A Porous Silicon-Based Optical Interferometric Biosensor

See allHide authors and affiliations

Science  31 Oct 1997:
Vol. 278, Issue 5339, pp. 840-843
DOI: 10.1126/science.278.5339.840


A biosensor has been developed based on induced wavelength shifts in the Fabry-Perot fringes in the visible-light reflection spectrum of appropriately derivatized thin films of porous silicon semiconductors. Binding of molecules induced changes in the refractive index of the porous silicon. The validity and sensitivity of the system are demonstrated for small organic molecules (biotin and digoxigenin), 16-nucleotide DNA oligomers, and proteins (streptavidin and antibodies) at pico- and femtomolar analyte concentrations. The sensor is also highly effective for detecting single and multilayered molecular assemblies.

Biosensors have been developed to detect a variety of biomolecular complexes, including oligonucleotides (1-4), antibody-antigen interactions (5, 6), hormone-receptor interactions (7), enzyme-substrate interactions (8, 9), and lectin-glycoprotein interactions (10). In general, biosensors consist of two components: a highly specific recognition element and a transducer that converts the molecular recognition event into a quantifiable signal. Signal transduction has been accomplished with electrochemical (11), field-effect transistor (12), optical absorption, fluorescence, interferometric (13) and other devices (14). Here we describe an optical interferometric transducer scheme based on inexpensive and readily available optically flat thin films of porous silicon (PSi). This material has been used for highly sensitive detection of small molecules (biotin and the steroid digoxigenin), short DNA oligonucleotides (16-nucleotide oligomers), and proteins (streptavidin and antibodies). Most notably, the sensor can be highly effective in detecting multiple layers of biomolecular interactions, termed “cascade sensing,” including sensitive detection of small molecules.

Recent studies have shown that certain electrochemical etches of single-crystal p -type (boron-doped) Si wafers produce microporous material (PSi) that displays well-resolved Fabry-Perot fringes in its reflectometric interference spectrum (15). In our sensor (Fig. 1), reflection of white light at the top and bottom of the PSi layer results in an interference pattern that is related to the effective optical thickness (product of thickness L and refractive index n ) of the film by Eq. 1 (16)Embedded Image(1)where m is the spectral order and λ is the wavelength of light. Binding of an analyte to its corresponding recognition partner, immobilized on the PSi substrate, results in a change in the refractive index of the layer medium and is detected as a corresponding shift in the interference pattern. Electrochemical etching of Si generates a thin (1 to 5 μm) layer of porous Si on the Si substrate with cavities as wide as 200 nm in diameter, providing a large surface area for biomolecular interaction inside the porous layer. The films are uniform and sufficiently transparent to display Fabry-Perot fringes in their optical reflection spectrum (17).

Figure 1

Schematic of the PSi-based optical interferometric biosensor. The silicon oxide surface of the porous layer can be modified to express various molecular recognition elements (such as oligonucleotides, biotin, or antibodies). Reflection of white light (W-lamp source) at the top and bottom of the PSi layer results in an interference pattern (Fabry-Perot fringes). The reflectometric interference spectrum is sensitive to the refractive index of the PSi matrix. Interactions of the molecular species with their recognition partners immobilized on the surface induce a change in the refractive index of the nanocrystalline semiconductor, giving rise to wavelength shifts in the fringe pattern that can be easily detected [charge-coupled device (CCD) camera] and quantified.

We used DNA oligonucleotide–derivatized PSi films to test the selectivity and limits of detection (18). In the presence of complementary DNA (cDNA) sequences (DNA concentrations ranging from 2 × 10−15 to 2 × 10−6 M), pronounced wavelength shifts in the interference pattern of the PSi films were observed (Fig. 2, A and B). Under similar conditions but in the presence of non-cDNA sequences, no significant shift in the wavelength of the interference fringe pattern was detected—only minor amplitude fluctuations were observed. We used fluorescence spectroscopy to independently investigate the surface coverage of immobilized DNA on PSi and the rate of analyte diffusion into the matrix. Solutions of fluorescein-labeled cDNA oligonucleotides were placed in fluorescence cuvettes, and the DNA-derivatized PSi sample was then added to the cell without stirring. At the lowest DNA concentrations used, the fluorescence intensity of the samples decreased to an asymptotic limit in 40 min (similar equilibration times were observed in the interferometric measurements described above). The data indicate that 1.1 × 10−12 mol of DNA is covalently bound in the region beneath a 1-mm2 spot on the PSi layer (calculated from standardized fluorescence titration curves). The data obtained from the reflectometric interference measurements also provided a similar coverage number. The lowest DNA concentration measured with the PSi interferometric sensor was 9 fg/mm2(Fig. 3). For comparison, the detection limits of current technologies are as follows: 1 pg/mm2 for interferometry, 5 pg/mm2 for grating couplers, and 0.3 pg/mm2 for surface plasmon resonance (19).

Figure 2

Interferometric reflectance spectra of DNA-modified PSi layers. Experiments were measured for two DNA sequences (DNA-A: 5′-pGC CAG AAC CCA GTA GT-3′ and DNA-B: 5′-CCG GAC AGA AGC AGA A-3′) and corresponding complementary strands (DNA-A′ and DNA-B′). For clarity, only one set of data is shown in each case. (A) The Fabry-Perot fringes from a PSi surface derivatized with DNA-A (“before hybridization,” red trace) shift to shorter wavelength upon exposure to a 2 × 10−12 M solution of DNA-A′ (the cDNA sequence to DNA-A) in 1 M NaCl(aq) (“after hybridization,” blue trace). The net change in effective optical thickness (from 7986 to 7925 nm) upon DNA-A′ recognition is represented by the difference between the two interference spectra (“difference,” green trace). (B) The control experiment, showing the Fabry-Perot fringes of a DNA-A–derivatized PSi surface before and after exposure to a 2 × 10−12 M solution of DNA-B (non-cDNA sequence) in 1 M NaCl(aq). No wavelength shift was observed up to the measured concentration of 10−9 M DNA-B.

Figure 3

Change in effective optical thickness in a DNA-A–modified PSi layer as a function of DNA-A′ (the cDNA sequence) concentration [25°C, 1 M NaCl(aq); equilibration time, 30 min]. The changes correspond to a net decrease in effective optical thickness upon hybridization by DNA-A′. The value of the change in effective optical thickness is defined as the measured effective optical thickness before hybridization minus the effective optical thickness 30 min after addition of the solution of DNA-A′. The sensor is responsive to femtomolar DNA concentrations.

We used biotinylated PSi to investigate sensing of multiple layers of biomolecular interactions (cascade sensing) and small molecule detection (20). Exposure of a biotinylated PSi sample to a streptavidin-containing solution resulted in a large blue shift of the interference fringes that corresponded to a decrease in the measured effective optical thickness (the lowest streptavidin concentration used was 10−14 M) (Fig. 4A). Control experiments in which a biotinylated PSi sample was exposed to inactivated streptavidin (streptavidin presaturated with biotin) did not display perceptible shifts in the interference pattern. The biotin-streptavidin monolayer surface was treated with biotinylated antibody to mouse immunoglobulin G (IgG, from goat IgG). Binding of this secondary antibody to the surface was indicated by a further decrease in effective optical thickness of the monolayer (the lowest tested concentration used was 10−12 M) (Fig. 4B). Treatment of the secondary antibody sample with antibody to digoxigenin (mouse IgG) at a concentration of 10−8 M caused a further decrease in the effective optical thickness of the monolayer (Fig. 4C). Remarkably, the interaction of digoxigenin (10−6 M), a steroid with molecular weight (MW) of 392, with the IgG antibody to digoxigenin bound to the PSi surface was also detected (Fig. 4D).

Figure 4

Cascade sensing and reflectometric interference spectra of a multilayered molecular assembly. (A) A biotinylated PSi sample (red) treated with a 5 × 10−7 M streptavidin solution (blue) (effective optical thickness decreased from 12,507 to 11,994 nm). (B) Streptavidin-bound PSi sample (red) treated with a 10−8 M solution of biotinylated goat IgG antibody (2° antibody to mouse antibody to digoxigenin) (blue) (effective optical thickness decreased from 11,997 to 11,767 nm). (C) A biotinylated IgG antibody (to mouse antibody to digoxigenin) bound to a PSi sample (red) treated with a 10−8 M solution of mouse IgG antibody to digoxigenin (1° antibody) (blue) (effective optical thickness decreased from 11,706 to 11,525 nm). (D) A mouse IgG antibody to digoxigenin bound to a PSi sample (through the complex shown) (red) treated with 10−6 M digoxigenin solution (blue) (effective optical thickness decreased from 11,508 to 11,346 nm). The red traces correspond to the optical measurement made before exposure to the analyte of interest, the blue traces correspond to the measurement made after exposure, and the difference spectra (the difference between the red and blue traces) are represented as the green trace. All experiments were performed in 0.5 M aqueous NaCl at 25°C.

To rule out the possibility of nonspecific interaction, we subjected a nonbiotinylated surface to the same experimental protocol as described above. No measurable change in the effective optical thickness was observed on treatment with streptavidin, secondary antibody, primary antibody, and digoxigenin. As in all affinity-based sensors, interference arising from nonspecific interactions with the recognition elements is to be expected. Interference from nonspecific adsorption to the surface-bound antibodies was not tested in the current study. We were also able to detect the relatively small biotin molecule (MW = 244) at concentrations as low as 10−12 M with biotin-streptavidin–modified PSi. The sensitivity of the system described is remarkable, especially in light of its ability to detect multiple layers of large- and small-molecule interactions even in cases where the recognition sites are apparently far removed (on the order of nanometers) from the Si surface.

Our hypothesis regarding this sensitivity is as follows. Selective incorporation or concentration of an organic analyte in the PSi layer can modify the refractive index in two ways: It should increase the average n of the medium in the pores by replacing water ( n = 1.33) with organic matter ( n = 1.45), and it can also decrease the n of PSi by modifying the carrier concentration in the semiconductor (21-25). A net increase in n is expected to shift the interference spectrum to longer wavelength, but in all of our binding studies, a shift to shorter wavelength was seen, an indication that the induced change in the semiconductor overwhelms the change in n occurring in the solution phase.

The reduction of n of the nanocrystalline Si substrate on binding of biomolecules is unexpected but is apparently responsible for the high sensitivity of this technique. The effect of interfacial capacitance on n is not easy to predict, especially for a material that is strongly absorbing in the wavelength region of observation (26). However, binding of molecules to semiconductor surfaces is known to modify carrier concentrations substantially (22, 23). For example, exposure of PSi to alcohols or water vapor can dramatically increase the conductivity of the porous layer (24), and binding of molecules to II-VI semiconductor surfaces substantially modifies carrier concentration in the space-charge region (27). In our system, molecular complexation presumably reduces interfacial capacitance, as predicted by the Gouy-Chapman double-layer model (28), and in turn expels charge carriers from the PSi fibrils into the bulk semiconductor. Reduction of the carrier density effectively reduces the n of the layer, shifting the Fabry-Perot fringes to higher energy.

We observed a similar effect in test experiments with aqueous NaCl solutions and unmodified oxidized PSi samples; the Fabry-Perot fringes initially shifted to the blue as NaCl was added to a sample immersed in deionized water, and then at higher NaCl concentrations (after establishment of a large double-layer capacitance), an increase in ion concentration caused a red shift in the fringes. Whether or not a double-layer capacitance-induced dielectric change is the correct mechanism, the change in n of the PSi layer is much greater than the change expected by replacement of water with biomolecules and provides higher sensitivity over existing optical interferometric detection schemes.


View Abstract

Navigate This Article