1. Introduction

A metasurface is a planar nanostructured surface that can control the properties of light, such as phase, amplitude, or polarizations by designing the geometry, the arrangement, or the material of meta-atoms. One notable application of optical metasurface is to generate structural coloration in miniaturized systems [1–3], where those supporting structural coloration are referred to as “color metasurfaces” in this paper. Structural coloration studies the color generation from the interaction of incident light with wavelength/subwavelength scale structures. Dynamic structural coloration leverages the tunable response of the optical elements to the change of environmental optical properties, such as the refractive index of the surrounding media. This ability to produce real-time optical response forms the foundation for colorimetric sensing, which has been widely used in the detection of humidity [4,5], pH [6], chemical compounds [7,8], or biological analytes [9–11]. One essence of using color metasurfaces is that they can be readily interrogated by using the human eye or a digital camera to read out the signal, whereas high-Q metasurfaces [12–15] require sophisticated spectrometers or advanced laser systems for high sensitivity. The sensitivity of the metasurface-based colorimetric sensor can be optimized by a well-designed optical response at visible wavelength.

Apart from high sensitivity detection, there is also a need to rapidly deliver analytes to the sensing surface without having to rely on slow Brownian diffusion. Most state-of-the-art biosensors face a general bottleneck due to Brownian diffusion limitation [16–19], which fundamentally hampers the overall performance of a biosensor. The process of transporting analytes onto the sensing surfaces via stochastic Brownian motion results in an unpredictable and long duration for incubation, ranging from several hours to multiple days, particularly in solutions with low analyte concentrations [16–20]. A long incubation time makes the sensing in low analyte concentration solution impractical, consequently hindering the utilization of a sensor's detection capacity. Hence, there is a need to develop an active method to rapidly transport analytes and speed up the incubation process.

2. Results

In this paper, we report a silicon-based color metasurface with high-contrast color responses that can be used for label-free and high-speed colorimetric sensing when combined with optofluidics. To address the diffusion limitations and promote active transport to the sensing sites, we have incorporated optofluidic flows to transport analytes [21–23]. The setup we used for this demonstration is schematically illustrated in Fig. 1(a). We employed a 973 nm laser to generate local heating atop the metasurface, which in turn created electrothermoplasmonic (ETP) flows to rapidly transport particles toward the metasurface as we applied an a.c. electric field across the microfluidic chamber [16,23–25]. This transport process is schematically illustrated in Fig. 1(b). The maximum temperature rise should be well-managed so that it is sufficient for generation of ETP flow but does no harm to biological particles like extracellular vesicles. Also, for the purpose of generating local heating, a thin layer of gold (8 nm) was deposited on the metasurface with a 3 nm chromium adhesion layer. ETP flows allow for rapidly transporting particles within a range of hundreds of microns and concentrating them on the illuminated surface. After the EVs were deposited onto the metasurface, we removed the liquid from the metasurface and the signal read-out was accomplished by a digital camera (Nikon DS-Fi1). The observed images were captured with true-color. As shown in Fig. 1(c) and (d), the existence of captured particles changed the local refractive index on the color metasurface and induced perceivable color change locally.

 

Fig. 1. (a) Experimental setup for fast incubation and color metasurface imaging. BS: beam-splitter. DM: short-pass dichroic mirror with 650 nm cutoff wavelength. (b) Schematic illustration of the metasurface’s color change triggered by EVs (extracellular vesicles) binding to the surface. Using optofluidic flows to rapidly concentrate particles and bring them down to the nanostructured substrate, the local color would change due to the refractive index contrast induced by the EV capture. Schematic of the color metasurface illuminated with white light (c) before EV concentration on the sensing surface, and (d) after EV concentration on the sensing surface. The color changes over the region with the captured EVs. The optimized metasurface enables a highly perceivable color change imaged by the CCD camera.

Download Full Size | PDF

The rest of the manuscript is organized as follows. First, we discuss the design of the color metasurface to generate structural colorations as the refractive index of the surrounding is tuned. Next, we discuss the optofluidic control enabled by ETP flow for rapid enrichment of EVs on the metasurface. Multiphysics modeling was used to simulate the temperature rise across the metasurface containing the chromium and gold coating layers, which contribute to the overall optical absorption and heat generation, as well as to model the ETP flow velocity. Finally, to test the feasibility of achieving read-out of EVs adsorption, we concentrated EVs, which are negatively charged, on the metasurface after functionalizing with cysteamine, which is positively charged. We note that for future applications involving the capture and quantification of EVs for diagnosis and treatment, the cysteamine layer may be replaced by a specific antibody targeting EV surface protein markers.

To achieve the proposed dynamic structural colorations, we utilized the strong dispersion and resonances of the array of amorphous silicon nanopillars at visible wavelengths [3]. The multipolar resonances of high-index materials have been shown to efficiently modulate light scattering [26–28] and create chromatic responses [1,3]. Of note, the ability to produce this vivid chromatic response is not a universal feature of arbitrary metasurfaces. In our optimized color metasurface, the silicon nanopillars have dimensions of 72 nm in height, 90 nm in radius and are arranged in a square lattice with 280 nm periodicity. The silicon nanopillars sit on a quartz substrate. This optimized metasurface harnesses the interplay between the symmetric and antisymmetric modes to achieve sensitive chromatic tunability.

To better interpret the interaction between symmetric and antisymmetric modes in the color metasurface, we employed a multipolar analysis and reconstructed the reflection spectrum based on the extracted multipolar information [29,30]. For the simplicity of analysis, we excluded the quartz substrate in the simulations reported in Fig. 2 to keep the symmetry of the nanopillars with respect to their mid-plane. The presence of the substrate in the actual experiments leads to discrepancies between the simulated spectral response shown in Fig. 2 and the measured spectra reported in Fig. 3. However, we need to point out that the performance of the color control and the mechanism of colorimetric EV detection remain unchanged [3]. We found that two dominant resonant modes exist in the visible wavelength range of interest. Their electric field distribution is illustrated in Fig. 2(a) and (b), with Fig. 2(a) illustrating the symmetric mode profile and Fig. 2(b) showing the antisymmetric distribution. Additional information about the calculation details of the multipolar analysis is provided in Supplement 1 section 1. Based on the symmetry of the multipolar radiation pattern with respect to the metasurface plane, the symmetric mode is primarily contributed by electric dipole (ED_x) and magnetic quadrupole (MQ_yz), while the antisymmetric mode arises from the interference of magnetic dipole (MD_y) and electric quadrupole (EQ_xz). Both the symmetric mode and antisymmetric mode undergo spectral shifts and change in strength as the environmental refractive index increases. As shown in Fig. 2(c) and (d), when the refractive index of the environment increases from air (n = 1) to EV (n = 1.38) [31,32], the dominating symmetric (electric dipole) response is enhanced and redshifted, while the antisymmetric (magnetic dipole) is suppressed. Consequently, as shown in Fig. 2(e) and (f), the reflection spectrum is redshifted with an enhanced value on the red wavelength end, which leads to a color change from green to red on the metasurface when the index of superstrate increases from air to EVs. The SEM micrograph in Fig. 3(a) displays the fabricated color metasurface comprised of amorphous silicon pillars on quartz substrate with the optimized geometry. The measured and simulated reflection spectra of this metasurface provided in Fig. 3(b) show good agreement for both water and air. The detailed description of the procedure for measuring the spectrum is included in Supplement 1 section 3. Notably, an increase of reflectance was observed in water at wavelengths greater than 600 nm, suggesting a change in color when placing the metasurface in water. We used a halogen lamp to illuminate the metasurface and observed its color under a microscope using an objective lens with a magnification of 40X and a numerical aperture of 0.75. The color appears green initially in air, as illustrated in Fig. 3(c), and changes into red upon submersion in water as illustrated in Fig. 3(d). This perceived color change is highly consistent with the color change predicted in the color space of Fig. 3(e), where the 1931 CIE (x,y) chromaticity coordinates were calculated from the measured spectrum in Fig. 3(b) [11,33]. The black circles marked (i) and (ii) indicate the colors reflected by the metasurface in air and water, respectively, and these colors are also highlighted in the corresponding insets of Fig. 3(e).

 

Fig. 2. Electric field distributions of the (a) symmetric and (b) antisymmetric modes at visible wavelength, respectively. From supplementary equations S1 to S5 of the multipole decomposition within one silicon pillar, we obtained amplitudes of the scattered plane waves originated from various multipolar components (c) in air (n = 1) and (d) in EVs (n = 1.38) respectively. From supplementary equation S6 and S7, we calculated reconstructed reflection spectra of the metasurface (e) and (f) using the respective nanopillar responses in air and in EVs shown in parts (c) and (d), respectively. In (e) and (f), ‘FDTD’ means the reflection spectrum numerically simulated from Lumerical FDTD software, and ‘Recon.’ refers to the spectrum calculated from the analytical model mentioned in Supplement 1 section 1.

Download Full Size | PDF

 

Fig. 3. (a) SEM micrograph of the fabricated metasurface for colorimetric detection. (b) Measured and simulated reflection spectra of the color metasurface shown in part (a) on quartz substrate with no metal films. The measured and simulated spectra show great agreement and indicate a vivid color change as expected when the environmental refractive index increases. (c) Optical images captured by digital camera when the metasurface is in air and (d) in water, respectively. The perceived colors are consistent with the colors predicted by the measured spectra shown in part (b). (e) CIE 1931 color space plot showing the color change predicted by the measured spectra of the well-tailored color metasurface versus a random metasurface. The color metasurface shows green (i) in air and pink (ii) in water, while the control group shows only a minor chromatic shift, from (iii) to (iv). The four insets show the colors corresponding to the coordinates in color space labelled as (i) to (iv).

Download Full Size | PDF

As a control group, we also fabricated another metasurface with an altered lattice constant and pillar radius. The silicon pillars of the control group had the dimensions of 72 nm in height, 120 nm in radius and 350 nm in lattice constant. The measured spectra of the control group are included in Supplement 1 section 2. Although the control group shows a variance of reflection signal beyond 700 nm wavelength, the calculated and observed color changes both demonstrated trivial color shift, as shown by the grey circles in the color space plot of Fig. 3(e). The calculated colors of the control group in air and water are also presented in the insets of (iii) and (iv), respectively. The optical images of the control group metasurface under the same imaging condition are also discussed in Supplement 1 section 2.

Once the color metasurface has demonstrated its ability to respond sensitively to the changes in the environmental refractive index, we proceeded to test the ability of this color metasurface to be integrated with optofluidics. Specifically, we focused on the possibility of inducing ETP flow into the sensing platform in this work. After depositing chromium and gold thin layers for the generation of ETP flows, we reconfirmed the color of the metasurface using a digital camera to ensure that it maintains its color-changing abilities as the index is tuned, as shown in Supplement 1 section 3.

The color metasurface was then integrated into a microfluidic chamber (120 µm in height), and a 2-mW focused laser beam (973 nm) with a focus spot diameter of 1.33 µm was illuminated onto the metasurface immersed in water to induce the heating effects. We carried out simulations of the temperature distribution within the illuminated metasurface (see Supplement 1, Section 4 for details). The maximal simulated steady-state temperature rise was 10 K, as shown in Fig. 4(a), making the final temperature in the microfluidic channel close to human body temperature and thus ensures the safety of delicate biological particles, such as proteins or EVs. For the geometry and material parameters of the microfluidic chamber used in our experiments, Joule heating due to the applied electric field was negligible [23].

 

Fig. 4. (a) Simulated steady-state temperature rise when using a 973 nm laser to illuminate on 3 nm chromium and 8 nm thick gold films coating on the color metasurface. The laser power was 2 mW and laser spot diameter was 1.33 µm, which are the values used in the experiments. (b) Trajectory map of ETP flow with 2 mW laser illumination and 10 V a.c. electric field. The frequency of the a.c. electric field is 3 kHz. The background colormap shows the radial velocity magnitude of the ETP flow in plane, and the orange lines are the extracted particle in-plane trajectories. The detailed procedure to track the trajectory of particles is provided in Supplement 1 section 5. (c) Color image of the metasurface region surrounding the laser-illuminated spot. It clearly shows local color changed due to the aggregation of EVs. The red circle indicates the spot size and position of illuminating laser. (d) Zoomed-in SEM image for the region shown in (c). The SEM image clearly shows the existence of EVs captured on the optical metasurface.

Download Full Size | PDF

We then measured the in-plane ETP flow velocities. An a.c. electric field with a peak-to-peak voltage of 10 V (i.e., an electric field intensity of 83,333 V/m) and a frequency of 3 kHz was applied to generate the ETP flow. As illustrated in Fig. 4(b), the measured velocity of ETP flows can reach over 30 µm/s across a radial distance of 150 µm, showing good agreement with the simulation results in Supplement 1 section 4. The process to measure the ETP flows is elaborated in section 5 of the Supplement 1 document. Since we are acquiring images of particle motion in the in-plane direction, only the in-plane velocity and not the out-of-plane velocity distribution was measured. The ETP flow forms a micro-vortex [34] in the microfluidic chamber to rapidly transport the suspended particles towards the hotspot. This active method of transporting particles overcomes the diffusion limits of Brownian motion and speeds up the incubation process as shown in the following. A more detailed Multiphysics modeling of the ETP flow is provided in Supplement 1 section 4.

We then performed an experimental demonstration of detecting EVs in low concentration solution (10⁷ EVs/mL). This concentration is below the normally reported concentration of EVs in blood, which is about 10⁹ to 10¹¹ EVs/mL [35–39]. To easily capture the concentrated EVs, we functionalized the color metasurface with cysteamine molecules before injecting the EV solutions. Cysteamine is a widely employed agent for capturing EVs on gold surfaces [40,41] due to its positive charge. Additional details regarding the functionalization process can be found in Supplement 1 section 3.

The ETP flow was maintained for a 2-minute incubation by illuminating the metasurface with laser and applying an a.c. field of 3 kHz. The induced ETP flow transported EVs towards the illuminated metasurface where the negatively-charged EVs were captured on the cysteamine-functionalized metasurface. Subsequently, the moisture (i.e. water) was removed leaving behind the captured EVs and the sample was imaged using a digital camera on an optical microscope. As depicted in Fig. 4(c), a clear color change was perceived under the microscope within a well-confined region centered on the laser-illuminated spot, suggesting the presence of EV clusters. The SEM image shown in Fig. 4(d) verifies the presence of EVs by showing nanoparticles with diverse sizes corresponding to the size range of EVs [42,43]. Supplement 1 section 3 includes the color image of a metasurface region located far away from the illuminated spot, where no color change occurred. After detection of EVs, the metasurface is reusable by removing the metal films (chromium and gold) using chromium and gold etchant. All the attached EVs on top of the metal films will be removed during the metal removal, and the Si metasurface can be reused for detection by depositing metal layers again. We repeated this deposit-removal cycles for three times, and the color images associated with EV detection after each of these cycles have been presented in Fig. S3.

3. Conclusion

To summarize, in this work, we have successfully demonstrated an innovative colorimetric EV detection approach based on silicon metasurface that yields structural coloration. This silicon metasurface exploits the response of multipolar optical modes to the changes in the refractive index of the surrounding medium to induce a vivid color change, allowing spectrometer-free and label-free sensing of extracellular vesicles. Furthermore, the combination of optofluidics and this color metasurface enables the detection of extracellular vesicles at femtomolar concentrations within a 2-minute incubation period.

As a proof-of-concept demonstration of the fusion of optofluidics and metasurface-based colorimetric EV detection, we have showcased the capability of optofluidics to overcome the diffusion limit and established a platform with simple optical configurations. Compared with our system, other reported detection techniques, which have also achieved femtomolar or even attomolar sensitivity [44–54], require longer process time than our platform to get a response. Notably, the aggregation of EVs by optofluidics driven by ETP flow in our configuration is highly efficient, and the EV assembly on the metasurface is easily perceivable to the human eye. Lowering the concentration of EVs is expected to result in fainter color shift, and at a certain concentration the signal is no longer legible to human eye. One imperative potential for future work to further push the sensitivity of this technology for even lower concentration is to use computer vision or machine learning [12,55] for color image post-processing. While the initial demonstration employed the capture of EVs based on their surface charge through the use of cysteamine, the use of immuno-specific antibodies would enable to capture and quantify disease-associated EVs by targeting disease-associated EV surface-bound protein markers and is a focus of future studies to build on this initial proof-of-concept. For instance, an array of color metasurfaces introduced in this study can be fabricated into a microfluidic chamber and each of the metasurfaces on the array can be functionalized with different types of immuno-specific antibodies. ETP flow can then be induced to all the metasurfaces for quick incubation.