Wavelength combining and separation
Three lasers put on one axis by two dichroic mirrors, then taken apart again by the same two coatings: colour used as a routing label.
Open in the canvas →Background
Two mutually incoherent beams of the same colour and polarization cannot be merged onto one axis without losing power: a 50:50 beamsplitter that overlaps them sends half of each the wrong way. (Two beams locked in phase can be — that is coherent beam combining, and it needs the phase held.) Two beams of different colour can, with no phase to hold, because a mirror can be made to tell them apart. A dichroic mirror is a multilayer interference coating that reflects one wavelength band and transmits another, with almost nothing absorbed[1]. Put it at 45° where a second laser crosses the first, and both leave along the same line.
The trick chains. Each further dichroic transmits everything already on the axis and reflects one more colour onto it, which is how the laser combiners that feed confocal microscopes and flow cytometers are built, and — with gratings or many more, much narrower filters — how wavelength-division multiplexing puts tens of channels into one optical fibre[2]. The same idea scales power as well as channel count: combining many lasers of slightly different wavelength onto one axis raises the brightness beyond what any single emitter gives, which is spectral beam combining[3].
Run backwards, the same coatings separate. A beam carrying several colours meets the dichroics in turn and each one peels off its own band — the detection side of a multicolour fluorescence microscope, or the demultiplexer at the far end of the fibre.
What limits a real system is the edge. A coating's transition from reflecting to transmitting has a finite width, it moves to shorter wavelengths as the angle of incidence grows, and at 45° it sits at different wavelengths for s- and p-polarized light[1]. Colours therefore have to be spaced further apart than the edge is wide, and every extra dichroic adds a little loss and a little leakage to every channel that passes it.
What this setup demonstrates
Three CW lasers — 640, 561 and 488 nm — are combined by two long-pass dichroic mirrors. The 640 nm beam starts on the axis. The first dichroic, with its edge at 600 nm, transmits it and reflects the 561 nm beam arriving from the side; the second, with its edge at 525 nm, transmits both and reflects the 488 nm beam. From there on the three colours are one beam.
A 10% pick-off sends a sample of that shared beam to a spectrometer, whose screen shows the three lines together. The rest meets the same two edges in the opposite order: 525 nm takes the blue off first, 600 nm takes the yellow, and the red continues straight on. Each detector reads 0.9 of one laser, at one wavelength.
Things to try. Select the first separating dichroic and drag its edge below 488 nm: the blue is no longer reflected there, travels on, and arrives at the 561 nm port together with the yellow. Do the same to the dichroic that brings the 488 nm laser in and that laser never joins the axis at all — it goes straight through its own combiner. Change a laser's wavelength across an edge and it changes port.
The dichroic here is an ideal step: everything on one side of the edge is reflected, everything on the other is transmitted, and nothing is lost. The edge does not shift with the angle of incidence or the polarization, and it has no width, so there is no leakage between ports and no minimum channel spacing. A real combiner has all three, and they are usually what sets how many colours it can carry.
Coating absorption, ghost reflections from the uncoated face, and the small sideways displacement a beam picks up crossing a tilted plate are not modelled either. The combined colours are drawn as one beam but stay incoherent with each other: this is power routing, not interference.
Related components
References
- RP Photonics Encyclopedia — Dichroic Mirrors
- Wikipedia — Wavelength-division multiplexing
- T. Y. Fan, “Laser beam combining for high-power, high-radiance sources,” IEEE Journal of Selected Topics in Quantum Electronics 11, 567–577 (2005)