SBIR/STTR Award attributes
Ultrafast lasers play an increasingly critical role in the generation, manipulation, and acceleration of electron beams for applications ranging from basic research to medicine to industry. Plasma-based accelerators in particular stand to benefit from improved precision and stability of high power lasers, which are poised to become the drivers for a novel class of high quality electron beam sources. Significant advances in average power, precision, and stability of facility-scale laser systems are required to realize these transformative technologies. A novel class of diagnostics and associated correction schemes will be developed to enable the stabilization of laser performance at kHz repetition rates. The proposed system will combine three elements: a precise and non-perturbative laser wavefront diagnostic, a fast and flexible machine-learning-based computation of the laser focal position, and an in-hardware deployment of the corresponding optical correction. The resulting correction scheme will be integrated with standard software suites to enable the control of focal position at kHz repetition rates in conjunction with experimental operation. These tools will be designed in collaboration with a world-class high intensity laser facility. The proposed work will demonstrate the synthesis of a novel diagnostic and correction scheme for precise control of laser focal position at repetition rates greater than 10 Hz. First, a non-perturbative laser wavefront diagnostic will be tested and refined to the required tolerance. Next, a family of adaptive algorithms will be designed and trained to provide efficient correction based upon these measurements. The resulting algorithms will be deployed on a proof-of-principle field programmable gate array hardware implementation, and the deployment scheme will be validated for use in experimental facilities. Finally, an experimental demonstration will be planned for Phase II. The proposed research will drive innovation in three distinct technologies. Laser technology is ubiquitous in modern industry and research; the proposed work will improve diagnostic quality and enable high repetition rate operation for emerging high power laser applications. Field programmable gate arrays are the industry standard for delivering real-time active feedback; the proposed work will develop a deployment pipeline for integrating this technology with advanced algorithms for controls, to the benefit of a diverse array of laser and accelerator applications. Machine learning promises to greatly improve complex machine performance, maximize operational duty cycle, and reduce costs associated with operator training and supervision; the proposed work will demonstrate an efficient methodology for implementing machine learning algorithms within a broadly applicable controls framework.
