We perceive the color of an illuminated object in our surrounding primarily since it selectively reflectsa certain color and absorbs the others. Beside these mechanisms, colors can be generated resortingto another mechanism called structural coloration. In structural coloration, the interaction of incidentwhite light with nanoscale features in the illuminated object leads to specific resonances in the reflectionlight each associated to a specific color. In recent years, due to several advantages of structural colorsover its counterpart such as resistivity to high-temperature, photobleaching and ultraviolet irradiationin one side, and huge advancement in the fabrication technology in other side, a significant researchhas been done to generate artificial structural colors using sub-wavelength nanophotonics structures.To implement these structural colors, both plasmonic and all-dielectric metasurfaces (MSs) have beenused due to their capability of manipulating the light-mater interaction at nanoscales. However, theintrinsic dissipative ohmic loss in metals used in plasmonic MSs significantly reduces the quality ofgenerated colors by broadening and weakening the resonance peak of the reflection which leads to smallcolor gamut with low color saturation. To circumvent this issue, all-dielectric MSs using high-index ortransparent dielectric materials have been used as an alternative to plasmonic-based structural colors .Here, for the first time to our knowledge, we designed and fabricated an all-dielectric MS comprisingof hafnia (HfO2) nanopillars (NPs) to generate colors. These NPs can support both electric dipole (ED)and magnetic dipole (MD) modes due to the low loss of HfO2. The MD mode can be coupled to thereflected light resulting in a sharp Fano-type resonance in the reflection spectrum which is desired forgeneration of pure and vivid colors. In the design step, we first simulated a set of MSs with differentdesign parameters using finite difference time domain (FDTD) solver in Lumerical. Then, we fed thecorresponding reflection responses as the training data to a pseudo-encoder module [2, 3, 4], withina deep learning architecture, and found the periodicity of the unit cell of our MS as the most effectivedesign parameter on the spectral properties of the sharp Fano-type resonances. Based on these results, weagain simulated a set of MSs with optimized design parameters to generate colors with high saturation.Finally, the designed MSs were fabricated using atomic layer deposition which enables fabrication ofnano-scale features with minimum surface roughness as required in our MSs. The experimental resultsare in good agreement with those of simulation.
 Shen, Yichen, Veronika Rinnerbauer, Imbert Wang, Veronika Stelmakh, John D. Joannopoulos, and
Marin Soljacic. ”Structural colors from Fano resonances.” ACS Photonics 2, no. 1 (2015): 27-32.
 Kiarashinejad, Yashar, Sajjad Abdollahramezani, and Ali Adibi. ”Deep learning approach
based on dimensionality reduction for designing electromagnetic nanostructures.” arXiv preprint
 Kiarashinejad, Yashar, Sajjad Abdollahramezani, Mohammadreza Zandehshahvar, Omid Hem-
matyar, and Ali Adibi. ”Deep Learning Reveals Underlying Physics of Light-matter Interactions in
Nanophotonic Devices.” arXiv preprint arXiv:1905.06889 (2019).
 Abdollahramezani, Sajjad, Hossein Taghinejad, Tianren Fan, Yashar Kiarashinejad, Ali A. Eftekhar,
and Ali Adibi. ”Reconfigurable multifunctional metasurfaces employing hybrid phase-change plasmonic
architecture.” arXiv preprint arXiv:1809.08907 (2018).