---
title: Meta-Optical Miniscopes
url: https://www.emergentmind.com/topics/meta-optical-miniscopes
type: topic
---

# Meta-Optical Miniscopes

Meta-optical miniscopes are miniature imaging systems that replace bulky refractive optics with metasurfaces or other engineered phase-only elements in order to obtain compact form factors, large field of view (FOV), long depth of focus (DOF), improved off-axis imaging, or encoded depth sensitivity. Across the literature, the term encompasses several related architectures: forward-viewing endoscopes using a single meta-lens for beam projection, coherent-fiber endoscopes with achromatized metasurface objectives, pupil-coded fluorescence miniscopes that recover 3D volumes from a single 2D measurement, extended-depth-of-focus camera modules that substitute computational refocusing for active focus control, and near-field platforms in which the resolving element itself is a meta-optical device held tens of nanometers from the sample [2212.11272] [2211.00808] [2010.05382] [2106.15807] [1608.04281].

## 1. System classes and optical architectures

A central architectural distinction is between image-forming meta-optics and coding meta-optics. In image-forming systems, the metasurface functions as a miniature objective or beam projector. The scanning fiber endoscope for short-wave infrared (SWIR) imaging uses a suspended single-mode fiber driven by a piezo tube to follow a spiral or elliptical scan, with a single meta-lens projecting the beam to a wide range of angles while diffuse backscattered light is collected by a return fiber to form an image point-by-point [2212.11272]. The meta-optical fiber endoscope (MOFIE) instead places an inverse-designed achromatized metasurface in front of a coherent fiber bundle so that the scene is directly imaged onto the bundle and relayed proximally to an RGB camera [2211.00808]. The “metascope” based on Miniscope v4 replaces only the objective lens with a planar metalens while retaining the excitation LED, dichroic mirror, tube lens, and electrowetting lens of the original optical train [2509.16373].

Coding architectures deliberately encode defocus or depth into the point spread function (PSF). Miniscope3D replaces the tube lens of a 2D Miniscope with an optimized multifocal phase mask located at the objective’s aperture stop, so that a single widefield exposure contains depth-dependent PSF structure and the 3D fluorescence volume is recovered by solving a sparsity-constrained inverse problem [2010.05382]. The visible EDOF meta-optic uses a rotationally symmetric, inverse-designed metasurface with a lens-like PSF that remains nearly invariant across wavelength and defocus, enabling broadband imaging after deconvolution with a single reconstruction kernel [2105.00160]. A related visible EDOF device demonstrates “varifocal functionality” without moving parts by using cubic wavefront coding to maintain PSF correlation over an 11 mm range, after which TV deconvolution restores detail [2106.15807].

A further architectural branch concerns tunability and near-field operation. MEMS-integrated Alvarez meta-optics realize focal tuning by laterally translating complementary cubic-phase metasurfaces with comb-drive actuators, yielding millimeter-scale focal-length tuning at low power [2111.07477]. In contrast, the “meta-instrument” is not a camera objective but a compact opto-mechatronic platform for nano-antenna arrays, superoscillatory lenses, hyperlenses, and solid immersion lenses that must be held tens of nanometers above the sample with sub-nanometer positional precision and a 100 kHz bandwidth [1608.04281].

## 2. Optical design principles and phase engineering

The most direct metasurface replacement of conventional optics uses the canonical focusing phase. In SWIR scanning fiber endoscopy, the initial phase is derived from a hyperboloid focusing phase designed to image a point at the back focal plane to infinity, with the thin-lens approximation written as $\Phi_0 \approx -\frac{\pi}{f \lambda}\rho^2 \approx -\frac{\pi}{d \lambda}\rho^2$ for fiber-to-lens spacing $d = 0.4 \,\mathrm{mm}$ and $\lambda = 1310 \,\mathrm{nm}$ [2212.11272]. To improve large-angle behavior, that work optimizes a centrosymmetric polynomial phase mask,
$$
\Phi(\rho) = \sum_{k=1}^{n} A_k \rho^{2k},
$$
with $n=3$, so that spot size is minimized across seven off-axis fiber positions and off-axis aberrations such as spherical aberration, astigmatism, and field curvature are mitigated [2212.11272].

Visible miniaturized systems employ several distinct phase strategies. The multifunctional Miniscope-v4-derived metascope reports four phase classes: a hyperbolic metalens,
$$
\varphi(r,\lambda) = -\frac{2\pi}{\lambda}\left(\sqrt{f^2+r^2}-f\right),
$$
a “square” metalens based on the thin-lens parabolic phase law,
$$
\varphi(r,\lambda) = -\frac{\pi r^2}{\lambda f},
$$
an inverse-designed EDOF metalens whose phase is optimized by maximizing the sum of logarithmic intensities along a target depth interval, and a double-helix (DH) metalens whose two-lobed PSF rotates approximately linearly with defocus [2509.16373]. The EDOF metasurface for broadband visible imaging uses a multiscale inverse-design objective that maximizes the minimum on-axis intensity along a prescribed axial interval, formulated as
$$
\max_{p_s}\min_{p\in P} I(p),
$$
with gradients computed through a locally periodic approximation and an adjoint formulation [2105.00160]. The fast visible EDOF meta-optic adds a cubic phase term to a lens phase,
$$
\phi(x,y)=\phi_{\mathrm{lens}}(x,y)+\alpha(x^3+y^3),
$$
which trades peak sharpness for PSF invariance across defocus [2106.15807].

Depth encoding in pupil-plane miniscopes follows a different logic. Miniscope3D uses a thin array of nonuniformly spaced convex microlenses with multiple focal lengths placed at the aperture stop, so that different microlenses come into best focus as object depth changes and the PSF’s spot pattern varies with axial position [2010.05382]. The full phase surface is parameterized as the point-wise maximum of convex spherical sags plus local Zernike tilt and astigmatism terms while excluding the power term to preserve focal length [2010.05382]. This suggests that “meta-optical” in the miniscope literature includes both nanophotonic metasurfaces and free-form pupil-coded phase masks, provided that the optical function is defined by sub-aperture phase engineering rather than by bulk refractive curvature.

## 3. Computational imaging, inverse problems, and learned correction

Many meta-optical miniscopes are designed jointly with reconstruction algorithms. The inverse-designed visible EDOF meta-optic models the captured image as $f = Kx + n$, or under a shift-invariant approximation $y = h * x + n$, and reconstructs images using Wiener deconvolution,
$$
\hat{X}(\omega)=\frac{H^*(\omega)}{|H(\omega)|^2+S_{nn}(\omega)/S_{xx}(\omega)}\,Y(\omega),
$$
as well as TV-regularized reconstruction [2105.00160]. Its central computational advantage is not merely EDOF, but a nearly invariant PSF across approximately 290 nm of the visible, which allows a single reconstruction filter for all channels [2105.00160].

Miniscope3D is an explicitly volumetric inverse problem. Its forward model is
$$
\mathbf{y}=\sum_z \mathbf{H}_z \mathbf{x}_z + \mathbf{n},
$$
where $\mathbf{H}_z$ is a depth-dependent, field-varying PSF operator. Because miniature objectives violate shift invariance, the system uses a low-rank weighted-convolution approximation,
$$
h[u,v;x,y,z] \approx A[u,v]\sum_{r=1}^{K} w_r(x,y,z)\,g_r[u-x,v-y;z],
$$
with $K \approx 10$–20 obtained by SVD of sparsely calibrated PSFs [2010.05382]. Reconstruction solves
$$
\min_{\mathbf{x}\ge 0}\; \tfrac{1}{2}\|\mathbf{H}\mathbf{x}-\mathbf{y}\|_2^2+\lambda\|\nabla \mathbf{x}\|_1
$$
using FISTA with 1–3k iterations [2010.05382].

A different form of computational compensation appears in metalens endoscopy. MetaScope identifies two optical degradations specific to a GaN metalens camera: intensity decay described by
$$
I_{\mathrm{total}}(\theta)=\eta_{\mathrm{meta}}(\theta)\cdot \cos^4(\theta)\cdot I_0
$$
and chromatic blur arising from wavelength-dependent PSFs in
$$
I_{\mathrm{output}}(x,y)=\sum_{\lambda}\left[T(\lambda)\cdot(\eta_{\mathrm{meta}}(\theta)\cdot \cos^4(\theta)\cdot I_0(\lambda)\ast \mathrm{PSF}(\lambda,x,y))\right]+\epsilon
$$
[2508.03596]. It responds with Optics-Informed Intensity Adjustment (OIA), which injects channel-efficiency and spatial-attenuation priors into an attention module, and Optics-Informed Chromatic Correction (OCC), which uses a Gaussian-mixture latent deformation model to aggregate dispersed energy through learned offsets [2508.03596]. The framework also introduces gradient-guided distillation from DINOv2 features of convex-lens ground truth [2508.03596].

Not all systems currently exploit computational inversion, but several explicitly anticipate it. The multifunctional metascope reports raw imaging only and proposes Wiener, Richardson–Lucy-like, or learned reconstruction as future work for its EDOF and DH modes [2509.16373]. The visible EDOF camera module similarly replaces active focusing with software-defined varifocality, demonstrating that EDOF meta-optics can shift optical complexity from mechanics to deconvolution [2106.15807].

## 4. Representative implementations and reported performance

Reported systems span SWIR endoscopy, visible fiber endoscopy, fluorescence microscopy, wide-FOV visible miniscopes, photoacoustic microscopy, and near-infrared camera modules. The table summarizes representative quantitative outcomes that recur in discussions of meta-optical miniscopes.

| System | Modality | Reported outcomes |
|---|---|---|
| SWIR scanning fiber endoscope [2212.11272] | 1310 nm forward-viewing endoscopy | FOV $\sim 70^\circ$, DOF $\sim 15\,\mathrm{mm}$, center resolution $\sim 140\,\mu\mathrm{m}$ at 15 mm, optical track $1.2 \rightarrow 0.86\,\mathrm{mm}$ |
| MOFIE [2211.00808] | Full-color coherent-fiber endoscopy | Rigid tip $3.75 \rightarrow 2.5\,\mathrm{mm}$, FoV $\sim 30^\circ$, DoF $>30\,\mathrm{mm}$, PSF FWHM $\sim 1.2\,\mu\mathrm{m}$, real-time video $\sim 14$ FPS |
| Miniscope3D [2010.05382] | Single-shot 3D fluorescence microscopy | Height $\sim 17\,\mathrm{mm}$, weight $\sim 2.5\,\mathrm{g}$, $2.76\,\mu\mathrm{m}$ lateral and $15\,\mu\mathrm{m}$ axial resolution, $900\times700\times390\,\mu\mathrm{m}^3$ at 40 volumes/s |
| Visible EDOF meta-optic [2105.00160] | Broadband visible imaging | 1 mm aperture, f/1, NA $\approx 0.44$, DOF $\sim 0.4\,\mathrm{mm}$, bandwidth $\sim 290\,\mathrm{nm}$, focusing efficiency $\approx 47\%$ |
| Multifunctional metascope [2509.16373] | Modified Miniscope v4 | Objective TTL $6.7 \rightarrow 2.5\,\mathrm{mm}$, WD $0.7 \rightarrow 2.0\,\mathrm{mm}$, EDOF $\sim 340\,\mu\mathrm{m}$, square-lens DOF $\sim 150\,\mu\mathrm{m}$ |
| Meta-optics PAM [2412.11733] | Photoacoustic microscopy excitation optics | Glass-free excitation path, 500 $\mu$m diameter metalenses, GM contrast maintained to $\pm 500\,\mu\mathrm{m}$ defocus |
| MAL NIR camera [2509.23599] | Wafer-level NIR camera | FOV $101.5^\circ$, TTL $3.39\,\mathrm{mm}$, F/1.64, volume $\approx 0.02\,\mathrm{cm}^3$ |

Several of these implementations illustrate different optimization targets rather than a single performance frontier. The SWIR scanning fiber endoscope achieves parity with a state-of-the-art refractive spherical doublet in full FOV and center resolution while reducing the rigid optical track by 28% and keeping edge-of-field resolution degradation below a factor of 2, compared with approximately 3× for the refractive design [2212.11272]. MOFIE emphasizes miniaturization of the distal rigid tip and full visible color fidelity; its object-plane resolution is approximately $50\,\mu\mathrm{m}$ because the coherent fiber bundle, not the metasurface optics, is the limiting sampler [2211.00808]. Miniscope3D prioritizes uniform volumetric resolution across a large axial range rather than raw 2D sharpness, and reports more than 2× better lateral and axial resolution throughout a 10× larger usable depth range than existing miniature single-shot volume-capture implementations [2010.05382].

The multifunctional metascope makes the trade-off between functionalities explicit. At focus, the square metascope resolves approximately $2.46\,\mu\mathrm{m}$ line width, corresponding to resolution $\sim 4.92\,\mu\mathrm{m}$ and 203.2 lp/mm, whereas the EDOF variant resolves approximately $4.38\,\mu\mathrm{m}$ line width at focus but maintains resolvable structure at $z=\pm 200\,\mu\mathrm{m}$ and extends the measured DOF to approximately $340\,\mu\mathrm{m}$ [2509.16373]. The DH mode provides passive depth sensing near the field center; in one calibration example, a measured rotation-angle difference produced an estimated depth difference of $\Delta z \approx 45\,\mu\mathrm{m}$, close to a $50\,\mu\mathrm{m}$ ground truth [2509.16373].

Photoacoustic microscopy with meta-optics broadens the concept beyond camera-like image formation. There the metalens sits in the excitation path only, yet it still functions as a miniscope-enabling element because it completely omits conventional glass optics in the excitation path and yields a planar, 500 $\mu$m diameter, 1.6 $\mu$m thick focusing component [2412.11733]. The grating metalens sacrifices lateral spot width for an axially elongated focus and maintains high lateral resolution and signal strength up to $\pm 500\,\mu\mathrm{m}$ from the focal plane [2412.11733].

## 5. Materials, fabrication, packaging, and manufacturability

The reported devices use a broad material palette, reflecting the wavelength range and fabrication priorities of each application. SWIR endoscopy uses crystalline silicon nanoposts on sapphire with $h=1000\,\mathrm{nm}$, lattice period $p=600\,\mathrm{nm}$, and 12 equally spaced phase levels defined by electron-beam lithography in ZEP-520A, followed by an $\sim 50\,\mathrm{nm}$ $\mathrm{Al_2O_3}$ hard mask and fluorine-based reactive-ion etching [2212.11272]. MOFIE and several visible miniscopes use silicon nitride on quartz or fused silica, with square posts or nanopillars chosen for polarization-insensitive operation and manufacturability [2211.00808] [2509.16373]. The broadband visible EDOF meta-optic uses 600 nm PECVD SiN on 500 $\mu$m fused silica with 400 nm period and minimum feature 100 nm, fabricated by JEOL JBX6300FS lithography and ICP etching [2105.00160].

Manufacturability is a recurring theme because miniaturization is only practically useful if packaging tolerances and process variability are manageable. The NIR meta-aspheric lens (MAL) combines a planar amorphous-silicon metalens on D263T Eco glass with a polymer aspheric refractive surface formed by laser direct writing and nanoimprint lithography, then bonds the two wafers with micrometer-level precision using active alignment and edge bonding [2509.23599]. Its reported wafer-level flow requires only one dicing step and no additional mechanical fixtures, and thousands of MAL units per 8-inch wafer are demonstrated [2509.23599]. This suggests that one trajectory for meta-optical miniscopes is not a purely flat-optic replacement, but a monolithically integrated hybrid element in which the metasurface supplies high-order and field-dependent correction while a molded or imprinted asphere supplies low-order power.

At the other end of the fabrication spectrum, photoacoustic microscopy with meta-optics deliberately uses a resist-only PMMA platform on BK7 glass, patterned as cuboid nanoholes with feature sizes 60–155 nm and unit-cell pitch approximately 350 nm [2412.11733]. The paper emphasizes the process simplification relative to $\mathrm{TiO_2}$, $\mathrm{Si_3N_4}$, or a-Si stacks [2412.11733]. Miniscope3D uses two-photon polymerization to print a free-form multifocal phase mask, then bonds it to the GRIN back surface with optical epoxy; slight tip, tilt, or misalignment is then absorbed by calibration rather than by sub-micron assembly tolerances [2010.05382].

High-throughput actuatable meta-optics have their own process constraints. The MEMS Alvarez meta-optic integrates 2-$\mu$m-thick PECVD silicon nitride nanopost metasurfaces with SOI comb-drive actuators, employing i-line stepper lithography rather than e-beam patterning for the metasurface itself and flip-chip bonding with an anisotropic conductive film spacer that produces an axial gap of approximately $10\,\mu\mathrm{m}$ and alignment errors of at most $3\,\mu\mathrm{m}$ translation and $0.03^\circ$ rotation [2111.07477]. Near-field meta-instruments similarly stress integrated mechanics and sensing: fiber interferometers, RF capacitive sensing, titanium springs, carbon rods, V-groove clamps, and MEMS piston stages are all part of the required infrastructure for operating meta-optical elements at 10–50 nm working gaps [1608.04281].

## 6. Trade-offs, misconceptions, and future directions

A recurrent misconception is that replacing refractive optics with metasurfaces automatically improves all performance metrics simultaneously. The literature does not support that interpretation. Instead, reported systems redistribute trade-offs. In the SWIR scanning fiber endoscope, miniaturization and off-axis improvement are obtained while center resolution remains comparable rather than superior to the refractive baseline [2212.11272]. In the multifunctional visible metascope, EDOF improves robustness across axial position but raw image sharpness and contrast decrease because the PSF acquires side lobes [2509.16373]. In visible EDOF imaging, the nearly invariant PSF supports a single deconvolution kernel across color, but the in-focus PSF is broader than that of a matched hyperboloid metalens and the diffraction-limited 1 mm metalens with matched geometric parameters would have FWHM $1.46\,\mu\mathrm{m}$ rather than the measured $24.9\,\mu\mathrm{m}$ [2105.00160].

Chromaticity remains one of the defining constraints. Several devices are explicitly single-wavelength or narrowband: the SWIR meta-lens endoscope is optimized for 1310 nm [2212.11272], the multifunctional metascope is centered at 530 nm [2509.16373], the photoacoustic microscope is strictly 532 nm [2412.11733], and the MAL NIR camera targets a 30 nm band near 940 nm [2509.23599]. Full-color operation has been demonstrated in MOFIE by inverse-designing the metasurface to maximize multichromatic MTF volume across the visible and in the visible EDOF optic by accepting a computational reconstruction step [2211.00808] [2105.00160]. A plausible implication is that future meta-optical miniscopes will continue to diverge into two lines: physically achromatized image-forming devices and narrowband or partially achromatized devices paired with computational correction.

Another misconception is that “flatness” alone solves system miniaturization. Some applications are limited by the optical element, but others are limited by supports, air gaps, fiber bundles, scan actuators, sensors, or acoustic transducers. MOFIE notes that further rigid-tip reduction depends on thinning the 500 $\mu$m carrier substrate to approximately 100 $\mu$m [2211.00808]. The SWIR SFE estimates that a thinner substrate and reduced fiber-to-lens spacing could reduce the optical track from 0.86 mm to approximately 0.4 mm, but the catheter still must satisfy rigid tip length $<10\,\mathrm{mm}$ and diameter $\lesssim 1\,\mathrm{mm}$ [2212.11272]. The near-field meta-instrument shows an even stronger version of this point: the optical element can only function if gap sensing, thermal drift control, and high-bandwidth MIMO positioning are also miniaturized [1608.04281].

Future directions in the cited work are notably consistent. System-level co-design recurs in visible EDOF imaging, multifunctional miniscopes, and metalens endoscopy: the remaining tube lens, dichroic, electrowetting lens, fiber NA, fiber-to-lens spacing, and reconstruction algorithm should all be optimized jointly with the metasurface phase [2105.00160] [2509.16373] [2508.03596]. Tunability is another major direction, exemplified by MEMS Alvarez meta-optics with focal tuning from 5.8 mm to 2.7 mm, corresponding to approximately 200 diopters, at voltages below 40 V and estimated dynamic power below $1\,\mu\mathrm{W}$ [2111.07477]. Space compression is a more radical path: multi-color spaceplates replace free-space propagation with a much thinner multilayer structure and report achromatic compression ratios up to 4.6 at three visible channels, which could reduce tube length without altering magnification if integrated downstream of an objective [2312.02378].

Taken together, these results define meta-optical miniscopes less as a single device category than as a design paradigm: optical power, aberration correction, depth encoding, focus invariance, beam steering, and even free-space compression are reassigned from bulk lens assemblies to planar or pupil-engineered structures, often with computational reconstruction and, in some cases, MEMS actuation. The reported devices already cover intravascular SWIR imaging, full-color fiber endoscopy, single-shot 3D fluorescence microscopy, visible wide-FOV and EDOF head-mounted imaging, narrowband near-infrared camera modules, metalens endoscopy with learned correction, and photoacoustic microscopy with glass-free excitation optics [2212.11272] [2211.00808] [2010.05382] [2509.16373] [2509.23599] [2508.03596] [2412.11733].

Source: https://www.emergentmind.com/topics/meta-optical-miniscopes