Sample
Represents an illuminated specimen with configurable transmission and stacked signal channels, including two-beam signals that appear only while both pulses reach the spot together.
Open in the canvas →In the real world
A specimen can transmit or absorb excitation light and generate an optical signal. Fluorescence and coherent nonlinear signals arise through different processes; nonlinear microscopy includes two-photon fluorescence, second-harmonic generation, and coherent anti-Stokes Raman scattering[1].
In OpticalSetup
Orient the sample across the beam: at zero rotation its long axis is horizontal, so a horizontal incoming beam needs a 90° sample rotation. Choose the specimen mode, excitation transmission, and desired channels. The model supports stacked fluorescence, Raman, phase contrast, two- and three-photon fluorescence, second-order (χ⁽²⁾), THG, CARS and stimulated Raman signals.
The χ⁽²⁾ channel covers both second-order processes at once, as one susceptibility does: each beam's second harmonic, and the sum frequency of any two different colours on the spot. There is no separate sum-frequency channel; a scene saved with one is read as this channel, and a scene that carried both keeps its second-harmonic channel — the surviving one covers both processes, so the retired entry's own settings go with it rather than the specimen emitting each signal twice.
Two-beam signals need the pulses together. Sum frequency, CARS and stimulated Raman only happen while both pulses are at the spot, so the model gates them on the Gaussian overlap of the two arrivals and drops them below 2 % of the peak — the second harmonic of each beam, which needs one beam only, is unaffected. Arrival is each beam's own optical path plus its emission offset, judged beam to beam, so a millimetre of unmatched arm is 3.3 ps and enough to switch a picosecond CARS signal off; the workbench says so rather than leaving an empty detector to interpret. Only trains at the same repetition rate are mixed, and Needs pulse overlap switches the requirement off per channel for a schematic that is about the signal rather than the timing.
That gate is one qualitative envelope for every two-beam channel, not a process-specific delay response. A real CARS signal in the instantaneous non-resonant limit weights the pump twice, going as Ip²Is rather than as the product of two intensities, and a resonant vibration adds dynamics of its own; none of that is modelled here. Each beam's own path is grouped across the sampling rays that draw it, so the spread of arrival times across a focused cone is not treated as a timing spread — two beams on different paths, however, stay separate, and a colour arriving on two arms pairs with whichever arm meets the pulse.
The example produces a fluorescent signal under 488 nm illumination. Adjust its emission and transmission, then use a filter and detector to distinguish emitted light from the excitation. Parametric channels use a forward lobe with an optional weaker backward contribution.
Signals are bounded qualitative proxies, not measured cross-sections or calibrated conversion efficiencies. Each channel is bounded by its own authored efficiency, not by a shared energy budget: signals never deplete the excitation, so stacking channels — or adding a second colour, which adds a mixed pair — adds drawn signal power rather than dividing it. The sum frequency of a pair scales with the shorter beam's intensity alone, not with the product of the two. Stimulated Raman transfer is a normalised display of the modulation, not a Raman gain or loss law: whatever contrast the modulated beam has is stretched to the channel's full authored excursion — a loss on the shorter-wavelength pump, a gain on the longer-wavelength Stokes — timed to the photons of that beam that actually reach the spot, and a steady beam transfers nothing. It covers one unmodulated beam receiving the modulation of one other beam through one modulator; with both beams modulated, more than one modulated partner, or a modulated beam behind several modulators, no transfer is drawn and the specimen's timing readout says why. Sum frequency is gated on arrival time alone — no phase matching, polarization condition or focusing overlap is checked — and the pulses' own path spread across a focused cone is not treated as a timing spread. The model does not reconstruct a three-dimensional specimen or predict photochemistry, bleaching, or a diffraction-limited point-spread function.