A circular dichroism (CD) spectrometer is an important tool in sensing chiral molecules possessing differential optical responses to circular polarization of light. These molecules are very common in biological and organic compounds, and, hence, the CD spectrometer is used in biological material studies [1], protein and DNA structural analysis [2], and stereochemical detection [3]. Typical CD spectrometers perform sequential measurements of left- (LCP) and right-circular (RCP) polarizations [4,5]. They are large in size and involve complex hardware to switch laser polarization and manage sequential data acquisition. Here, we seek an alternative chiroptical spectroscopy technique that performs spatial separation of LCP and RCP. This can dramatically reduce the size of CD spectrometers and provide real-time sensing.

Chiroptical spectroscopy built on the differential response between LCP and RCP requires chiral structures to obtain the L/RCP beams. With the advent of nanotechnology, multilayer metamaterial structures with strong chirality have been demonstrated [6–9]. Generally, the design and fabrication of such bulk chiral structures is complicated because they require layer-by-layer fabrication with each layer oriented and aligned with respect to the previous layer. The complexity can be drastically reduced by using metasurfaces [10]—metamaterials with reduced dimensionality [11].

Metasurfaces consist of a 2D array of plasmonic nanoantennas (NAs); each antenna is tailored to obtain a local change to the phase and/or polarization of the incident light. This approach has been successfully used in light bending [10], flat lenses [12], and wave plates [13]. In addition, metasurfaces have been used to obtain bi-anisotropic [14,15] and chiral effects [16–19].

In this Letter, we introduce a metasurface that exhibits the photonic spin Hall effect (PSHE) [20–23]. It reflects different spins (circular polarizations) in opposite directions, exhibiting mirror-symmetric reflection angle dispersion, as depicted in Fig. 1. Hence, it generates separate spectra for LCP and RCP. Thus, by using a single, deeply subwavelength scale and lightweight metasurface, we eliminate the need for a tunable light source that switches from LCP to RCP, a bulky natural chiral medium, and other more complicated detection schemes. This unique functionality can be obtained by using a broadband source, such as a xenon lamp or LED, which includes equal components of LCP and RCP, which are then spatially separated by the metasurface for straightforward detection. Not only does this metasurface enable a very compact device (130 nm thickness), but it also allows for real-time sensing due to simultaneous collection of the LCP and RCP data at all wavelengths. First, we explain the methodology of the metasurface design, followed by experimental results of the device implementation. Then, we describe a projection of the proposed design principles onto another metasurface device that performs optical rotation of incident light.

 

Fig. 1. Illustration of the metasurface used as a CD spectrometer using the photonic spin Hall effect. The spin components of the incident broadband source are reflected in opposite directions, and each wavelength component is reflected at a different angle. As a result, LCP and RCP spectra are obtained simultaneously. Colors are used for illustration and do not represent the wavelength values used in this work.

Download Full Size | PDF

To obtain a spin-dependent response, an array of anisotropic elements is used to achieve different phase gradients in response to LCP and RCP light, so that these circularly polarized components are reflected in different directions. This technique was introduced using polarization gratings [24,25] with thick layers to achieve a $\pi$ phase delay between the major and minor axes. Later, it was proposed to use anisotropic subwavelength aperture antennas [26], which are much thinner but have poor power efficiency (only a few percent). By using gap–plasmon [27–32] (GP) NAs instead, it is possible to obtain the same function with a compact subwavelength structure and dramatically improve power efficiency. With this design we are overcoming the poor power efficiency of single-layered plasmonic metasurfaces such as those made with v-antennas [33]. There have also been earlier demonstrations of effective GP structures in the microwave regime [34,35]. The unit cell of a GP structure is shown in Fig. 2(a). It has a bottom gold layer that works as a reflecting mirror, a top 30 nm thick gold NA, and a 50 nm thick dielectric (alumina, ${\mathrm{Al}}_2{\mathrm{O}}_3$) spacer. The metal/dielectric/metal sandwich enables the excitation of a compact GP wave [28]. The incident plane wave couples to this slow GP wave, which accumulates a large optical phase over the very short length of the NA. By tuning the aspect ratio ($L_x/L_y$) of the antenna’s geometry, we can indeed achieve a phase delay of $\pi$ between the reflection coefficients of the GP NA along its major and minor axes.

 

Fig. 2. (a) Schematic of a unit cell of a gap-plasmon-based antenna structure consisting of gold/alumina/gold structure. Silicon substrate carries the metasurface and plays no role in the operation. (b) Top view of the unit cell, with nanoantenna dimensions $L_x=280\mathrm{nm}$ and $L_y=230\mathrm{nm}$. (c) Top view of the unit cell, with the nanoantenna tilted at an angle $\alpha$. (d) Simulation results of circular copolarized reflection power (red) and cross-polarized reflected power (blue). (e) FE SEM image of the metasurface, with dashed rectangle to demonstrate one period of the structure.

Download Full Size | PDF

To present the operation of the entire device, we begin with analysis of a single NA. Let the complex coefficient of reflection for a single NA, as shown in Fig. 2(b) in the $x$ and $y$ axes be $r_x$ and $r_y$, respectively. Then, for the NA tilted at an angle $\alpha$ in Fig. 2(c), it is straightforward to obtain its reflection matrix in circular basis using the Jones calculus [36] as

So, for an RCP incident wave, the reflected wave $E_{\text{ref}}$ takes the form

Therefore, we have two reflection terms, a copolarized term with an abrupt phase term of $e^{-i2\alpha }$ ($e^{i2\alpha }$) for RCP (LCP) incidence, and a cross-polarized term with no phase gradient. We are interested in the first term only, where the phase shift can be controlled by tuning $\alpha$. Therefore, we tailor our NA dimensions to minimize $r_x+r_y$ and maximize $r_x-r_y$, which imposes the out-of-phase requirement for $r_x$ and $r_y$. The dimensions of the NA in Fig. 2(b) are chosen to optimize the performance of the structure in the near infrared (NIR) region. Figure 2(d) shows the simulation results of the copolarized reflected power component ${|r_x-r_y|}^2/4$ and the cross-polarized component ${|r_x+r_y|}^2/4$ using a finite element solver with the Johnson–Christy material values for gold [37]. Simulation reveals that we have a broadband NIR range where the majority of the reflected power (up to $\sim 50\%$) is in the copolarized term, which can be controlled by $\alpha$ according to Eqs. (2) and (3). In our structure, we use a periodic array of four antennas rotated at angles $\alpha =0°,45°,90°,135°$, as shown in Fig. 2(e), which is the field emission scanning electron microscope (FE SEM) image of the metasurface. The four NAs form a linear phase distribution from 0 to $2\pi$ ($-2\pi$) for a reflected LCP (RCP) beam across a period of $P=1.8\mathrm{\mu m}$. Therefore, by applying the generalized law of reflection [10], we obtain the angle of reflection from the metasurface as

The metasurface is fabricated on top of a silicon substrate, where the bottom 50 nm gold and the 50 nm alumina layers are implemented using electron beam deposition, and the 30 nm thick gold antennas are patterned using a standard electron beam lithography and lift-off process. The experimental setup used to test the metasurface is shown in Fig. 3(a). It consists of a tunable monochromatic source, a polarizer, and a retarder to obtain circularly polarized incident beams. Measurements are taken using a spectroscopic ellipsometer device, which allows rotation of the detector to collect the reflected ray as a function of reflection angle ${\theta }_r$. Figure 3(b) shows the measurement taken for both LCP and RCP for different values of wavelengths in the range $\lambda =1.2-1.7\mathrm{\mu m}$ plotted as a function of reflected angle ${\theta }_r$. The reflected power is 40% at $\lambda =1.5\mathrm{\mu m}$, and decreases gradually on both sides, as predicted by simulation in Fig. 2(d). Complete separation of LCP and RCP spectra is demonstrated. Lithographic patterning of gold nanostructures increases the Au effective losses due to electron scattering at finer polycrystalline grain boundaries and additional surface roughness features. Efficiency can be enhanced with encapsulated thermal annealing, which could significantly reduce the electron damping factors and at the same time preserve the shape of nanostructures [38].

 

Fig. 3. (a) Schematics of the experimental setup for testing the metasurface. A tunable monochromatic source, a polarizer, and a retarder are used to obtain circularly polarized incident beams for different wavelengths. Measurements are taken using a rotating arm device which allows rotation of detector to collect the reflected ray as a function of reflection angle ${\theta }_r$. (b) Experimental results of reflected power for LCP and RCP incident beams at different wavelengths as a function of reflected angle showing discrimination of LCP and RCP spectra.

Download Full Size | PDF

The structure of the CD spectrometer can be modified to obtain another device that rotates the polarization angle (PA) of linearly polarized light. Linearly polarized light with a PA $\phi$ as shown in the inset of Fig. 4 can be written in terms of its circular components as

 

Fig. 4. Inset: PA $\phi$ between the E-field and the horizontal ($x$) axis. (a) Schematics of one period of the metasurface consisting of two rows, where each row splits the incident beam by reflecting LCP (orange) and RCP (purple) into opposite sides. Alternating rows reflect opposite spins on the same side due to opposite gradient of antenna orientations. Displacement of alternating rows by a quarter period causes RCP phase delay with respect to LCP in both sides of reflections by $\pi /2$. (b) Schematics of the whole metasurface, which performs optical rotation to the reflected beams by 45° due to induced phase shift between different spin components. (c) FE SEM of the metasurface with dashed lines representing one period. (d) Reflected power from the metasurface for the left reflected beam as a function of wavelength and reflection angle ${\theta }_r$, showing that for each wavelength, maximum intensity occurs at $\sin {\theta }_r=-\lambda /P$.

Download Full Size | PDF

A similar technique was proposed in our previous work [39], but the power efficiency of the GP antennas is now 1 order of magnitude higher. Figure 4(c) presents an FE SEM image of the fabricated sample. Both reflected beams undergo the same PA rotation, so it is sufficient to present the results of the left beam only. It is also clear that the structure is twofold symmetric about the normal axis—i.e., the structure is the same if rotated upside down—confirming that the two beams should remain the same. The efficiency of the metasurface is 24% divided equally between two beams. Figure 4(d) shows the total power ($P_{\mathrm{T}}$) reflected from a vertically polarized beam as a function of wavelength $\lambda$ and reflection angle ${\theta }_r$. The result confirms the generalized Snell’s reflection angle [10] of $\sin {\theta }_r=-\lambda /P$, at which the maximum reflected intensity occurs for each wavelength. For each wavelength in the range $\lambda =1.2–1.7\mathrm{\mu m}$, we analyze the polarization state of the reflected beam at the angle of maximum intensity.

The metasurface is tested using the same experimental setup in Fig. 3(a). The polarizer is used to provide incident beams with PA values of ${\phi }_{\mathrm{i}}=0°,45°,90°,-45°$. Then the analyzer is used to obtain the reflected power filtered at the PA ${\phi }_0$ defined by Eq. (6) and at the orthogonal polarization state. Figures 5(a)–5(d) show that the power ratio of the data filtered at the required PA is almost 100%, and there is nearly no power at the orthogonal PA. The ratio between power at the required polarization state to the orthogonal state is around 20 dB at $\lambda =1.5\mathrm{\mu m}$, and the ratio deteriorates at the edge of the bandwidth due to a decrease in the output power.

 

Fig. 5. Polarization state results for incident PA values of ${\phi }_{\mathrm{i}}=0°,45°,90°,-45°$. For each case, the ratio of the power at both the required polarization state and its orthogonal state with respect to the total power are plotted. Almost all the power is in the polarization state, confirming the relation ${\phi }_0={\phi }_{\mathrm{i}}+45°$, and almost no power is at the orthogonal state. For each wavelength, polarization measurement was obtained at the angle of maximum reflected power of $\sin {\theta }_r=-\lambda /P$.

Download Full Size | PDF

In conclusion, a metasurface is designed and implemented to work as an extremely thin and small CD spectrometer in NIR. The photonic spin Hall effect is used for real-time spatial separation of LCP and RCP spectra of an unpolarized source, such as a lamp. It eliminates the need for switching the light source polarization and any necessary hardware required for this operation. With a GP NA, the power efficiency of the metasurface is up to 40%—an order of magnitude larger than similar structures—because it can be optimized to focus the reflected energy in the required mode of operation. Another metasurface based on the same GP structure is designed and implemented to rotate the PA of a linearly polarized light by 45°, over a broadband NIR region.