1. Introduction

Fluorescence lifetime imaging microscopy (FLIM) provides crucial information on molecular dynamics of biological systems to elucidate cellular function [1–6]. Important cellular information, such as cytoplasm matrix viscosity, pH level, and cellular metabolism, has been characterized using lifetime-resolved methods [7–13]. Additionally, lifetime imaging is used to differentiate chromophores with similar emission spectra but different lifetimes [14,15].

Based on the techniques used to acquire FLIM data, there are three major methods, the frequency domain, the time domain, and the pump-probe ones. In frequency-domain FLIM, the intensity of the excitation light is continuously modulated while the lifetime is extracted from the Fourier spectrum of the intensity and the phase shift of the emitted fluorescence signal. The frequency domain method is the earliest FLIM method. It works well when the fluorescence signal is strong and there is a single lifetime.

The time domain methods, such as TCSPC and gated camera, are intuitive in data interpretation and have found wide applications in recent years. TCSPC is a very sensitive method that achieves very high signal-to-noise ratio through photon counting. It may, however, take relatively long time to acquire a complete FLIM data set, limited by avoiding pile-up in photon counting. Gated camera can achieve higher frame rate, however, at the expense of low signal-to-noise ratio. In general, cost and acquisition time are among the main concerns in applying time domain methods. Both the frequency and the time domain methods analyze the spontaneous emission of fluorescence and often require high NA optics for efficient signal collection.

Alternatively, the pump-probe method extracts lifetime from the stimulated emission and is not limited by the NA of optics, and thus the working distance. The recent advancements in light sources have opened new opportunities for the pump-probe method, as also reported in this study. With the use of all-semiconductor lasers, it is possible to realize full electronic control, with the advantages of miniaturization, cost-effectiveness, robustness, and versatility.

Fluorescence usually occurs via spontaneous emission, which is an incoherent process with large angle of divergence. A high numerical aperture (NA) lens is thus required to collect the emitted photons effectively [16,17]. However, a high NA lens also sets the limits on the field of view and the working distance. Additionally, a highly sensitive detector with large gain, such as PMT or APD, is needed to detect the weak spontaneous signal.

For comparison, stimulated emission microscopy has recently demonstrated as an effective alternative to observe “dark” or short lifetime (subnanosecond) fluorophores [18–22]. Stimulated emission is also shown to work at extended working distance due to the accompanying high coherence and small divergence [23]. A photodiode, with good quantum efficiency and high dynamics range, is sufficient as a detector, due to the high intensity of the stimulated beam.

In this work, we have realized stimulated emission FLIM on ATTO647N dye with a pair of synchronized gain switched diode lasers as the excitation and stimulation light sources, respectively. In addition to compactness and cost-effectiveness, the all-semiconductor and fully electronic configuration allow FLIM measurements in a highly flexible and versatile way. The stability and warm-up time is also greatly improved, when compared with large frame lasers. The variable repetition rate would allow probing of dyes with a broad range of lifetimes. The reduced peak power due to the longer pulse width also better matches the fluorophores in avoiding adverse effects, such as saturation, bleaching, and nonlinear excitation.

2. Experimental set-up

The schematic of the stimulated emission based FLIM setup is shown in Fig. 1 . Two pulsed diode lasers (LDH-635 and LDH-700, Picoquant, Berlin, Germany) with wavelength 635 nm and 700 nm are used to excite and induce stimulated emission, respectively. The selected wavelengths match well with the peaks of absorption (644 nm) and emission (669 nm) spectra of ATTO647N dye (ATTO-TEC, Germany). The repetition rate is set at 80MHz and can be easily changed. The two laser beams are combined with a dichroic mirror (720SP, Semrock, USA) before being directed into the galvono-mirror scanning unit (FV300, Olympus, Japan) for scanning imaging. The average power of irradiation from the excitation and the stimulation beams measured at the sample are approximately 0.1 mW and 2.58 mW, respectively. Taking into account that the laser beam doesn’t fill up the pupil of the aperture, the final spot-size on the sample is estimated to be 25 μm². The power density on the sample from the excitation and the stimulation beams are approximately 400 W/cm² and 1 × 10⁴ W/cm², respectively.

 

Fig. 1 Schematic of stimulated emission based FLIM setup. LD1, LD2: laserdiode. LC1, LC2: laser driver. W1, W2: half-wave plate. P1, P2, P3: polarizer. EOM: Electro-optical modulator. R: retro-reflector. PH: pinhole. DM: dichroic mirror. OL: objective lens. S: sample. F: filter. D: detector. LA: lock-in amplifier.

Download Full Size | PDF

The collecting objective is placed at a slightly out-of-focus position so as to minimize the collection of spontaneous fluorescence. A doublet lens with a focal length of 40 mm is used after the collecting objective to collimate the stimulation beam within the silicon detector’s active region (DET36A, Thorlabs, USA). The optical setup minimizes both the collection of spontaneous emission and the effect of thermal lensing, which are frequently encountered in the pump-probe measurements [24].

In the measurements, the key to alignment is to ensure the overlapping of excitation and stimulation spatially and temporally. Spatial overlapping is easily achieved by observing the two beams’ profiles and their overlap between the scanning unit and the objective with a CCD camera. A time correlated single photon counting (TCSPC) module is used to synchronize the excitation and the stimulation pulses coarsely [4,25]. Fine tuning of temporal overlapping is achieved by adjusting the length of the trigger signal cable at an interval of 2 cm, which corresponds to time delay of approximately 0.1 ns. Very fine temporal delay adjustment is carried out by moving the retro-reflector to change the optical path length. The pulse duration (~100 ps), however, does not require much fine adjustment. To achieve maximum signal, the polarizations of the excitation and the stimulation beams are kept linear and paralleled to each other.

The excitation beam is modulated by an electro-optical modulator (EOM, ConOptics, USA) from 10 to 100 KHz. A narrow bandpass filter centered at 700 nm is placed in front of the silicon photodiode to reject the excitation beam and the spontaneous fluorescence that is outside of the filtered band of stimulated emission. The output from the photodiode is then demodulated with a lock-in amplifier (SR830, Stanford research, USA) to recover the relatively weak stimulated emission signal from the strong background presented by the stimulation beam [26]. The authenticity of the signal is verified by blocking the excitation and/or the stimulation beams, as well as changing the relative time delay between the lasers. A high signal-to-noise ratio of 600:1 is achieved. The fluorescence lifetime is obtained by fitting the signal decaying curve as a function of time delay introduced electronically between the excitation and the stimulation pulses. For imaging, the FV300 scanning unit is set on the slow scanning mode and an image size of 256 × 256 pixels is acquired. The scan frequency is set at 500 Hz to match the time constant of the lock-in amplifier. It takes a bit over 2 minutes to acquire an image.

The ATTO647N dye solution with various concentrations (1~100 μΜ) is used as the specimen. The solution is injected into a gap sandwiched between two cover slips with 2 mm separation. The dye solution fills a region of 2 mm diameter, while the scanned area is approximately 600 μm × 600 μm.

3. Results and discussion

The images reconstructed from stimulated emission at various time delays are shown in Fig. 2 , in which a dye concentration of 100 μM is used. Predictably, the maximum signal intensity is obtained at no time delay, as shown in Fig. 2(a). The intensity decreases gradually with the increased time delay as the population of excited atoms reduces by transiting to lower energy states, as shown in Figs. 2(b) to 2(d).

 

Fig. 2 Images reconstructed from stimulated emission at (a) 0 ns time delay; (b) 1.5 ns time delay; (c) 3 ns time delay; (d) 4.5 ns time delay. Note that the intensity of these images is color coded. The regions of very low intensity are non-fluorescent defects on the surface of the sample container. The scale bar is 100 μm.

Download Full Size | PDF

As in previous studies [18,23], the signal of stimulated emission exhibits linear dependence on the excitation and the stimulation powers, $∆\mathrm{I}=\mathrm{I}\left(\mathrm{d}\right)-\mathrm{I}\left(0\right)\propto {\mathrm{N}}_0{\mathrm{I}}_{\mathrm{E}}{\mathrm{I}}_{\mathrm{S}}$, when the maximum intensity is below saturation.

Figure 3(a) shows the stimulated emission signal as a function of pulse delay. The acquisition is similar to gated camera based FLIM, in which complete images are acquired at various time delays. The lifetime fitting results in a value of 3.4 ns. The measurements are repeated several times under various intensities of excitation and stimulation. For comparison, the reported values are 1.8 ns and 3.4 ns [27,28]. The corresponding FLIM image is shown in the inset of Fig. 3, which is obtained through pixel-by-pixel fitting.

 

Fig. 3 (a) Stimulated emission signal as a function of time delay between the excitation and the stimulation laser pulses. The decay curve is obtained from the region of interest (marked in inset). Inset: the corresponding FLIM image. The areas of very short lifetime are non-fluorescent defects on the surface of the sample container (r.f. Fig. 2). The scale bar is 100 μm. (b) The standard deviation of the signal as a function of the modulation frequency on the excitation beam.

Download Full Size | PDF

The dependence of standard deviation (or fluctuation) of the signal on the modulation frequency is shown in Fig. 3(b). The decreasing of the standard deviation as a result of increasing modulation frequency is attributed to the reduction in 1/f noise [29], with quantum shot noise being the ultimate limit. In Fig. 3(b), it seems the noise reaches minimum when the modulation frequency reaches 80~100 KHz.

4. Conclusion and outlook

In addition to compactness and cost effectiveness, the many advantages in using the all-semiconductor lasers over large frame mode-locked laser include stability and reduced warm-up time, flexible repetition rate, and optimized peak power due to longer pulse width. The long term stability is highly correlated with the size of a laser system and the time for it to reach thermal equilibrium. The warm-up time of a semiconductor laser is much shorter, due to its compactness. It is also difficult to change the repetition rate of a mode-locked laser, since it is originated from the cavity length. For comparison, the repetition rate of the gain-switched semiconductor laser is fully adjustable electronically. This feature would allow the easy probing of fluorophores with various lifetimes, especially the long ones. The peak power of a mode-locked laser is usually very high and is often used for nonlinear excitation. In the case of single photon excitation or stimulated emission, the high peak power may, however, lead to the adverse effects, such as saturation, bleaching, and nonlinear excitation. In the work of stimulated emission depletion microscopy (STED) [30], it has been reported that stimulated emission is best enacted with pulses of longer duration, to better match the lifetime of fluorophores. The gain switched semiconductor laser effectively fulfills the above consideration. As a result, the quality of signals using all-semiconductor laser configuration improves, when compared with the one from the Ti:sapphire laser.

In most fluorescence detection, the coherence of the fluorescence is either non-existent (spontaneous) or not fully exploited (stimulated emission) to provide additional information. By combining with interferometric measurements, one of the prospective applications of stimulated emission detection is optical coherence tomography (OCT). The inclusion of stimulated emission would add the fluorescence contrast available to OCT in investigating stained samples.

In conclusion, we have shown that semiconductor lasers and full electronic control can greatly reduce the barrier in implementing stimulated emission for a time-resolved configuration. The improved sensitivity and simplified configuration would render this technique a practical tool in coming biomedical researches. The interplay of fluorescence contrast, coherence, and time-resolving would also promise unprecedented opportunities.