Interferometric Scattering Microscopy (iSCAT)

• iSCAT is a label-free optical technique that detects nanoscale objects via interference between weak scattered and strong reference fields.
• It utilizes engineered pupil geometries, homodyne amplification, and advanced vectorial diffraction models to achieve nanometer localization with microsecond temporal resolution.
• The quantitative framework of iSCAT enables precise particle sizing, mass estimation, and dynamic tracking, impacting studies from biomolecular analysis to cellular imaging.

Interferometric scattering microscopy (iSCAT) is a label-free optical microscopy technique that detects nanoscopic objects by measuring the interference between a weak field scattered by the object and a stronger reference field, typically reflected from a nearby interface. Across its modern implementations, iSCAT is characterized by homodyne amplification of weak scattering, nanometer-scale localization, microsecond-to-millisecond temporal resolution, and compatibility with wide-field, common-path, and engineered-pupil geometries. The technique has developed from a sensitive nanoparticle-imaging modality into a broader quantitative framework for tracking, sizing, mass estimation, multiparametric characterization, chemical readout, and ultrafast dynamics, while also motivating rigorous treatments of vectorial point spread functions, aberrations, and information-theoretic precision limits (Taylor et al., 2018).

1. Interferometric signal formation and contrast

The core iSCAT observable is the detected intensity generated by interference between a reference field and a scattered field. In its standard form,

$$I_{\mathrm{det}} \propto |E_{\mathrm{ref}}+E_{\mathrm{sca}}|^2 = |E_{\mathrm{ref}}|^2 + |E_{\mathrm{sca}}|^2 + 2|E_{\mathrm{ref}}||E_{\mathrm{sca}}|\cos\phi ,$$

where $\phi$ is the phase difference between reference and scattered fields. For small scatterers, the $|E_{\mathrm{sca}}|^2$ term is often negligible, so the signal is dominated by the interference cross-term, which is linear in the scattering amplitude rather than the scattering intensity (Taylor et al., 2018).

This linearization is central to the sensitivity of iSCAT. In the Rayleigh regime, the particle polarizability is commonly written as

$$\alpha = 4\pi a^3\left(\frac{\epsilon_1-\epsilon_m}{\epsilon_1+2\epsilon_m}\right),$$

with scattering cross-section

$C_{\mathrm{sca}}=\frac{k^4}{6\pi}|\alpha|^2 .$

The experimentally used normalized contrast can be expressed as

$$C=\frac{I_{\mathrm{det}}-I_{\mathrm{ref}}}{I_{\mathrm{ref}}} =2\frac{|s|}{|r|}\cos\phi+\frac{|s|^2}{|r|^2},$$

which makes explicit that the phase and the ratio of scattered to reference amplitude jointly determine detectability (Mahmoodabadi et al., 2020).

A major practical theme in iSCAT has therefore been reference-field management. In numerical-aperture-shaped iSCAT, a partially transmissive spatial mask near the back focal plane attenuates the strong low-NA reflected background much more than the high-NA scattered light. The extinction contrast is enhanced approximately as $T^{-1/2}$, where $T$ is the mask transmissivity. However, for shot-noise-limited detection, the signal-to-noise ratio does not increase: the gain in raw contrast is offset by the photon-number penalty. This distinction has become important in interpreting contrast-enhanced iSCAT configurations, because improved visibility in raw camera images is not identical to a better shot-noise limit (Cole et al., 2016).

2. Vectorial point spread functions, aberrations, and axial encoding

A central conceptual shift in the field has been the move from Gaussian approximations of the image spot to explicit modeling of the interferometric point spread function (iPSF). A robust vectorial diffraction model based on the Richards–Wolf formalism describes the detector field in high-NA reflection iSCAT and incorporates stratified media, polarization, and aberration through an optical path difference term. In this treatment, aberrations arise from refractive index mismatches at the sample–coverslip interface and from high-NA oil-immersion objectives imaging into aqueous media, and they modify both the amplitude and the phase of the scattered field (Mahmoodabadi et al., 2020).

These aberrations are not merely parasitic. They induce an asymmetric, extended axial iPSF, shift the maximum iSCAT signal away from nominal geometric focus, and generate a slowly varying Gouy phase. As a result, the iPSF can remain highly sensitive over an axial range on the order of $10\,\mu\mathrm{m}$, far exceeding the depth of focus associated with conventional intensity-based PSFs. Ring structure and side lobes evolve uniquely with axial displacement, so the lateral shape itself becomes a 3D encoding variable rather than a nuisance distortion (Mahmoodabadi et al., 2020).

The same requirement for explicit field modeling appears in multiscale settings. A theoretical and experimental study of samples whose structural dimensions differ by $4$–$\phi$0 orders of magnitude showed that sub-nanometer surface roughness on glass coverslips or mica can substantially modify the absolute signal and the PSF of a gold nanoparticle. In simplified cell-mimetic environments, the PSF becomes position-dependent and even asymmetric, especially near dielectric interfaces such as a nucleus. These results establish that high-fidelity iSCAT depends on modeling coherent background contributions from roughness, layered substrates, and cellular structure rather than treating the background as a featureless offset (Lin et al., 2020).

3. Three-dimensional localization and single-particle tracking

The performance of iSCAT in single-particle tracking follows directly from its interferometric sensitivity and the indefinite photostability of scattering labels. In supported lipid bilayers, tracking of $\phi$1 gold nanoparticles attached to GM1 ganglioside or DOPE lipids achieved $\phi$2 spatial precision at $\phi$3 temporal resolution, with trajectories longer than $\phi$4 localizations. Those data resolved strong transient confinements within domains as small as $\phi$5, multiple mobilities, and deviations from normal diffusion, demonstrating that high-speed iSCAT can access membrane dynamics beyond what mean-square-displacement analysis alone reveals (Hsieh et al., 2013).

Subsequent work extended axial tracking from the previously typical few-hundred-nanometer regime to tens of micrometers. A robust 3D localization strategy for high-speed iSCAT videos combines experimentally calibrated iPSF stacks, radial variance transform (RVT) extraction of 1D radial profiles, Pearson-correlation mapping against model profiles, and graph-theoretical trajectory assignment. Physically connected regions in the temporal correlation map are scored, branch ambiguities are resolved with Dijkstra’s algorithm, and adaptive polynomial fitting refines the axial coordinate below the model grid spacing. This framework demonstrated continuous 3D tracking of gold nanoparticles as small as $\phi$6 diffusing in water at $\phi$7 temporal resolution, with an average axial localization error of $\phi$8 for $\phi$9 particles (Kasaian et al., 2024).

Pupil engineering has addressed a separate limitation of conventional iSCAT: rapid oscillations of signal-to-noise ratio with axial position. In spiral-phase iSCAT, a spiral phase mask placed in the Fourier plane redistributes the phase of the scattered field uniformly across phase space while leaving the reference phase flat. The resulting spiral-phase iPSF does not blink through contrast nulls; instead, it winds through distinct shapes as the particle moves axially. Numerical and experimental results showed nanometer-scale localization precision in all three dimensions, with typical lateral precision of about $|E_{\mathrm{sca}}|^2$0–$|E_{\mathrm{sca}}|^2$1, axial precision of about $|E_{\mathrm{sca}}|^2$2–$|E_{\mathrm{sca}}|^2$3, an amplitude error as low as $|E_{\mathrm{sca}}|^2$4, and 3D tracking at $|E_{\mathrm{sca}}|^2$5 frames per second (Brooks et al., 2024).

A complementary line of work has replaced framewise heuristic processing by full-image probabilistic inference. A Bayesian framework that fits a parameterized iPSF model with Hamiltonian Monte Carlo yields posterior distributions for the particle’s 3D position and polarizability, together with uncertainties and parameter correlations. The same framework was used to track a diffusing particle in three dimensions, infer a diffusion coefficient directly without calculating a mean-square displacement, and quantify DNA ejection from an individual lambda phage virus. The significance is methodological: iSCAT images can be treated as a generative inverse problem rather than a contrast-extraction problem (Wit et al., 2023).

4. Contrast engineering, phase control, and speckle suppression

Because iSCAT sensitivity is often limited by background contrast rather than by the existence of a detectable scattering field, a large fraction of recent work has focused on engineered contrast. One approach generalizes the reference model itself. A modified theory of iSCAT that includes unexpected reflections from all optical interfaces—objectives, waveplates, beam splitters, and other elements, not only the sample–glass interface—predicts position-dependent variations in both intensity and phase of the reference field. In that framework, oblique illumination suppresses many off-axis reflections, and experimental measurements showed about a $|E_{\mathrm{sca}}|^2$6-fold improvement in signal-to-background ratio with only minor loss of particle peak intensity. A scan-based phase-map calibration using fixed particles and a piezo stage then compensates residual position-dependent phase through

$|E_{\mathrm{sca}}|^2$7

and the combination of intensity and phase correction yielded up to $|E_{\mathrm{sca}}|^2$8-fold improvements in signal-to-background ratio without rotational scanning (Zhang et al., 3 Nov 2025).

Another strategy exploits the defocus symmetry of the background. For speckles originating from surface undulations modeled as thin dielectric nanodisks, the contrast evolves anti-symmetrically with defocus,

$|E_{\mathrm{sca}}|^2$9

whereas nanoparticle signals are not anti-symmetric. Defocus-integration iSCAT therefore acquires a stack of images symmetrically around focus and numerically integrates them along $$\alpha = 4\pi a^3\left(\frac{\epsilon_1-\epsilon_m}{\epsilon_1+2\epsilon_m}\right),$$0:

$$\alpha = 4\pi a^3\left(\frac{\epsilon_1-\epsilon_m}{\epsilon_1+2\epsilon_m}\right),$$1

The anti-symmetric speckle cancels, whereas the nanoparticle contribution persists. The anti-symmetry coefficient was reported as $$\alpha = 4\pi a^3\left(\frac{\epsilon_1-\epsilon_m}{\epsilon_1+2\epsilon_m}\right),$$2–$$\alpha = 4\pi a^3\left(\frac{\epsilon_1-\epsilon_m}{\epsilon_1+2\epsilon_m}\right),$$3 for speckle, compared with $$\alpha = 4\pi a^3\left(\frac{\epsilon_1-\epsilon_m}{\epsilon_1+2\epsilon_m}\right),$$4 for dielectric nanoparticles and $$\alpha = 4\pi a^3\left(\frac{\epsilon_1-\epsilon_m}{\epsilon_1+2\epsilon_m}\right),$$5 for gold nanoparticles, and experimentally the method increased SNR by $$\alpha = 4\pi a^3\left(\frac{\epsilon_1-\epsilon_m}{\epsilon_1+2\epsilon_m}\right),$$6 for dielectric particles and $$\alpha = 4\pi a^3\left(\frac{\epsilon_1-\epsilon_m}{\epsilon_1+2\epsilon_m}\right),$$7 for gold nanoparticles, without hardware modification (Jiao et al., 2024).

Fourier-plane phase control has pushed the same idea further. A photothermal phase plate placed on the conjugate back focal plane was used to synchronize the phase difference between all scattering components and the reference beam in a high-NA microscope. Synchronizing the Fourier-plane phase to a constant improved the PSF, produced near-perfect circular symmetry, and enhanced the interference contrast by more than $$\alpha = 4\pi a^3\left(\frac{\epsilon_1-\epsilon_m}{\epsilon_1+2\epsilon_m}\right),$$8 for a $$\alpha = 4\pi a^3\left(\frac{\epsilon_1-\epsilon_m}{\epsilon_1+2\epsilon_m}\right),$$9 gold nanoparticle. Localization uncertainty improved from $C_{\mathrm{sca}}=\frac{k^4}{6\pi}|\alpha|^2 .$0 to $C_{\mathrm{sca}}=\frac{k^4}{6\pi}|\alpha|^2 .$1 in $C_{\mathrm{sca}}=\frac{k^4}{6\pi}|\alpha|^2 .$2. When the synchronized phase was set to $C_{\mathrm{sca}}=\frac{k^4}{6\pi}|\alpha|^2 .$3, the background speckles became anti-symmetric with defocus, enabling defocus-integration suppression with a mean SBR enhancement of $C_{\mathrm{sca}}=\frac{k^4}{6\pi}|\alpha|^2 .$4 and detection of immobilized particles as small as $C_{\mathrm{sca}}=\frac{k^4}{6\pi}|\alpha|^2 .$5 on a glass substrate (Lin et al., 17 Oct 2025).

5. Simulation frameworks and fundamental precision bounds

The increasing complexity of iSCAT experiments has made rigorous forward models indispensable. Multiscale electromagnetic modeling has combined Lorentz reciprocity, vectorial diffraction theory, and finite-element methods to compute the angular spectrum of arbitrary samples embedded in planar multilayer structures and to propagate those fields through high-NA imaging systems. This framework explained how local topography, particle material, substrate roughness, and cellular heterogeneity reshape the interferometric signal and PSF, and it provided a route toward high-fidelity iSCAT in real applications (Lin et al., 2020).

A separate unified simulation platform implemented iSCAT, coherent bright-field, and dark-field image formation in a Matlab toolbox using a boundary element method for scattering from arbitrary particle geometries and stratified substrates, together with a fully vectorial imaging model based on Richards–Wolf propagation. The reported simulations agreed quantitatively with experiments for different particle shapes and excitation angles, and they reproduced contrast enhancement near the Brewster angle. The broader implication is that iSCAT image formation is no longer restricted to dipole or Gaussian approximations; arbitrary sample geometries and optical stacks can be handled within a single numerical workflow (Hitzelhammer et al., 2024).

Such forward models have enabled rigorous lower bounds on estimator performance. Using classical Cramér–Rao bounds for shot-noise-limited images and quantum Cramér–Rao bounds for the scattered optical state, one study showed that iSCAT provides significantly better axial localization precision than coherent bright-field microscopy (COBRI) and dark-field microscopy for the same number of collected scattered photons, while all three schemes have similar transverse precision per scattered photon. For mass photometry, the quantum limit is

$C_{\mathrm{sca}}=\frac{k^4}{6\pi}|\alpha|^2 .$6

where $C_{\mathrm{sca}}=\frac{k^4}{6\pi}|\alpha|^2 .$7 is the number of collected scattered photons (Dong et al., 2021).

More recent theory has refined this picture for off-axis illumination. Quantum Cramér–Rao calculations for oblique illumination predicted an anisotropic information gain in iSCAT: for a glass–water interface, the $C_{\mathrm{sca}}=\frac{k^4}{6\pi}|\alpha|^2 .$8-localization quantum bound improved from $C_{\mathrm{sca}}=\frac{k^4}{6\pi}|\alpha|^2 .$9 on-axis to $$C=\frac{I_{\mathrm{det}}-I_{\mathrm{ref}}}{I_{\mathrm{ref}}} =2\frac{|s|}{|r|}\cos\phi+\frac{|s|^2}{|r|^2},$$0 off-axis, and for glass–air interfaces the axial bound improved by a factor of about $$C=\frac{I_{\mathrm{det}}-I_{\mathrm{ref}}}{I_{\mathrm{ref}}} =2\frac{|s|}{|r|}\cos\phi+\frac{|s|^2}{|r|^2},$$1. The same analysis also showed that rotating coherent scattering microscopy (ROCS), despite its higher spatial resolution, has worse localization precision than iSCAT with oblique illumination, and that Brewster-angle illumination can increase contrast while worsening the localization limit. This has crystallized a recurrent caution in the field: spatial resolution and raw contrast are not sufficient metrics for localization precision (Hitzelhammer et al., 3 Oct 2025).

6. Quantitative extensions and emerging modalities

Several variants of iSCAT have broadened the set of observables beyond position and intensity. Axial profiling of iSCAT contrast, combined with a vectorial PSF model and $$C=\frac{I_{\mathrm{det}}-I_{\mathrm{ref}}}{I_{\mathrm{ref}}} =2\frac{|s|}{|r|}\cos\phi+\frac{|s|^2}{|r|^2},$$2-stack fitting, was used to determine the height of the scattering dipole and thereby the size of spherical dielectric nanoparticles beyond the Rayleigh limit. The method was reported to measure sizes from tens of nanometers and beyond the Rayleigh regime with nanometer precision, and was also applied to fluorescent nanodiamonds, for which a reasonable size estimate and a correlation between fluorescence signal and particle size were observed (Žambochová et al., 2022).

Dual-angle interferometric scattering microscopy (DAISY) extends iSCAT by combining backward scattering images with forward scattering acquired by twilight off-axis holography. The forward channel directly yields polarizability through the complex scattered field, while the backward-to-forward ratio constrains size through an optical form factor. DAISY thus enables simultaneous quantification of size and polarizability without requiring a priori information about the surrounding medium or diffusion-based super-resolution analysis. It was used to optically differentiate biomolecular fractal aggregates from spherical particles in fetal bovine serum at the single-particle level (Olsén et al., 2023).

Chemical specificity has been added through bond-selective iSCAT. In that configuration, a pulsed mid-infrared laser excites molecular vibrations, and the resulting photothermal modulation of the particle polarizability is detected by a visible probe in a wide-field common-path interferometric geometry. A thin film layered substrate reduces reflected light while preserving a reference field, and virtual lock-in camera demodulation extracts the photothermal signal. The method was demonstrated on polymer beads and on microorganisms including Staphylococcus aureus, Escherichia coli, and Candida albicans, providing simultaneous nanoscale interferometric sensitivity and bond-selective contrast (Yurdakul et al., 2021).

Time-resolved material dynamics have motivated femtosecond iSCAT. Femto-iSCAT combines femtosecond pump–probe microscopy with interferometric scattering and a partial reflector that attenuates the reference field more than the scattered field. The reported transient image contrast exceeded previous approaches by more than $$C=\frac{I_{\mathrm{det}}-I_{\mathrm{ref}}}{I_{\mathrm{ref}}} =2\frac{|s|}{|r|}\cos\phi+\frac{|s|^2}{|r|^2},$$3 orders of magnitude, while the temporal resolution reached $$C=\frac{I_{\mathrm{det}}-I_{\mathrm{ref}}}{I_{\mathrm{ref}}} =2\frac{|s|}{|r|}\cos\phi+\frac{|s|^2}{|r|^2},$$4. Applications included transport of different energy carriers at interfaces, heterogeneous hot-electron relaxation in single plasmonic resonators, and edge-state dynamics in optoelectronic semiconductors (Lyu et al., 2022).

Cellular micromotion has motivated yet another reformulation. ChiSCAT uses chaotic speckle illumination in a reflective high-NA common-path geometry to maximize sensitivity to motion in any point and any direction, at the price of losing visually interpretable images. Unsupervised learning based on matched filtering and motif discovery then recovers recurrent motion patterns and detects action potentials. In blebbistatin-paralyzed cardiomyocytes, the unsupervised analysis recovered action potentials in $$C=\frac{I_{\mathrm{det}}-I_{\mathrm{ref}}}{I_{\mathrm{ref}}} =2\frac{|s|}{|r|}\cos\phi+\frac{|s|^2}{|r|^2},$$5 of regions, compared with $$C=\frac{I_{\mathrm{det}}-I_{\mathrm{ref}}}{I_{\mathrm{ref}}} =2\frac{|s|}{|r|}\cos\phi+\frac{|s|^2}{|r|^2},$$6 for supervised matched filtering (Trelin et al., 2024).

At the level of integrated label-free cellular imaging, bidirectional quantitative scattering microscopy (BiQSM) combines backward-scattering iSCAT and forward-scattering quantitative phase microscopy using off-axis digital holography with bidirectional illumination and spatial-frequency multiplexing. The resulting FS and BS data are simultaneous and spatiotemporally consistent, and the dynamic range was reported to be $$C=\frac{I_{\mathrm{det}}-I_{\mathrm{ref}}}{I_{\mathrm{ref}}} =2\frac{|s|}{|r|}\cos\phi+\frac{|s|^2}{|r|^2},$$7 times wider than QPM alone. This enables simultaneous visualization of microscale cellular architecture and nanoscale dynamics, with FS–BS correlation analysis used to probe proteins, lipids, and membranes (Horie et al., 19 Mar 2025).