Rotating Coherent Scattering Microscopy (ROCS)

• ROCS is a coherent-scattering imaging technique that rotates an oblique illumination beam about the optical axis to create rotationally averaged images with a sharper point spread function.
• Its incoherent summation of azimuthally resolved iSCAT images suppresses interference fringes, providing improved spatial resolution at the expense of reduced 3D localization precision per scattered photon.
• ROCS is best suited for applications prioritizing enhanced image contrast, resolution, and robustness in dense samples, in contrast to off-axis iSCAT which excels in per-photon localization accuracy.

Rotating coherent scattering microscopy (ROCS) is a coherent-scattering imaging modality that, in the formulation analyzed in "Fundamental Bounds for Off-Axis Illumination in Interferometric and Rotating Coherent Scattering Microscopy" (Hitzelhammer et al., 3 Oct 2025), is a particular realization of off-axis coherent scattering microscopy. It illuminates a nanoscopic particle near an interface with an oblique beam whose incidence direction is rotated around the optical axis, and forms the final image by incoherently summing the intensity patterns associated with the different azimuthal illumination directions. Within coherent localization microscopy, ROCS is closely related to interferometric scattering microscopy (iSCAT): both rely on elastic scattering of coherent light and on the presence of a reference field generated by reflection at the interface, but ROCS removes much of the azimuth-dependent interference structure present in single-angle iSCAT frames. The consequence emphasized in recent theory is that ROCS can exhibit a narrower effective point spread function (PSF) and higher spatial resolution while nevertheless delivering worse three-dimensional localization precision per scattered photon than off-axis iSCAT (Hitzelhammer et al., 3 Oct 2025).

1. Concept and placement within coherent scattering microscopy

In the analysis of (Hitzelhammer et al., 3 Oct 2025), ROCS is constructed from an iSCAT-like geometry in which a particle lies on a glass substrate and is illuminated by an oblique plane wave at polar angle $\theta$. For each illumination direction, specified by an azimuthal angle around the optical axis, one computes an iSCAT image $I^{(\theta,z_f)}(x,y)$; the ROCS image is then the incoherent sum of these images over all azimuthal directions at fixed $\theta$ and focus position $z_f$ (Hitzelhammer et al., 3 Oct 2025). This rotational integration corresponds experimentally to rotating the incidence direction and integrating the camera intensity over one full rotation.

This definition places ROCS between conventional iSCAT and dark-field-like coherent scattering schemes. In iSCAT, the detected signal is the interference between a strong reference field reflected from the interface and a weak field scattered by the particle. In ROCS, each instantaneous frame is generated by the same basic interference physics, but the final data are not interferometric in the same way as iSCAT because the azimuth-resolved interference fringes are washed out by incoherent averaging. The rotationally averaged signal is therefore distinct from a strictly coherent detection scheme even though its constituent frames are coherent-scattering images (Hitzelhammer et al., 3 Oct 2025).

The experimental motivations cited for ROCS in the earlier literature, including Ruh et al. 2018, Jünger et al. 2022, and Iqbal et al. 2025, are increased spatial resolution, boosted image contrast, and greater robustness to uncontrolled interference from dense samples. These reported advantages concern image quality and visualization. The central theoretical question posed in (Hitzelhammer et al., 3 Oct 2025) is narrower and more specific: when the performance criterion is three-dimensional localization precision per scattered photon, does ROCS outperform or underperform iSCAT? The answer is underperform.

2. Image formation and vectorial forward model

For a single illumination direction, coherent scattering at the detector is described by the standard interferometric intensity expression

$$I(\mathbf{r}') \propto |\mathbf{E}'_{\mathrm{tot}}(\mathbf{r}')|^2 = |\mathbf{E}'_{\mathrm{ref}}|^2 + |\mathbf{E}'_{\mathrm{sca}}|^2 + 2\Re\!\left[\mathbf{E}'_{\mathrm{ref}}\cdot \mathbf{E}'_{\mathrm{sca}}{}^{*}\right],$$

with primed quantities denoting image-side fields (Hitzelhammer et al., 2024). In the ROCS setting of (Hitzelhammer et al., 3 Oct 2025), the relevant single-frame quantity is the iSCAT intensity $I^{(\theta,\varphi,z_f)}(x,y)$ for illumination polar angle $\theta$, azimuth $\varphi$, and focus position $z_f$. The rotationally averaged ROCS image is

$$I_{\mathrm{ROCS}}^{(\theta,z_f)}(x,y)=\frac{1}{2\pi}\int_0^{2\pi} I^{(\theta,\varphi,z_f)}(x,y)\,d\varphi.$$

Because this averaging is performed over intensities rather than fields, relative phases between different azimuths are not preserved, and the outer fringe structure characteristic of single-angle iSCAT is partially suppressed (Hitzelhammer et al., 3 Oct 2025).

The forward model used to analyze ROCS is fully vectorial and high-NA. In (Hitzelhammer et al., 3 Oct 2025), the scattered field is computed either with a full boundary element method (BEM) or with a quasistatic dipole approximation, and imaging is propagated through the objective with a vector Debye-type integral. The back focal plane (BFP) field is represented by $I^{(\theta,z_f)}(x,y)$0 over objective angles $I^{(\theta,z_f)}(x,y)$1 and $I^{(\theta,z_f)}(x,y)$2, with maximum collection angle $I^{(\theta,z_f)}(x,y)$3 fixed by numerical aperture through $I^{(\theta,z_f)}(x,y)$4. The same general formalism underlies the simulation framework in "Unified simulation platform for interference microscopy" (Hitzelhammer et al., 2024), where incoming fields are written as a plane-wave decomposition,

$I^{(\theta,z_f)}(x,y)$5

and the imaging system is modeled with a fully vectorial Richards–Wolf integral.

Within that framework, ROCS corresponds to a specific illumination geometry: a narrow ring of coherent plane waves in the pupil that is rotated azimuthally. For a fixed ring position, an idealized pupil field can be written as

$I^{(\theta,z_f)}(x,y)$6

with $I^{(\theta,z_f)}(x,y)$7 set by the illumination NA and $I^{(\theta,z_f)}(x,y)$8 specifying polarization (Hitzelhammer et al., 2024). A simulated ROCS acquisition then consists of solving the scattering problem for a sequence of azimuths $I^{(\theta,z_f)}(x,y)$9, computing the corresponding images $\theta$0, and forming the intensity average

$\theta$1

3. Fisher information, Cramér–Rao bounds, and quantum limits

The localization analysis in (Hitzelhammer et al., 3 Oct 2025) is formulated in terms of Fisher information (FI), the Cramér–Rao bound (CRB), quantum Fisher information (QFI), and the quantum Cramér–Rao bound (QCRB). The parameters of interest are the particle coordinates $\theta$2. Assuming Poisson-distributed counts in the image plane, the Fisher information matrix is

$\theta$3

where $\theta$4 is the image region of integration and the position derivatives are computed numerically by finite differences (Hitzelhammer et al., 3 Oct 2025). The CRB lower-bounds the variance of any unbiased estimator:

$\theta$5

and, when cross-terms are ignored or small, the standard-deviation bound is

$\theta$6

A central normalization throughout (Hitzelhammer et al., 3 Oct 2025) is per scattered photon collected by the imaging system. This removes absolute photon number from the comparison and turns the bound into a statement about information extracted per photon rather than about total photon budget. For ROCS, the formulas are unchanged in form, but $\theta$7 is replaced by the rotationally averaged ROCS image. The change in image structure changes the spatial derivatives $\theta$8 and therefore the available Fisher information.

The QCRB provides the measurement-independent lower limit determined by the quantum state of the scattered field. With the scattered light modeled as a superposition of coherent states over the objective aperture, the QFI for parameter $\theta$9 is

$z_f$0

with normalization

$z_f$1

The corresponding bound is

$z_f$2

By construction, any classical detection scheme, including iSCAT and ROCS, must satisfy $z_f$3 (Hitzelhammer et al., 3 Oct 2025).

4. Off-axis illumination and redistribution of localization information

The first major theoretical result of (Hitzelhammer et al., 3 Oct 2025) is that off-axis illumination can increase the available information for localization, but anisotropically. The paper visualizes information transport with an FI-flow field for the $z_f$4 coordinate,

$z_f$5

Under on-axis illumination, the FI is roughly symmetrically distributed in forward and backward directions, so high NA is required to collect most of it. Under off-axis illumination at $z_f$6 and $z_f$7, the FI flow is redistributed and enhanced in the backward direction, increasing the fraction collected by the same detection system (Hitzelhammer et al., 3 Oct 2025).

For a gold nanosphere near a glass–water interface, the QCRB for the $z_f$8 direction, defined as the in-plane direction of tilt, improves from approximately $z_f$9 under on-axis illumination to a minimum of approximately $$I(\mathbf{r}') \propto |\mathbf{E}'_{\mathrm{tot}}(\mathbf{r}')|^2 = |\mathbf{E}'_{\mathrm{ref}}|^2 + |\mathbf{E}'_{\mathrm{sca}}|^2 + 2\Re\!\left[\mathbf{E}'_{\mathrm{ref}}\cdot \mathbf{E}'_{\mathrm{sca}}{}^{*}\right],$$0 under off-axis illumination, corresponding to a gain factor of approximately $$I(\mathbf{r}') \propto |\mathbf{E}'_{\mathrm{tot}}(\mathbf{r}')|^2 = |\mathbf{E}'_{\mathrm{ref}}|^2 + |\mathbf{E}'_{\mathrm{sca}}|^2 + 2\Re\!\left[\mathbf{E}'_{\mathrm{ref}}\cdot \mathbf{E}'_{\mathrm{sca}}{}^{*}\right],$$1. In contrast, the $$I(\mathbf{r}') \propto |\mathbf{E}'_{\mathrm{tot}}(\mathbf{r}')|^2 = |\mathbf{E}'_{\mathrm{ref}}|^2 + |\mathbf{E}'_{\mathrm{sca}}|^2 + 2\Re\!\left[\mathbf{E}'_{\mathrm{ref}}\cdot \mathbf{E}'_{\mathrm{sca}}{}^{*}\right],$$2-direction QCRB is nearly flat as a function of illumination angle, indicating little or no improvement. For the axial coordinate $$I(\mathbf{r}') \propto |\mathbf{E}'_{\mathrm{tot}}(\mathbf{r}')|^2 = |\mathbf{E}'_{\mathrm{ref}}|^2 + |\mathbf{E}'_{\mathrm{sca}}|^2 + 2\Re\!\left[\mathbf{E}'_{\mathrm{ref}}\cdot \mathbf{E}'_{\mathrm{sca}}{}^{*}\right],$$3, the behavior depends on refractive-index contrast: at a glass–water interface the QCRB does not improve with $$I(\mathbf{r}') \propto |\mathbf{E}'_{\mathrm{tot}}(\mathbf{r}')|^2 = |\mathbf{E}'_{\mathrm{ref}}|^2 + |\mathbf{E}'_{\mathrm{sca}}|^2 + 2\Re\!\left[\mathbf{E}'_{\mathrm{ref}}\cdot \mathbf{E}'_{\mathrm{sca}}{}^{*}\right],$$4, whereas at a glass–air interface it improves from approximately $$I(\mathbf{r}') \propto |\mathbf{E}'_{\mathrm{tot}}(\mathbf{r}')|^2 = |\mathbf{E}'_{\mathrm{ref}}|^2 + |\mathbf{E}'_{\mathrm{sca}}|^2 + 2\Re\!\left[\mathbf{E}'_{\mathrm{ref}}\cdot \mathbf{E}'_{\mathrm{sca}}{}^{*}\right],$$5 to approximately $$I(\mathbf{r}') \propto |\mathbf{E}'_{\mathrm{tot}}(\mathbf{r}')|^2 = |\mathbf{E}'_{\mathrm{ref}}|^2 + |\mathbf{E}'_{\mathrm{sca}}|^2 + 2\Re\!\left[\mathbf{E}'_{\mathrm{ref}}\cdot \mathbf{E}'_{\mathrm{sca}}{}^{*}\right],$$6, a gain factor of approximately $$I(\mathbf{r}') \propto |\mathbf{E}'_{\mathrm{tot}}(\mathbf{r}')|^2 = |\mathbf{E}'_{\mathrm{ref}}|^2 + |\mathbf{E}'_{\mathrm{sca}}|^2 + 2\Re\!\left[\mathbf{E}'_{\mathrm{ref}}\cdot \mathbf{E}'_{\mathrm{sca}}{}^{*}\right],$$7 (Hitzelhammer et al., 3 Oct 2025).

The same anisotropic behavior appears at the classical level in iSCAT. In glass–water, $$I(\mathbf{r}') \propto |\mathbf{E}'_{\mathrm{tot}}(\mathbf{r}')|^2 = |\mathbf{E}'_{\mathrm{ref}}|^2 + |\mathbf{E}'_{\mathrm{sca}}|^2 + 2\Re\!\left[\mathbf{E}'_{\mathrm{ref}}\cdot \mathbf{E}'_{\mathrm{sca}}{}^{*}\right],$$8 decreases with increasing $$I(\mathbf{r}') \propto |\mathbf{E}'_{\mathrm{tot}}(\mathbf{r}')|^2 = |\mathbf{E}'_{\mathrm{ref}}|^2 + |\mathbf{E}'_{\mathrm{sca}}|^2 + 2\Re\!\left[\mathbf{E}'_{\mathrm{ref}}\cdot \mathbf{E}'_{\mathrm{sca}}{}^{*}\right],$$9 for a wide range of particle and focus positions, with gains of approximately $I^{(\theta,\varphi,z_f)}(x,y)$0–$I^{(\theta,\varphi,z_f)}(x,y)$1; $I^{(\theta,\varphi,z_f)}(x,y)$2 worsens with increasing $I^{(\theta,\varphi,z_f)}(x,y)$3; and $I^{(\theta,\varphi,z_f)}(x,y)$4 generally worsens at larger $I^{(\theta,\varphi,z_f)}(x,y)$5. In glass–air, axial behavior can improve. The physically relevant point is that off-axis illumination redistributes rather than uniformly increases information. Improvements along one coordinate can be accompanied by degradation along another (Hitzelhammer et al., 3 Oct 2025).

5. ROCS versus off-axis iSCAT

The direct ROCS–iSCAT comparison in (Hitzelhammer et al., 3 Oct 2025) shows that a narrower PSF does not imply a better localization bound. The iSCAT PSFs under oblique illumination have a central lobe surrounded by interference rings whose structure changes strongly with illumination angle. The ROCS PSF is obtained by rotating the oblique illumination around the optical axis and summing the corresponding iSCAT images incoherently. The result is narrower and largely lacks the outer interference rings visible in iSCAT. This narrowing is consistent with earlier ROCS literature emphasizing improved spatial resolution.

Localization behavior is different. For equal photon budgets, ROCS yields systematically worse localization CRBs than off-axis iSCAT. For a representative case at $I^{(\theta,\varphi,z_f)}(x,y)$6, the ROCS value of $I^{(\theta,\varphi,z_f)}(x,y)$7 is more than twice the iSCAT value. Because $I^{(\theta,\varphi,z_f)}(x,y)$8, ROCS would require approximately five times more photons to achieve the same $I^{(\theta,\varphi,z_f)}(x,y)$9 precision. Across defocus values, the ROCS curves for $\theta$0 and $\theta$1 lie above the corresponding iSCAT curves; ROCS becomes rotationally symmetric, so its $\theta$2 and $\theta$3 precision are identical, but both are inferior to the best iSCAT direction aligned with the illumination tilt (Hitzelhammer et al., 3 Oct 2025).

The mechanism is the loss of high-information interference structure. In iSCAT, the redistribution of intensity within the interference rings as the particle moves carries substantial Fisher information. ROCS suppresses those rings through incoherent azimuthal averaging. That operation sharpens the central lobe and can improve visual resolution, but it reduces the image gradients $\theta$4 that enter the FI integrals. The paper’s conceptual conclusion is therefore explicit: spatial resolution, understood through PSF width, and localization precision, understood through CRB, are not equivalent metrics (Hitzelhammer et al., 3 Oct 2025).

This conclusion also reframes other common proxies for performance. The same study notes that contrast and resolution are not reliable indicators of localization capability: Brewster-angle iSCAT can have higher contrast yet worse CRBs, and ROCS can have a sharper PSF yet lower FI per photon. A sparse-sample experiment aimed at single-particle, high-precision 3D localization therefore favors off-axis iSCAT without incoherent rotation. ROCS remains relevant when dense samples or uncontrolled interference make stable iSCAT fringes difficult to exploit.

6. System parameters, simulation workflows, and design space

ROCS performance inherits the same dependencies that govern off-axis coherent scattering microscopy more broadly: illumination angle $\theta$5, numerical aperture, wavelength $\theta$6, refractive-index contrast, detector sampling, and noise statistics. In (Hitzelhammer et al., 3 Oct 2025), increasing NA improves localization but with diminishing returns; for on-axis iSCAT, increasing $\theta$7 from $\theta$8 to $\theta$9 improves CRBs by only a factor of approximately $\varphi$0 near focus. The analysis is performed at $\varphi$1 for glass–water ($\varphi$2 versus $\varphi$3) and glass–air ($\varphi$4 versus $\varphi$5) interfaces. Shot-noise-limited Poisson statistics and sufficiently fine detector sampling are assumed throughout, with all bounds normalized per scattered photon (Hitzelhammer et al., 3 Oct 2025).

The modeling infrastructure of "Unified simulation platform for interference microscopy" (Hitzelhammer et al., 2024) makes this design space explicit for ROCS. The scattered fields are computed with a full Maxwell solver based on BEM in stratified media, using the MATLAB nanobem toolbox, and are propagated through a fully vectorial imaging model. The framework handles arbitrary particle shapes, layered substrates, and high-NA illumination, and introduces an arbitrary pupil-plane operator $\varphi$6 that can represent waveplates, attenuators, beam blocks, or ROCS-specific pupil shaping. This makes ROCS naturally expressible as ring-selected illumination plus angular scanning in the BFP.

That same formalism is also suited to Brewster-angle analysis. For p-polarized illumination at a planar interface, the reflected reference field is minimized near the Brewster angle, and simulations in (Hitzelhammer et al., 2024) show a pronounced contrast enhancement near that angle. In a ROCS setting, the ring radius can therefore be chosen so that the corresponding incidence angle lies near the Brewster angle, reducing specular background while maintaining measurable scattering. The framework is validated in (Hitzelhammer et al., 2024) against experiments on a 55 nm gold sphere and a 100 nm silver nanocube, and it is explicitly positioned as a route to simulate ROCS PSF width, symmetry, defocus dependence, contrast, and Fisher-information-based localization behavior.

The practical trade-off remains the same as in the bound analysis. Optimization of iSCAT can exploit illumination angles that maximize information along a chosen coordinate, especially along the tilt direction. ROCS cannot retain that directional advantage because azimuthal rotation restores rotational symmetry by averaging over the same raw photons. This suggests a design principle rather than a paradox: ROCS is appropriate when resolution, contrast, and robustness to complex backgrounds dominate, whereas off-axis iSCAT is appropriate when per-photon localization precision is the primary objective.

7. Related orientation-sensitive extensions and future directions

A related development is cylindrical-polarization-based interferometric scattering microscopy (cypiSCAT), which is described as conceptually close to ROCS because both use elastic coherent scattering from anisotropic nano-objects to infer orientation at high temporal resolution (Vala et al., 13 Jan 2026). cypiSCAT differs from classic ROCS by encoding orientation directly into a single interferometric PSF through a composite vortex half-wave plate placed in a pupil-conjugate plane. The reference wave passes through a central isotropic region, while the scattered field is transformed into a cylindrically polarized beam. Their interference produces a dipolar PSF whose orientation reports the particle’s in-plane angle.

The experimentally demonstrated performance in (Vala et al., 13 Jan 2026) reaches frame rates up to $\varphi$7 frames per second, exposure time $\varphi$8, static angular localization precision of approximately $\varphi$9, and torque sensitivity down to approximately $z_f$0. The method intrinsically suppresses isotropic background because the transformed isotropic scattered field is orthogonal to the reference and therefore does not contribute to the interferometric term. The paper also develops a full rotational-Langevin analysis pipeline from angular trajectories to $z_f$1, rotational friction $z_f$2, and torque $z_f$3 (Vala et al., 13 Jan 2026).

For ROCS, these results do not alter the localization bounds established in (Hitzelhammer et al., 3 Oct 2025), but they indicate a nearby design direction. A plausible implication is that ROCS-style systems aimed at rotational dynamics could move beyond purely temporal encoding by combining off-axis coherent scattering with preserved interferometric reference waves and pupil engineering. The data in (Vala et al., 13 Jan 2026) suggest that vortex-based PSF design, anisotropy gating, and explicit stochastic torque analysis are compatible with coherent-scattering microscopy and may extend ROCS from a resolution- and contrast-oriented modality toward calibrated orientation and torque metrology. Within the present theoretical picture, however, the central distinction remains: incoherent rotational averaging improves effective PSF width and robustness, but it discards phase-sensitive structure that is valuable for localization.