Chapter 6 · Lesson 4 of 4 · about 12 minutes
Localization tests and selecting the campaign
A focused pulsed laser, scanned across a die, locates the regions that produce an upset, a transient or a functional interrupt. That helps in understanding a failure signature and in testing a recovery path before or alongside accelerator work, and the laser run needs as clear an engineering question as the beam run.
What you need to know
A focused pulsed laser deposits charge where the beam is focused. Scanned across a die, it builds a map of where an upset, a transient or a functional interrupt begins. Single-photon absorption at wavelengths near 800 nm deposits charge along the beam from the surface down and is blocked by metal. Two-photon absorption near 1260 nm, entering through a polished backside, deposits charge only at the focus and reaches under the metal stack. The laser answers questions an accelerator cannot: which block of a complex device produces the interrupt, whether a candidate mitigation removes it, and what the transient looks like at a known location and time. It does not answer how often. Laser pulse energy is not an LET, the charge distribution differs from an ion track, and reflections and carrier dynamics vary across the die, so a device-specific calibration against ion data is required before any laser threshold is quoted in LET. A region that is dark to the laser may be shielded by metal rather than immune.
The beam is only part of what the campaign costs. A latchup run of 10⁶ ions per cm² at 10³ ions per cm² per second (a low flux, so that each high-current event can be seen and cleared) takes 17 minutes of beam. Each run also costs a change of ion, a tune, a dosimetry check, and a sample or bias change, and ten minutes is an optimistic allowance for that overhead. The U1 plan of lesson 6.1, six LETs at two bias conditions on two parts, comes to 24 runs at about 27 minutes each, about 11 hours. That is more than a shift, with nothing in reserve for the surprise that a first run usually produces. The plan should therefore rank the runs by the uncertainty each one resolves and hold back beam time and parts for the branch the team expects least.
The facility is chosen for the evidence it can produce. For heavy ions, that means the LET range and the range at depth (chapter 6 lesson 3), the cocktail change time, flux control down to the low fluxes a SEFI-prone device needs, beam size and uniformity over the die or board, the dosimetry method, vacuum or in-air operation (which decides cabling, thermal control and how a heated latchup test is built), and whether the operators will accept the test hardware. For protons, it means the energies, the degrader options, the dose per unit fluence and the activation of the parts. For a pulsed laser, it means optical access (front or back side, thinning, polishing), wavelength, pulse energy calibration and scan coverage. For all of them it means the current schedule, the sample preparation lead time and the facility's data interface. The best facility is the one that can produce the evidence the application needs within a credible total plan, and that is frequently not the one with the highest LET.
Worked example 6.4. The beam-time budget of the U1 campaign
| Item | Value | Note |
|---|---|---|
| Runs planned | 24 | six LETs × two bias conditions × two parts |
| Fluence per run | 1.0 × 10⁶ | a clean run bounds σ below 3.0 × 10⁻⁶ cm² at 95 % |
| Beam time per run | 17 min | at 1.0 × 10³ ions per cm² per second |
| Overhead per run | 10 min | ion change, tune, dosimetry, sample or bias change |
| Total | 10.7 h | more than one shift; no reserve |
The plan therefore runs the second part at the second bias condition only if the first part shows events. That brings the committed time to 8.0 hours and leaves the rest as reserve. The first run, at the highest LET, maximum input and 125 °C, decides the branch. If U1 shows high-current events there, the plan continues down in LET to find the onset and adds the recovery demonstration. If it does not, the lower LETs are not needed for latchup and the beam goes to the SET characterization of U3 instead. With both branches and their run counts written into the plan, the team can decide during the shift without working the campaign out again. The reserve is two parts and two hours; without it, a destructive outcome on the first part would end the campaign before the recovery path is exercised.
Exercise
Give the laser run a defined objective (a location, a mechanism or a recovery path) and record which of its findings still need confirmation with ions.
Hold a readiness review with the facility and the test team against the measurements the plan actually needs.
Going deeper
Pulse energy is not an LET
Single-photon and two-photon absorption each deposit charge in a spatial distribution that differs from an ion track. Reflections, carrier diffusion, focusing and nonlinear absorption change both the charge deposited and the charge collected. Converting a laser threshold to an ion threshold or an on-orbit rate requires a correlation measured on that device and a statement of what the scan covered; there is no general conversion from laser energy to LET.
A laser sensitivity map locates mechanisms. It does not give a heavy-ion cross section or a bound on destructive effects.
Spend the budget on the uncertainty that matters
Plan the campaign in stages, with the branches for a clean result, a rare event, an unexpected mode and a destroyed part written down in advance. More devices, more fluence and more operating conditions resolve different uncertainties, so trade them against each other deliberately. Counting statistics do not substitute for sampling the population, and more samples do not reach a sensitive region the ion cannot penetrate or a signature the instrumentation cannot see.
Record which questions remain open after the campaign the project can afford, and which design, procurement or analysis action answers each one.
The common mistake
A location that is silent under the laser may be hidden by metal rather than immune.
Money spent, calendar time used and fluence delivered are three different things, and none of them measures assurance on its own.
Takeaway
- Because the laser controls position and timing, it can find causes that broad-beam counts cannot.
- The best facility is the one that can produce the evidence the application needs within a credible total plan.
Check yourself
A laser scan of U2 finds the SEFI origin in the refresh controller and shows that a firmware change removes it. What is established, and what is not?
Established: the location, and a mitigation that works for the injected charge at that location. Not established: the heavy-ion or proton cross section of the remaining SEFI modes, or that the controller is the only origin. The ion run that confirms the rate is still needed.
Why is a run with no events still charged to the budget as a full run?
Because what the run can tell you is set by its fluence. The bound it supports depends on the live fluence delivered, so a run cut short to save time supports a weaker bound.
What makes a heated latchup test harder in vacuum?
Thermal control and cabling. In vacuum the device heats and cools by conduction only, feedthroughs are limited, and the heater, thermocouple and current monitor must be qualified for the chamber. An in-air facility simplifies this, at the cost of the range the lower-energy ions lose in air.
A two-photon backside scan of U1 finds no location at which a laser pulse produces a high-current event, and the team proposes to drop the heavy-ion latchup runs. What is missing?
A laser scan locates mechanisms; it does not give a heavy-ion cross section or a bound on destructive effects. Pulse energy is not an LET, the charge distribution differs from an ion track, and reflections and carrier dynamics vary across the die, so without a device-specific calibration against ion data a silent scan cannot be placed on the LET axis the mission rate needs. The scan also has to state what it covered, and the ion run remains the evidence for how often and for the bound at the fluence the plan set.
To fit the U1 campaign into one shift, a colleague proposes to raise the flux tenfold. What does that change, and what does it risk?
It shortens the 17 minutes of beam per run but leaves the ten minutes of overhead per run (ion change, tune, dosimetry, sample or bias change) untouched, so the saving is smaller than it looks. The low flux was chosen so that each high-current event can be seen and cleared before the next ion arrives; at a higher flux events can arrive faster than the limiter clears them, and the run then measures recovery under pileup rather than the cross section (lesson 7.2). The plan instead cuts the committed time to 8.0 hours by running the second part at the second bias condition only if the first part shows events, and keeps the rest as reserve.
Why does the U1 plan put the harshest run (the highest LET, maximum input, 125 °C) first, rather than starting at low LET and working up to the onset?
Because that run decides the branch. If U1 shows high-current events there, the plan continues down in LET to find the onset and adds the recovery demonstration; if it does not, the lower LETs are not needed for latchup and the beam goes to the SET characterization of U3. Starting at the bottom spends runs before the question that shapes the rest of the shift has been answered. The cost is that the first part may be lost to a destructive outcome, which is why two parts and two hours are held in reserve; without them, that outcome would end the campaign before the recovery path is exercised.
Further reading
- S. Buchner, F. Miller, V. Pouget and D. McMorrow, "Pulsed-laser testing for single-event effects investigations," IEEE Trans. Nucl. Sci., vol. 60, no. 3, pp. 1852-1875, 2013.
- R. Koga, "Single-event effect ground test issues," IEEE Trans. Nucl. Sci., vol. 43, no. 2, pp. 661-670, 1996.
- JEDEC JESD57A, "Test procedures for the measurement of single-event effects in semiconductor devices from heavy-ion irradiation," 2017.
- ESCC Basic Specification No. 25100, "Single event effects test method and guidelines."
- NASA, "Avionics Radiation Hardness Assurance (RHA) Guidelines," NASA/TM-20210018053, 2021. NTRS 20210018053.