Temporal stability of the broadband source, such as supercontinuum (SC), is the key enabling factor for realizing high performance ultrafast serial time-encoded amplified microscopy (STEAM). Owing to that the long-pulse SC (picosecond to nanosecond) generation generally results in an ultrabroadband spectrum with significant pulse-to-pulse fluctuation, only the ultrashort-pulse (femtosecond) SC sources, which offer better temporal stability, have been employed in STEAM so far. Here we report a simple approach to achieve active control of picosecond SC stability and to help extend the applicability of SC in STEAM from the femtosecond to the picosecond or even nanosecond regimes. We experimentally demonstrate stable single-shot STEAM imaging at a frame rate of 4.9 MHz using the CW-triggered picosecond SC source. Such CW-stabilized SC can greatly reduce the shot-to-shot fluctuation, and thus improves the STEAM image quality significantly.
© 2011 OSA
Enhancing the temporal resolution (i.e. frame rate) in optical imaging without sacrificing the detection sensitivity has long been an imperative and yet challenging task, particularly in those applications where ultrafast measurements are essential, e.g. to unravel the high-speed dynamical phenomena in living biological cells  and to perform high-throughput cellular/molecular diagnostics in the microfluidic platforms . While the CCD/CMOS technologies remain as the mainstay in a myriad of optical microscopy applications, serial time-encoded amplified microscopy (STEAM) has recently been demonstrated as a completely new optical imaging modality which achieves ultrafast optical imaging without compromising the sensitivity [3–7]. Such intriguing feature stems from the fact that STEAM is able to optically amplify the images, which are encoded in the broadband optical pulses, and to capture them with a single-pixel detector. This eliminates the need for the CCD/CMOS imagers and overcomes the fundamental limitations of the existing CCD/CMOS technologies, i.e. a trade-off between imaging sensitivity and frame rate [3–7].
Imaging in STEAM is achieved by mapping the spectrally-encoded image shots respectively into individual optically-amplified temporal pulses. For high-performance STEAM, the light source in the system should possess two essential features: (1) it should exhibit a broadband spectrum—a perquisite of spectrally-encoding imaging [3–8], and (2) it should offer a high degree of shot-to-shot temporal amplitude stability—to minimize the instability in the frame-to-frame image quality. In this regard, supercontinuum (SC) pulses represent an appealing light source for STEAM primarily because of its ultra-wideband spectrum, which can span more than an octave . Despite of it, the utility of SC in ultrafast optical imaging could be hindered by its temporal instability, which is a rather subtle issue. Although it has been known that SC stability can be improved by using femtosecond pump pulses, such ultrashort pulses require dedicated dispersion control and are often sensitive to perturbations. Longer-pulse SC (e.g. picoseconds or nanoseconds) is hence more practical and indeed has been widely utilized in many biophotonics applications . However, long-pulse SC suffers from significant pulse-to-pulse fluctuation which originates from modulation instability (MI)—a nonlinear process initiated spontaneously from noise . Such noisy behavior explains why the long-pulse SC sources have been precluded in all the previously reported STEAM systems, which instead employed the femtosecond SC sources [3–7].
Different schemes have been proposed and implemented to control the SC instability, such as feedback mechanism , dispersion engineering , pulse seeding approach [12,13], or THz intensity modulation of the input pulse . In contrast to the prior works, we have recently demonstrated a simple triggering mechanism to enhance and stabilize picosecond SC generation by introducing a minute continuous-wave (CW) light within the MI gain spectral regions . This CW-triggering technique does not require phase-locking with the pump and precise time-delay tuning (as opposed to the pulse-seeding scheme [12,13]). Hence, it represents a handy approach to achieve active control of SC stability and to help extend the applicability of SC in STEAM from the femtosecond to the picosecond or even nanosecond regimes. In this paper, we report stable single-shot STEAM imaging at a frame rate of 4.9 MHz using the CW-triggered picosecond SC source. More importantly, such CW-stabilized SC can greatly reduce the shot-to-shot fluctuation, and thus improves the image quality significantly.
2. Working Principles
The present STEAM system consists of three key parts as shown in Fig. 1 . The first part is the stabilized CW-triggered picosecond SC source (Fig. 1(a)). It is realized by injecting an additional weak CW light trigger (~50 dB weaker than the pump) with a wavelength within the MI gain spectrum during the SC generation . In this CW-triggering approach, the MI growth can be initiated by the controllable CW, rather than noise. The subsequent soliton fission (which is a key process leading to drastic spectral broadening in SC generation) is then triggered in a more deterministic fashion. Hence, the SC generation is greatly accelerated and suffers significantly less pulse-to-pulse amplitude fluctuation—an essential condition for performing high-speed and stable single-shot STEAM. We emphasize that this CW-triggering technique represents a simple and practical approach to achieve active control of the SC properties as it does not require phase-locking with the pump and precise time-delay tuning (as opposed to the pulse-seeding scheme [12,13]).
The second stage of the STEAM system is to perform space-wavelength mapping in which the spatial information of a sample is encoded onto the spectrum of the stabilized SC pulse using a spatial disperser. While STEAM can perform either one-dimensional (1-D) or two-dimensional (2-D) imaging [3–7], we here employ a diffraction grating as a 1-D spatial disperser to perform 1-D STEAM (Fig. 1(b)). We note that such 1-D STEAM configuration naturally finds a compelling application in high-speed and sensitive imaging flow-cytometry .
In the last part of the system, the image information in the spectrally-encoded SC pulse is then mapped into a temporal waveform by amplified dispersive Fourier transformation (ADFT) (Fig. 1(c)). ADFT is essentially a process of wavelength-time mapping by virtue of GVD and optical amplification in a dispersive fiber [16,17]. It has the advantages of not only compensating for the inherent loss associated with GVD, but also providing the optical gain to enhance the detection sensitive in STEM. Finally, the image-encoded temporal pulse can be detected by a single-pixel photodiode, and captured by a high-speed real-time oscilloscope. Details of the working principles of STEAM can be referred to Refs. [3,4].
Because the pulse-to-pulse fluctuation of the SC is now stabilized by the CW-triggering scheme, the corresponding frame-to-frame STEAM image quality would be greatly improved. Longer-pulse SC (e.g. picoseconds or longer) is generally recognized to give rise to an unstable source which makes it unsuitable for high-speed real-time measurements, e.g. STEAM. In this regard, this demonstration is significant in that it extends the applicability of SC in STEAM from the femtosecond to the picosecond or even nanosecond regimes, in which MI plays an important role in the SC generation.
3. Experiments and results
Intense pump pulses delivered by a picosecond mode-locked fiber laser (pulse width of 5.8 ps, peak power of 22 W), as well as a wavelength-tunable CW source (:80 μW) are coupled together into a 50 m highly-nonlinear dispersion-shifted fiber (HNL-DSF) (zero-dispersion wavelength: 1554 nm, dispersion slope: 0.035 ps/nm2/km, nonlinear coefficient: 14 W−1 km−1). As shown in Fig. 2(a) , the SC spectrum is greatly enhanced when the weak CW trigger wavelength is tuned to the MI gain sidebands (1500–1510 nm and 1610–1620 nm), where it experiences large MI gain. Notably, when the CW-trigger is at 1615 nm, the SC spectrum shows a clear signature of the onset of soliton fission which results in drastic spectral broadening in the SC generation. Such enhanced SC spectrum spans from ~1400 nm to ~1700 nm, which is much wider than the untriggered SC spectrum, and the SC power is increased by ~20 dB on both the redshifted and blueshifted sides of the SC spectrum. Figure 2(b) shows the amplitude histogram of 781 filtered SC pulses (1620–1650 nm) obtained by real-time pulse-amplitude statistical measurements which employ a real-time oscilloscope (4 GHz, 20 GS/s) . In the untriggered case, it shows a clear long-tail distribution, which signifies the occurrence of the optical rogue waves . In contrast, when the CW-trigger is added, it alters the SC amplitude statistics to an almost-Gaussian distribution and shows a:50% reduction in the standard deviation. Such improvement in pulse-to-pulse stability is also readily evident from the real-time pulse traces in both untriggered and CW-triggered cases (insets of Fig. 2(b)). Such CW-stabilized picosecond SC pulses are favorable for high-speed and single-shot STEAM imaging.
As described earlier that the key feature of STEAM is the mapping of an image into a serial time-domain waveform by a two-step approach: space-wavelength mapping and wavelength-time mapping. In the space-wavelength mapping, we employ a diffraction grating (600 lines/mm) together with an imaging lens (focal length of 50 mm) to deliver the spatially-dispersed SC beam, termed as spectral-shower [3–5], onto the sample. The spectrum of the back-scattered spectral-shower is then encoded with the spatial information of the sample. Afterwards, a dispersion compensation fiber (DCF) (GVD: –356 ps/nm) is used to temporally disperse the pulses via GVD and thus maps the spectrally-encoded information of the SC pulse into a temporal waveform. Within the DCF, Raman amplification is also implemented to overcome the inherent dispersive loss in the DCF. The Raman pump (with a power of 1.2 W) at 1535 nm is chosen to amplify the STEAM signal within the wavelength band from 1620 nm to 1650 nm. An optical gain within this spectral band is measured to be 13 dB and the optical signal-to-noise ratio (OSNR) is measured to be ~13 dB across the same spectral band. We note that multiple pumping scheme could be adopted to further enhance the gain and the gain bandwidth .
We use the present STEAM system based on the stabilized CW-triggered SC source to image the test barcode pattern in order to illustrate its basic functionality, as shown in Fig. 3 . The test barcode is printed on a transparency film. We verify the space-time mapping operation of STEAM by comparing the image-encoded spectrum (generated by space-wavelength mapping) and the image-encoded temporal waveform (generated by wavelength-time mapping) (Fig. 3(a)). It is clear that the spectral shape closely resembles the temporal waveform shape. Figure 3(b) shows another test barcode pattern image captured by STEAM using the stabilized CW-triggered SC.
It has been reported that the image resolution of STEAM, unlike the typical optical microscopes, depends not only on the imaging lenses, but also on the properties of other components that are unique to STEAM, namely the spectral resolution of the diffraction grating, the GVD of the fiber, and the bandwidth of the back-end digitizer . It appears in Fig. 3(a) that some of the high spatial frequency contents of the image are lost in the temporal waveform when compared with the image-encoded spectrum. This phenomenon evidently illustrates the GVD is the key limiting factor of the actual resolution of the present STEAM system . To achieve a better image resolution, we employ another longer DCF (with a GVD of –867 ps/nm). As a result, the measured spatial resolution can be improved down to 31.3 μm. This is in excellent agreement with the theoretical calculation, governed by the aforementioned three factors (i.e. the spectral resolution of the grating, the fiber GVD, and the digitizer bandwidth), which is 30.8 μm . The resolution can be greatly further improved by using a high numerical aperture (NA) imaging lens. We remark that such STEAM system performance evaluation can be made possible only if the source exhibits reasonably good pulse-to-pulse temporal stability. With a stable broadband source, he single-shot real-time barcode image can be clearly captured (as shown in Fig. 3(a) and (b)), that is in this case offered by the stabilized CW-triggered long-pulse (picosecond) SC source.
We further examine the impact of the SC source’s temporal stability on the STEAM image quality. We image the USAF-1951 standard resolution target using STEAM with two different SC sources: the untriggered SC and the CW-triggered SC. As a proof-of-concept demonstration, the present STEAM system operates in the single-shot line scan mode (along the x-direction) at a rate of 4.9 MHz (effectively determined by the repetition rate of the source). The two-dimensional (2-D) images are obtained by translating the sample in the orthogonal direction. We note that such 1-D (line-scan) STEAM configuration naturally fits in the applications involved microfluidic flow cell imaging . By taking the advantage of that the biological cells (e.g. blood cells) flowing along the microfluidic channel, the 1-D spectral shower can be used to interrogate and line-scan the individual cells to acquire the 2-D image without any beam scanning of the spectral shower. This could find a compelling application in high-speed and sensitive imaging flow-cytometry .
Figures 4(a)–(d) show the comparisons of the STEAM images of the resolution target (elements 2-4, 2-5, and 2-6) between the cases of using the untriggered SC and the CW-triggered SC. Clearly, the STEAM images taken by the CW-triggered SC (Figs. 4(a) and (c)) result in a better image quality in terms of the image noise and image contrast when compared with the images taken by the untriggered SC (Figs. 4(b) and (d)). It is mainly attributed to the fact that untriggered SC suffers serious temporal instability. We remark that each line scan(in the x-direction) represents the single-shot STEAM raw data and each scan takes only 12.8 ns, which is essentially the duration of each temporally-dispersed pulse after the ADFT process.
We also perform STEAM of the lens paper. For the case of STEAM using the untriggered SC, the fiber structures in the lens paper are hardly visualized and the image suffers serious noise contamination (Fig. 5(a) ). In contrast, the STEAM image taken by the CW-triggered SC (Fig. 5(b)) shows a greatly improved image contrast—demonstrating the impact of the CW-stabilized picoseconds SC source on the imaging performance of STEAM. This is, to the best of our knowledge, the first demonstration of STEAM using longer-pulse SC (as opposed to be femtosecond SC) which is stabilized by a simple CW-triggering mechanism. This demonstration is significant in that longer-pulse SC sources generally give poorer temporal stability especially when the SC is initiated by MI. Hence, they are incapable of performing real-time ultrafast and single-shot measurements. This also explains why the longer-pulse SC sources have been precluded in all the previously reported STEAM systems, which instead employed the femtosecond SC sources [3–7]. Here, the improvement in the temporal stability of SC by the CW-trigger opens up a wider range of choices of the ultrabroadband SC sources (from femtosecond to picosecond or even nanosecond pulses) for high-performance STEAM.
Note that the STEAM system is essentially a confocal microscope. It is because the present STEAM system eliminates the out-of-focus information by the core of the collection fiber (the fiber collimator in Fig. 1(b)) rejects the light scattered from the out-of-focus planes of the sample. Indeed, we can notice that the out-of-focus features in the STEAM image (Fig. 5(b) red circle) are missing when compared with the conventional bright-field microscope image (Fig. 5(c) red circle).
In conclusion, we have experimentally demonstrated for the first time, to the best of our knowledge, STEAM using the stabilized long-pulse SC (picosecond) source, which is enabled by a simple active CW-triggering scheme. Such triggering approach enhances and stabilizes the long-pulse SC without relying on the intricate techniques such as precise time delay tuning, phase-locking, dedicated feedback control. More importantly, this demonstration extends the applicability of SC in STEAM from the femtosecond to the picosecond or even nanosecond regimes. Specifically, we demonstrated that the image quality of ultrafast single-shot STEAM (at a rate of 4.9 MHz) can be greatly enhanced by using such CW-triggered SC. We note that the wavelength range (~1550 nm) of the present SC source is yet to be optimized for practical flow cell imaging. Nevertheless, the present demonstration can be translational to the shorter near infrared (NIR) wavelength regime where CW-triggered SC can be realized by the highly-nonlinear fibers, e.g. photonic crystal fibers, and off-the-shelf CW laser diodes which are readily available for the 800 nm to 1μm operation. Such shorter wavelength regime is more favorable for cellular imaging using STEAM as better diffraction limited resolution can be achieved. By further improving the image resolution [5,16], we expect that STEAM could find a wide range of appealing applications in which real-time, high-speed and high-throughput imaging is essential, e.g. imaging flow-cytometry. Furthermore, we note that the present stabilized SC can also be applicable to all other applications where stable, real-time, and ultrafast optical measurements, such as ADFT-based spectroscopy  and optical time-stretch signal processing , are critical.
The work in this paper is partially supported by grants from the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No. HKU 7179/08E, HKU 7183/09E and HKU 717510E). The authors also acknowledge Sumitomo Electric Industries for providing the HNL-DSF and Alnair Laboratories for providing the variable bandwidth tunable bandpass filter.
References and links
2. J. V. Watson, Introduction to Flow Cytometry (Cambridge University Press, 2004).
4. K. Goda, K. K. Tsia, and B. Jalali, “Amplified dispersive Fourier-transform imaging for ultrafast displacement sensing and barcode reading,” Appl. Phys. Lett. 93(13), 131109 (2008). [CrossRef]
6. A. Mahjoubfar, K. Goda, A. Ayazi, A. Fard, S. H. Kim, and B. Jalali, “High-speed nanometer-resolved imaging vibrometer and velocimeter,” Appl. Phys. Lett. 98(10), 101107 (2011). [CrossRef]
9. J. M. Dudley and J. R. Taylor, Supercontinuum Generation in Optical Fibers (Cambridge University Press, 2010).
10. P. M. Moselund, M. H. Frosz, C. L. Thomsen, and O. Bang, “Back-seeding of higher order gain processes in picosecond supercontinuum generation,” Opt. Express 16(16), 11954–11968 (2008). [CrossRef] [PubMed]
14. G. Genty, J. M. Dudley, and B. Eggleton, “Modulation control and spectral shaping of optical fiber supercontinuum generation in the picosecond regime,” Appl. Phys. B 94(2), 187–194 (2009). [CrossRef]
16. K. Goda, D. R. Solli, K. K. Tsia, and B. Jalali, “Theory of amplified dispersive Fourier transformation,” Phys. Rev. A 80(4), 043821 (2009). [CrossRef]
17. D. R. Solli, J. Chou, and B. Jalali, “Amplified wavelength–time transformation for real-time spectroscopy,” Nat. Photonics 2(1), 48–51 (2008). [CrossRef]
19. M. N. Islam, “Raman amplifiers for telecommunications,” IEEE J. Sel. Top. Quantum Electron. 8(3), 548–559 (2002). [CrossRef]
21. J. Chou, O. Boyraz, D. Solli, and B. Jalali, “Femtosecond real-time single-shot digitizer,” Appl. Phys. Lett. 91(16), 161105 (2007). [CrossRef]