Supercontinuum laser
Emits a pulsed beam across a configurable spectral band, for exploring wavelength separation, filtering, and broadband illumination in one setup.
Open in the canvas →In the real world
A supercontinuum source produces light spanning hundreds of nanometres — often the whole visible range and beyond — while retaining the spatial coherence and collimation of a laser beam. It is, in effect, white light that behaves optically like a laser: it can be focused to a diffraction-limited spot and coupled into a single-mode fibre, neither of which a lamp of comparable bandwidth allows.
The broadening is not produced by the gain medium. A pump laser — typically a mode-locked oscillator delivering high peak power — is launched into a strongly nonlinear medium, most often a photonic crystal fibre engineered so that its zero dispersion wavelength sits near the pump. Over a few centimetres, a cascade of nonlinear processes redistributes the pump energy across a vastly wider spectrum: self-phase modulation broadens it initially, then soliton fission, Raman self-frequency shift, and dispersive wave generation extend the edges.
Because the process is pump-driven, the output inherits the pump's pulse train: a supercontinuum is emitted as pulses at the pump's repetition rate, not as steady light, even though it looks white. Spectral flatness and pulse-to-pulse stability vary considerably with how far into the anomalous-dispersion regime the source is driven.
Supercontinuum sources became practical laboratory instruments after photonic crystal fibre made it possible to place the zero-dispersion wavelength wherever the available pump happened to be, rather than the other way round. They are now standard in broadband spectroscopy, optical coherence tomography, and as tunable excitation sources for fluorescence microscopy, where a single box replaces a rack of discrete laser lines.
In OpticalSetup
Interference
With Beam with size and Interference enabled, split copies of this source interfere through ideal beamsplitters, fully reflective flat mirrors, delay lines and phase objects. The calculation integrates the actual spectrum: fringe contrast changes with arm mismatch, and the spectrometer sees the changed spectrum. Independent sources add as powers. New sources enable this; older saved sources keep it off until you enable it. The Coherence length readout gives the arm difference at which the fringes fall to half contrast for the spectrum as set.
The combined temporal field is not reconstructed. After recombination, duration, autocorrelation and oscilloscope traces are unavailable; pulse animation stops at the recombination surface. Material and coating phase outside the supported ideal optics is not inferred. A spectral or path-budget limit is reported explicitly. See the model and bounds.
The Supercontinuum laser replaces a single wavelength with a spectrum minimum and maximum, and emits a flat-top band between them. Downstream wavelength-selective elements — filters, dichroics, etalons, the spectrometer — integrate against that true flat profile rather than a centroid, so a 20 nm bandpass placed on a 400 nm-wide source transmits the fraction of power it actually overlaps.
Dispersive elements (prisms, gratings) sample the band at several discrete wavelengths and fan them out individually, each carrying its own wavelength-derived colour — which is why a prism turns this source into a visible rainbow even though the undispersed beam is drawn as a single broadband white line.
It carries the same pulse train as the Pulsed Laser, since a real supercontinuum inherits its pump's timing. Its pulse duration and envelope (Gaussian or sech²) are set directly rather than derived: in a real source they depend on the pump and on the nonlinear fibre that generated the continuum, neither of which is modelled here, and a continuum fresh out of the fibre is typically heavily chirped and picoseconds long. The duration you set is therefore the input to the propagation estimate, not a claim that the continuum is transform-limited.
The one limit the setting cannot cross is the transform limit. No pulse can be shorter than its own spectrum allows, so the duration has a floor set by the band, which the inspector shows as Transform limit. For a band hundreds of nanometres wide that floor is around a femtosecond and never gets in the way. A narrow band raises it: 690–700 nm cannot carry a Gaussian pulse shorter than 71 fs. Narrowing the band or changing the envelope lifts a duration that has fallen below the new floor. Widening the band never shortens the duration you set. The two endpoints are kept at least 10 nm apart: typing one past the other stops it a step short, because a band of zero width has no transform limit at all.
Broad-band temporal spread
A centre-wavelength GDD is not extended across this flat band. For every traced length of catalogue glass, OpticalSetup evaluates the Sellmeier group index at both authored endpoints and accumulates their signed group-delay difference. The displayed duration is the input duration and that endpoint spread added in quadrature. This is a robust first arrival-to-last arrival estimate for the declared flat spectrum, and detector, scope, probe and pulse packets use the same accumulated value.
The authored duration acts as a floor. A compressor can take the path's spread back out, but not the chirp the source was authored with, so a fully compensated continuum returns to the duration you set, not towards its transform limit. Compensation is also only ever partial: a single GDD value cancels the glass's dispersion at one wavelength, while the Sellmeier curvature across hundreds of nanometres remains. Through 100 mm of N-BK7, a 400–900 nm continuum spreads to about 20 ps; a compressor set to exactly cancel the centre-wavelength GDD still leaves about 4 ps, and the best a single GDD reaches is under 1 ps.
Once a filter, dichroic or AOTF cuts into the continuum, the duration is shown as unavailable, whether the filter sits before the glass or after it. The accumulated spread belongs to endpoints the filter may have removed, and reconstructing the slice would need each wavelength's own delay history rather than a single number.
The spectrum is an idealized flat top, not a measured shape with the peaks, dips, and edge roll-off of a real continuum, and its shape does not change with pump power. No broadening is simulated: the band is declared, not generated from a pump and a nonlinear fibre. Pulse-to-pulse spectral noise, a real limitation of these sources, is not represented. The endpoint estimate is not a propagated complex field: it cannot reproduce sub-pulses, wavelength-dependent intensity, non-monotonic group delay, higher-order phase inside the band, or nonlinear evolution. It is a bounded temporal-span estimate, not a pulse-reconstruction claim.