππ π‘ππ = 1.5π, and other parameters were default. Observe that the electrode eventually stabilizes in the air-gap. The figure below plots the position of the actuating electrode as a function of time. 16 The figure below shows the πΆ − π‘ characteristics for the simulation. Default parameters were used in this case. Observe that the electrode eventually snaps down to the dielectric. Simulation stops as soon as the pull-in event is detected. The figure below plots the position of the actuating electrode as a function of time. 17 Transient analysis: Spring-mass with ramp voltage This section summarizes the results obtained for transient ramp voltage simulation using the spring-mass model. The figure below shows the πΆ − π‘ characteristics for the simulation. Default parameters were used in this case. Observe that the electrode eventually snaps down to the dielectric. Simulation stops as soon as the pull-in event is detected. The figure below plots the position of the actuating electrode as a function of time. The figure below plots the variation of applied voltage as a function of time. 18 For this simulation, one may also plot the variation of capacitance and beam position with respect to voltage. These results are not shown here. 19 Transient analysis: Cantilever with numerical model and step voltage This section summarizes the results obtained for transient step voltage simulation on the cantilever geometry. Similar qualitative results may be obtained for fixed-fixed beam and circular geometries as well. The figure below shows the πΆ − π‘ characteristics for the simulation. Default parameters were used. Observe that the electrode eventually stabilizes in the air-gap. The figure below plots the position of the point on actuating electrode with maximum displacement as a function of time. 20 The figure below shows the πΆ − π‘ characteristics for the simulation. In this case, ππ π‘ππ = 8π, and other parameters were default. Observe that the electrode eventually snaps down to the dielectric. Simulation stops as soon as the pull-in event is detected. The figure below plots the position of the point on actuating electrode with maximum displacement as a function of time. 21 Transient analysis: Cantilever with numerical model and ramp voltage This section summarizes the results obtained for transient ramp voltage simulation on the cantilever geometry. Similar qualitative results may be obtained for fixed-fixed and circular geometries as well. The figure below shows the πΆ − π‘ characteristics for the simulation. Default parameters were used in this case. Observe that the electrode eventually snaps down to the dielectric. Simulation stops as soon as the pull-in event is detected. The figure below plots the position of the point on actuating electrode with maximum displacement as a function of time. 22 The figure below plots the variation of applied voltage as a function of time. For this simulation, one may also plot the variation of capacitance and beam position with respect to voltage. These results are not shown here. 23 AC analysis: Spring-mass This section summarizes the results obtained for AC voltage simulation using the spring-mass model. The figure below shows the πΆ − π‘ characteristics for the simulation. In this case, πππ = 1π, and other parameters were default. The figure below plots the position of actuating electrode as a function of time. 24 The figure below plots the variation of applied voltage as a function of time. 25 AC analysis: Cantilever with numerical model This section summarizes the results obtained for AC voltage simulation on the cantilever geometry. Similar qualitative results may be obtained for fixed-fixed and circular geometries as well. The figure below shows the πΆ − π‘ characteristics for the simulation. Default parameters were used. The figure below plots the position of the point on actuating electrode with maximum displacement as a function of time. The figure below plots the variation of applied voltage as a function of time. 26 27 Exercises Switch design Design a cantilever switch having a pull-in voltage of approximately 40π, pull-out voltage of approximately 3π and off-capacitance of 100ππΉ. Change only the beam thickness (π»), dielectric thickness (ππ ) and width (π), keeping other parameters as default. Use the “compact model” simulation mode to iterate between guesses. Verify by running the “numerical simulation” that your designed switch works as expected. Hint1: Changing ππ will affect πππ , but has minimal effect on πππΌ . Hint2: Changing π affects capacitance, but does not change the actuation voltages. Answer: π = 85ππ, ππ = 50ππ, π» = 3.75ππ. Resonance Estimate the natural resonance frequency of a spring-mass system (using the default parameters) at π = 0.5π and π = 2.0π. For this, apply a step of the given voltage. The initial oscillation frequency (before it gets damped out) is the natural resonance frequency at the given voltage. Verify the correctness of the estimated resonance frequencies by observing the AC capacitance change (ΔπΆ) in the cyclo-stationary state for AC simulation with frequencies of 0.9ππ πΈπ , ππ πΈπ and 1.1ππ πΈπ , and using ππ΄πΆ = 0.01π. For this, measure the peak-to-peak difference in capacitance when the oscillation frequency stabilizes to the forcing input frequency (also called the cyclo-stationary state). ΔπΆ at resonance should be maximum because the forcing input frequency matches the natural oscillation frequency, resulting in positive feedback. Answer: At π = 0.5π, ππ πΈπ = 2852π»π§. ΔπΆ(0.9ππ πΈπ ) = 0.199ππΉ, ΔπΆ(ππ πΈπ ) = 0.212ππΉ and ΔπΆ(1.1ππ πΈπ ) = 0.163ππΉ. Note that ΔπΆ maximizes near ππ πΈπ . At π = 2.0π, ππ πΈπ = 2210π»π§. ΔπΆ(0.9ππ πΈπ ) = 0.189ππΉ, ΔπΆ(ππ πΈπ ) = 0.206ππΉ and ΔπΆ(1.1ππ πΈπ ) = 0.180ππΉ. 28