Multiphoton microscopy

Multiphoton microscopy is a laser-scanning fluorescence technique in which a molecule is excited by absorbing two (or three) photons simultaneously, each carrying half (or a third) of the transition energy. A dye normally excited at 460 nm (2.70 eV) can be excited by two photons at 920 nm (1.35 eV each). Because the probability of two-photon absorption scales with the square of intensity, appreciable excitation occurs only within the focal volume of a high-numerical-aperture objective, roughly 0.3–0.5 µm across and 1–2 µm deep for an objective of NA near 1. Denk, Strickler and Webb demonstrated it in 1990; it is now a standard method for imaging living tissue, especially the brain.

Why femtosecond pulses are needed

The two-photon signal from one fluorophore is proportional to $\langle I^2 \rangle$, the time average of the squared intensity. For a pulse train of repetition rate $f$ and pulse duration $\tau$, the average of $I^2$ exceeds that of continuous light at the same average power by about

$$\frac{\langle I^2 \rangle_\text{pulsed}}{\langle I^2 \rangle_\text{CW}} \approx \frac{1}{f\,\tau}.$$

With $f$ = 80 MHz and $\tau$ = 100 fs the duty cycle is $8 \times 10^{-6}$, so the enhancement is 125,000. At 10 mW average power at the sample, each pulse carries

$$E_p = \frac{P_\text{avg}}{f} = \frac{10\ \text{mW}}{80\ \text{MHz}} = 0.125\ \text{nJ},$$

and the peak power is $E_p/\tau \approx$ 1.25 kW (about 1.1 kW for a sech² pulse shape). Focused to a spot of radius 0.49 µm (0.61λ/NA at 800 nm, NA 1.0), that is a peak intensity near $1.7 \times 10^{11}$ W/cm², against an average intensity near 1.3 MW/cm². This is why a mode-locked Ti:sapphire laser, tunable from about 700 to 1000 nm, became the standard source, now joined by fixed-wavelength ultrafast fiber lasers near 920 nm and 1040 nm.

Optical sectioning without a pinhole

In a confocal microscope the excitation beam bleaches and excites fluorophores throughout the cone above and below the focus, and a pinhole rejects the out-of-focus fluorescence. In a two-photon microscope the total signal from each plane falls off with distance from focus, so the excitation itself is confined and no pinhole is needed. All emitted photons, including those scattered on the way out, can be collected by a large-area detector placed close to the objective (non-descanned detection). This also confines photobleaching to the focal plane.

Depth in tissue

Near-infrared light scatters less than visible light in tissue, and the quadratic dependence means scattered excitation photons contribute little background. Two-photon imaging in mouse cortex commonly reaches several hundred micrometers and, under favorable conditions, close to 1 mm; the limit is set by out-of-focus background generated near the surface when the power is raised to compensate for scattering losses. Three-photon excitation, using sources near 1300 nm and 1700 nm (windows where the combination of scattering and water absorption is low), has a cubic intensity dependence that suppresses this background further. Its effective excitation wavelengths are a third of the source, about 433 nm and 567 nm, and it has imaged structures beneath the cortex in an intact mouse brain. Three-photon systems usually run at lower repetition rates (around 1 MHz) with higher pulse energies, typically from an optical parametric amplifier or a soliton-shifted fiber source.

Label-free contrast

The same pulses generate second-harmonic generation in non-centrosymmetric structures such as collagen fibrils and muscle myosin, and third-harmonic generation at interfaces and refractive-index changes such as lipid droplets. These signals appear at exactly half or a third of the excitation wavelength and need no stain.

Pitfalls

Dispersion in the objective and scan optics stretches a 100 fs pulse and reduces the signal; a prism or chirped-mirror precompensator restores it. Because signal scales with power squared, a 10% drop in power at the sample costs about 19% of the two-photon signal, so laser power drift appears directly in quantitative data. Heating from water absorption and nonlinear photodamage set the usable power ceiling, which falls as imaging depth increases. Two-photon absorption spectra are often broader than and shifted from twice the one-photon peak, so the excitation wavelength is best chosen from published two-photon cross-section data.

Common questions

What is the difference between two-photon and confocal microscopy?

Both produce optically sectioned images by scanning a focused laser. Confocal microscopy excites a whole cone of fluorophores with visible light and uses a pinhole to reject out-of-focus light; two-photon microscopy uses near-infrared femtosecond pulses that excite only at the focus, needs no pinhole, and penetrates scattering tissue more deeply.

Why is the resolution of two-photon microscopy lower?

The excitation wavelength is about twice as long, which enlarges the focal spot. The squared dependence narrows the effective point spread function by about $\sqrt{2}$, which only partly offsets the longer wavelength, so lateral resolution is somewhat worse than confocal imaging with one-photon excitation of the same dye.

What laser is used for two-photon microscopy?

Most systems use a tunable mode-locked Ti:sapphire laser delivering about 100–140 fs pulses at roughly 80 MHz, or a femtosecond fiber laser at a fixed wavelength. Powers at the sample range from a few milliwatts near the surface to tens of milliwatts at depth.

References: W. Denk, J. H. Strickler, W. W. Webb, Science 248, 73 (1990); W. R. Zipfel, R. M. Williams, W. W. Webb, Nature Biotechnology 21, 1369 (2003); F. Helmchen, W. Denk, Nature Methods 2, 932 (2005); N. G. Horton et al., Nature Photonics 7, 205 (2013).