Voltage-Switched Metalens Gives Photoacoustic Microscope Two Focal Depths

A team at POSTECH has built a photoacoustic microscope that changes its focal depth by changing a drive voltage. The lens never moves. According to the team's paper in Light: Science & Applications, a silicon nitride nanostructured metalens switches between two diffraction-limited focal modes at a visible wavelength, with switching voltages of 0.87 V and 1.03 V and a 250 ms response time. The two foci together give 1.2 mm of axial coverage while keeping lateral resolution at 3.7 μm. The work was led by Junsuk Rho and Chulhong Kim.

The result is a modest but concrete step. It does not make a continuously tunable lens. It makes a two-state one, and that is enough to ease a basic trade-off in optical-resolution photoacoustic imaging.

01The depth-of-focus problem

Photoacoustic microscopy illuminates tissue with pulsed laser light. Absorbers such as hemoglobin heat up and emit ultrasound, and a transducer records it. In the optical-resolution variant, the light is focused tightly to get micrometre-scale lateral resolution. Tighter focusing shortens the depth of focus, so only a thin axial slab is sharp at a time.

The usual workaround is a Bessel or other non-diffracting beam, which keeps a narrow core over a longer axial range. The cost is side lobes. Chulhong Kim's group describes them as reducing signal-to-noise ratio and spreading laser energy over a larger volume, which lowers signal intensity. The POSTECH approach keeps a conventional single focus and moves it electronically between two planes. Each image is formed with a compact focal spot and no side-lobe pedestal.

02How the metalens encodes two focal lengths

The lens is a flat array of silicon nitride (SiNx) nanostructures. It is designed for 532 nm light, where hemoglobin absorbs strongly, and it focuses right- and left-circularly polarized light to two different focal lengths, F₁ and F₂. The helicity of the incoming light selects the focus.

The material was tuned for this job. From a set of deposition conditions, the team chose a SiH₄/N₂ ratio of 1.52, giving a refractive index of 2.4 and an extinction coefficient of 0.0027 at 532 nm. That means high index contrast for strong phase control and very little absorption at the green wavelength. The lens is also optimized for use in water, because photoacoustic imaging needs the optics to work underwater to limit signal loss. The optimum geometry the authors report is a nanostructure height of 1000 nm and a period of 340 nm.

Two design details stand out in the paper:

• The selected meta-atom behaves as a half-wave plate, which is what makes the response depend on circular polarization handedness.
• Because the library of meta-atoms is limited, the propagation phase is discretized into eight levels. Coarse phase quantization is a known source of efficiency loss and stray light. The paper does not quantify that loss in the material I reviewed.

The authors designed two phase maps corresponding to focal lengths of 4.5 mm and 5.2 mm. These are for the lens in air. In water the foci lengthen, and the measured foci sit at 6.05 mm and 6.96 mm, the stretch coming from the change of surrounding medium. The two foci are separated by 0.9 mm axially. The 1.2 mm total coverage is larger than that separation because each focus has its own usable depth range.

03The voltage side: a liquid-crystal retarder

Nothing in the metalens itself responds to voltage. The switching comes from a separate liquid-crystal (LC) cell upstream, which sets the polarization handedness. The paper explains that the LC molecules realign in response to an applied AC voltage and act as a phase retarder, with a retardation set by the cell thickness and the birefringence profile. According to Phys.org's account, the cell uses the same type of liquid crystal found in smartphone displays. That is a mature, cheap, low-voltage technology. The sub-2 V drive levels fit that picture.

Changing the retardation flips the beam between right- and left-circular polarization, and the metalens focuses each to a different depth. I could not confirm from the sources reviewed whether any additional waveplate sits in the chain, so I leave that detail out.

The same group has used this pattern before: an LC cell paired with a metalens, as in their earlier work on electrically switching between bright-field and edge-enhanced imaging on a millisecond scale. The new work moves it into a biomedical modality.

04Packaging and the eye demonstration

Photoacoustic systems need optical and acoustic axes to coincide. Here the team mounted the metalens on a ring-type ultrasound transducer, which avoids physical interference and keeps the optical and acoustic waves coaxial. Because the lens is a thin flat element, it fits in the aperture of the ring, so no extra alignment stage is needed.

For validation, the group imaged eyes with chemically induced corneal injury. The paper's abstract says rat eyes, monitoring neovascularization across the entire eye depth range. The Phys.org release summary says mice, and I could not reconcile the two. Per Phys.org, the shallow mode showed new corneal vessels, and the deep mode selectively resolved vessels near the iris, about 2.2 mm below the surface. The authors also report that vessel diameters and lesion areas measured in the two modes agree closely, which suggests minimal polarization-dependent artifacts. That is a check worth making, since polarization-multiplexed lenses can differ between channels.

Resolution differs slightly by mode. Phys.org gives 3.1–3.7 μm lateral resolution, and the paper's headline figure is the 3.7 μm worst case.

05What it does and does not change

The advantages are mostly about mechanics. Without a z-stage, voice coil or tunable-lens barrel, there are fewer parts to drift, wear or vibrate. A 250 ms switch is also fast compared with moving a scan head. Rho says removing mechanical motion could enable more compact and stable imaging systems for biomedical and industrial uses. That is a reasonable expectation, but the paper shows it only for one benchtop configuration.

The limits are also clear:

• Two states, not a continuum. Anything between or beyond the two foci is out of reach, and the 1.2 mm range is fixed by the metalens design. Adding more foci would need more polarization channels or spatial multiplexing, each of which shares the aperture or the light.
• Sequential acquisition. Each focal mode needs its own scan, so the full-depth image takes at least two passes.
• Wavelength lock. The meta-atoms are tuned to 532 nm. Metalenses are strongly dispersive, so multi-wavelength functional imaging, such as oxygen saturation, would need extra design work. The sources do not say whether that has been tested.
• Eye scale. The tested depth range is about 2 mm in a small rodent eye. Human ophthalmic imaging involves longer working distances and motion, and no human data are reported. The authors themselves note that the current implementation can be improved, though the passage I saw does not list the specific changes.

Next steps are likely to be more focal states, higher efficiency from finer phase quantization, and tests beyond the anterior segment. Diabetic retinopathy is the motivating target, but the work stops at an animal corneal-burn model. Whether a two-depth, voltage-switched lens can compete with other depth-extension methods in clinical use will depend on signal-to-noise ratio, scan time and ease of alignment in a handheld probe. The paper does not yet show those comparisons.