Model-order reduction (MOR) is a numerical approach that has been successfully applied across many disciplines to obtain computationally efficient reduced-order models (ROMs). Despite these successes and the potential to avoid physical assumptions and approximations, MOR has not yet been explored for the development of compact photocurrent models, where traditional analytic techniques dominate. In this article, we demonstrate the viability of MOR in this context by using proper orthogonal decomposition to develop a hybrid analytic-numerical compact model, which replaces analytic solutions of the ambipolar diffusion equation with numerical ROM solutions. We demonstrate our model by comparing its predictions with those of a state-of-the-art analytic model using photocurrent measurements obtained at the Little Mountain Test Facility.

Hetero-Junction Bipolar Transistors (HBT) have several advantages over Silicon Bipolar Junction Transistors (BJT) in radiation environments. One advantage is an intrinsic hardness to displacement damage causing radiation. The generally smaller size of HBTs compared to BJTs also means that less photocurrent is generated by these devices. A disadvantage of the smaller size is less ability to dissipate heat due to smaller surface areas and contacts. This report describes simulations intended to study the initial heating of HBT transistors due to ionizing radiation events and the subsequent heating caused by feedback in the devices when responding to these events.

The effect of a linear accelerator's (LINAC's) microstructure (i.e., train of narrow pulses) on devices and the associated transient photocurrent models are investigated. The data indicate that the photocurrent response of Si-based RF bipolar junction transistors and RF p-i-n diodes is considerably higher when taking into account the microstructure effects. Similarly, the response of diamond, SiO 2 , and GaAs photoconductive detectors (standard radiation diagnostics) is higher when taking into account the microstructure. This has obvious hardness assurance implications when assessing the transient response of devices because the measured photocurrent and dose rate levels could be underestimated if microstructure effects are not captured. Indeed, the rate the energy is deposited in a material during the microstructure peaks is much higher than the filtered rate which is traditionally measured. In addition, photocurrent models developed with filtered LINAC data may be inherently inaccurate if a device is able to respond to the microstructure.

Photocurrent generated by ionizing radiation represents a threat to microelectronics in radiation environments. Circuit simulation tools that employ compact models for individual electrical components (SPICE, e.g.) are often used to analyze these threats. Historically, many photocurrent compact models have suffered from accuracy issues due to the use of empirical assumptions, or physical approximations with limited validity. In this paper, an analytic model is developed for epitaxial diode structures that have a heavily-doped sub-collector. The analytic model is compared with both numerical TCAD calculations and the compact model described in reference [1]. The new analytic model compares well against TCAD over a wide range of operating conditions, and is shown to be superior to the older compact model [1]. The methods put forth in this paper could also be applied to model devices with similar physics, such as photonic and power devices.