Letter

Vol. 40, No. 23 / December 1 2015 / Optics Letters

5447

Light source for narrow and broadband coherent Raman scattering microspectroscopy MAXIMILIAN BRINKMANN,1,2,* SVEN DOBNER,1,2

AND

CARSTEN FALLNICH1,2

1

Institute of Applied Physics, Corrensstr. 2, 48149 Münster, Germany Cells-in-Motion Cluster of Excellence (EXC 1003-CiM), University of Münster, Münster, Germany *Corresponding author: [email protected]

2

Received 1 September 2015; revised 15 October 2015; accepted 20 October 2015; posted 21 October 2015 (Doc. ID 249116); published 17 November 2015

We present a light source that is well adapted to both narrow- and broadband coherent Raman scattering (CRS) methods. Based on a single oscillator, the light source delivers synchronized broadband pulses via supercontinuum generation and narrowband, frequency-tunable pulses via four-wave mixing in a photonic crystal fiber. Seeding the four-wave mixing with a spectrally filtered part of the supercontinuum yields high-pulse energies up to 8 nJ and the possibility of scanning a bandwidth of 2000 cm−1 in 25 ms. All pulses are emitted with a repetition frequency of 1 MHz, which ensures efficient generation of CRS signals while avoiding significant damage of the samples. Consequently, the light source combines the performance of individual narrow- and broadband CRS light sources in one setup, thus enabling hyperspectral imaging and rapid single-resonance imaging in parallel. © 2015 Optical Society of America OCIS codes: (180.5655) Raman microscopy; (190.4970) Parametric oscillators and amplifiers; (320.7140) Ultrafast processes in fibers. http://dx.doi.org/10.1364/OL.40.005447

Coherent Raman scattering (CRS) depicts a powerful contrast mechanism for label-free and chemically selective imaging, based on the coherent excitation of vibrational resonances of a sample. Over the years, several sophisticated techniques to utilize CRS have been developed. Concerning their spectral excitation bandwidth, these CRS schemes can be categorized into two main groups: narrowband and broadband schemes. In narrowband schemes, a single vibrational resonance of a sample is excited by a pump and a Stokes pulse, both with a spectral bandwidth narrower than the resonance. Given that the frequency difference of the two pulses matches the frequency of the addressed vibration, either a blueshifted signal can be measured as in coherent anti-Stokes Raman scattering (CARS) [1], or an energy transfer from the pump (energy loss) to the Stokes pulse (energy gain) can be observed, as in stimulated Raman scattering (SRS) [2]. As the detection in these techniques is based on a fast photodetector, high-speed acquisition with pixel dwell times in the order of microseconds is indeed possible; however, only a single resonance can be 0146-9592/15/235447-04$15/0$15.00 © 2015 Optical Society of America

addressed at once [3]. In contrast, in broadband CRS schemes, the combination of a spectrally narrow pump pulse and a spectrally broad probe pulse excites all resonances of a sample simultaneously. The aforementioned Raman-induced energy transfer, associated with each excited resonance, then causes amplitude modulation of the probe spectrum, which is measured in femtosecond stimulated Raman scattering (FSRS) [4] with an optical spectrum analyzer. Therewith, the complete Raman spectrum of a sample can be extracted at once; however, limited by the readout time of the optical spectrum analyzer, the acquisition time is at least one order of magnitude longer than in narrowband CRS techniques [5]. Thus, broadband CRS schemes are advantageous when spectral information is desired, e.g., to identify an unknown sample or to observe spectral dynamics of a sample, while narrowband CRS schemes are favorable when a high acquisition speed is desired, e.g., when tracking the spatial dynamics of a specific substance in a scanning microscope. Consequently, to harvest the advantages of both schemes, the alternating application of narrow- and broadband schemes to one sample and, thus, a light source adapted to both schemes is desirable. In order to satisfy this aspiration, we have recently set up a laser system that meets various demands of CRS spectroscopy of biological samples: a master oscillator and power amplifier (MOPA) emits pulses with a central wavelength of 1033 nm and high energy up to 5 μJ, from which all other desired pulses are derived. A mediocre repetition rate of 1 MHz carefully balances the applicable average and peak power to avoid optical damage, especially in biological samples [6,7]. For comparison: in order to reach sufficient peak power to efficiently induce CRS with systems exhibiting a much higher repetition rate, high average power is required, which impairs delicate samples through heat accumulation [8]. On the other hand, when using a laser system with a lower repetition rate of a few kHz, a thermal impact is indeed negligible, but the maximal applicable peak power is limited through multiphoton-induced damage [9]. In [10], we have shown that our laser system is well suited for broadband CRS methods via supercontinuum (SC) pulses generated in GdVO4 . In order to expand the scope of the light source toward narrowband CRS schemes, in this contribution we show how our

5448

Vol. 40, No. 23 / December 1 2015 / Optics Letters

laser system could be advanced to also emit narrowband and frequency-tunable pulses. The generation of the narrowband, frequency-tunable pulses is accomplished by means of a fiber-based optical parametric amplifier (FOPA), working on the basis of four-wave mixing (FWM) in a photonic crystal fiber (PCF). In FWM, which is an optical χ 3 process, an optical pump field interacts with the fiber to be converted into spectrally up- and downshifted signal and idler fields, respectively [11]. Light sources based on FWM have been applied to CRS: for example, spontaneous FWM, meaning that the signal and idler fields are generated out of quantum noise, allows to construct all-fiber light sources [12]. However, pulses generated by spontaneous FWM inherently show strong fluctuations and exhibit bandwidths wider than typical Raman linewidths [13]. These disadvantages can be overcome by stimulating the FWM, which has been done with an external frequency-tunable laser [14] or by means of optical feedback [15]. In the concept presented here, the supercontinuum pulses were used to seed the FWM. Therewith, the need for an external seed laser or optical feedback was eliminated, while the convenience and flexibility of an amplifier was preserved. The experimental setup is illustrated in Fig. 1. In order to drive the described SC generation and FWM, the system was based on a high-energy MOPA (model Satsuma from Amplitude Systems). The MOPA output (pulse energy, 5 μJ; pulse duration, 400 fs; repetition rate, 1 MHz) was split into two arms to generate at first the pulses required for broadband CRS: broadband probe pulses were obtained via SC generation, narrowband pump pulses were obtained via spectral filtering and subsequent amplification. The SC generation in the first arm was accomplished by focusing 1 μJ of pulse energy using a lens with a focal length of 75 mm into a 4 mm long crystal of GdVO4 . As can be seen in Fig. 2, relative to the pump wavelength of 1033 nm, the spectrum of the generated pulses covered the entire molecular vibrational spectrum. Divided by a short-pass filter (DM1), the short-wavelength (1 μm) part was used to seed FWM in a PCF. In addition, we were able to measure pulse-to-pulse amplitude fluctuations of the SC pulses as low as 1% RMS, owing to the spectral broadening being closely linked to coherent processes such as filamentation and self-phase modulation [16]. In comparison, supercontinuum sources based on optical fiber usually suffer from much higher fluctuations up to 50% [17], when pumped with pulses with a temporal duration of several 100 fs and above, as the spectral broadening is then dominated by noise-induced FWM [11]. Parallel to the generation of the SC pulses, in the second arm of the setup, narrowband pump pulses had to be derived from the MOPA pulses. For this task, two approaches have been presented: spectral compression of a negatively chirped pulse in a PCF due to self-phase modulation has been reported to be efficient but suffers from low stability [18]. Another approach based on the spectral compression due to a narrow phasematching bandwidth in a second-harmonic generation process would yield pulses with twice the central frequency of the MOPA pulses [19]. Thus, we decided to derive the narrowband pump pulses by spectrally filtering the MOPA pulses by means of a folded Fourier filter, based on a transmission grating (1200 lines∕mm), a focusing mirror (f  500 mm), and a mechanical slit. Although this filter dissipated a large fraction of the input pulse energy (∼95.5% at the following setting), it allowed us to freely choose the central frequency and spectral width of the generated pulses. In general, we generated pulses with a spectral full width at half maximum (FWHM) of 0.5 nm, i.e., slightly narrower than a typical Raman line width, as such pulses ensure a sufficient spectral resolution for CRS and yet high peak powers at moderate average powers [1]. Subsequent to the spectral filter, the pulse energy would suffice for the application to CRS but needed to be amplified to drive the effect of FWM. For this purpose a divided pulse amplification (DPA) scheme as described in [20] was utilized. In this technique, the original pulse was divided into 32 temporally separated copies of itself, each with a reduced peak power, by a set of five birefringent YVO4 crystals. The pulse copies were individually amplified in an Yb-doped fiber (1 m of Liekki Yb1200-25/250DC) and subsequently recombined coherently by the reverse process to form one intense pulse. In this way, we were able to amplify the pulse energy up to 800 nJ before nonlinearity started to distort the pulse spectrum, which was about 26 times the energy as without the DPA. The amplified pulses exhibited a spectral FWHM bandwidth of

Spec. power density (dBm/nm)

Relative wavenumber (cm-1) -6117

-2324

0

1570

2702

3557

1033

1233

1433

1633

0 -10 -20 -30 -40 633

833

Wavelength (nm)

Fig. 1. Diagram of the experimental setup. HWP, half-wave plate; PBS, polarizing beam splitter; GdVO4 , gadolinium vanadate crystal; DM, dichroic mirror; LP, long-pass filter; MO, microscope objective; DPA, divided pulse amplification.

Fig. 2. Spectrum of the supercontinuum pulses, generated in gadolinium vanadate. The power of the spectral components around 1030 nm was reduced by a notch filter. On the upper x axis, the wavenumber relative to 1033 nm is depicted.

Vol. 40, No. 23 / December 1 2015 / Optics Letters

Letter 0.5 nm, as shown in Fig. 3(a), and, assuming a Gaussian pulse profile, a temporal pulse duration of 3.4 ps. This corresponds to a time-bandwidth product of 0.48, thus indicating nearly transform-limited pulses. Moreover, pulse-to-pulse amplitude fluctuations of less than 2% RMS could be measured, meaning that the fluctuations increased by a factor of 2 compared with the MOPA output. This increase is two times less than originally reported for the mentioned alternative based on SPM-induced spectral compression [18]; however, just recently, a source for synchronized femto- and pico-second pulses based on SPM-induced spectral compression and a DPA scheme has also been presented with an improved noise figure [15]. The latter, as well as our here presented results, render the DPA as the method of choice to derive high-power pico- from femtosecond pulses. Finally, the frequency-tunable pulses for narrowband CRS were generated by FWM in a PCF, pumped with an average power of 50 mW (i.e., 50 nJ of pulse energy) of the DPA pulses and seeded with the spectrally filtered long wavelength part of the SC. To be more precise, the seed pulses were obtained by cutting the desired frequency component out of the SC spectrum using a standard folded Fourier filter, consisting of a reflection grating (600 lines∕mm) and a focusing mirror (f  300 mm). In this way, the frequency of the seed pulses could be tuned by turning the grating. The spectral bandwidth of the seed pulses could be set via the slit width. The seed pulses, with an average pulse energy of 15 pJ and a spectral bandwidth of 1.5 nm, were then temporally and spatially overlapped with the DPA pulses via a dichroic mirror (DM2) and coupled into 10 cm of PCF (SC-1040-5.0 from NKT Photonics). We chose this PCF because its zero dispersion wavelength at 1040  10 nm, which is close to the central wavelength of the DPA pulses, ensured a broad phase-matching bandwidth [11]. The resulting measured tuning curve of the system is shown in Fig. 3, thus indicating that the FOPA pulses could be tuned across almost 2000 cm−1 while having a minimum pulse energy of 1 nJ. The dip around 1140 nm in the tuning curve was a result of a relatively low gain and seed power at this spectral position. Although the FWM gain maximum was located at about 1200 nm, the lower gain below 1140 nm was compensated for by the high-energy content of the seed supercontinuum in this range. Thus, this combination of a SC-seeded FOPA did not only allow to access the whole molecular fingerprint region (700–1500 cm−1 ) with sufficient pulse energy for CRS [7], but also wavenumbers as low as 80 cm−1 , which represent an important spectral region for investigations on complex biomolecules [21].

5449

A sample spectrum of the FOPA pulses is shown in Fig. 3(b), which depicts a spectral FWHM bandwidth of about 2 nm or 10 cm−1. Furthermore, a numerically deconvolved, measured intensity cross correlation between these pulses and the MOPA pulses exhibited an FWHM duration of 1.64 ps, which represents an estimate for the duration of the FOPA pulses, as shown in Fig. 3(c). In order to minimize the temporal walk-off between amplified pulses of different wavelengths, we only used 10 cm of PCF. As also shown in Fig. 4 and calculated following reference [22], the maximum walk-off between amplified pulses did amount to about 0.6 ps. Consequently, no temporal delay adjustment was necessary when tuning the wavelength, leaving the tuning speed to be solely determined by the turning speed of the grating. In our case, about 25 ms were required to scan across the complete tuning bandwidth of almost 2000 cm−1 . This would allow us to tune the wavelength in a frame-by-frame manner in a time equal to the acquisition time for one image, when imaging with 20 images per second. Moreover, pulse-to-pulse amplitude fluctuations were measured to be below 3% RMS, which, as will be shown later, allowed precise CARS measurements. In respect to SRS, for which amplitude noise is more crucial than for CARS, the relative intensity noise (RIN) of the FOPA pulses was measured to −119 dBc∕Hz at a potential modulation frequency of 500 kHz. Alternatively, the DPA pulses exhibited a RIN of −141 dBc∕Hz, which should allow sensitive SRS measurements by detecting the stimulated Raman loss of the pump, if necessary, with a balanced detection scheme [2]. In order to extend the tuning range even further, especially into the CH stretch region around 3000 cm−1 , we tried to simply increase the FWM bandwidth by scaling up the pump power. However, this resulted in a spectrally broadened and structured signal pulse, due to saturation effects such as backconversion and to a higher spontaneous FWM background. An alternative mechanism to shift the gain of the FOPA toward higher wavenumbers is based on tuning the pump wavelength away from the ZDW into the normal dispersion regime of the PCF [11]. In [23], it was shown that, by tuning the pump wavelength by 6 nm across the emission bandwidth of ytterbium, wavenumbers across the CH stretch region can theoretically be reached with a similar PCF. However, we did not test this mechanism, as the grating used in the Fourier filter for the seed pulse showed an insufficient diffraction efficiency above 1300 nm.

Wavelength (nm)

-12 -18 1032 1033.5

Wavelength (nm)

-6 -12

1080

600 650 700 -1

1240

1305

1.64 ps

0.6 0.4

1.0

Power of FOPA pulses

0.2

Temporal walk-off of FOPA pulses

-18

Wavenumber (cm )

1182

10.0

1 (c)

0

1128

2 4 6 8

Time (ps)

Fig. 3. (a) Spectrum of the DPA pulses. (b) Exemplary spectrum of FOPA pulses. (c) Intensity crosscorrelation of FOPA and 400 fs long MOPA pulses, deconvolved with a 400 fs long Gaussian pulse.

0.1 0

400

800

1200

1600

0.0

Walk-off (ps/10 cm)

-6

10.3 cm

-1

Power (mW)

0.5 nm

0 (b)

Intensity (a.u.)

0 (a)

Spec. pow. dens. (dB/nm)

Spec. pow. dens. (dB/nm)

1035

2000

Wavenumber (cm-1)

Fig. 4. Measured tuning curve of the FOPA pulses (red curve), together with the calculated temporal walk-off between pulses of different wavelengths in the PCF (blue curve).

Vol. 40, No. 23 / December 1 2015 / Optics Letters

Spectral intensity (a.u.)

5450

1 CARS spectrum of Cinnamomum cassia Spontaneous Raman spectrum

1 -1

0.5

14 cm

0 1550 1750 0

0

400

800

1200

Wavenumber (cm-1)

1600

2000

Fig. 5. (a) Measured CARS spectrum of cinnamomum cassia (raw data) with a detailed view in the inset, together with (b) a corresponding measured spontaneous Raman spectrum.

Nevertheless, concerning the tuning range, this system is comparable with other advanced light sources proposed for CRS but offers about 40 times higher pulse energy as frequencydoubled Er-doped fiber lasers [19], about 10 times more peak power than similar systems based on FWM [23] and a tuning speed, which is about one order of magnitude higher than reached with temperature phase-matched OPOs [24]. In order to demonstrate the suitability of the system for narrowband CRS, we applied the laser source to measure the spectral fingerprint region of cinnamomum cassia via CARS. Thereby, the pulses of the DPA and the FOPA acted as the pump and Stokes pulses, respectively. Using the same microscopic setup as in [10], the CARS signal was detected in the forward direction and filtered out of other spectral components. By tuning the FOPA pulses as in Fig. 4, while monitoring their power to ensure a constant total excitation power of 5 mW, we obtained the spectrum of cinnamomum cassia, as shown in Fig. 5(a). The figure shows the raw measured data, divided by the power of the FOPA pulses. No further postprocessing was applied. Even though a CARS spectrum does not directly resemble a spontaneous Raman spectrum due to the complex nature of a CARS signal, comparing the CARS spectrum to the additionally shown spontaneous Raman spectrum of the same sample [Fig. 5(b)] shows that the measured CARS spectrum accurately comprises all vibrational lines. Furthermore, as can be seen in the inset of Fig. 5, the narrowest resolved line in the CARS spectrum had a width of 14 cm−1 , which depicts an upper limit for the spectral resolution of the light source. Although this forward CARS measurement did not reveal novel findings, it can be seen as a proof-of-principle demonstration of the capabilities of the light source. Especially together with advanced detection methods, such as epi- or polarization-resolved detection [1], more complex samples will become investigatable. In summary, we have presented an improved concept for the generation of broadband SC pulses, synchronized with tunable narrowband pulses. The system was based on a MOPA, which delivered pulses with 5 μJ pulse energy and 400 fs pulse duration. Focusing about 20% of the MOPA power into a GdVO4 crystal yielded a stable SC, ranging from less than 0.7 μm to more than 1.6 μm. Additionally, 3 ps long pulses were derived from the MOPA and amplified with a DPA scheme to 800 nJ. Frequency-tunable pulses with a bandwidth below 10 cm−1 and a pulse energy of up to 8 nJ were generated by FWM in a PCF, pumped with the DPA pulses. Seeding the effect of FWM with the long wavelength (>1 μm) part of the

Letter SC allowed us to scan a bandwidth of almost 2000 cm−1 in 25 ms. The vibrational CH stretch region is also theoretically reachable by tuning the wavelength of the DPA pulses. A repetition rate of 1 MHz of the whole system ensured an efficient generation and detection of CRS signals, while the impacts of thermal and nonlinear damage on the sample were both kept low. Due to its parameters, the presented light source is highly adapted to narrowband CRS methods and is, as previously shown, additionally well suited for broadband CRS methods. Moreover, offering synchronized pulses at three different wavelengths, the light source should even enable advanced CRS techniques such as time-resolved broadband CARS to investigate, for instance, processes such as vibrational dephasing [25]. The original MOPA pulses should certainly also be applicable to multiphoton-fluorescence and harmonic-generation imaging, offering the system for multimodal imaging. Funding. Cells-in-Motion Cluster of Excellence (EXC 1003 – CiM); Deutsche Forschungsgemeinschaft (DFG); University of Münster, Germany.

REFERENCES 1. M. Müller and A. Zumbusch, ChemPhysChem 8, 2156 (2007). 2. C. W. Freudiger, W. Yang, G. R. Holtom, N. Peyghambarian, X. S. Xie, and K. Q. Kieu, Nat. Photonics 8, 153 (2014). 3. C. L. Evans, E. O. Potma, M. Puoris’haag, D. Côté, C. P. Lin, and X. S. Xie, Proc. Natl. Acad. Sci. USA 102, 16807 (2005). 4. D. W. McCamant, P. Kukura, and R. A. Mathies, Appl. Spectrosc. 57, 1317 (2003). 5. B. Marx, L. Czerwinski, R. Light, M. Somekh, and P. Gilch, J. Raman Spectrosc. 45, 521 (2014). 6. Y. Fu, H. Wang, R. Shi, and J.-X. Cheng, Opt. Express 14, 3942 (2006). 7. C. Xu and F. W. Wise, Nat. Photonics 7, 875 (2013). 8. T. Shimada, W. Watanabe, S. Matsunaga, T. Higashi, H. Ishii, K. Fukui, K. Isobe, and K. Itoh, Opt. Express 13, 9869 (2005). 9. M. H. Niemz, Laser-Tissue Interactions (Springer, 1996). 10. S. Dobner and C. Fallnich, J. Chem. Phys. 140, 084201 (2014). 11. J. M. Dudley and S. Coen, Rev. Mod. Phys. 78, 1135 (2006). 12. M. Baumgartl, M. Chemnitz, C. Jauregui, T. Meyer, B. Dietzek, J. Popp, J. Limpert, and A. Tünnermann, Opt. Express 20, 4484 (2012). 13. D. Solli, G. Herink, B. Jalali, and C. Ropers, Nat. Photonics 6, 463 (2012). 14. S. Lefrancois, D. Fu, G. R. Holtom, L. Kong, W. J. Wadsworth, P. Schneider, R. Herda, A. Zach, X. Sunney Xie, and F. W. Wise, Opt. Lett. 37, 1652 (2012). 15. E. S. Lamb and F. W. Wise, Biomed. Opt. Express 6, 3248 (2015). 16. M. Bradler, P. Baum, and E. Riedle, Appl. Phys. B 97, 561 (2009). 17. K. L. Corwin, N. R. Newbury, J. M. Dudley, S. Coen, S. A. Diddams, K. Weber, and R. S. Windeler, Phys. Rev. Lett. 90, 1 (2003). 18. E. R. Andresen, J. Thøgersen, and S. R. Keiding, Opt. Lett. 30, 2025 (2005). 19. V. Kumar, M. Casella, E. Molotokaite, D. Polli, G. Cerullo, and M. Marangoni, J. Raman Spectrosc. 43, 662 (2012). 20. L. J. Kong, L. M. Zhao, S. Lefrancois, D. G. Ouzounov, C. X. Yang, and F. W. Wise, Opt. Lett. 37, 253 (2012). 21. G. Giraud, J. Karolin, and K. Wynne, Biophys. J. 85, 1903 (2003). 22. A. B. Fallahkhair, K. S. Li, and T. E. Murphy, J. Lightwave Technol. 26, 1423 (2008). 23. M. Chemnitz, M. Baumgartl, and T. Meyer, Opt. Express 20, 26583 (2012). 24. M. Jurna, J. P. Korterik, H. L. Offerhaus, and C. Otto, Appl. Phys. Lett. 89, 251116 (2006). 25. Y. J. Lee and M. T. Cicerone, Appl. Phys. Lett. 92, 10 (2008).

Light source for narrow and broadband coherent Raman scattering microspectroscopy.

We present a light source that is well adapted to both narrow- and broadband coherent Raman scattering (CRS) methods. Based on a single oscillator, th...
565B Sizes 0 Downloads 13 Views