IC-Processed Polysilicon Micromechanics: Technology, Material, and Devices

Abstract: 
Movable pin joints, gears, springs, cranks, and slider structures with dimensions measured in micrometers have been fabricated using silicon microfabrication technology. These micromechanical structures, which have important transducer applications, are batch-fabricated in an IC-compatible process. The movable mechanical elements are built on layers that are later removed so that they are freed for translation and rotation. A new undercut-and-refill technique that makes use of high surface mobility of silicon atoms undergoing chemical vapor deposition is used to refill undercut regions in order to form restraining flanges. Typical element size and masses are measured in millionths of a meter and billionths of a gram. The process provides the tiny structures in an assembled form, avoiding the nearly impossible challenge of handling such small elements individually.
The lateral thermal conductivity of heavily doped low pressure chemical vapor deposited (LPCVD) polycrystalline silicon films is measured using polycrystalline silicon microbridges elevated three micrometers above a silicon substrate. The bridges, lightly doped in their central regions and heavily doped elsewhere, are fabricated using a sacrificial silicon-dioxide layer and phosphorus out-diffusion from doped oxide. Voltage-current characteristics measured on the bridges both under high vacuum and in silicone oil are used to calculate lateral thermal conductivity in the polycrystalline silicon. The experimental values for the thermal conductivity of heavily doped polycrystalline silicon range from 0.29 to 0.34 Wcm^-1K^-1 and average 0.32 Wcm^-1K^-1. These values agree closely with results obtained by a second method that employs uniformly doped polycrystalline silicon bridges. In the second method, high-vacuum, voltage-current characteristics are measured and the indicated thermal conductivities for two samples are 0.29 and 0.30 Wcm^-1K^-1, respectively.
A new polysilicon bridge-slider structure, in which one end of the bridge is fixed and the other is connected to a plate sliding in two flanged guideways, is designed and fabricated to study the strain at fracture of LPCVD polysilicon. In the experiments, a mechanical probe is used to push against the plate end, compressing and forcing the bridge to buckle until it breaks. The distance that the plate needs to be pushed to break the bridge is recorded. Nonlinear beam theory is then used to interpret the results of these axially loaded bridge experiments. The measured average fracture strains of as-deposited LPCVD polysilicon is 1.72%. High-temperature annealing of the bridge-sliders at 1000degC for one hour decreases the average fracture strain to 0.93%. These values substantially exceed reported fracture-strain values for large-volume samples of either mono or polycrystalline silicon. It is found that high-temperature annealing reduces both the magnitude and the distribution of the fracture strain in polysilicon.
A point-force, load-deflection method using a stylus-type surface profiler to determine the Young's moduli (Ey) of thin-film microstructural materials is introduced. In this method both force and deflection in microstructures are measured simultaneously by the profiler to provide a convenient and accurate means to obtain Young's modulus directly. Measurements on two types of micromechanical structures are described: a doubly supported bridge and a bridge-slider (a beam with one fixed end and one end which can slide in a flanged housing). The influence of residual stress in the doubly supported beam is described and accounted for theoretically to interpret measurements made on low-stress silicon-nitride films. For this material Ey is 373 GPa. In the bridge-slider structure, residual strain is relaxed to zero. Measurements on a polycrystalline-silicon bridge slider show a value for Ey of 123 GPa in unannealed material that is doped heavily with phosphorous, and grown at 656degC.
The polysilicon bridge-slider structure is also used to study static friction in microstructures. In the experiments, a mechanical probe pushes against the end of the slider, compressing the bridge, and making it buckle until the sliding plate becomes stuck in the guideways. The distance that the plate needs to be pushed to make the plate stick is recorded. Nonlinear beam theory is then used to calculate the static-friction coefficient from the position at which sticking occurs. It is found that the average friction coefficient between two layers of unannealed polysilicon is 1.40; between two layers of annealed polysilicon it falls to 1.21. For either type of sample, there is a minimum static-friction coefficient of roughly 0.8 and a maximum coefficient that is nearly 2.0.
Integrated stylus-force microgauges made of low-residual-stress silicon nitride using surface-micromachining technology have been designed, built, and tested. These gauges employ polysilicon piezoresistors as sensing elements and are calibrated using standard weights made from probe tips. The predictions of beam theory taking account of both residual and applied stress in the beam support are in good agreement with experimental results. The gauges cover a stylus-force range from 0.1 to 50 mg with an accuracy of roughly +/-1 mg.
Micromotors having rotors with diameters of 120 um have been fabricated and driven electrostatically to continuous rotation. These motors were built using processes derived from IC microcircuit fabrication techniques. Initial tests on the motors show that friction plays a dominant role in their dynamic behavior. Observed rotational speeds have thus far been limited to several hundred rpm which is a small fraction of what should be achievable if only natural frequency were to limit the response. Experimental starting voltages (60 volts at minimurn and above 100 V for some structures) are at least an order-of-magnitude larger than had been expected. Continuous motor motion has been observed for as long as one minute under three-phase bias at 200 V. Observations of reverse as well as forward rotor rotation with respect to the driving fields can be explained in terms of the torque/rotor-angle characteristics and friction for the motors.
A frictional-torque model consisting of two components, one a constant term and the second a position-dependent term, is used to analyze the motion of a variable-capacitance IC-processed micromotor. Values for the position-dependent frictional torque are calculated from a two-dimensional electrostatic simulator using a non-concentric rotorhub model. The constant toque, calculated from the experimental starting voltage, is much higher than can be explained by static friction and is possibly due to the coulombic attraction between charge sheets. Predictions of the theory are in agreement with experiment. Dynamic frictional coefficients for polysilicon surfaces adjacent to silicon-nitride surfaces are calculated to range from 0.21 to 0.38 in the micromotors
Publication date: 
August 31, 1989
Publication type: 
Ph.D. Dissertation
Citation: 
Tai, Y. (1989). IC-processed Polysilicon Micromechanics: Technology, Material, and Devices. United States: University of California, Berkeley.

*Only registered BSAC Industrial Members may view project materials & publications. Click here to request member-only access.