By making structures smaller than the wavelength of light, it is possible to make surfaces where the optical behavior is not dictated by the choice of material and macroscopic shape, as it is for example with traditional lenses. These metasurfaces, enabled by increasingly advanced nanofabrication capabilities, have been used to tailor light behavior in a myriad of ways, ranging from achromatic flat lenses to advanced holograms and orbital angular momentum lasers [1,2]. Now, the race is on to make dynamic metasurfaces whose optical responses can be changed in a controllable way, preferably fast and with large efficiency .
Recently, we demonstrated one such platform by combining gap surface plasmon (GSP for short) metasurfaces with micro-electromechanical systems (MEMS) . Explained briefly: A GSP metasurface basically consists of metallic nanostructures separated a short distance away from a metallic substrate, illustrated in the figure below. The shape and size of these nanostructures determines the optical resonances of the system and can be used to make reflective metasurfaces that are broadband and have high efficiency . Meanwhile, MEMS are, as the name indicates, microscopic mechanical systems that can be controlled electrically. In our case we used a gold MEMS mirror as the metallic substrate, where piezoelectric membranes holding the mirror are used to change the distance between the mirror and nanostructures, thus letting us regulate the optical response in a controllable manner.
Illustration of the unit cell used in this work. A gold nanobrick is fabricated on a glass substrate and placed in close proximity to a gold substrate.
The key observation behind the paper can be seen in the figure below, which shows simulated reflection amplitudes for nanobricks with different dimensions, for three different nanostructure-substrate separation distances (and for light with wavelength 800 nm). Also shown with contour lines the difference in reflection phase shift between two orthogonal linear polarization states. For a specific nanobrick size the difference in phase varies as the separation is altered. Thus, the system can be used as a dynamic wave plate and be used to switch between different polarization states of light. For example, by choosing the separation such that the phase difference is 90 degrees, the device will function as a quarter wave plate and can be used to convert between linear and circular polarization states, while at 180 degrees the device will function as a half wave plate and convert an incident polarization state to the orthogonal one.
Simulated reflection amplitudes for different nanobrick dimensions for three mirror-nanobrick separation distances. Contour lines show the difference in the reflection phase for two orthogonal polarization states. The wavelength of the light is 800 nm.
Having observed this, the question was which geometry to choose. We wanted to have a dynamic wave plate with full 360-degree phase-difference coverage and with good and even reflection efficiency at all separations. From the figure above it is clear that some geometries, like 200 nm length and 50 nm width gives large phase difference at 20 nm separation but have poor reflectance at a separation of 300 nm. We discovered, however, that some specific geometries have large reflectivity for all separation distances while additionally being able to cover the full 360-degree range of phase differences. Furthermore, we noticed that these geometries would allow the device to function for even larger mirror-nanostructure separations, simplifying the fabrication process, with one drawback being reduced bandwidth. For these geometries, one polarization state is efficiently reflected at the nanobrick layer, while the orthogonal state passes through, is reflected off the mirror and goes back out through the nanobrick layer. Thus, one polarization state is unaffected by the separation distance, while the phase change of the orthogonal state can be controlled.
In the end, also taking into account fabrication considerations such as how close we could manage to place the nanostructures, we opted for gold nanobricks with 200 nm length and 100 nm width, with thickness 50 nm. The structures were fabricated at the Center for Nano Optics at the University of Southern Denmark, before being shipped to SINTEF in Oslo, Norway, where they were mounted together with the gold-coated MEMS mirrors and promptly returned to Denmark for optical characterization.
At first, the results were somewhat promising but not quite perfect. We were able to demonstrate the devices operating as mirrors, quarter wave plates and half wave plates, but not three-quarter wave plates as in simulations. Several possible causes were investigated, such as reflections from the glass substrate on which the nanostructures were fabricated, incorrect gold thickness and tilted mirrors. The explanation was found after noticing that the simulations looked more like the experimental results when simulating very long bricks, at which point we pulled apart one of the devices and examined it in a scanning electron microscope (which had been down for service earlier). Sure enough, the structures were longer than intended and even connected in several areas as can be seen in the image below.
SEM image of a sample which did not work. Dark area is a glass substrate, light areas are 50 nm-thickness nanostructured gold.
Consequently, we then proceeded to fabricate a new round of devices, making sure that the nanostructures matched the design. This time around, the measurements turned out exactly as expected from simulations – to such a large extent that group leader Sergey Bozhevolnyi told everyone to enjoy the results but not get too used to it. An example can be viewed in the video below, from the supplementary material of the paper, where simulated and measured polarization states are shown while the mirror-nanostructure separation is changed.
All the technical details as well as further references can be found in the paper, which is openly available at https://www.nature.com/articles/s41467-022-29798-0
Video (part of the article's supplemental material) illustrating how the resultant polarization state changes when the mirror-nanostructure gap is varied, showing close agreement between simulations (lines) and measurements (markers).
 Chen, W.T., Zhu, A.Y. & Capasso, F. Flat optics with dispersion-engineered metasurfaces. (2020). Nat Rev Mater 5, 604–620. https://doi.org/10.1038/s41578-020-0203-3
 Kamali, S., Arbabi, E., Arbabi, A. & Faraon, A. (2018). A review of dielectric optical metasurfaces for wavefront control. Nanophotonics, 7(6), 1041-1068. https://doi.org/10.1515/nanoph-2017-0129
 Shaltout, A. M., Shalaev, V. M., & Brongersma, M. L. (2019). Spatiotemporal light control with active metasurfaces. Science, 364(6441), eaat3100. https://doi.org/10.1126/science.aat3100
 Meng, C., Thrane, P. C., Ding, F., Gjessing, J., Thomaschewski, M., Wu, C., ... & Bozhevolnyi, S. I. (2021). Dynamic piezoelectric MEMS-based optical metasurfaces. Science Advances, 7(26), eabg5639. https://doi.org/10.1126/sciadv.abg5639
 Ding, F., Yang, Y., Deshpande, R. & Bozhevolnyi, S. (2018). A review of gap-surface plasmon metasurfaces: fundamentals and applications. Nanophotonics, 7(6), 1129-1156. https://doi.org/10.1515/nanoph-2017-0125