Open Access
Add to BibSonomy
Abstract
The integration of fluorescence sensing directly into the fluidic channel in lab-on-a chip systems using thin film Si detectors enables on-chip multi-target medical diagnostics and biochemical analyses. This paper reports on the experimental demonstration and theoretical analysis of a filter-free thin film fluorescence sensor designed for integration into the channel of a fluidic platform. Static tests of this optical sensor show repeatable detection of 6-Hex fluorophore concentrations from 300 nM to 20 µM, with an average signal-to-noise ratio of 26 dB-50 dB, which agrees well with the theoretical model.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
The integration of highly sensitive and specific sensors for fluorescence targets in fluidic systems has the potential to enable miniature, portable, point of care diagnostics for a wide variety of analytes. Currently, most reported systems are specific to the target and application for which they are developed. However, a more generalized system with different programmable specific targets, identified in real time, would enable flexible, quickly adaptable sensing as targets rapidly change or emerge, such as coronaviruses, which can use optical fluorescence detection integrated with PCR [1]. Such a cyberphysical system could even incorporate “smart” self-monitoring and decision options that could affect real-time system function, (a cyberphysical system as identified by the National Science Foundation [2]). If these characteristics were integrated into a lab-on-a-chip (LOC) system, with portability, cost effectiveness, high sensitivity, and high specificity, these systems could be rapidly adapted to emerging needs.
An integrated LOC includes functions such as sample preparation and analysis on the same platform, and a number of studies have reported integration of sensors for detection in LOC platforms [3–6]. Optical sensing is a well-established technique for biochemical sensing that is non-invasive, highly sensitive, and provides localized detection. It can also provide real-time detection and has been integrated into a variety of fluidic systems [6]. For example, a Si light emitting diode and PD have been integrated into a fluidic microchannel to measure a shift in the interference spectra [7] . Fluorescence sensing, in particular, is highly specific and sensitive to tagged biomarkers [8] and has been used extensively in LOC diagnostic systems. Optical sensing systems that have been integrated with microfluidic platforms include the monolithic integration of a GaAs-based fluorescence sensor utilizing a PIN photodetector with a VCSEL emitter grown on top of the PD, which was optically coupled to a fluidic channel using filters and lenses [9]. Amorphous Si PDs and organic PDs have also been integrated with fluidic channels, utilizing filters or lenses, and/or mirrors to improve weak signals [10–13]. A common drawback to these reported optical sensing systems is the use of integrated and/or external optics and/or low signal to noise ratios (SNRs) that limit system flexibility and portability, for point-of-care use. In particular, the use of filters and dichroic mirrors constrains the wavelength range of operation, limiting the optical system to specific fluorophores.
This paper describes an optical thin film Si fluorescence sensor designed for integration directly into the channel of a fluidic platform. The sensor does not use filters, mirrors, or lenses, has a high SNR, and can be integrated into continuous flow or digital droplet fluidics systems. The optical sensor can operate over a wide range of wavelengths without re-design or further physical adjustments, enabling flexible use and rapid adaptation to emerging targets. Herein is reported the fabrication and test of an optical fluorescence sensor, designed for integration into LOC fluidic platforms, directly into the fluidic channel, without altering or obstructing the fluid flow. Thin film (10 µm thick) Si photodetectors (PDs) were bonded on the bottom of a pyrex plate designed for integration as the upper surface of a fluidic LOC system. The excitation pump was incident on the fluid through the via in the back of the annular PD, with the detector inside and facing the fluid. The integration of an absorbing optical beam dump for the pump light on the bottom of the fluidic channel eliminates the need for filters, mirrors and/or lenses for blocking the pump beam from the PD for fluorescence sensing. The sensor measurements are reported for 10 µL droplets on a standard fluidics system bottom plate (silicon nitride/Si). The fluorophore concentration measurements herein range from 300 nM to 20 µM and they agree well with simulations of this integrated annular thin film Si PD sensor.
2. System design and fabrication
The integrated Si fluorescence sensor was comprised of an annular thin film Si PD bonded to a pyrex plate with an optical fiber pump input from the back of the top pyrex plate, as shown in Fig. 1. The optical pump was matched to the 6-HEX fluorophore pump wavelength at 532 nm and entered the fluidic droplet from the back of the PD, through a patterned aperture in the thin film single crystal Si. Of note is that the fluorophore and pump wavelength can be changed without change to the optical system, as the Si PD has high broadband visible responsivity. Thus, the system is quickly adjustable to different fluorophores and different targets (also note that multiple PDs can be integrated into the system for multiple targets in a single system). The pump light was provided by a laser and delivered to the sensor system through an optical fiber, as show in Fig. 1. The thin film (10 µm thick) Si PD was integrated onto the top plate of the fluidics system, and the bottom plate was a silicon nitride-coated Si substrate. The droplet containing the fluorophore was located between the top plate and the bottom plate of the system, and the gap between the plates (droplet height) was 160 µm, as set by a gasket adhered to the bottom plate.
Fig. 1. Cross-sectional schematic of the integrated thin film Si fluorescence sensor (not to scale).
The shape of the PD is annular, as shown in Fig. 2, to allow the optical pump light to be coupled into the droplet through the aperture defined in the center of the PD. This enables efficient pumping of the droplet and minimizes the pump signal reaching the detector by transmitting the pump beam through the droplet and into a bottom plate or surface designed to absorb the pump beam with minimal reflection of the pump back into the PD. There is also a ring of metal patterned around the PD aperture to protect the inner portion of the PD from the reflected pump, where the simulation of this design indicates that the majority of the reflected pump occurs. The active area of the PD is the area between the outer edge of the metal ring and outer ring of the top contact to the PD. The back contact to the PD is patterned on the other side of the PD, and thus, is not visible in the photomicrograph in Fig. 2. This optical design is aimed at achieving a high SNR for integrated LOC fluorescence detection without any additional optical elements such as filters, mirrors, or lenses.
2.1 Top plate fabrication
A thin film Si PD (10 µm thick) was bonded to a pyrex plate that would become the top plate of the sensing system. The thin film Si PD was fabricated from silicon-on-insulator (SOI) material, which had a 10 µm thick p-Si device layer (3 × 1014 cm−3, Addison Engineering), with 1 µm of buried oxide on a 400 µm bulk Si handle wafer. After RCA cleaning, a spin-on glass phosphorus (n-type) dopant (Phosphorofilm, Emulsitone Chemicals) was spun onto the wafer and a diffusion anneal at 1000 °C was used to create a p-n junction with a junction depth of ∼600 nm. Next, a sloped etch was performed using a Deep Reactive Ion Etcher (DRIE) with C4F8, SF6 and O2 to define the PD active area mesa. The inner radius and outer radius of the mesa were designed to be 125 µm and 675 µm, respectively. A Ti (80 nm)/ Ni (50 nm)/Au (200 nm) contact was deposited and patterned on the phosphorus doped Si to create the top contact. Then the wafer surface (top) was bonded to a temporary carrier wafer using Waferbond adhesive in a vacuum oven. The bulk Si SOI handle substrate wafer was then removed using DRIE with a mixture of 9:1 SF6 and O2. This was followed by the removal of the buried oxide layer with a wet etch using buffered oxide etch solution. Next, a broad area (back) contact of Al (100 nm)/ Ti (80 nm) /Ni (50 nm)/Au (200 nm) was deposited onto the exposed p-Si side of the PD device layer.
The thin film Si PD was then bonded to a 500 µm thick pyrex top plate. First, a Ti (80 nm)/ Au (200 nm) bonding pad was patterned onto the pyrex top plate. The thin film Si PD was then heterogeneously metal-metal bonded to the bonding pad on the pyrex substrate with a 1 kg weight applied at 250 °C.
After bonding, the temporary carrier wafer was removed from the sample by melting the Waferbond and cleaning with TCE and a O2 RIE etch. To electrically connect the top (n-side) contact on the PD to a lead on the top plate (without shorting the contact on the back), a polyimide interlayer dielectric (ILD), PI 2611 (HD Microsystems), was spun, cured, and patterned using a mixture of 1:9 CHF3 and O2 RIE etch. Then, a Ti (80 nm)/ Au (240 nm) top contact lead was deposited and patterned, and a 65 nm silicon nitride AR coating was deposited using PECVD. A photomicrograph of a thin film PD bonded to a pyrex top plate is shown in Fig. 2.
2.2 Bottom plate fabrication
A PECVD silicon nitride (65 nm)/Si (600 µm) bottom plate was used to assess the performance of the PD for fluorescence sensing. The silicon nitride thickness was designed to be an AR coating at the pump wavelength, to enable the pump beam to be transmitted to the absorbing Si substrate. To be consistent with use in a fluidic system, a hydrophobic CYTOP solution (20% by weight, Bellex International Corp.) was spin coated and cured on the nitride-coated bottom plate to produce a 50 nm film hydrophobic film.
Adhesive laser patterned gaskets were used between the bottom and top plates to contain the droplets and to maintain a fixed separation between the top and bottom plate (i.e. droplet height). These SecureSealTM Adhesive Spacer gaskets (ThermoFisher Scientific) were adhered to the bottom plate using the gasket adhesive, which resulted in a gasket height of 160 µm.
3. Experimental assembly
The Si PDs were first characterized after fabrication, and before testing with the bottom plate. The current-voltage (I-V) curve, dark current and responsivity of each of the thin film PD was measured, and each had a clear diode characteristic, dark current of less than 25 pA (averaging 500 data points), and a typical responsivity at a wavelength of 556 nm of 0.36 A/W. The measured I-V characteristic of the PD is shown in Fig. 3. These results are comparable to previously reported single crystal thin film Si PDs on pyrex [14].
To test the integrated thin film Si fluorescence sensor, the top plate and the bottom plate were brought together with the test liquid dispensed inside the gasket. First, the top plate was adhered to a printed circuit board (PCB) for mechanical support. The PCB also had an aperture for optical pump delivery and contact pads for electrical connections to the PD contact pads. A photograph of the top plate ready for test is shown in Fig. 4. A Keithley source measurement unit (SMU) was used for to measure the PD photocurrent during the experiments.
To deliver pump light to the droplet, a 50 µm core/125 µm cladding multimode fiber (Thorlabs, M14L05) was coupled to a 4.5 mW, 532 nm wavelength laser using a collimator and neutral density filters. The output end of the fiber was aligned to the back of the PD aperture to maximize throughput through the aperture, and was epoxied to the back of the pyrex top plate. The bottom plate with the gasket was mounted on a translation stage for alignment to the top plate.
The fluids under test had the fluorophore 6-Hex (Anaspec, Inc), which was first dissolved in dimethyl sulfoxide (DMSO), and then further diluted with DI water, in fluorophore concentrations ranging from 300 nM to 20 µM. The DMSO content in the highest fluorophore concentration was less than 5%, and it decreased with decreasing fluorophore concentrations. The fluorophore excitation and emission wavelengths were 532 nm and 556 nm, respectively. In each test, 10 µL of fluid was pipetted onto the bottom plate and then the bottom plate and the top plate were brought together until the top plate touched the gasket that was adhered to the bottom plate. Once the droplet was loaded, fluorescence was measured, and the droplet was removed, followed by a rinse in DI water to clean the bottom plate and PD. Measurements were carried out from low concentration of fluorophores to higher concentration, with the PD operating under zero applied bias (photovoltaic mode).
4. Experimental results and data analysis
There were multiple integrated top plates/PDs and bottom plates that were tested to assess the system performance. Table 1 lists the top plate and bottom plate used for each of these tests. Note that these tests refer to testing of the fabricated PD bonded to a top plate (TPi) with a nitride-coated Si bottom plate (NPj). Each of the top plates and bottom plates were fabricated separately. The input power was kept constant at 18 µW at λ=532 nm for all experiments.
Table 1. Summary of different tests discussed in the paper.
Figure 5 shows the results from all of the system measurement tests. For each fluorophore concentration, 500 measurements were averaged during each experiment, each with an integration time of 1 µs, and the measured standard deviation is shown as the error bars for each measurement. Figure 6 shows a magnified version of the same data to highlight the measured photocurrent at the lower fluorophore concentrations. The three tests carried out with TP1 as the top plate, even with two different sets of bottom plates (NP1 and NP2), have data points that agree well. However, the second top plate, TP2, had a higher measured photocurrent, which is likely due to fabrication variability, such as in the alignment of the metal ring around the PD aperture, and the thickness of the AR coating, and alignment of the fiber to the PD aperture. The 300 nM detected fluorophore concentration is the lowest fluorophore concentration that has been detected with an integrated system without using filters, lenses, or mirrors, to the best of the authors’ knowledge. With the demonstrated experimental setup, the measurement resolution is 0.1 µM. This resolution limit was experimentally determined and can be improved by reducing measurement variability, and by further optimizing the system.
Fig. 5. Measured photocurrent as a function of fluorophore concentration using fabricated thin film Si detector.
Fig. 6. Measured photocurrent as a function of fluorophore concentration (0–1 µM) using fabricated thin film Si detector.
The average SNR for each test was calculated using Eq. (1), where ${I_{fluorophore}}$ is the mean measured photocurrent when testing the fluorophore droplet, ${I_{DIWater}}$ is the mean photocurrent measured with a DI water droplet, and ${\sigma _{fluorophore}}$ is the standard deviation for the measurements for that fluorophore concentration. The average photocurrent measured over all tests with the DI water droplet was 7.5 nA. The average SNR across all concentrations and tests was 43 dB, with the SNR for all tests ranging between 26 dB and 53 dB, which compares well to reported SNRs for other integrated fluorescence systems [9].
Two sources of unintentional photocurrent that contribute to the noise floor are the back illumination current, which is the current that exists when the fiber illuminates the system without a sample under test., and the reflected pump current, which has been minimized, but is not zero. To compensate for these photocurrents, the DI water measurement can be used to approximately calibrate the sensing system by subtracting the DI water measurement from the data for each data set, since the DI water measurement contains all of the artifacts of the background signal such as dark current, back illumination current, and an approximate value of the pump signal as detected at the PD. Figure 7 shows the data from Fig. 5, with the corresponding DI water data subtracted from each of the measurements. Figure 8 shows the same data on a magnified scale to observe the effect of the DI water subtraction at lower fluorophore concentrations.
Fig. 7. Measured photocurrent with DI water subtracted as a function of fluorophore concentration for experiments using fabricated thin film Si detector.
Fig. 8. Measured photocurrent with DI water subtracted as a function of fluorophore concentrations 0–1 µM for experiments using fabricated thin film Si detector.
The data for all of the experiments is in better agreement, indicating that the DI water calibration of the system has the potential to address variation in system performance due to fabrication and experimental variance.
5. Modeling the system
Fluorescence modeling for the integrated thin film Si PD fluorescence sensor system described herein was carried out using the commercial ray tracing software Zemax OpticStudio using non-sequential ray tracing and the Zemax photoluminescence and scattering models. The photoluminescence model was used to set the absorption, emission, and quantum yield spectrum for the fluorophores based upon published literature on the 6-HEX fluorophore [15]. The nitride-coated Si bottom plate was modeled with a specular reflectance of 4% and scattering at 5% due to surface roughness and thickness variations. Figure 9 compares the output of the modeled fluorophore sensing system (open circles) to the photocurrent measurements. Figure 10 shows the same data on a magnified scale to show the comparison at the lower fluorophore concentrations.
Fig. 9. Experimental data for all tests compared to Zemax simulations for all measured concentrations using the fabricated thin film Si sensor.
Fig. 10. Experimental data for all tests compared to Zemax simulations for concentrations 0–1 µM using the fabricated thin film Si sensor.
As is expected from the modeling, each data set shows a fairly linear trend for the measured photocurrent as a function of the fluorophore concentration. The deviations from the linear trend in the measured data, within each experiment, could be attributed to the mechanical movement while changing concentrations, as well as any biofouling caused by using only a DI water rinse as an intermediate step between different fluorophore concentrations. The experimental data could be improved by automating the process of droplet delivery for testing, as well as by using appropriate washing buffers and hydrophobic coatings on the PD. In addition to the slight deviation from the expected linear trend as shown within each experiment, the variability between multiple experiments can be attributed to fabrication variability.
To compare the modeled fluorophore photocurrent to the measured results, in which the DI water was subtracted from the measured data points, in the model, the theoretical pump photocurrent was subtracted from the total photocurrent. These theoretical results were compared to the experimental results (with the DI water subtracted), as shown in Fig. 11, and as a magnified view for the lower fluorophore concentrations in Fig. 12. At lower fluorophore concentrations, the fluorescence-based photocurrent is smaller than at higher fluorophore concentrations, thus the calibration had a larger impact upon the theoretical model comparison to the measured data at lower concentrations, since the photocurrent due to background illumination and dark current is a higher percentage of the total photocurrent at lower fluorophore concentrations.
Fig. 11. Experimental data for all tests with measured DI photocurrent subtracted compared to Zemax simulations with pump photocurrent component subtracted, for all measured concentrations.
Fig. 12. Experimental data for all tests with measured DI photocurrent subtracted compared to Zemax simulations with pump photocurrent component subtracted, for concentrations 0-1 µM.
For the lowest measured 6-HEX concentration in the system, 300 nM, the noise levels in the measured average photocurrent signal (measured by the standard deviation of the 500 consecutive data points sampled), is at least 100 times lower than the average detected signal. Assuming similar noise levels and device performance and minimizing measurement variability, the theoretical limit of detection is estimated to be 50 nM without compromising the SNR. The theoretical limit of detection of our device could be improved by optimizing factors such as the system geometry, light delivery, and engineered antireflection/hydrophobic coatings.
6. Conclusions and future work
Integrating Si PD-based fluorescence sensing into LOC systems has great potential for flexible, rapidly adaptable sensing. Herein, the design, fabrication, test, and simulation of an annular thin film Si PD for integrated fluorescence sensing in microfluidic platforms is reported. Droplets with fluorophore concentrations varying from 300 nM-20 µM were tested, and the average SNR was between 26 dB and 50 dB over all of the tests. The lowest fluorophore concentration detected in this system was 300 nM, which is one of the lowest reported detection limits for filter-free system with such high SNR, to the best of authors’ knowledge. A theoretical model was developed using non-sequential ray tracing and a photoluminescence model in Zemax OpticStudio, and the agreement between theory and experiment was promising for the measured photocurrent data and for the measured data with the DI water subtracted as an approximate calibration to account for fabrication variations. In particular, the calibration improved the theoretical model comparison to the measured data at lower concentrations significantly, since the unintentional photocurrent noise floor is a higher percentage of the total photocurrent at lower fluorophore concentrations. We estimate that, for a fully optimized system, the theoretical limit of detection for the system would be 50 nM without affecting the SNR.
Funding
National Science Foundation (CNS-1135853).
Acknowledgement
We thank Dr. Talmage Tyler II and Dr. Richard Fair for their help in developing this work with their inputs for fabrication and experiments, and Duke University for supporting this work. This work was performed in part at the Duke University Shared Materials Instrumentation Facility (SMIF), a member of the North Carolina Research Triangle Nanotechnology Network (RTNN), which is supported by the National Science Foundation (ECCS-1542015) as part of the National Nanotechnology Coordinated Infrastructure (NNCI).
Disclosures
The authors declare no conflicts of interest.
References
1. “U-TOP TM COVID-19 Detection Kit For in vitro diagnostic use only,” vol. 2. 2020.
2. “National Science Foundation Cyber-Physical Systems.” [Online]. Available: https://www.nsf.gov/news/special_reports/cyber-physical/.
3. C. Monat, P. Domachuk, and B. J. Eggleton, “Integrated optofluidics: A new river of light,” Nat. Photonics 1(2), 106–114 (2007). [CrossRef]
4. L. Luan, M. W. Royal, R. Evans, R. B. Fair, and N. M. Jokerst, “Chip scale optical microresonator sensors integrated with embedded thin film photodetectors on electrowetting digital microfluidics platforms,” IEEE Sens. J. 12(6), 1794–1800 (2012). [CrossRef]
5. J. P. Lafleur, A. Jönsson, S. Senkbeil, and J. P. Kutter, “Recent advances in lab-on-a-chip for biosensing applications,” Biosens. Bioelectron. 76, 213–233 (2016). [CrossRef]
6. N. M. M. Pires, T. Dong, U. Hanke, and N. Hoivik, “Recent developments in optical detection technologies in lab-on-a-chip devices for biosensing applications,” Sensors 14(8), 15458–15479 (2014). [CrossRef]
7. P. Le Minh, J. Holleman, J. Berenschot, N. Tas, and A. Van Den Berg, “Monolithic integration of a novel microfluidic device with silicon light emitting diode-antifuse and photodetector,” Eur. Solid-State Device Res. Conf., pp. 451–454 (2002).
8. B. Kuswandi, J. Nuriman, W. Huskens, and Verboom, “Optical sensing systems for microfluidic devices: A review,” Anal. Chim. Acta 601(2), 141–155 (2007). [CrossRef]
9. E. Thrush, O. Levi, L. J. Cook, J. Deich, A. Kurtz, S. J. Smith, W. E. Moerner, and J. Harish, “Monolithically integrated semiconductor fluorescence sensor for microfluidic applications,” Sens. Actuators, B 105(2), 393–399 (2005). [CrossRef]
10. T. Kamei, B. M. Paegel, J. R. Scherer, A. M. Skelley, R. A. Street, and R. A. Mathies, “Integrated Hydrogenated Amorphous Si Photodiode Detector for Microfluidic Bioanalytical Devices,” Anal. Chem. 75(20), 5300–5305 (2003). [CrossRef]
11. G. Williams, C. Backhouse, and H. Aziz, “Integration of Organic Light Emitting Diodes and Organic Photodetectors for Lab-on-a-Chip Bio-Detection Systems,” Electronics (Basel, Switz.) 3(1), 43–75 (2014). [CrossRef]
12. F. Lefèvre, A. Chalifour, L. Yu, V. Chodavarapu, P. Juneau, and R. Izquierdo, “Algal fluorescence sensor integrated into a microfluidic chip for water pollutant detection,” Lab Chip 12(4), 787–793 (2012). [CrossRef]
13. H. Robbins, K. Sumitomo, N. Tsujimura, and T. Kamei, “Integrated thin film Si fluorescence sensor coupled with a GaN microLED for microfluidic point-of-care testing,” J. Micromech. Microeng. 28(2), 024001 (2018). [CrossRef]
14. S. Dhar, D. M. Miller, and N. M. Jokerst, “High responsivity, low dark current, heterogeneously integrated thin film Si photodetectors on rigid and flexible substrates,” Opt. Express 22(5), 5052 (2014). [CrossRef]
15. “AAT BioQuest Spectrum Viewer (HEX).” [Online]. Available: https://www.aatbio.com/spectrum/HEX.
