Open Access
Add to BibSonomyAbstract
Due to its high molecular specificity, Raman spectroscopy is a well-established analytical tool. Usually the inelastically scattered Raman light is spectrally dispersed by a spectrometer. Here, we present an alternative method, using an optical fiber as dispersive element. As the group velocity within the fiber is wavelength-dependent, different Raman bands arrive at different times at the detector. In combination with time-correlated single-photon counting, Raman spectra can be measured in the time domain. As detector we implemented a Superconducting Nanowire Single-Photon Detector (SNSPD), which possesses a timing accuracy of about 20 ps. Within this contribution we show first results of Raman spectra measured in the time domain using gradient index fibers of varying length.
© 2015 Optical Society of America
1. Introduction
Raman Spectroscopy is a commonly used analytical method allowing the identification of organic, inorganic or biological compounds, because every material shows its own typical Raman spectrum [1]. In order to resolve such spectra, the inelastically scattered (Raman) light has to be analyzed by spectrometers. Usually, a grating or prism is deployed as dispersive element, translating the intensity of polychromatic light to a spatial coordinate as a function of wavelength. The spatially dispersed spectrum is measured using multichannel detector arrays (e.g. CCD cameras) or scanned with single point detectors, respectively.
For most applications, the main figure of merit of any spectrometer is its spectral resolution. However, higher wavelength resolution fundamentally implies lower signal-to-noise ratio since the available polychromatic signal is distributed to an increasing number of wavelength intervals. In case of grating spectrometers, a complex interplay between the diffractive power of the grating, the detector spacing and the width of the entrance slit defines the spectral resolution and the Signal-to-Noise Ratio (SNR). Altering the spectral resolution usually requires elaborate setup changes including grating exchange. The presence of movable parts makes a system sensitive to mechanical disturbances, e.g. vibrations induced by the surrounding. For applications that require mobility, systems without movable parts are preferable.
The aim of this work is to demonstrate an alternative method for wavelength resolved measurements of Raman spectra with the prospect of a less complex and more robust system setup [2]. For that, we exploit the wavelength dependent traveling time of light signals passing through optical fibers and measure the arrival time of photons by utilizing Time-Correlated Single-Photon Counting (TCSPC [3],). With this technique we are able to detect Raman scattered photons with different, wavelength dependent time delays. The repeated measurement of the same experiment provides correlation of single-photon events to specific time slots. Thus, the resulting histogram of all events represents the Raman spectrum on a time axis. The translation from the variation of the intensity of the optical signal in time into a wavelength resolved spectrum is done afterwards by postprocessing of the recorded data.
In case of such Fiber Dispersed Raman Spectroscopy (FDRS), the obtainable temporal resolution depends on the excitation wavelength, the pulse length, the temporal resolution of the detection setup including the TCSPC electronics and the single-photon detector itself and the dispersion and the length of the fiber. In principle, that approach promises to achieve an adjustable spectral resolution simply by installing optical fibers with different lengths. Neither the light throughput nor the alignment of the setup would be affected, which promises the construction of a very user-friendly and robust system.
In order to proof this and to explore its limitation, we deployed a single-photon detector with best possible timing accuracy. The use of a Superconducting Nanowire Single-Photon Detector (SNSPDs) with a timing accuracy down to less than 20 ps. allows us to study the fiber dispersion closely. Moreover, the use of relatively short fibers in the order of a few meters is an additional benefit.
Based on the principle of FDRS, in the following necessary fiber lengths as well as needed temporal resolution are estimated for the intended spectral resolution. Subsequently, the working principle of SNSPDs will be explained and their design possibilities are discussed. After a description of the complete setup, the first successful proof-of-concept is demonstrated. Based on these measurements, further improvements and future prospects are discussed considering potential point-of-care application of efficient and robust Raman spectrometers.
2. Basics
2.1. Transition time of light through optical fibers
For an optical fiber, the transit time τ of a pulse is determined by the photon group velocity vg [4] of the central core material
where L is the length of the fiber. The spectral dispersion of the fiber, commonly called material dispersion, is determined by the variation of the group velocity vg with wavelengthwhere c and λ are the speed and wavelength of light in vacuum, and n is the wavelength-dependent index of refraction. The Sellmeyer data of sample number 9 of Butov et al. [5] for the index of refraction of doped silica have been used to estimate the spectral resolution that can be obtained for a given fiber length.Figure 1(a) compares the arrival time of photons at the detector for different lengths of fiber. For the calculation of the data points, the corresponding Raman peaks of cyclohexane (see Fig. 5(a), left) were used, assuming 532 nm as excitation wavelength. The zero-point is set to the arrival time of the longest wavelength, i.e. at 629 nm as it arrives first at the end of the fiber. Figure 1(b) demonstrates that the spreading of the Raman bands can easily be changed by adjusting the fiber length. However, as we will show later, the basic principle “the longer the better” is not valid, because the fiber influences the laser pulses. The group velocity depends non-linearly on the wavelength and so does the transit time. The temporal resolution decreases for increasing wavelength. Figure 1(a) can be used to approximate the necessary fiber length for a given temporal resolution of the detection system (see paragraph 4.2). The curves in Fig. 1(b) show a spectral resolution of 1 nm or 3 nm for a mean wavelength of about 629 nm (a resolution of 1 nm corresponds to 25.2 cm−1 and 3 nm to 75.5 cm−1). With this, the necessary fiber length to resolve the double peak structure within the CH-stretching region of cyclohexane around 629 nm (see Fig. 5(a)) can be estimated.
Fig. 1 (a) Wavelength vs. transit time difference through different fiber lengths. Symbols correspond to Raman bands of cyclohexane, see spectrum in Fig. 5(a). (b) Difference in transit time for a wavelength difference for mean wavelength of 629 nm of 1 or 3 nm and fiber lengths of up to 300 m.
2.2. Operation principle and characteristics of implemented SNSPD
SNSPDs are well suited for single-photon detection [6,7] because of several features: (i) the detection efficiency (DE) can be as higher than 90% [8–11], (ii) the dark count rate (DCR) is typically less than 100 Hz at a relatively high DE [12], (iii) the response spectrum can span from visible light to infrared [12,13], and (iv) the temporal jitter can be as low as 16 ps [6,9,11,14]. However, the investigation of timing jitter is still subject of topical research. In consequence, strongly varying numbers for the timing jitter have been reported [15,16]. Due to these characteristics, SNSPDs are utilized in many fields where efficient, fast, and precise detection of single-photons is required. For example, SNSPDs were already applied in experiments in laser ranging [17] and fiber based temperature sensing experiments [18].
The principle of photon detection is as follows, see Fig. 2(a): The material is kept below its critical temperature (Tc) and a bias current (Ib) near the critical current (Ib = 90-95% Ic) is applied. A photon hits the meander line and is absorbed. The energy of the photon leads to a breaking of Cooper pairs and local heating, thus a small area gets normal conducting. The normal conducting area is indicated by the darker color. The normal conducting domain spreads until a normal conducting belt arises and a voltage pulse can be measured. With most of the bias current flowing through the amplifier, the non-superconducting region cools and returns to the superconducting state.
Fig. 2 Working Principle of SNSPD (a) Single-photon detection scheme: First, a photon arrives at the detector (left), is absorbed (middle) and leads to a normal conducting belt over the whole wire width. (b) Scanning electron micrograph of an SNSPD, the sensitive area (meander) is about 5 x 5 µm2
Obviously it is not possible to directly observe the temporal behavior of optical signals in single-photon regime. Hence, TCSPC (see section 3.3) is used to determine the time response.
The implemented SNSPD (see Fig. 2(b)) was fabricated using a standard fabrication process at the KIT (Karlsruhe Institute of Technology, Karlsruhe/Germany) [19,20]. The detector was optimized for the highest achievable temporal resolution instead of highest detection efficiency, for which usually film thicknesses of less than 5 nm are required. In order to achieve this, the critical current must be as high as possible [21]. Hence, we used a film thickness of 6.5 nm. Additionally this offers the advantage that only a two-stage amplifier chain is sufficient for the measurements (see also discussion in section 4.1).
At the operating point of 0.9 Ic, the used detector has ~1% detection efficiency (DE) at 532 nm. As is shown in [19] for a similar device (slightly thinner film), the DE should be nearly constant for the observed wavelength range. The dark count rate (DCR) at this operating point was <1 s-1. However, we should mention, that the characterization was done at a different optical setup than the present one. Hence, the achievable values might vary slightly from the here given values.
3. Experimental environment
The measurement setup as shown in Fig. 3 can be separated into three parts: the excitation path with the pulsed laser and optical fibers, the detection path with the cryogenic system, and the TCSPC registration.
Fig. 3 Measurement setup consisting of three parts: the excitation path with the pulsed laser and optical fibers, the detection path with the cryogenic system, and the TCSPC readout (see text above).
3.1. Raman spectroscopy setup
The Raman spectroscopy setup (see Fig. 3) comprises an Nd:Vanadate picosecond (ps)-laser system (High Q picoTrain, High Q Laser, Austria) with a repetition rate of 80 MHz and a pulse length of a about 7 ps. The 532 nm second harmonic of the ps-laser with a resulting spectral bandwidth of about 0.3 nm, which corresponds to 10.6 cm−1, is used for Raman excitation. The laser pulses are directed to the optical system using a short fiber (about 100 cm). After collimation (collimator 1) and passing a laser line filter (filter 1), the pulses are redirected (mirror 1, 2) and focused onto the sample with a 20x microscope objective. This objective also collects the Raman backscattered light (indicated by the dotted dark green line in Fig. 3). The dichroic mirror allows the Raman scattered light to pass the mirror whereas reflected light from the excitation laser cannot pass. An additional notch filter (filter 2) further reduces the number of excitation photons in the signal path in order to suppress fiber based background generation [22] in the collecting fiber. Subsequent lenses (lens 1, 2) are used in a telescope configuration to reduce the beam diameter and to improve the confocality of the whole system. Afterwards the Raman scattered light is coupled (collimator 2) into the collection fiber (fiber 2). This fiber acts as the dispersive element and has a length of several meters, depending on the intended temporal resolution and the dispersion of the actual type of fiber. It can be connected directly via FC/PC adapters to a spectrometer or the cryogenic part of the measurement system.
3.2. Cryogenic system
Inside the cryostat, a fiber guides the Raman photons to the cold stage where the SNSPD is located. The fiber is fixed to a micro-focus optics, which is attached to a three axis piezoelectric stage. The spot on the detector has a minimum diameter of 50 µm. In contrast to that, the sensitive area of the SNSPD is only 5 x 5 µm2, i.e. only a small fraction of the collected Raman photons are currently detected by the system.
In order to reduce thermal fluctuations, the detector area is shielded by aluminum with only a small hole in front of the detector to allow the focused light to reach the detector. Sufficient cooling and with that a low dark count rate [12] is ensured by gluing the detector to a copper holder by conductive silver. Via bonding wire, the detector is connected to a bias tee, where a DC-bias source and RF readout of the detector are attached. The complete holder is mounted onto the cold plate of a liquid helium bath cryostat [19].
3.3. Time-Correlated Single-Photon Counting (TCSPC)
TCSPC is based on the detection of single-photons of a periodic signal, on the measurement of the detection times and on the reconstruction of the spectrum from the individual time measurements. The principal TCSPC components are schematically shown in Fig. 3.
To apply TCSPC for resolving optical signals in the single-photon range, the experiments must be repeated many times under the same conditions. Hence, a pulsed optical setup (see Fig. 3) is needed. The laser driver also provides the electrical “sync” signal for the photon arrival time measurement. This trigger signal is also fed to the TCSPC electronics (Simple-Tau-130-EM-DX, Becker & Hickl, Germany).
The signal from the SNSPD is amplified by a two-stage amplifier chain that offers an effective bandwidth of about 2.1 GHz. The signal from the input is subsequently fed to a time-to-amplitude converter (TAC), operating in the reversed start-stop configuration. Here, the time measurement is started when a photon is detected and stopped with the next reference pulse from the laser. As a consequence, the time axis is reversed after the completion of a measurement. Making use of the statistical photon distribution during the signal, a histogram is created, which represents the original optical signal.
4. Results and discussion
4.1. Fiber characterization
For the proof of principle tests, we used a gradient index fiber with a core diameter of 50 µm (Optigrade 550 OM4 standard, j-fiber, Jena/Germany). For the design wavelength [23,24] of the fiber at about 850 nm all modes travel with the same velocity; however, differing wavelengths result in different velocities. Autocorrelation measurements at 690 nm (Autokorrelator PulseCheck 700-1100 nm, APE Berlin/ Germany) showed that a 6 ps excitation pulse broadened to a pulse length of 19 ps when travelling through 45 m of an OM4 fiber.
However, this is still acceptable for initial tests; especially taking into account the benefit of simplification of the collection of photons and their coupling into the fiber due to the relatively large core diameter and a numerical aperture of 0.22.
4.2. Temporal accuracy of the system
The achievable temporal resolution of the system was determined by the characterization of every single component that leads to inaccuracy due to a temporal jitter caused by finite slope and bandwidth limitations. The total jitter is mainly the sum of three different contributions following the Gaussian error propagation: (i) contribution of the detector itself, depending on the internal processes that lead to the development of the normal conducting domain; (ii) contribution due to the pulse to pulse jitter of the laser and (iii) contribution from the TCSPC readout electronics plus the used amplifiers.
All three components show a Gaussian distribution of their jitter. Hence, the overall system jitter can be derived from and as follows:
is the standard deviation of a Gaussian distribution and can be converted into the commonly used full width at half maximum (FWHM) value.To make an estimation of the contribution of the SNSPD and the other components, we characterized the TCSPC system. 18.4 ps (FWHM) originate from the laser and are calculated by the pulse-to-pulse jitter from the data sheet ( ± 50 kHz at a repetition rate of 80 MHz).The electronic part of the system was measured separately. First, the TCSPC electronics itself was characterized by using the instrument response function (IRF) measuring system proposed by Becker & Hickl [25]. The signal from a pulse generator was split into two paths. One cable was connected to the path leading to the CFD-input of the electronics (in our experiments connected to the SNSPD) and has a fixed length of 1 m. The second path had a length of 2.2 m which resulted in a maximum observable time frame of about 6 ns. To satisfy the requirements of a minimum pulse width and amplitude of the electronics, we adjusted the settings accordingly at the pulse generator. As a test we checked the form and width of the resulting histogram and compared the results with that of the manufacturer. Both histograms showed a Gaussian distribution and had a FWHM of 6.9 ps.
Further, the contribution of the 2-stage amplifier chain was measured electronically with a slightly adjusted setup. The amplifiers were added into the later signal path of the SNSPD (CFD input). To make sure that the amplifiers were used in their specifications and do not reach saturation, the signal was damped accordingly. The FWHM value of the response histogram was increased to about 11 ps, see Fig. 4.
Fig. 4 Measured histograms of the timing jitter of the TCSPC electronics (red curve) with a FWHM value of 7 ps. The histogram in black shows the result if amplifiers are implemented with FWHM value of 11 ps.
In order to estimate the contribution of the SNSPD, we needed the histogram of a measurement with the SNSPD connected to the amplifier chain. Hence, we used a very short pulsed laser (pulse length of ~150 fs) with a repetition rate of 75 MHz. Due to this short pulse width, the contribution to the IRF is negligible. As trigger, the same optical pulse that triggered the SNSPD was used. Hence, there is no pulse-to-pulse jitter. With that, we achieved a timing resolution of the IRF of 22.5 ps (FWHM). Using Eqs. (3) and (4), setting σLaser = 0, the contribution of the SNSPD alone can be calculated to 19.8 ps.
The measured timing jitter value of 19.8ps stems from an older experiment. Due to material degradation effects, the same detector achieved a slightly lower critical current during the Raman experiments. Since the timing jitter depends on the absolute achievable value of Ic [21], the contribution to the jitter from the SNSPD is slightly higher here, at ~22 ps (FWHM).
Taking now into account jitter values of FWHMTCSPC + amplifiers ≈11 ps, FWHMSNSPD ≈22 ps and FWHMlaser ≈18.4 ps, the system itself exhibits a total jitter of FWHMtotal ≈31 ps.
4.3. Comparison of spectra over wavelength and timescale
For our Raman measurements, fiber lengths of about 15, 30 and 45 m and different test samples (bulk materials, liquids in a quartz cuvette (Hellma Analytics, Germany) or polymer (Eppendorf Safe-Lock tubes, Sigma-Aldrich, USA)) were used.
In Fig. 5, a comparison of spectrometer (Shamrock 303i, Andor, UK) based Raman spectra (intensity over wavenumber/wavelength) on the left versus fiber dispersed spectra (intensity over wavenumber/time) on the right is plotted, measured with the here presented method. For the measurements, the collection fiber (fiber 2 in Fig. 3) was either connected to the spectrometer or the cryostat. This way, the experimental conditions were kept constant for both experiments. Exemplarily, we show the results for three different materials, namely two liquids (cyclohexane (topmost row) and methanol (middle row)) and one solid (polypropylene (lowest row)). For these measurements, about 45 m of a graded index multimode fiber were used as dispersive element. The laser power on the sample was between 11 and 13 mW.
Fig. 5 Measurement results: left: spectrometer based spectra, i.e. intensity vs. wavenumber/wavelength, right side fiber dispersed spectra with 45 m of dispersive fiber, i.e. intensity (≡ counts) vs. wavenumber/transit time, Rows from top to bottom (measurement time 300 s) (a) cyclohexane (inset: measurement time 60 s), (b) methanol and (c) polypropylene (Eppendorf safe-lock tube).
The transition time through the fiber is decreasing for increasing wavelengths. Hence, the time dependent spectrum is a mirrored image compared to the spectrometer based one. In order to simplify the comparison, the time axis is inverted. For a better overview and comparison between the different measurement results, t = 0 s corresponds to ~629 nm (as discussed in section 2.1), which is the upper boundary of the measured wavelength range and thus arrives first at the detector. When we compare the spectrometer based (integration time 1 s) and the fiber dispersed spectra, several things can be observed. First of all, the fiber dispersed spectrum (Fig. 5, right) resembles the equivalent classical spectrum, see left column of Fig. 5, but is resolved over timescale (upper x-axis). This shows that it is possible to record Raman spectra by utilization of our fiber dispersed Raman spectroscopy technique. Besides, not only the Raman bands appear at the expected position, but in contrast to [26] also the different peak intensities are reproduced.
For all presented fiber dispersed spectra, the integration time was 300 s. The only exception here is the inset in Fig. 5(a) on the right. In that case, the measurement time was only 60 s, resulting in a decrease of the signal to noise ratio (S/N) from 737 to a still useful value of 289 at 2900 cm-1. Here, we use the definition that the S/N ratio is defined as the difference of peak and background signal, divided afterwards by the square root of the background signal (definition used by Horiba Scientific). At the current state of development, the required measurement time is obviously impractically long for the targeted applications. However, many optimization possibilities concerning detector [9], the optical system and the fiber allow for future improvement.
The translation of the time-scale to wavenumber-scale (lower x-axis used for fiber dispersed spectra of Fig. 5) is done by a calibration curve. In our case, the cyclohexane-sample was used to calculate the calibration curve, shown in Fig. 6. The black solid line represents the calibration curve itself, the red triangles the Raman bands of the cyclohexane and the black squares correspond to the Raman band positions of methanol.
Fig. 6 Calibration curve for correlation of wavenumber and time-axis of fiber dispersed measurements. Raman band positions of cyclohexane and methanol are correctly represented.
5. Discussion
Notwithstanding the successful proof-of-concept, a detailed analysis of the recorded fiber dispersed spectra reveals a deviation from the expected spectral resolution. The example of cyclohexane (compare Fig. 5(a)) shows, that the closely spaced Raman bands at 627.7 and 630.7 nm (spectral distance ~3 nm) cannot be resolved by using 45 m of the OptiGrade 550 fiber. This can be explained neither by the spectral width of the pulsed laser excitation (compare to section 3.1) nor an insufficient temporal resolution. At 629 nm, a wavelength separation of ~3 nm would correspond to about 45 ps temporal distance, which is within the estimated timing accuracy of our detection system (compare to section 4.2).
A reasonable explanation for that is intermodal dispersion that occurs in the used fiber. Whilst the chromatic dispersion of the fiber material is actively used to our advantage, intermodal dispersion due to different mode velocities also takes place [24,27].
During the measurements that are shown in Fig. 5, the count rates were between 1.5 - 4 kHz for the different substances. Still, the achieved maximum count number is much lower for the SNSPD measurements despite the much longer measurement time of 300 s than that of the spectrometer (measurement time of 1 s). This is caused on one hand by the high differences in quantum efficiency of the used detector types. While the CCD camera has a quantum efficiency of more than 90%, the SNSPD had an efficiency of less than 1%.). On the other hand, only a small fraction of the collected photons from the experiment could be used for the detection itself. This is due to the mismatch of the detector area of 5 x 5siore µm2 and a minimum spot diameter of 50 µm. While the CCD could use all photons that were in the fiber, most of the light was lost in the area around the detector. But this trade off was made deliberately (see also section 4.1).
In Fig. 7 measurement results with three different fiber lengths are shown. As expected, the Raman scattering signal is stretched to a longer time frame when the fiber length is increased. A comparison between the measurements with 15 and 30 m fiber length clearly shows that the resolution is improved by the use of the longer fiber. Especially in the region between 500 and 550 nm, where four separated Raman bands in the spectrum (marked by red arrows in Fig. 7) are expected, it can be seen that these four peaks are already well separated in the time domain measurement with 30 m fiber. Further lengthening of the fiber to 45 m leads to the expected additional signal stretch. However, the resolution is not improved further as can be clearly seen at the constant modulation depth between the two peaks.
Fig. 7 Comparison of measurements with different fiber lengths with cyclohexane as sample, black = 15 m, red = 30 m, blue = 45 m. Red arrows show the area where the improvement of the resolution by the use of a longer fiber can be seen. The separation of the two Raman bands is better for the 30 m long fiber. Further lengthening to 45 m does not improve the resolution anymore and leads not only to a signal stretch but also to a broadening of the Raman bands.
In order to avoid the influence of intermodal dispersion, in future implementations of the FDRS concept the multimode fiber can be replaced by a single-mode fiber [26]. With that, the temporal resolution of 31 ps, as determined in section 4.2, should be achieved. An estimation of the expected improvement of the temporal resolution, and with that of the wavelength resolution, is shown in Fig. 8. Here, calculated spectra are compared assuming the current temporal resolution (about 47 ps) and the expected temporal resolution using a single mode fiber (31 ps), respectively. Shown is the temporal range around t = 0, which correlates to about 629 nm in the spectrum. In this range the two Raman bands at 627.7 and 630.7nm cannot be distinguished, as it is correctly reproduced by our calculated result with 47 ps temporal resolution. If we assume a temporal resolution of 31 ps, a separation of these two Raman bands is already possible.
Fig. 8 Calculation of Raman peaks with different temporal resolution of the complete system; red: the present resolution of about 47 ps; black: the optimized resolution of 31 ps.
However, the limits of this technique are given by the maximum count rate (i.e. depends on the detection efficiency), the jitter of the whole detection system and the repetition rate of the system which defines the maximum observable time window. For TCSPC, a high repetition rate of the excitation process is desirable for short acquisition times. But this repetition rate of the excitation laser limits the maximum achievable temporal resolution by the time interval between two pulses, 12.5 ns in our case. The current temporal resolution is determined by the overall system jitter which sums up to about 30 ps with the used setup (see section 4.2). As a rough estimation, the maximum resolution of the here described setup would allow about 416 of such time slots fitting into the 12.5 ns of the pulse-to-pulse distance. This results in a maximum resolution for Raman spectra (0 – 3400 cm−1) of about 8 cm−1 which is comparable to commercial setups. There are several ways to improve that number. One is to use a lower repetition rate of the laser in combination with an adapted length of the fiber in order to stretch the signal to the longer time window. Also the pulse-to pulse-jitter could be eliminated by adding an appropriate (optical) delay so that the optical pulse that caused the SNSPD pulse and that of the laser trigger are identical. This would lead to a further reduced IRF. The other way for improvement is the optimization of the detector [9].
6. Conclusion
We have successfully demonstrated a new concept of a Raman spectrometer based on dispersion in optical fibers. As a proof of concept we performed Raman measurements on selected samples (cyclohexane, methanol, polypropylene) on pulsed laser excitation at 532 nm wavelength. The scattered Raman signal was dispersed in a graded index fiber (OptiGrade 550) and analyzed with a SNSPD in combination with TCSPC. The achieved temporal resolution of 47 ps corresponds to a wavelength resolution of 3.42 nm (wavenumber resolution 149.25 cm−1) at 629 nm.
The current result is mainly limited by intermodal dispersion of the used fiber with a length of 45 m, which deteriorated the intrinsic instrument response function from 31 ps to 47 ps. That issue can be solved by using a single-mode fiber, which would improve the wavelength resolution to 2.25 nm (wavenumber resolution 56.7 cm−1) at 629 nm without any further changes of the setup. The relatively long acquisition time of spectra is caused by a small quantum efficiency of less than 1% of the used SNSPD and due to the size mismatch between the used fiber and the detector. With an adapted setup, e.g. plasmonic structures [28] or utilizing waveguide coupling [9], the detection efficiency can be improved to values up to more than 90% resulting in a significant reduction of the acquisition time. With the higher DE, the characteristics for the whole measurement system would be improved drastically.
Future developments of the concept would allow for a spectral resolution which can be increased up to the fundamental limit of the excitation bandwidth (with the actual setup ~10 cm−1) simply by increasing the length of the dispersive fiber. Here, the use of a SNSPD with high temporal resolution allows keeping the fiber length as short as possible, thus minimizing absorption losses in the fiber. Our own results as well as findings from other groups show that there might be still potential for even lower timing jitter, whereas the expected physical limit is in the range of the electron-electron interaction time of about 10 ps [29]. With this knowledge, the given fiber dispersion and the excitation wavelength, the optimal fiber length for a specific wavenumber resolution can be estimated.
In principle, also conventional detectors such as avalanche diodes or photomultiplier tubes can be used instead of SNSPDs, whereas their reduced temporal resolution can be compensated by longer fibers. However, using SNSPDs is still advantageous for another reason. The acquisition time of a spectrum for a given spectral resolution and signal-to-noise ratio is mainly defined by the repetition rate of the pulsed excitation. Whereas the pulse rate of commercial laser modules can be in the range of 100 MHz, conventional detectors are typically limited to count rates of max. 10 MHz. In comparison, SNSPDs enable extremely fast count rate up to several hundred Megahertz [14,19], which would shorten the acquisition time at least by one order of magnitude. Admittedly, to exploit that performance, also advancement of current TCSPC electronics is necessary, whose performance was currently developed to match only the capabilities of conventional detectors.
For future applications of a Raman spectrometer based on superconducting detectors it is also important to solve issues related to cryogenic cooling. Although such cooling up to now was treated as a serious handicap, current developments of compact, efficient cryo-coolers based on the Stirling principle enable efficient plug-in instruments [30], paving the road to user-friendly and robust devices.
At the same time, using conventional detectors, a parallel category of instruments with higher robustness at the expense of longer acquisition times could be preferential in point-of-care applications where spectrometers have to be used outside a controlled laboratory environment. Here, the inherent ruggedness against loss of adjustment is a substantial advantage over conventional spectrometers based on gratings or prisms.
Acknowledgment
We thank the Thüringer Kultusministerium TMBWK for supporting us by funding the project FluoRaman FKZ 11030-715 and the Bundesministerium für Bildung und Forschung BMBF for the project Fiber Health Probe 13N12525.
References and links
1. M. Schmitt and J. Popp, “Raman spectroscopy at the beginning of the twenty-first century,” J. Raman Spectrosc. 37, 20–28 (2006).
2. G. Petrov, V. Yakovlev, V. Shcheslavskiy, “Raman spectroscopy without spectrometer,” CLEO:2011 - Laser Applications to Photonic Applications, CLEO:2011 - Laser Applications to Photonic Applications, Optical Society of America, 2011, PDPB6.
3. W. Becker, A. Bergmann, M. A. Hink, K. König, K. Benndorf, and C. Biskup, “Fluorescence lifetime imaging by time-correlated single-photon counting,” Microsc. Res. Tech. 63(1), 58–66 (2004), doi:. [CrossRef] [PubMed]
4. W. B. Whitten and H. H. Ross, “Fiber Optic Waveguides for Time-of-Flight Optical Spectrometry,” Anal. Chem. 51(3), 417–419 (1979). [CrossRef]
5. O. V. Butov, K. M. Golant, A. L. Tomashuk, M. J. N. van Stralen, and A. H. E. Breuls, “Refractive index dispersion of doped silica for fiber optics,” Opt. Commun. 213(4-6), 301–308 (2002), doi:. [CrossRef]
6. R. H. Hadfield, “Single-photon detectors for optical quantum information applications,” Nat. Photonics 3(12), 696–705 (2009), doi:. [CrossRef]
7. C. M. Natarajan, M. G. Tanner, and R. H. Hadfield, “Superconducting nanowire single-photon detectors: physics and applications,” Supercond. Sci. Technol. 25(6), 063001 (2012), doi:. [CrossRef]
8. S. Miki, T. Yamashita, H. Terai, and Z. Wang, “High performance fiber-coupled NbTiN superconducting nanowire single photon detectors with Gifford-McMahon cryocooler,” Opt. Express 21(8), 10208–10214 (2013). [CrossRef] [PubMed]
9. W. H. P. Pernice, C. Schuck, O. Minaeva, M. Li, G. N. Goltsman, A. V. Sergienko, and H. X. Tang, “High-speed and high-efficiency travelling wave single-photon detectors embedded in nanophotonic circuits,” Nat Commun 3, 1325 (2012), doi:. [CrossRef] [PubMed]
10. F. Marsili, V. B. Verma, J. A. Stern, S. Harrington, A. E. Lita, T. Gerrits, I. Vayshenker, B. Baek, M. Shaw, R. P. Mirin, and S. W. Nam, “Detecting single infrared photons with 93% system efficiency,” Nat. Photonics 7(3), 210–214 (2013). [CrossRef]
11. C. Schuck, W. H. P. Pernice, O. Minaeva, M. Li, G. Gol’tsman, A. V. Sergienko, and H. X. Tang, “Matrix of integrated superconducting single-photon detectors with high timing resolution,” IEEE Trans. Appl. Supercond. 23(3), 2201007 (2013), doi:. [CrossRef]
12. M. Hofherr, D. Rall, K. Il’in, A. Semenov, H.-W. Hübers, and M. Siegel, “Dark count suppression in superconducting nanowire single photon detectors,” J. Low Temp. Phys. 167(5), 822–826 (2012). [CrossRef]
13. M. Hofherr, D. Rall, K. Il’in, M. Siegel, A. Semenov, H.-W. Hübers, and N. A. Gippius, “Intrinsic detection efficiency of superconducting nanowire single-photon detectors with different thicknesses,” J. Appl. Phys. 108(1), 014507 (2010). [CrossRef]
14. M. Tarkhov, J. Claudon, J. P. Poizat, A. Korneev, A. Divochiy, O. Minaeva, V. Seleznev, N. Kaurova, B. Voronov, A. V. Semenov, and G. Gol’tsman, “Ultrafast reset time of superconducting single photon detectors,” Appl. Phys. Lett. 92(24), 24111 (2008). [CrossRef]
15. D. Liu, S. Miki, T. Yamashita, L. You, Z. Wang, and H. Terai, “Multimode fiber-coupled superconducting nanowire single-photon detector with 70% system efficiency at visible wavelength,” Opt. Express 22(18), 21167–21174 (2014). [PubMed]
16. D. Rosenberg, A. J. Kerman, R. J. Molnar, and E. A. Dauler, “High-speed and high-efficiency superconducting nanowire single photon detector array,” Opt. Express 21(2), 1440–1447 (2013). [CrossRef] [PubMed]
17. S. Chen, D. Liu, W. Zhang, L. You, Y. He, W. Zhang, X. Yang, G. Wu, M. Ren, H. Zeng, Z. Wang, X. Xie, and M. Jiang, “Time-of-flight laser ranging and imaging at 1550 nm using low-jitter superconducting nanowire single-photon detection system,” Appl. Opt. 52(14), 3241–3245 (2013).
18. M. G. Tanner, S. D. Dyer, B. Baek, R. H. Hadfield, and S. Woo Nam, “High-resolution single-mode fiber-optic distributed Raman sensor for absolute temperature measurement using superconducting nanowire single-photon detectors,” Appl. Phys. Lett. 99(20), 201110 (2011), doi:. [CrossRef]
19. J. Toussaint, R. Grüner, M. Schubert, T. May, H.-G. Meyer, B. Dietzek, J. Popp, M. Hofherr, M. Arndt, D. Henrich, K. Il’in, and M. Siegel, “Superconducting single-photon counting system for optical experiments requiring time-resolution in the picosecond range,” Rev. Sci. Instrum. 83(12), 123103 (2012), doi:. [CrossRef] [PubMed]
20. A. Semenov, B. Günther, U. Böttger, H.-W. Hübers, H. Bartolf, A. Engel, A. Schilling, K. Il’in, M. Siegel, R. Schneider, D. Gerthsen, and N. A. Gippius, “Optical and transport properties of ultrathin NbN films and nanostructures,” Phys. Rev. B 80(5), 054510 (2009), doi:. [CrossRef]
21. L. You, X. Yang, Y. He, W. Zhang, D. Liu, W. Zhang, L. Zhang, L. Zhang, X. Liu, S. Chen, Z. Wang, and X. Xie, “Jitter analysis of a superconducting nanowire single photon detector,” AIP Advances 3(7), 072135 (2013), doi:. [CrossRef]
22. I. Latka, S. Dochow, C. Krafft, B. Dietzek, and J. Popp, “Fiber optic probes for linear and nonlinear Raman applications - current trends and future development,” Laser Photon. Rev. 7(5), 698–731 (2013), doi:. [CrossRef]
23. M.-J. Li and D. A. Nolan, “Optical transmission fiber design evolution,” J. Lightwave Technol. 26(9), 1079–1092 (2008), doi:. [CrossRef]
24. R. Olshansky and D. B. Keck, “Pulse broadening in graded-index optical fibers,” Appl. Opt. 15(2), 483–491 (1976). [CrossRef] [PubMed]
25. W. Becker, The bh TCSPC Handbook, 4th Edition (Becker Hickl GmbH, 2010) .
26. Z. Meng, S. Cheng, G. I. Petrov, J. A. Jo, and V. V. Yakovlev, “Raman spectroscopy using time-correlated photon-counting detection,” Proc. SPIE 8572, 85721G (2013). [CrossRef]
27. N. R. Teja, M. A. Babu, T. Prasad, and T. Ravi, “Different types of dispersion in an optical fiber,” Int. J. Sci. Res. Pub. 2(12), 1–5 (2012).
28. M. Csete, Á. Sipos, A. Szalai, F. Najafi, G. Szabó, and K. K. Berggren, “Improvement of infrared single-photon detectors absorptance by integrated plasmonic structures,” Scient. Reports 2, (2013).
29. K. Il’in, M. Lindgren, M. Currie, A. Semenov, G. Gol’tsman, R. Sobolewski, S. Cherednichenko, and E. Gershenzon, “Picosecond hot-electron energy relaxation in NbN superconducting photodetectors,” Appl. Phys. Lett. 76(19), 2752 (2000), doi:. [CrossRef]
30. C. Wang, G. Thummes, and C. Heiden, “A two-stage pulse tube cooler operating below 4 K,” Cryogenics, Bd. 37(3), 159–164 (1997). [CrossRef]
