Integrated semiconductor fluorescent detection

Integrated semiconductor fluorescent detection system for biochip and biomedical applications Evan Thrush*a, Ofer Levi*a, Ke Wang*a, Mark A. Wistey*a, James S. Harris Jr.*a, Stephen J. Smith*b a Solid State and Photonics Laboratory, Stanford University; bDept. of Molecular and Cellular Physiology, Stanford University ABSTRACT As biological analysis systems scale to smaller dimensions, the realization of small and portable biosensors becomes increasingly important. The innovation of integrated fluorescence sensors is now possible due to the development of optoelectronics over the past decade. We present the monolithic integration of vertical cavity surface emitting lasers (VCSELs), PIN photo-detectors and optical emission filters to be used as a fluorescence sensor. The integration will drastically reduce cost and size of fluorescence detection systems. Also, parallel sensing architectures of more than one hundred channels will be possible. The sensor will be utilized for near-infrared fluorescence detection. This spectral range is compatible with standard AlGaAs optoelectronic technology and will also reduce background fluorescence from complex bio-fluids such as blood. PIN heterostructure photodetectors have been fabricated and tested. Photodetector experiments show extremely low dark current of less than 500fA/mm, quantum efficiency greater than 85 percent and linear detector response. Optical simulations have been preformed to evaluate the performance of a proximity sensing architecture. These simulations predict a detection sensitivity lower than 10000 fluorescent molecules in a detection area of 104 µm2. Keywords: Biochips, fluorescence, sensors, infrared dyes, VCSELs, semiconductor lasers, photodetectors, AlGaAs

1. INTRODUCTION The integration and scaling of biological analysis systems will find wide applicability in the areas of bio-warfare, clinical medicine and biological experimentation. Medical diagnostics still rely upon labor intensive, timely, and expensive laboratory techniques. The experience of waiting for days or even weeks for the results of a test for infectious disease is a common reality of modern day medical practice. Small portable micro-total analysis systems (µTAS) promise to provide immediate point of care services that would facilitate detection of cancer, disease and bio-warfare agents1,2,3. Moreover, integrated biological analysis systems will provide high throughput experimentation and give scientists an understanding of complex biological processes. This will naturally lead to the discovery of innovative drugs and medical treatments. A variety of sensing schemes have been developed for molecular detection, such as electrochemical, optical absorption, and interferometric sensing4,5,6. However, fluorescence sensing remains the most widely used methodology in biotechnology. Separation technologies, such as capillary array electrophoresis and micro-array technology use fluorescent labeling for the detection of DNA and proteins. Fluorescence detection offers exquisite sensitivity and compatibility with standard biochemical reactions, such as polymerase chain reaction (PCR). Traditional bio-fluorescence readers use bulky and discrete elements, which are expensive and require large footprint and precise alignment7. The advantages of integrated biological analysis systems are reduced when these systems rely upon large and fragile optical sensing equipment. Integrated on-chip sensing architectures make portable and robust medical care equipment practical. As a result of the recent explosion in optoelectronics for telecommunications, a variety of interesting and useful integrated optical sensing architectures can now be realized8,9,10,11. The theme of our research is exploring sensing architectures that hold potential to be inexpensive, parallel, and micro-scale solutions. This paper presents progress towards the monolithic integration of vertical cavity surface emitting lasers (VCSELs), PIN photodetectors, and emission filters to achieve fluorescence sensing.

Biomedical Nanotechnology Architectures and Applications, Darryl J. Bornhop, et al., Editors, Proceedings of SPIE Vol. 4626 (2002) © 2002 SPIE · 1605-7422/02/$15.00

289

2. SYSTEM METHODOLOGY VCSELs have been an exciting area of research over the past decade. VCSELs emit normal to the wafer surface, which makes them attractive for systems level research. Vertical oriented devices facilitate parallelism or pixilated architectures. A variety of technologies, such as micro-optics and silicon electronics, can be wafer bonded to VCSEL arrays to form interesting hybrid systems, such as optical busses and transceiver arrays. VCSEL technology can also be relatively inexpensive due to a high level of integration, wafer level testing, excellent beam divergence characteristics and simple packaging12. Unfortunately as one moves to shorter wavelengths from the near-infrared regime, VCSELs become increasingly difficult to fabricate13,14. Green, blue and yellow VCSELs have not been successful; although, promising work continues in this area. The lack of visible emission limits the applicability of VCSELs to many relevant biological experiments, such as those involving green fluorescent protein (GFP). However, VCSELs can be successfully applied to experiments utilizing near-infrared dyes. This research will use dyes that have absorbance peaks in the range 750-800nm and emission peaks in the range of 780-830nm. Commercially available dyes (for example, IR-800, Li-Cor Inc.) meet these specifications and are useful for protein and DNA analysis. In recent years, the popularity of near-infrared dyes has increased due to low background fluorescence of biological samples in the infrared15. In analyzing complex bio-fluids, such as blood, near-infrared dyes have a significant advantage over visible dyes16. Figure 1 illustrates some general sensor architectures that are possible with vertical oriented optical devices, such as VCSELs, PIN photodiodes, and emission filters. The imaging architecture utilizes micro-optics, refractive or diffractive, for focusing the laser beam and collecting the fluorescence. The proximity sensor allows the laser beam to propagate freely to the biological sample, and a large area detector, which surrounds the VCSEL, collects fluorescence. The waveguide architecture utilizes gratings to couple the laser beam into the waveguide, where the evanescent tail of the waveguide mode excites the biological sample. Then, the fluorescence is collected by the waveguide and coupled out onto the detector1.

Fig. 1: Schematics of system architectures: A) imaging, B) proximity and C) waveguide.

290

Proc. SPIE Vol. 4626

In regards to the micro-optical design and overall system design, current work is being directed to filter through the many possible design alternatives to find the most optimal and practical solutions. Nonetheless, it is constructive to discuss these general architectures, shown in Figure 1, to get a feeling for the advantages and disadvantages of the different approaches. Imaging architectures will be limited by the inherent inefficiency of micro-optics, which typically have numerical apertures (NA) less than 0.2. However, imaging schemes are less susceptible to scattered background due to the ability to have pinhole detectors or confocal architectures7. The imaging architecture obviously requires more sophisticated fabrication technologies, but the increased level of integration due to the small area detectors may reduce the costs when compared to the proximity and waveguide architectures. The proximity sensor will be able to achieve a larger numerical aperture, which will be limited by the emission filter angular sensitivity and geometrical considerations. However, the proximity sensors will suffer from increased background, dark current, and cross talk due to the large area detector. In evanescent waveguide sensors, the evanescent tail probes only the first 100-200nm from the waveguide surface. This technology is most useful when combined with affinity-based technologies, where the analyte can be attached to the waveguide surface1. Typically, waveguide architectures require more laser power because only the evanescent tail of the waveguide mode excites fluorescence. One advantage of waveguide architectures is the high collection efficiency of emitted photons possible with high numerical aperture waveguides. In contrast to a “starring” or transmission architecture, where the pump source is emitting directly into the detector, it is advantageous to have the laser and detector located on the same substrate for several reasons. First of all, these onesided architectures essentially decouple the optoelectronic design from the biochip design. One could imagine having a re-usable optical monitoring chip, whereas the biochip is disposable. Also, with these one-sided approaches, the importance of the biochip’s optical quality is reduced. Perhaps the most important advantage of one-sided architectures is the significant reduction in scattered background from the laser. By not shining the laser directly into the detector, significant spatial filtration is achieved, which is extremely important for high sensitivity applications7,9.

3. OPTOELECTRONIC DESIGN Achieving lasing, photo-detection and filtration from one GaAs substrate in a practical and inexpensive way is a design challenge. We propose a device design that utilizes existing VCSEL technology. Figure 2 shows a schematic of the optoelectronic design. The VCSEL on the left includes two mirrors or distributed Bragg reflectors (DBRs) and a laser gain region. Adjacent to the VCSEL, a simple PIN photodetector is realized by adding an intrinsic GaAs region underneath the standard VCSEL epitaxial structure. The PIN photodetector utilizes the N-doped DBR as both an emission filter and electrical contact. The filtering behavior of the DBR is illustrated in Figure 3. The DBR is highly reflecting at the lasing wavelength of 780nm (R > 99.999%) and relatively transparent at the Stokes’ shifted dye emission. It is important to note that this design achieves a high quality photodetector and emission filter by one simple modification to a typical VCSEL design. This will result in reduced costs and higher yield when compared to other integration schemes.

Fig. 2: Optoelectronic device design. A VCSEL, PIN photodetector and emission filter are monolithically integrated.


291

DBR Filter, Emission and Absorption Charactierstics 1.2

Emission Arbitrary Units

1

Absorption Reflectivity

0.8 0.6 0.4 0.2 0 750

775

800

825

850

875

Wavelength (nm)

Fig. 3: Spectral characteristics of DBR filter and near-infrared fluorescent dyes.

The optoelectronics will be based on conventional AlGaAs technology. Intracavity contacted 780nm VCSELs with optical output powers in excess of 10mW should be possible13,14,17. In contrast to telecommunications applications, where one is coupling into optical fiber, multi-mode behavior is acceptable for sensing applications. An extremely high quality PIN heterostructure photodetector is possible with this technology as shown in Sec. 4. Typical AlGaAs DBRs can be grown to be at least 99.99% reflecting. Interference filters have an angular dependence, so this must be taken into account for non-collimated fluorescence collection. Due to the high index contrast of AlGaAs, the angular sensitivity of the DBR is drastically reduced. Computer simulations show that filtration of 105 is possible over a numerical aperture of .7 (Fig. 4). Nonetheless, additional spatial filtration will be needed in conjunction with the filter to achieve reasonable sensitivity, as shown with optical simulations (Sec.5).

Angular Dependence of DBR -2 log(Transmittance)

TM TE

-3 -4 -5 -6 0

20

40

60

80

Angle (Degrees)

Fig. 4: Illustrates the angular dependence of the DBR for both TM and TE modes of radiation.

292


As stated before, VCSEL performance degrades significantly as one moves to shorter and shorter wavelengths. This practically limits the dye absorption peak to be greater than 750nm. There exists another constraint on the dye characteristics that should be noted. The PIN detector utilizes a GaAs active region that is transparent to wavelengths greater than 877nm. Therefore, the emission peak of the dye should be less than 830nm to achieve efficient collection of the fluorescence. However, more exotic materials systems with lower bandgaps may be used as the detector active layer, such as InGaAs quantum wells. There is an engineering tradeoff between materials issues and dye constraints that must be weighed for a particular application. Unfortunately, this technology can only be used in single dye experiments; however, more exotic integration schemes and sophisticated dyes could make multi-dye experiments a reality. Scientists often tag molecules with different colors to achieve multi-molecule detection. However, the small number of spectrally unique dyes limits the spectral tagging of molecules. Due to the large emission line widths of organic molecules, detection of more than ten different molecular probes is a challenge. However, spatial mapping can be used to detect thousands of molecules simultaneously, such as in micro-array technology or affinity assays. The proposed technology will achieve multi-analyte detection by scanning or arraying the sensors. For example, 100 element spatial arrays will be possible with this technology. Another use of multi-dye experiments is for establishing a biological control or reference. As the fidelity of micro-array and biotechnology improves, biological controls can be achieved through other approaches, such as in Affymetrix arrays. Another advantage of the micro-array technology or affinity assays is the inherent gain in molecular concentration. As compared with the bulk bio-fluid, the spatially defined molecular probes integrate or increase the molecular concentration at the expense of hybridization time1. For some applications, this concentration gain may make PCR amplification unnecessary to achieve detection of rare analytes, simplifying preparation chemistry.

4. Photodetector Results Significant progress has been made towards the realization of the photodetector. Figure 5a shows the PIN photodetector epitaxial layer structure that was grown by molecular beam epitaxy (MBE). A standard liftoff process is used to define the top ring contact. Then, deep reactive ion etching is used to electrically isolate adjacent detectors by etching into the P+ substrate, forming mesa structures. A scanning electron microscope (SEM) image of the photodetector is shown in Fig. 5b.

Fig. 5a: Schematic of PIN mesa photodetector

Fig. 5b: SEM photograph of photodetector

A standard probe station is used to make electrical contact to the photodiodes in order to perform dark current and series resistance measurements. Current and voltage characteristics are measured with a semiconductor parameter analyzer (HP 4155B). Extremely low values of dark current are measured (Fig. 6). Low dark current is important for high sensitivity applications because dark current causes shot noise. These extremely low values of dark current will most likely make the detector shot noise negligible when compared to the detector readout circuitry noise and other parasitic noises. This will enable system architectures that utilize large area detectors, such as the proximity sensor. Since the detector dark current scales linearly with diameter, surface states on the detector perimeter are the dominant source of


293

dark current generation. After the mesa etch exposes the active region perimeter, the dark current is extremely sensitive to processing parameters. One must avoid oxidation of the photodetector sidewalls to maintain the low detector dark current. In future work, surface passivation methods, such as sulfur passivation, may be utilized to reduce the dark current even further.

Detector Dark Current at -5 V Bias Dark Current (fA)

600 400 200 0 0

200

400

600

800

1000

Detector Diameter (µm) Fig. 6: Dark current measured as a function of detector diameter.

DBR resistance can be extremely high and degrade device performance. This high resistance is caused by the discontinuous valence and conduction band structure of DBRs. Detectors utilizing a P-doped DBR result in unacceptably high values of detector series resistance, greater than 1kΩ for a detector with 25 µm diameter. As a result, we use an N-DBR filter, as shown in Figure 5. Due to the lower effective mass of electrons and the smaller band offsets in the conduction band, the detector measurements show much less detector series resistance. Detector series resistance is measured to be about 50 ohms for a detector with 25 µm diameter. The series resistance is negligible for detectors larger than 100µm diameter. The spectral response of the photodiode is measured. The response of the DBR is normalized out of the spectral response measurement. In other words, the spectral responsivity [A/W] is calculated by dividing the photocurrent by the optical power transmitted through the DBR. Unfortunately, the uncertainty in the DBR growth results in an uncertainty in the normalization. Over the spectral range of 800-870nm, the responsivity is found to be about .56 A/W with a standard deviation of .1A/W. This is a quantum efficiency of greater than 85percent, which is to be expected from a heterostructure PIN photodiode that utilizes a large absorbing region. It is also found that the spectral response is independent of the applied reverse bias, which is to be expected for photodiodes operating at low speeds and power. The photodetector shows an extremely linear response (Fig. 7). The experiment is performed with a 150 µm diameter illumination spot on a 200 µm diameter detector. Linear detection over nine orders of magnitude is measured, from 1pW to 1mW. Saturation effects are not observed, so the dynamic range may even be larger. This dynamic range should be more than adequate for most quantitative fluorescence experiments. The linearity does not depend on the applied voltage, which is expected again for low speeds and power.

294


Photocurrent (mA)

10

0

10

-2

10

-4

10

-6

10

-8

-10

10

10

-10

10

-8

10

-6

-4

10

-2

10

10

0

Optical power (mW) Fig. 7: Detector response vs. optical input power

5. OPTICAL SIMULATIONS The optoelectronics described in the above sections can be utilized in a staggering number of system architectures. This fact in conjunction with the many design tradeoffs and parameters needed to be considered has inspired our use of computer modeling to find optimal system designs. A ray-tracing program called ASAP from BRO (Tucson, Arizona) is being used to perform the optical simulations. The purpose of this section is to present some preliminary simulation results and give the reader a flavor of our optical simulation capabilities. The proximity sensor is simulated (Fig. 8). The detector inner and outer diameters are 100 µm and 1mm, respectively. The biological sample is located on a slide that is .5 mm from the sensor surface. The optical output power from the laser is 1mW. The laser beam diverges out to illuminate the sample with approximately a 50 µm diameter spot. We assume that 5000 molecules are within this illumination spot. Simulation results show that the photocurrent detected is 1pA. Cross-talk from adjacent pixels is .1pA, which is a worse case scenario and can be reduced by a variety of geometry and signal processing schemes. With our present scattering models, the simulation shows that only 5nW of laser radiation will be incident on the detector surface. This is an exciting result and shows that significant spatial filtration may be possible with the proximity architecture.


295

Fig. 8: Simulation of proximity detection architecture. Upper left: VCSEL shown in the center and detector surrounds it. Dye emission from bio-chip surface, located 500 µm from device, is shown as well. Upper right: intensity profile of VCSEL emission at bio-chip surface. Lower right: intensity profile of dye emission at detector surface.

5. CONCLUSIONS In conclusion, integrated bio-fluorescence sensor architecture and optimization considerations are described above. Monochromatic and high power light emitters integrated with high quality photodetectors should be possible. Preliminary experimental results show low detector dark current of less than 500fA/mm can be achieved. With a simple proximity sensor that utilizes no-optics, detection of fewer than 10,000 molecules in a detection area of 104 µm2 may be possible. Integration of a VCSEL, photodiode and optical emission filter on one GaAs enables simple and easy integration of several detection channels. The potential applications of such a sensor are numerous, ranging from µTAS systems to micro-array readers.

ACKNOWLEDGEMENTS We would like to acknowledge the Defense Advanced Research Project Agency (DARPA) as the primary funding agency for this project. Also, the National Science Foundation is supporting this work through a Graduate Research Fellowship. We would like to thank Breault Research Organization (BRO), Tucson, AZ for licensing ASAP to us for educational purposes.

REFERENCES 1. G. L. Duveneck, M. Pawlak, D. Neuschafer, E. Bar, W. Budach, and U. Pieles, "Novel bioaffinity sensors for trace analysis based on luminescence excitation by planar waveguides," Sensors and Actuators B 38-39, pp. 8895, 1997.

296


2. A. Manz, N. Graber, and H. M. Widmer, "Miniaturized total chemical-analysis systems: a novel concept for chemical sensing," Sensors and Actuators B pp. 244-248, 1990. 3. K. E. Petersen, W. A. McMillan, G. T. A. Kovacs, M. A. Northrup, L. A. Christel, and F. Pourahmadi, "Toward next generation clinical diagnostic instruments: scaling and new processing paradigms," Journal of Biomedical Microdevices 1, pp. 71-79, 1998. 4. R. E. Kunz, "Miniature integrated optical modules for chemical and biochemical sensing," Sensors and Actuators B 38-39, pp. 13-28, 1997. 5. E. Verpoorte, A. Manz, H. Ludi, A. E Bruno, F. Maystre, B. Krattiger, H. M. Widmer, B. H. van der Schoot, and N. F. de Rooij, "A silicon flow cell for optical-detection in miniaturized total chemical-analysis systems," Sensors and Actuators B 66, pp. 66-70, 1992. 6. A. T. Woolley, K. Lao, A. Glazer, and R. A. Mathies, "Capillary electrophoresis chips with integrated electrochemical detection," Anal.Chem. 70, pp. 684-688, 1998. 7. G. Ocvirk, T. Tang, and D. J. Harrison, "Optimization of confocal epifluorescence microscopy for microchipbased miniaturized total analysis systems," Analyst 123, pp. 1429-1434, 2001. 8. M. T. Gale, R. E. Kunz, and H. P. Zappe, "Polymer and III-V transducer platforms for integrated optical sensors," Optical Engineering 34, pp. 2396-2406, 1995. 9. J. C. Roulet, R. Volkel, H. P. Herzig, E. Verpoorte, N. F. de Rooij, and R. Dandliker, "Microlens systems for fluorescence detection in chemical microsystems," Optical Engineering 40, pp. 814-821, 2001. 10. M. Wiki and R. E. Kunz, "Wavelength-interrogated optical sensor for biochemical applications," Optics Letters 25, pp. 463-465, 2000. 11. G. M. Yee, N. I. Maluf, P. A. Hing, M. Albin, and G. T. A. Kovacs, "Miniature spectrometers for biochemical analysis," Sensors and Actuators A 58, pp. 61-66, 1997. 12. Henryk Temkin and Larry A. Coldren Carl W.Wilmsen. Vertical-cavity surface-emitting lasers: design, fabrication, and applications. Cambridge University Press, New York, 1999. 13. W. W. Chow, K. D. Choquette, M. H. Crawford, K. L Lear, and G. R. Hadley, "Design, fabrication, and performance of infrared and visible vertical-cavity surface-emitting lasers," IEEE J. Quantum Electronics 33, pp. 1810-1824, 1997. 14. J. D. Lambkin, "Red VCSEL technology," Compound Semiconductor 6, pp. 65-70, 2000. 15. D. B. Shealy, M. Lipowska, J. Lipowski, N. Narayanan, S. Sutter, L. Strekowski, and G. Patonay, "Synthesis, chromatographic separation, and characterization of near-infrared-labeled DNA oligomers for use in DNA sequencing," Anal.Chem. 67, pp. 247-251, 1995. 16. K. F. Buechler, P. McPherson, J. Anderberg, K. Nakamura, S. Lesefko, and B. Noar, "Microarray immunoassays in the microfluidic triage protein chip," Proceedings of the Fifth International Conference on Miniaturized Chemical and Biochemical Analysis Systems (µTAS 2001), Monterey, CA USA, pp. 42-44, 2001. 17. J. A. Lott and H. K. Shin, "Red and deep red vertical cavity surface emitting lasers with top and bottom AlGaAs/Al-oxide distributed Bragg reflectors and AlGa-oxide current apertures," 23rd International Symposium on Compound Semiconductors, 23-27 Sept.1996, St.Petersburg, Russia pp. 377-380, 1996. * [email protected]; phone 1-650-725-2774; fax 1-650-723-4659; http://snow.stanford.edu/~ethrush/; Stanford University, 420 Via Ortega, Stanford, CA, USA 94305


297