All-dielectric hybrid VIS-NIR dual-function metaoptic

Metasurfaces are a promising technology that can serve as a compact alternative to conventional optics while providing multiple functions depending on the properties of the incident light, such as the wavelength, polarization, and incident angle. Here, we demonstrate a hybrid visible/near-infrared dielectric metaoptic capable of reflecting 940 nm light in a specified direction while allowing transmission of visible light (450–750 nm). This dual functionality is achieved by combining an aperiodic distributed Bragg reflector with dielectric meta-atoms. Experimental demonstration is also reported, showing an anomalous reflection of near-infrared light within a 20° full field-of-view and the transmission of wavelengths from 450 nm to 750 nm.


Introduction
Metasurfaces are planar optical elements composed of artificially fabricated nanostructures with tailored optical responses.Recently, they have attracted considerable interest owing to their potential control of the amplitude, phase, and polarization of light [1][2][3][4].This manipulation of light properties with metasurfaces has found diverse applications, including wavefront control [5,6], dispersion compensation [7,8], Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. and metalens imaging [9][10][11].Their flat form factor is particularly well-suited for the development of miniature optical and electronic devices [12,13] targeting precise functionalities.Consequently, metasurfaces have demonstrated potential as alternatives to traditional optical elements, like lenses [14,15], gratings [16,17], polarizers [18,19], and holograms [20,21].
Metasurfaces, when combined with dielectric Bragg reflectors (DBRs), can offer additional functionality and control over the manipulation of light [16,22,23].DBRs consist of alternating layers of material with different refractive indices.They are commonly used in photonics due to their high reflectivity and ability to reflect light at a specific wavelength while transmitting other wavelengths, making them useful in applications like laser mirrors [24] and optical filters [25].Here, we propose the combination of a metasurface and an aperiodic DBR to develop an all-dielectric hybrid visible and near-infrared (VIS-NIR) metaoptic with different control over the NIR and visible wavelengths.
Figure 1 shows the schematic diagram of our designed dualfunction metaoptic.It comprises periodic meta-atoms on top of a thin film substrate.The single DBR substrate can reflect NIR light and transmit visible light, but it lacks the ability to change the direction of the reflected NIR light.The single metasurface layer functions as a transmissive grating, designed to optimize the diffraction efficiency of a specific diffraction order.Combining the metasurface and DBR substrate, this metaoptic can anomalously reflect NIR light within a full field-of-view (FOV) of 20 • while allowing visible light to transmit directly through.The transmission of visible light is enabled by both the DBR's properties and the dielectric material used in the metasurface layer.The anomalous reflection [26] of NIR light is achieved by the diffraction enabled by the top meta-atoms, and the reflection angle depends on the period of these metaatoms.We co-optimized the thicknesses of the selected DBR layers and the geometry of the meta-atoms to simultaneously enhance the reflection efficiency of NIR light [16] and introduce more flexibility to optimize the metaoptic for dual functionality.The combination of this aperiodic thin film substrate and the meta-atoms offers a promising avenue for the development of novel optical devices with enhanced performance and functionality.

Simulations
All simulations use the rigorous coupled-wave analysis method in RSoft (Synopsys).The DBR substrate consists of silicon nitride (Si 3 N 4 ) and silicon dioxide (SiO 2 ) as the high and low index material.The meta-atom material is titanium dioxide (TiO 2 ) due to its high refractive index (n = 2.36 @ 940 nm) and low loss.We started with a periodic DBR with each layer's thickness fixed to a quarter wave thickness.The central wavelength of the reflectance peak was chosen to be 940 nm, a typical wavelength for NIR illuminators.The thickness of each Si 3 N 4 (n = 1.96 @ 940 nm) layer is 120 nm, and each SiO 2 (n = 1.46 @ 940 nm) layer is 160 nm.This configuration produced a reflectance spectrum with a flat reflectance around 940 nm and less than a 20% reflectance within the visible spectrum (450 nm-750 nm) (see appendix A).
To design the periodic meta-atoms on top of the DBR substrate, we used a 1330 nm by 400 nm unit cell.The pitch size along the y-axis was chosen to be less than 450 nm to avoid the diffraction of visible light along the direction that is orthogonal to the grating vector, which we set to be along the x direction.Without loss of generality, we chose the anomalous reflection angle/1st-order diffraction angle to be 45 • for normally incident NIR light.The pitch size along the grating vector direction was determined to be 1330 nm by the grating equation.
A parameter sweep of the radius and height for a single TiO 2 nanopillar placed on top of the DBR substrate was first carried out (see appendix B).From the sweep results, the reflection amplitude of the single pillar remains near 1 for different radii and heights.When the pillar has a height larger than 400 nm, its reflection phase varies in a range larger than 2π with different radii, which is necessary to achieve a linear grating phase ramp over the unit cell.Three nanopillars were used in one unit cell to build the gradient phase, considering the size required for each pillar to support the local phase variance.The starting point design of the metasurface layer was decided based on the parameter sweep results.Next, a general optimization was established using the multi-variable optimization and scanning tool in RSoft.The optimization variables included radii, heights, the locations of the three pillars in a unit cell, and the thicknesses of the first two layers of the basic DBR.As discussed in He et al [16], the thicknesses of the first several layers of the thin film substrate determine the reflection phase, which, combined with the phase variance caused by the meta-atoms, helps increase the anomalous reflection efficiency.We allowed the thicknesses of the first two layers of the DBR substrate to vary, aiming to increase the anomalous reflection efficiency of NIR light while maintaining good visible light transmission.To optimize the metaoptic towards this dual functionality, we used one TE polarized source with a wavelength ranging from 400 nm to 940 nm, illuminating from the meta-atoms' side at the normal incidence.For each simulation, we placed a detector to measure the 1st-order diffraction efficiency on the reflection/meta-atoms side and another to measure the 0th-order diffraction efficiency on the transmission/substrate side.When building the merit function, our goal was to maximize the 1st-order diffraction efficiency on the reflection side for 940 nm and the average 0th-order transmission for the incident wavelength ranging from 450 nm to 750 nm, using the same weights for each.
The optimized design parameters are listed in the caption of figure 2(a).We evaluated the angular bandwidth of this optimized design for diffracting NIR light by varying the incident angle from −10 • to 10 • .Referencing the result in figure 2(c), over the full FOV of 20 • , the average diffraction efficiency is approximately 40% for this optimized design.Interestingly, the maximum efficiency appears at 5 • angle of incidence instead of normal incidence.This might be a result of co-optimizing the NIR reflection and the VIS transmission.Figure 2(e) shows the transmission of the optimized device under an on-axis illumination, and it is about 50% on average for the full visible spectrum.The reason for the lower transmission of the metaoptic is the diffraction of visible light in transmitting through the metaoptic because of the large grating period (1330 nm) compared to the visible wavelengths (450 nm-750 nm).We also evaluated the transmission angular bandwidth for this optimized design, and the results are depicted in appendix C.
To demonstrate the effectiveness of modifying the thicknesses of the first two layers of the DBR substrate, we simulated the metaoptic with the optimized meta-atoms in the above design and the basic periodic DBR substrate, as illustrated in figure 2(b).Figure 2(d) displays a maximum efficiency of only 30% for this geometry, and it is significantly decreased compared to the device with the aperiodic DBR.The average diffraction efficiency within the angular bandwidth (full FOV of 20 • ) is around 20%.We also compared the transmission efficiency of these two designs (see figure 2(f)), showing no significant change in the 0th-order transmission efficiency.This matches well with our expectation that the additional control of the reflection phase achieved by the thickness variation of the top two DBR layers enables a higher reflection efficiency in the NIR wavelength.

Fabrication
Figure 3(a) illustrates the metaoptic fabrication process.In the first step, a Si 3 N 4 /SiO 2 DBR mirror was epitaxially grown on a fused silica wafer (University Wafer) using plasmon-enhanced chemical vapor deposition (PECVD).In the second step, the undiluted ZEP-520A (Zeon Chemicals), a positive-tone electron beam resist, was spin-coated onto the DBR at a speed of 1800 rpm for 90 s to achieve the desired resist thickness of 600 nm, which was set to be larger than the TiO 2 nanopillar's height for over-etching in a later step.The resist was then baked at 170 • C for 2 min.The DisCharge H2Ox2 (DisChem) was spin-coated on the resist at a speed of 2000 rpm for 60 s to avoid charging effects during exposure.The patterns were exposed using an accelerating voltage of 100 kV (JEOL 9500).After exposure, the samples were first washed with DI water for 30 s to remove the DisCharge H2Ox2.The resist was developed in ZED-N50 for 3 min, after which the sample was transferred to MIBK for 2 min to stop the developing process.It was then rinsed with IPA for 30 s and finally blown dry by N 2 gas.In the third step, a 200 nm thick TiO 2 film was coated on the sample using the atomic layer deposition (ALD, Cambridge Savannah 200) as it provided a TiO 2 film with the highest refractive index (n = 2.36 @ 940 nm) and lowest absorption of light.A standard two-pulse system of water and the TDMAT precursor was used for running cycles, with each cycle consisting of a 0.015 sec water pulse followed by a 25 s delay and a 0.7 s TDMAT pulse followed by the same 25 s delay.The deposition rate was approximately 0.06 nm per cycle.The chamber was maintained at 100 • C and the precursor was heated and maintained at 70 • C throughout the process.In the fourth step, reactive ion etching (RIE) was carried out using inductively coupled plasma (ICP) RIE (Oxford Instrument) with a mixture of CHF 3 and O 2 gas (52 and 2 standard cubic centimeters per minute, respectively) at an RF power of 15 W and ICP power of 2500 W. The etch rate was calibrated to be about 30 nm min −1 by etching a planar TiO 2 thin film deposited on a Si wafer.The estimated etching time was then calculated based on the thickness of the deposited TiO 2 layer from the previous step.After etching, the thickness of the nano-pillar was checked from the cross-section scanning electron microscope (SEM) image, and the etching time was adjusted to achieve the target thickness.Finally, the sample was cleaned under oxygen plasma to remove the degraded resist, and the remaining resist was removed using methylene chloride.Figure 3(b) shows the sample top view under a SEM.The inset is the zoomed-in image with a slanted view.Nanopillars with three different sizes are periodically arrayed in the picture.Figure 3(c) presents the crosssection of the device, which shows that the first two layers of the substrate have different thicknesses from the rest of the layers.

Testing and discussion
The diffraction efficiency of the fabricated metaoptic for near IR light was measured using the setup presented in figure 4(a).A Fianium WhiteLase micro supercontinuum (SC) laser was combined with a 940 nm, 10 nm full-width-half-maximum bandpass filter as a light source.A linear polarizer filtered the desired TE polarization for the metaoptic.The sample was mounted on a custom-built stage, which combines three translation stages along the x, y, and z directions and a rotation stage.The custom stage controls all degrees of freedom required to align the sample.Additionally, two other rotation stages were aligned center-to-center (circular plate in figure 4(a)).The custom-built stage with the sample was mounted on the arm of one of the two rotation stages to control the incident angle.The efficiency was measured for angles of incidence from −10 • to 10 • .On the other arm of the two rotation stages, a lens was mounted to collect the light from the desired diffraction order and to focus it on a Si photodiode.
To better match the plane wave illumination used in simulations, the metaoptic sample with dimensions of 1 mm by 1 mm was overfilled by a Gaussian profile illumination beam.A knife edge experiment measured the Gaussian beam profile, which was then used to calculate the ratio of the sample area compared with the beam size, which defines the fill factor.The amount of power striking the metaoptic could be determined by multiplying the total input power by the fill factor.The diffraction efficiency was then calculated using the intensity collected by the first Si photodiode divided by the input power on the sample.The SC laser passes through a band pass (BP) filter with a center wavelength of 940 nm and reflects from mirrors 1 and 2, then passes through a linear polarizer (LP).The 70/30 beam splitter (BSP) transmits 70% of the light for the metaoptic NIR reflection efficiency measurement and reflects 30% of the light for knife edge measurement.The 1st diffraction order on the reflection side of the sample is captured by the lens and detector on the collection arm; (b) measured 1st-order diffraction efficiency for the metaoptic sample as a function of the angle of incidence; (c) the experimental setup used to measure the transmission of the metaoptic sample: The SC laser reflects from mirrors 1 and 2, and then passes through the LP.Light transmits through a pinhole and illuminates the metaoptic.A collimator collects the transmitted light and sends it to the fiber spectrometer; (d) the measured 0th-order transmission spectrum of the fabricated metaoptic sample.
The measured 1st-order diffraction efficiency within a 20 • full FOV for the fabricated metaoptic is shown in figure 4(b).The observed efficiency increases from about 8% at a −10 • angle of incidence to about 40% at a 3 • angle of incidence.The efficiency remains roughly constant at 40% from 3 • to 6 • incident angle and drops slowly with larger incident angles.This measured efficiency peak position is consistent with the simulated result shown in figure 2(c).Still, the maximum efficiency value is found (as expected) to be slightly lower than the simulated one.The main reason for this difference is fabrication error.The diameters of the three kinds of pillars were measured to be within ± 5 nm of the designed ones, which leads to the efficiency drop.Furthermore, as seen in figure 3(c), the cross-section of the device shows that the three kinds of pillars have slightly different heights, and the tips of the pillars are not all flat, especially the narrowest pillars.These fabrication defects are also expected to decrease the diffraction efficiency.In appendix D, we present a tolerance analysis demonstrating that the efficiency curve in figure 4(b) can indeed be explained by the observed fabrication errors.Another possible discrepancy is that the laser is not perfectly Gaussian, so the power on the grating, which was calculated by assuming a Gaussian profile for the illuminating beam, may somewhat differ in reality.In conclusion, it is demonstrated that the metaoptic shows an average diffraction efficiency of 25% for the NIR light within a 20 • full FOV from measurement.
The transmission of the fabricated metaoptic was measured using the setup illustrated in figure 4(c).The same white light laser source was used.Since the light source spot is larger than the metaoptic area, a pinhole with a 0.5 mm diameter was placed in front of the fabricated device to let the laser beam underfill the sample (1 mm × 1 mm).This ensures that all the light collected by the collimator and the fiber illuminates the metaoptic area.The lens and photodiode on the collection arm were replaced with a collimator coupling the light into a fiber spectrometer (Ocean Optics HR2000CG-UV-NIR).We measured the spectrum of the beam collected by the fiber with and without the metaoptic sample plate in the setup.The transmission was calculated by dividing the measured spectrum with the sample by the collected spectrum without the sample.
Figure 4(d) shows the measured 0th-order transmission for wavelengths from 450 nm to 750 nm.The measured transmission is about 40% on average, and it shows a peak of 60% at a wavelength of ∼575 nm.The transmission drops to 20% at the two edges of the wavelength range.The measured transmission spectrum indicates that the fabricated metaoptic sample is semi-transparent.The measured curve is also consistent with the line shape in figure 2(e) but with a 10% overall transmission drop.The difference can be explained by the same reasons as the diffraction efficiency mismatch.An additional reason could be that the deposited materials are not purely transparent due to defects during fabrication, leading to increased absorption.
One of the promising applications of the demonstrated metaoptic is eye tracking in augmented reality (AR) devices.Currently, most AR devices apply two or more cameras mounted on the frame to avoid blocking the user's view and track the user's pupil motion [27].However, this approach adds hardware complexity and requires a large amount of computational processing power, time, and sophisticated imaging algorithms to assess the gazing direction from side-view images.With a dual-function device like the metaoptic placed in front of the user, not only is the see-through capability enabled, but also the scattered NIR light from the user's eye can be reflected into the direction of the camera mounted on the temple to capture a front view image of the pupil movement.The metaoptic achieves 28% NIR reflection efficiency at normal incidence and 40% transmission on average in the visible, making it well-suited for eye tracking.Nevertheless, to apply the metaoptic to eye tracking for AR devices, it is necessary to reduce the diffraction of visible light, as mentioned at the end of the simulation section.Song et al minimized these visible light diffractions by applying an absorptive material in the non-local metasurface [28] and achieved a transmission for the visible light of 85%.However, this approach is not suitable for eye tracking due to the limited full field of view in [28], which is only 0.5 • .The metaoptic wider full field of view of 20 • is better suited for this purpose.Additionally, the diffraction efficiency for redirecting NIR light in [28] is only 10% at normal incidence, whereas the metaoptic achieves 28%, a factor of 3 improvement.This improvement could be advantageous, especially considering the limited amount of NIR light that may be safely directed toward the eyes.Minimizing these visible light diffractions to increase the overall transmission while maintaining high NIR diffraction efficiency for eye tracking applications can be explored in future work.

Conclusion
In this paper, we proposed and reported a hybrid VIS-NIR metaoptic design that offers independent functionality across different wavelength bands.The dual-function design is based on the combination of an aperiodic DBR and dielectric metaatoms.We demonstrated by simulations, fabrication, and measurements that this metaoptic device reflects NIR light into the desired angle while preserving transparency for the VIS light.We expect this dual-function metaoptic to unlock new possibilities for optical imaging and wearable devices, particularly in applications such as eye tracking systems in AR displays.

Figure 1 .
Figure 1.Illustration of the proposed device.It is composed of a Si 3 N 4 /SiO 2 thin film substrate and a TiO 2 nanopillar array that conspire to reflect NIR light in a specific direction while also allowing for the transmission of visible light.

Figure 2 .
Figure 2. Design of a hybrid VIS-NIR dual-function metaoptic: (a) The unit cell for the final optimized meta-atoms layer combined with the aperiodic DBR substrate.The height of the pillars is 540 nm.The diameters for the three pillars are 130 nm, 290 nm, and 350 nm.The center-to-center distance between the smallest and medium pillars is 260 nm, and between the medium-sized and the largest pillars is 480 nm.The DBR has 21 layers; the optimized thickness of the first layer (Si 3 N 4 ) is 260 nm, and the optimized thickness of the second layer (SiO 2 ) is 220 nm.The rest of the periodic part of the DBR substrate has a thickness of 120 nm for Si 3 N 4 layers and 160 nm for SiO 2 layers; (b) the unit cell combining the meta-atoms layer of the optimized design in (a) and a basic periodic DBR substrate; (c), (d) simulated 1st order diffraction efficiency as a function of the angle of incidence with 940 nm TE-polarized light illumination for the metaoptic in (a), (b); (e), (f) simulated transmission spectrum of the metaoptic in (a), (b).The semi-transparent line represents the raw data from the simulation, and the solid red line is the smoothed line using the smoothdata function in MATLAB.

Figure 3 .
Figure 3. Metaoptic fabrication: (a) fabrication steps for the metaoptic device: the DBR film is first coated on the fused silica substrate using PECVD, and the inverse pattern is generated on the ZEP 520A resist using electron beam lithography.The TiO 2 thin film is coated on the resist pattern conformally using ALD.The ICP-RIE is used to etch the top layer of TiO 2 and expose the resist.Finally, the resist is removed using both O 2 plasma and solvent; (b) a top-down SEM image of the fabricated device; the inset is a zoomed-in image at a slanted view; (c) the cross-section image of the fabricated device.

Figure 4 .
Figure 4. Metaoptic measurements: (a) the experimental setup used to measure the diffraction efficiency: The SC laser passes through a band pass (BP) filter with a center wavelength of 940 nm and reflects from mirrors 1 and 2, then passes through a linear polarizer (LP).The 70/30 beam splitter (BSP) transmits 70% of the light for the metaoptic NIR reflection efficiency measurement and reflects 30% of the light for knife edge measurement.The 1st diffraction order on the reflection side of the sample is captured by the lens and detector on the collection arm; (b) measured 1st-order diffraction efficiency for the metaoptic sample as a function of the angle of incidence; (c) the experimental setup used to measure the transmission of the metaoptic sample: The SC laser reflects from mirrors 1 and 2, and then passes through the LP.Light transmits through a pinhole and illuminates the metaoptic.A collimator collects the transmitted light and sends it to the fiber spectrometer; (d) the measured 0th-order transmission spectrum of the fabricated metaoptic sample.

Figure D1 .
Figure D1.Fabrication artifacts simulation: (a) 3D representation of the geometry for simulating fabrication errors; the radius of the largest pillar is 185 nm, the radius of the middle pillars is 155 nm, and the radius of the smallest pillar is still 65 nm; (b) 2D image in the x-y plane of (a); (c) Comparison of the 1st order diffraction efficiency among the simulated result for the geometry in (a), the simulated result for the nominal design, and the measured result for the fabricated sample.