Advanced CAD System for Electromagnetic MEMS
Interactive Analysis (Academia)

Contractor: Stanford University
Agreement Number: F30602-96-2-0308
Project Number: HJ1500-3221-0599
DARPA Order Number: N-7-0504/AO E117/12
Contract Period: September 13, 1996 - September 12, 1999
Date: October 16, 1998
Sponsor: Department of the Air Force
Technical POC: Robert W. Dutton
Report Prepared by: Robert W. Dutton

RESEARCH HIGHLIGHTS:

• Designed unified parameter extraction test bed to extract parameters that are consistent among 3 extraction methods.
• Designed Canonical Benchmark Problem that alleviates dielectric charging problems allowing: both electrical (C-V) and interferometric characterization; 2D simulation and verification.
• Defined unified and consistent parameter extraction methodology for MCNC MUMPs structures which will be automated (web-based).
• Prototype of domain decomposition code properly decomposes real geometries (canonical test structures of the RF switch & comb drive) into regions of 1-D, 2-D, and 3-D simulation.
• New hybrid approach to web-based software interface combines Java and existing technology (HTML, CGI, perl) provides an Internet mechanism--- compatible with Netscape 4.07---for creating MEMS geometry and grid.

Task 2

• Assembly of second-generation fatigue/tensile test tool completed.
• Strain rate sensitivity observed in tensile tests for aluminum beams.
• First-generation dedicated stress sensors now being fabricated.

Task 3

• Successful release of structures consisting of a combination of materials including bi- and tri-layers of Ti/W and silicon nitride, used in an effort to "cancel" stresses in MEMS films, has been achieved.
• Key issues pertinent to an oxygen plasma release (relevant to a wide variety of potential MEMS processes) have been carefully characterized and previous (show stopper) problems have been solved.
• Numerous metal films (Cr, Ti, Pt, and Ir) were deposited, using electron beam evaporation, and results comparing Ion Beam Assistance (vs. without using the ion beam) on film stress are reported.

I. RESEARCH STATUS:

Task 1: Software Development (Prof. Robert W. Dutton)

Develop and demonstrate fully-integrated, FEM-based prototype solver capabilities to model behavior and fabrication process-dependency of MEMS devices. This will include capabilities to consider multi-physics and materials dependencies as well as other process induced factors--geometry effects due to deposition/etching. The overall tool integration strategy will be to develop and test key components that overcome limits currently seen to be "bottlenecks" in commercial systems. Application-specific lumped modeling will be developed and used to guide the overall direction of the CAD efforts, including improved parameter extraction schemes that cross-link device performance with layout parameters that can in turn support design (and hopefully optimization) of practical MEMS devices (i.e., RF switch).

Approach:

1. Layered access in process specification across levels of: layout, process specification, geometry definition, grid and constitutive models
2. Enhanced element technology for discretization of multi-physics systems of equations (e.g. hp-adaptivity)
3. Unified approach to interface geometry and gridding through use of servers.
4. Demonstration of an integrated MEMS simulation capability that supports: internet access to modeling, open interface standards and parameter extraction (for macromodeling) in support of the RF switch application.
5. Prototyping of benchmark (canonical) test structures--both computationally and experimentally--that demonstrates capabilities (and error limits) of models and tools.

Progress:

Measurements of the bowing of fixed-fixed beams has been used to extract the residual strain and thickness of beams fabricated in the MCNC MUMPs process. The extracted thickness is consistent with that measured using an interferometer. Pull-in voltages are then used to extract Young's modulus and residual stress independently. The residual stress is consistent with that extracted from bowing measurements. Alternatively, the sacrificial gap can be extracted instead of the residual stress. Capacitance-voltage measurements are then used to determine the effective thickness of the nitride dielectric. The simulation fit, however, is not very good at this point.

The goal is to automate the entire process to allow the extraction of the 5 main parameters necessary to model a device in the MCNC MUMPs process: Young's modulus, residual strain, layer thickness, sacrificial gap and dielectric thickness. A Web-based interface will automatically extract these parameters from measurements of fixed-fixed beam bowing, pull-in voltages and CV curves.

The Canonical Benchmark Problem was redesigned to allow the device to be simulated purely in 2D if necessary. This allows the parameter extraction framework to be directly extended to simulate this device. Double- pull-in-electrode structures allow sharp measurements of pull-in voltages as a function of the bias applied to the second electrode. This serves as a backup to capacitance measurements which might be affected by dielectric charging and surface roughness. The first samples should be available in mid-October and the remainder in mid-December.

The effects of capacitances, resistances and residual charges, on the performance of microelectromechanical devices was also studied [1]. Parasitic capacitances are shown to limit the performance of a capacitor-stabilized full-gap positioner [1]. The behavior of general three-electrode structures is also analyzed.

The domain decomposition technique detailed in [4] has shown great promise to utilize a hierarchical representation of process simulation data to generate geometry of interest to the MEMS community. The technique calls for dividing a wafer into 1-D, 2-D, and 3-D simulation regions. In the 1-D regions, fluxes calculations predict the thickness of material deposited. 2-D regions utilize the 2D levelset geometry server developed at Stanford. In the 3-D regions, two options are being explored. The first is to utilize a pseudo-3D technique [2], the second is to perform full 3-D levelset simulation of the process simulation using commercially available tools (Taurus by TMA).

Software Interoperability: Work continues investigating VRML as a platform independent, portable representation of MEMS geometry. Although recent versions of some commercial vendors software output VRML files (MEMCAD 4.2 by Microcosm), some known software bugs (specifically in Ideas used by Microcosm) have significantly delayed the testing of the VRML geometry. "Work arounds" for the bugs have been found, and work is underway to test geometric manipulations and meshing of the VRML geometry.

Due to security issues involved in web-based applications, several new approaches have been successfully explored. First, a user friendly method of file upload (using HTML/CGI) was utilized to circumvent the Java security restriction of accessing files on the local (client) system. In addition, due to firewall restrictions, a new server model (utilizing CGI) was developed that allows the successful client execution of geometric and meshing engines. Since this approach requires no special software beyond the current release of commercial web browsers (Netscape 4.07), platform independent, Internet based prototyping of geometry and meshing has been achieved.

Difficulties/Problems:

• Simulated capacitance-voltage curves still do not fit measurements very well. The effects of compliant surface roughness and dielectric charging are being incorporated into simulations.
• Simulated deposition of multiple layers of material has uncovered additional robustness issues in the pseudo-3D technique.
• Combining results of hierarchical simulation results has posed some problems in the solid modeler (Shapes) and alternative methods of combining data are being explored.

Goals targeted for the next quarter:

• Measure and simulate canonical benchmark in 2D using Abaqus.
• Characterize boundary conditions and other topographical discontinuities of relevance to the canonical benchmark.
• Improve robustness of pseudo-3D technique to handle multiple depositions.
• Begin evaluation of commercial 3D process simulation tools (i.e. Taurus).
• Generate realistic geometry and mesh of canonical test structures (RF switch, comb drive) and prepare for analysis.

References:

1. E. Chan, "Effects of capacitors, resistors and residual charges on the static and dynamic performance of electrostatically-actuated devices", submitted to SPIE Design, Test and Microfabrication of MEMS/MOEMS 1999.
2. N.M. Wilson, Z.K. Hsiau, R.W. Dutton, and P.M. Pinsky, "A Heterogeneous Environment for Computational Prototyping and Simulation Based Design of MEMS Devices", SISPAD `98, September 2-4, 1998.
3. N. M. Wilson, R. W. Dutton, and P. M. Pinsky, "Utilizing Existing TCAD Simulation Tools to Create Solid Models for the Simulation Based Design of MEMS Devices", IMECE, November 15-20, 1998.
4. N. M. Wilson et al "title TBD" being submitted to ASME

   Task 2. Characterization of MEMS Material Models (Prof. John Bravman)

   Objective:

   Develop extraction schemes for materials properties using simplified MEMS test structures. In combination with the CAD tools efforts, the physical models (including microstructure) will be evaluated and modified as needed. The scope of material characterization efforts include both standard metal layers (possibly graded as well) and reliability issues such as plastic yield and fatigue. The extraction and validation steps for physical models is a key linking between the three sub-tasks.

   Approach:

   1. Mechanical characterization of MEMS materials with dependencies on processing and microstructure.
   2. Reliability study including fracture and fatigue.
   3. Feedback to the constitutive models in the FEM-based modeling Task (#1).

   Progress:

   1. Fatigue Study
   Assembly of second-generation test tool has been completed. Calibration of piezoelectric actuator, load cell and thermocouple have been performed. Custom written LabVIEW software has been modified to allow faster data acquisition to monitor tests at higher strain rates. Since very small loads (mN range) are to be measured, the minimal drift with temperature in load cell output becomes important. To combat this problem the load cell output is being amplified. To provide a constant temperature environment a thermally insulated box has been built, which keeps temperature constant to within two one hundredths of a degree over a 24 hour period, in an air-conditioned room with two degrees overall temperature variation.

   Tensile/fatigue test samples of 2 and 4 micron thickness have been micromachined from aluminum films using the third generation mask set. The main innovation in this mask set is the pop-out design of the dies. This design allows for easy detachment of the dies from the wafer without subjecting the samples to the cooling liquid in the wafersaw. This much more benign process and resulted in a dramatically increased sample yield per wafer. Also the dies in the third generation mask set have a larger window to allow easier sample prep for TEM analysis, and they contain a thinner silicon support which decreases cutting time once the samples are clamped. This support can now be severed within a fraction of a second by just touching it with a rotating diamond blade.

   Tensile tests on the new samples have been performed at strain rates between 0.0002 and 2 strain/sec. To provide better statistical backing at least five samples from different wafers have been tested at all strain rates. From these results a strain rate dependence of the yield strength is observable.

   Difficulties/Problems:

   • Load cell output variations with minimal temperature changes (hundredths of degrees) still cause problems for longer term tests. An additional temperature compensation (on top of the manufacturer supplied solution) is being investigated.
   2. In-situ Stress Measurement

   First-generation dedicated stress sensors are currently being fabricated. The sensor consists of two wafer levels, a membrane wafer and a tip wafer. Six masks from the existing 7-mask set are being used. Processing has gone smoothly so far. Supporting membrane of low stress LPCVD silicon nitride has been defined in the membrane wafer. Tip and vent have been defined in the tip wafer. In 2/3 of the membrane wafers being processed, corrugations on the electrode side of the membrane have been included. The other 1/3 are processed without corrugations, but simply with square-shaped recesses to avoid tip-to-membrane contact. Corrugations were initially designed to ensure sufficient flexibility of the supporting membrane, and hence a reasonably low deflection voltage (below 100V) for tunneling. However, corrugations will also result in thickness non- uniformity of the supporting membrane, which might cause stress non-uniformity in the film to be deposited on top. We hope to compare the tunneling effects from these two configurations, i.e. with and without corrugations, so that we can optimize the supporting membrane for accurate stress measurement.

   A 3" sputtering system will be used for initial ex-situ stress measurements on the existing sensors. Tantalum is to be used as the test material. Ta films with different thicknesses have been deposited under different deposition conditions. Stresses in these films have been measured by wafer curvature. By increasing the sputtering current, we are able to bring the film stress from slightly tensile to 2GPa tensile. By decreasing the ambient Argon gas pressure, we can bring the film stress from slightly tensile to 2GPa compressive. Now that we have determined the relationship between film stress and deposition conditions, the system is ready to be used for ex-situ trials.

   Goals targeted for the next quarter:

   1. Fatigue Study
      
      • Perform relaxation experiments and fatigue tests at varying displacements (strains). Correlate fatigue test load drops with load drops in relaxation experiments. Investigate surface morphology changes stemming from cyclic loading with AFM. Start fabricating Iridium test beams.
   2. In-situ Stress Measurement
      • Continue with stress sensor processing. This includes fabrication of the membrane electrodes and tip electrodes, release of the supporting membrane, bonding, and dicing. We plan to experiment with bonding, and dicing through two-wafer stacks.
      • Conduct ex-situ stress measurement on the existing sensors in the 3" sputtering system. We shall start with tensile Ta. The problem of unstable tunneling of the existing sensors still needs to be resolved.
   References:
   1. G. Cornella, R.P. Vinci, R. Suryanarayanan Iyer, R.H. Dauskardt, J.C. Bravman, "Observations of Low Cycle Fatigue of Al Thin Films for MEMS Applications", to be published in MRS Spring Symposium Proceedings, Volume 518: Microelectromechanical Structures for Materials Research, 1998.
   2. R.P. Vinci, G. Cornella, R. Suryanarayanan Iyer, R.H. Dauskardt, J.C. Bravman, "Low Cycle Fatigue in Thin Film Al Beams", Presented at the Ringberg Symposium on Mechanics and Design of Small-Scale Materials and Systems, Tegernsee, Germany, June 18, 1998.

   Task 3. MEMS Device Modeling and Design (Prof. Greg Kovacs)

   Objective:

   Design and implementation of MEMS structures, including materials parameter extraction, will be used to test the CAD and physical models based on MEMS devices of interest to DARPA. The experimental measurements made with these structures will facilitate the critical evaluation of models, physical parameters and overall simulation accuracy of CAD for MEMS devices. The test vehicles from this work also support canonical benchmarking of both new materials for MEMS and accuracy of Composite CAD in MEMS applications, specifically the RF switch.

   Approach:

   1. Design/implementation of test structures for materials characterization.
   2. Concurrent building of prototype MEMS application devices (RF switch).
   3. Measurement, parameter extraction, and testing of CAD models.

   Progress:

   The overall improvement of process flow and thermal cycling of RF switch fabrication (in support of the Canonical Test Device work), including the use of new, promising MEMS materials has resulted in practical test devices as well as several very promising new options for materials that could dramatically enhance switch performance. Two key accomplishments in this area are as follows:

   • We have successfully released structures consisting of a combination of materials including bi- and tri-layers of Ti/W and silicon nitride in an effort to "cancel" stresses from the films. The measured bow (deflection relative to the beam edge) of the doubly supported beams indicates that the upper layer of Ti/W of the tri-layer structures reduces the overall deformation compared to the Ti/W-Nitride films we had explored earlier.
   • We have also explored key issues that are pertinent to an oxygen plasma release (relevant to a wide variety of potential MEMS processes). Our previous experiment had shown that the overall curvature of the samples released at 200 Watts and at 100 Watts were not consistent, perhaps due to plasma-induced heating. By halving the power yet again (50 Watts) there has been a marked decrease in the amount of deflection of the doubly supported beams.
   In terms of extraction methodology for mechanical properties of the test devices, the TESTAROSA array will be dramatically simplified (several structures have been determined to be unnecessary) and the overall approach is being more directly focused toward the Canonical Test Devices (and the overall approach discussed in Task #1). Consistent and significant deformation caused by stress gradients in the RF/MEM films suggests that using the Canonical Devices is in fact the most promising method to be able to generically test any thin film, regardless of its stress state. Support of specific DoD applications such as AFRL, using these structures, is discussed below (see Technology Transition).

   In addition to the process improvements discussed above, significant progress has been made in the deposit and measurement of unique films, including Iridium and Tungsten. The evaporator has been characterized as well as alternatives in the lift-off process. Numerous metal films (Cr, Ti, Pt, and Ir) were deposited in the Innotec electron beam evaporator, comparing Ion Beam (vs. without beam) Assistance. All films (except Ti) produced without Ion Beam Assistance were measured to have enormous tensile stresses. Highlights of specific materials results include:

   • Cr and Ir had stress values so large that the stress measurement tool could not accurately determine the value. (Cr & Ir: >1,800 MPa)
   • Tensile stress in Cr and Ir deposited with Ion Beam Assistance were reduced to 100 MPa and < 300 MPa, respectively.
   • Gun current dependences (spacing and angle) have been studied extensively and will require more refining of the process.
   • Stress values for Pt: 350-400 MPa; Ti 250 MPa (compressive) were measured.
   • Pt is the only metal to date that we can reproducibly shift from a tensile stress state to a compressive state. Values attained to date are 350-400 MPa in tension, a group with stresses in the range -25 to +100 (the compressive/tensile breakpoint), and -250 MPa compressive.

   Difficulties/ Problems

   • Our first attempt at depositing Ir onto the Testarosa mask set did not anticipate the incredibly high temperatures inside the evaporator. The organic materials intended to allow clean lift-off do not hold up at temperatures in excess of 200 degrees C. This quarter we will use more thermally robust lift-off structures.
   Goals for Next Quarter:
   • Successfully deposit Ir film on Testarosa design set and parameter extraction for these films
   • Complete process flow development for RF Switch, including fabrication of first set of RF MEMS structures
   • Analyze Ti/W-Nitride structures to determine consistent behavior
   Specific targets for the Innotec:
   • The goal of this portion of the project is to find the ion gun operating parameters that will give us the ability to modify the stress state between compressive and tensile states so that we can produce controlled stress gradient films that will not curl or buckle, when released.
   • To accomplish this we intend to continue to move the gun closer to the substrates. Work is now underway to modify the chamber to allow us to position the gun within 1 inch from the substrate (currently we are only able to position the gun to within 10 inches from the substrates).
   • Further depositions trials will be continued for Ir and W, our best candidate materials for the RF switch. Optimization of the Pt process for thicker film applications(1-2 microns) will also continue.
   Technology Transfer:
   We have also attracted the attention of TI/Raytheon to contribute to a combined Stanford/ Industry RF MEMS fabrication run. Their designs (for DoD applications), along with the AFRL contribution will be combined with Stanford's to permit exploration of unique metals for use with RF MEMS. We have also begun discussions with SPAWAR regarding their interests in RF switches and MEMS modeling.

   II. Financial Information: AF F30602-96-2-0308

   ********************Still processing university papers********************
   

   Project Balance as of September 30, 1998:

   Fund usage	Amount
   Total Amount Funded	$....
   Salaries and Wages	$...
   Staff Benefits	***
   Travel	***
   Expendable Mat./Services	***
   Repairs and Maintenance	***
   	--------
   Total Direct Costs	$...
   Capital Equipment	***
   Student Aid (Tuition)	***
   Indirect Costs	$...
   Total Expenses to date	$...
   	----------
   Funds to date, 9/30/98	$ ...

   ---

   Back to the MEMS home page