How modeling engineers could reconcile high costs with the pursuit of model accuracy.
As power electronics shrink in size, the demands on power, frequency, and efficiency grow exponentially. The semiconductor industry is leaning heavily into wide bandgap materials like gallium nitride (GaN) and silicon carbide (SiC) to help meet these demands. Recent research projects that the global GaN semiconductor devices market will grow at a CAGR of 25.4% from 2023 to 2030. However, the challenges remain when modeling and simulating their unique characteristics.
This article will explore the growing impact of GaN, emerging challenges in device modeling, and how modeling engineers could reconcile high costs with the pursuit of model accuracy.
Wide bandgap semiconductors (WBGs) are making waves.
As the name suggests, they have a larger bandgap – the energy difference between insulating and conducting states – setting them apart from traditional materials like silicon. What does it mean for power supplies?
Fig. 1: Comparison of GaN vs. traditional silicon semiconductors.
Thermal conductivity represents a material’s ability to transfer heat. Lower thermal conductivity implies material inefficiencies at high-power applications because the more heat a material retains, the more its electrical characteristics change.
With a thermal conductivity of 2 W/cmK for GaN, they can efficiently operate up to 200°C, surpassing silicon’s 150°C.
The electric breakdown field indicates the material’s dielectric strength after it takes in enough energy to breach the bandgap. With a breakdown field is 3.3 MV/cm, GaN outperforms silicon’s 0.3 MV/cm. The higher breakdown field makes GaN almost 10 times more capable of handling higher voltages.
Electron mobility describes how quickly electrons can move through a material. GaN boasts electron mobility of 2,000 cm2/Vs, allowing for switching frequencies up to 10 times higher than silicon, making it suitable for high-frequency applications.
As a result, GaN emerges as a frontrunner for high-frequency, low-power, and high-temperature applications — all while optimizing energy consumption.
For the successful integration of new technology into our established systems and designs, rigorous models are indispensable. And GaN, with its high production costs, makes this imperative even more pronounced.
To maximize the efficiency in GaN device modeling, there are seven pivotal factors:
Electrical trapping effect: When modeling GaN devices, the electrical trapping effect is often a major concern. Whether originating from surface states or bulk defects, it can significantly reduce power efficiency and lead to current collapse.
Thermal behavior: GaN devices can operate at high power densities, leading to localized heating effects. While GaN has good thermal stability, localized hotspots can compromise both the performance and the longevity of the devices.
Material inhomogeneity: The quality of GaN epilayers and heterostructures can vary, leading to material inhomogeneities. Incorporating these variances into models is pivotal to reflecting real-world performance accurately.
High electric fields: While GaN devices can withstand high electric fields, this also means that phenomena like impact ionization can become significant, especially near device breakdown. Modeling these high-field effects is crucial for designing reliable devices.
Interface quality: Especially within GaN-based heterostructures or metal contacts, interface quality can drastically change device performance. Models must accurately represent potential challenges, such as trap states or alloy disorders at these interfaces.
Quantum effects: Due to the high electron mobility and narrow channels (in the case of High Electron Mobility Transistors, HEMTs), traditional drift-diffusion models may not suffice. This underscores the growing demand for advanced modeling methods that consider quantum corrections or even full quantum transport.
Lack of standardized parameters: Perhaps one of the more overarching challenges is the lack of a standardized parameter set for GaN compared to its well-established silicon counterpart. For device modeling engineers, this means navigating uncharted waters and emphasizing the need for a highly flexible modeling environment.
I’ll explain the strategies using an analogy many of us are familiar with – Star Wars.
Fig. 2: Stay on target with design and modeling goals.
This is a snapshot Mark Knutson, one of our modeling experts, recently took about a jogging trail near his house, marked with symbols reminiscent of TIE fighters.
Think of each of these symbols as individual milestones in a device modeling project. Ideally, these markers should form a straight line. However, most project milestones, much like the painted symbols, often deviate from the intended path. When the path is curved, it throws more challenges or more physics at the modeling engineer to navigate. But the prime directive should remain clear: stay on target.
A novice in the field might be tempted to think modeling is as straightforward as feeding 50 parameters into an optimizer to get the desired results. However, we’re witnessing an influx of new models, inevitably accompanied by a more extensive repertoire of parameters. Especially when we dive deeper into the realm of wide bandgap semiconductors and grapple with intricate nuances like trapping effects, such a simplistic approach is inadequate.
Thus, it becomes essential for device modeling engineers to embrace a more expansive viewpoint to stay on target. This means they must be diligent in understanding the underlying causes, finding ways to mitigate them, and determining the best-fitted models for their objectives.
In essence, while the universe of device modeling expands, the real challenge is not merely to sail through but to do so with a keen eye on the destination. This calls for necessitates an environment that is both flexible and adaptive to the ever-evolving requirements.
A flexible modeling environment is paramount in wide bandgap semiconductor device modeling, particularly given the rapid evolution of GaN materials and their unique characteristics. Recognizing the complexities and potential of GaN materials, we’ve meticulously refined our flagship device modeling platform, IC-CAP, to maximize the flexibility and versatility of device modeling environments.
As we move beyond traditional silicon to materials like GaN, device modeling engineers need a tool that can easily adapt to these new device models.
IC-CAP is ahead of the curve by offering complete coverage for GaN devices, from the traditional, empirical-based, Angelov-GaN model to the most recent physics-based models like the ASM-HEMT and MVSG models. For those keen on precision, IC-CAP offers the Artificial Neural Network (ANN)-based DynaFET model.
Fig. 3: IC-CAP’s GaN CMC ASM-HEMT model extraction example.
With the IC-CAP parameter extraction language (PEL) or the Python programming environment, modeling engineers can directly customize their own IPs, tailor models, fine-tune simulations, and orchestrate workflows to specific technology specs. Often it is as simple as going to another browser to get the examples immediately, making IC-CAP an open and flexible working environment.
The IC-CAP Wafer Professional (WaferPro) module enables automated on-wafer measurement and characterization. With a specialized test plan workspace inside the IC-CAP platform, engineers can efficiently measure post-process data via a wide range of Keysight expertise and instruments to enhance model accuracy.
The future of power comes from wide bandgap semiconductors (WBGs).
To wrap it up with our Star Wars analogy, just as Luke Skywalker precision-dropped the photon into the Death Star, Keysight’s mission with IC-CAP is to equip device modeling engineers with a flexible modeling sandbox, ensuring they consistently “stay on target” while navigating the vast and evolving expanse of the wide bandgap semiconductor landscape.
 |
 |
 |
 |
 |
 |
 |
 |
 |  |
Leave a Reply