top of page

Laser sources for 2- and 3-photon excitation

Contributed by Marco Arrigoni, Director of Marketing at Coherent.

Laser source requirements for 2-photon microscopy

Two-photon excitation of biological samples requires short pulses, duration < 1 ps, high repetition rates, average powers compatible with sample viability, and wavelengths that match the excitation spectra of a many different probes.

The pulse duration requirement originates from the non-linearity of two-photon excitation, a process with a low probability that requires a large flux of photons to compensate for the small excitation cross section. Pulse durations commonly used are between 75 and 250 fs, where shorter pulses result in more efficient excitation but with a higher risk of non-linear sample or even ablative damage.

Such short pulses can be produced only by operating a laser in the mode-locked regime, where many longitudinal modes of the laser are locked in phase, resulting in a train of ultra-short pulses emitted at a repetition rate equal to c/2L where c is the speed of light and L is the length of the laser resonator. For practical considerations, most mode-locked lasers operate at fixed repetition rates of 50-100 MHz, that correspond to resonator lengths of 1.5 to 3 meters. Lowering the repetition rate would require resonator lengths of many meters, impractical with conventional lasers based on discrete optics.

These repetition rates satisfy the usual speed requirement of laser scanning microscopy. For example, imaging a sample with 512 x 512 pixels at a refresh rate of 4 Hz with at least one laser pulse per pixel would require a laser repetition rate of about 1 MHz.  Keeping into consideration also the “dead times” of the scanners as they cover an area slightly larger than the desired image, “blanking” at the edges and the need to have multiple excitation pulses in each pixel to decrease noise, it becomes apparent that repetition rates > 10 MHz are advantageous, even more so when using faster resonant scanning.

The most commonly used indicators are excited in the single-photon regime with wavelengths between 400 nm and 550 nm (violet to green light). In the two-photon regime this corresponds roughly to wavelengths between 750 nm and 1,100 nm (near infrared), that happen to be a perfect match for tunable lasers based on a Titanium-sapphire active medium.

Finally, it has been shown in many different studies that most samples tolerate average laser powers of 100-200 mW at these near infrared wavelengths, a power level that enables bright images with a large variety of imaging and functional fluorescent probes.


Combining all these parameters, the recipe for the ideal laser source for two-photon excitation could be summarized as:

  • 700-1,100 nm tuning range

  • 1-4 Watts of average power (considering that some microscopes may have a throughput as low as 10%)

  • 50-100 MHz repetition rate

  • 75-150 fs pulses


Titanium Sapphire (TiS) lasers, first introduced in the early nineties and substantially improved in the early to mid-2,000s with the introduction of single-box automated devices (Chameleon by Coherent and Mai Tai by Spectra-Physics), contributed to the rapid and exponential growth of two-photon microscopy. It is estimated that there are over 5,000 of these TiS lasers used for two-photon microscopy in laboratories across the globe.

Laser source requirements for 3-photon microscopy

3-photon microscopy requires a substantial paradigm shift in the laser specification. First, the much lower probability of a 3-photon event requires a considerably higher peak power. This can be achieved either reducing the pulse duration or increasing the energy content of each pulse. In practice, very short pulses (10-50 fs) are very difficult to manage as they propagate through optical materials - or simply air - because they undergo group velocity dispersion (GVD), that is more difficult to compensate when the pulses are < 50 fs. In addition, shorter pulses require a larger laser bandwidth to support them. Eventually the bandwidth of the pulse will become wider than the excitation spectrum of the probe, resulting in a loss in efficiency.

These considerations lead to the use of 40-50 fs pulses for 3-photon excitation, a good balance between high peak power and reasonably simple pulse management. These short pulses are a factor 2-4 shorter than the pulse duration commonly used for two-photon microscopy, and therefore are too long to fully compensate for the much smaller three-photon excitation cross section. It is therefore necessary to increase the peak power by scaling up the energy of each pulse. The energy per pulse typical of a laser for two-photon microscopy is 20-40 nJ. Increasing this energy by a factor 40-50 means that the pulses for 3-photon microscopy should be 1-2 mJ at the laser output, leading to 100-400 nJ on the sample, considering all the microscope losses.

Pump Lasers for 3-Photon Microscopy

Using the well-validated assumption (in the two-photon regime) that the average power on the sample should be kept below 400 mW to ensure sample viability, and assuming a maximum of 400 nJ per pulse, the laser repetition rate should be ~1 MHz. For faster imaging, the repetition rate may be increased to 4-5 MHz if the energy per pulse on the sample is proportionally decreased.

We will address wavelength considerations later, but the main point here is that - until recently – designing a laser able to produce several mJ of energy per pulse at 1 MHz repetition rate was a major challenge. In fact, a TiS laser with a 1 MHz repetition rate would require a cavity length of tens of meters and, to make things worse, producing short pulses at microjoule energy levels cannot be achieved with conventional mode-locking because the resulting peak power is above the damage limit of the TiS medium. To circumvent this problem, Gerard Mourou and Donna Strickland invented chirped-pulse amplification (CPA) in the 1990s and – as a result - were awarded the 2018 Nobel prize for Physics. In the CPA scheme, the pulses produced by a mode-locked laser are down-selected, then stretched to tens of picoseconds of duration, amplified to maintain peak power well below the damage threshold and finally re-compressed to the original femtosecond pulse duration, but with much higher peak power.

Other considerations lead to the conclusion that it’s technically difficult (and expensive) to build a TiS laser that satisfies the energy and repetition rate required by three-photon excitation. Fortunately, a fiber laser based on Ytterbium active medium (Yb laser) can bypass these problems. This is because the use of an active fiber distributes the amplification of the pulse over a longer distance (the fiber can be coiled to result in a compact configuration) providing sufficient gain also at higher repetition rates that are unachievable with a TiS amplifier based on chirped-pulse amplification.

Fig 1 shows the typical architecture of such a fiber amplifier. A 50 MHz diode-pumped “seed” fiber laser produces a few milliwatts of power, an integrated acousto-optical modulator (AOM) “gates” the pulses dividing the repetition rate by N (=1,2,3…) resulting in a train of pulses at 50, 50/2, 50/3 MHz. These pulses are amplified in successive diode-pumped fiber stages of the Yb fiber amplifiers. After that they get “stretched” (chirped) to hundreds of picoseconds to avoid damage to the active fiber. At the output of these amplification stages, the pulses are recompressed with a grating in a free-space configuration.

Figure 1. Block diagram of the main elements of a high-power Ytterbium fiber laser


This architecture is extremely flexible and powerful as it can produce average powers of 40-60 watts supported at any repetition rate from 1 MHz to 50 MHz. The energy per pulse is limited by the onset of damage to the optical fiber, typically around 100 mJ per pulse.

Yb lasers have become very popular for many industrial processes requiring femtosecond pulses at 1 micron wavelength or its green and UV harmonics. In life sciences, they are popular for 2-photon optogenetics with red-shifted opsins. In most cases the stimulation (or silencing) of large populations of neurons is accomplished using one or two spatial light modulators (SLMs), where the modulators split the incoming powerful beam into many “beamlets”, each one targeting one neuron in the three-dimensional sample volume.

Tunable sources for 3-photon microscopy

Yb lasers operate at a fixed wavelength of 1030-1070 nm and can support pulses of >250 fs. Moving from 2- to 3-photon excitation, the wavelength required to excite probes based on green fluorescent protein changes from ~920 nm to ~1300 nm, for red-shifted probes, it changes from ~1100 nm to ~1700 nm. As we noted before, efficient 3-photon excitation requires pulses of 40-60 fs. Yb lasers are very flexible with respect to repetition rate but are unable to produce sufficiently short pulses at these excitation wavelengths and so they are used to pump an optical parametric amplifier (OPA), a non-linear device that can produce pulses meeting all the requirements of 3-photon excitation.

In an OPA, a beta-barium borate (BBO) crystal is pumped by laser light and converts the pump wavelength into two wavelengths – both longer than the pump wavelength – that satisfy conservation of energy and momentum of the pump pulses.  This can be expressed as:


where l denotes the wavelength and p, s and i identify the pump wavelength and the two wavelengths produced by the OPA that are conventionally called signal and idler wavelengths. For historical reasons, the shorter of the two wavelengths is called signal wavelength while the longer is the idler wavelength.

When ls=li (and therefore equal to 2lp) the process is called degenerate, which is the inverse process of second harmonic generation and governed by the same laws of physics. Usually OPAs are pumped by the second harmonic of the Yb laser output (usually via an LBO crystal), that is, with green light at 515-520 nm.  In the degenerate case ls=li=2lp~1030 nm. This means that the signal wavelength can in principle be adjusted between 1030 nm and a shorter wavelength, longer than the 515 nm limit. As the signal wavelength becomes shorter, the idler wavelength must become longer and eventually will not be supported by the optical transmission window of the crystal. To select li=1300 nm, energy conservation requires the signal wavelength be ls= 860 nm.

Fig.2 Block diagram of the

main elements of an OPA.


​​​​​The configuration of a typical OPA is shown in Fig 2. A small part of the pump beam is used to generate a white light continuum in a thin plate through a combination of non-linear processes including Raman shifting and four-wave mixing. This light has, as the name implies, a very broad spectrum that can be used to initiate or “seed” the parametric generation process in the crystal. Most of the pump power from the Yb laser is used to pump one or more crystals that amplify the desired wavelength of this white light seed.

Given the white light nature of the continuum, how does the OPA select the two output wavelengths? This is achieved by phase-matching, i.e. adjusting the angle between the optical axis of the crystal and the pump beam. By tilting the crystal, it is possible to select which pair of wavelengths from the white light will be amplified, as far as they satisfy the previously stated energy and momentum conservation. OPAs have a manual or motorized and computer-controlled stage that sets the crystal angle to produce the signal wavelength of choice.

The specific design of the OPA enables the production of pulses that are much shorter than the pump pulse, by taking advantage of a “non-collinear” design, where the pump beam and the signal beam propagate at different angles in the crystal. The non-collinear propagation mechanism is such that a larger bandwidth of the signal wavelength can be amplified, resulting in pulses substantially shorter than the pump pulses. Pulses as short as 20-25 fs can be produced in an OPA (and shorter than 10 fs in an OPCPA). The non-collinear design is necessary to achieve these pulse durations when using a 200-350 fs pump but leaves a ~300 nm gap in the tuning range, typically between 940 and 1250 nm. Fortunately, this region is not relevant for three-photon excitation.

Conversion of the pump beam from the Yb laser to its second harmonic is typically achieved with a 40-50% efficiency, however, the subsequent parametric generation is less efficient, resulting in total conversion efficiency from laser to the two combined OPA wavelengths of 10-12%. To make things worse, since the idler beam is at a longer wavelength than the signal beam, this already low efficiency is unevenly split between signal and idler because of the different wavelengths (photon energies). Therefore, the idler may be as low as 3-4% of the pump beam at 1035 nm and at longer wavelengths, like 1700 nm, the efficiency decreases even further. For this reason, it is important that the pump laser provides a few tens of mJ per pulse at the desired OPA repetition rate. In fact, parametric generation is a non-linear process that requires a minimum energy to overcome noise and reach an acceptable efficiency. Once the minimum input energy has been achieved, generally 10-15 uJ, the performance of the OPA scales linearly with an increase in pump energy until the optical aperture becomes the limiting factor.. For this reason, the pump requirements for a 1 MHz OPA (usually 20-40 Watts) are different from the pump requirements for a 4 MHz OPA (50-60 watt). The high pulse energies of Yb lasers are critical in this regard, with a few tens of mJ per pulse Yb pump yielding just a few mJ per pulse at 1300 or 1700 nm. Fig. 3 shows a typical tuning curve for an OPA pumped with 60 Watts at 1 and 4 MHz.


Figure 3. Typical power vs wavelength

for a Coherent Opera-F OPA pumped

with 60 W (4MHz) and 40 W (1 MHz)

The pulses at the output of the OPA may be chirped negatively or positively, depending on wavelength and beam output (signal or idler). This means that the pulse duration is longer than the pulse duration supported by the bandwidth of the pulse. In addition to this chirp, the optics in the microscope will add positive group velocity dispersion resulting in an even longer pulse. If the chirp is negative, any optics in the path between the OPA and the sample may reduce the (negative) chirp or even result in a positive chirp.  In most cases, there will be a residual chirp that should be compensated by an external compressor. This device uses either a pair of prisms or so-called negative dispersion mirrors (NDMs). The compressor-induced chirp should be adjustable as the total chirp from the OPA to the sample is generally a function of the wavelength and the optics, especially modulators and objective lenses.

Considering the usual losses of a microscope system and the conversion efficiency of an OPA in the near infrared, it is necessary to use a pump laser with at least 40 Watts of output power to get about 1 Watt at 1,300 nm. With the typical losses of a multiphoton microscope, the power on the sample will likely be between 100 and 300 mW, compatible with damage-free imaging. Pulses optimized for durations of 50-70 fs will produce the necessary three-photon excitation.

Alternatives to OPAs

An OPA pumped in the green spectrum by an Yb laser is the most popular approach to 3-photon excitation, but several approaches have been demonstrated in the literature. Alternatives include

(1) an optically pumped chirp pulse amplifier (OPCPA) pumped by a Yb laser. Conceptually similar to an OPA, an OPCPA is usually optimized to produce even shorter pulses - as short as < 10 fs. The White Dwarf OPCPA-based 3-photon laser is available from Class5 Photonics.

(2) a Yb laser-pumped OPCPA can also be used to simultaneously generate 1700 and 2600 nm signal and idler beams, with frequency doubling of the idler beam yielding simultaneous beams at 1300 and 1700 nm (Guesmi et al., 2018).

(3) other non-linear processes in fibers, such as self-soliton frequency shifting that transforms the output wavelength of an Yb or Er amplifier, producing a beam at ~1700 nm (Horton et al., 2013).

bottom of page