## Abstract

An optical method is presented for in the situ monitoring of biomolecular films via reflection microscopy on patterned substrates. The method is based on measuring the reflection coefficient of a composite consisting of a substrate, a patterned optical layer, the thin film to be monitored and the cover medium. The optical layer is patterned so that an array of squares is surrounded by the bare substrate. The reflectance difference between the optical layer squares and the bare substrate is the observable, whose fractional changes reveal the thickness of the film through a simple analytical expression. The periodic image is recorded by a digital microscope, and through Fourier transform techniques, the normalized differential reflectance of the patterned optical composite is calculated as the contrast factor of two dimensional bit map. The method is demonstrated by measuring a protein binding assay inside a microfluidic module placed under a microscope.

© 2013 OSA

## 1. Introduction

Methods for in situ and accurate monitoring of molecular film growth on solid substrates are valuable tools for the analytical science. They allow study of deposition rates, adsorption or desorption processes and association/dissociation kinetics in the case of binding bioassays. Especially in the case of protein binding or DNA hybridization assays, label free detection in real time provides valuable results without the interfering effects of fluorophore or nanoparticle labels. If existing equipment already used for general purpose testing, such as optical microscopes, can be employed with minimal additions to monitor the thin film growth the benefits are obvious. There is a variety of transducers capable of measuring material deposition or film growth. We choose to work with optical transducers to take advantage of the galvanic isolation between the excitation and the detection electronics and the versatility in choosing wavelength while manipulating light through the laws of optics [1]. A wealth of data regarding biomolecular intereactions has been derived from label free methods such as surface plasmon resonance [2,3], waveguide based interferometry [4], reflectometric interference spectroscopy [5], or ellipsometric methods such as oblique-incidence reflectivity difference [6,7]. In these optical arrangements, dedicated set-ups involving light sources (lasers, LEDs or white light sources) and detecting devices (photodetectors or spectrometers) are precisely coupled through focusing optics or optical fibers while often polarizers and other optical components intervene. Several of these functions are summarized in a modern optical microscope equipped with a digital camera and if such an instrument can be employed for label free detection the benefits are obvious. Here, we demonstrate how a reflection microscope can be used to monitor the film thickness on a patterned substrate by using the digital images as reflectance bit maps that reveal the optical path changes due to the film growth. An optical layer on top of a substrate is patterned as an array of squares while bare substrate surrounds the squares. The biomolecular film is grown on top of the patterned substrate. The optical layer thickness is chosen so as to maximize, for a specific spectral band, the sensitivity of the optical composite reflection coefficient to any film deposition while in the bare substrate areas the reflection coefficient has zero sensitivity to thin film growth. This way the differential reflectance between the optical layer square islands and the bare areas will change during film growth and its fractional change provides the adlayer thickness through a simple formula. The differential reflectance is very accurately calculated by two dimensional Fourier transform of the patterned substrate digital image. In the transform inverse space the response will sharply peak at the point inversely proportional to the real space periodicity. Such response is immune to parasitic reflections originating in the cover optical windows of the fluidic structure and translates into a direct measure of the adlayer thickness. To demonstrate the method, a digital microscope will be employed as a biomolecular film growth monitor, the only additions being the use of a microfluidic structure to supply the appropriate reagents to a patterned substrate placed on the microscope chuck. The next section outlines the theoretical analysis of the differential reflectance and its optimization in terms of choosing the appropriate optical layer and substrate. In the subsequent sections the optical and fluidic setup will be described and then the experimental results will be presented and discussed.

## 2. Differential reflectance at normal incidence as an analytical tool

This section will outline the theory of maximizing differential reflectance changes induced by the growth of a film on top of an optical layer already in place on a substrate. The term differential refers to the reflectance dependence on the optical layer thickness. The use of patterned substrates with optimized dual layer thicknesses and the employment of microscope images to isolate the film induced normalized reflectance changes allows calculation of the grown film thickness through a simple analytical expression.

#### 2.1. Optimization of the optical stack in terms of the reflectance sensitivity to film growth

An optical stack consisting of two semi-infinite media (cover medium, n_{0}, and substrate, n_{3}), the film under study (n_{1}, d_{1}), and the optical layer (n_{2}, d_{2}) is shown in Fig. 1
.

The overall reflection coefficient for light incident at a normal angle through the cover medium is [8]

_{i}(i = 1,2) are the normalized optical lengths in the film and optical layer, andwhere r

_{i}(i = 1,2,3) are the Fresnel coefficients, and

Our objective is to maximize the magnitude of the derivative ∂R/∂d_{1} and use this parameter in normalized terms as an observable for the calculation of the film thickness. The selection of materials to optimize ∂R/∂d_{1} is usually restricted to the optical layer (n_{2}, d_{2}) and the substrate (n_{3}) since the cover medium is determined by the process to be followed for the film growth. For instance, if a protein binding assay is to be monitored, the cover medium will be the biological sample or assay buffer with n_{0}≈1.34, while the protein layer has a refractive index n1 usually between 1.38 and 1.45. Therefore the emphasis from now on will be on how to maximize the magnitude of the derivative ∂R/∂d_{1} in terms of (n_{2}, d_{2}) and n_{3}. By differentiating Eq. (1) with respect to d_{1}, we obtain

_{1}<< 1, that is, we are dealing with very thin films. Then Eq. (5) becomes

_{1}= 0 (or n

_{0}= n

_{1}) since in this case there is no optical differentiation between the cover medium and the film under growth. More importantly, the same sensitivity vanishes when the optical layer (δ

_{2}, n

_{2}) is absent since in this case either δ

_{2}= 0 or r

_{3}= 0 andIn a patterned substrate, therefore, where the optical layer has been selectively removed, the bare substrate areas exhibit zero sensitivity to the film growth. On the contrary, in the areas bearing the optical layer the sensitivity of R to the grown film thickness is maximized if appropriate thickness δ

_{2}is chosen. To find the value of δ

_{2}(δ

_{2m}) that maximizes the absolute value of the derivative in Eq. (6) we set

In most of the cases |FR|«1 and |CR|«1 so that Eq. (10) simplifies to cos(2δ_{2m})≈0, that is the optimum thickness is such that a photon that makes a round trip through the optical layer will experience a phase difference very close to π(1/2±m), with m an integer. Considering that |FR|«1 and |CR|«1 hold, or equivalently |r_{1}|, |r_{2}|«1, then the extrema for the derivative in Eq. (6) become

_{2m}= (1/2 + 2m)π, or d

_{2}= (1/8 + 1/2m)λ

_{0}/n

_{2}, and plus when 2δ

_{2}= (3/2 + 2m)π, or d

_{2}= (3/8 + 1/2m)λ

_{0}/n

_{2}, with m an integer.

#### 2.2. Differential reflectance as an intrinsic parameter in structured substrate characterization

Looking at the simple formula in Eq. (11) one would think that it would be enough for monitoring of d_{1} to just measure the reflected beam on a one dimensional optical stack with optimized optical layer thickness. However, unwanted reflections in fluidic covers or various interfaces interfere with the measured sensitivity of R to d_{1}. In fact they will underestimate of the actual value of the d_{1} induced reflectivity changes if normalized with respect to the initial R value at d_{1} = 0. In order to suppress unwanted reflection, we propose another method based on the measurement of the reflectance difference between sites with different optical layer thicknesses. In fact, the optical structure is partitioned into two interpolated array areas: one with no optical layer (bare substrate) where the initial reflectivity is R(0,0) and another bearing an optical layer with the optimized optical thickness δ_{2m} = $2\pi {n}_{2}{d}_{2}{}_{m}/{\lambda}_{0}$ where the initial reflectivity is R(0,d_{2m}). If now the observable is the differential reflectance:

*R*with respect to d

_{1}, and since the derivative of the first term of the right hand side in Eq. (12) is zero, we obtain with the help of Eq. (6)

_{2m}thickness is from Eq. (1)

_{2m})≈0 and that the absolute values of r

_{1}and r

_{2}are close to zero, an assumption usually valid for a realistic choice of materials, then

The sign in Eq. (16) is plus when d_{2} = (1/8 + 1/2m)λ_{0}/n_{2}, and minus when d_{2} = (3/8 + 1/2m)λ_{0}/n_{2}. The accuracy of Eq. (16) is shown in Fig. 2
with δ_{2} = δ_{2m} = 3π/4 and it is compared to the exact expressions obtained by dividing the right hand sides of Eq. (13) and Eq. (14). Two different values and far apart were chosen for n_{3}: 4 (silicon) and 1.46 (quartz). The value for the optical layer refractive index n_{2} was swept from 1.4 to 2.2 covering the range of transparent dielectrics, while n_{0} = 1.34 (bioassay buffer). As shown in Fig. 2, Eq. (16) is quite accurate and points out that the normalized differential reflectance is inversely proportional to the (n_{0}-n_{2}) difference and nearly independent of the substrate (n_{3}). This is true especially for n_{3} in the vicinity of n_{2}, but even for n_{3} = 4 it is still an excellent approximation for n_{2} less than 1.5. Judging by the results from two different values of the film refractive index, n_{1} = 1.4 and 1.45, we conclude that the sensitivity is proportional to the n_{1}-n_{0} difference (0.06 and 0.11 for the upper and the lower set of curves, respectively), as predicted by Eq. (16).

The significance of Fig. 2, or Eq. (16), lies in the fact that the film thickness d_{1} can be calculated dividing the fractional change of the measured differential reflectance by the sensitivity values in Fig. 2, provided of course that δ_{1}«1 as is the case for biomolecular films of few nm. Using the simplified Eq. (16) we obtain:

_{0}, n

_{0}, n

_{1}, n

_{2}). Finally, the values of cos(2δ

_{2m}) from Eq. (10) are shown in Fig. 3 which proves that the approximation cos(2δ

_{2m})≈0 is very good for the range of values explored here.

#### 2.3 Differential reflectance measurement through Fourier transform on periodic microscope images and choice of materials

Measuring ΔR requires simultaneous measurement of the reflectance at different sites and this is best done by optical reflection microscopy. The fractional changes of the differential reflectance are the same as the fractional changes of the contrast in the microscope image of the same patterned substrate. In fact, if a periodic pattern is chosen with interpolated areas having an optical layer thickness of either 0 or d_{2m}, then Fourier transform techniques applied on the periodic image bit map can isolate the changes in ΔR with a high degree of accuracy by isolating the changes in the image contrast factor. Now, as for the choice of materials to maximize the non-normalized reflectance sensitivity to d_{1}, Eq. (11) points to an optical stack with a relatively high |r_{3}| provided r_{3}^{2}<1/3. This can be achieved by choosing substrate material of high n_{3}, such as silicon. Such a substrate will have the added benefit of being opaque to optical frequencies, so that there are no interfering beams reflected from the back substrate side. The choice of the optical layer (n_{2}) comes from optimizing the normalized differential sensitivity in Eq. (16). Small values of the n_{0}-n_{2} difference lead to higher normalized differential reflectance. When n_{0} = n_{2}, ΔR_{0} = 0 and normalization with respect to ΔR_{0} is no longer possible. Also in the case r_{1,2} ≈0, implying ΔR_{0}≈0, noise related uncertainties in the measured value of ΔR_{0} can lead to large uncertainties in determining d_{1} through Eq. (16). Another issue when ΔR_{0} approaches zero is that it will be nearly impossible to focus on the patterned optical stack under the microscope. For the aforementioned reasons small n_{0}-n_{2} values are desirable provided the patterned optical stack is still imageable. If Si is chosen as a substrate, then SiO_{2} is a natural choice for the optical layer not only in terms of easiness of fabrication but also in terms of its refractive index proximity to the cover medium. In practical cases the most usual cover media is either vacuum or air (n_{0}≈1) or water buffer solutions (n_{0}≈1.34) in the case of bioassays. Therefore the 1.46 n_{2} value of SiO_{2} seems to satisfy the above mentioned conditions.

Having outlined the basic considerations in terms of geometry and choice of materials, the demonstration of the method will be presented in the sections to follow and consists of a monitoring the progress of a protein binding assay through optical microscopy on a silicon wafer with a patterned silicon dioxide film. Therefore from now on n_{0} = 1.34, n_{2} = 1.46, n_{3} = 4. For the protein refractive index two values were considered: n_{1} = 1.4 and n_{1} = 1.45. In a microscope a spectral band is more practically realized through the use of band pass optical filters as opposed to monochromatic light. The use of a narrow spectral band instead of monochromatic light has negligible effect on the differential reflectance sensitivity to film growth, as is shown in Fig. 4
. Here, the normalized differential reflectance sensitivity, calculated from Eq. (13) and Eq. (14) at λ_{0} = 450 nm, is plotted along with the averaged differential reflectance sensitivity defined as $(\Delta R({d}_{1},{d}_{2}{}_{m})-\Delta R(0,{d}_{2}{}_{m}))/({d}_{1}\Delta R(0,{d}_{2}{}_{m}))$ with d_{1} = 1nm and with ΔR calculated by averaging R(d_{1},d_{2m}), R(d_{1},0) and R(0,d_{2m}) from (1) over the spectral range 425-475nm. Therefore, the conclusions derived through the above equations obtained for monochromatic light still hold for narrow spectral bands where now λ_{0} is the band center. Finally, for maximum sensitivity the oxide thickness is chosen as d_{2m} = 3/8(λ_{0}/n_{2}) or δ_{m} = 3π/4. With λ_{0} = 450 nm, d_{2m} assumes the value of 115 nm. Now, Eq. (16) holds with the minus sign which implies that protein binding decreases the differential reflectance.

## 3. Experimental method

The instrument employed for the reflectance measurements was an optical microscope equipped with a digital camera (OLYMPUS DP71). Apart from the bioreagents, the only additions to perform the experiments, was a fluidic module as described below.

#### 3.1 Optical and bio-fluidic set up

The silicon wafer was oxidized at 1030 C° and a 100 nm SiO_{2} layer was grown. The oxide layer was patterned through lithography and etching in buffered HF solution, so that a periodic pattern of oxide islands was created with a pitch of 20 microns (Fig. 5
). The wafer was silanized with APTES and coated with biotinylated Bovine Serum Albumin (b-BSA). The total combined thickness of the optical layer (SiO_{2} + APTES + b-BSA) is estimated near 110 nm, very close to d_{2m} = 115 nm.

The wafer was placed under the microscope in a fluidic structure consisting of two Al plates, a quartz plate, a PDMS gasket, and plastic tubing (Fig. 6 ). The top Al plate had a quartz window so that the patterned silicon wafer could be monitored in real time by focusing through the quartz cover and the buffer solution while biomolecular solutions were pumped through the fluidic channels by a micropump. The microscope light was filtered in the blue (425-475nm, Fig. 6(a)) and an objective 10X was used. The gasket was made by casting PDMS in a mold bearing the plastic tubing for the supply of reagents. After the molding was complete the part of the tubing left in the reaction areas (shown as the four diamond shaped openings in Fig. 6) was removed by a sharp blade. Four screws were holding the two Al plates in place and provided the required pressure so that the PDMS gasket would seal the reaction areas. The digital files (12 bit vertical resolution, 1 picture (3072x3072 bits) per 4 seconds) of the pictures taken by the digital microscope camera were processed to monitor the reaction between the immobilized b-BSA and the streptavidin in the solution.

#### 3.2 Coating and Biomolecular reagents

Biotinylated BSA was prepared according to a published method [9]: The protein was reacted with sulfo-NHS-LC-biotin in a 0.3 M sodium carbonate buffer (pH 9.2) at a molar ratio of 1:120 for 2 h at room temperature. Then, the mixture was extensively dialyzed against 0.1 M NaHCO3 solution, pH 8.5, containing 0.15 M NaCl and 15 mM sodium azide and was stored at 4 °C. Biotinylated BSA was diluted at a concentration of 25 mg/l in 50 mM phosphate buffer, pH 7.4. The solution was then applied to the patterned wafer surface and incubated for two hours at room temperature. Then, the wafer was washed with 50 mM phosphate buffer, pH 7.4 and subsequently immersed in a blocking solution of 10 g/l BSA solution in 50 mM phosphate buffer, pH 7.4 for one hour in order to cover the remaining free protein binding sites. After blocking the wafer was flashed with de-ionized water and blown dry before inserting in the fluidic structure of Fig. 6. The same blocking solution was used as the assay buffer during the reaction with streptavidin.

## 4. Experimental results and discussion

#### 4.1. Contrast factor through Fourier transform

The digital pictures taken are treated as bitmaps of the reflectivity across the imaged area. From the reflectivity map we isolate the differential reflectance that is, the difference in reflectance of the bare substrate from the oxide regions. This reflectance is free from interfering effects coming from the fluidic covers and the microscope internal optics and depends only on the intrinsic optical stack reflectance. For the optical stack chosen, the oxide reflectance increases with binding while the bare silicon reflectance stays practically the same resulting in a contrast decrease of the overall image. The differential reflectance is proportional to the contrast of the two-dimensional image bitmap. The two dimensional plots shown in Fig. 7(a,b) demonstrate the contrast change of the periodic pattern before and after the reaction. The blue squares are the silicon oxide square areas where the reflectance is lower compared to the reflectance of the surrounding bare silicon. The contrast shown is larger before the reaction because of the lower reflectance oxide regions (deep blue areas, Fig. 7(a)) compared to the same after the reaction (blue areas, Fig. 7(b)). Although in Fig. 7(a,b) the two dimensional plots show pictorially the reaction effects on the contrast, in a way allow visualization of the reaction, it takes digital signal processing through Fourier transform of the two dimensional image bit map to derive the reflectance difference at a good enough resolution by isolating the image contrast factor. At the point corresponding to the periodicity of the image in the direct domain, the transformed signal will contain a sharp peak (resonant peak) way above the noise level of the non resonant frequencies (Fig. 7(c)). This peak is directly proportional to the differential reflectance sought after. Parasitic signals, like the unwanted reflections from fluidic interfaces, will not affect the signal since they do not have the same periodicity. At the same time, normalization with respect to the incident light intensity is possible if the resonant peak is divided by the Fourier transform signal at zero frequency. The later signal is a measure of the average reflected light intensity and directly proportional to the incident intensity. Changes of the resonant peak over time provide a direct measure of the film built up and its thickness.

#### 4.2 Bioanalytical results

The normalized differential reflectance obtained as the resonant peak in the Fourier transform domain (Fig. 7(c)) is shown in Fig. 8
. The magnitude of the peak stays more or less constant in the first 120 pictures where only the reaction buffer (10 g/l BSA) flows (Fig. 8). Then, after the introduction of streptavidin in the buffer solution and the beginning of the specific reaction, it sharply declines and after assuming a maximum slope it starts leveling off. In the maximum slope region the reaction is controlled by the diffusion of the streptaviding towards the coated silicon surface and by the binding rate constant. As the reaction proceeds, the available binding sites diminish and the streptavidin film growth starts leveling off. The saturation tendency is more evident in the highest molarity plot (20 nM) which saturates at a differential reflectance drop of 18%. From Fig. 2, on normalized differential reflectance sensitivity, we conclude that such a drop corresponds to a streptavidin layer thicknesses of 8.5 nm in case n_{1} = 1.4 or 4.6 nm in case n_{1} = 1.45nm.

Such numbers are in excellent agreement with streptavidin layer thicknesses after saturation reaction with biotin coated surfaces [10]. The 5 nM reaction starts the saturation turn after the first 600 pictures and keeps approaching the 20 nM levels, while the 10 nm plot saturates at even higher thicknesses compared to the 20 nM. Such saturation effect anomalies appear [11,12] possibly due to second order effects favoring thicker saturation layers at more moderate analyte concentrations. Nevertheless, the maximum slopes do correspond with the analyte (streptavidin) concentration. The three slopes in descending order are: 1.4x10^{−3}, 5.7x10^{−4} and 2.8x10^{−4} per shot, in good agreement with the concentration order.

#### 4.3 Multianalyte capabilities

The above described method for the monitoring of biomolecular binding can be extended to cover cases where a number of different probes have been immobilized on the patterned silicon wafer leading to an equal number of different reactions taking place simultaneously, provided the appropriate mixture is supplied. In this case the field to be illuminated and monitored is segmented into the corresponding spotting areas, while the algorithm is modified so that each area is independently processed through Fourier transform to isolate the differential reflectance for each different spotting area. In such a case of course, the space bit count per area will be reduced resulting in higher noise levels. This places an upper limit to the number of different spotted areas that can be monitored simultaneously. However, such an effect can be counterbalanced if a higher spatial resolution (pixel number) camera is employed.

## 5. Conclusions

A method for monitoring molecular films was presented based on measurements of the differential reflectance of a substrate with a patterned optical layer. The film growth changes the reflection coefficient of the optical composite in a way decisively determined by the optical layer thickness. The optical layer is partitioned into two periodic arrays of maximum and zero reflectance sensitivity with regard to the molecular film growth. The maximum sensitivity regions are the optical layer islands and the zero sensitivity areas is bare silicon. The difference in the reflectance of the two regions is taken as the observable since it is free from interfering effects originating in fluidic cover or parasitic reflections. The tool for the simultaneous measurement of the reflectance from a multitude of points was an optical microscope equipped with a 12 bit digital camera. The differential reflectance was obtained by isolating the contrast factor of the periodic digital images processed through Fourier transform. By calculating the resonant peak fractional changes in the Fourier transform domain, the molecular film built-up was monitored in time through the theoretical analysis developed in section 1 and the simple formula of Eq. (16). As a demonstration experiment, label free protein-protein binding is monitored. Streptavidin solutions of various concentrations reacted with biotinlylated BSA coated on silicon wafer with a patterned SiO_{2} film. The only instrument employed here was a digital microscope plus an optical window equipped fluidic structure for the supply of reagents. After the introduction of the streptavidin solutions the signal changed proportionally to the analyte concentration and saturated at values in agreement to literature data for the same reaction. Multiplexed assay performance is possible if the illuminated field is partitioned into areas where different biomolecular probes have been immobilized and where independent resonant peaks are calculated.

## Acknowledgments

We thank Dr. P. Petrou and Dr. S. Kakabakos for the supply of bioreagents.

## References and links

**1. **A. P. F. Turner, “Biochemistry. Biosensors--sense and sensitivity,” Science **290**(5495), 1315–1317 (2000). [CrossRef]

**2. **X. W. Guo, “Surface plasmon resonance based biosensor technique: a review,” J Biophotonics **5**(7), 483–501 (2012). [CrossRef]

**3. **S. Herranz, M. Bockova, M. D. Marazuela, J. Homola, and M. C. Moreno-Bondi, “Surface plasmon resonance biosensor for parallelized detection of protein biomarkers in diluted blood plasma,” Anal. Bioanal. Chem. **398**(6), 2625–2634 (2010). [CrossRef]

**4. **M. S. Luchansky and R. C. Bailey, “Silicon photonic microring resonators for quantitative cytokine detection and T-cell secretion analysis,” Anal. Chem. **82**(5), 1975–1981 (2010). [CrossRef]

**5. **A. Brecht, G. Gauglitz, and J. Polster, “Interferometric immunoassay in a FIA-system - a sensitive and rapid approach in label-free immunosensing,” Biosens. Bioelectron. **8**(7-8), 387–392 (1993). [CrossRef]

**6. **X. Zhu, J. P. Landry, Y. S. Sun, J. P. Gregg, K. S. Lam, and X. Guo, “Oblique-incidence reflectivity difference microscope for label-free high-throughput detection of biochemical reactions in a microarray format,” Appl. Opt. **46**(10), 1890–1895 (2007). [CrossRef]

**7. **J. Y. Wang, J. Dai, L. P. He, Y. Sun, H. B. Lu, K. J. Jin, and G. Z. Yang, “Label-free and real-time detections of the interactions of swine IgG with goat anti-swine IgG by oblique-incidence reflectivity difference technique,” J. Appl. Phys. **112**(6), 064702 (2012). [CrossRef]

**8. **O. S. Heavens, *Optical Properties of Thin Solid Films* (Dover Publications, 1991) Chap.4.

**9. **M. Zavali, P. S. Petrou, S. E. Kakabakos, M. Kitsara, I. Raptis, K. Beltsios, and K. Misiakos, “Label-free kinetic study of biomolecular interactions by white light reflectance spectroscopy,” Micro & Nano Lett. **1**(2), 94–98 (2006). [CrossRef]

**10. **S. Busse, V. Scheumann, B. Menges, and S. Mittler, “Sensitivity studies for specific binding reactions using the biotin/streptavidin system by evanescent optical methods,” Biosens. Bioelectron. **17**(8), 704–710 (2002). [CrossRef]

**11. **P. Ihalainen and J. Peltonen, “Immobilization of streptavidin onto biotin-functionalized Langmuir-Schaefer binary monolayers chemisorbed on gold,” Sens. Actuat. B **102**(2), 207–218 (2004). [CrossRef]

**12. **M. Zavali, “Direct Study of biomolecular interactions through white light reflection spectroscopy” Senior Thesis, University of Ioannina, Greece, (2006).