Electronics drive nearly every aspect of modern technology, from everyday smartphones to deep-space missions, autonomous vehicles, and AI chat tools. However, when exposed to extreme environments such as radiation, extreme temperatures, and high pressures, these systems can degrade, fail, or behave unpredictably, posing serious challenges for their reliability and longevity.
Our research focuses on making microelectronics more reliable in extreme environments. We combine fundamental physics, advanced testing methods, and system-level analysis to understand and prevent failures before they happen. By developing predictive models and radiation-hardening strategies, we ensure that critical electronic systems keep working when failure is not an option.
Research Thrusts
Thrust 1: High-Throughput Laser-Based Testing for Radiation Effects
The Challenge:
Particle accelerators are the gold standard for radiation testing, but there are only 5 heavy ion accelerators in the US. This leads to long wait times and high costs. However, we can use focused laser pulses to emulate the effects of space radiation on electronic systems.
Why It’s Hard:
Lasers deposit charge differently than charged particles. For decades, the community debated: Can we trust laser data? When do lasers accurately predict particle effects? Getting quantitative correlation required understanding wavelength-dependent absorption, charge distribution profiles, and device-specific responses.
Current Questions:
- How do we make laser testing predictive without particle validation?
- What are the fundamental limits of inducing the same response using these alternative techniques?
- Can we extend our findings to emerging sources (pulsed electrons, X-rays)?
Why It Matters:
In addition to speeding up microelectronics qualification for use in space systems, high-throughput laser testing can generate the massive datasets needed for machine learning discovery. This is the foundation enabling our next two research thrusts.
Thrust 2: Data-Driven Discovery of Radiation-Hardened Design Principles
The Challenge:
Traditionally, engineers design radiation-hardened chips through trial-and-error: build a layout, test it, watch it fail, figure out what went wrong, try again. This process can take years. We need to move from “test and hope it works” to “design with confidence it will work.”This requires a better understanding of which physical features actually matter.
Why It’s Hard:
Radiation failures depend on dozens of interacting factors: transistor spacing, doping profiles, well configurations, guard ring placement, substrate contacts, power distribution. Traditional physics modeling can simulate these, and some researchers have worked on this. However, exploring the full design space is computationally intractable. And without massive datasets, machine learning can’t discover patterns.
What We’re Building:
- Explainable AI frameworks that extract interpretable design rules from experimental data
- Physics-informed ML models that discover which features govern radiation tolerance
- First ML-based mitigation of radiation upsets in RF systems (IEEE Meritorious Paper Award, 2021)
Current Questions:
- Can we discover design rules for radiation failures from data?
- How do we ensure ML-discovered principles are physically meaningful?
- Do principles discovered on simple structures transfer to complex integrated systems?
The Vision:
Instead of testing several layout variants, designers input their layout and get predicted radiation vulnerability with recommended fixes. Some available tools already do this in a limited capacity. However, we’re building more generalized methods that work across fabrication processes and for multiple types of failure.
Thrust 3: Multi-Scale Predictive Frameworks for Radiation Effects in Complex Systems
The Challenge:
Modern electronics are complex! 3D stacks, heterogeneous chiplets, co-packaged photonics… their complexity means we are reaching a point where we cannot test comprehensively. For example, accelerators can produce particles to reach the components in the middle of a stack. However, we still need to answer: Will this system survive in orbit?
Why It’s Hard:
Radiation effects span six orders of magnitude in scale and time:
- Atomic: Charge generated in femtoseconds, nanometer volumes
- Device: Charge transport across microns in picoseconds
- Circuit: Voltage transients propagating in nanoseconds
- System: Bit flips, crashes manifesting in milliseconds
Understanding one scale doesn’t predict the others. Simulating everything at the atomic level is impossible. We need frameworks that connect scales and predict the reliability of untestable systems.
What We’re Building:
- Multi-scale models connecting atomic radiation interactions to system failures
- Frameworks that combine TCAD device physics + circuit simulation + data-driven system prediction
- Methods to validate predictions when we can’t test the final system
Current Questions:
- How do we bound uncertainty when predicting systems we’ve never tested?
- Can principles from testable 2D structures predict behavior in untestable 3D stacks?
- How do different technologies (Si, SiGe, GaN, photonics) interact under radiation in heterogeneous systems?
The Vision:
With predictive models that consider multiple failure mechanisms, we would be able to describe the radiation behavior of complex 3D systems before building them. But this requires solving fundamental challenges: connecting physics across six orders of magnitude in scale, validating predictions without test data, and quantifying uncertainty when we can’t measure ground truth.
Application Domains
Our research addresses critical radiation effects challenges:
🛰️ Space Systems
Radiation-hardened electronics and photonics for satellites, planetary rovers, deep-space probes, and space-based infrastructure operating in harsh radiation environments
🖥️ Terrestrial Computing
Radiation tolerance for data centers, AI accelerators, and high-performance computing, where cosmic rays and terrestrial neutrons cause soft errors
✈️ Avionics & Critical Systems
Radiation-hardened electronics for aerospace and mission-critical applications where reliability is essential
🔮 Emerging Technologies
Extending radiation effects prediction to novel material systems, photonic integrated circuits, heterogeneous chiplets, and 3D integration for radiation environments
Publications
Recent highlights:
A. Ildefonso et al., “Using Machine Learning to Mitigate Single-Event Upsets in RF Circuits and Systems,” IEEE Trans. Nucl. Sci., vol. 69, no. 3, pp. 381-389, 2022. (IEEE Meritorious Paper Award)
A. Ildefonso et al., “Optimizing optical parameters to facilitate correlation of laser- and heavy-ion-induced single-event transients in SiGe HBTs,” IEEE Trans. Nucl. Sci., vol. 66, no. 1, pp. 359-367, 2019. (IEEE Outstanding Conference Paper & Outstanding Student Paper)
Full publication list: Google Scholar
Opportunities
For PhD Students:
In HERMES Lab, you’ll work at the intersection of experimental physics, machine learning, and electronic design for radiation environments. You’ll gain expertise in:
- High-throughput radiation effects testing (laser characterization, particle testing)
- Data science and machine learning for physics discovery
- Multi-scale modeling of radiation effects from quantum mechanics to system behavior
- Radiation-aware design and optimization
Interested? Contact Dr. Ildefonso: aildefon@iu.edu