Open Access
Add to BibSonomy
Abstract
To perform waveguide-enhanced Raman spectroscopy (WERS) or fluorescence spectroscopy in a compact device, the optical fibers to couple the passive photonic circuit to the laser source and detector require attachment directly to the die. This necessitates the integration of edge couplers and waveguide-based filters to isolate the fiber background emission from the on-chip signal, while efficiently coupling the pump laser and detector to the input and output fibers, respectively. In this work, we experimentally demonstrate the successful integration of four-port lattice filters with sensing spirals and inverse-taper edge couplers in a passive photonic circuit. We further show that the four-port lattice filter enables the collection of backscattered on-chip Stokes signal, improving and simplifying overall system performance.
1. Introduction
Waveguide-enhanced Raman spectroscopy [1,2] (WERS) takes advantage of the long interaction length and strong collection efficiency of low-loss nanophotonic waveguides to perform chip-scale chemical [3,4] and biological [5] sensing. On-chip evanescent-field fluorescence spectroscopy also shows promise for the on-chip detection of biomolecules. [6,7] Though photonic integrated circuits (PICs) offer the potential for full integration of sources, filters, sensing regions, and spectrometers, current technological limitations require an off-chip source and detector to be fiber-coupled to the PIC. Such an architecture requires not only low-loss, broadband couplers, but also high-performance broadband waveguide filters to separate the pump and background fiber emission from the PIC Raman signal.
In this work, we experimentally demonstrate the successful integration of multistage lattice filters with broadband edge couplers and a sorbent-coated WERS sensing region in a passive PIC. This device requires no external focusing optics to couple the laser source and spectrometer to the die. Instead, an array of polarization maintaining single-mode fibers are aligned directly to the waveguide facets of the PIC. Custom inverse-taper edge couplers [8] are designed for the TM mode at wavelengths between 1064 nm and 1300 nm. An eight-stage, four-port lattice filter [9] is used on both the input and output of the sensing region, enabling the collection of both forward-scattered and backscattered [10,11] Raman signal with high signal-to-background ratios across the entire wavelength range of interest, even with input optical fibers as long as 35 m. One recent study successfully connected two single mode fibers to a die with a single directional coupler to collect backscatter Raman emission [12]. However, this study uses a conventional narrowband directional coupler, limiting the collection of broadband Stokes signals and the suppression of reflected pump light.
2. Experimental methods
The fabrication of the PIC is performed at AIM Photonics using the Passives multi-project wafer (MPW) process. We use a 220 nm thick silicon nitride (SiN) core layer [13] fully etched and clad in silicon dioxide (SiO$_2$). The waveguides are 800 nm wide except where otherwise specified. The layout is generated using Synopsys Optodesigner. As shown in Fig. 1, the device comprises two sensing spirals in a region in which much of the top cladding has been removed and filled with a sorbent material. One of the sensing spirals is connected to lattice filters at both the input and the output. The filtered spiral is 9.2 cm long, and the unfiltered spiral is 14.5 cm long. Both devices connect to inverse-taper edge couplers for coupling to an array of polarization-maintaining (PM) single-mode fibers (Oz Optics VGA with 6 $\mu$m core diameter fibers) with the slow-axis aligned out of plane (vertically). Figure 2 depicts images of each of the aforementioned components, taken with a microscope. We use the TM$_{00}$ mode in this work due to its stronger evanescent penetration into the sorbent material, and weaker background generation from the SiN core. [3]
Fig. 1. Left: An array of polarization-maintaining single-mode optical fibers. Right: The photonic circuit with integrated edge couplers, lattice filters, and sensing spirals. (not to scale)
Fig. 2. Microscope images of (a) The array of edge couplers; (b) The 9.2 cm long sensing spiral, prior to FPOL deposition; and (c) The eight-stage lattice filter.
The eight-stage lattice filters [9] comprise eight pairs of broadband directional couplers and differential delay sections. These pairs are then terminated by a ninth directional coupler. The asymmetric couplers are designed to have approximately 3% power coupling between 1064 nm and 1300 nm [9,14]. The delay is chosen to be 2.192 $\mu$m for a filter resonance at 1064 nm and a passband extending past 1300 nm. Such a lattice filter should transmit 1064 nm in the cross port, and transmit wavelengths between 1080 nm and 1300 nm in the through port.
As shown in Fig. 1, eight edge couplers with a spacing of 127 $\mu$m connect the lattice filters or spirals to the external fiber array. The architecture enables both forward-scatter (port 2 $\rightarrow$ port 4, referred to as FS LF) and backscatter (port 2 $\rightarrow$ port 3, referred to as BS LF) waveguide emission signal to be collected from the spiral connected to the lattice filters, and only forward-scattered Raman signal to be collected from the unfiltered spiral (port 6 $\leftrightarrow$ port 7, referred to as No LF). This design should filter out Raman emission generated from the optical fibers connected to the filtered spirals, but not from the unfiltered spirals. The two outermost edge couplers (port 1 $\leftrightarrow$ port 8) simply connect to each other in a waveguide loop-back for alignment of the fiber array and edge coupler testing.
Individual die are spin-coated with fluoropolyol (FPOL), a standard hydrogen-bond acidic sorbent polymer, [15,16] to fill the sensing trenches for WERS. Samples are mounted and aligned to the fiber array using a multiaxis stage. For alignment, a fiber-coupled 1064 nm pump laser first passes through a bandpass filter (LL01-1064), then through a length of PM fiber (PM980-XP), before coupling to one of the loop-back waveguides on the sample via the fiber array. The output of the loop-back is measured with a power meter to optmize alignment. WERS measurements are obtained by passing the appropriate output from a the fiber array through a longpass filter (LP02-1064RU) and linear polarizer (LPNIR100-MP2) before measurement on a benchtop spectrometer with a liquid-nitrogen cooled detector. White light spectra [17] are obtained using the same detector, but with the input connected to a polarized broadband white-light source instead of the laser.
3. Results and discussion
After alignment of the fiber array to the sample using the laser, white light is transmitted through the loop-back and detected by the spectrometer. A linear polarizer at the white light source can be set to excite either the TE$_{00}$ (electric field polarized parallel to the fiber fast axis) or TM$_{00}$ (electric field polarized parallel to the slow axis) waveguide mode. The measured spectrum is normalized by the spectrum obtained by connecting the white light source directly to the spectrometer, and divided by two to represent coupling loss per facet. These normalized edge coupling spectra are plotted in Fig. 3, along with the calculated spectra for each mode. To calculate the coupling loss, we perform a modal overlap integration between the fiber mode and the inverse taper mode, and add reflection loss from the silica/air interfaces based on the effective index of the fiber and waveguide modes. The modes are found using beam-propagation analysis. In addition, we include substrate loss in Fig. 3(b) found from a mode-solver model that uses a perfectly-matched layer below the bottom SiO$_2$ cladding.
Fig. 3. Measured and calculated coupling efficiency between the inverse taper edge coupler and optical fiber.
At shorter wavelengths, the measured coupling loss per facet agrees well with the calculated loss, but at longer wavelengths the measured loss is as much as 2 dB different from the calculated loss. At shorter wavelengths, the loss is dominated by mode-mismatch, which is accurately captured by our model. However, at longer wavelengths, we believe the loss is dominated by substrate leakage that is very sensitive to small variations in the waveguide width and thickness. Since the substrate loss is dominated by the narrowest parts of the taper, shortening these sections should help to decrease the coupling loss in future designs.
To characterize the eight-stage lattice filter, a filter identical to the ones placed between the sensing spirals and the edge couplers is placed alone between two edge couplers elsewhere on the die. The same white light technique as that used to characterize the edge couplers is used to measure the transmission of the TM$_{00}$ mode through the cross and through ports of the lattice filter. The measured signal is normalized by the signal collected through a straight waveguide placed between two identical edge couplers. This measured spectrum is shown in Fig. 4(a). Measurements indicating transmission greater than 100% are due to small coupling errors between the normalizing waveguide and the lattice filter. The data indicate 75% transmission in the cross port at 1064 nm, but only 25% in the through port. In addition, transmission in the through port is $>$90% between 1080 nm and 1300 nm (Stokes shift of 300 cm$^{-1}$ to 1700 cm$^{-1}$) and is $<$10% in the cross port over this range.
Fig. 4. Measured and calculated through and cross transmitted power for the 8-stage lattice filter. The measured spectrum is normalized by the transmission of a straight waveguide with identical edge couplers.
An imperfect lattice filter will affect the WERS SNR in both forward-scatter and backscatter geometries. Though the exact SNR reduction depends on the amount of waveguide background and the source of measurement noise, among other factors, the typical SNR reduction will be of order the laser loss in the cross port combined with the signal loss in the through port. Here, that reduction is approximately 2.5 dB.
The lattice filter is designed and modeled using a finite-element mode solver (Comsol Multiphysics) combined with a transfer matrix coupled-mode model. [14] The simulation is shown in Fig. 4(b) and agrees very well with the measurement. The primary difference, the depth of the extinction at the laser wavelength, is due to the actual coupling coefficient of the directional couplers in the lattice filter differing from the targeted coupling. Simulations show that a 2% coupling error (that is, 5% coupling instead of 3%) can lead to the observed extinction at the laser wavelength.
The measured lattice filter spectrum shows that in a backscatter WERS measurement, $>$90% of the Stokes signal from the sensing spiral will be coupled to the appropriate output port. In a forward-scatter measurement, in which lattice filters are placed at both the input and output of a sensing spiral (see Fig. 1), the background emission from the input fiber in the Stokes region would be attenuated by $>$10 dB, and the pump laser in the output fiber would be attenuated by $>$6 dB (plus any propagation loss in the sensing region). As we will show below, this performance is sufficient to prevent fiber background emission from masking both forward-scatter and backscatter WERS signal from the sensing spiral.
To assess the capability of the lattice filter with an actual WERS measurement, we acquire emission spectra from the three different sensing paths described above, as shown in Fig. 5: backscatter with lattice filters (BS LF, port 2 $\rightarrow$ port 3), forward-scatter with filters (FS LF, port 2 $\rightarrow$ port 4), and forward-scatter without filters (No LF, port 6 $\rightarrow$ port 7). For each sensing path, spectra were taken with three different fiber lengths preceding the device: 5 m, 15 m, and 35 m. All three sensing paths show emission from the coated waveguides, as evidenced by a number of Raman peaks associated with the sorbent coating (at 760 cm$^{-1}$ and 1005 cm$^{-1}$ [16]) and a broad fluorescence background (emitted primarily by the SiN [3]). Figure 5(D) depicts an emission spectrum of only a 35 m fiber, indicating that the fiber’s emission is consistent with Raman scattering from SiO2. [18] As the fiber length increases, fiber Raman emission begins to appear in the waveguide spectra in Figs. 5(B) and (C).
Fig. 5. Measured emission spectra with increasing fiber lengths from each sensing path: A) Backscatter with lattice filters (BS LF, port 2 $\rightarrow$ port 3), B) Forward-scatter with filters (FS LF, port 2 $\rightarrow$ port 4), and C) forward-scatter without filters (No LF, port 6 $\rightarrow$ port 7). Panel D) shows an emission spectrum of a 35 m fiber, for reference.
To quantify the contribution of the fiber background emission to the total signal shown in Fig. 5, a linear regression with respect to fiber length was applied to the spectra. First, to account for small differences in connector loss for each length of fiber, the spectra were normalized by collected pump power in the spectrometer. Then, for each wavelength, the linear fit gives a slope representing fiber emission (counts/s*m) and a y-intercept representing waveguide emission (counts/s). Figure 6 plots the results of this process for each sensing path, assuming a 35 m long fiber.
Fig. 6. Measured emission spectra separated into fiber and waveguide contributions from each sensing path: A) Backscatter with lattice filters (BS LF, port 2 $\rightarrow$ port 3), B) Forward-scatter with filters (FS LF, port 2 $\rightarrow$ port 4), and C) forward-scatter without filters (No LF, port 6 $\rightarrow$ port 7).
Within the error of our regression, in the backscatter spectra shown in Fig. 6(A) there is no contribution from a fiber as long as 35 m. Superimposed on the waveguide emission spectrum is the through port ripple of the input lattice filter (from emission propagating from the sensing spiral to port 3). The forward-scatter spectrum in Fig. 6(B) also shows the through port ripple of the output lattice filter (from waveguide emission propagating from the sensing spiral to port 4). This sensing path also shows significant contribution from the fiber emission (which also shows the input filter ripple of the cross port) in the total collected emission signal. Last, the sensing path with no lattice filters (Fig. 6(C)) shows the strongest contribution of fiber emission with respect to the total collected emission. Note that the spiral measured in Fig. 6(C) is a different length than that measured in Fig. 6(A) and Fig. 6(B), resulting in a different level of loss and waveguide emission.
These data show that a lattice filter integrated with a sensing spiral significantly improves the rejection of fiber background signal compared to a sensor without a lattice filter. In addition, the lattice filter enables the efficient collection of backscatter WERS, which not only is significantly lower in fiber background than forward-scatter, but also is almost entirely free of pump power: The measured pump power from the LF FS path is 250 times higher than that of the backscatter path. Reducing the collected pump power is important for detector protection and for decreased emission in the collection fiber.
Our data indicate that the collected waveguide emission in backscatter is about three times weaker than the waveguide emission collected in forward-scatter. There is no fundamental difference between the emission rates in forward scatter vs. backscatter WERS. However, the fiber arrays are known to have over 3 dB of coupling loss variation from fiber to fiber due to imperfect core-cladding concentricity and nonuniform cladding diameters. Thus, we believe that this measured signal difference is simply due to small differences in loss in the edge coupling, with additional loss from the fiber connectors. In addition, it is important to note that though 35 m of fiber is longer that what would be likely be used in a fiber-coupled PIC, using such long lengths helps to quantify the fiber emission contribution to the overall signal. Decreased emission from a shorter fiber would still be significant in forward-scattering compared to the waveguide Raman emission from a trace analyte molecule in the sorbent material. [2,3]
4. Conclusions
We have demonstrated that lattice filters can be successfully integrated with edge couplers and sensing waveguides in a passive PIC for WERS. The measured ratio of fiber emission to waveguide emission is signifcantly lower in lattice-filter-coupled devices than in non-filtered devices. In addition, the fiber emission is virtually undetectable in backscatter measurements, which are only possible with a 4-port input filter such as a lattice filter. Also, we have measured significantly lower laser power in the backscatter configuration compared to forward-scatter, which is important to reduce additional emission in the output fiber.
Future lattice filter designs show significantly improved performance, both at the laser wavelength and in the Raman passband: $>$20 dB extinction of the laser in the through port, and $>$20 dB extinction of the Raman signal in the cross port. These improvements, coupled with the integration of low-fluorescence SiN sensing waveguides, [19] attached fiber arrays, [20] and compact flow-cells, will enable high-fidelity, compact WERS systems.
Funding
Office of Naval Research (NRL-WU1X06); Defense Threat Reduction Agency (HDTRA1033531).
Disclosures
The authors declare no conflicts of interest.
References
1. A. Dhakal, A. Z. Subramanian, P. Wuytens, F. Peyskens, N. L. Thomas, and R. Baets, “Evanescent excitation and collection of spontaneous Raman spectra using silicon nitride nanophotonic waveguides,” Opt. Lett. 39(13), 4025–4028 (2014). [CrossRef]
2. S. A. Holmstrom, T. H. Stievater, D. A. Kozak, M. W. Pruessner, N. Tyndall, W. S. Rabinovich, R. A. McGill, and J. B. Khurgin, “Trace gas Raman spectroscopy using functionalized waveguides,” Optica 3(8), 891–896 (2016). [CrossRef]
3. N. F. Tyndall, T. H. Stievater, D. A. Kozak, K. Koo, R. A. McGill, M. W. Pruessner, W. S. Rabinovich, and S. A. Holmstrom, “Waveguide-enhanced Raman spectroscopy of trace chemical warfare agent simulants,” Opt. Lett. 43(19), 4803–4806 (2018). [CrossRef]
4. H. Zhao, B. Baumgartner, A. Raza, A. Skirtach, B. Lendl, and R. Baets, “Multiplex volatile organic compound Raman sensing with nanophotonic slot waveguides functionalized with a mesoporous enrichment layer,” Opt. Lett. 45(2), 447–450 (2020). [CrossRef]
5. A. Dhakal, P. C. Wuytens, F. Peyskens, K. Jans, N. L. Thomas, and R. Baets, “Nanophotonic waveguide enhanced Raman spectroscopy of biological submonolayers,” ACS Photonics 3(11), 2141–2149 (2016). [CrossRef]
6. R. Bernini, N. Cennamo, A. Minardo, and L. Zeni, “Planar waveguides for fluorescence-based biosensing: Optimization and analysis,” IEEE Sens. J. 6(5), 1218–1226 (2006). [CrossRef]
7. F. J. Aparicio, E. Froner, E. Rigo, D. Gandolfi, M. Scarpa, B. Han, M. Ghulinyan, G. Pucker, and L. Pavesi, “Silicon oxynitride waveguides as evanescent-field-based fluorescent biosensors,” J. Phys. D: Appl. Phys. 47(40), 405401 (2014). [CrossRef]
8. V. R. Almeida, R. R. Panepucci, and M. Lipson, “Nanotaper for compact mode conversion,” Opt. Lett. 28(15), 1302–1304 (2003). [CrossRef]
9. N. F. Tyndall, T. H. Stievater, D. A. Kozak, M. W. Pruessner, S. A. Holmstrom, and W. S. Rabinovich, “Ultrabroadband lattice filters for integrated photonic spectroscopy and sensing,” Opt. Eng. 57(12), 127103 (2018). [CrossRef]
10. A. Raza, S. Clemmen, P. Wuytens, M. de Goede, A. S. K. Tong, N. L. Thomas, C. Liu, J. Suntivich, A. G. Skirtach, S. M. Garcia-Blanco, D. J. Blumenthal, J. S. Wilkinson, and R. Baets, “High index contrast photonic platforms for on-chip Raman spectroscopy,” Opt. Express 27(16), 23067–23079 (2019). [CrossRef]
11. T. H. Stievater, N. F. Tyndall, D. A. Kozak, M. W. Pruessner, C. A. Roberts, R. A. McGill, B. L. Miller, E. P. Luta, M. Z. Yates, E. D. Emmons, P. G. Wilcox, and J. A. Guicheteau, “Backscatter waveguide-enhanced raman spectroscopy for trace chemical agent detection using a handheld spectrometer,” Proc. SPIE 11416, 1141603 (2020). [CrossRef]
12. D. M. Kita, J. Michon, and J. Hu, “A packaged, fiber-coupled waveguide-enhanced raman spectroscopic sensor,” Opt. Express 28(10), 14963–14972 (2020). [CrossRef]
13. E. W. Ong, N. M. Fahrenkopf, and D. D. Coolbaugh, “SiNx bilayer grating coupler for photonic systems,” OSA Continuum 1(1), 13–25 (2018). [CrossRef]
14. Z. Lu, H. Yun, Y. Wang, Z. Chen, F. Zhang, N. A. F. Jaeger, and L. Chrostowski, “Broadband silicon photonic directional coupler using asymmetric-waveguide based phase control,” Opt. Express 23(3), 3795–3808 (2015). [CrossRef]
15. J. W. Grate, A. Snow, D. S. Ballantine, H. Wohltjen, M. H. Abraham, R. A. McGill, and P. Sasson, “Determination of partition coefficients from surface acoustic wave vapor sensor responses and correlation with gas-liquid chromatographic partition coefficients,” Anal. Chem. 60(9), 869–875 (1988). [CrossRef]
16. N. F. Tyndall, T. H. Stievater, D. A. Kozak, M. W. Pruessner, B. J. Roxworthy, W. S. Rabinovich, C. A. Roberts, R. A. McGill, B. L. Miller, E. Luta, and M. Z. Yates, “Figure-of-merit characterization of hydrogen-bond acidic sorbents for waveguide-enhanced raman spectroscopy,” ACS Sens. 5(3), 831–836 (2020). [CrossRef]
17. N. F. Tyndall, T. H. Stievater, D. A. Kozak, M. W. Pruessner, and W. S. Rabinovich, “Mode-crossing spectroscopy for photonic waveguide characterization,” APL Photonics 4(10), 106107 (2019). [CrossRef]
18. F. Galeener and J. M. Jr., “Raman studies of the thermal oxide of silicon,” Solid State Commun. 37(9), 719–723 (1981). [CrossRef]
19. A. Dhakal, P. Wuytens, A. Raza, N. Le Thomas, and R. Baets, “Silicon nitride background in nanophotonic waveguide enhanced Raman spectroscopy,” Materials 10(2), 140–141 (2017). [CrossRef]
20. P. Yin, J. R. Serafini, Z. Su, R.-J. Shiue, E. Timurdogan, M. L. Fanto, and S. Preble, “Low connector-to-connector loss through silicon photonic chips using ultra-low loss splicing of smf-28 to high numerical aperture fibers,” Opt. Express 27(17), 24188–24193 (2019). [CrossRef]
