Data processing: generic control systems or specific application – Specific application – apparatus or process – Product assembly or manufacturing
Reexamination Certificate
2000-04-19
2003-05-20
Von Buhr, Maria N. (Department: 2125)
Data processing: generic control systems or specific application
Specific application, apparatus or process
Product assembly or manufacturing
C700S121000, C702S084000, C257S048000
Reexamination Certificate
active
06567715
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to post-process analysis and characterization of micromechanical devices and assemblies, and in particular to automated means for certification of MEMS fabrication processes at the substrate level.
BACKGROUND OF THE INVENTION
Microelectromechanical systems and their close relatives (which shall be known collectively as MEMS) are becoming attractive for a wide variety of applications, owing to their high functionality, small size, and potentially high reliability and low cost. However, despite the development of a large number of demonstration devices and a small number of commercial devices, design and processing of MEMS-based apparatus remains largely a trial and error process.
Variations in the material properties or gross structure of the MEMS devices can destroy the devices, or simply result in greatly reduced operational lifetimes. For example, two comb-driven micromotors which are apparently identical in structure and operational characteristics when new can show many orders of magnitude difference in the number of rotational cycles required to cause failure. Without sufficient control on the factors leading to such difficulties, widespread commercial application of MEMS technology is not to be expected.
A simplified example of a MEMS fabrication process would involve the deposition, patterning, and planarization of several material layers. The layers comprise such materials as polysilicon, silicon oxides, and silicon nitride. As these materials have different chemical properties, it is possible to use patterned composite structures to form the desired parts from (usually) polysilicon, and then to release the parts from the surrounding materials, primarily using chemical etching. Typical MEMS components have layer thickness of a few microns, and lateral dimensions from perhaps 10 to 1000 microns.
It is easy to list a variety of factors which can alter the functionality of a MEMS device after release. Perhaps the most obvious is the residual stress in the polysilicon components. This residual stress varies from point to point in the processed polysilicon, resulting in residual stress gradients which cause such components to warp in a manner determined by their geometrical structure and the magnitude and direction of the residual stress gradients. Sometimes warping can either be contained or compensated for by the design of the individual component. In other cases, for example meshing gears, where the thickness to diameter ratio can be as small as 1 to 1000, warping of any substantial magnitude cannot be allowed. Even when design can reduce problems associated with warping, a processing glitch which changes the residual stress from the design value can result in fabrication of components with reduced operational lifetimes. Similar problems can appear if the residual stress is not constant over the substrate, if the layer thicknesses vary from the design values, if the layer thicknesses are not consistent across the substrate, or if the surrounding materials are not entirely removed in the release process.
All of the above process errors interfere with the operation of the MEMS device through changing the expected degree of component warping, and do so while the component dimensions remain inside tolerances. Clearly, changes in component geometry will also cause warping to change. In principle such changes can also result from unexpected changes in the elastic properties of the polysilicon from which they are formed. Such changes can occur through a faulty polysilicon deposition (e.g. if voids or impurities are produced), if the surrounding materials are not completely removed in the release process, or (effectively) if geometric flaws in the components produce stress concentrations.
Although component warping is a serious problem, it is not the only route through which process variations can produce faulty MEMS devices. If a gear revolves around a simple hub, the rate of wear will depend on many factors, among which are the levels of friction and stiction between the surfaces, and the clearances, vertical and in shaft-aperture diameter difference, between the moving gear and the fixed hub. Process problems which can alter the expected levels of friction and stiction include producing rough or smooth interaction surfaces on release, not entirely removing the surrounding materials on release, growth of atypical material at the polysilicon/supporting material interfaces, and various types of unwanted surface contamination. The geometric clearances between the hub and the gear can be altered dramatically by small changes in layer thicknesses, by unexpected warping (e.g., the gear may warp so that it is forced up against the top of the hub), and by the presence of surrounding materials which are not entirely removed during the release process.
The above has only begun to touch on the enormous variety of MEMS processing problems that can lead to reduced lifetime of the resulting devices, or even to complete non-functionality. However, it is sufficient to suggest that a major barrier against routine commercial production of complex MEMS devices exists. This problem is made more difficult owing to the difficulty of detecting such process failures, especially in the absence of immediate structural failure when the device is released. An inappropriately warped part can be hidden in an internal portion of the device, where the degree of warping cannot be measured. If it is visible, the detection of an extra half-micron of component warp on a released component which is free to move and vibrate is not an easy challenge. Detection of residual material left after release is very difficult, as the changes in dimensions can be very small. Similarly, surface contamination is difficult to detect, particularly on a substrate with hundreds of highly complex multi-layer devices.
As the MEMS devices being fabricated cannot easily be measured or tested to reveal flaws resulting from processing defects, there is a need for standard test structures which will reveal clearly when such procession defects appear. In addition, in the design process there is a need for the precise measurement of mechanical properties, including the levels of friction and stiction expected in various regimes. This can also be addressed through the use of standard test structures.
Some work has appeared along these lines. However, the results of such measurements, even when carried out on a single substrate, are inconsistent. A group at MIT has examined the use of pull-in cantilever and fixed-fixed beam test structures to evaluate the quality of MEMS devices which are co-fabricated. They used the well-known transition which occurs as an attractive force is applied electrostatically between a supported member and a surface, between gradual bending of the member toward the surface, and abrupt collapse of the member onto the surface. The diagnostic quantity is the pull-in voltage (the voltage at which abrupt collapse occurs), the collapse being detected by the electrical contact made between the member and the surface.
A detailed mechanical model of the apparatus and the electrostatic forces associated not only with deflection of the cantilever, but of the highly nonlinear effects which enter in as the distance between the cantilever and the surface varies (that distance itself being a function of position along the cantilever) must then be used to extract, e.g., elastic properties from the pull-in voltage. However, the same process failures with which a commercial fabricator has to be concerned will act to change the interpretation of that measurement. Residual stress gradients and anchors which do not hold the beam parallel to the wafer surface will change the level of stress needed to get to the pull-in condition, and hence will alter the pull-in voltage with no other signature. Changes in component geometry or material from the expected can seriously alter the results of a pull-in measurement, again with telltale signatures. Finally, pull-in measurements gi
DeBoer Maarten P.
Jensen Brian D.
Miller Samuel L.
Sinclair Michael B.
Smith Norman F.
Dodson Brian W.
Sandia Corporation
Von Buhr Maria N.
LandOfFree
Method and system for automated on-chip material and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and system for automated on-chip material and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system for automated on-chip material and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3021665