Super-resolution Imaging of Limited-size Objects

Improvement of label-free far-field resolution of optical imaging is possible with prior knowledge of the object such as its sparsity or accumulated by a posteriori examination of a similar class of object1-4. We show that the sole knowledge of the object’s limited size is another fundamental resource to achieve resolution beyond the Abbe-Rayleigh diffraction limit: a higher resolution can be achieved with smaller objects. To prove this, we developed an imaging method that involves the representation of light scattered from the object with orthonormal field-of-view-limited Slepian-Pollak functions and experimentally demonstrated λ/8 resolution of sub-wavelength objects. Our method requires no assumption of the shape and complexity of the object and its labelling allowing a wide range of applications in the studies of nanoparticles and isolated microorganisms.

💡 Research Summary

The paper presents a label‑free far‑field imaging technique that surpasses the classical Abbe‑Rayleigh diffraction limit by exploiting only the prior knowledge that the object’s size is limited. The authors build on the mathematical properties of Slepian‑Pollak (prolate spheroidal wave) functions, which form an orthogonal basis for any field confined to a finite spatial region (the field‑of‑view, D) and band‑limited to a finite spectral region (K). By expanding the complex scattered field of a sub‑wavelength object in this basis, each coefficient c_i can be retrieved from the measured projection d_i divided by a transfer factor γ_i that characterises how the i‑th Slepian‑Pollak mode propagates through the imaging system. While γ_i decays rapidly with mode order, making high‑order coefficients vulnerable to noise, the authors quantify the fundamental limits using quantum‑information theory. They derive a quantum Fisher information bound and a corresponding Cramér‑Rao inequality that relate the achievable resolution Δ, the field‑of‑view size D, the number of reliably estimated coefficients M, and the total photon budget P_tot via Δ≈D/M and 2 M γ_i≈P_tot. This analysis shows that reducing D (i.e., knowing that the object is small) dramatically lowers the photon number required to resolve finer details, turning object‑size limitation into a resource for super‑resolution.

To realize this in practice, the authors implement a shot‑noise‑limited spatial N‑mode tomography scheme. A reconfigurable mask placed in the Fourier plane of a 4f system selectively mixes a chosen Slepian‑Pollak mode with a reference mode, converting their sum into a plane wave that is detected by a single‑pixel photon counter. By repeating the measurement with different mask configurations, they construct an N × N diagonal repetition matrix R that controls the number of shots per mode. The system’s transfer matrix T (including all optical distortions and crosstalk) is calibrated using a known point or line source. Measured complex amplitudes are then corrected by applying a vector linear least‑mean‑square filter, essentially the inverse of T, to retrieve the true coefficients.

Experimentally, the method is applied to platinum nanoparticles fabricated on sapphire using focused‑electron‑beam‑induced deposition. The particles are smaller than 0.8 λ (λ = 638 nm). For each particle, 13 Slepian‑Pollak coefficients (2‑D case) or 6 coefficients (1‑D case) are measured from four illumination directions. The reconstructed images display features that conventional imaging with a numerical aperture of 0.9 (the objective used) cannot resolve at all. Spectral analysis shows an effective NA ranging from 2.43 to 3.6 (average ≈ 3.16), far exceeding the physical NA of the microscope. Point‑spread functions measured after reconstruction have full‑width‑half‑maximum values of roughly λ/6 (2‑D) and λ/8 (1‑D). A Siemens‑star test pattern yields a resolution of about λ/7 for 2‑D imaging and λ/8 for 1‑D, confirmed by Rayleigh‑type criteria, PSF width, and spectral extrapolation.

The authors also demonstrate that even channels with extremely low energy transfer ratios (|γ_i/γ_1|² < 10⁻⁵, crosstalk dominant) can be recovered with acceptable accuracy, indicating robust information transfer through the Slepian‑Pollak channel. Limitations arise from the need for many repeated measurements to approach the shot‑noise limit and from background scattering caused by surface roughness outside the field of view. Nonetheless, the work overturns the long‑standing belief that high‑order Slepian‑Pollak coefficients are experimentally inaccessible, proving that a simple prior—knowledge of maximal object size—suffices to achieve deep super‑resolution without labeling or invasive probes.

In conclusion, the study provides both a rigorous information‑theoretic framework and a practical experimental platform for label‑free far‑field super‑resolution imaging. By converting object‑size limitation into a quantitative resource, it opens pathways for high‑resolution studies of nanoparticles, isolated microorganisms, and other specimens where labeling is undesirable or impossible, potentially impacting fields ranging from biomedical microscopy to nanometrology.