RASTMIN: Single-Molecule Localization by Raster Scanning a Minimum of Light

Single-molecule localization microscopy (SMLM) achieves nanometer-scale precision, with structured illumination single-molecule (SML-SSI) methods such as MINFLUX reaching ~1–2 nm precision with only a few hundred detected photons. However, MINFLUX and related SML-SSI approaches are technically complex compared to camera-based SMLM, limiting their widespread adoption. This work introduces RASTMIN (single-molecule localization by RASTer scanning a MINimum of light), an SML-SSI method that matches MINFLUX localization precision but can be implemented on standard scanning fluorescence microscopes (e.g., confocal or multiphoton) with minor modifications. The research objective is to demonstrate the principle, theoretical limits, and experimental performance of RASTMIN, and to show its applicability to super-resolution imaging by leveraging fluorophore blinking.

MINFLUX has demonstrated exceptional photon efficiency and precision (~1–2 nm) in model systems (e.g., DNA origami), extended to 3D in fixed and live cells, and combined with fluorescence lifetime detection. Parallelized SML-SSI variants increase throughput at the cost of resolution. Compared to camera-based SMLM, SML-SSI methods (especially MINFLUX) are more complex technically. Theoretical considerations emphasize Poisson-limited detection; real cameras often exhibit super-Poissonian noise, degrading precision relative to ideal detector assumptions. Prior work articulated the dependence of localization precision on the excitation pattern geometry, photon budget, signal-to-background ratio (SBR), and sampling strategy.

Principle: RASTMIN illuminates the sample with a spatially modulated excitation field I(r) featuring a central intensity minimum (ideally zero). Two images are used: (1) a high-SNR, high-resolution reference image of the excitation minimum (obtained by averaging raster scans over a stable emitter), and (2) a low-SNR, low-resolution raster scan over the sub-diffraction region containing a single emitter, yielding photon counts n = [n1, n2, …, nK] at pixel positions r = [r1, r2, …, rK]. The emitter position is inferred by comparing the single-molecule image to the reference (e.g., via maximum-likelihood estimation). Theoretical performance: The Cramér–Rao lower bound (CRB) for localization uncertainty σCRB was computed for square rasters of side L and a doughnut-shaped I(r) with FWHM = 300 nm. For L = 100 nm, N = 1000 detected photons, K = 6×6 pixels, and SBR = 4, σCRB(x,y) maps show ~1 nm uncertainty at the center, increasing outward, remaining <5 nm over an area larger for RASTMIN than for MINFLUX under equivalent conditions. Average σCRB over the central region scales with photon number and L, with σCRB ∝ L; improvements are ultimately limited by achievable SBR. Increasing K (number of exposures/pixels) increases information content under Poisson noise, improving precision substantially up to ~K ≈ 36, with marginal gains beyond; practical limits arise from scanner positioning precision. Temporal considerations: Localization photon efficiency depends on beam geometry and exposure positions, not measurement time. Temporal resolution is governed by scanning speed over distance, not the number of exposures K. For imaging, total acquisition is dominated by fluorophore ON/OFF kinetics, not scan speed. Implementation: RASTMIN was realized on a home-built confocal microscope with an optical scanner and single-photon counting detector (APD). Two modifications were added: (1) a vortex phase plate with suitable polarization optics to generate a toroidal (doughnut) focus; (2) active drift correction for lateral and axial stabilization. Typical stabilization precisions: lateral 0.8–1.3 nm (σDC), axial 1.5–2.0 nm. Experimental procedure: Single ATTO647N fluorophores on DNA origami were pre-located by low-power scanning (~5 µW at the objective back focal plane). RASTMIN frames were acquired by raster scanning a sub-diffraction region over the target molecule with parameters typically K = 6×6, L = 100 nm, frame time ~130 ms (SBR ≈ 4). For each molecule, 100 frames were collected; localization precision was estimated as the standard deviation over 100 localizations. Photon counts per frame were varied by changing excitation power (e.g., 3, 12, 24 µW). Super-resolution imaging with blinking: DNA origami with six Alexa Fluor 647 sites arranged in a 3×2 grid (15 and 20 nm spacings) were measured. Frames (L = 100 nm, K = 6×6) were recorded at 20 ms per frame over ~6 minutes. Emission events were identified via an intensity threshold on integrated counts. Multi-emitter frames were excluded using intensity variance; first/last frames of events were discarded to avoid partial sequences; valid frames per event were summed to yield one localization. Localizations could be filtered by total photon count (e.g., N > 450) to ensure precision. Typical average N ~ 1900 for accepted events.

• Theory: For L = 100 nm, N = 1000, K = 6×6, SBR = 4, σCRB at the excitation center is ~1 nm; precision <5 nm extends beyond the excitation pattern and is larger for RASTMIN than for MINFLUX under matched conditions. Average σCRB improves with decreasing L (σCRB ∝ L) and higher N; RASTMIN and MINFLUX exhibit equivalent photon efficiency and outperform ideal camera-based localization at the same SBR. Increasing K improves precision markedly up to ~36; gains beyond are marginal; recommended K ≈ 16–100.
• Experimental single-molecule localization: With SBR ≈ 4 and frame time 130 ms, varying power yielded mean photon counts per frame of ~276, 971, and 2166, with measured single-molecule localization uncertainties consistent with theory plus drift error. Example histograms showed σ ≈ 3.2 nm at N ≈ 239, 1.4 nm at N ≈ 1036, and 1.0 nm at N ≈ 2212. Accounting for drift correction (lateral σDC ≈ 0.8–1.3 nm), measured σTOT matched √(σCRB^2 + σDC^2). RASTMIN achieved ~1–2 nm precision for moderate photon counts (N ≈ 700–2000).
• Super-resolution imaging: RASTMIN combined with blinking reconstructed the designed 3×2 fluorophore array on DNA origami with 15 and 20 nm spacings. The distribution width at individual sites yielded an experimental localization precision of ~2 nm; the theoretical average precision for the dataset (mean N ~ 1900) was ~1.5 nm.
• Practical performance: RASTMIN sequences can be completed in ~1–20 ms with standard galvanometric scanners, meeting the requirement to be faster than typical fluorophore ON-times (10–500 ms).

The results demonstrate that RASTMIN achieves MINFLUX-level localization precision with substantially reduced experimental complexity by leveraging standard scanning microscopes augmented with a doughnut-like excitation profile and active drift stabilization. Theoretical and experimental data confirm that RASTMIN’s photon efficiency matches MINFLUX, with σCRB scaling with L and SBR setting the ultimate limit. The larger area of high-precision localization compared to MINFLUX under matched conditions is advantageous for robust localization. Importantly, temporal resolution is governed by scanning speed rather than the number of exposures K, allowing flexible trade-offs without compromising photon efficiency. Integrating RASTMIN with fluorophore blinking enables super-resolved imaging with ~2 nm precision at individual sites, validating its applicability to nanostructures such as DNA origami. The approach democratizes nanometer-precision SML by removing specialized hardware requirements of MINFLUX while preserving performance, provided nm-scale drift stabilization is ensured.

RASTMIN, a raster-scanned minimum-of-light SML-SSI method, delivers nanometer localization precision (~1–2 nm) comparable to MINFLUX using standard laser-scanning fluorescence microscopes with minimal modifications (vortex phase plate for a toroidal focus and active drift correction). Theory and experiments confirm high photon efficiency, predictable scaling (σ ∝ L), and practical viability for super-resolution imaging via fluorophore blinking, successfully resolving 15–20 nm spacings on DNA origami with ~2 nm precision. Future work could extend RASTMIN to 3D localizations, integrate lifetime or spectral dimensions, explore parallelization strategies balancing throughput and resolution, and apply the method in live-cell contexts while maintaining nm-scale stabilization.

• Localization improvements with decreasing L are ultimately limited by achievable SBR.
• Increasing K beyond ~36 yields marginal precision gains and may be constrained by scanner positioning precision.
• Requires active drift correction with ~1 nm lateral stability (and ~1.5–2.0 nm axial), which may be challenging in some setups/samples.
• Temporal resolution depends on scanner speed; for imaging, total acquisition remains limited by fluorophore blinking dynamics, resulting in minute-long acquisitions similar to other SMLM techniques.
• Camera-based implementations are disadvantaged by super-Poissonian noise; RASTMIN relies on single-photon counting detectors for optimal performance.