Open Access
Add to BibSonomyAbstract
A low-cost, compact Raman spectrometer suitable for the on-line water monitoring applications is explored. A custom-designed concave grating for wavelength selection was fabricated and tested. The detection of the Raman signal is accomplished with a time-gated single photon avalanche diode (TG-SPAD). A fixed gate window of 3.5ns is designed and applied to the TG-SPAD. The temporal resolution of the SPAD was ~60ps when tested with a 7ps, 532nm solid-state laser. To test the efficiency of the gating in fluorescence signal suppression, different detection windows (3ns-0.25ns) within the 3.5ns gate window are used to measure the Raman spectra of Rhodamine B. Strong Raman peaks are resolved with this low-cost system.
© 2014 Optical Society of America
1. Introduction
Raman spectroscopy is a powerful characterization technique for the analysis of chemical compositions. It has been widely used for pharmaceutical and environmental applications [1, 2]. Since water is a weak Raman scatterer, Raman spectroscopy is more applicable in detection of chemical or biological contaminants in water samples. The main issues associated with the Raman spectra detection are the low efficiency of Raman scattering and the strong accompanying fluorescence. To achieve high detection efficiency, commercial Raman spectrometers are equipped with expensive and sophisticated wavelength selectors and detectors, which limit most of the commercial Raman spectrometer mainly for laboratory use. Portable Raman spectrometers are also commercial available, such as the SciAps (Inspector 300) and Snowy Range Instruments (CBEx). However, most of these Raman spectrometers cannot address the fluorescence problem, unless a near infrared excitation source is used (Inspector 500, CBEx 1064). In this work, we target a low-cost, portable Raman spectrometer to explore the potential of Raman spectroscopy for on-line water monitoring applications. In particular, we focus on the miniaturization of the spectrometer and the suppression of the background fluorescence.
The wavelength selector (filter or monochromator) is an essential part of a Raman system. It determines the spectral resolution, but it is also the largest component of the entire Raman instrument. A monochromator typically contains a planar grating for wavelength separation and two mirrors for light collimation and focusing. The optical paths between the mirrors and the grating determine the size of the system. To minimize the system`s size, we designed and fabricated a concave grating. Unlike the planar grating, the concave grating can perform both functions of light separation and focusing without the need for extra mirrors [3]. As a consequence, a concave grating based system is more compact and the size of the Raman spectrometer is significantly reduced.
An important challenge of Raman spectroscopy applications is how to detect the weak Raman signal if strong fluorescence background is present. Given that the excitation band of most of the natural fluorophores is in the ultraviolet region or visible region, a direct and efficient way is to use longer wavelength excitation. A typical application is the Fourier transform Raman spectrometer (FT-Raman), which is able to obtain fluorescence free Raman signal by using a near infrared (NIR) source (1064nm) [4]. However, considering the cost of NIR detectors and the dependence of the Raman scattering efficiency on the excitation wavelength (λ−4), a short wavelength excitation is preferred for a low-cost and compact system. To suppress the background fluorescence, and also to guarantee a high enough Raman scattering efficiency, the time gated detection is used in our system.
The fluorescence signal rejection of the time-gated detection technique depends on the different temporal responses of Raman scattering and fluorescence emission to the excitation source. Raman scattering is an instantaneous response to a laser excitation, while fluorescence is emitted with a temporal distribution characterized by the so-called fluorescence lifetime, which ranges from hundreds of picoseconds to tens of nanoseconds, depending on the fluorophores. Therefore, if we excites a sample with very short laser pulse, then the fluorescence reduced Raman spectra can be obtained by synchronizing the detection with the short excitation.
Ultra-fast gating (<10ps) has been achieved by the Kerr gate [5], but this is mainly used in laboratory because of the complex system setup. Both intensified charge-coupled devices (ICCDs) [6] and photomultiplier tubes (PMTs) [7] have been used for the time-gated Raman spectrometer with sub-nanosecond gating. These detectors are extremely sensitive, but are also expensive and not easy to integrate with the electronic control circuits. To reduce the cost and improve the integration capability of the detector, a time-gated single photon avalanche diode (TG-SPAD) based CMOS pixel circuit is used.
With the ability for single photon detection and sub-nanosecond temporal response, the CMOS SPAD has been used in various applications such as the fluorescence lifetime imaging (FLIM) [8], near infrared spectroscopy (NIRS) [9] and light detection and ranging (LIDAR) [10]. The fast gating capability makes the SPAD a low cost alternative to ICCDs and PMTs. The TG-SPADs have also been proposed recently for fluorescence rejection in Raman spectroscopy. Controlled by external pulse generators, sub-nanosecond detection window were achieved in [11–13]. The time positions of recorded photons were also measured in [13] by an off-chip time-to-digital converter (TDC). Comparing to [11] and [13], the SPAD of this work is gated using on-chip pulse generators with a fixed gate window of 3.5ns, which offers a minimum detection windows of ~200ps. The temporal resolution of the TG-SPAD was measured to be ~60ps with a very short laser pulse. The pixel circuit is specially designed to provide for the measurement of the photon arriving time within the 3.5ns gate window. The concave grating was fabricated in-house by a low-cost holographic process, which further reduces the cost of the Raman system. Combining the concave grating and the TG-SPAD, a Raman system was built, and strong Raman peaks were resolved by this low-cost and small-size system.
2. System design
Figure 1 shows the block diagram of a concave grating based Raman spectrometer. The incident light from a laser passes through a dichroic mirror and is focused onto the sample by a lens. The fluorescence, Stokes, anti-Stokes and Rayleigh scattered signals are collected by the lens and then separated by the dichroic mirror, with shorter wavelength components (anti-Stokes Raman and Rayleigh) transmitted and longer wavelength components (Stokes scattering and fluorescence) reflected [14]. The reflected signal is then coupled into an optical fiber by the second lens and then forwarded to the Rowland circle of the concave grating. As shown in Fig. 1, the Rowland circle is a virtual circle, internally tangential to the concave grating, and with a diameter equal to the radius of the grating curvature. In Rowland configuration, when the entrance slit is positioned on the Rowland circle, the diffraction beam from the grating will also be focused onto the Rowland circle, but at different positions according to the wavelength. A detector is positioned on the Rowland circle to acquire the spectrum. Taking into account the curved focal plane of the Rowland circle and the planar active surface of the SPAD, a portion of the light on the SPAD that is measured is inevitably out of focus. This degrades the spectral resolution and broadens the spectral lines.
Based on Fig. 1, design considerations of the Raman spectrometer include the excitation source, illumination and collection optics, wavelength selector and detection system. Considering the wavelength dependence of Raman scattering efficiency and the spectral response of CMOS SPADs [15, 16], a 532nm solid state laser is used in the test setup. The Raman shifts of most chemical and biological samples are in the range of 500cm−1 to 2000cm−1, corresponding to wavelength range from ~545nm to 595nm for the selected 532nm laser. According to this specific wavelength band, the concave grating can be designed for optimum diffraction efficiency and spectral resolution.
2.1 Concave grating design and fabrication
The design considerations of a concave grating spectrometer include the grating constant, radius, incident angle, and entrance slit width. Given the wavelength band, these parameters can be determined through the simulation of the diffraction efficiency and spectral resolution. Since the Stokes signal is delivered by an optical fiber to the concave grating, the entrance slit width is then determined by the fiber’s diameter. The details for the design of a concave grating are provided in our previous work [3].
The concave grating was fabricated on a plano-concave lens (Thorlabs-LC1258-Ø25mm-38.6mm radius of curvature) by the holographic technique. Figure 2(a) shows the fabrication setup. The lens was spin-coated with a thin layer of photoresist (S1808) and placed on the sample holder on one side of an L-stage, on which the other side has a mirror mounted. Collimated UV laser beam (325nm) of certain size (~10 cm) is incident on the L-stage, covering both the lens and the mirror. The beam reflected from the mirror interferes with the rest of un-reflected beam, generating the interference pattern of desired pitch that depends on the tilt angle of the UV laser beam. The pattern is recorded in the photoresist after exposing the lens to the interference pattern for a period of time determined by the light intensity. To form the grating structure, the photoresist was developed using a developer (CD30). Finally, the surface of the lens with the patterned stripes of photoresist was sputter-coated with 2.5nm Cr and 20nm Au to make the surface reflective. The grating constant of the concave grating fabricated by this way can be calculated from
where λ is the wavelength of the recording UV laser beam, and θ is the tilt angle between the L-stage and the incident beam.
Fig. 2 (a) Setup of the holographic fabrication using an L-stage; (b) SEM image of the fabricated concave grating surface; (c) Comparison of the lens before and after fabrication of the grating pattern; (d) Spectra from the concave grating illuminated by a halogen lamp.
The entire L-stage is held on a rotational stage, by which the tilt angle can be adjusted to obtain different grating constants. Figure 2(b) shows the SEM image of the grating surface, indicating that the periodic grating structure is well formed on the concave lens. The grating constant is ~980nm. The comparison of the lens before and after fabrication of the grating pattern is shown in Fig. 2(c). The rainbow appearing after the fabrication confirms the desired diffraction of grating. The concave grating was tested by a halogen lamp, and the 0th order (white) and the 1st order (rainbow) images are seen in Fig. 2(d).
2.2 Pixel design and operation
The SPAD in our pixel circuit is designed and fabricated in the standard 130nm CMOS technology of IBM. More information about the design of the SPAD can be found in [16]. Most of the SPADs used are operated in the free running mode, which means that the SPADs are always in the “on” state until they are ‘fired’ by photons. After the firing, the SPAD can be either passively or actively quenched, then later reset again to the on-state. In contrast, the time-gated SPADs (TG-SPADs) are turned on in a very narrow gate window, and only photons coincident with this window can be detected. Consequently, the gating signal of the SPAD dominates the operation of the pixel circuit.
To achieve fast gating of the SPAD, three on-chip pulse generators (P1, P2 and P3) are designed, shown in Fig. 3. The external trigger signal is shaped by an on-chip Schmitt trigger to a nearly rectangular waveform with target levels of 3.3V and 0V (ground, GND). The shaped signal is then fed into the three monostable pulse generators, which generate three negative pulses. The pulse widths of the three generators are different, because of the different number of delay stages in each signal path. Each delay stage consists of at least an inverter and a capacitor. The pulse width is determined by the product of the time constant of the delay stage and the number of stages connected in series. Thus, the pulse widths were adjusted during the design either by varying the capacitances in the delay stages and the number of serially connected stages.
The sequence of the three pulses was specially designed to gate the SPAD and form the readout pulse from the pixel. Figure 4 shows the controlling circuitry of the active pixel and its timing diagram. The gate window starts at the falling edges of P1 and P2, which simultaneously turn on M1 and turn off M2. The pMOSFET M1 pulls up the cathode of the SPAD to VDD, since the nMOSFET M2 isolates the cathode from GND at this time; thus, performing RESET of the SPAD at high reverse bias for a short duration of ~500ps of P1. Then, M1 is turned off, and the SPAD is ready to sense an avalanche event for the remaining time of 3.5ns of P2 (shaded area shown in Fig. 4(b)), since the duration of P2 is ~4ns. The gate window is ended at the rising edge of P2, which turns on M2 and pulls the SPAD cathode voltage down to GND, below the breakdown threshold of the SPAD.
When a detection event happens within the gate window, electron-hole pairs are generated and multiplied in the high-field depletion region of the SPAD. A large avalanche current discharges the SPAD (and the other capacitances connected to the cathode) down to the diode breakdown voltage, which is depicted as the step-down drop Vex of the cathode voltage VC in Fig. 4(b). Unlike the free running operation, transistor M1 in the TG-SPAD pixel is in the off state during the gate window, so the cathode is not reset to VDD until the start of the next gate window.
A gated readout circuit is designed to sense the cathode voltage drop only within the gate window. This readout circuit consists of three series-connected transistors (M3, M4 and M5). Outside the readout gate, the high level of P3 turns off M3 and connects VR to GND. The readout starts at the falling edge of P3, which turns on M3 and disconnects VR from GND. If no detection happens within the gate, the high level of VC disconnects M4 from VDD, and VR is maintained to be zero. In contrast, if a detection event occurs within the gate, the voltage drop of VC will turn on M4 and pulls up VR from GND to high level (VDD-2VTH). The readout ends at the rising edge of P3 which resets VR to zero. The unlabeled transistors at the output are two inverters that form the levels between VDD and GND and provide for the signal buffering and matching to a 50Ω load at the output of the pixel.
The Out pulse duration is equal to the time interval between the photon arrival and the end of the gate. As depicted in Fig. 4(b), the Out pulse width varies with the different photon arrival times in the different gate windows (4 gates are given), and if no photon is detected, then there is no pulse (the 3rd gate). Thus, this TG-SPAD pixel can be used to obtain the photon arrival time within the gate.
2.3 Performance of the TG-SPAD pixel
A photomicrograph of the fabricated TG-SPAD is shown in Fig. 5(a). The chips were packaged and mounted on PCB with DC and RF connectors [Fig. 5(b)] for characterization. A schematic representation of the experimental setup for the evaluation of the TG-SPAD pixel is shown in Fig. 5(c). The bias voltages are provided by an Agilent B1500A Semiconductor Device Analyzer (SDA), which also monitors the bias currents during the experiments. The input trigger signal is provided by a pulse pattern generator (MP1763B, Anritsu). Its bit rate is set by the frequency of a Synthesized Sweeper (83752A, Agilent). The output Out is measured by a high-speed real-time oscilloscope (LeCroy SDA 18000 Serial Data Analyzer).
Fig. 5 (a) Micrograph of the fabricated chip; (b) Test PCB with DC and RF connectors; (c) Setup of the illumination measurement; (d) Measurement of the photon arrival time.
To characterize its dark count probability per gate and the afterpulsing probability, the TG-SPAD pixel was tested in dark with gating frequencies of up to 100MHz. The dark counts were significantly reduced at lower gating frequencies, since the dark counts that potentially occur during the long gate off time were not detected. In addition, benefit from the low parasitic capacitance (~220fF) at the cathode node, the dark count probability per gate was found to be stable for longer deadtime (≥20ns), which implies low afterpulsing probability (<3%).
To test the optical performance of the TG-SPAD pixel, the chip is illuminated with a high speed pulsed laser. As shown in Fig. 5(c), the trigger signals from the pattern generator are used to provide the synchronization, one to drive the PCB and the other to trigger the laser driver (PicoQuant: PDL-800-B) and the laser head (PicoQuant: LDH-D-C-510). To adjust the delay in cables, a delay unit (Optronics-TRRC1) is connected between the pattern generator and the laser driver. To test the efficiency of the TG-SPAD pixel for measurement of the photon arrival time, the laser is operated at a repetition rate of 80MHz. Figure 5(d) gives the measured photon arrival time within the fixed gate window (3.5ns), the FWHM of which is ~165ps. From the datasheet of the laser, the FWHMlaser≈130ps. Quadratic subtraction yields that the added FWHM during the test is
This value is a conservative, and includes jitter from generator, oscilloscope, laser, delay unit and the TG-SPAD. Thus, the TG-SPAD pixel has FWHMTG-SPAD<100ps. Another experiment using a 7ps pulsed laser in the next section implied that the FWHMTG-SPAD≈60ps.
3. Time-Gated Raman spectrum measurements and discussions
Figure 6 shows the setup of the time gated Raman spectrometer combining the concave grating and the TG-SPAD. A diode-pumped solid state laser (Passat Compiler 355, 7ps pulse width, ~160µJ/pulse @100Hz) was used as the excitation source. This laser can emit three wavelengths (1064nm, 532nm and 355nm) simultaneously, and the 532nm was used for measurement. This laser is multifunctional, expensive, and was selected to demonstrate the functionality of our system with our customized wavelength selector and detector. A cheaper and smaller laser diode chip will be used for our final integrated system. Since the maximum operating frequency of the laser is 200Hz, and the jitter of the laser output is above 100ns when triggered by an external source, then the setup shown in Fig. 5(c) cannot be used for this laser.
To synchronize the narrow laser pulse with the 3.5ns gate, the laser beam is split into two paths by a beam splitter (70:30). The 70% path is used to illuminate the sample in a cuvette, and the 30% path is sensed by a high-speed photodiode (Thorlabs, DET10A). The output (1ns rising edge) from a high speed photodiode is used to trigger the oscilloscope and the TG-SPAD. To control the position of the gate window with respect to the laser pulse, a delay unit is used to adjust the position of the gate window. Since a single TG-SPAD pixel is used in the experiments, the chip is mounted on a three-axis translation stage (Thorlabs, NanoMax TS 313D). The translation stage offers 4mm travel range with coarse and fine adjustments of resolutions of 10µm and 1µm, respectively. The chip is aligned at the Rowland circle of the concave grating, and then moved in steps of 10−50μm on the tangent of the circle to acquire the data for the spectrum from 520nm to 600nm. This arrangement of the setup corresponds to the required Raman shift up to 2100 cm−1.
To test the efficiency of the gating for the suppression of the background fluorescence signal, a standard fluorescence dye (Rhodamine B) is selected as a characterization target. As shown in Fig. 6, both the fluorescence emission and scattering signals (Rayleigh and Raman) are collected by a multimode optical fiber (200µm diameter, 0.22NA), without using either an emission filter or a dichroic mirror. The other end of the fiber is positioned on the Rowland circle of the concave grating, which separates the light spatially and focuses the diffraction beam onto the SPAD.
As mentioned above, Raman scattering is an almost instantaneous response, while fluorescence is emitted with a temporal distribution characterized by the fluorescence lifetime of the material. Therefore, to effectively suppress the background fluorescence, the width and the position of the detection window of the SPAD are of great importance. The fluorescence lifetime of Rhodamine B was reported to be ~1.7ns [17] when dissolved in water, which implies the majority of the fluorescence signal is emitted within the 3.5ns gate. In addition, the emission peak of fluorescence is ~583nm, overlapping with the Raman peaks if excited by the 532nm laser. Therefore, detection of the Raman peaks in this situation is very challenging.
To find out the dependence of the fluorescence suppression on the width of the detection window, different detection window widths are tried. As shown in Fig. 7, the detection window is defined as the time interval between the excitation and the end of the gate window, and the delay unit is used to adjust the width of the detection widow, until the excitation is within the gate window. Since the gate window is fixed to 3.5ns by the on-chip pulse generators, then the maximum detection window is 3.5ns.
The spectrum of Rhodamine B is measured under 6 different detection windows, varying from 3ns to 250ps. The data for each position is collected from 10000 acquisitions, which corresponds to 10000 laser shots in total. The dark count in these 10000 acquisitions is less than 1, and it is negligible compared with the counts in the spectra. The dark count probability, together with the detection efficiency of the SPAD, play the dominant roles in the detection limit of the system. The spectra are plotted as number of counts vs. the Raman shift, shown in Fig. 8. The peak intensity of the Rayleigh scattering does not change much with the detection window, because Rayleigh scattering is an instantaneous response to the excitation.
A strong fluorescence emission in a broad band is observed for the detection window of 3ns. However, in contrast to the Rayleigh scattering, the intensity of the fluorescence emission is significantly decreased by narrowing the detection window, proving the efficiency of the narrowing of the detection window for suppression of fluorescence signals.
In addition to the broad fluorescence emission band, a strong Raman peak is observed to be superimposed with the fluorescence signal for the 3ns detection window. Actually this peak is a superposition of two peaks, since the peak at detection window 3ns gradually splits into two peaks when the detection window is reduced to 250ps. Figure 9(b) gives a better perspective for the evolution of the Raman peaks as function of the detection window.
To obtain the positions of the Raman peaks, the measured data from Fig. 8 are processed as follows. First, the fluorescence background is extracted, and the background levels (curves in Fig. 9(a)) reduce as the detection window is decreased. Second, the background is subtracted to obtain the Raman spectra (symbols) in Fig. 9(b). Third, the Raman spectra are fitted with Gaussian functions, curves in Fig. 9(b), and from the parameters of the Gaussian approximations, the Raman shift and FWHM of the Raman peaks are obtained. The vertical dashed lines in Fig. 9(b) depict the average positions (6 detection windows) of two Raman peaks, R1 and R2. We observe stable values for the Raman shift of these peaks R1 and R2 at different detection windows, since the two Raman peaks are well aligned with the vertical dashed lines. The average values of R1 and R2 are 1372cm−1 and 1284cm−1, respectively, which are in good agreement with the results (1365cm−1, 1290cm−1) measured for Rhodamine B by surface enhanced Raman spectrometer [18]. The good agreement (within ± (6−8) cm−1 between ours and published data) verifies that the proposed spectrometer with concave grating and TG-SPAD is suitable for Raman spectroscopy.
Another observation in Fig. 9(b) is the FWHM of the Raman peaks, in the order of 100 cm−1 for longer detection windows, as deduced from the curves of the Gaussian approximations in Fig. 9(b). As discussed in [3], the entrance slit (width and numerical aperture) plays an important role in system’s spectral resolution. To achieve high spectral resolution, narrow entrance slits (10μm-25µm) are employed in commercial spectrometers. However, the entrance slit width of our setup is determined by the multimode fiber diameter (200µm). The large diameter fiber was chosen through a trade-off between spectral resolution and light intensity that can be coupled and transmitted, since high intensity is also important for the measurement. In addition to the wide entrance slit, the curved focal plane of the grating also contributes to the low spectral resolution. Since the SPAD is linearly moved on the translation stage, then some fraction of the spectrum is measured at the positions that are out of focus, which degrades the spectral resolution.
Although some weak Raman peaks have not been resolved by this low cost and compact setup, two strong Raman peaks have been successfully detected, and this proves the efficiency of our approach of using concave grating for wavelength separation and time gated SPAD for fluorescence suppression. When these are combined and the acquired data properly processed, then the small-size and simple optical assembly of concave grating with TG-SPADs at the Rowland circle, is feasible for a portable Raman spectrometer for environmental applications in remote areas. The overall performance deduced from our prototype is summarized in Table 1, and state-of-the-art CMOS TG-SPADs are also provided as a comparison. Because of the narrow depletion region and the polyimide layers on top of the active region in this standard CMOS process, our PDE is lower than those from the HV CMOS process. However, our SPADs possess better temporal resolution and lower current consumption. Most importantly, without using commercial high-sensitivity and high-resolution spectrometers as in [13] and [11], our customized concave grating is used to resolve the spectra in this work. This results in a decrease in the cost and size of the system, which are of great importance for the on-line applications of the Raman spectrometer.
Table 1. Summary of the state-of-the-art CMOS TG-SPADs
4. Conclusion
In summary, this work is targeting a low-cost and compact system to measure the Raman spectrum. To reduce the cost and size, a 2.5cm concave grating with less than 4cm radius of curvature is designed and fabricated by a low-cost, in-house holographic technique. The dispersion properties of the concave grating are tested by applying a white light source from a halogen lamp. The Raman spectrum from the concave grating is detected by a CMOS SPAD, which is operated in the time-gated mode to suppress the strong background fluorescence. The efficiency for fluorescence rejection is verified by measuring the Raman spectrum of a standard fluorescence dye. The fluorescence suppression is significant for short detection windows of 250ps. The short detection window has also allowed for resolution of two Raman peaks, which are in good agreement with published data, although these peaks combine at long detection windows of 2-3ns.
References and links
1. T. Vankeirsbilck, A. Vercauteren, W. Baeyens, G. Van der Weken, F. Verpoort, G. Vergote, and J. P. Remon, “Applications of Raman spectroscopy in pharmaceutical analysis,” TrAC Trends Analyt. Chem. 21(12), 869–877 (2002). [CrossRef]
2. C. J. Strachan, T. Rades, K. C. Gordon, and J. Rantanen, “Raman spectroscopy for quantitative analysis of pharmaceutical solids,” J. Pharm. Pharmacol. 59(2), 179–192 (2007). [CrossRef] [PubMed]
3. Z. Li, M. J. Deen, Q. Fang, and P. R. Selvaganapathy, “Design of a flat field concave-grating-based micro-Raman spectrometer for environmental applications,” Appl. Opt. 51(28), 6855–6863 (2012). [CrossRef] [PubMed]
4. J. R. Ferraro, Introductory Raman Spectroscopy. (Academic Press, 2003).
5. F. Knorr, Z. J. Smith, and S. Wachsmann-Hogiu, “Development of a time-gated system for Raman spectroscopy of biological samples,” Opt. Express 18(19), 20049–20058 (2010). [CrossRef] [PubMed]
6. F. Ariese, H. Meuzelaar, M. M. Kerssens, J. B. Buijs, and C. Gooijer, “Picosecond Raman spectroscopy with a fast intensified CCD camera for depth analysis of diffusely scattering media,” Analyst (Lond.) 134(6), 1192–1197 (2009). [CrossRef] [PubMed]
7. J. V. Sinfield, O. Colic, D. Fagerman, and C. Monwuba, “A low cost time-resolved raman spectroscopic sensing system enabling fluorescence rejection,” Appl. Spectrosc. 64(2), 201–210 (2010). [CrossRef] [PubMed]
8. D. U. Li, J. Arlt, J. Richardson, R. Walker, A. Buts, D. Stoppa, E. Charbon, and R. Henderson, “Real-time fluorescence lifetime imaging system with a 32 x 32 0.13microm CMOS low dark-count single-photon avalanche diode array,” Opt. Express 18(10), 10257–10269 (2010). [CrossRef] [PubMed]
9. A. Dalla Mora, A. Tosi, F. Zappa, S. Cova, D. Contini, A. Pifferi, L. Spinelli, A. Torricelli, and R. Cubeddu, “Fast-gated single-photon avalanche diode for wide dynamic range near infrared spectroscopy,” IEEE J. Sel. Top. Quantum Electron. 16(4), 1023–1030 (2010). [CrossRef]
10. C. Niclass, M. Soga, H. Matsubara, M. Ogawa and M. Kagami. “A 0.18µm CMOS SoC for a 100m-range 10fps 200× 96-pixel time-of-flight depth sensor,” IEEE Int Solid-State Circuits Conf Dig Tech Papers, 488–489 (Jan. 2013).
11. Y. Maruyama, J. Blacksberg, and E. Charbon, “A 1024 x 8, 700-ps time-gated SPAD line sensor for planetary surface exploration with laser raman spectroscopy and LIBS,” IEEE J. Solid-State Circuits 49(1), 179–189 (2014). [CrossRef]
12. I. Nissinen, J. Nissinen, A. Lansman, L. Hallman, A. Kilpela, J. Kostamovaara, M. Kogler, M. Aikio, and J. Tenhunen, “A sub-ns time-gated CMOS single photon avalanche diode detector for Raman spectroscopy,” European Solid-State Devices Research Conference (ESSDERC), 375–378 (2011) [CrossRef]
13. J. Kostamovaara, J. Tenhunen, M. Kögler, I. Nissinen, J. Nissinen, and P. Keränen, “Fluorescence suppression in Raman spectroscopy using a time-gated CMOS SPAD,” Opt. Express 21(25), 31632–31645 (2013). [CrossRef] [PubMed]
14. M. J. Deen and E. D. Thompson, “Design and simulated performance of a CARS spectrometer for dynamic temperature measurements using electronic heterodyning,” Appl. Opt. 28(7), 1409–1416 (1989). [CrossRef] [PubMed]
15. M. M. El-Desouki, D. Palubiak, M. Deen, Q. Fang, and O. Marinov, “A novel, high-dynamic-range, high-speed, and high-sensitivity CMOS imager using time-domain single-photon counting and avalanche photodiodes,” IEEE Sens. J. 11(4), 1078–1083 (2011). [CrossRef]
16. D. Palubiak, M. M. El-Desouki, O. Marinov, M. Deen, and Q. Fang, “High-speed, single-photon avalanche-photodiode imager for biomedical applications,” Sensors Journal, IEEE 11(10), 2401–2412 (2011). [CrossRef]
17. N. Boens, W. Qin, N. Basarić, J. Hofkens, M. Ameloot, J. Pouget, J. P. Lefèvre, B. Valeur, E. Gratton, M. vandeVen, N. D. Silva Jr, Y. Engelborghs, K. Willaert, A. Sillen, G. Rumbles, D. Phillips, A. J. Visser, A. van Hoek, J. R. Lakowicz, H. Malak, I. Gryczynski, A. G. Szabo, D. T. Krajcarski, N. Tamai, and A. Miura, “Fluorescence lifetime standards for time and frequency domain fluorescence spectroscopy,” Anal. Chem. 79(5), 2137–2149 (2007). [CrossRef] [PubMed]
18. T. Vo-Dinh, L. R. Allain, and D. L. Stokes, “Cancer gene detection using surface‐enhanced Raman scattering (SERS),” J Raman Spectrosc 33(7), 511–516 (2002). [CrossRef]
19. E. Vilella, O. Alonso, A. Montiel, A. Vilà, and A. Dieguez, “A low-noise time-gated single-photon detector in a HV-CMOS technology for triggered imaging,” Sens. Actuators A Phys. 201, 342–351 (2013). [CrossRef]
