Multimodal multiphoton microscope
One femtosecond beam, two contrast mechanisms: second-harmonic generation for non-centrosymmetric structure and two-photon fluorescence for labelled molecules.
Open in the canvas →Background
Multiphoton microscopy trades one photon of high energy for two of low energy arriving at essentially the same instant. Because the probability of that coincidence scales with the square of the instantaneous intensity, it happens at a useful rate only where the light is most concentrated — inside the focal volume — and nowhere else along the beam. Optical sectioning therefore comes for free: no pinhole is needed, because almost nothing outside the focus is excited in the first place. The long excitation wavelength also scatters less in tissue and is absorbed less by it, which is why multiphoton imaging reaches depths that confocal microscopy cannot.
That nonlinearity is also why the source has to be a femtosecond laser. What matters is peak intensity, and for fixed average power the peak scales inversely with pulse duration: compressing 1 ns of energy into 150 fs raises the peak by four orders of magnitude, generating signal at powers a sample can survive. An 80 MHz train delivers a pulse every 12.5 ns, fast enough to dwell on each pixel for many pulses while letting excited states relax in between.
Two different nonlinear processes are available at once, and they report on different things:
- Second-harmonic generation is a coherent, energy-conserving conversion of two photons into one at exactly half the wavelength. It is forbidden in any material with a centre of symmetry, so it appears only where the structure itself is non-centrosymmetric and ordered — collagen, myosin, microtubules, actin-rich structures. No label is involved and no energy is deposited, so it is both endogenous and gentle.
- Two-photon excited fluorescence is an absorptive process: the molecule really is promoted to an excited state and re-emits after relaxing, so the emission is incoherent, Stokes-shifted, and broad. It reports on whatever is labelled — GFP and its relatives are the workhorses.
Around 920 nm is a good compromise between the two: it drives two-photon excitation of GFP efficiently while putting the second harmonic at 460 nm, clear of the fluorescence band and easy to separate with a narrow bandpass.
What this setup demonstrates
A complete beam path from source to detectors: a 920 nm, 150 fs, 80 MHz laser, a half-wave plate and polarizer for power control, alignment mirrors, a galvanometer pair for scanning, and a scan lens and tube lens relaying the scan mirrors to the objective's back pupil.
The sample is a nonlinear specimen carrying two channels, SHG and GFP two-photon fluorescence, and the microscope collects both on either side of it:
- Forward — the second harmonic is collected by the condenser objective, separated from the excitation by a short-pass dichroic, and isolated by a 460/20 nm bandpass onto a PMT. Selecting that PMT shows a signal at exactly 460 nm: half of 920, as second-harmonic generation requires. A camera on the same forward path records the transmitted 920 nm excitation.
- Epi — fluorescence is collected back through the excitation objective, reflected off the short-pass dichroic, and read by a second PMT. This one reports around 507 nm across a 484–530 nm band: GFP emission, not a harmonic of anything. Epi-collection is the practical choice for fluorescence in thick tissue, where scattered emission is recovered but forward transmission is not.
The two channels are distinguishable by their spectra alone, which is the point: one narrow line locked to exactly half the excitation wavelength, one broad Stokes-shifted band that moves with the fluorophore rather than with the laser.
The nonlinear signals are generated by the sample element as configured channels with a set conversion efficiency, not computed from a susceptibility, an intensity, or a focal volume — so doubling the power does not quadruple the SHG here as it would on a bench. The excitation is traced geometrically, so the focal spot is a ray crossing rather than a diffraction-limited volume, and the optical sectioning that defines multiphoton microscopy is implied by the geometry rather than computed. Emission is launched into a fixed set of rays rather than the full 4π of a real fluorophore, and photobleaching, saturation, and the pulse broadening the excitation would really suffer through the objective are not modelled. The galvanometers are shown static; the scan relay is drawn correctly but the image is not formed.