1. Introduction

Two-photon microscopy (TPM) of biological tissues expressing fluorescent proteins has been studied extensively, because TPM can be used to obtain three-dimensional (3-D) images with sub-micrometer spatial resolution. In general, TPM uses femtosecond optical pulses [1–3]; however, by simple consideration of the principle of two-photon absorption and subsequent fluorescence processes, picosecond pulses should have the same capability as femtosecond pulses when the peak power and average power are the same. We have successfully imaged biological specimens, using picosecond optical pulse sources, based on semiconductor laser diodes (LDs) [4–6]. An advantage of this approach is that the picosecond pulses do not suffer from temporal broadening caused by dispersion in the optical components.

Compact, stable, ultrashort optical pulse sources, having a wavelength of ~1 μm, can be used to excite dominant fluorescent proteins present in biological specimens through a two-photon absorption process [2]. Additionally, light at this wavelength is not absorbed by water, oxyhemoglobin, deoxyhemoglobin, or melanin in the tissue [7]. We have developed 1-μm-band picosecond pulse sources, based on mode-locked LDs (MLLDs) using InGaAs quantum-well (QW) laser diodes, and have demonstrated their effectiveness in deep-site TPM of biological tissue [6,8]. MLLDs offer several advantages: optical pulses below several picoseconds in duration can be generated easily, due to the saturable absorber section incorporated into the LD device, and the synchronization of multiple MLLDs is possible using sinusoidal current modulation [6]. However, the precise balance control of gain and saturable absorption is required for stable operation of an MLLD. Additionally, the MLLD cavity should be sufficiently short to ensure long-term stability; thus, the pulse repetition is set to frequencies in the gigahertz to sub-gigahertz range. An optical gating device can also be incorporated to control the optical pulse repetition rate [6]. For TPM applications, the pulse repetition is generally from tens to one hundred megahertz to simultaneously satisfy the high peak power and low average power requirements.

Gain-switched LDs (GSLDs) [9] could potentially be ideal optical pulse sources for TPM, due to their reliability, ease of use, and long-time stability in the optical power and the laser oscillation mode. Moreover, in principle, the GSLDs can generate the optical pulses at flexible pulse repetition rates from single shot to over a gigahertz frequency. However, the temporal width of optical pulses generated from a GSLD is usually in the range of a few tens to several tens of picoseconds; these pulse durations are too long for TPM applications. It is noted, however, that the potential limit of ultrashort optical pulse generation from GSLDs is still not clear, because the detailed physical processes involved in the gain switching (GS) operation of a semiconductor laser have yet to be fully understood. We recently revealed that sub-5-ps optical pulses could be obtained by simply extracting the short-wavelength edge components of the optical output from a gain-switched 1.55-μm distributed feedback LD (DFB-LD), having a 20-GHz modulation frequency [4,5].

An important condition for generating shorter optical pulses from a GSLD is to create a very high carrier density inside the LD active layer via an intense transient excitation. Under this excitation, the optical output spectrum is broadened to cover almost the entire optical gain bandwidth for a cleaved facet LD (having a Fabry–Perot cavity). We have confirmed that a shorter wavelength component shows shorter temporal duration and earlier oscillation [10]; the initial lasing process occurred preferentially in the higher photon-energy region. In contrast, for a DFB-LD, which shows single-mode operation even under strong GS, the spectrum shows very different dynamic behaviors. A remarkable short wavelength shift occurs at the beginning of the GS laser oscillation; this is due to the refractive index decrease by carrier plasma effects under high-density carrier excitation [11,12]. Once the laser oscillation is initiated, the oscillation resonance wavelength shifts to the longer wavelength side, corresponding to a rapid carrier density decrease during GS-pulse generation. Under the conditions described above, Fourier-transform-limited (FTL) sub-5-ps optical pulses were obtained for a 1.55-μm DFB-LD by extracting the short-wavelength spectral components using an optical band pass filter (BPF) [11,12].

In this paper, we describe a newly developed 1064-nm DFB-LD, which operates based on the strong GS used in the previous studies discussed above. We examined two methods for obtaining 7-ps optical pulses by spectral shaping: (i) the extraction of short-wavelength components by the BPF and (ii) compensation of frequency down-chirping in the mid-wavelength portion of the laser oscillation spectrum using a narrow-band chirped fiber Bragg grating (CFBG). We also demonstrated that a combination of the proposed 1064-nm GSLD and the proper optical amplification scheme can provide desirable optical pulse sources for TPM applications.

2. Experimental setup for picosecond pulse generation

We designed and fabricated a 1064-nm DFB-LD, having the high-frequency characteristics of a 13-GHz cut-off frequency. The active layer of the DFB-LD consisted of three InGaAs QWs. The device length was shortened to 250 μm to reduce the photon lifetime. Figure 1 shows a schematic diagram of the experimental setup for generating picosecond optical pulses by GS operation. The temperature of the LD was set to 20, 25, and 30°C by thermoelectric control. This temperature change resulted in a shift in the control laser oscillation wavelength, allowing the flexibility of the device to be tested with wavelength variation. Electrical pulses of 140-ps duration and 8-V amplitude were used to drive the DFB-LD for creating very high carrier densities necessary for short optical pulse generation. The GSLD pulse repetition rate was variable between 1 and 50 MHz in the present experimental setup. We mainly set the repetition rate of the electrical pulses at 10 MHz, a frequency appropriate for TPM applications. A low-power Yb-doped fiber amplifier (YDFA) was used to amplify the optical pulses before spectral filtering. This amplification was used for various optical measurements, including second harmonic generation (SHG) intensity autocorrelation to measure the optical pulse width. The spectrum of the amplified optical pulses was then processed using two different methods. The first method involved the extraction of the short-wavelength edge components using the BPF with sharp-edged filtering characteristics (full-width half-maximum (FWHM): 0.5 nm). The second method used spectral chirp compensation in the mid-wavelength region of the output spectrum. In this approach, a newly designed narrow-band CFBG, with a dispersion of −30 ps/nm and a reflection bandwidth of 0.3 nm, was used. The optical spectra and pulse widths were measured by an optical spectrum analyzer having a resolution bandwidth of 0.01 nm (Advantest Q8384) and an SHG intensity autocorrelator (Oyokoden Femtowave, model 980), respectively.

 

Fig. 1 Schematic diagram of the experimental setup for strong gain switching (GS) of the 1064-nm distributed feedback laser diode (DFB-LD) and subsequent spectral handling.

Download Full Size | PDF

3. Short-pulse generation with the extraction of a short-wavelength edge component

Figure 2(a) shows the optical spectra of the pulses generated from the 1064-nm DFB-LD with strong GS operation by a solid line for three different LD operating temperatures: 20, 25, and 30°C. The peak wavelengths of the optical pulses were 1064.7 nm at 20°C, 1065.0 nm at 25°C, and 1065.4 nm at 30°C. The pulse widths ranged from 14.4 to 15.0 ps. Spectral shaping was applied to the three pulses. The peak spectral components dominated the entire pulse duration, as shown in the previous report [12], and corresponded to quasi-steady state output when the excitation pulse duration was much longer. Therefore, by extracting the short-wavelength components, optical pulses with short duration were obtained.

 

Fig. 2 (a) Optical spectra for the outputs from the gain-switched LD (GSLD) (blue solid line), and transmission curves of the tunable bandpass filter (BPF) (red dotted line) when the LD temperature is tuned to 20, 25, and 30°C. (b) Optical spectra and (c) SHG intensity autocorrelation traces for the optical pulses of short-wavelength edge components, extracted using the BPF.

Download Full Size | PDF

Optical spectra for the outputs from the GSLD and the transmission curves of the tunable BPF are shown in Fig. 2(a) when the LD temperature was tuned to 20, 25, and 30°C. Figures 2(b) and 2(c) show the optical spectra and SHG intensity autocorrelation traces of the pulses after spectral filtering, respectively. For this pulse width measurement, the average optical power was set to 50 μW by controlling the YDFA gain. For all of the cases, the pulse duration was ~7.5 ps (using sech² fitting) and the time-bandwidth product ranged from 0.39 to 0.46. These results demonstrate that nearly FTL 7.5-ps optical pulses were obtained by tuning of the BPF center wavelength; thus, the short-wavelength edge components experienced minimal spectral chirping. However, the present pulse width was longer than that of the 1.55-μm GSLD, which provided sub-5-ps optical pulses [11]. One possible reason for this is that the response frequency of the present 1064-nm DFB-LD was lower than the 20-GHz value for the 1.55-μm DFB-LD.

Figures 3(a) and 3(b) show the SHG intensity autocorrelation traces and central pass wavelength dependence of the optical pulse width for spectral filtering at various wavelengths, respectively. The temperature of the LD was set to 25°C. The pulse width was below 8 ps in the short-wavelength region below 1064.5 nm. However, the pulse width increased as the central wavelength increased above 1064.5 nm. This increase in the pulse width indicates that the pulses undergo nonlinear down-chirping, as was also reported for the 1.55-μm GSLD [12], because the shorter wavelength components of the optical pulses are emitted earlier in the transient lasing process.

 

Fig. 3 (a) SHG intensity autocorrelation traces and (b) central pass wavelength dependence of the optical pulse width for spectral filtering at various wavelengths.

Download Full Size | PDF

4. Short-pulse generation with spectral chirp compensation

In another attempt to shorten the optical pulses generated from the 1064-nm GSLD, we extracted the mid-wavelength portion of the GSLD output spectrum and compensated for linear chirp using a CFBG. As explained in Sec. 3 the mid-wavelength portion has the frequency chirping in contrast to the short-wavelength edge component; therefore, the chirp compensation can result in obtaining optical pulses having temporal durations shorter than the initial value. For this purpose, we designed and fabricated a narrow-band CFBG, optimized for optical pulse generation from the present 1064-nm GSLD.

Figures 4(a) and 4(b) show the optical spectra and SHG intensity autocorrelation traces for the optical pulses reflected from the narrow-band CFBG, respectively, for three different LD temperatures (20, 25, and 30°C). The CFBG was designed to obtain the best performance for the 1064-nm LD temperature of 25°C. The corresponding SHG autocorrelation traces in Fig. 4(b) indicated that linear chirp compensation by the CFBG, having a dispersion of −30 ps/nm, worked well for the mid-portion of the optical spectrum at 25°C. Using a Gaussian-shaped fit, the resulting pulse width was measured at 7.1 ps. The time-bandwidth product of 0.47 indicated that the optical pulses were nearly FTL. In contrast, for LD operating temperatures of 20 and 30°C, the pulse widths were longer than 10 ps. Thus, the CFBG was not effective for the short- and long-wavelength components of the spectrum.

 

Fig. 4 (a) Optical spectra and (b) SHG intensity autocorrelation traces for the optical pulses reflected from the narrow-band chirped fiber Bragg grating (CFBG).

Download Full Size | PDF

Taken together, the results showed that frequency chirping of the optical pulses generated from the strongly driven DFB-LD was quite nonlinear. However, the extraction of the short-wavelength edge components always produced non-chirped short pulses. The mid-wavelength portion of the GSLD output spectrum exhibited a linear down-chirping feature; this suggested that the appropriate combination of band extraction and linear chirp compensation could also produce FTL picosecond pulses.

5. Two-photon microscopy application of amplified 7-ps optical pulses

The present approach of generating picosecond optical pulses from a 1064-nm GSLD was intended for TPM application for yellow fluorescent biological specimens. We performed TPM imaging with optical pulses generated using two different methods explained in Sec. 3 and 4, respectively. We obtained similar signal-to-noise ratios in the imaging for both cases because generated optical pulses have the almost same durations and spectral forms. In the following we showed the TPM images with optical pulses obtained using the CFBG methods. We attempted to amplify 7-ps optical pulses to a high peak power (>20 kW), using a two-stage optical fiber amplifier scheme. For this, we utilized a conventional single-mode YDFA as the pre-amplifier and a large-core double-clad YDFA as the main amplifier; the basic configuration of the present amplifier is given in a previous report [6]. After amplification, the average optical power increased to 1.6 W (at the amplifier exit); thus, the peak power of the 7-ps optical pulses was estimated to be ~23 kW for the 10-MHz repetition rate. Although the optical pulse spectrum was broadened and distorted mainly by self-phase modulation inside the amplifier fibers, we confirmed that the pulse width had not broadened, as expected for the picosecond pulses.

First, we observed paraformaldehyde (PFA)-fixed biological specimens to evaluate our amplified optical pulses as a trial imaging experiment. We performed TPM imaging on PFA-fixed brain slices of H-line mouse expressing enhanced yellow fluorescent protein (eYFP) in the neocortex layer V pyramidal neurons. Our two-photon microscope consisted of an inverted microscope (Olympus IX71) and a laser beam scanning unit (Olympus FV300). Figure 5 shows a picture of two-photon fluorescent imaging of the brain slices of the H-line mouse, obtained with 10-MHz amplified optical pulses. The 800 × 600 pixel picture was obtained from one scan (scan time: 3.8 s). Fine structures of the eYFP-expressing neurons were observed. The optical-pulse repetition rate dependence of the TPM images was also examined at 5, 10, 20, and 50 MHz. A brighter picture was obtained for a lower pulse repetition rate, reflecting the higher peak power of the optical pulses under a constant averaged optical power, as was previously demonstrated for fluorescent beads by using 774-nm optical pulses [4]. It is noted that notable photon bleaching was not observed in these TPM imaging.

 

Fig. 5 Two-photon fluorescence imaging of the enhanced yellow fluorescent protein (eYFP)-expressing neurons in the PFA-fixed brain slices of the H-line mouse.

Download Full Size | PDF

Next, we demonstrated in vivo deep TPM imaging of the H-line mouse brain expressing eYFP using a properly adjusted in vivo TPM setup [8]. We have successfully performed the in vivo imaging at the depth of several hundred micrometers. Figure 6 shows an in vivo TPM imaging at 600 μm depth in the brain of the H-line mouse. It took 8 s for obtaining this picture (512 × 512 pixels). The eYFP-expressing neorcotex layer V pyramidal neurons were clearly observed with high axial resolution and weak background.

 

Fig. 6 In vivo deep two-photon fluorescence imaging of the eYFP-expressing neurons in the brain of the H-line mouse.

Download Full Size | PDF

The present results indicate that the 1-μm wavelength optical pulse source, based on a GSLD, has high potential for TPM applications, comparing favorably with optical sources based on mode-locked lasers including MLLDs. Operation stability and repetition rate flexibility, which arise from precise electronic control, provide additional advantages.

6. Conclusions

We have successfully generated 7-ps optical pulses by intensive GS operation of a 1064-nm DFB-LD, incorporating spectral filtering and chirp compensation. Two methods were examined for shortening the optical pulse duration: the first method extracted the short-wavelength edge components using a BPF; the second method involved frequency chirp compensation around the mid-wavelength portion of the broadened laser oscillation spectrum, using a narrow-band CFBG. Both methods resulted in a pulse duration of ~7 ps. However, in the case of short-wavelength edge extraction by a BPF, we confirmed that tuning of the BPF could follow the central wavelength shift of the DFB-LD for producing FTL optical pulses. This result indicates that the short-wavelength edge extraction method can be a universal scheme for FTL ultrashort optical pulse generation from a gain-switched DFB-LD. The CFBG method can provide better short pulse performance than a BPF, if the chirp compensation and bandwidth selection are properly designed.

After amplification of the 7-ps optical pulses to over 20-kW peak power, we adapted them for TPM of yellow fluorescent biological specimens. Clear two-photon imaging was confirmed for eYFP-expressing neurons in the fixed brain slices of the H-line mouse. Furthermore, a successful experiment was also confirmed for in vivo TPM imaging of the eYFP-expressing neurons in the deep region of the H-line mouse brain. These results indicate that the present GSLD technology can provide very promising picosecond optical pulse sources for TPM including in vivo deep three dimensional two-photon bio-imaging. The effectiveness of the flexible pulse repetition rate in GSLD has also been examined by showing that a lower repetition rate gives brighter images, reflecting the higher peak power of the optical pulses under a constant averaged optical power. Further progress in LD devices and GS technology would enable the generation of optical pulses of a few picoseconds duration in the wavelength region near 1 μm. Therefore, the present results indicate that the 1-μm band picosecond GSLD will be a core device for advanced optical pulse sources for TPM, providing a simple and reliable approach that can be controlled by high-speed electronics.