果冻传媒

To meet the growing demands of modern industry, high-precision mechatronic systems like wafer scanners, used in the production of integrated circuits, are essential. Current advanced wafer stages use a dual-stroke mechanism with a magnetically levitated moving-coil motor for coarse positioning and a short-stroke motor for fine adjustments. However, the heat load and cable requirements limit the throughput. For his PhD research Yorick Broens explored a new approach, using a moving-magnet planar motor. This design offers better isolation from environmental factors and a lighter moving mass, enhancing throughput but increasing motion control complexity. To address challenges in nanometer accuracy positioning he developed new control methods. These methods have shown significant enhancements in lithographic scanning performance. 

To guarantee high throughput and high reliability for mechatronic system systems, the silicon wafer is positioned under the projection optics with sub-nanometer accuracy using the wafer stage module. The currently employed state-of-the-art wafer stage configuration is comprised of a dual-stroke mechanism, combining a magnetically levitated moving-coil motor for coarse positioning of the wafer with a short-stroke motor for precise positioning. However, limitations in the moving-coil design, such as heat load and the need for cable connections between the moving-body and the environment for powering the coils, impact its maximum throughput.

Advanced motion control methods

Consequently, recent research explores planar motors using a moving-magnet setup with a single stroke only, aiming to replace the current dual-stroke configuration. The moving-magnet configuration offers advantages over the moving-coil configuration, such as complete environmental decoupling of the mover and the potential for a smaller, lighter moving body, allowing for increased throughput. Despite the benefits of the moving-magnet configuration, the use of this layout introduces increased complexity in motion control. These motor configurations display highly complex nonlinear multi-physical effects that can only be approximately modeled based on first-principles knowledge. The combination of electromagnetic interactions, low-frequency flexible dynamics, and position-dependent effects present a significant challenge in terms of achieving desirable performance using the currently employed state-of-the-art motion control design approaches. To overcome these challenges, Yorick Broens developed a novel toolset of advanced motion control methods, allowing for nanometer accurate positioning of moving-magnet planar actuator systems.

Improved commutation approach

To address the complex nonlinear position-dependent electromagnetic interactions, an improved commutation approach for moving-magnet planar actuators was introduced, aiming to enhance the existing model-based algorithm by improving its representation of the underlying physics. The proposed improvements include: a novel static electromagnetic alignment approach for precise magnet plate alignment with respect to the coil arrays, a learning-based feedforward approach to account for position-dependent variations in electromagnetic interactions, directly compensating in the model-based commutation for improved system performance, and a dynamic regulatory method to compensate for remnant effects and model-based commutation errors. Performance of the proposed methods is experimentally demonstrated on a moving-magnet planar actuator prototype, showing an improvement of approximately 65% in MA position tracking performance.

Frequency domain auto-tuning approach

Besides that, a novel frequency domain autotuning approach for structured linear-parameter-varying (LPV) multiple-input multiple-output (MIMO) feedback controllers was proposed, addressing position-dependent effects without relying on accurate parametric models of the system. The contributions include: a stability check using frequency response functions that is capable of handling diagonal and full block controller structures, a modular LPV MIMO structured feedback control parameterization for autotuning, and a real-time implementation approach preserving the physical interpretation of the controller in terms of maintaining the continuous-time controller parameterization for the corresponding discrete-time controller.

Extensions to the control framework

And finally, to address low-frequency flexible dynamics, the researcher proposes two extensions of the state-of-the-art rigid-body control framework, allowing for active deformation control of the magnet-plate. The extension includes: an output-based modal observer extension of the state-of-the-art control configuration, allowing for active control of flexible dynamics by utilizing the available over-actuation in the system, an error-based modal observer extension of the state-of-the-art control configuration, allowing for computational efficient algorithms for active vibration control, and augmentation of the proposed output-based modal observer with the active regulation of the commutation frame, allowing for extreme positioning accuracy.

All proposed developments have been validated on an experimental moving-magnet planar actuator prototype, resulting in a factor 38 improvement in the x,y scanning performance and a factor 36 improvement in z scanning performance, compared to current state-of-the-art control approaches.

Title of PhD thesis: . Supervisors: Prof. Roland Toth, and Prof. Hans Butler.