1. Introduction

After the first demonstration of the two-photon microscope (TPM) by Denk, Strickler, and Webb in 1990 [1], TPMs have been extensively used in noninvasive biomedical sectioning imaging for live cells and tissues. Even though many different TPM systems were introduced and have been continuously improved over the last 10 years [2–5], most of them are based on femtosecond Ti:Sapphire lasers utilizing their superb performances such as high peak power, high pulse energy, and great stability. Lately, several attempts have been reported on developing new TPMs with compact and portable light sources. For example, two-photon imaging of actin filaments in PtK2 cells labeled with Alexa Fluor 488 was reported with a gain-switched laser source [6], where a frequency doubled semiconductor laser produced a kilowatt peak power at a 770 nm wavelength with a pulse width of 5 ps. Another TPM with a gain-switched semiconductor laser was demonstrated at a 980 nm wavelength with 3.5 ps pulse width and a kilowatt-level peak power for imaging Glomeruli and convoluted tubules from a mouse kidney section [7]. A fiber laser is an alternative compact light source for a TPM. A frequency doubled mode-locked Er³⁺ doped fiber laser generating ultrashort pulses at a 810 nm wavelength with a pulse width < 150 fs, and a 10 kilowatt-level peak power was used for two-photon fluorescence correlation spectroscopy (FCS) [8]. A mode-locked fiber laser combined with a photonic crystal fiber (PCF) as a nonlinear wavelength shifting element was presented recently as a tunable excitation light source for a TPM [9]. A practical multiphoton imaging system based on a commercially available mode-locked fiber laser operating at a 1030 nm wavelength was demonstrated lately [10].

An optical source with a short pulse width is desirable in a TPM to enhance the two-photon efficiency and suppress thermal damage in a biological sample. A recently available Yb³⁺ doped fiber (YDF) is an excellent gain medium for a short pulse fiber laser source due to its large gain and broad gain bandwidth. Moreover, the long emission wavelength of an YDF over 1 µm guarantees less scattering and higher penetration depths than a conventional Ti:Sapphire laser source in tissue imaging [11]. It also reduces the autofluorescence background. Therefore, a compact and robust laser source based on an YDF combined with fiber optic light delivery is an attractive option for a next-generation TPM and its endoscopic applications. However, the circulation of femtosecond optical pulses in a fiber laser and the delivery of short pulses through an optical fiber are frequently accompanied by unwanted nonlinear optical side effects such as a self-phase-modulation or Raman frequency shift. Because of these nonlinear effects in an optical fiber, the output pulse energy of a mode-locked fiber laser source is normally less than a few tens of picojoules in the positive dispersion region of an optical fiber at around a 1550 nm wavelength [12]. A mode-locked pulse with higher pulse energy can be obtained with a highly dispersive fiber cavity design at the negative dispersion region of an optical fiber near the 1060 nm wavelength using a chirped-pulse amplification scheme. Recently, a fiber laser with all-normal-dispersion fibers was reported to produce mode-locked optical pulses with a large pulse energy of 20 nJ, which is comparable to those of Ti:sapphire lasers [13]. Pulse broadening in an optical fiber used in light delivery or a thick objective lens in an imaging system is another important factor that needs to be precisely controlled for the realization of a future TPM with a compact femtosecond fiber laser source.

In order to overcome these problems, we present a TPM system based on a mode-locked fiber laser with variable pulse control. A homemade mode-locked fiber ring laser was built with an YDF as a gain medium at a 1060 nm wavelength. The fiber laser consisted of all-normal group velocity dispersion (GVD) fibers without any dispersion compensating components in the laser cavity. The laser source was equipped with an external grating pair pulse compressor for continuous control of frequency chirping and for the pulse width of the output pulses. We demonstrate that the optimum condition for nonlinear excitation of a sample could be achieved by monitoring the photocurrent of a GaAsP photodiode (PD) placed at the sample position of a TPM system.

2. Mode-locked fiber laser with variable pulse width control

The schematic diagram of the TPM system with a continuous pulse width control is shown in Fig. 1 . It consists of three parts: (a) a mode-locked fiber laser, (b) an external pulse compressor, and (c) a laser scanning microscopy setup. The gain medium of the laser was a 30-cm-long YDF. The small-signal absorption coefficient of the YDF was 1200 dB/m at a 976 nm wavelength (Yb1200-4/125, LIEKKI). The ring cavity of the laser was made of a wavelength division multiplexer (FWDM-9803-N-B-1, AFR), an optical isolator (PSSI-06-P-N-B-1, AFR), a polarization controller, and a single mode fiber (HI1060 specialty fiber, Corning). Total cavity length was estimated to be 5.5 m. The YDF was pumped by a 200-mW pump laser (P161-600-980A, EM4 Inc.) at a 980 nm wavelength. With the proper positioning of the polarization controller, self-starting mode-locked pulses at a 1060 nm wavelength were obtained by nonlinear polarization rotation [14]. Because the average output power of the laser was 50 mW with a 36 MHz repetition rate, we had a relatively large pulse energy at 1.4 nJ. The output pulse of the laser was positively chirped with a long pulse width of 13 ps, which is due to the design of the laser cavity with highly dispersive all-normal GVD fibers. The mode-locked pulses were amplified by an external YDF amplifier (YDFA) also consisting of all-normal GVD fibers. The average optical power of the laser was amplified to 300 mW by the YDFA without much nonlinear spectral broadening through this optical amplification process due to the advantage of this chirped pulse amplification process [15].

 

Fig. 1 Schematic diagrams of our custom built two-photon microscope system. (a) a passively mode-locked fiber ring laser producing 36 MHz pulses with an external Yb³⁺ doped fiber amplifier, (b) a variable double-pass grating-pair pulse compressor, (c) a confocal two-photon microscope with a point-scanning galvanometer system. LD: laser diode, WDM: wavelength division multiplexer, YDF: Yb³⁺ doped fiber, YDFA: Yb³⁺ doped fiber amplifier, PBS: polarization beam splitter, PMT: photomultiplier tube.

Download Full Size | PDF

The pulses undergo large normal dispersion and thus need external compensation to obtain femtosecond pulses. We used a double-pass grating pair pulse compressor to compensate for the large frequency chirping in the amplified laser pulses [16]. By adjusting the distance between the two gratings, the pulse width of the amplified excitation laser pulse was compressed from 13 ps to about 385 fs while the optical power was reduced from 300 mW to about 80 mW after passing through the double-pass pulse compressor made of two identical diffraction gratings with 600 grooves per millimeter (20RG600-1000-2, Newport).

Figure 1(b) shows the schematic diagram of a pulse compressor. α and β are the angles of the incident and diffracted beams from the surface normal of the first grating along the clockwise direction. α is negative, while β is positive as a sign convention for the angles. G1 and G2 are two identical diffraction gratings; L_sep is the perpendicular distance between the two gratings. The grating equation can be expressed as

Figure 2(a) and 2(b) show the optical spectrum and the intensity autocorrelation traces of the amplified pulses before and after the grating-pair pulse compressor. The 3-dB spectral bandwidth was 14 nm. The full width at half maximum (FWHM) of the autocorrelation trace after the compression was 385 fs. Assuming a Gaussian pulse, the 3-dB pulse width of the compressed laser source was estimated as 270 fs. The ideal transform limited pulse width corresponding to this spectral bandwidth was 120 fs for a Gaussian pulse. The compressed pulse width exceeded the Fourier-transform limited pulse width, and we believe that this was due to higher order dispersion from the optical fibers. The compressed laser source delivered up to an average power of 80 mW at a repetition rate of 36 MHz. The corresponding energy for each pulse was 2.2 nJ with a peak power of 7 kW.

 

Fig. 2 (a) Spectrum and (b) autocorrelation traces of our fiber laser source before and after pulse compression.

Download Full Size | PDF

3. Two-photon laser scanning microscope system

Amplified and compressed mode-locked fiber laser output pulses were used for the excitation laser source of a custom-built two-photon laser scanning microscope setup illustrated in Fig. 1(c). Our laser scanning microscope (LSM) consisted of five major components: a scan head, two relay lenses (a scan lens and a tube lens), dichroic mirror, an objective lens, and a photomultiplier tube (PMT) as a detector. One of the most important considerations in building a custom-built LSM is an efficient beam path design for both the excitation and collection of lights. Our LSM is based on an infinity-corrected optical system with a scanning head (6215HSM40B, Cambridge Technology Inc.) which consists of two closely spaced galvanometric scanning mirrors. The scanning head is controlled by a driver board (PositionPro2, Cambridge Technology Inc.) which is synchronized with an image acquisition system. The back focal plane of the objective lens and the center of these two galvanometric mirrors are approximately positioned at the telecentric planes connected with two relay lenses: the scan lens (f = 50 mm, NT49-356-INK, Edmund Optics) and the tube lens (f = 150 mm, NT49-362-INK, Edmund Optics) [17, 18]. Focus control along the z-axis of the LSM was accomplished with a computer controlled stepper motor (DRV001, Thorlabs). The fluorescent light from a specimen was collected by the same objective lens. The fluorescent light was separated from reflected excitation light by a dichroic mirror (900dcsp-laser, Croma) and an IR filter (FF01-750/SP, SEMROK), which were placed in front of the PMT (R-3896, Hamamatsu). The dichroic mirror reflects the excitation laser at a 1060 nm wavelength and transmits light between 400 and 800 nm wavelengths. The distance from the rear aperture of the objective lens to the cathode of the PMT was only about 40 mm. This distance needs to be as small as possible in order to maximize the detection efficiency for the ballistic or scattered fluorescence.

We used a frame grabber (PCI-1409, National Instrument) for high-speed 2D image acquisition. The PMT signal does not include a dark reference level which is required to determine digitized signal intensities in a frame grabber. This dark reference level known as the clamping interval should be identical in each horizontal line during the image acquisition period. We have previously introduced a simple masked illumination method as an effective scheme to obtain a consistent clamping level and used it for this LSM system [19]. Image acquisition, display, and laser beam scanning were all controlled with a custom-built LabVIEW program that can be easily modified and adjusted for various imaging applications. Our scanning system was designed to acquire an image of 400 × 400 pixels. The pixel duration time was 6.25 µs, and the frame rate was 1 frame per second with unidirectional scanning. To obtain the spatial resolution of the imaging system, we used fluorescent beads 200 nm in diameter (F-8810, 580/605 nm, Molecular Probes) as point-like objects. A two-photon image of spin-coated fluorescent beads was obtained with a 60 × objective lens (60 × /1.2W, UplanSAop, Olympus) with a 0.065 µm pixel spacing along the lateral direction and a 0.2 µm step size along the axial direction. The obtained fluorescence intensity profiles of a single bead were fit to Gaussian profiles. As shown in Fig. 3 , the FWHMs of the fitted intensity profiles along the lateral and axial directions were 0.44 µm and 1.34 µm, respectively. These values are comparable to the theoretical estimates of the resolution based on NA, the emission wavelength, and the refractive index of water (e.g. R_lateral = 0.7 λ_em /NA = 0.35 µm; R_aixal = 2.3 λ_em n/(NA²) = 1.24 µm).

 

Fig. 3 (a) Lateral and (b) axial resolution. The intensity profiles of measured fluorescence intensities were fitted with Gaussian functions to obtain the FWHMs of measured intensities to obtain the lateral and the axial resolutions.

Download Full Size | PDF

4. Pulse optimization procedure monitored with a GaAsP photodiode

In order to maximize the efficiency of two-photon excitation (TPE), the pulse width of an excitation laser should be minimal at a specimen. The pulse width of an excitation laser can be altered in a microscope system due to the GVD of any optical components, such as a fiber optic delivery line, a scan lens or a tube lens, a beam splitter, dichroic filter, and an objective lens [20]. Since objective lenses and spectral filters are frequently changed during measurements, there should be a proper pulse optimization scheme to monitor and adjust the GVD of a microscope system. Because we used a grating pair pulse compressor in front of the TPM, the frequency chirping of the input laser pulse can be continuously adjusted with this pulse compressor such that the frequency chirping induced by the GVD of the TPM can be properly canceled. This optimum pulse condition would produce the maximum peak pulse intensity and would produce the maximum two-photon absorption (TPA) or excitation in a nonlinear medium placed at the sample position of our TPM. We used a GaAsP semiconductor photodiode (G1116, Hamamatsu). TPA in a semiconductor photodiode (PD) can be observed when the incoming photon energy is higher than the half of the bandgap energy of the PD and is lower than the bandgap energy of the PD [21]. The maximum detectable wavelength of the PD by single-photon absorption was about 680 nm and is suitable for the monitoring of TPA with a 1060 nm wavelength light in our TPM setup [22].

The GaAsP PD was placed in the sample position for our TPM just after the objective lens (10 × /0.25dry, Plan N, Olympus), and the induced photocurrent was measured for pulse and continuous wave (CW) operation as a function of the excitation laser power. Figure 4(a) shows the TPA response of the PD for two different spot sizes of the incoming pulsed laser beam: 2.5 μm and 250 μm. Photocurrents produced by beams with 2.5 and 250 μm spot sizes are plotted with green solid triangles and red solid circles, respectively. Photocurrents produced by the 2.5 μm spot were 4 or 5 orders of magnitude larger than those produced by the 250 μm spot. This clearly shows that the photocurrent produced by the PD for the 2.5 μm spot size was mostly due to two-photon absorption instead of single-photon absorption. In order to verify the effect of single-photon absorption due to material defects or band tail, we repeated the same measurement for the photocurrent as a function of the excitation laser power using a CW laser output with a 250 μm spot size. The black solid squares in Fig. 4(a) show the photocurrent produced by the CW laser beam. Since the peak intensity of the CW laser was lower than that of the pulsed laser by 5 orders of magnitude, the TPA can be neglected in this case. This linear relation between the photocurrent of the PD and the input power of the CW laser is mostly due to single-photon absorption. As expected from the difference between single-photon absorption and TPA, the slope of the black line is about the half of the red line [21]. The black solid squares connected with black lines in Fig. 4(a) can be considered as the background photocurrent signal by single-photon absorption. Figure 4(a) shows that the two-photon signals were well above this background signal even when the power of the pulsed laser was as small as 1 mW.

 

Fig. 4 (a) PD signal as a function of the excitation power for CW and pulse operation. The squares are for a spot size of 250 μm in CW excitation and the circles (triangles) are for a spot size of 250 μm (2.5 μm) in pulsed excitation. (b) Normalized PD signals as a function of the grating separation length. Solid lines are least-square fitting curves with quadratic functions.

Download Full Size | PDF

Figure 4(b) shows the TPA efficiency monitored by the GaAsP PD as a function of the distance between the two gratings used for the pulse compression in our TPM system. The monitoring PD was placed at the beam waist of the incoming laser light. We tested three different lenses: a 10 × objective lens, and two 60 × objective lenses (60 × /0.9W, LUMPlanFl/IR, Olympus and 60 × /1.2W). The TPA curves for these four different focusing elements were normalized and fitted with quadratic functions to find the optimum grating distance for each of these lenses. The optimum distances corresponding to 10 × /0.25dry, 60 × /0.9W, and 60 × /1.2W objective lenses were 18.58 mm, 18.82 mm, and 18.88 mm, respectively. Since the changing ratio of the GDD with respect to the perpendicular distance L_sep of the two gratings was 11,260 fs²/mm for our grating pulse compressor, the GDD change corresponding to 0.3 mm variation in the L_sep becomes 3378 fs². Pulse broadening associated with this GDD change would be about 13 fs for a Gaussian pulse with 14 nm spectral width at a 1060 nm wavelength. This is about 5% pulse broadening with respect to the optimum pulse width of 270 fs. This small variation in pulse width is because the GVD is very low at a 1060 nm wavelength and has a rather large pulse width. However, it should be noted that pulse broadening would be sharply increased when the center wavelength and the pulse width become smaller and the spectral width becomes larger. For example, when the Coherent Chameleon Ultra laser system generating 140 fs pulses with 20 nm spectral width at a 1060 nm wavelength was used, pulse broadening increased to 14% for the same case. Moreover, if the center wavelength is tuned to 700 nm, pulse broadening increased up to 30% because the GDD for common glass materials at a 700 nm wavelength becomes twice as large as the 1000 nm wavelength [23].

By measuring the photocurrent induced by the TPA in incorporating a GaAsP PD, we obtained the optimum prechirping condition for various different objective lenses in our TPM system. This is a very simple and sensitive procedure which can be performed just before an actual TPM measurement. Since a semiconductor photodiode is immune to photobleaching and is sensitive to a wide spectral range, it can be applied to many other TPM systems for pulse optimization. We expect that this simple process would be particularly helpful in observing vulnerable bio samples in long-term live cell observations or in vivo imaging.

5. Two-photon imaging at a 1060 nm wavelength

In order to demonstrate the 3D two-photon imaging with our prechirped pulses at a 1060 nm wavelength, we obtained sectioned imagines of nerve cells. We used the masked illumination method that we have proposed before [19]. Figure 5 shows 2D and 3D two-photon images of Vybrant DiI-stained dorsal root ganglion (DRG) cells for axonal tracing of neuronal pathways using anterograde and retrograde transport. Vybrant DiI (V22885, 549/565 nm, Molecular Probes) has been used for anterograde and retrograde neuronal tracers. The sample preparation for retrograde tracing from the axon terminal to the cell body was as follows: First, a right distal sciatic nerve of a 5-week-old Spraque-Dawely (SD) rat was uncovered and soaked in a DiI solution. Second, after 1 week, the samples were taken from the SD rat’s L4 and L5, which are the roots of the sciatic nerve after fixing myocardial perfusion with 4% paraformaldehyde. The 3D two-photon image of DRG cells in Fig. 5(a) was obtained from an object volume of 60 μm × 60 μm × 40 μm, which consists of 200 2D frames. Each 2D frame has 400 × 400 pixels, and the acquisition time was 1 frame per second. The step size between the frames was 200 nm along the z-axis. A high-performance 3D visualization software (Aviso standard, VSG) was used to display Fig. 5. The excitation laser power out of a 60 × /1.2W objective lens was 7 mW at the sample. Note that the Vybrant DiI as a lipophilic membrane stain diffuses laterally to stain the entire cell. Figure 5(b)-5(d) shows two-photon images for the Vybrant DiI-stained membrane of a DRG cell for three particular frames of the 135th xy-plane, 100th xz-plane, and 255th yz-plane. We can clearly identify the retrograde neuronal tracer in the DRG cell, which was taken from the distal sciatic nerve.

 

Fig. 5 (a) 3D two-photon image of Vybrant DiI-stained DRG cells. (b), (c), and (d) 2D two-photon images from the 135th xy-plane, 100th xz-plane, and 255th yz-plane, respectively.

Download Full Size | PDF

6. Conclusions

We have demonstrated a new TPM system based on a mode-locked Yb³⁺-doped fiber laser operating at a 1060 nm wavelength. No dispersion compensating element was used in the fiber laser, and the pulse width of the laser was 13 ps with large frequency chirping. Mode-locked pulses are compressed with an external grating pair pulse compressor for efficient TPE. The pulse width of the laser was compressed up to 270 fs after the pulse compressor, and the average power and repetition rate of the laser were 80 mW and 36 MHz, respectively. The lateral and axial resolutions of our TPM system were 0.44 µm and 1.34 µm, respectively. Furthermore, we have introduced an efficient method to obtain a minimum pulse width at the sample position of a TPM system by monitoring the two-photon induced photocurrent in a photodiode. Various focusing lenses were tested with this optimum pulse width monitoring system, and we have found that the pulse widths vary as much as 14 fs at a 1060 nm wavelength due to dispersion by the focusing lenses. By using this two-photon efficiency monitoring scheme with an external grating pair pulse compressor, we can obtain the optimum prechirped pulses for a given objective lens or any other optical components added in a TPM system, such as a beam splitter or dichroic filter. The feasibility of our system was demonstrated by taking 3D two-photon images of Vybrant DiI-stained DRG cells with a volume of 60 μm × 60 μm × 40 μm and 400 × 400 × 200 pixels. We believe that this simple but elegant TPM system equipped with variable pulse width control and a straightforward two-photon efficiency monitoring scheme is a very useful and affordable system for practical live cell and tissue imaging applications.