Due to the strong absorption by the ozone layer, ultraviolet (UV) components of solar irradiation with wavelengths shorter than 280 nm are almost entirely blocked from reaching the Earth’s surface. This spectral range, known as the solar-blind UV region, is virtually free from sunlight interference. As a result, the solar-blind band offers a unique and valuable spectral window that can be leveraged as an important indicator for detecting natural hazards and human activities, such as fire occurrence, ozone depletion, dynamite explosions, missile launches, and electric leakage from high-voltage transmission lines [1,2]. Because of the low background noise, solar-blind imaging is crucial for early warning or detection of hazards, featuring both low false-alarm rates and high sensitivity. Moreover, solar-blind imaging boasts a wide range of bio-applications, including the detection of nucleic acid concentration and the analysis of proteins and DNA secondary structures [3–6]. On the other hand, owing to the short wavelength of solar-blind light, solar-blind imaging can achieve high-contrast and high-resolution imaging [7]. The conventional solar-blind imaging system is typically constructed by integrating photodetectors with a series of optical components, such as lenses, filters, and polarizers [8]. Over the past few decades, semiconductor based photodetectors have evolved significantly from the first generation of silicon (Si) photodetectors to wide bandgap semiconductor based solar-blind photodetectors, such as gallium oxide and diamond [9,10]. In contrast to the rapid advancements in high-performance photodetectors, current solar-blind imaging systems remain heavily dependent on cascading refractive optical components with curved surfaces, resulting in bulky and heavy setups that impede system miniaturization and integrations with on-chip systems [11].

During the past decade, a new type of planar optics, consisting of a two-dimensional (2D) array of artificially engineered nano-antennas, known as metasurfaces, has emerged as a powerful alternative to conventional optics [12–14]. This technology offers numerous advantages, including compactness, light weight, ultra-high spatial resolution in subwavelength scales [15–18], etc. However, in comparison to the extensive research on metasurfaces in the visible and infrared regions, the development of metasurfaces within the solar-blind regions lags behind due to limited material selection that must meet strict criteria, i.e., high transparency and a large index, as well as manufacturability of nanostructures with high-aspect-ratios, etc. For instance, commonly used high-index dielectrics for creating metasurfaces in visible and infrared ranges, such as gallium nitride (GaN) [19,20], titanium dioxide (${\mathrm{TiO}}_2$) [21,22], silicon nitride (${\mathrm{Si}}_3{\mathrm{N}}_4$) [23], and Si [24,25], suffer from significant absorption loss in the solar-blind region due to inter-band transitions. On the other hand, most of the commonly used solar-blind transparent materials (e.g., ${\mathrm{MgF}}_2$, ${\mathrm{CaF}}_2$, and ${\mathrm{SiO}}_2$) [26] have very small refractive indices, which would dramatically increase the complexity and difficulty of fabricating solar-blind metasurfaces due to the stringent requirements for high aspect ratios [27]. To address these limitations, several novel material systems have been recently exploited to develop metasurfaces in the solar-blind region with promising performance [28–31]. For instance, solar-blind metasurfaces including focusing metalenses and metasurface-holograms with working wavelengths down to 266 nm were realized using nanostructured hafnium oxide (${\mathrm{HfO}}_2$) films on quartz substrates [29]. A resin embedded with zirconium dioxide (${\mathrm{ZrO}}_2$) nanoparticles was developed as a printable material with both high refractive index and low extinction coefficient, which demonstrates considerable potential for applications in metasurface holography, spanning from the near-UV to solar-blind range [30]. Moreover, metasurfaces have also been employed for deep UV (DUV) light harvesting, generation, and biosensing [31–33]. Lately, aluminum nitride (AlN) metalenses were demonstrated for DUV imaging and laser microfabrication applications [34].

Diamond is widely recognized for its exceptional properties, including the highest thermal conductivity, a high laser-induced damage threshold, extreme hardness, and remarkable chemical inertness. Notably, its large bandgap ($\sim 5.4\mathrm{eV}$) and high refractive index ($\sim 2.5$) make it an ideal material for solar-blind applications [35]. In particular, the recent progress in diamond metasurfaces presents a timely opportunity to develop highly efficient metasurfaces for use in solar-blind imaging components. Despite this potential, there has been no report on the exploration of diamond based metasurfaces for solar-blind imaging applications to date.

In this study, diamond metalenses enabled solar-blind imaging was demonstrated. As an example, two representative functionalities, i.e., bright-field imaging and spiral phase contrast imaging, were showcased. Our findings prove the feasibility of achieving solar-blind imaging in a simple and compact form using a single layer of sub-micron thin diamond metalenses. Furthermore, this approach can leverage the exceptional properties of diamond to enable a practical and capable solution for developing miniaturized, lightweight, and robust solar-blind imaging systems, which would be suitable for operation even in harsh environments.

Bright-field imaging and phase contrast imaging are two examples of the most commonly used techniques in modern imaging systems. Bright-field imaging enables to directly display the morphology of the objects based on their amplitude profiles, while phase contrast imaging allows for the visualization of transparent objects by transforming their phase contrast into intensity variation. Specifically, to realize bright-field imaging, a diamond metalens with a hyperbolic phase profile $[{\phi }_{\mathrm{hyperbolic}}(x,y,f)$; see the left panel of Fig. 1] was designed according to Eq. (1), which can convert the incident plane wave into a spherical wavefront [36]: $${\phi }_{\mathrm{hyperbolic}}(x,y,f)=\frac{2\pi }{\lambda }(f-\sqrt{x^2+y^2+f^2}),$$(1)where $\lambda$ is the wavelength of incident light, $f$ is the focus length, and ($x$, $y$) denotes the coordinates of a given point at the metasurface plane. Furthermore, spiral phase contrast imaging can be conveniently realized by further converting the diamond metalens into a vortex metalens through a simple superimposition of a spiral phase with a topological charge of $l$ on the hyperbolic phase [37,38]: $${\phi }_{\mathrm{vortex}}(x,y,f)=\frac{2\pi }{\lambda }(f-\sqrt{x^2+y^2+f^2})+l\times \arctan \left(\frac{y}{x}\right).$$(2)

By doing this, the vortex metalens can perform the functionalities of imaging and radial Hilbert transform simultaneously. The underlying principle of spiral phase contrast imaging can be mathematically formulated as follows [39,40]. Taking the case of $l=1$ as an example, a first order vortex metalens was designed to create a focused spiral wavefront along the light propagation direction (see the right panel of Fig. 1). The output image field ($U_o$) can be described as a geometrically amplified convolution of the diffraction-introduced point spread function (PSF) $h$ with the input target field $U_i$ [41–43]: $$U_o(x,y)=h(x,y)\otimes \frac{1}{|M|}{U}_i(\frac{x}{M},\frac{y}{M}),$$(3)where $M$ represents the magnification of the system.

Given the adopted spiral phase filter with $l=1$, the PSF $h(x,y)$ is obtained through the Fourier transform of the pupil function $\exp (i\theta )·\mathrm{circ}(r/R)$, and can be expressed as follows [44]: $$h(\rho ,\varphi )=\frac{\pi R}{2\rho }[J_0\left(\frac{kR\rho }{z_2}\right)H_1\left(\frac{kR\rho }{z_2}\right)-J_1\left(\frac{kR\rho }{z_2}\right)H_0\left(\frac{kR\rho }{z_2}\right)]\exp (i\varphi ),$$(4)where ($r$, $\theta$) and ($\rho$, $\varphi$) denote the polar coordinates of the planar metalens and imaging plane, respectively. $R$ is the radius of the metalens, $J_0(J_1)$ is the zeroth (first) order Bessel function of the first kind, $H_0(H_1)$ is the zeroth (first) order Struve function, and $z_2$ is the distance from the metalens to the image plane.

By taking the input field distribution $U_i=|U_i|\exp$($i\psi$) into Eqs. (3) and (4), the output image can be obtained as $$U_o\propto \exp (i\psi ){\mathit{g}}_{\mathrm{amp}}\exp (i{\delta }_{\mathrm{amp}})+i{U}_i{\mathit{g}}_{\mathrm{ph}}\exp (i{\delta }_{\mathrm{ph}}),$$(5)where ${\mathit{g}}_{\mathrm{amp}}$ and ${\mathit{g}}_{\mathrm{ph}}$ represent the amplitude and phase gradients, and ${\delta }_{\mathrm{amp}}$ and ${\delta }_{\mathrm{ph}}$ are the polar angles of the corresponding gradients. Thus, the output field is proportional to the gradients ${\mathit{g}}_{\mathrm{amp}}$ and ${\mathit{g}}_{\mathrm{ph}}$. Therefore, the output field $U_o$ reflects the 2D field differentiation of the input image. In this way, the spiral phase contrast imaging can be employed to highlight regions where the amplitude or phase distribution of the target objects will be disrupted, providing a straightforward and efficient method to extract the edge profiles of an input image. Apparently, such an edge enhancement technique facilitated by spiral phase contrast imaging significantly reduces the data processing required for filtering key information and preserving the core geometric features, making it particularly advantageous in the fields of image processing, computation, and machine vision. However, conventional spiral phase contrast imaging relies on the use of bulky and complex two-lens $4f$ systems in combination with a spatial light modulator (SLM) to generate a spiral phase profile on the Fourier plane, resulting in heavy and complicated systems with limited resolution.

To overcome the aforementioned limitations, a solar-blind metalens and a vortex metalens were designed and fabricated using diamonds to replace the conventional imaging systems for bright-field imaging and edge-enhanced imaging, respectively. For this purpose, isotropic diamond nano-resonators with circular cross sections were chosen as the meta-atom building block [Fig. 2(a)], which enables the creation of meta-optics with a polarization independent response due to their centrosymmetric structures. A single-crystal diamond substrate grown by chemical vapor deposition (CVD) was employed due to its exceptional optical quality, which exhibits a wide transmission window from the infrared down to the solar-blind region with an absorption edge located around 226 nm, indicating their suitability for broadband applications extending to the solar-blind region [Fig. 2(b)]. Moreover, the obtained CVD diamond has a high refractive index larger than 2.4 and a small extinction coefficient across a broad spectral range, making it highly suitable for the development of high-index metasurfaces [Fig. 2(c)]. To design the diamond metalens and diamond vortex metalens, the meta-atom library is developed by conducting numerical analysis on the complex transmission coefficient of the diamond nanopillars with a fixed height of 450 nm as a function of their diameters. In this way, phase modulations of light at the wavelength of 261 nm can be achieved in a complete $2\pi$ range by simply changing the radius of the diamond nanopillar while maintaining a relatively high transmission efficiency [Fig. 2(d)]. Furthermore, to simplify the design of the diamond metasurface, the desired phase profiles of the designed metalenses were evenly divided into five discrete phase levels. Accordingly, five diamond nanopillars with radii of 45 nm, 49 nm, 59 nm, 70 nm, and 83 nm were selected to modulate the incident solar-blind light, achieving a complete $2\pi$ phase coverage.

Prior to the nanofabrication of metalenses, diamond substrates were mechanically polished on both sides to achieve a small root mean square (RMS) roughness of approximately 2.09 nm, i.e., less than $\lambda /100$, where $\lambda$ is the operation wavelength of the metasurface [Fig. 2(e)]. The fabrication process of the diamond metalens is detailed in Fig. 3(a). Firstly, the double-side polished single-crystal diamond underwent a 3 h immersion in aqua regia solution ($\mathrm{HCl}\text{:}{\mathrm{HNO}}_3=3:1$) to eliminate any remaining metal on its surface; then, ultrasonic baths in acetone, alcohol, and deionized water were applied to remove any organic residues from the diamond substrate surface; afterwards, a 50 nm nickel (Ni) layer was utilized as a hard mask. This layer was deposited onto the diamond surface by electron beam evaporation and patterned using an electron beam lithography system through a lift-off process. Following this, the metalens pattern was transferred into the diamond substrate by reactive ion etching (RIE) with oxygen plasma at an RF power of 200 W. Finally, the Ni mask was removed through chemical etching using a 1:2 solution of $\mathrm{HCl}\text{:}{\mathrm{HNO}}_3$. Figures 3(b) and 3(c) present the optical microscope images and scanning electron microscope (SEM) images of the fabricated diamond imaging metalens and vortex metalens, respectively. The well-defined circular cross-sections and varied dimensions of the obtained diamond nanopillars demonstrate the feasibility of nanofabrication techniques developed in our work. However, we note that the actual diameters of diamond nanopillars tend to be larger than the designed values, which can be attributed to fabrication imperfections.

The focusing performance of the fabricated diamond metalens and vortex metalens was experimentally evaluated using a home-built characterization setup, as shown in Fig. 4(a). The experimental setup included a continuous 261 nm laser serving as the excitation source, along with a UV camera (BGS-USB3-LT665) combined with a UV objective ($40\times$ magnification, LMU-40-UVB) and a tube lens, which were mounted on a motorized linear translation stage with a movement precision of 1 μm (EM-LSS90-150C1). This configuration was used to capture the beam profiles of the incident laser as a function of its propagation distance along $z$-axis after it passed through the diamond metalens, where the metalens plane was defined as $z=0\mathrm{\mu m}$. The left panels of Figs. 4(b) and 4(c) summarize the intensity distributions of the excitation laser along the $xz$ plane, where the focal lengths of the diamond metalens and vortex metalens were determined to be approximately 510 μm, in close agreement with the designed value (500 μm). The beam profiles near the focal plane of the focused laser generated by the diamond metalens and vortex metalens were enlarged and are presented in the right panels of Figs. 4(b) and 4(c), respectively. The diamond metalens demonstrates efficient focusing of the incident laser into a solid bright spot. Meanwhile, the diamond vortex metalens produces a doughnut-shaped intensity distribution featuring a dark annular center, demonstrating the existence of a phase singularity in the focused light with a non-zero topological charge number due to its spiral phase profile. The focusing efficiencies of the fabricated diamond metalenses were determined by calculating the ratio of the integrated light intensity within the focal spot (where the intensity falls to 90% of its maximum value) at the focal plane to the integrated light intensity at the metalens plane. In this way, the focusing efficiencies of the metalens and the spiral metalens were measured to be approximately 19.4%, and 18.5%, respectively, indicating their suitability for solar-blind imaging applications. Moreover, it is worth noting that the efficiency of diamond metalenses can be significantly enhanced by increasing the number of phase levels and improving fabrication precision to minimize phase errors.

To experimentally evaluate the imaging performance of the fabricated diamond metalenses, target samples containing various pre-defined patterns were placed in front of the metalenses at a distance of 1100 μm (more than twice their focal length). Subsequently, the resulting images generated by the metalenses under illumination from a 261 nm laser were then relayed onto a UV camera using a combination of an objective and a tube lens for recording, as depicted in Fig. 5(a). Specifically, target samples were fabricated by patterning a 140 nm thick Au film deposited on a sapphire substrate into the letters of “Diamond” and “Metalens” as well as two distinct patterns, as shown in Fig. 5(b). The obtained diamond metalenses, measuring $200\mathrm{\mu m}\times 200\mathrm{\mu m}$ in size, exhibit a theoretical diffraction-limited resolution of approximately 678 nm, corresponding to their numerical aperture (NA) of $\sim 0.19$. As shown in Fig. 5(c), the diamond metalens enables the acquisition of high-quality bright field images in the solar-blind region, while the diamond vortex metalens allows for efficient edge enhancement when imaging the same object [Fig. 5(d)]. Furthermore, leveraging the outstanding material properties of diamond, diamond metalenses are expected to be well-suited for imaging in challenging conditions, such as high laser power, corrosive environments, and abrasive wear.

In summary, we have experimentally demonstrated solar-blind UV imaging using diamond metalenses, taking advantage of their ultra-compact design and exceptional light manipulation capabilities. Specifically, two representative functionalities, namely, bright-field imaging and spiral phase contrast imaging, were successfully showcased. Our approach paves a new way for achieving solar-blind imaging in a simple and compact manner with a single layer of sub-micron thin diamond metasurfaces. Benefiting from the exceptional material properties of diamond, we anticipate that diamond metalenses based solar-blind imaging systems will be highly suitable for applications in harsh and demanding environments.