Observational Techniques & Instrumentation
Charge Transfer Inefficiency: Trailing Smears in a CCD
When a fresh CCD reads out, roughly 0.999996 of every electron packet survives each hop between pixels — but after years in orbit, radiation-carved defects in the silicon can steal a percent or more from a faint galaxy, dragging a comet-like smear behind it toward the readout amplifier. This is charge transfer inefficiency (CTI): the failure of a CCD to shift a charge packet perfectly from one pixel to the next during readout, leaving a trail of belatedly released electrons.
CTI is the dominant instrumental limit on precision imaging from space telescopes such as Hubble, Gaia, Chandra, and Euclid. It corrupts photometry, astrometry, and the exquisitely faint galaxy-shape measurements that weak gravitational lensing depends on — and because the damage accumulates with cosmic-ray dose, it grows relentlessly over a mission's life.
- RegimeCCD readout in radiation-damaged detectors
- Key numberCTE ≈ 0.999996 fresh; CTI = 1 − CTE grows to ~10⁻⁵–10⁻⁴/transfer
- Driven byLattice defect traps (divacancy ~0.17 eV, P-V E-center ~0.44 eV) from proton/electron irradiation
- PhysicsShockley–Read–Hall capture/emission; τ_emit ∝ T⁻² exp(ΔE/kT)
- Observed withHST ACS/WFC & WFC3, Gaia, Chandra ACIS, Euclid VIS (extended pixel-edge tests, trap pumping)
- Matters forFaint-source photometry/astrometry; weak-lensing galaxy shapes
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
What CTI is and why it matters
A CCD does not read all its pixels at once. Instead, it moves charge packets in bucket-brigade fashion: applied clock voltages shift each packet, row by row (the parallel/vertical direction) and then pixel by pixel along the serial register (the horizontal direction), until every packet reaches a single readout amplifier. A megapixel image therefore requires millions of individual charge transfers, and the fidelity of each hop is measured by the charge transfer efficiency (CTE). Its complement, CTI = 1 − CTE, quantifies the fraction lost per transfer.
Even a tiny per-transfer loss compounds. A source in the far corner of a 2k×4k detector may endure several thousand transfers before it is digitized, so a CTI of only 10⁻⁴ can bleed a large fraction of a faint packet into a trailing smear. Because the effect is worst for the faintest sources and the pixels farthest from the amplifier, CTI biases photometry, shifts measured positions (astrometry), and distorts source shapes — precisely the quantities precision astronomy cares about most.
The mechanism, step by step
Silicon collects photoelectrons in potential wells defined by the clock electrodes. A perfect lattice would pass every electron along untouched. But energetic particles — relativistic protons and electrons trapped in Earth's radiation belts, plus solar and cosmic-ray events — knock silicon atoms out of place, creating stable defects: vacancies, the divacancy, and the phosphorus–vacancy E-center. Each defect introduces an energy level inside the band gap that can capture an electron from a passing packet and hold it.
The capture-and-release is governed by Shockley–Read–Hall (SRH) statistics. A trap grabs an electron on a capture timescale set by the packet's electron density; it then re-emits after a characteristic emission time constant τ that depends on the trap's depth ΔE below the conduction band and the temperature: τ ∝ T⁻² exp(ΔE/kT). If τ is longer than the interval between clock shifts, the electron is released too late — into a pixel that trails behind its home pixel in the readout direction. That belated release is the smear.
Characteristic numbers, scales, and the key relation
Scientific-grade CCDs ship with CTE ≈ 0.999996, i.e. CTI ≈ 4×10⁻⁶ per transfer. Radiation damage in low-Earth orbit steadily fills the silicon with traps, pushing CTI toward ~10⁻⁵–10⁻⁴ over years; Hubble's ACS and WFC3 have tracked this rise across two solar cycles. The two most important species are the divacancy, a shallow trap ~0.17 eV below the conduction band that dominates CTI at temperatures below ~250 K, and the deeper E-center at ~0.44 eV.
The controlling equation is the SRH emission-time relation τ_e ∝ T⁻² exp(ΔE/kT). It has a counterintuitive consequence: colder detectors have longer emission times, so at very low temperatures charge stays trapped through many transfers and can be lost entirely — while at warmer temperatures τ shrinks toward the transfer time and much of it is returned promptly. Operators therefore choose an operating temperature that balances dark current against trap release, and shorten the effective 'distance to readout' by clocking and injection tricks.
How CTI is detected and measured
The classic diagnostic is the extended-pixel-edge response (EPER) and first-pixel-response test: overclock the CCD past the last real pixel and read out the deferred charge that trickles out of traps as an exponential tail — its amplitude directly measures CTI. Astronomers also monitor CTI in flight using warm pixels and cosmic-ray hits as natural point sources, watching how their trails lengthen with detector position and epoch.
The most incisive laboratory method is trap pumping (pocket pumping): charge is clocked back and forth across a region so that individual traps repeatedly capture and release, producing tell-tale dipole signatures. By varying the clock timing and temperature and fitting SRH theory, one measures each defect's emission time constant, energy level, and capture cross-section — and locates traps to sub-pixel accuracy. These characterizations feed pixel-based correction codes (for Hubble and Euclid) that read the trailed image and iteratively reconstruct the original charge distribution before science measurements are made.
Where CTI operates — and what it is not
CTI afflicts essentially every long-lived CCD in space: Hubble's ACS/WFC and WFC3/UVIS, the Chandra ACIS X-ray spectrometer (where CTI also degrades energy resolution because event energy is proportional to collected charge), ESA's Gaia astrometry mission, and Euclid's VIS imager. In each, the damage dose scales with time in the belts and spikes after large solar events and coronal mass ejections.
It is distinct from several look-alikes. Blooming is charge overflowing a saturated well into neighbours, independent of readout. Bleeding/smearing from a frame-transfer or shutterless readout is geometric, not trap-driven. Persistence in infrared detectors is a latent image from a different physical mechanism (charge in the diode, not bucket-brigade transfer). CTI is uniquely (a) directional — always trailing toward the readout node — (b) nonlinear in flux, hitting faint packets hardest, and (c) history-dependent, since a bright prior source can fill traps and temporarily protect the next one.
Open questions and significance
CTI is arguably the single most important instrumental systematic for space-based weak-lensing cosmology. Trailed charge imprints a coherent, spurious ellipticity aligned with the readout direction that can masquerade as cosmic shear; Euclid's VIS science requires this be corrected to Δe well below 10⁻⁴. Because CTI is not a convolution, it cannot be removed by standard PSF-deconvolution shape methods — it must be undone pixel by pixel, forward-modelling the trap physics.
Open challenges remain. Trap populations and their time constants drift with temperature history, annealing, and clock waveforms, so correction models must be continually recalibrated in flight. Mitigation strategies — charge injection to pre-fill traps, optimized clocking, and warmer/colder operating points — trade one systematic for another. And as detectors accumulate dose over a mission's life, corrections must extrapolate into damage regimes never tested on the ground. Getting this right underpins the shape measurements from which Euclid and the Roman Space Telescope hope to pin down the equation of state of dark energy.
| Property / case | Typical value | Notes |
|---|---|---|
| Fresh CTE (per transfer) | 0.999996 (CTI ≈ 4×10⁻⁶) | Manufacturer spec for scientific-grade CCDs before irradiation |
| Divacancy trap level (below E_c) | ~0.17 eV | Dominates CTI at low temperature (< ~250 K); short emission time |
| Phosphorus–vacancy E-center | ~0.44 eV | Deeper trap; frozen out on cooling (released only slowly when cold); important in n-channel devices |
| Emission time constant (T dependence) | τ ∝ T⁻² exp(ΔE/kT) | Colder ⇒ longer τ ⇒ more charge lost per fast transfer |
| Transfers to amplifier (full frame) | ~1000–4000 | A packet in a far corner can suffer thousands of trapping chances |
| Weak-lensing shape tolerance | Δe ≲ 10⁻⁴ (spurious ellipticity) | Sets Euclid VIS CTI-correction budget for cosmic-shear science |
Frequently asked questions
What is the difference between CTE and CTI?
They are complements of the same quantity. Charge transfer efficiency (CTE) is the fraction of a charge packet successfully shifted from one pixel to the next during readout; charge transfer inefficiency is CTI = 1 − CTE, the fraction lost. A fresh scientific CCD has CTE ≈ 0.999996, so CTI ≈ 4×10⁻⁶ per transfer. Astronomers quote whichever is more convenient, but they describe the same trap-limited process.
Why does the smear always trail toward the readout amplifier?
Charge is clocked in a fixed direction toward a single output amplifier. A trap captures electrons from a packet, then re-emits them a short time later — by which point the clocks have already moved on to the next transfer step, so the released electrons land in pixels the packet has just vacated, i.e. behind it relative to its travel direction. The trail therefore always points back along the readout path, never ahead of the source.
What actually creates the charge traps?
Radiation. Energetic protons and electrons from Earth's radiation belts, plus solar-particle events and cosmic rays, displace silicon atoms and create stable lattice defects — chiefly the divacancy (a trap ~0.17 eV below the conduction band) and the phosphorus–vacancy E-center (~0.44 eV). Each defect adds an energy level inside the band gap that can capture and later release electrons, following Shockley–Read–Hall statistics.
Why does cooling the CCD sometimes make CTI worse?
Because the trap emission time constant follows τ ∝ T⁻² exp(ΔE/kT), which grows as temperature falls. Colder detectors hold trapped electrons longer, so charge stays captured across more transfers and is more likely to be lost from its home packet entirely. Warmer operation shortens τ so electrons are returned promptly, but that raises dark current — so operators pick a temperature that balances the two.
How do astronomers correct for CTI in real data?
They characterize the trap population — densities, energy levels, and emission time constants — using extended-pixel-edge (overclock) tests and, in the lab, trap pumping. That physical model drives a pixel-based correction code that reads the trailed image and iteratively reconstructs the pre-transfer charge distribution. Hubble's ACS and Euclid's VIS both use such forward-modelling corrections, applied before photometry, astrometry, or galaxy-shape measurement.
Why is CTI such a problem specifically for weak gravitational lensing?
Weak lensing infers dark matter and dark energy from tiny, coherent distortions in galaxy shapes. CTI adds its own coherent, spurious ellipticity aligned with the readout direction that mimics real cosmic shear, and it hits faint galaxies hardest because a fixed number of traps steals a larger fraction of a small packet. Since CTI is nonlinear and not a convolution, standard PSF methods cannot remove it — it must be undone pixel by pixel to reach Euclid's Δe ≲ 10⁻⁴ requirement.