Integrated surface plasmon resonance biosensors promise to enable compact and portable biosensing at high sensitivities. To replace the far field detector traditionally used to detect surface plasmons we integrate a near field detector below a functionalized gold film. The evanescent field of a surface plasmon at the aqueous-gold interface is converted into photocurrent by a thin film organic heterojunction diode. We demonstrate that use of the near field detector is equivalent to the traditional far field measurement of reflectivity. The sensor is stable and reversible in an aqueous environment for periods of 6 hrs. For specific binding of neutravidin, the detection limit is 4 μg/cm2. The sensitivity can be improved by reducing surface roughness of the gold layers and optimization of the device design. From simulations, we predict a maximum sensitivity that is two times lower than a comparable conventional SPR biosensor.
©2009 Optical Society of America
Despite widespread demand there remains an unmet need for cost effective biosensors. Applications in research laboratories, home and point of care diagnostics, process industries, environmental monitoring, security and bio-defense, require the measurement of bio-analytes with high specificity and minimal time lag between sample collection and measurement readout. Among commonly used sensing methods, surface plasmon resonance (SPR) achieves relatively high sensitivity (0.5 ng/cm2), and provides the benefits of label free detection and real time measurement of binding kinetics, while integration with microfluidics reduces the sample size and enables high throughput. SPR biosensors are highly versatile tools, being routinely used to examine protein - protein, antibody - antigen, and receptor - ligand interactions. However, they are also large, difficult to transport and relatively costly, due to their dependence on precise calibration and alignment of the internal optics.
The disadvantages of SPR biosensors could be overcome by integrating the devices in solid state. As in other optical applications such as lasers, solid state integration can significantly reduce the footprint of the device, and improve the environmental stability in the presence of vibrations and temperature changes. The resulting device contains three components: an optical pump, a near field detector for surface plasmons, and an aqueous interface with the bio-analytes.
In this work, we demonstrate integration of a near field surface plasmon detector and the biorecognition interface of a traditional SPR sensor. The near field detector is used to replace the conventional far field optical detector. It is integrated directly with an Au/water interface that supports surface plasmons and acts as the binding site for bio-analytes. The integration of the optical detector is arguably the key challenge confronting the integration of a SPR biosensor. The remaining optical element, the optical pump, can be replaced by a microcavity light emitting diode (LED)[5, 6], or vertical cavity surface emitting laser (VCSEL), placed under the near field detector and tuned for the wavevector of the plasmon resonance. Integration with an LED or VCSEL is not performed here; instead the integrated detector is characterized using an external laser.
2. Results and discussion
Conventional SPR sensors consist of a gold film deposited on glass and immersed in water. The system is optically excited through the glass by a pump laser. When the angle of the laser beam incident on the Au/water interface hits the resonance, surface plasmons are generated at the Au/water interface and the reflected light drops markedly. The resonant angle is a sensitive function of the refractive index of all media within the range of the surface plasmon, typically ~50 nm in Au and ~200 nm in water. Thus, analyte binding events at the Au/water interface modify the coupling of light into the surface plasmon and are detected from variations in the optical reflectivity.
We seek to replace the far field measurement of reflected light with a direct near field measurement of the surface plasmons themselves. The sensitivity of conventional SPR sensors is maximized for an approximately 50-nm-thick layer of Au deposited on glass. But the electromagnetic field is negligible at the bottom of the 50-nm-thick Au layer, preventing near field detection of the surface plasmons below the Au. Thus, to efficiently detect surface plasmons at the aqueous interface with minimal change in sensitivity, we split the gold layer and insert a semiconductor; see Fig. 1.
To determine the ideal properties of the semiconductor, we calculate the sensitivity of a model near field SPR detector. The top and bottom gold layers are 20-nm-thick. The top surface of the device is immersed in buffer with a refractive index n=1.38. The substrate is glass with a refractive index n=1.72. The Poynting vector within the model device is calculated using a transfer matrix method assuming plane wave incident light. To detect the surface plasmon in the near field the semiconductor must exhibit strong optical absorption. We assume that the semiconductor is 50-nm-thick with an extinction coefficient k=0.2. The sensitivity of the detector is calculated from the relative change in absorption within the semiconductor given the introduction of an interfacial 5-nm-thick protein layer with refractive index n=1.40.
As shown in Fig. 2(a), we find that the sensitivity of the model device is maximized for semiconductor refractive indices between n=1.3 and n=1.8. The relative change in absorption within the semiconductor is 30% for the optimal choice of refractive index. When combined with the requirement for strong optical absorption, this calculation supports the choice of organic semiconductors for this application. For example, the archetype organic photovoltaic material copper phthalocyanine (CuPC) exhibits n=1.7 and k=0.2 at λ=650 nm.
To compare the sensitivity of the near field detector to that of a conventional SPR, we also calculated the relative change in reflection from a 50-nm-thick Au layer. The same 5-nm-thick protein later with refractive index n=1.40 causes a 60% change in reflection, suggesting that the near field detector should exhibit roughly half the sensitivity of a conventional device. The calculated sensitivity of the near field detector should not be considered as a limit, however, since its structure contains opportunities for design optimization. For example, the bottom Au contact can be replaced by a lower loss Ag electrode.
Consistent with the refractive index guidelines of Fig. 2(a), we design a practical organic semiconductor-based photovoltaic detector. The anode is a 20-nm-thick gold layer. The donor material within the organic photovoltaic is a 10-nm-thick film of CuPC. The acceptor material is a 10-nm-thick film of buckminsterfullerene (C60). To increase optical absorption, a 20-nm-thick bulk heterostructure consisting of a 1:1 mixture of the donor and acceptor materials is deposited between the donor and acceptor layers. The cathode consists of an 8.5-nm-thick layer of bathocuproine (BCP) and a 20-nm-thick top gold layer. See Methods for complete fabrication details.
The electric field within the CuPC/C60 device is simulated in Fig. 2(b) as a function of the incident angle of optical excitation. Off resonance, the incident light is primarily reflected. But at the resonance, the incident light excites a surface plasmon that propagates in the plane of the Au and organic layers, significantly decreasing the reflected light and enhancing optical absorption within the photovoltaic. This enhancement in absorption is apparent in the strong electric field throughout the organic layers at the resonance condition.
The dependence of the photocurrent and reflectivity on the angle of incidence of the incoming light is measured and compared with the simulation in Fig. 3. Devices were immersed in a saline buffer (HEPES, GE Healthcare, Piscataway, NJ) typical of biosensing applications and exposed to a 1 mW laser at λ=670 nm; see Methods. Both data and simulations show an increase in photocurrent and a decrease in reflected light at the resonance condition (approximately 58°). The photocurrent at the surface plasmon resonance is approximately two times higher than the off resonance baseline due to enhanced absorption. We conclude that the reflectivity and photocurrent are equivalent measures of surface plasmon generation. However, the resonance width for the experimental plots exceeds the theoretical prediction, slightly lowering the sensitivity. The discrepancy is likely due to the 10 nm surface roughness of the Au layers within the device. Surface roughness lowers sensitivity by enhancing scattering of the surface plasmons, which decreases their lifetime and hence increases the angular width of the resonance. The scattering losses could be alleviated by careful preparation of the Au surfaces.
Next, the sensor response is examined within a microfluidic system; see Methods. To test the stability, sensitivity and reversibility of the sensor, water pulses of 30 and 60 s in length are injected into a constant flow of HEPES buffer. As shown in Fig. 4, the slight change of refractive index during the water pulses is detected by the sensor. We observe a simultaneous change in reflectivity and current with proportional amplitude of the two quantities. The sensor shows reversibility at the end of the water pulse and good stability with negligible drift of the baseline. The water pulse response was tested for several incidence angles for the incoming light to find the maximal sensitivity angular coordinate for later binding assays.
Next, we performed a specific binding assay for biotin-neutravidin, an archetypal evaluation of sensing platforms. The surface of the sensor was first immersed in water for 2-3 hours with a 5:1 molar mixture of PEG (polyethylene glycol) acid disulfide and biotin PEG disulfide (Polypure, Oslo, Norway). The purpose of the functionalization is to space out the biotin moieties to avoid steric hindrance and spatial overlap between neutravidin binding sites. The PEG backbone prevents protein absorption on Au, minimizing non specific interaction with the surface. Normally, this functionalization is performed in ethanol because functionalization in water decreases the surface coverage due to the hydration volume around the ethylene glycol moieties. The removal of water from the polyethylene glycol chains is thermodynamically unfavorable, and it prevents close packing of the polymer as well as surface access to protein species present in solution. But the water-based assembly is necessary here because the organic photovoltaic materials are weakly soluble in ethanol. Higher sensitivity could be obtained for surface functionalization with carboxyl methyl dextran hydrogel, which contains more binding sites for neutravidin within the range of the surface plasmon.
For selective detection of neutravidin, any remaining non specific binding sites on the Au surface were passivated with a 1 mg/ml solution of casein. Then, in a constant flow of 250 μl/min HEPES buffer, the sensor was exposed to sequential pulses of 250 μg/ml neutravidin (Pierce Biotechnology, Rockport, IL) of 125 μl injection volume each. Figure 5 shows a simultaneous response in the reflectivity and short circuit current when casein and neutravidin bind irreversibly to the functionalized surface of the sensor. The detection limit of the near field device is 4 μg/cm2.
As a control, we performed the same binding experiment using a conventional reflectivity-based detection of surface plasmons on a 50-nm-thick gold layer, except that the surface functionalization was performed in ethanol. We obtained a sensitivity from the conventional approach that is approximately three times better than the near field detector. This result is approximately consistent with the theoretical analysis accompanying Fig. 2 where we expected a factor of two reduction in sensitivity in the near field detector. The additional loss in the actual near field detector is likely due to inferior surface functionalization because of the restriction to water rather than ethanol. However, both the near field detector and our control reflectivity measurement exhibit absolute sensitivities that are substantially inferior to optimized SPR detectors. The fault is largely due to our experimental setup which does not control for fluctuations in the laser power or environmental variables such as temperature. Despite the relatively poor absolute sensitivity observed in our devices, experiment and theory are consistent in the relative sensitivity reduction associated with the near field geometry. This consistency suggests that optimized near field detectors can be employed in the majority of applications for surface plasmon resonance detectors.
Finally, we consider the stability of near field surface plasmon detectors. Illumination is typically applied to contemporary biosensor chips for no longer than a few hours. Thus, we expect that the stability of organic photovoltaic cells is sufficient for application in solid-state SPR detectors. The shelf life must be much longer than the operation life, but encapsulated organic photovoltaic cells have exhibited shelf lives exceeding 6000 hours. Our devices were not encapsulated and were tested within 24 hours of surface functionalization. We observed stable photocurrent throughout the 6 hour duration of our experiments with the top gold layer immersed in a saline buffer. Submerging the sensor in saline buffer solutions does not affect the electrical performance. In all cases the gold contact where the binding takes place was grounded. Diode characteristics in either air or buffer remained unchanged for anode bias in the -1V to +1V region, indicating that there are no leakage currents in solution. Although we did not observe stability problems in our experiments, C60 is known to exhibit photo-induced degradation in the presence of oxygen. Consequently, we also experimented with another acceptor, 3,4,9,10-perylene tetracarboxylic bisbenzimidazole (PTCBI). The combination of PTCBI and CuPC forms extremely stable photovoltaic devices. We observed similar device performance from PTCBI/CuPC, however, we found the use of PTCBI significantly increased the density of short circuit defects in these relatively thin devices.
To conclude, conventional SPR detectors measure the optical reflection in the far field. In this work, we replace the far field detector with a near-field detector positioned below the Au binding surface. The correlation between far field reflectivity and photocurrent from the near field detector observed in Figs. 3–5 demonstrates that the near field detectors can replace the traditional far field approach. Based on numerical simulations we predict a factor of two decrease in the sensitivity of our near field detector compared to a conventional SPR detector. Our experiments observe a factor of three reduction in sensitivity relative to a reflectivity-based control, with the additional losses in the near field detector most likely due to incomplete surface functionalization. Possibilities for improving the sensitivity of the near field detector relative to the conventional reflectivity-based approach include using longer wavelength light, the selective replacement of Au by Ag, and reductions in surface roughness of the metal layers. When combined with a microcavity LED or VCSEL, the near field detector should allow the integration of SPR biosensors into thin film devices, improving portability and environmental stability, potentially lowering costs, and introducing a new approach to the unsolved problems of biosensing.
3.1 Sensor fabrication and experimental setup
Devices were fabricated using thermal evaporation under vacuum (~10-6 Torr). First, a 20-nm-thick gold anode with a 3-nm-thick chrome adhesion layer was deposited through a shadow mask onto a flint glass substrate (SF10 glass, Schott AG) with a refractive index n=1.72. The organic photovoltaic materials CuPC, BCP, C60 (Sigma-Aldrich, St. Louis, MO), and PTCBI (Sensient Imaging Tech Gmbh, Wolfen, Germany) were used after thermal gradient purification. The 20-nm-thick top gold contact was patterned using a shadow mask. The active area of the device is approximately 0.79 mm2.
After fabrication, the photovoltaic cells were optically coupled using an index matching fluid (Cargille Laboratories) to a hemi-cylindrical prism made from the same material as the substrate (SF10 glass, Schott AG). The prism was mounted on a translation stage (Thorlabs, Newton, NJ) above a motorized rotation stage (AF Optical, Fremont, CA) aligned so that the motional rotation axis coincides with the symmetry axis of the cylindrical prism. The active region of the sample was placed on the prism axis and a λ=670 nm laser beam was collimated, p-polarized and focused on the same active region. The incident angle of the incoming laser beam was varied by rotating the prism. The angular dependence of the photocurrent and the reflectivity as monitored by a silicon photo-detector, were measured with a Keithley 2602 dual source-meter. For simulation purposes, the refractive indices and extinction coefficients of the gold used in the calculation were measured using an Aquila nkd-8000 (Aquila Instruments Ltd., Blackburn, UK).
Microfluidic masters were made of 2150 SU-8 negative photoresist (MicroChem Corp., Newton, MA), spun at 2000 rpm on a piranha cleaned silicon wafer for an expected thickness of approximately 0.28 mm. The wafer was soft baked in a convection oven for 7 minutes at 65 °C for thermal stress reduction followed by a 90 minute bake at 95 °C. The photoresist was patterned through a chrome mask by ultraviolet light exposure with a 370 mJ/cm2 dose, baked for 5 minutes at 65 °C and 30 minutes at 95 °C. Development was performed on a spinner with SU-8 developer until all unexposed material was removed. PDMS molds (Sylgard kit, Dow Corning, Midland, MI) were made with a mixture of 10:1 elastomer to primer ratio that was baked overnight at 65 °C. The rest of the microfluidic components, tubing, manual valve, adapters and connectors were purchased from Upchurch Scientific (Oak Harbor, WA). The final volume of the flow chamber was 1.2 μl, while tubing and connectors accounted for 3 μl.
All the sensing and binding measurements solutions were delivered to the surface of the sample using an Agilent 1100 HPLC autosampler. Exposure to water and biomolecular species was performed in step functions of various temporal lengths by adjusting the flow rate (250 μl/min) and the injection volume (125 μl). Before and after each injection of water, casein and neutravidin, the device was rinsed thoroughly in a buffer identical with the one used to dilute the biological samples. Although square water pulses are sent from the injection coil of the autosampler, the device response indicates that the pulse is modified by diffusion and parabolic flow patterns inside the connecting microfluidic tubing and chamber. Reflectivity and photocurrent were sampled with a period of 1 second.
This work was supported by the Institute for Soldier Nanosciences at MIT. CW would like to thank the Drapers Company, London for a travel scholarship.
References and links
1. L. S. Jung, C. T. Campbell, T. M. Chinowsky, M. N. Mar, and S. S. Yee, “Quantitative interpretation of the response of surface plasmon resonance sensors to adsorbed films,” Langmuir 14, 5636–5648 (1998). [CrossRef]
2. J. Homola, S. S. Yee, and G. Gauglitz, “Surface plasmon resonance sensors: review,” Sens. Actuators B 54, 3–15 (1999). [CrossRef]
4. J. K. Mapel, M. Singh, M. A. Baldo, and K. Celebi, “Plasmonic excitation of organic double heterostructure solar cells,” Appl. Phys. Lett . 90, 3, (2007). [CrossRef]
5. K. Celebi, T. D. Heidel, and M. A. Baldo, “Simplified calculation of dipole energy transport in a multilayer stack using dyadic Greenℙs functions,” Opt. Express 15, 1762–1772 (2007). [CrossRef] [PubMed]
6. C. L. Mulder, K. Celebi, K. M. Milaninia, and M. A. Baldo, “Saturated and efficient blue phosphorescent organic light emitting devices with Lambertian angular emission,” Appl. Phys. Lett . 90, 3 (2007). [CrossRef]
7. K. Iga, “Surface-emitting laser - Its birth and generation of new optoelectronics field,” IEEE J. Sel. Top. Quantum Electron . 6, 1201–1215 (2000). [CrossRef]
8. B. Johnsson, S. Löfås, and G. Lindquist, “Immobilization of proteins to a carboxymethyldextran-modified gold surface for biospecific interaction analysis in surface plasmon resonance sensors,” Anal. Biochem . 198, 268–277 (1991). [CrossRef] [PubMed]
9. S. Zhang, L. Berguiga, J. Elezgaray, T. Roland, C. Faivre-Moskalenko, and F. Argoul, “Surface plasmon resonance characterization of thermally evaporated thin gold films,” Surf. Sci . 601, 5445–5458 (2007). [CrossRef]
10. E. T. Castellana, S. Kataoka, F. Albertorio, and P. S. Cremer, “Direct writing of metal nanoparticle films inside sealed microfluidic channels,” Anal. Chem . 78, 107–112 (2006). [CrossRef]
11. P. Harder, M. Grunze, R. Dahint, G. M. Whitesides, and P. E. Laibinis, “Molecular conformation in oligo(ethylene glycol)-terminated self-assembled monolayers on gold and silver surfaces determines their ability to resist protein adsorption,” J. Phys. Chem . B 102, 426–436 (1998). [CrossRef]
12. E. Ostuni, R. G. Chapman, R. E. Holmlin, S. Takayama, and G. M. Whitesides, “A survey of structure-property relationships of surfaces that resist the adsorption of protein,” Langmuir 17, 5605–5620 (2001). [CrossRef]
13. S. Lofas and B. Johnsson, “A novel hydrogel matrix on gold surfaces in surface-plasmon resonance sensors for fast and efficient covalent immobilization of ligands,” J. Chem. Soc., Chem. Commun ., 1526-1528 (1990).
15. W. J. Potscavage, S. Yoo, B. Domercq, and B. Kippelen, “Encapsulation of pentacene/C-60 organic solar cells with Al2O3 deposited by atomic layer deposition,” Appl. Phys. Lett . 90, 3 (2007). [CrossRef]
16. A. Hamed, Y. Y. Sun, Y. K. Tao, R. L. Meng, and P. H. Hor, “Effects of oxygen and illumination on the in situ conductivity of C-60 thin-films,” Phys. Rev . B 47, 10873–10880 (1993). [CrossRef]
17. S. R. Forrest, “Ultrathin organic films grown by organic molecular beam deposition and related techniques,” Chem. Rev . 97, 1793–1896 (1997). [CrossRef]