Acousto-optic tunable filter
Selects one or more spectral lines and passes them straight through — multiplexed, with every line open at once, or sequential, stepping through them one at a time. The beam depleted of those lines is deflected to a configurable angle and can be shown or hidden.
Open in the canvas →In the real world
An AOTF selects colours electronically. Sound travelling through a crystal compresses and rarefies it, and since the refractive index follows density, an acoustic wave is a moving index grating that light can diffract from. That much it shares with an acousto-optic modulator. The difference — and the whole point of the device — is that an AOTF arranges the interaction so only one wavelength at a time can diffract from a given tone.
It does this in a birefringent crystal, usually tellurium dioxide, in a geometry where the diffracted light emerges in the orthogonal polarization state. Because the two states have different refractive indices, the momentum-matching condition between the optical and acoustic waves is satisfied at only one optical wavelength per acoustic frequency. An AOM diffracts whatever you send it; an AOTF picks a line out of it.
Change the RF drive frequency and you change the line. That is the tuning mechanism, and it is purely electronic — no filter wheel to rotate, no grating to turn. The RF power sets the diffraction efficiency, so the same device controls how much of that line gets through.
Multiplexed and sequential drive
The property that makes AOTFs indispensable is that the crystal does not have to be driven with one tone. Apply several RF frequencies simultaneously and each selects its own wavelength, with its own amplitude setting that line's intensity independently. That is multiplexing: every selected line is present in the output at the same time. One small crystal thereby replaces a rack of shutters, filters, and attenuators — which is why the laser combiner in a confocal or multiphoton microscope is almost always an AOTF.
Driving one tone at a time instead, stepping from line to line, is sequential operation. Only one wavelength is present at any instant. Because switching takes microseconds, the sequence can be faster than a pixel dwell, so a scan can step excitation wavelengths line by line or even pixel by pixel and build a separate image per colour — which is exactly what multiplexed drive cannot do, since there every colour arrives at once and the detector cannot tell them apart.
Typical passbands are one to a few nanometres — narrow enough to isolate one laser line from its neighbour. The light that is not selected is not absorbed; it simply fails to diffract and continues on, which means a real installation always has somewhere for it to go, usually a beam dump.
Beyond microscopy, AOTFs appear in hyperspectral and multispectral imaging, Raman instruments, fluorescence spectroscopy, and space-borne instruments where a filter wheel's mass and mechanism are unwelcome.
In OpticalSetup
The element is built around a list of selected lines, one per RF tone, in the same spirit as the SLM's stacked functions. A fresh AOTF has a single line; press Add line to stack more, up to sixteen. Each line carries its own wavelength and efficiency, so a weak line and a strong one can be selected together, while the passband is a property of the device and is shared by all of them — as it is in a real crystal, where the resolution follows from the interaction length rather than from which tone is applied.
That passband is a sinc², the phase-matching response of the acousto-optic interaction itself. The width set in the inspector is its full width at half maximum; either side of the central lobe it passes through true zeros, with sidelobes at 4.7% and 1.7% of the peak between them. Those sidelobes are the device's real rejection floor: a line sitting in the first one still gets a few percent through, which no rectangular passband would ever show.
Where the light goes
The selected lines leave along the incoming axis, so the selection stays on the optical axis and the rest of a setup can be built downstream of it in a straight line. Everything not selected — the beam depleted of those lines — is deflected to an angle you choose.
That depleted beam is hidden by default, because in a working instrument it goes straight into a dump and drawing it only clutters the figure. Turn on Show depleted beam while designing and it is traced to wherever it actually goes, so you can confirm nothing downstream is sitting in its path and put a beam dump there. Hiding it changes only the drawing: the power accounting is the same either way.
Driving the lines
Multiplexed drive opens every selected line at once, so all of them are in the output together and a spectrometer downstream shows the whole set. Sequential drive steps through them one at a time at a rate you set, so exactly one line is present at any instant and the spectrometer shows it change as the sequence advances.
The sequence runs on the canvas clock, slowed to a step or two a second — a real driver steps at kilohertz, far too fast to read — in the same illustrative spirit as a scanning galvo or a chopper wheel. Each line is fully open while it is its turn; the sequence chooses which line, not how much of it gets through.
What the numbers do
Selection is exact and conserves energy. A 20 nm window on a 420–700 nm supercontinuum passes 20/280 of the power and the depleted port carries the rest, summing to one. Efficiency multiplies on top, so three multiplexed lines at 0.9 selected from three matching laser lines deliver 2.7× a single line's worth. Narrow selections work too: a 0.5 nm line out of that supercontinuum is 0.18% of the beam and still traces correctly rather than being discarded as negligible.
The geometry is the reverse of a physical device. In a real AOTF the selected light is the diffracted first order and leaves at an angle, while the remainder passes straight through as the zeroth order. This element draws the opposite assignment — the selection continues along the incoming axis and the remainder is deflected — because it keeps a multi-line selection on the optical axis where the rest of a setup is built. The power accounting is identical either way; only which port is bent differs.
The passband is set directly rather than following from an acoustic frequency: in a real device one RF tone fixes the selected wavelength, the diffraction angle, and the polarization rotation together through phase matching, so a combination set here need not correspond to any crystal. Its shape is modelled — the sinc² below — but its sidelobes are truncated at the third zero rather than continuing to fall away forever, so the deepest rejection a real device gives out in the far wings is not reproduced. The polarization rotation itself is not modelled, so the selected light leaves in the state it arrived and cannot be cleaned up with a polarizer the way a real one is. There is no relation between RF power and efficiency, no acoustic transit time — lines switch instantly — and no crystal transmission range, so a line can be selected at any wavelength the source provides.