Chapter 6 · Lesson 3 of 4 · about 14 minutes
Heavy ions and protons: what each test establishes
A high LET on the beam table is only the start. The ion has to pass through the lid, metal, substrate or other overburden and reach the region that produces the effect, and a high surface LET can leave a deep latchup structure untested.
What you need to know
A heavy-ion facility offers ion species and energies. The device responds to the linear energy transfer and the range at its sensitive depth, and the two are coupled: an ion slows as it penetrates, its LET rises toward the Bragg peak near the end of its range, and then it stops. The LET printed on the cocktail table is the surface value in silicon at normal incidence. If a lid, passivation, metallization, a flip-chip bump field or an unthinned substrate lies between the beam and the sensitive volume, the ion arrives with less range and a higher LET than printed, or does not arrive at all. The first task in selecting a beam is therefore to build the material stack of the prepared sample and to compute, with a stopping-power tool, the LET and residual range at the sensitive depth for every ion in the plan.
The figure shows range against LET for a 10 MeV per nucleon cocktail (indicative values; the facility's current table governs). Range falls as LET rises. The highest-LET ion in such a cocktail has about 90 µm of range in silicon, which is enough for a delidded front-side device but not for 100 µm of remaining substrate behind a flip chip. Latchup and power-device mechanisms involve structures tens of micrometers deep (the wells and substrate of CMOS, the epitaxial layer and body of a vertical MOSFET). A screen that reports no latchup with an ion that stopped above the well has tested nothing. Common practice asks for tens of micrometers of residual range beyond the sensitive depth, more for latchup and for power devices, and for the LET to be quoted at that depth.
Tilting the device raises the path length through a thin sensitive volume by 1/cos θ, and the effective LET quoted for a tilted run is LET/cos θ. The same factor applies to the overburden: at 60°, every micrometer of overburden costs two micrometers of range. The effective-LET shortcut also has a physical limit. It describes energy deposited along a longer chord through a thin slab, and it generally fails for latchup, for power devices and for any structure in which the collection depth is comparable to the lateral size, because there the ion species, the track and the direction each matter on their own. Keep the species, energy, angle and material stack in the data record, since a derived LET alone cannot be reinterpreted later.
Protons answer a different question. Above a few tens of MeV, a proton upsets a device mainly through nuclear reactions in or near the sensitive volume, and the recoils from those reactions deposit the charge. The cross section rises with energy to a plateau that is typically four or five orders of magnitude below the heavy-ion saturation value. Trapped protons dominate the single-event rate of soft devices in low Earth orbit, so a proton cross section is needed wherever the mission crosses the belts or sees solar events. A facility typically offers a high energy (around 200 MeV, where the cross section is near its plateau) and one or two lower energies (30 to 100 MeV, where most of the trapped-proton fluence above the floor lies) to establish the shape. The exposure also deposits dose (chapter 6 lesson 2), so the parts are consumed. In highly scaled technologies, low-energy protons (below about 2 MeV) can upset by direct ionization, which a high-energy run does not sample. Where no proton data exist, the heavy-ion curve is used to estimate the proton sensitivity by proxy methods that are generally conservative; the site's proton proxy tool implements one.
Worked example 6.3. A xenon run that never arrived
U1 is tested from the front with its lid removed, and its latchup-sensitive wells sit under 50 µm of overburden (passivation, metallization and the oxide stack of a thick-metal power process are taken as 50 µm of silicon-equivalent material for this example). The plan wants an effective LET near 120 MeV·cm²/mg for the high-temperature latchup screen and proposes xenon (surface LET 58.78, range 90 µm) at 60°.
| Angle | Path through the overburden | Nominal effective LET | Range left for the sensitive region |
|---|---|---|---|
| 0° | 50 µm | 59 | about 40 µm |
| 30° | 58 µm | 68 | about 32 µm |
| 45° | 71 µm | 83 | about 19 µm |
| 60° | 100 µm | 118 | none: the ion stops in the overburden |
At 60° the path through the overburden is 100 µm, longer than the range, so the ion stops before the wells. The run would report no latchup at a nominal LET of 118 that no structure experienced. At 45° the ion reaches the wells with about 19 µm of range to spare, near its Bragg peak, so its LET at depth is above the surface value and the effective-LET label no longer describes the deposition in a deep well. A better plan uses a higher-energy cocktail (more range per LET) or a thinner sample preparation, quotes the LET at the sensitive depth, and keeps the latchup claim at normal incidence or small angles.
A question to carry through the course
When, if ever, can the dose applied to a part during a proton SEE test be used as a proxy for a TID test?
Things to weigh here:
- The parts the proton run consumes, and whether they come from the flight lot.
- The dosimetry: the fluence is converted to dose through the stopping power in silicon, and the dose is uniform through the package at 200 MeV but not at low energy.
- What the run plan would have to add before the exposure could count as a dose test: a worst-case bias, parametric measurements between exposures, an anneal step.
Exercise
Before you approve an ion, check that it reaches the active region with range to spare and meets the requirements of the mechanism under test.
For each proton run, state whether it supplies a response curve, an exercise of a mitigation, or only an early check for susceptibility.
Going deeper
Do not reduce every response to LET
Two ions with the same LET can differ in range, in track structure and in the destructive response they produce. Tilting changes the path through the overburden, the projected area and the direction of the track. The effective-LET approximation assumes a thin planar slab, and it can fail for nonplanar geometry, for collection at several nodes and for power devices. Those cases need data at several angles and a physical model that fits the structure.
Keep the ion species, energy, angle of incidence and material stack in the data set rather than only a derived LET.
Model the secondary spectrum
A single proton energy samples one recoil distribution and one relation between energy and depth. High-Z structures near the sensitive volume and the device geometry can change the nuclear secondaries and the energy dependence. Inferring a heavy-ion bound from proton results, or the reverse, requires a validated model for the mechanism. Direct ionization by low-energy protons is a further pathway that a recoil-only picture misses.
Proton measurements reduce a stated model uncertainty; they do not replace heavy-ion coverage.
The common mistake
A surface LET with a cosine correction does not by itself show that the ion was equivalent to the mission particle at the sensitive depth.
A proton run with no events does not show that the part is free of destructive heavy-ion effects.
Takeaway
- An ion that stops short of the sensitive region produces a clean result that looks like hardness.
- A proton-rich environment needs proton data, and each kind of source leaves different mechanisms untested.
Check yourself
A report states "no SEL to an effective LET of 120 MeV·cm²/mg (Xe at 60°)". What should the reviewer ask for?
The ion energy, the sample preparation and overburden, and the range left at the sensitive depth. At 60° the path through the overburden doubles, and a 90 µm ion may not reach the wells. For latchup the effective-LET construction is generally not valid in any case.
Why is a proton run needed for U2 even though its heavy-ion curve is complete?
Trapped protons dominate the rate of a soft part in low Earth orbit (chapter 9 lesson 2). Their cross section comes from nuclear reactions and cannot be read off the heavy-ion curve without a proxy model, and proton-specific effects (direct ionization at low energy, SEFI modes) may differ.
What does the proton exposure of U2 cost beyond beam time?
The parts themselves: 10¹⁰ protons per cm² at 200 MeV deposits a dose of order a kilorad, and more at lower energy, so proton-tested parts are generally retired. If the same parts serve other runs, the dose has to be tracked.
In worked example 6.3 xenon at 45° reaches the U1 wells with about 19 µm of range to spare. A colleague records the run as a valid latchup screen at an effective LET of 83. What is wrong with that label?
Two things. With 19 µm left the ion is near its Bragg peak, so its LET at the wells is above the surface value of 58.78, and the 83 label (the surface LET divided by cos 45°) does not describe the deposition in a deep well; 19 µm is also short of the tens of micrometers of residual range that common practice asks for, more for latchup. Beyond that, the effective-LET construction is a thin-slab shortcut, and it generally fails for latchup, where the collection depth is comparable to the lateral size and the species, track and direction each matter. The better plan quotes the LET at the sensitive depth, uses a higher-energy cocktail or a thinner preparation, and keeps the latchup claim at normal incidence or small angles.
A flip-chip part arrives with 100 µm of substrate behind the active layer, and the test house proposes a latchup screen with its 10 MeV per nucleon cocktail, tilted to raise the effective LET. What does the range figure say?
The highest-LET ion in such a cocktail has about 90 µm of range in silicon, so it stops short of the wells, and tilting makes the problem worse, since the path through the overburden grows by 1/cos θ (at 60° every micrometer of overburden costs two micrometers of range). A clean result from that setup would show only that the ion never arrived. The screen needs a higher-energy cocktail with more range per LET, or a thinned substrate, and the LET quoted at the depth of the wells.
Why does a proton plan pair an energy near 200 MeV with one or two energies between 30 and 100 MeV, instead of taking the single highest energy the facility offers?
The cross section rises with energy to a plateau, and near 200 MeV it is close to that plateau, so the high-energy point sets the level. Most of the trapped-proton fluence above the floor lies between 30 and 100 MeV, where the curve is generally still rising, so the lower energies fix the shape in the part of the spectrum that produces the rate; a single high-energy point leaves that shape unmeasured. Neither run samples direct ionization by protons below about 2 MeV, which can matter in highly scaled technologies and is a separate question.
Further reading
- 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."
- R. Koga, "Single-event effect ground test issues," IEEE Trans. Nucl. Sci., vol. 43, no. 2, pp. 661-670, 1996.
- P. E. Dodd and L. W. Massengill, "Basic mechanisms and modeling of single-event upset in digital microelectronics," IEEE Trans. Nucl. Sci., vol. 50, no. 3, pp. 583-602, 2003.
- F. W. Sexton, "Destructive single-event effects in semiconductor devices and ICs," IEEE Trans. Nucl. Sci., vol. 50, no. 3, pp. 603-621, 2003.
- R. Ladbury, J.-M. Lauenstein and K. P. Hayes, "Use of proton SEE data as a proxy for bounding heavy-ion SEE susceptibility," IEEE Trans. Nucl. Sci., vol. 62, no. 6, pp. 2505-2510, 2015.
- Space-RHA, "Proton proxy tool," space-rha.com/proton-proxy-tool.
- Space-RHA, "SEL LET screening tool," space-rha.com/sel-let-tool.