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
Sponsor: Department of the Air Force
Technical POC: Robert W. Dutton
Task 1--
Task 2--
Task 3--
The results on Task 1 represent significant progress both in support of the overall inter-operability goal as well as in definition (and execution) of creating the canonical test structures for the RF switch. The results in Tasks 2 and 3 continue to make strong progress in the demonstration of new materials, including fatigue analysis, for MEMS--currently using aluminum. Ongoing efforts are targeted to shift to other metals with more advantageous physical and electrical properties.
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). The development (and implementation) of a canonical test structure suite of devices will be a key means to test and exploit the results of this task.
Approach:
Progress:
A mask set of various test structures was submitted to MCNC for fabrication in May. This design contains test structures to extract Young's modulus, residual stress, stress gradient and boundary conditions. Three, and possibly four, different methods will be used to extract consistent parameters from these structures. The goal is to identify all the necessary parameters to calibrate a 3-D simulator. Previously, various parameter extractions were performed independently and discrepancies among methods were not investigated.
Another mask set was submitted in early July containing designs for possible Canonical Benchmark Problems. This design alleviates dielectric charging problems allowing both electrical (capacitance-voltage) and interferometric (Zygo) characterization at the expense of more complicated topography. The more complex topography will allow us to assess the efficacy of the topography simulator under development. This mask also contains structures that allow better measurements of the thickness of the various layers.
3-D contact electro-mechanical simulations in both MEMCAD and IntelliCAD were investigated. The results agree to within 5% of 2-D quasi-analytic Abaqus simulations. However, very high mesh density is required on the contact surfaces in 3-D simulations.
Significant progress has been made in the area of geometric modeling. The technique previously detailed using NURBS to represent physically deposited materials [1] had serious limitations due to the lack of robustness in the solid modeler (SHAPES) to handle the needed boolean operations. A new approach, detailed in [2], shows great promise and appears to have the robustness required for MEMS geometric applications. At the lowest level, the solid modeler being used represents all solids using a Boundary Representation (B-rep). In a B-rep, you define objects as regions on underlying curves or surfaces enclosed by specified boundaries. We begin by creating a surface mesh from the physical deposition (SPEEDIE) data. Using this surface mesh as the boundary, a solid can be created which can be handled robustly by the solid modeler. The trade-off is that using the B-rep solid is computationally more expense then some of the alternatives [2].
The impact on mesh generation of using the B-rep for the solids is now being investigated. Preliminary results indicate that this technique has increased the density of the mesh unnecessarily. However, improvements detailed in the goals section on reducing unneeded SPEEDIE simulations will greatly reduce the problem of high mesh refinement due to B-rep solids.
Software Interoperability: Significant progress has been made in the area of utilizing VRML as a portable representation of MEMS geometry. With the advent of the B-rep technique described above, the fundamental problem with importing VRML geometries has been resolved. In addition, several commercial vendors (Microcosm, Intellisense) have or will have VRML export capabilities by September 1998. Progress continues on the Web-based graphical user interface, written in Java, to manage the geometry and mesh generation. Recent releases of commercial web browsers (Netscape 4.5) continue to increase the support of Java and will hopefully lead to full support of our GUI on all major software platforms by the end of the year.
Difficulties/Problems:
Preliminary analysis of the bowing of fixed-fixed beams on the Testarosa test bed show that other factors besides stress, stress gradient and boundary conditions need to be accounted for.
Goals targeted for the next quarter:
References:
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:
Progress:
A. Fatigue Study-- Initial experiments have been performed with the goal of identifying differences, if any, that exist between static and cyclic stress relaxation. Previously, low cycle fatigue tests were performed using 2 micron thick Al beams. These were carried out using a multiple step test method, in which each specimen was cycled with several blocks of constant maximum displacement amplitude. A load drop and its corresponding stabilized state were observed for each displacement block. Recently, static stress relaxation measurements were made at a number of constant displacement amplitudes (chosen to correspond with the cyclic tests). Stress relaxation under static conditions was much more rapid than under cyclic conditions. These results were presented in April at the MRS Spring Symposium [3] and in June at the Ringberg Symposium on Mechanics and Design of Small-Scale Materials and Systems [4].
Stress relaxation under cyclic loading was also observed even in the "elastic" behavior regime, implying significant strain rate sensitivity (or local microplasticity at stress concentrators). This is contrary to the limited strain rate sensitivity seen under monotonic loading in Q2. Strain rate sensitivity data extracted from the recent static stress relaxation tests also indicate significant strain rate sensitivity. Further measurements must be performed to clarify this issue because high frequency (high strain rate) testing will be required in the future. Strain rate sensitivity (and anelastic deformation in the "elastic" regime) could have significant impact on long term reliability of Al MEMS structures, and should be included in predictive CAD models.
In an effort to feed back to the constitutive models in the FEM-based modeling task, established phenomenological relaxation laws have been matched with both the static and cyclic relaxation behavior. Work on more mechanistic models is continuing with the goal of extracting activation energies and volumes. This would allow us to tie microstructure to performance.
AFM and TEM analysis are crucial for more accurate model development. The basic test sample design has been altered to facilitate microscopic characterization. The basic geometry of the microbeams is unchanged, but the surrounding sample support is enlarged for easier handling. The new design also allows separation of dies without wafer saw cutting, previously a source of significant surface contamination and sample failure. Initial tests of the new design are encouraging; full fabrication and testing will be completed in Q4.
The second-generation fatigue test tool design is complete, and many of the components have been fabricated. A strain gage has been integrated into the piezoelectric actuator, and an LVDT for strain measurement has been added directly to the sample grips. This should reduce error in strain measurement, one of the weaknesses of the previous design. We will complete assembly once the piezoelectric actuator components arrive.
B. In-situ Stress Measurement-- Sample IR detectors have been procured from Prof. Kenny. External feedback circuitry has been built on a proto-board. Electrical testing (isolation test and membrane deflection test) have been conducted on 302 sensors to identify the ones capable of tunneling. Successful tunneling has been observed under reasonable deflection voltages (30-60V) for only 28 sensors. Yield was low due to misalignment during dicing at IC Sensors and to certain design issues. Of the 28 IR sensors, 6 are available for testing as stress sensors because they have ruptured membranes (necessary for pressure sensing but not for stress measurement). Several of the working devices are unstable during tunnelling, another issue to be resolved.
In preparation for developing first-generation dedicated stress sensors, the 7-mask set and process flow for the current IR sensors have been re-examined. Key issues are the design of membrane electrode and membrane corrugation rings, as well as certain critical steps in the process flow.
Difficulties/Problems:
References:
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:
Progress:
In our last report we considered the effects of the plasma etch release process. In particular we were concerned with the effects of excessive ion bombardment and heating of the structures during release. To minimize and control these effects we used a ground plane shield above the samples during the release process to eliminate the effects of ion bombardment. In addition, we also reduced the RF power used during the release, to minimize the impact of localized heating on the samples. The results of these experiments all showed similar degrees of curvature for the cantilever beams, which indicates that the primary cause of deformation is not dependent on the release process parameters that we used, but is primarily dependent on the "as deposited" material properties of the film and substrate itself. As the film grows the stress state of the film varies and gives rise to locked in stresses that creates a stress gradient within the film. We now believe that the out of plane deformations of the cantilever beams is due primarily to this stress gradient, and thus we will now be pursuing deposition methods that will reduce this gradient (i.e. Ion Beam Assisted Deposition, or through the deposition of intentionally stressed multilayered films).
We fabricated and are now testing structures built from multilayers of Al/Si and silicon nitride over Ti/W. Preliminary results show that the Ti/W silicon nitride films have large uncompensated stress gradients. Further work is in progress to test these structures with just the Ti/W and silicon nitride films alone. From the measured stresses we will fabricate a triple layer of Ti/W and silicon nitride to build structures which co not curl up by cancelling the stress in the first two layers. In other work we have also deposited Iridium on a Testarosa wafer for eventual dicing and release to characterize both the film's stress state and the step coverage.
Difficulties/Problems:
Although we are still pursuing stress gradient free films, it is apparent that stress gradients will be difficult to completely control and as a result extraction of the mechanical properties from simple beam mechanics becomes problematic. Complete elimination of stress gradients may not be entirely necessary to determine the mechanical properties.
The Dutton group is investigating a method that will measure the curvature of a singly constrained beam and apply this data to electrostatic pull in tests of doubly constrained beams to determine the mechanical properties of the thin films. We are currently in the process of characterizing the accuracy and applicability of this method, which will allow us to determine stress gradient, in-plane stress, Young's Modulus and boundary conditions of the mechanical structures.
The stress gradient is not an exclusive property of the film's composition, but is also a property of the deposition process. We will now be focusing on the control of the deposition parameters that will reduce or control the gradient. This work will involve depositing the Ir and W films under a variety of ion bombardmant conditions to find the parameters that will give the lowest gradients.
Goals targeted for the next quarter: