Acousto-optic deflector (AOD)
Steers first-order light to a set deflection angle, held static or swept, with wavelength-dependent scanning and an optional zero order.
Open in the canvas →In the real world
An acousto-optic deflector steers a laser beam by changing a frequency. A piezoelectric transducer bonded to a transparent crystal launches a sound wave through it; the travelling compression makes a moving grating of refractive index, and light crossing that grating is diffracted. Change the drive frequency and the acoustic wavelength changes with it, so the diffracted beam leaves at a different angle — a scanner with nothing in it that moves[1].
It is the same interaction an AOM uses. The difference is entirely in the drive: a modulator is run at a fixed frequency and varying power, to switch a beam on and off, while a deflector is run at constant power and varying frequency, to point it[1]. The deflection angle follows from the Bragg condition, and for an isotropic medium it is
Three things follow from that one expression, and between them they explain every specification on an AOD datasheet.
The angles are small. Sound is slow and its wavelength is enormous next to light's, so the ratio is tiny. A 1064 nm beam in fused silica — sound speed 5.9 km/s — driven at 100 MHz deflects by 18 mrad, about one degree[1]. The usable range of a real deflector is "rather small — a few degrees"[1], with published devices quoting scan angles from roughly 5 to 60 mrad[2]. Anyone needing more puts a telescope after it, which trades beam width for angle.
Slow crystals are better. Since the angle goes as 1/v, a material with a slow sound wave gives more deflection for the same frequency range. This is why the standard choice for the visible and near infrared is tellurium dioxide driven on its slow shear mode, where sound travels at about 620 m/s — roughly a tenth of the speed in fused silica[1][2]. Fused silica is used in the ultraviolet and germanium in the mid infrared, in each case because the crystal has to be transparent before anything else matters[1].
The scan is chromatic. The angle is proportional to wavelength, so two colours entering together leave at different angles. That is a nuisance for a broadband beam and the whole point of an AOTF, which uses the same physics to select colours rather than to steer them.
Resolvable spots
The number that actually matters when choosing a deflector is usually not the scan angle but how many distinguishable directions fit inside it[1]. A beam cannot be pointed more precisely than its own divergence, so the resolution is the scan range divided by that divergence — equivalently, the time the sound takes to cross the beam multiplied by the frequency range it is driven over[1].
That product is why a deflector wants a wide, well-collimated beam: widening it lengthens the aperture time and buys resolution. It is also why resolution and speed pull against each other. The device cannot settle faster than sound crosses the beam, so the same choice that gives many spots makes each jump slower, and a crystal chosen for its slow sound wave is slow in both senses[1].
Two deflectors mounted at right angles steer in two dimensions[1]. Because the beam can be sent to any angle in the range as fast as it can be sent to the neighbouring one, an AOD pair can address points in an arbitrary order rather than rastering through them — the basis of random-access scanning in multiphoton microscopy, where the interesting neurons are visited and the space between them is not.
What a datasheet reports
Diffraction efficiency is typically 50–80%, sometimes near 90%, and lower at longer wavelengths[1]. It is polarisation dependent, and it peaks at the centre of the frequency range and falls away toward both ends — which is why it should be checked at the edges of the scan and not only in the middle[1]. Some devices compensate by raising the drive power at the extremes, and beam-steered designs use a phased array of electrodes to swing the acoustic wave direction and hold efficiency across a wider scan[1].
The undiffracted zero order carries whatever was not deflected. It does not move with the drive and is normally dumped[1].
One effect is worth knowing because it is invisible: the diffracted beam comes away shifted in optical frequency by exactly the drive frequency, since it has scattered from a moving grating. For a deflector this is "usually irrelevant"[1] — 80 MHz on a 532 nm beam is a shift of 7.6×10⁻⁵ nm — but it is the same effect an AOM is bought for.
In OpticalSetup
The deflector is specified the way you would choose one: by the angles. Set the centre deflection and, for a scan, the total scan angle swept around it. The drive frequency behind those angles is left implicit — reading θ = λf/v forwards, a scan linear in frequency is linear in angle, so the angles are the honest parameterisation and the crystal never has to be named.
The defaults are a real device: 4° of centre deflection and 2° of scan is what a TeO₂ slow-shear deflector gives at 532 nm on an 80 MHz drive across a 40 MHz bandwidth. The wavelength scaling is kept, referenced to the design wavelength, so a beam at twice that wavelength deflects twice as far and a broadband beam fans out — which is the chromatic behaviour a real deflector has.
Four drives are available. Static holds one angle. Triangle sweeps and retraces, sawtooth sweeps and flies back, and random step addresses one angle per step in an unpredictable order and holds it until the next — the random-access mode, rather than a sweep. The scan runs on the shared simulation clock at the rate set in kilohertz, so it stays phase-locked to pulses and to any other modulator on the bench.
The scan rate is bounded by the same physics that sets the resolution. The inspector reports the access time for the aperture in use, taken as 1.5 µs per millimetre for TeO₂ slow shear, and the rate that implies: a 20 mm aperture takes 30 µs to fill and so cannot be re-pointed faster than about 33 kHz, which is why catalogue random-access cycle rates sit between roughly 40 and 170 kHz[2] rather than in the megahertz. Ask for more and the readout says the crystal cannot settle that fast.
The angles are configured, not derived. Nothing here knows a crystal, an acoustic velocity, or an RF bandwidth, so a combination set on this element need not correspond to any device that could be built — and the ceilings allowed are deliberately looser than reality so an illustrative sketch stays readable. Real deflectors reach a few degrees at most.
Diffraction efficiency is a flat user-set fraction across the whole scan. A real one peaks at the centre frequency and falls away toward both ends, which is the specification that most often decides whether a device is usable, and it is polarisation dependent, which is not modelled either. There is no relation between drive power and efficiency.
The optical frequency shift is not applied. It is real, but at 7.6×10⁻⁵ nm for 80 MHz at 532 nm it is more than a thousand times finer than the finest wavelength difference anything in this workbench resolves; the AOM, which exists for that shift, still carries it.
Access time is reported but not enforced: the beam jumps instantly between angles, with no settling and no transient while the acoustic wave refills the aperture. The number of resolvable spots — arguably the figure that decides a real deflector's worth — is not computed at all, and neither is the cylindrical lensing a fast scan produces when different parts of the beam see different acoustic frequencies at once. Multi-tone drive, which addresses several angles at once, is not available: one drive, one deflected beam.
Related components
References
- “Acousto-optic Deflectors,” RP Photonics Encyclopedia
- Gooch & Housego — Acousto-optic deflectors: product specifications for TeO₂ and fused-silica devices