1. Introduction

Since the first successful demonstration of two-photon excitation fluorescence imaging of bio-tissues [1], a variety of novel nonlinear-optic multiphoton imaging (MPI) technologies have been developed [2]. Many of these technologies need a high-peak-power (over a kilowatt) ultrashort optical-pulse source at the heart of an imaging system to efficiently induce nonlinear-optic effects in bio-specimens. At present, most conventional optical-pulse sources are mode-locked solid-state lasers. However, even the current mode-locked solid-state lasers are large, expensive, and not maintenance-free. This makes it difficult for MPI to become a common technology. To render nonlinear-optic imaging technologies reliable and widely useful, we must be able to provide very stable, compact, and inexpensive optical-pulse sources. Semiconductor laser diodes (LDs) have had great success as reliable, low cost, and high-performance light sources in information technology (IT) applications. However, LDs can also be highly functional light sources in scientific and technological measurement systems. In this paper, we discuss potential applications of mode-locked LDs (MLLDs) in nonlinear bio-imaging.

2. Optical pulse source configuration

In the present optical-pulse sources, MLLDs are the core devices used to generate stable optical pulses of a few picoseconds in duration. Subsequent optical amplification by low-noise pre-amplifiers and low-nonlinear-effects main-amplifiers increases the peak power of the optical pulse to over a kilowatt, and the optical pulses can thus be used for nonlinear MPI purposes. We previously demonstrated the operation of a kilowatt-peak-power optical-pulse source of several picoseconds duration using a gain-switching method on the basis of a combination of a high-speed-response (>15 GHz bandwidth) 1.55 µm LD and a short-pulse electric-pulse generator. The second-harmonic 0.77 µm optical pulses were used successfully for two-photon excitation fluorescence imaging (TPI) of bio-specimens [3].

However, the presently available optical-pulse sources are intended for operation at a wavelength around 1 µm. One major reason for this is that they are used to excite green-fluorescent protein (GFP) in bio-specimens through a two-photon absorption process. GFP is used widely in the field of bio-imaging because of its high expression efficiency [4]. The GFP two-photon absorption peak wavelength is approximately 960 nm [5,6], and the efficient excitation wavelength for TPI is in the range of 900–1000 nm. The basic principle for high-peak-power amplification is the same as in our previous report. Note, however, that MLLDs are very attractive optical-pulse-generating devices because we can easily synchronize optical pulses of a few picoseconds duration from different MLLD oscillators through low-jitter hybrid mode-locking [7]. This enables multicolor pulse excitation of bio-specimens and can induce several different kinds of nonlinear effects useful for bio-imaging. This is a notable advantage of MLLDs compared to other optical-pulse sources, including all-fiber-optic devices.

 

Fig. 1. Schematic configuration of the high-peak-power 980-nm optical-pulse source. The MLLD-laser oscillator is in the area enclosed by the dashed line. BPF: tunable band-pass optical filter; BS-LD: bisection-laser diode; SOA: semiconductor optical amplifier; YDFA: Yb-doped fiber amplifier; LPF: long-wavelength-pass optical filter. The inset shows a SHG autocorrelation waveform for the optical pulses from the MLLD indicating a 3-ps pulse duration assuming a sech² shape.

Download Full Size | PDF

Starting from a 500 MHz MLLD, the optical-pulse repetition rate is decreased to a subharmonic frequency by using a semiconductor optical amplifier (SOA) as the optical gating device. Subsequently, optical-pulse amplification to kilowatt-peak-power level becomes possible by employing optical-fiber amplifiers, as described for the gain-switching LD light source [3]. We have adopted this configuration for MLLDs operating at 980 and 1030 nm. Figure 1 illustrates the configuration for a high-peak-power 980 nm source. It consists of a MLLD for short-pulse generation, an SOA for subharmonic optical gating and pre-amplification, and a two-stage Yb-doped fiber amplifier (YDFA) for main amplification. The 980-nm LD chip used has a separate confinement heterostructure with a GaAs/InGaAs strained-layer double-quantum-well active layer. Characterized by a ridge waveguide structure, the bisection LD has a gain-section length of approximately 500 µm and a saturable-absorber (SA) section length of 30–50 µm. A bilayer antireflection (AR) coating was deposited on the cleaved facet of the gain-section side, resulting in a residual reflectivity of less than 10^-3.

The external-cavity length was set to 300 mm for an optical-pulse-repetition frequency of 500 MHz. For hybrid mode-locking operation, the SA section was biased at −0.2 V for proper SA adjustment. The gain section was driven by a forward direct current of 14 mA, modulated by a 500 MHz microwave signal with an amplitude of 2 V at the 50 Ω termination. An optical isolator was inserted after the MLLD to eliminate any backward-reflected light. Under hybrid mode-locking conditions, the oscillation wavelength was tunable in the range of 965–990 nm with a 2-nm-bandwidth tunable band-pass optical filter (BPF) inside the cavity. The pulse width generated was mostly 2–3 ps, the average output power was about 0.6 mW (0.4–0.6 W peak power), and the timing-jitter was less than 1 ps. An example for the second-harmonic-generation (SHG) intensity autocorrelation waveform is shown in the inset of Fig. 1.

3. Sub-harmonic optical gating and high-peak-power amplification

Optical pulses with a low repetition rate in the megahertz range and a high peak power were generated by subharmonic gating of the SOA inside the fiber amplifier. Our SOA chip had the same 900-µm ridge-waveguide structure as the MLLD chip, but without the SA section. The SOA was driven synchronously by electronic pulses; these pulses were of 1.2 ns duration and had an amplitude of 4.4 V (at the 50 Ω termination) at a frequency of 1–100 MHz. Under these conditions, with an appropriately chosen incident optical-pulse polarization, we confirmed that the leaked (not gated) optical-pulse height was suppressed to less than 10^-3. This high on/off extinction ratio is very important for concentrating the energy stored in the fiber amplifier, to produce the gated optical pulses by saturation amplification and to enable the kilowatt-peak-power optical pulses. The SOA also acted as a 10-dB net-gain preamplifier for the gated optical pulses. Because of losses in the fiber coupling and the optical isolator, the average optical power fed into the first Yb-doped fiber amplifier (YDFA) was approximately 40 µW.

Using the two-stage YDFA that we designed and built, the pulses were amplified after optical gating at the subharmonic repetition rate. With the exception of the Yb-doped fiber (YDF), we used free-space optics rather than fiber optics to reduce nonlinear optical effects due to amplification in the main YDFA. For the first YDFA, the 1.5-m-long YDF was excited by a forward-pumping LD (at a wavelength of 940 nm and a maximum power of 180 mW). For the second YDFA, the same type of YDF was co-excited by two orthogonally polarized LDs (at a wavelength of 915 nm and a maximum power of 180 mW) in a forward-pumping geometry. The Yb³⁺-weight concentration in the YDFs was 900 ppm. To eliminate optical feedback and unintentional laser oscillation, optical isolators were placed in front of the input ports of the first and second YDFAs, and the output end of the second YDFA was angle-cleaved. A 2-nm-wide BPF was inserted between the two YDFAs to reduce amplified spontaneous emission (ASE). To remove the residual pumping-laser light after the second YDFA, we used a long-wavelength-pass (>950 nm) optical filter (LPF). This resulted in a maximum output power of approximately 50 mW for a pulse-repetition rate of 10 MHz. The amplified optical-pulse power ratio was estimated at approximately 75%, with the remainder being due to ASE components. A further decrease in the pulse-repetition rate resulted in an increase in ASE, and thus we used a repetition rate of 10 MHz in the MPI experiment. Although the optical-pulse spectrum was broadened by a factor of 5 due to self-phase modulation inside the main YDFA, the pulse duration was not notably broadened because of a short YDF length, and thus the maximum peak power was kept at above 1 kW.

Except for the design of the MLLD chip and the optical components used, the 1030-nm optical-pulse source was developed using the same concept. However, the wavelength tuning range of the 1030-nm optical-pulse source was approximately 1020–1045 nm, while the 980-nm source was tuned only in the 977–982 nm range, determined by the gain width of the YDF used.

4. Two-photon fluorescence imaging

We examined the feasibility of the 980 and 1030 nm optical-pulse sources for TPI. Although the kilowatt-peak-power optical pulses obtained were successfully applicable to TPI of bio-specimens stained with various fluorescent dyes, very clear TPI for the specimens expressing GFP was not obtained with 1030-nm optical pulses. Therefore, we mainly used 980-nm optical pulses for TPI with GFP fluorescence. Figure 2 shows a TPI example of mouse brain neurons expressing GFP compared to a conventional single-photon confocal microscope image taken with a 488-nm continuous-wave laser. In this experiment, we attempted to image living brain neurons expressing GFP in a mouse (Acc. No. CDB0430T). In these neurons, GFP is fused to protein kinase C (PKC), an enzyme important in the regulation of several neuronal functions [8,9]. The optical transmission efficiency from the output port of the main YDFA to the cover glass of the focus objective (60X, 1.2 NA, 280 µm WD) was 17% at a wavelength of 980 nm. As shown in Fig. 2 (b), single-photon images were sometimes blurred due to the scattering of fluorescence originating from the other branches close to the focal plane. The high spatial resolution of the TPI allowed us to investigate the intracellular three-dimensional structure of the neurons [8,9]. Although nonlinear optical-pulse compression to the femtosecond temporal region is quite possible, this is not necessary for TPI. Our present results clearly indicate that picosecond optical pulses are suitable for practical use without pulse-broadening problems caused by optical dispersion in a microscope.

 

Fig. 2. (a) Two-photon and (b) single-photon excitation fluorescence imaging of the same branch in mouse brain neurons expressing GFP. The two-photon image was taken with 0.4 kW peak-power (at the surface of the specimen) optical pulses at a 10-MHz repetition rate, and was processed using a Kalman image filter. The single-photon image was taken with a 488-nm argon-ion laser.

Download Full Size | PDF

5. Second-harmonic-generation imaging

We subsequently adopted these optical-pulse sources for nonfluorescent MPI. Among various nonfluorescent MPI methods, SHG imaging (SHI) would be most convenient for bio-specimens without using any fluorescent dyes or fluorescence proteins. A result of SHI is shown in Fig. 3 (a) for a specimen of unstained mouse adipose tissue; an auto-fluorescence TPI result is also displayed in Fig. 3 (b) for comparison. The 980-nm optical-pulse source was again used to take these MPI photos. The result shows clear differences between the performance of SHI compared to TPI. The differences can be attributed to the collagen structures inside the cells, which can efficiently produce SHG signals.

 

Fig. 3. (a) SHG imaging and (b) auto-fluorescent TPI of the same region in mouse adipose tissue. Both images were taken with 250-W peak-power (at the surface of the specimen) optical pulses at a 16-MHz repetition rate, and were processed using a Kalman image filter.

Download Full Size | PDF

6. Sum-frequency-generation with two synchronized optical pulse sources

We also demonstrated all-electrically-controlled timing synchronization of the 980 and 1030-nm optical pulses. The synchronization was confirmed through sum-frequency generation (SFG) in a BiB₃O₆ crystal, as shown in Fig. 4, and was subsequently applied to SFG microscopy. Figure 5 indicates a demonstration of SFG imaging. We used a periodically-poled stoichiometric LiTaO₃ (PPSLT) crystal instead of a bio-specimen as the imaging object to exclude unintentional two-photon fluorescence, and employed a narrow-band optical filter (in actual, this is a combination of a pair of optical filters, Semrock FF01-504/12 and Semrock FF01-500/15, giving 9nm transmission bandwidth) to assure SFG signal detection excluding SHG signals. Although SFG imaging may not provide very different information from SHI, the low-jitter synchronized operation of two (and more) different optical-pulse sources can be suitably extended to other nonlinear multiwavelength-excitation imaging technologies, including coherent anti-Stokes Raman scattering (CARS) [10].

 

Fig. 4. Example of all-electrically-controlled timing synchronization of the 980- and 1030-nm optical pulses; prism-resolved photo of SFG and SHG pulses generated in a BiB₃O₆ crystal.

Download Full Size | PDF

 

Fig. 5. SFG imaging for PPSLT crystal surface. The periodic SFG intensity contrast corresponds to the poling periodicity.

Download Full Size | PDF

7. Summary

In summary, we have developed all-electrical timing-controllable optical-pulse sources using external-cavity MLLDs operating at wavelengths around 980 and 1030 nm. For each wavelength, with a combination of a SOA as the optical gating device and a two-stage YDFA as the main amplifier, we obtained kilowatt-peak-power optical pulses at a repetition rate of 10 MHz. Using these optical pulses, we obtained clear TPI for bio-specimens, especially for mouse brain neurons expressing GFP. The feasibility of SHG- and SFG-imaging applications was also demonstrated. Note, however, that if we develop high-power SOAs at the relevant wavelengths, the present optical-pulse source in which fiber amplifiers are used as the main amplifier will be modified to an all-semiconductor laser configuration [11], and thus optical pulses at a variety of wavelengths can be obtained using many sets of MLLDs and SOAs. Therefore, we expect optical-pulse sources using MLLDs to see widespread adoption in practical nonlinear multiphoton microscopy.