Dynamics and Control of Flexure-based Large Range Nanopositioning Systems.

Abstract

The objective of this thesis is to demonstrate desktop-size and cost-effective flexure-based multi-axis nanopositioning capability over a motion range of several millimeters per axis. Increasing the motion range will overcome one of the main drawbacks of existing nanopositioning systems, thereby significantly improving the coverage area in nanometrology and nanomanufacturing applications.

A single-axis nanopositioning system, comprising a symmetric double parallelogram flexure bearing and a traditional-architecture moving magnet actuator, is designed, fabricated, and tested. A figure of merit for the actuator is derived and shown to directly impact the system-level trade-offs in terms of range, resolution, bandwidth, and temperature rise. While linear feedback controllers provide good positioning performance for point-to-point commands, the tracking error for dynamic commands prove to be inadequate due to the nonlinearities in the actuator and its driver. To overcome this, an iterative learning controller is implemented in conjunction with linear feedback to reduce the periodic component of the tracking error by more than two orders of magnitude. Experimental results demonstrate 10 nm RMS tracking error over 8 mm motion range in response to a 2 Hz bandlimited triangular command.

For the XY nanopositioning system, a lumped-parameter model of an existing XY flexure bearing is developed in order to understand the unexplained variation observed in the transfer function zeros over the operating range of motion. It is shown that the kinematic coupling, due to geometric nonlinearities in the beam mechanics, and small dimensional asymmetry, due to manufacturing tolerances, may conspire to produce complex-conjugate nonminimum phase zeros at certain operating points in the system's workspace. This phenomenon significantly restricts the overall performance of the feedback control system. After intentional use of large asymmetry is employed to overcome this problem, independent feedback and iterative learning controllers are implemented along each axis. Experimental results demonstrate 20 nm RMS radial tracking error while traversing a 2 mm diameter circle at 2 Hz.