Phase-sensitive plasmonic biosensor using a portable and large field-of-view interferometric microarray imager

Nanophotonics, and more specifically plasmonics, provides a rich toolbox for biomolecular sensing, since the engineered metasurfaces can enhance light–matter interactions to unprecedented levels. So far, biosensing associated with high-quality factor plasmonic resonances has almost exclusively relied on detection of spectral shifts and their associated intensity changes. However, the phase response of the plasmonic resonances have rarely been exploited, mainly because this requires a more sophisticated optical arrangement. Here we present a new phase-sensitive platform for high-throughput and label-free biosensing enhanced by plasmonics. It employs specifically designed Au nanohole arrays and a large field-of-view interferometric lens-free imaging reader operating in a collinear optical path configuration. This unique combination allows the detection of atomically thin (angstrom-level) topographical features over large areas, enabling simultaneous reading of thousands of microarray elements. As the plasmonic chips are fabricated using scalable techniques and the imaging reader is built with low-cost off-the-shelf consumer electronic and optical components, the proposed platform is ideal for point-of-care ultrasensitive biomarker detection from small sample volumes. Our research opens new horizons for on-site disease diagnostics and remote health monitoring.


Phase sensitivity derivation of plasmonically enhanced interferometric interrogation
In this section we present the effect of the plasmonic phase and intensity modulations on the light that propagates through the optical elements of our interferometric imager, and we derive the phase sensitivity expression as a function of plasmonic resonance parameters. Moreover, we estimate the phase sensitivity of both the plasmonic and the control transparent substrates by evaluating the derived expression with the numerically computed plasmonic parameters.
The following notation is used in this section: Wavelength Time Speed of light in vacuum Angular frequency, 2 / Wave vector, 2 / Refractive index of the material to be detected The intensity of the incident light from the LED source on the plasmonic sensor surface is given by: The wave equations of the collimated, X and Y polarized and sheared beams incident on the plasmonic surface can be represented in the plane wave form: The transmitted light through the plasmonic Au-NHA surface is both intensity and phase modulated. The intensity and phase modulation functions vary spatially depending on the effective refractive index change associated with the material to be detected. The modulation functions are distinguished by "ON" and "OFF" tags, corresponding to the light transmitted through the patterned area (the sensing region) and bare area (the reference region), respectively. While the modulation functions of the reference region are determined by the bulk refractive index of the top media only, the sensing region functions also depend on the refractive index change that the material induces. Moreover, the modulation functions of Au-NHAs are polarization independent due to Au-NHAs' symmetric geometry.
The intensity modulation functions (also shown in Fig.2d) are given by: An interferogram is formed on the image sensor, once the light traverses the second savart plate and the polarizer (with polarization axis orthogonal to the first polarizer). A typical interferogram in Figure 1S shows four spatial regions of different intensities. Figure 1S: A typical optical path difference map of a silica circular spot measured on plasmonic Au nanohole array showing four regions corresponding to the double crescent that forms due to the interference of the sheared beams.
where the phase shift function between the signal and the reference is: Multiple image acquisitions (30 in our case) at different tilt positions of the top savart plate introduce a phase bias parameter ( ), which is only added to the -polarized component of the beam, and modulates the intensities in the R2 and R4 regions ( 2 = 4 ): Using the intensity variations in a series of interferograms taken at different values of , where 0 ≤ ≤ 2 , the phase shift function, ( , ), can be computed. This method is called phase-shifting interferometry (PSI), and is detailed in Terborg et. al 1 .
The phase shift value is then converted into optical path difference (OPD), in nm units, to convey a physical meaning using the relationship below: The OPD sensitivity of the plasmonic chips on the LIM system then can be defined as: Note that the transmission phase and intensity spectra only shift to higher wavelengths as increases without experiencing any change in shape. The rate of this red-shift is defined as bulk sensitivity, . Therefore, ( , ) can be written in terms of ( , ) as follows: Let = − ( − ) and we have: where we define a phase derivative parameter: The phase sensitivity can then be computed at the extraordinary transmission (EOT) peak wavelength, , which dominates the phase information measured by the image sensor by providing the largest number of photons. Therefore, Given that in our plasmonic Au-NHA system, ~656 [ ] and using the equation 19, we can estimate the phase sensitivity of the Au-NHA plasmonic system for small refractive index variations on the bare sensor around EOT peak in air medium ( = 1) as: On the other hand, the phase sensitivity for silica layers ( 2 = 1.45) as a function of its thickness ( 2 ) on a transparent control substrate with air background can be stated as: where In order to state the OPD sensitivity in conventional RIU units, we calculated the effective bulk refractive index change (∆ ) that a thin silica layer creates with air background. A silica film thickness to ∆ conversion relation is calculated using numerically computed Au-NHA bulk sensitivity and the silica thin film induced shift as: Figure 2S).
After rewriting equation 22, the phase sensitivity on transparent substrates becomes: [ / ] (24) Figure 2S: Au-NHA computed EOT peak sensitivity to a) varying silica layer thickness b) bulk refractive index change.

OPD contrast statistical data extraction
For statistical evaluation, OPD contrast data is extracted from large-area OPD maps using a graphical user interface (GUI), which was custom designed to handle microarrays with different geometric parameters (see Figure 3S). The OPD map, where each pixel represents 4.4 x 4.4 µm 2 area, is color scaled to increase visibility and the array to be considered is manually chosen on the GUI. After defining the grid parameters, unit cells, where a single microspot resides, are identified. Within each unit cell, the highest and the lowest valued "N" pixels are averaged and their difference is taken as "OPD contrast". The number of pixels (N) considered for the contrast computation depends on the microarray element dimensions (5% of the spot area) in order to keep consistency among different microarrays.
Using a set of interferograms, the computational OPD map generation results in a twin image (see Figure 1S), where the regions R2 and R4 have similar magnitude but opposite signed OPD values. Since we compute the OPD contrast using both of these regions, the experimentally acquired OPD contrast is about twice the original numerical estimate. Thus the numerically estimated sensitivities become ×

Statistical characterization of the fabricated plasmonic chips
In order to understand the effect of the process variations on the relevant plasmonic properties, we evaluated a large set of fabricated Au-NHA plasmonic chips by analyzing their transmission intensity spectra acquired using a conventional spectrophotometer connected to the light outlet of an inverted microscope. When the Au-NHA chips are illuminated by a broadband light source, they transmit a narrow bandwidth corresponding to their EOT resonance, shown in Figure 4S. We extracted the characteristic parameters of the resonance peak, such as its spectral position ( ) and bandwidth, which can be quantified either by considering the full width half max (FWHM) or the half width half max (HWHM) extracted from the sharper side of the curve (see Figure 4S). The results of this statistical study are presented in Table 1S. We measured three spectra per chip, 44 chips per wafer and four wafers, and computed the average standard deviations of the parameters within the chips, wafers and batch. We found that the deviation in the parameters are sufficiently small and only slightly increase from chip to the batch.