Abstract
Real thermionic device performance asymptotically approaches ideal behavior in the saturated and Boltzmann regimes at high and low voltages. The maximum power point occurs between the two ideal regimes in a transition region where the output current density is a fraction of the ideal output current due to space charge in the interelectrode gap. The operating mode of a power-generating thermionic device is determined by the operating conditions of the device and the device specifications. These conditions and specifications include the electrode temperatures, operating voltage, device geometry, and the electrode material properties.
The focus of this paper is to develop a physics-based model for the transition region between the two regimes and to incorporate data from real devices. We based the modeling on classical models of current transport between electrodes. The saturated regime is characterized by the Child-Langmuir law for thermionic emission and Langmuir space-charge theory is implemented for the charge-affected regime to characterize the transition behavior in real diodes. We modify the conventional solution for the interelectrode motive to include an electron reflection probability parameter at the collector and establish a general functional form for the current flux variation with voltage in the transition region. We then utilize machine learning tools to interpret the behavior of real data in the context of the developed framework, and to develop a performance predictive tool that leverages the advantages of a physics-inspired machine learning methodology.
The model is demonstrated to fit performance data for four different real, small gap, vacuum thermionic devices to within 12%. We used the framework and the functional form suggested in tandem with experimental data to develop a performance predictive model for the power-generating operating regimes which reflects the underlying physics and is consistent with performance trends in the data for real devices. The sensitivity of the device performance to changes in work function, electron reflection probability, and interelectrode gap width is also discussed.
