Traditional refractive lenses are often bulky, so modern optics increasingly develop and use diffractive optical elements (DOE) to achieve miniaturization, lightweight, and cost-effectiveness. However, over the past two decades, a new technology called metasurfaces has emerged, and the optical components produced based on this technology are known as metalenses. Metalenses have significantly expanded and extended the functionality of traditional DOEs. In addition to providing better beam shaping, they have the potential to replace refractive imaging lenses and even integrate polarization control and other features. In this article, we compare the differences among metalenses, traditional DOEs, and refractive lenses, and provide insights into the advantages of metalenses.
Traditional refractive lenses rely on different curvatures across the lens to bend light using the lens’s refractive index and curvature, following Snell’s law. Metalenses and traditional DOEs are both based on diffractive optics (Huygens’ principle) and have “flat” features on a macroscopic scale. They no longer rely on lens curvature to control light propagation. Instead, they belong to the category of “flat” diffractive lenses.
Metalenses, due to their consistently flat nano-structures, are superior to traditional DOEs under similar functions. Furthermore, metalenses offer additional potential capabilities that traditional DOEs cannot achieve, such as controlling light’s polarization, intensity, and dispersion.
Traditional refractive lenses depend on Snell’s law (as shown in Figure 1), which requires different thicknesses across the lens to create the required surface curvature, thus bending the light. As a result, optical systems tend to be bulky, as shown in Figure 2.
Metalenses and traditional DOEs are based on diffractive optics, utilizing Huygens’ principle to achieve more complex beam manipulation. According to Huygens’ principle, light propagation is determined by the wavefront surface, and the direction of light propagation is perpendicular to the wavefront surface. The subsequent wavefront can be seen as the envelope of secondary wave sources on the preceding wavefront.
For instance, as shown in Figure 3 (left), if a beam of horizontal collimated light travels along the x-axis, its wavefront is parallel to the y-axis. We can assume that the wavefront phase is a constant C1 when reaching the y-axis. At this point, the secondary wave sources along the y-axis generate wavefronts represented by green arcs, and their envelopes form a new wavefront (shown as the blue dashed line) after passing through the y-axis. This wavefront remains parallel to the y-axis, meaning the light continues along the x-axis.
Now, as shown in Figure 3 (right), placing a traditional DOE or metalens along the y-axis modifies the phase of the secondary wave sources (commonly known as phase shift), changing it from C1 to C2(y) = C1 + delta(y). The green arcs show the new phase envelope, resulting in a new arc-shaped wavefront (red dashed line), with the corresponding propagation direction now being focused light (red solid line).
Above, we explained how to focus collimated light beams using the principle of diffractive optics to reshape light propagation. In practice, beam shaping can take many forms, such as reshaping VCSEL beams into dot, line, or surface patterns, which traditional refractive lenses cannot achieve.
Although traditional DOEs and metalenses are both designed based on diffractive optics, they have different methods of implementing phase modulation, leading to differences in their performance.
Traditional DOEs are macroscopically flat but have uneven microstructures, typically with serrated or multistep shapes. Traditional DOEs appear flat at the macroscopic level, but at the microscopic scale, they have uneven surface protrusions, often in the form of triangular serrations or multistep structures.
Serrated DOEs are typically used for gratings and imaging applications, while multistep DOEs are commonly used for some gratings, imaging, and beam shaping (e.g., shaping into dot arrays), making them the most prevalent type of traditional DOE in the market. These multistep DOEs have different height steps, known as two-step, four-step, or eight-step DOEs. Light passing through N different height steps achieves discrete phase shifts of 2π/N, 4π/N, …, 2π, resulting in N different discrete phase shifts.
Traditional DOEs with different height steps have the following characteristics:
The efficiency of multistep DOEs increases with the number of steps. This is because the ideal wavefront is always smooth and continuous, but real multistep DOEs can only produce N distinct discrete phase shifts. Thus, after passing through a multistep DOE (as shown by the green solid line in Figure 5), the phase can only approximate the smooth ideal wavefront (as indicated by the red dashed line in Figure 5). The more height levels there are, the closer the approximation to the ideal smooth phase, leading to higher efficiency.
The greater the number of steps in a multistep DOE, the more difficult and costly it is to manufacture. This is because whether using nanoimprint or photolithography to fabricate DOEs, accurately producing steps of different heights at specific positions becomes increasingly challenging as the number of steps increases, thereby increasing manufacturing errors and costs. Consequently, traditional DOEs on the market rarely exceed eight steps, and manufacturing errors range from 30 nm to several micrometers.
Traditional DOEs suffer from shadowing effects, resulting in narrow angular bandwidth. Shadowing effects occur not only in multistep DOEs but also in serrated DOEs. Figure 6 illustrates a serrated traditional DOE used for a blazed grating function. Due to the shadowing effect, its theoretical diffraction efficiency limit decreases as the grating emergent angle θ increases [8].
Additionally, a study by Zeiss [1] pointed out that the efficiency of serrated DOEs decreases rapidly with increasing incident angle due to the shadowing effect. This limitation is a direct consequence of modifying phase by changing height. Therefore, traditional DOEs are not suitable for scenarios involving large-angle light incidence or emission. For instance, if used as a beam shaping lens for large fields of illumination (FOI), their efficiency is low. Similarly, when traditional DOEs are used as imaging lenses with large numerical apertures (NA), the shadowing effect causes narrow angular bandwidth, which restricts the amount of light passing through the optical system at oblique incidence, affecting the uniformity of image brightness and thus the imaging quality (e.g., relative illumination) [1].
As shown below, metalenses have truly “flat” surfaces at the nanoscale, with consistent heights. Unlike traditional DOEs, which achieve phase modulation through varying microstructure heights, metalenses achieve phase modulation through different nano-structures (e.g., varying the rotation angle or geometric dimensions of nano pillars). Two types of nano pillar arrangements are used to achieve phase modulation: PB phase [2] and transmission phase [3].
The rotation angles or geometric dimensions of the nano pillars must be properly designed to discretely sample the required 2π phase shift, similar to different height levels in DOEs. The red curves below represent the relationship between PB phase, transmission phase, and rotation angles or nano pillar diameters, while the green curves represent the transmittance of the nanostructures. Regardless of the phase modulation method, the micro- and nano-structures of metalenses have uniform heights, which eliminates the shadowing effect found in traditional DOEs.
The features of metalenses include:
The nano pillars are spaced at approximately half the wavelength, and the smaller the spacing, the higher the efficiency of the device. Similar to multistep DOEs, the more densely packed the nano pillars are, the closer the approximation to the ideal smooth wavefront, resulting in higher efficiency. As shown below, when the period of the nano pillar arrangement is reduced, the sampling becomes denser, which better approximates the ideal smooth wavefront and thus improves efficiency.
However, when selecting the nano pillar arrangement period, manufacturing constraints must be considered. To achieve enough different rotation angles or sizes to discretely cover the 2π phase shift, the arrangement period is often close to half the wavelength.
Unlike traditional DOEs, metalenses have a much larger angular bandwidth. As discussed in the DOE section, traditional DOEs suffer from a shadowing effect that leads to a narrow angular bandwidth. The figure below provides a visual comparison of the deflection efficiency of refractive lenses, serrated DOEs, and metalenses at different wavelengths, angles of incident (AOI), and target angle of emergence (α) [1]. The black boundary represents the 60% efficiency boundary, and its vertical width for a given wavelength represents the angular bandwidth.
Taking the refractive lens as the benchmark, traditional DOEs have comparable angular bandwidth only when the angle of emergence is less than 8°, while metalenses still maintain similar angular bandwidth to refractive lenses at 53°. This guarantees high efficiency for metalenses in large incident angle scenarios (e.g., large FOV imaging) or large emergent angle scenarios (e.g., large FOI shaping, large NA lenses).
For example, the imaging results of traditional DOEs and metalenses used as imaging lenses are shown below. The metalens provides uniform imaging, whereas the traditional DOE shows a significant decrease in intensity at the edge of the image, resulting in vignetting.
Metalenses often exhibit stronger dispersion, resulting in narrow spectral performance. The nano pillars designed for discrete sampling of 2π at a specific wavelength may have significant resonance absorption or phase changes for another wavelength, causing a sharp drop in efficiency. For instance, as shown in Figure 9, the metalens has a narrower spectral range than refractive lenses and DOEs, with efficiency dropping sharply for wavelengths below 405 nm or above 660 nm. Even within the 405-660 nm range, the strong dispersion of the metalens results in different effects at different wavelengths, such as varying emergent angles for the same incident angle. A typical example is the noticeable difference in focal length for different wavelengths in a standard focusing metalens.
In applications requiring achromatic performance, multiple metalenses or additional refractive lenses are often needed to compensate for dispersion, or a customized design with more complex nano pillars is used to achieve achromatic performance [6].
In optical applications, multiple components are sometimes needed to achieve various functions simultaneously, resulting in bulky systems and increased costs. Metalenses, unlike refractive lenses and traditional DOEs, can integrate multiple optical functions into a single device, such as filtering or polarization control, helping reduce system size further.
The following is an example of how metalenses integrate polarization control: If the nano pillars are cylindrical, their cross-section is circular, and they have rotational symmetry of 90 degrees, meaning they are not sensitive to the polarization of incident light. However, if elliptical or rectangular nano pillar cross-sections are used, the same nano pillar can be designed to apply different phase shifts to different polarized incident light, providing additional polarization selectivity alongside beam shaping or imaging. This is a capability not possessed by refractive lenses or traditional DOEs.
The diagram below shows the concept of a metalens focusing horizontal and vertical polarized light onto different photodetectors, along with a schematic of the nano pillar arrangement for such a polarization-selective metalens [7].
At this point, we have compared the differences between traditional DOEs and metalenses in terms of altering light propagation to achieve beam shaping or imaging. Below is a summary table that combines their differences in mass production processes and other aspects.
Table 1: Comparison Between Traditional DOEs and Metalenses
| Feature | Traditional DOE | Metalens |
|---|---|---|
| Mass Production Process | Nanoimprint, grayscale lithography | CMOS mass production process |
| Material Cost | Low: PMMA, PC, PET Medium: Glass | Medium |
| Manufacturing Precision | Errors between 30 nm to several microns | ~10 nm error |
| Product Consistency | Nanoimprint: Medium Lithography: High | High |
| Mass Production Yield Rate | Medium | High |
| Compatibility with Reflow and Wave Soldering | No: PMMA, PC, PET Yes: Glass | Yes |
| Operating Temperature Range | Narrow: PMMA, PC, PET Wide: Glass | Wide |
| Operating Humidity Range | Narrow: PMMA, PC, PET Wide: Glass | Wide |
| Angular Bandwidth | Narrow | Wide |
| Efficiency for Large FOI Shaping (High Incidence Angle, e.g., >10°) | Low | High |
| Efficiency for Large FOV Imaging (High Incidence Angle, e.g., >10°) | Low, significant vignetting | High, no significant vignetting |
| Spectral Bandwidth | Very wide | ~250 nm (e.g., in visible light) |
| Multi-functional Integration | No | Yes |
This metalens transforms VCSEL or EEL sources into wide field-of-illumination line patterns, such as those used in robotic obstacle avoidance. It achieves a uniform and fine line beam using a single lens, ensuring high manufacturing precision and consistency. Traditional approaches use refractive lenses, like collimating the beam to 2 mm in diameter and then passing it through a Powell lens. However, alignment or manufacturing errors can lead to inconsistent beam profiles, which the metalens approach effectively avoids.
This metalens shapes VCSEL sources arranged in an array into tens to thousands of dot arrays. It is primarily used for facial recognition, SLAM in robotics, and other applications requiring high signal-to-noise ratios. Compared to a multi-step DOE solution, which may or may not use a collimator lens, the metalens design achieves 5-10% higher efficiency and over 50% better contrast.
There are two main types of homogenizing products: multi-step DOE homogenizers and engineered diffusers. The engineered diffuser is not based on diffraction, unlike the DOE, and has a surface structure of micro-lens bumps that scatter light by refraction. When comparing metalenses to DOE-type homogenizers, the metalens shows higher efficiency and better suppression of 0th order diffraction. Compared to engineered diffusers, the metalens provides more uniform intensity distribution with sharper boundaries, although engineered diffusers may have slightly higher overall efficiency.
Alpha Cen has developed a near-infrared metalens for Time-of-Flight (ToF) imaging that allows a significant reduction in module height—from 12.6 mm to 5.6 mm—compared to traditional lens groups. Below is a comparison of ToF imaging module with the metalens versus refractive lens solutions.
The quality of metalenses is highly dependent on precise manufacturing. For example, the near-infrared metalens is manufactured on glass or SiO2 substrates with Si nano pillars. The height, diameter, and refractive index of the nano pillars must conform to design specifications, as deviations can lead to stray light and reduced efficiency. Controlling the nano pillar diameter with high accuracy is the most challenging part of the production process, contributing to the largest discrepancies between actual test values and design values.
For instance, in a 940 nm homogenizing metalens, significant deviations in nano pillar diameter from the design can result in bright 0th order diffraction spots, which are unacceptable for most beam shaping applications. Only by maintaining tight control over nano pillar diameter can the device eliminate these stray light and meet client specifications.
The graph below illustrates the relationship between 0th order energy percentage and nano pillar diameter error in a 940 nm metalens, assuming no other manufacturing errors. As shown, a 15 nm diameter deviation can result in a 3% 0th order energy proportion, which is unacceptable for most beam shaping applications.
Alpha Cen’s line shaping, dot projector, and homogenizing metalenses keep 0th order energy below 2% due to rigorous production iterations and tight control over manufacturing errors, ensuring consistent performance that meets design expectations.
Metalenses represent a significant advancement in optical technology, offering numerous advantages over traditional DOEs and refractive lenses. They provide better efficiency, integration capabilities, and superior performance in a variety of applications, from imaging to beam shaping. Alpha Cen’s metalens technology is already demonstrating success in various practical applications, achieving both miniaturization and functional enhancements in optical systems.
[1] Decker, M. et al. Imaging performance of polarization-insensitive metalenses. ACS Photonics 6, 1493–1499(2019).
[2] Pan, M., Fu, Y., Zheng, M. et al. Dielectric metalens for miniaturized imaging systems: progress and challenges. Light Sci Appl 11, 195 (2022).
[3] Khorasaninejad, M. et al. Metalenses at visible wavelengths: diffraction-limited focusing and subwavelength resolution imaging. Science352, 1190–1194 (2016).
[4] Chen, W.T., Zhu, A.Y. & Capasso, F. Flat optics with dispersion-engineered metasurfaces. Nat Rev Mater 5, 604–620 (2020).
[5] M. Khorasaninejad et al. Metalenses at visible wavelengths: Diffraction-limited focusing and subwavelength resolution imaging.Science 352, 1190-1194 (2016).
[6] Chen, W.T., Zhu, A.Y., Sanjeev, V. et al. A broadband achromatic metalens for focusing and imaging in the visible. Nature Nanotech 13, 220–226 (2018).
[7] Arbabi, E. et al. Full-stokes imaging polarimetry using dielectric metasurfaces. ACS Photonics 5, 3132–3140 (2018).
[8] Hessler, T., Rossi, M., Kunz, R. E., & Gale, M. T. (1998). Analysis and optimization of fabrication of continuous-relief diffractive optical elements. Applied Optics, 37(19), 4069.
[9] S. Banerji et al., arXiv: 1907.06251 (2019)
Metalens Design & Fabrication Services: From Concept To Mass Production
