1. Introduction

Spectroscopy measurements exhibit a wide range of applications in industry, medicine, environment and other fields. In general, CCD or photomultiplier tube (PMT) is the photodetector option of a spectrometer. CCD is involved in a simultaneous multichannel measurement process, and it features high measurement speed, but suffers from low gain, low time resolution and relatively large readout noise. CCD should also be cooled down to low temperature for low-level light detection. By contrast, PMT features high gain (~10⁶), high time resolution and low readout noise. This tube operates at room temperature for even single photon detection. Nevertheless, PMT is bulky and fragile, and it suffers from low measurement speed due to a slow single-channel wavelength scanning measurement process. Recently, single-photon avalanche diode (SPAD) array, as a pixel-type position-sensitive detector, has become available in standard complementary metal-oxide semiconductor technology [1]. The SPAD array shows advantages of both CCD (i.e., rapid measurement) and PMT (i.e., high gain, high time resolution and operation at room temperature). It is applied in laser Raman spectroscopy and laser-induced breakdown spectroscopy [1]. However, the SPAD array suffers from large number of output channels, which leads to complex readout electronics and high cost.

Silicon photomultiplier (SiPM) displays attractive advantages of high gain, high photon detection efficiency (PDE), excellent timing resolution, photon number discrimination, low operating voltage, compactness and convenience for integration [2]; It has been considered a replacement of PMT in high-energy physics [3], nuclear medicine systems [4], Raman spectroscopy [5], fluorescence spectroscopy [6], and other low level light detections [7]. Several position sensitive (PS) SiPM structures show few output electrodes (e.g., four electrodes) and complex structures integrated between the micro avalanche photodiode (APD) cells in past years [8–11]. These structures are limited with few-photon detection [8, 9], slow and position-related signal rising edge [10], and large position measurement error (PME) [11]. To overcome these drawbacks of the PS-SiPMs, the authors reported a 2D tetra lateral PS-SiPM with an intrinsic continuous cap resistive layer for charge division and integrated quenching resisters in the epitaxial layer recently [12]. We demonstrated the position resolution of 182.9 µm with mean photoelectron number (MPEN) of 52 [12]. Nonetheless, given that the divided signals between four cathodes may be reduced to a level comparable with the background noise, the signal-to-noise ratio (SNR) is the main challenge for the 2D PS-SiPM to realize single-photon detection with high position resolution.

In the present study, we improved the SNR of both the PS-SiPM and the measurement methodology. Results showed that a one-dimensional (1D) PS-SiPM exhibited excellent photon number and position discriminations for few-photon detection. By using this novel photodetector with only three output electrodes, easy fabrication technology and simple readout electronics, Raman spectroscope of carbon tetrachloride (CCL₄) was obtained at room temperature without a slow wavelength scanning process.

2. Mechanism and theory

Figure 1 shows the schematic structure of the 1D PS-SiPM in this study. The device operates above breakdown voltage and some micro APD cells are triggered by the incident photons as shown in the right section of Fig. 1. The photo-induced avalanche charges are divided between cathodes 1 and 2 through the continuous cap resistive N + + layer. The spot position is determined by the weight of the output charges between the two cathodes. By contrast, the light spot position is determined by the physical position of each SPAD pixel for the SPAD array in [1], in which large amount of output channels and complex readout electronics are necessary. Correspondingly, the barycenter position of incident photons in the x-direction for 1D PS-SiPM can be obtained [13] by the following equation:

 

Fig. 1 Schematic structure of the one-dimensional position-sensitive silicon photomultiplier (1D PS-SiPM). Left: cross section. Right: top view. The green point (x, y) represents the incident photons position.

Download Full Size | PDF

The uncertainty of the measured position mainly comes from that of the incident photons’ barycenter position and that of the device with electronic noise. Therefore, the position resolution of the measurement system (PR_System) is determined by following equation:

The standard uncertainty of the measured position (σ_X) according to Eq. (1), is determined by the standard uncertainty of Q₁ and Q₂ (i.e., σ_Q1 and σ_Q2) contributed by the noise component in the measurement channels (denoted by noise charges Q_n1 and Q_n2). Thus, the σ_X can be deduced by the error transfer rules [14] and Eq. (1) as follows:

When the photons are incident at x = 0, the mean collected charges and the standard uncertainty of the noise charge of the two measured channels are equal to each other (i.e., <Q₁> = <Q₂>, σ_Qn1 = σ_Qn2 = σ_Qn) because the two measured channels are symmetrical in this case. Therefore, PR_Device at x = 0 can be simplified from Eq. (4) as follows:

3. PS-SiPM description and measurement setups

The 1D PS-SiPM was provided by Novel Device Laboratory (Beijing, China), with effective size of 2.2 mm × 2.2 mm, geometrical fill factor of ~41%, pitch and density of the micro APD cells of ~10 μm and ~10⁴/mm², respectively. As shown in Fig. 1, each cell is connected in parallel by intrinsic continuous cap resistive layer (N + + ) and isolated by the gap depletion regions; each cell is also in series with a quenching resistor formed in epitaxial silicon layer. Fabrication was started from the p-type <100> epitaxial layer on P + silicon substrate. Initially, a thick field oxide was thermally grown on the wafer, which was selectively removed using buffered Hydrogen Fluoride. Subsequently, Boron ion (B⁺) implantation was performed for the P type enriched region. After a thermal drive-in process, the N + + region was formed by heavy Phosphide ion (P⁺) implantation, followed by an annealing process. An optimized antireflecting coating was achieved by Plasma Enhanced Chemical Vapor Deposition. The contacts and metallization were realized to terminate the fabrication process. Finally, the device was packaged using TO-5 metal holder without a glass cap. The peak PDE of a regular SiPM device with the same micro APD cell structure is 12.4% at 460 nm [15].

An attenuated picosecond laser (GK laser, Amberpico-Q-10), with wavelength of 532.43 nm, linewidth of 0.15 nm, pulse width of 15 ps (FWHM), and repetition rate of 100 kHz, was used as the light source. The laser light was divided by a beam splitter. To characterize the device (Fig. 2(a)) by adjusting the distance between the optical fiber prober (diameter~50 μm) and the device surface, the FWHM of incident light spot in the x-direction was measured to be ~64.8μm by using the knife-edge scanning method. The device was fixed on a micro positioner with a positional accuracy of 1 μm. Both the device and the micro positioner were placed in a black metal box to avoid the environmental electromagnetic and light interference. For the Raman spectroscopy measurement of CCl₄ as shown in Fig. 2(b), a laser beam through the beam splitter was used for Raman scattering excitation. The scattered light from the sample was collected by an objective lens (80X, NA = 0.75, Olympus). Afterward, a dichroscope (LPD01-532RS, Semrock) was used to reflect the 532.43 nm laser and filter the anti-Stokes photons. The remaining light was collected by a lens; prior to this, a notch filter was used to remove the Rayleigh scattering and stray light. The Raman photons were incident on the entrance slit of the monochromator (Zolix, Omni-λ3028i) with the slit widths of 0.5 mm. The 1D PS-SiPM, fixed in a black metal box, was coupled to the exit slit of the monochromator.

 

Fig. 2 Schematic of the setup for (a) device characterization and (b) Raman spectroscopy. The attenuator for the pulse laser is not shown in the figure. The yellow part illuminated by the laser (arrow) is the device under test. All the measurements were performed at 20°C.

Download Full Size | PDF

The breakdown voltage of the device was 26.5 V, and the over voltage was chosen to be 3.3 V for all measurement. The avalanche signals from cathodes1 and 2 were amplified by two identical high-speed amplifiers (FEMTO, HSA-Y-2-40, Gain 40 dB, Input Impedance 50 Ω), and acquired by two identical channels of the 4 GHz bandwidth oscilloscope (Lecroy Wave Runner 640Zi). A synchronizing signal from the PIN photodiode (UPD-200-UD, ALPHALAS) was used as a trigger to reject the dark counts of the device.

4. Results and discussion

4.1 Characteristics of the device

The photoelectron pulse area distributions recorded by cathode 1 with the incident light spot positions at (−1000 μm, 0) and (1000 μm, 0) are shown in Fig. 3. The laser pulse amplitude was reduced by the attenuator to MPEN of ~9. A total of 50000 pulse events were recorded for each position. This result showed that the device could discriminate the number of incident photons. When the light spot is close to cathode 1, the discrimination capability of photon number becomes improved; in the opposite direction, the discrimination capability worsens. This result can be explained by that the shared charge signal becomes weak when the light spot is farther away from cathode 1, thereby decreasing the SNR. Furthermore, the peak-to-valley ratio of the photoelectron pulse area distribution becomes small. The measured photoelectron pulse area distribution, as shown in Fig. 3, deviates from the standard Poisson distribution. This deviation can be explained by the poor photon number resolution in case of relatively large number of photons due to the fluctuation of photoelectron pulse area.

 

Fig. 3 Photoelectron pulse area distributions recorded by cathode 1 at (a) (−1000 μm, 0) and (b) (1000 μm, 0). The mean photoelectron number was ~9.

Download Full Size | PDF

Figure 4(a) shows the photoelectron pulse area distribution recorded by cathode 1 when the photons (MPEN, ~9) were incident at position (0, 0). The mean avalanche pulse area recorded by cathode 1 for the single photoelectron (1 p.e.) events (<S_{1, 1p.e.}>) is 1.21 pV·s. Thus, the mean total avalanche pulse area corresponding to a single photoelectron, i.e., <S_1p.e.>, is 2.42 pV·s (i.e., double of <S_{1, 1p.e.}>). Accordingly, the mean total avalanche pulse area triggered by n photoelectrons (n = 1, 2, 3…), i.e., <S> in Eq. (6), is equal to n·<S_1p.e.>. The device gain is estimated to be <S_1p.e.>/R_S/e, or ~3 × 10⁵ (load impedance R_S is 50 Ω). As shown in the inset of Fig. 4(a), the FWHM of the statistical histogram of the pedestal area, i.e., FWHM_Sn in Eq. (6), is fitted to be 0.34 pV·s. Figure 4(b) illustrates that the theoretical curve of PR_Device, calculated by Eq. (6), is improved from 393.4 μm to 56.2 μm with the increase of photoelectron number from 1 to 7. This improvement can be attributed to that the position resolution of 1D PS-SiPM is mainly limited by SNR, rather than the pixel size as in the SPAD array in [1]. The theoretical curve of PR_Device is in good agreement with the measured values of PR_System which is represented with the diamonds in Fig. 4(b) and obtained through the FWHM of the position distribution based on Eq. (1) by selecting photoelectron number events from 1 to 7. The small difference between PR_System and PR_Device can be explained by that the FWHM_photons in Eq. (2), i.e., FWHM of the barycenter distribution for the incident photons in the x-direction, is equal to the FWHM of the incident light spot (~64.8 μm) for the single photon detection, and rapidly decreased with the increased incident photon number. The measured PR_System for single photoelectron detection was 399.6 μm; according to Eq. (2), the corresponding PR_Device should be 394.3 μm which is in good agreement with the theoretical value of 393.4 μm. In addition, the PR_System and PR_Device for 7 photoelectrons detection were 65.1 and 56.2 μm, respectively; this means that the FWHM_photons for 7 photoelectrons detection was equal to a reasonable value of 32.9 μm.

 

Fig. 4 (a) Photoelectron pulse area distributions recorded by cathode 1. The inset shows the corresponding full-width at half maximum (FWHM) of the statistical histogram of the pedestal area. (b) The dependence of the position resolution on the photoelectron number. In both (a) and (b), the photons were incident at the (0, 0).

Download Full Size | PDF

The single photon PR_Device for different light spot positions was obtained through the FWHM of the position distribution based on Eq. (1) by selecting single photoelectron (1 p.e.) events and using Eq. (2). As shown in Fig. 5, the single photon PR_Device is best at x = 0, and get worse as the light spot is close to one of the cathodes.

 

Fig. 5 Measured single photon PR_Device at different (x,y) positions with the fitted curves to represent the variation trend.

Download Full Size | PDF

The PME of the device, i.e., the deviation between the true light spot position and the measured one from Eq. (1), was investigated. The active area of the device was scanned from (−1000 μm, −1000 μm) to (1000 μm, 1000 μm) with a step of 100 μm in the x-direction and 200 μm in the y-direction by the incident light spot with the MPEN of ~0.11 (single-photon level). A total of 20000 sets of data for S₁ and S₂ at each measured position were recorded. As shown in Fig. 6, the discrimination of incident light spot position, even at single photon level, is demonstrated. Quantitatively, the average value and the mean square error of PME data for all the measured light spot positions are derived to be 29.6 and 27.3 μm, respectively, or the PME is 29.6 ± 27.3 μm (~1.3% of the length of the device). For this 1D device, the detected light spot position in the x-direction, as determined by Eq. (1), is independent on the positions of the light spot in the y-direction.

 

Fig. 6 Light spot position vs. true light spot position measured at different y-points. The fitted straight lines represent the linearity between the measured and true light spot positions.

Download Full Size | PDF

4.2 Feasibility of Raman spectroscopy measurement

Given that the 1D PS-SiPM can detect only the barycenter position of incident photons, it is hard to realize accurate spectroscopic measurement in conjunction with a grating monochromator when two or more photons with different colors are incident at different positions of the detector simultaneously. Therefore, the feasibility of Raman spectroscopic measurement using the 1D PS-SiPM was investigated by adjusting the intensity of the exciting pulse laser source so that single photoelectron is induced from Raman photons and recorded by the detector.

To calibrate and optimize the grating monochromator, the Raman spectrum of CCl₄ in the range of 534.70–558.70 nm was first measured with a single channel wavelength scan process and the 1D PS-SiPM (the exit slit width was 0.5 mm). The results are shown as the inset in Fig. 7(a). Considering that the reciprocal linear dispersion of the monochromator is ~2.0 nm/mm and the length of the device is 2.2 mm, the maximum spectrum width of the light from the monochromator that can be received by the device is ~4.4 nm. This width corresponds to the wavelength range covering Raman peaks of CCl₄ at 538.68 and 541.48 nm, i.e., wavenumber shifts of 218 and 314 cm⁻¹ accordingly for 532.43 nm laser excitation. The central wavelength from the exit slit of the monochromator was thus fixed at 540.08 nm, i.e., the intermediate value between the aforesaid two Raman peaks.

 

Fig. 7 (a) Spatial distribution of single Raman photons corresponding to the Raman peaks of CCl₄ at 538.68 and 541.48 nm. The inset shows the Raman spectroscopy of CCl₄ between 534.70 and 558.70 nm measured with regular wavelength scanning method and the same 1D PS-SiPM. (b) The Raman spectroscopy of CCl₄ for the peaks at 538.68 and 541.48 nm, as transformed from (a).

Download Full Size | PDF

Measurement of Raman spectroscopy of CCl₄ was performed by using the 1D PS-SiPM and setting the exit slit of monochromator to 2.2 mm in width without the slow wavelength scanning process. The spatial distribution of the incident single Raman photons was detected by selecting all the single photoelectron (1 p.e.) events. Results are shown in Fig. 7(a). The two peaks of the spatial distribution at −711.0 and 745.5 μm, with FWHM of 793.7 and 725.7 μm, correspond to the two Raman peaks at 538.68 and 541.48 nm, respectively. The single photon PR_System was determined by both the single photon PR_Device and FWHM of the spatial distribution of the single Raman photons from the monochromator (exit slit width of 2.2 mm) which was expected much larger than that discussed in Chapter 4.1 with a fiber probe (diameter of 50 μm) as a light source. Consequently, the FWHM of 793.7 and 725.7 μm, i.e., the single photon PR_System, is significantly larger than the single photon PR_Device (393.4 μm); the detected position of single Raman photons ranges from −1500 μm to + 1500 μm, and exceeds the device length (2.2 mm). Through calibration by comparing the two Raman peaks at 538.68 and 541.48 nm with the measured spatial distribution peaks at −711.0 and 745.5 μm, the spatial distribution of the Raman spectroscopy shown in Fig. 7(a) can be transformed to that in wavelength distribution shown in Fig. 7(b). In addition, the reciprocal linear dispersion of the monochromator is ~1.92 nm/mm, which is considerably close to ~2.0 nm/mm, a specification value provided by manufacturer. Figure 7(b) shows that, the spectroscopy resolution, i.e., the FWHM of the wavelength distribution, for the Raman peaks at 538.68 and 541.48 nm is 1.45 and 1.33 nm, respectively.

In this study, the 1 p.e. dark count rate (DCR) of the 1D PS-SiPM is ~1.88 MHz and the measured time gate is ~14 ns (t, close to the full width of single photon pulses). A total of ~10000 single photon (1 p.e.) events were selected from a total of 50000 pulse events, and the corresponding MPEN is ~2.5. Given that the dark events follow the Poisson distribution [16], the probability of 1 p.e. dark pulse appearing within the measured time gate calculated using the equation P_dark(t) = 1−exp(−DCR × t) is 0.026. Given that the detected photoelectron number also follows Poisson distribution [17], the probability of photoresponse absence is approximately 0.082 considering the equation P(0) = exp(−MPEN). The 1 p.e. dark count contributing to all selected single photon (1 p.e.) events is equal to the product of the total pulse events number, the probability of 1 p.e. dark pulse appearing within the measured time gate, and the probability of photoresponse absence; this dark count is approximately 107 (i.e., 50000 × 0.026 × 0.082). The ratio accounting for all selected 10000 single photon (1 p.e.) events is ~1%. Therefore, the effect of dark pulses on accurate measurement of Raman spectroscopy with gating technique can be ignored.

5. Conclusion

The feasibility for fast measurement of Raman spectroscopy with a 1D PS-SiPM operated at room temperature was verified. The single photon detector with only three output electrodes, exhibits the advantages of both CCD (rapid measurement) and PMT (high gain, fast photon response, and simple readout electronics). The device, with a size of 2.2 mm × 2.2 mm and micro APD cell pitch of ~10 μm, performed both photon number and position discriminations. The PME was 29.6 ± 27.3 μm in single photon level and single-photon position resolution was 393.4 μm. With the increase of the photoelectron number from 1 to 7, the position resolution was improved from 393.4 μm to 56.2 μm monotonically. This novel device concept shows potential as a low-cost and high-performance detector for various laser spectroscopies.