2.1.

SC Light Generation

Figure 1 presents a schematic diagram of our SC light pulse source. SC light was generated by a PCF from incident 774-nm light pulses with 1-kW peak power and 5-ps duration. The 774-nm light pulses were the second-harmonic (SH) output of amplified 1548-nm gain-switching semiconductor laser diode (LD) pulses, and the details of this light pulse source are described in Ref. 10. For second-harmonic generation (SHG), we used a 5-mm-long bulk crystal of a periodically poled $\mathrm{MgO}$ -doped ${\mathrm{LiNbO}}_3$ (PPMgLN) device. Previously, we reported the successful application of these SH light pulses with 1-kW peak power for TPI.¹⁰ Here, we used these SH light pulses to generate stable SC light with a maximal bandwidth exceeding one octave. We have already described 300-nm-wide ( $1∕4$ octave) SC light generation,¹¹ and the great improvement reported here is due to an SH conversion efficiency increase to over 50% by further suppressing nonlinear optic spectral broadening inside the fiber amplifier for 1548-nm light pulses.¹²

Fig. 1

Schematic configuration of supercontinuum (SC) light generation utilizing a gain-switched laser diode (LD)-based 1-kW-peak-power light pulse source. PPMgLN: periodically poled $\mathrm{MgO}$ -doped ${\mathrm{LiNbO}}_3$ ; PCF: photonic crystal fiber; BPF: optical band pass filter.

The typical optical spectrum of the SC light generated is shown in Fig. 2 . In this case, we injected 1-kW-peak-power SH pulses into a 50-m-long PCF that had a zero-dispersion wavelength near $800\mathrm{nm}$ .¹¹ Since the coupling efficiency of light pulses to the PCF was approximately 25%, the peak power of the light pulses inside the PCF was $250\mathrm{W}$ . At an intensity $10\mathrm{dB}$ lower than the peak, spectrum broadening exceeding one octave was observed. Although the short-wavelength region in Fig. 2 was limited to $600\mathrm{nm}$ because of the spectrometer measurement spectral region, we also confirmed violet light components in a prism-resolved spectral measurement.

Fig. 2

A typical optical spectrum for supercontinuum light generated from a 50-m-long photonic crystal fiber. Dotted lines indicate the sliced-out wavelengths around 920 and $1030\mathrm{nm}$ .

2.2.

Optical Filtering and Amplification

For the TPI application, we extracted infrared light components using an approximately 10-nm-wide full width at half maximum optical filter because the average optical power must be kept above a few tens of microwatts to amplify the filtered SC light to a power sufficient for TPI using optical fiber amplifiers. Fortunately, the spectral purity of the excitation light pulses is not an important requirement in TPI, which means that it is not necessary to generate Fourier-transform-limited light pulses. Indeed, the measured pulse width was $2\mathrm{ps}$ , which is the same value as that obtained with a 2-nm-wide optical bandpass filter. For 1030-nm light amplification, a laboratory-made Yb-doped fiber amplifier (YDFA) that consisted of a preamplifier and a main amplifier was used, in which the active fiber length was approximately $4\mathrm{m}$ . Except for the Yb-doped gain fibers, all of the optical couplings were built using free-space optics geometry to dispense with coupling fibers in order to avoid the nonlinear optical effects induced in these fibers. The measured maximum output power from the 1030-nm YDFA was approximately $100\mathrm{mW}$ with excitation from a 980-nm LD. The light pulse was broadened to 7-ps duration by fiber dispersion. Based on careful measurement, we found that approximately 20% of the light power accumulated between light pulses as amplified spontaneous emission (ASE) at a 10-MHz repetition frequency. Therefore, when the pulse repetition rate was $10\mathrm{MHz}$ , the peak power of the 7-ps light pulse was estimated to be $1.1\mathrm{kW}$ at average light power of $100\mathrm{mW}$ .

A similar procedure was adopted for generating high-peak-power 920-nm light pulses. In this case, 1-ps, 920-nm SC light pulses were amplified with a custom-made Nd-doped fiber amplifier (NDFA) pumped by an 810-nm LD. Since the active fiber in this NDFA was up to $16\mathrm{m}$ long, the light pulse was broadened to $8\mathrm{ps}$ at the end of the NDFA because of fiber dispersion. Figures 3a and 3b show an SHG intensity autocorrelation trace and an optical spectrum for amplified 920-nm light pulses. Because we found that the ASE is more severe in the NDFA than in the YDFA, we increased the light pulse repetition rate to $50\mathrm{MHz}$ to reduce the influence of the ASE. At this repetition rate, the estimated ASE ratio was 50%, and the peak power of the light pulses was $250\mathrm{W}$ when the average output power was $200\mathrm{mW}$ .

Fig. 3

(a) Second-harmonic generation (SHG) intensity autocorrelation trace for 920-nm light pulses indicating an 8-ps duration ( ${\mathrm{sech}}^2$ shape assumed), and (b) the corresponding optical spectrum.

2.3.

Advantages of Our Light Source

Here, we should clarify some of the advantages of our light pulse source for TPI. Although kW-peak-power ps and fs optical pulses have been used in nonlinear multiphoton bioimaging (MPI), including two- or three-photon excitation fluorescence,^{13, 14} second- or third-harmonic generation (SHG, THG),^{15, 16} and coherent anti-Stokes Raman scattering (CARS),¹⁷ TPI is the most widely used imaging technology. For TPI, mode-locked Ti:sapphire lasers (MLTLs) are generally used ^{14, 15, 16, 17}; however, they are expensive and require a great deal of maintenance. Furthermore, the tunable wavelength range of conventional MLTLs is between 700 and $900\mathrm{nm}$ . Although we demonstrated a TPI with a stable turn-key 774-nm light pulse source, as shown in Fig. 1,¹⁰ a drawback of this light pulse source was that the wavelength was fixed at an SHG of $1548\mathrm{nm}$ . Therefore, we tried to expand the light pulse wavelength so that it was widely selectable by generating SC light. In our light pulse source scheme, necessary spectral components can be extracted, and control of the power level of an optical fiber amplifier can increase the light power to the level necessary for TPI. Note that the maximal LD light power used in our scheme is still limited to a few $100\mathrm{mW}$ , and this low average-power feature will allow a compact light source once we establish the entire design. Furthermore, the light pulses are synchronized to the electronic pulses driving the LD, so the repetition rate is flexibly variable simply by changing the electronic pulse frequency. This feature is useful for obtaining the best quality of TPI pictures, as demonstrated in Ref. 10.

3.1.

TPI of Mouse Kidney Tissues Stained with Alexa Fluo 488

To test the feasibility of our light pulse source, a section of mouse kidney stained with Alexa Fluor 488 (Molecular Probes F24630; Invitrogen, Carlsbad, California) was used as the TPI specimen. The clear high-resolution image of the distal convoluted tubules and collecting tubules in the renal medulla indicated that TPI was successful, using 1030-nm optical pulses from our SC light source. At a 10-MHz repetition rate, the average measured optical power used was $40\mathrm{mW}$ from YDFA, giving an approximately 5-mW average power (70-W peak power) on the sample surface considering the 20% ASE ratio.

3.2.

TPI of Mouse brain Neurons Containing GFP

Next, we attempted to image fluorescent proteins in mouse brain neurons using TPI. Fluorescent proteins, such as green, yellow, and red fluorescent proteins (GFP, YFP, and RFP), are widely used to stain specific molecules in particular cell organelles both in vivo and in vitro.¹⁸ We examined the fluorescence of GFP, which is genetically bound to intracellular protein kinase C (PKC).¹⁹ Although PKC is a key enzyme that modulates the function of neurons in the brain,²⁰ little is known about the subcellular distribution of its activities. Since the two-photon absorption wavelength range of GFP is generally shorter than $1\mu \mathrm{m}$ , we tried using 920-nm light pulses.

Figure 4 shows the TPI of $\gamma$ -type PKC in a mouse Purkinje neuron. At a 50-MHz repetition rate, 160-mW average power (200-W peak power) from the light source, and consequent average power of approximately $16\mathrm{mW}$ (20-W peak power) on the sample surface considering the 50% ASE ratio, we detected GFP signals from the cell body [Fig. 4a] and dendrites [Fig. 4b]. We found that $\gamma$ -type PKC was uniformly distributed in the intracellular space around the cell body. By contrast, the GFP signal in the dendrite was not uniform and was localized as bright spots on the branch. Dendrites are the primary locus of synaptic inputs and their integration, which are responsible for the activities of various ion channels, pumps, and transporters. Since these membrane-associated proteins are highly regulated by intracellular molecules, including PKC, the dendritic localization of the enzyme revealed using TPI suggests that $\gamma$ -type PKC functions at its target proteins. These results clearly demonstrate that our SC light pulse source provides a useful wavelength region for the TPI of fluorescent proteins and redder dyes.

Fig. 4

Two-photon fluorescence images of the GFP bound to $\gamma$ -type PKC in the soma (a) and dendrite (b) of a cerebellar Purkinje cell excited using amplified 920-nm light pulses at a 50-MHz repetition rate and 160-mW average power from the light source (16-mW average power and 20-W peak power on the sample surface considering the 50% ASE ratio). The shape of a Purkinje cell is shown schematically in the inset, and the regions for TPI are indicated by squares in the inset.