Chapter 4 · Lesson 6 of 6 · about 26 minutes
Displacement damage
An energetic recoil can knock an atom out of its lattice site. Some of the damage heals quickly; some defects remain and change the electrical or optical behavior of the device. Displacement damage matters most where the function depends on carrier lifetime, light generation, low noise or charge transfer.
The space-rha.com displacement-damage model: recoil cascade, energy partition, NIEL, and the device views for a bipolar transistor, an LED, a CCD and a solar cell.
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.
A displaced atom can change what the device collects.
Damage energy is not a count of electrically active defects.
Vacancies, interstitials, complexes and clusters each have different electrical effects. The particle species and energy set the distribution of recoils, and the material, its impurities, the temperature and the annealing decide which defects survive. A damage quantity is useful once it has been shown to correlate with the device response in question.
Identify the material, the function, the spectrum at the device and the endpoint before choosing a NIEL curve or an equivalent fluence.
Evidence to retain. Record the species and energy, the NIEL basis for that material, the fluence at the device, the temperature, the annealing, the ionizing dose that accumulated at the same time, and the correlation between the damage quantity and the endpoint.
Mission context. An instrument can lose detection efficiency or gain a little at a time without ever producing a discrete single-event effect.
Defects can remove useful carriers or create unwanted ones.
A defect can recombine a signal carrier before it is collected, or it can generate charge that shows up as leakage or dark signal. The same damage can therefore reduce the signal and raise the noise. Which of the two matters depends on the device and on how it is operated.
Lifetime sets the collection distance.
A shorter minority-carrier lifetime means a shorter diffusion length, so a carrier generated far from the collection region is less likely to reach it and contribute to the signal. In a depleted region, the same generation centers raise the leakage instead. Bias, temperature and the occupancy of the defects decide which of these dominates.
Compare the collection distance, the depleted volume, the integration time and the measurement temperature between the test and the application. One lifetime measurement does not give every endpoint.
Evidence to retain. Measure the dark and the illuminated behavior the application relies on, at controlled bias and temperature, and record the integration time and the annealing.
Mission context. A cooled detector can keep acceptable noise at one temperature and exceed its dark-signal budget in a warmer operating mode.
A bipolar device can lose gain as defects accumulate.
Minority carriers crossing the base can recombine at radiation-induced defects before they reach the collector. The same collector current then needs more base current. What that does to the circuit depends on the operating current, the feedback, the headroom and how much drive margin was available.
Separate bulk damage from surface ionization.
Displacement damage and TID can both reduce bipolar gain, through different paths, and a particle test may produce both at once. The change in gain depends on the operating current and the temperature, so one value of β at one current does not characterize the application.
Measure the gain, or the circuit performance, over the current range the application uses, and where the evidence allows it, say how much of the change is bulk damage and how much is surface ionization.
Evidence to retain. Record the collector and base currents, the temperature, the bias, the ionizing dose that accumulated alongside the displacement damage, the particle spectrum, the annealing and the exact process and lot.
Mission context. A current driver can lose compliance or overload an upstream stage as the base drive it needs increases.
Follow the complete optical signal path.
Displacement defects can reduce the light output of an emitter and the response of a detector. An optocoupler can degrade through both its emitter and its receiver, and the combined transfer can fall further than either change alone would suggest. What the system needs is enough received signal across temperature, drive current and the life of the mission.
Drive and temperature change the margin.
Nonradiative recombination competes with light emission in the emitter, and the gain or collection in the receiver can degrade as well. Current-transfer ratio, responsivity, bandwidth and noise are different endpoints, and a single CTR measurement at room temperature may not cover an application that depends on timing or on a threshold.
Test over the drive range, the load, the temperature and the decision threshold the application uses. Before transferring damage data from one optocoupler to another, compare the emitter material, the wavelength and the receiver technology.
Evidence to retain. Record the emitted and received signal, the CTR or responsivity, the drive current, the temperature, the wavelength, the load, the timing and the recovery or annealing history.
Mission context. An isolated command interface can drop below its receiver threshold after a gradual loss of CTR while both sides remain powered and otherwise functional.
Image quality can fail through a small defect population.
DD can create hot pixels, raise the dark signal and produce defect signals that vary with time. In a CCD, the traps can also interfere with charge transfer. The mean response can look fine while a small population of pixels damages the science measurement or the tracking performance.
CMOS dark signal and CCD charge transfer are different endpoints.
Dark-current nonuniformity, random telegraph signal, responsivity and charge-transfer inefficiency each need their own measurement. Cooling, integration time, clocking, the signal in the scene and the calibration all affect the image quality that results. TID can add surface contributions on top of the bulk displacement damage.
Keep the pixel-level distributions and maps, since the population that matters disappears into an average. Use integration times, temperatures, readout modes and illumination levels that represent the application.
Evidence to retain. Keep the per-pixel maps and distributions, the temperature history of the dark measurements, the integration time, the clocking, the signal level, the annealing and the revision of the processing algorithm.
Mission context. A star tracker can lose centroid accuracy to a sparse set of defective pixels while its average dark current stays within allocation.
Damage matters when carriers fail to reach the collector.
Solar cells and photodetectors depend on generating carriers and collecting them before they recombine. DD can reduce the collection and change the leakage and the operating curves. Depending on the device, the engineering endpoint may be end-of-life power, quantum efficiency, responsivity or noise.
Spectrum, junction design and operating point change the result.
Collection depends on how deep the carriers are generated, on the junction and depletion structure and on minority-carrier transport. A multijunction solar cell adds the current matching between its subcells. One reference-particle equivalence, or one material curve, may not represent every junction or every measured parameter.
Use the radiation spectrum the device will see and a damage basis matched to its material and response. Record how each endpoint was defined, the illumination spectrum, the temperature, the bias and the annealing.
Evidence to retain. Keep the spectra at the device, the equivalence method, the identity of the cell and its junctions, the I–V or spectral response, the temperature, the illumination and the annealing.
Mission context. A mission power budget has to account for the output the cell stack retains at end of life, while a detector budget may be set by noise rather than by signal loss.
A common damage scale is useful when the response follows it.
NIEL and displacement damage dose make it possible to compare exposures to different particles in a stated material. They do not guarantee that every defect population or every device parameter degrades in the same proportion. Equivalence is a model, and like any model it should be backed by data and used only within a stated scope.
Transport the spectrum first, then choose the damage basis.
The fluence that matters is the spectrum reaching the sensitive material, after the shielding, not the external fluence. State the NIEL curve, the material, the reference particle and energy, and the units. Then check whether the endpoint actually follows that damage quantity across the energies and particles involved, taking the annealing and the TID accumulated at the same time into account.
For a spectrum, integrate Φ(E)NIEL(E), and keep the energy-bin definitions and the units with the result. A single multiplier for one monoenergetic beam cannot stand in for a mixed field that has not been resolved.
Evidence to retain. Save the spectra, the units, the NIEL curve and its version, the reference basis, the integration convention and the range over which the correlation was checked. Tie the resulting damage quantity to data on the specific endpoint.
Mission context. A change in shielding changes the proton energies and the secondary fields at the part, so the equivalent fluence has to be recomputed from the new transported spectrum.
Going deeper
Cascades evolve before the device sees the final defect population.
Nuclear collisions create primary knock-on atoms, which create secondary ones. The energy is divided among ionization, atomic motion and lattice excitation, and vacancy and interstitial pairs that end up close together can recombine. The defects that remain, and the complexes they form with impurities, introduce energy levels with their own capture cross sections and thermal kinetics.
The threshold displacement energy depends on the material and on the crystal direction.
SRH kinetics links energy levels to generation and recombination.
Capture and emission rates depend on the energy level of the defect, its capture cross sections, the carrier densities and the thermal velocity. In the SRH approximation the electron and hole lifetimes share one denominator, and the sign of np − ni² decides whether the net result is recombination or generation. Clusters and interacting defects can behave in ways an isolated-level picture does not cover.
U = (np − ni²) / [τp(n + n₁) + τn(p + p₁)]. L = √(Dτ) assumes diffusion with a meaningful effective lifetime.
Base transport competes with recombination.
Bulk defects shorten the minority-carrier lifetime and can increase recombination in both the neutral and the depleted regions. The effect on the base transport factor depends on the base width, the diffusion length, the injection level and the electric fields. At larger damage, changes in carrier concentration or compensation may matter as well.
A lifetime law can show the trend, but β is not simply τ/τ₀. A quantitative answer for the actual transistor over its current range needs a calibrated transport and circuit model.
Defect-assisted recombination changes quantum efficiency.
Radiative emission competes with SRH recombination and other nonradiative processes. New defects can change the injection efficiency and the lifetime in the emitter, and the receiver adds its own collection and gain mechanisms. Defect populations and annealing differ from one material to another, which makes NIEL scaling across emitters and detectors uncertain.
The normalized product model in the laboratory assumes that the emitter, the coupling and the receiver can be treated as separate factors. It shows how the chain multiplies; it does not fit a fluence response or predict a distribution of CTR.
Capture and emission timing control the artifact.
A defect can generate charge, switch between metastable configurations, or capture a signal carrier and release it later. In charge transfer, how much charge is deferred depends on the emission time compared with the time a packet dwells at the trap, and on the occupancy of the trap. With several trap species and a signal history, a correction may work under some conditions and not others.
A quantitative CTI model needs the trap densities, their energies and capture cross sections, the clock timing and validation against measurement.
A diffusion equation becomes a device response only with boundaries.
Lifetime enters the transport, but the geometry, the surface recombination, the depletion fields and the optical generation profile set the probability of collection. In a solar cell, the resulting photocurrent combines with the change in dark current to reshape the I–V curve. A ratio of lifetimes is not a power-retention factor.
Transport has to be solved, or validated, with the actual boundaries and material parameters. The diffusion-length example in the laboratory isolates one physical dependence out of several.
Energy partition is an abstraction of defect production.
Nuclear stopping and the damage-energy partition do not specify which defects are isolated, which form clusters, which survive or which are electrically active. The correlation with device response can fail between recoil regimes, between materials, in nanoscale structures or under different annealing conditions. A stopping code that treats only Coulomb interactions can leave out hadronic channels that matter.
SRIM/TRIM is the right tool for stopping and range questions within its scope; a mixed field with nuclear reactions needs a validated coupled transport code. Molecular dynamics and defect calculations work at other scales and need their own validation.
The common mistake
The single assumed K in the lifetime model does not define a defect spectrum or a set of SRH parameters.
The degradation path in the laboratory is an assumed interpolation. It is not an axis of radiation exposure.
The NIEL values used as inputs are arithmetic examples. They are not taken from a published material database.
Takeaway
- Tie the damage metric to a response that has been demonstrated in the actual material and device.
- Say which endpoint is being protected: signal collection, gain, leakage, noise or timing.
- Review the gain at the current where the circuit operates, and separate the ionizing contribution from the displacement contribution.
- Assess the link margin for the whole optical chain at its operating conditions.
- Choose image-level and pixel-population endpoints that reflect the measurement the instrument is meant to make.
- Derive the response quantity for the specific material, junction and function.
- Keep the NIEL curve and the check that the response follows it, and not only the final equivalent fluence.
Check yourself
A team converts the mission proton spectrum to an equivalent fluence with a silicon NIEL curve and applies it to a GaAs LED. What is wrong, and what has to be checked before any equivalence is used?
NIEL and displacement damage dose compare exposures in a stated material, and a silicon NIEL curve does not transfer unchanged to GaAs or InGaAs (lesson 5.2); the damage basis has to match the device material and the endpoint. Equivalence is a model: it does not guarantee that every defect population or every device parameter degrades in the same proportion, so the response should be checked across the energies and particles involved, with a second energy generally used to check the scaling, and the annealing and the TID accumulated at the same time taken into account. Record the NIEL curve and its version, the reference particle and energy, the units and the range over which the correlation was checked, and not only the final equivalent fluence.
Why does the course insist on transporting the spectrum through the shielding before choosing the damage basis, and what happens to the equivalent fluence when the shielding changes?
The fluence that matters is the spectrum reaching the sensitive material after the shielding, not the external fluence; for a spectrum the damage quantity is the integral of Φ(E)NIEL(E), kept with its energy-bin definitions and units. A change in shielding changes the proton energies and the secondary fields at the part, so the equivalent fluence has to be recomputed from the new transported spectrum. A single multiplier for one monoenergetic beam cannot stand in for a mixed field that has not been resolved.
A proton test on a bipolar transistor reports the gain loss at one collector current, and the circuit operates the transistor at a much lower current. What does the report leave open?
The change in gain depends on the operating current and the temperature, so one value of β at one current does not characterize the application; the gain, or the circuit performance, has to be measured over the current range the application uses, and what the loss does to the circuit depends on the headroom, the feedback and the drive margin. A proton test also produces displacement damage and TID at once, which reduce the gain through different paths (bulk recombination and surface ionization) and scale differently to the flight environment, so where the evidence allows it, the record should say how much of the change is each. The ionizing dose that accumulated alongside the displacement damage belongs in the record.
A cooled detector meets its noise budget at the temperature of the test. Why can the same defect population push it over its dark-signal budget in a warmer operating mode, and what should be compared between the test and the application?
A defect can recombine a signal carrier before it is collected, or it can generate charge that shows up as leakage or dark signal; the same damage can reduce the signal and raise the noise, and bias, temperature and the occupancy of the defects decide which of the two dominates. A shorter lifetime shortens the diffusion length, so carriers generated far from the collection region are lost, while in a depleted region the same generation centers raise the leakage. Compare the collection distance, the depleted volume, the integration time and the measurement temperature between the test and the application; one lifetime measurement does not give every endpoint.
An irradiated imager shows a mean dark current well inside its allocation and is accepted. Why can the star tracker it serves still lose performance, and what should the test record keep?
Displacement damage can create hot pixels, raise the dark signal, produce defect signals that vary with time and, in a CCD, interfere with charge transfer, and the mean response can look fine while a small population of pixels damages the measurement; a star tracker can lose centroid accuracy to a sparse set of defective pixels while its average dark current stays within allocation. Keep the per-pixel maps and distributions, since the population that matters disappears into an average, together with the temperature history of the dark measurements, the integration time, the clocking, the signal level, the annealing and the revision of the processing algorithm, and use integration times, temperatures and readout modes that represent the application.
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 2024 Short Course, "Short Course"
- IEEE NSREC 2013 Short Course, "Short Course"
- Srour & Palko (2013), Displacement Damage Effects in Irradiated Semiconductor Devices
- Space-RHA visual explanations
- Space-RHA: NIEL tool