Chapter 4 · Lesson 3 of 6 · about 14 minutes
Power devices: burnout and gate rupture
Power devices can be destroyed by a single ion at a voltage below their electrical rating. The stress that matters includes the drain or collector voltage, the gate bias, switching transients, temperature, and the energy and path of the particle. A wide bandgap does not make a device immune to destructive SEE.
The space-rha.com power-device laboratory: silicon burnout, gate rupture and the SiC view, with bias, angle, range and temperature controls.
An original space-rha.com laboratory, embedded unchanged with its own controls; its onset values, gains and depths are teaching assumptions, not limits for flight hardware.
Rated voltage is not a radiation operating limit.
Burnout, gate damage and recoverable transients are different failures.
SEB in a Si MOSFET involves a regenerative carrier process and enough energy deposition to destroy the structure. SEGR is a failure of the dielectric under an oxide field that the event has enhanced. SiC and GaN devices have different structures and their own competing failure paths, so a silicon explanation, or a silicon voltage fraction, cannot be assumed to transfer to them.
Map the evidence across the bias space the application uses, including the off-state and switching stresses. After an event, measure the leakage, check the gate and confirm the device still works, because a current trip alone does not say whether the device was damaged.
Evidence to retain. Record the exact die technology, the terminal biases, the switching waveform, the gate drive, the temperature, the ion energy and range, what counted as a destructive event and what was measured afterward.
Mission context. A DC bus voltage that looks conservative can be exceeded at the device by turn-off overshoot or by a transient on the gate.
Going deeper
Electrothermal feedback can outrun external protection.
Carrier multiplication, redistribution of the field, parasitic conduction and local heating feed one another. Which of them controls the failure can depend on the drift design, the body structure, any heterointerfaces and the deposition profile of the ion. A local failure can be over in less time than a board-level limiter needs to remove the stored energy.
The field sketch in the laboratory is not a TCAD solution. A device-specific model needs its impact-ionization, transport and thermal assumptions checked against data taken over energy, range, angle and bias.
Takeaway
Base the derating on particle and bias data that apply to the device and on the actual waveform at its terminals, not on the datasheet rating alone.
Check yourself
A 200 V silicon MOSFET switches an 80 V rail (the step-test example of lesson 7.4). The designer argues that the rail is so far below the rating that no radiation derating is needed. What is wrong with that argument?
Rated voltage is not a radiation operating limit; a power device can be destroyed by a single ion at a voltage below its electrical rating. The derating has to come from particle and bias data that apply to the device: lesson 7.4 steps the drain voltage at LET 37 with the gate at 0 V, finds the highest passing voltage at 150 V and takes 75 % of it, 112 V, as the safe operating area, and the application is inside it because the off-state drain-source voltage, with its switching transients, stays below 100 V. The comparison has to use the actual waveform at the terminals, because turn-off overshoot or a transient on the gate can exceed a DC bus voltage that looks conservative.
A GaN transistor replaces a silicon MOSFET, and the team applies the same derated fraction of the rated voltage that served the silicon part. Why does the course reject that step?
A wide bandgap does not make a device immune to destructive SEE. SiC and GaN devices have different structures and their own competing failure paths, so a silicon explanation, or a silicon voltage fraction, cannot be assumed to transfer to them. The evidence has to be mapped across the bias space the application uses, including the off-state and switching stresses, for that device; for SiC and GaN the degradation is generally gradual leakage growth rather than a single rupture, and the angular dependence is part-specific (lesson 5.2).
A heavy-ion run on a power MOSFET ends with no current trips, and the part is reported as having survived at that drain voltage. What measurement is missing?
A current trip alone does not say whether the device was damaged. SEGR is a failure of the dielectric under an oxide field that the event has enhanced, and it can leave gate damage or raised leakage rather than an immediate burnout, so after the exposure the leakage has to be measured, the gate checked (lesson 7.4 adds a gate stress test after irradiation) and the function confirmed. The record has to say what counted as a destructive event and what was measured afterward.
Why are burnout and gate rupture argued by avoidance, with a derated safe operating area, rather than mitigated by a current limiter the way a latchup can be?
Carrier multiplication, redistribution of the field, parasitic conduction and local heating feed one another, and this electrothermal feedback can outrun external protection: a local failure can be over in less time than a board-level limiter needs to remove the stored energy. A latchup is a sustained path that the supply keeps feeding, which protection can interrupt (lesson 4.2); burnout and gate rupture are generally complete within the ion strike. The derated fraction covers the step size and the part-to-part variation, and it is a convention of the organization, written as such (lesson 5.2).
A derating test steps the drain voltage in DC with the gate held at 0 V, and the part is then used in a hard-switched actuator branch. Which conditions of the application does the test not cover?
The stress that matters includes the drain voltage, the gate bias, the switching transients and the temperature, together with the energy and path of the particle. A DC step with the gate held off covers the off-state at that gate condition; the switching waveform at the terminals, including turn-off overshoot and transients on the gate, the gate drive and the temperature have to be mapped as well, and the destructive path often starts in the driver, where a transient that turns the switch on into a short is reviewed under the analog and digital flows (lesson 5.2). The record should hold the exact die technology, the terminal biases, the switching waveform, the gate drive, the temperature and the ion energy and range.
Further reading
- NASA Avionics RHA Guidelines, NASA/TM-20210018053 (2021), §5 TID
- G. R. Allen, "Predicting On-Orbit Performance of Devices," RADECS 2025 Short Course, Slides 4–5, 10–11, 12–22
- IEEE NSREC 2020 Short Course, "Short Course"
- Sexton (2003), Destructive single-event effects in semiconductor devices and ICs
- Space-RHA: power derating tool
- G. R. Allen, "Predicting On-Orbit Performance of Devices," RADECS 2025 Short Course, Slides 2–10, 25–30