Introduction

Real-time capturing of nanoscale transformations within liquid environments remains a significant challenge, hindering real-time observation of dynamic processes essential for catalysis, nanoparticle synthesis, and materials science1,2,3. Processes occurring at the solid–liquid interface, such as catalysis and nanoparticle transformation, often exhibit complex, spatially heterogeneous reaction dynamics that govern reactivity, selectivity, and product formation2,4,5,6,7,8. Conventional single-particle spectroscopy techniques, such as localized surface plasmon resonance (LSPR)9,10, provide high sensitivity to nanoparticle morphology and local environment changes. However, they suffer from poor temporal resolution and throughput, limiting their application for rapid reaction processes. Liquid-phase transmission electron microscopy (TEM) offers nanometer-scale spatial resolution, but introduces non-native reaction pathways owing to the high-energy electron beam for imaging11,12. Atomic force microscopy (AFM) can resolve surface changes at the atomic scale, but its slow imaging speed prevents effective tracking of rapid reaction dynamics13,14,15,16. Optical microscopy has emerged as a promising, non-invasive alternative for studying chemical processes, from catalysis and nanoparticle growth to material synthesis17,18,19. However, due to the optical diffraction limit (ODL), nanoparticles smaller than the diffraction limit appear as blurred diffraction-limited spots, precluding direct observation of sub-particle structural changes2. Although recent advancements in plasmonic imaging have improved sensitivity and resolution, these approaches are still constrained by signal deconvolution challenges20,21, making it difficult to resolve intricate dynamics at the sub-particle scale.

Here, we present Rotary-Encoded Scattering Microscopy (RESM), an optical imaging technique that is capable of tracking sub-particle level transformations in real time. RESM dynamically engineers the PSF to encode sub-particle structural changes into distinctive scattering patterns, enabling real-time, label-free imaging of chemical reactions at the sub-particle level with high spatiotemporal resolution. Unlike conventional scattering-based approaches that rely solely on intensity changes, RESM uses rotary encoding to extract sub-particle dynamics from the rotational symmetry and shape of the scattering patterns. This method enables the real-time visualization of complex chemical reactions at heterogeneous interfaces, revealing reaction anisotropy and inferring the presence of intermediate states that have remained undetectable using existing techniques. The development of RESM addresses a critical gap in the field of optical imaging by providing a tool for probing chemical processes at the sub-particle scale in their native liquid environments.

Results

Principle of RESM for sub-particle imaging

The principle of RESM is based on dynamically encoding sub-particle structural information into characteristic scattering patterns that vary with the rotational symmetry of the nanoparticles through the use of azimuthally rotating p-polarized illumination. The incident light is directed at an angle greater than the critical angle through an oil-immersion objective (N.A. = 1.49), resulting in total internal reflection at the glass-solution interface. This configuration generates an evanescent wave on the solution side, selectively exciting nanoparticles near the interface and minimizing background scattering (Fig. 1a). The evanescent field interacts with nanoparticles near the interface, inducing dipole radiation that is modulated by the physical properties of the nanoparticles and the surrounding medium (Fig. 1b)22,23,24. The radiation pattern of these dipoles depends on their orientation relative to the interface. A dipole perpendicular to the interface radiates primarily downward at the critical angle, producing a characteristic doughnut-shaped pattern, whereas a dipole parallel to the interface emits mainly at lower angles, resulting in a solid spot (Fig. 1c, d)25,26. By maintaining p-polarization with a vortex wave plate27, RESM enhances the perpendicular polarization component, which maximizes the formation of the doughnut-shaped scattering pattern (Supplementary Information Section 1). The reflected light is effectively blocked at the Fourier plane of the optical path, and the scattered light is selectively collected. The distinct scattering patterns encode structural symmetry and dynamics of the nanoparticles (Fig. 1a). The use of azimuthal rotation ensures that the scattered field varies with the symmetry and orientation of the nanoparticle, resulting in distinctive doughnut-shaped or asymmetrical scattering patterns depending on the structural symmetry of nanoparticle. Mathematically, the scattering pattern intensity \(I\left(\rho,\phi \right)\) collected by the camera can be expressed as

$$I\left(\rho,\phi \right)=\int {\left|{{{\bf{E}}}}\left(\rho,\phi,\psi \right)\right|}^{2}{{{\rm{d}}}}\psi+{I}_{{bg}}$$
(1)

where \({{{\bf{E}}}}\left(\rho,\phi,\psi \right)\) is the electric field distribution at the radial and azimuthal coordinates \(\left(\rho,\phi \right)\) for a given azimuthal angle \(\psi\) of the incident light, and \({I}_{{bg}}\) is the background intensity.

Fig. 1: Imaging Principle of RESM.
Fig. 1: Imaging Principle of RESM.
Full size image

a Schematic diagram of the light path in RESM. p-polarized light with the incident angle greater than the critical angle undergoes total internal reflection at the glass/solution interface, simultaneously generating evanescent waves on the solution side. Azimuthally rotating p-polarized illumination (red arrow) was achieved by modulating the light’s azimuth with a scanning galvanometer. BFP: back focal plane. The scattered light of the nanoparticles at the interface is collected by the camera, while the reflected light is blocked by the iris. b Different polarized nanoparticles, and (c) corresponding equivalent dipoles, d Far-field patterns. The nanoparticles can be polarized by illumination light incident at various angles (\({\theta }_{{trans}}\)) (b). The scattering behavior of polarized nanoparticles can be equivalent to the radiation of dipoles in different polarization states (c). The red lines in c correspond to the radiation patterns in the lower half space of dipoles. The disparity in far-field patterns of dipoles in c reflects the change in dipole orientation on the interface (d). e Imaging principle of RESM. The exposure time is an integral multiple of the rotation period of the galvanometer. Nanoparticles with azimuth symmetry have the same response to excited light at all azimuths, thus superimposing a scattering pattern with rotational symmetry, while nanoparticles without such property superimpose a scattering pattern without rotational symmetry.

For nanoparticles with rotational symmetry, the response to azimuthal rotation illumination is consistent, producing a uniform doughnut-shaped scattering pattern. However, for nanoparticles lacking rotational symmetry or undergoing structural transformations, the response to illumination at different azimuths varies, leading to orientation-specific distortions in the scattering pattern (Fig. 1e). By synchronizing the modulation of the azimuthal angle with the sampling rate of camera, RESM integrates the structural responses over a complete rotation cycle, encoding detailed symmetry information into the scattering pattern. This dynamic mapping provides real-time insights into structural evolution at the sub-particle scale, enabling the detection of reaction anisotropy and morphological transformations that are challenging to detect with conventional intensity-based techniques.

Imaging of sub-particle transformation using RESM

To demonstrate the capability of real-time visualizing dynamic reaction at the sub-particle scale, we apply RESM to probe the oxidation of Ag NSs in FeCl3 solution as a model reaction system28. The overall reaction can be described as follows:

$${{{\rm{Ag}}}}+{{{{\rm{Fe}}}}}^{3+}\to {{{{\rm{Ag}}}}}^{+}+{{{{\rm{Fe}}}}}^{2+}$$
(2)
$${{{{\rm{Ag}}}}}^{+}+{{{{\rm{Cl}}}}}^{-}\to {{{\rm{AgCl}}}}$$
(3)

A microfluidic flow system was used to control the delivery of FeCl3 solutions to the Ag NSs immobilized on a glass substrate, ensuring consistent reaction conditions throughout the experiments (Supplementary Fig. 4). We firstly compared RESM with a conventional dark-field (DF) imaging configuration in capturing morphological transformations information of the reaction (Supplementary Fig. 1). For RESM, the initial scattering pattern of Ag NSs appears as a doughnut shape, resulting from dipole polarization under a p-polarized evanescent field (top panel in Fig. 2a). As the oxidation reaction proceeds, these patterns undergo dynamic distortion accompanied by pronounced intensity fluctuations (blue line in Fig. 2c, Supplementary Movie 1), suggesting a disruption of the rotational symmetry of particles. The result indicates complex anisotropic oxidation mechanisms occurring at the sub-particle level, which is distinct from a simple core-shrinking model29.

Fig. 2: Dynamic imaging of oxidation process of Ag NSs using RESM.
Fig. 2: Dynamic imaging of oxidation process of Ag NSs using RESM.
Full size image

a Snapshot sequence of Ag NSs during oxidation in FeCl3 solution. The top panel corresponding to RESM mode, while the bottom panel corresponding to DF mode. b Simulated scattering patterns of Ag NSs undergoing asymmetric oxidation. c, d Time series plot of normalized intensity of Ag NSs during oxidation in FeCl3 solution under DF and RESM model. The curves in (c) corresponding to the particles in (a). d is statistically obtained from the normalized intensity curves of dozens of Ag NSs. The top, middle, and bottom curves correspond to the first, second, and third quartiles of all particle intensities at each moment, respectively. e Normalized simulated scattering intensity curve of a Ag NS when oxidized in the manner shown in (b). Scalebar, 500 nm.

In contrast, DF imaging shows only a solid spot that decreases monotonically in intensity, with no discernible pattern evolution or symmetry breaking (bottom panel in Fig. 2a, c, Supplementary Movie 1). This suggests a gradual reduction in scattering ability but does not reveal any intermediate states or localized structural changes. The inability of DF imaging to resolve sub-particle dynamics is attributed to its reliance on intensity variations alone, without encoding structural information into the scattering pattern. DF aggregates the responses of nanoparticles to light of different wavelengths, polarizations, and incident angles into a single spot, resulting in signals that reflect only broad trends and mask finer structural changes, which is a problem prevalent in other similar imaging schemes, such as transmission bright field microscopy (Supplementary Information Section 2).

In contrast, the ability of RESM to modulate wavelength, polarization, and incident angle enables it to reveal detailed reaction mechanisms that remain imperceptible to traditional label-free optical techniques. By encoding structural information into the scattering patterns, RESM can distinguish between symmetrical and asymmetrical transformations, as well as dynamic changes in nanoparticle morphology. To analyze the dynamic distortions and intensity fluctuations observed in RESM scattering patterns, we performed statistical analysis on the response behaviors of multiple particles. Statistical analysis confirmed that the observed pattern dynamics are general across multiple particles (Fig. 2d).

To further validate our experimental observations, we conducted finite element simulations of the scattering patterns (Fig. 2b). Finite-element simulations further support these findings, reproducing the transition from symmetric doughnuts to asymmetric patterns as structural alterations are introduced in RESM. The simulations also reveal that the scattering intensity does not necessarily decrease monotonically but can exhibit fluctuations corresponding to changes in nanoparticle morphology (Fig. 2e and Supplementary Fig. 7). These results confirmed that RESM was able to dynamically map sub-particle transformations and directly correlate changes in the scattering patterns with chemical reactions occurring at the nanoscale.

Revealing concentration-dependent oxidation mechanisms using RESM

To elucidate the influence of FeCl3 concentration on the oxidation mechanisms of Ag NSs, we systematically varied the FeCl3 concentration and monitored the transformations using RESM. By extracting orientation mappings of scattering patterns from experimental image sequences (Fig. 3a), we visualized the dynamic evolution of Ag NSs during the reaction. This approach enabled us to track subtle changes in the symmetry and intensity of scattering pattern over time, providing insights into the concentration-dependent dynamics that govern oxidation pathways.

Fig. 3: Concentration-dependent oxidation behavior of Ag NSs in FeCl3 solution probed by RESM.
Fig. 3: Concentration-dependent oxidation behavior of Ag NSs in FeCl3 solution probed by RESM.
Full size image

a Schematic illustration of the process for extracting orientation mapping from a time sequence of RESM images. Orientation mappings (top panel of b, c) and normalized intensity curves (bottom panel of b, c) of Ag NSs during oxidation in FeCl3 solution with different concentration, where (b) correspond to a concentration of 10 µM, while (c) correspond to a concentration of 1 mM. The deposition of AgCl on the surface of Ag particle after a certain induction time (light blue patch in c) alters the reaction mechanism of Ag particle, shifting it from isotropic to anisotropic (light yellow patch in c).

At low FeCl3 concentrations (10 µM), the scattering intensity of the nanoparticle decreased monotonically with negligible change in orientation (Fig. 3b, Supplementary Movie 2). The scattering patterns maintained a symmetric doughnut shape throughout the reaction, implying that the dissolution of Ag occurred uniformly across the nanoparticle surface, and no AgCl nucleation sites or other structural intermediates were detected in this stage. This behavior suggests that the reaction was diffusion-controlled, with the radial diffusion of Fe3+ ions ensuring an even oxidation rate at all points on the nanoparticle surface30.

In contrast, oxidation in a 1 mM FeCl3 solution resulted in significant fluctuations in the intensity curve and rapid changes in scattering pattern orientation (Fig. 3c, Supplementary Movie 3), indicating a violent anisotropic reaction. The stability of the scattering pattern in control experiments with Au nanospheres under identical conditions (Supplementary Fig. 8) confirms that potential Fe(OH)3 precipitation does not cause interfacial artifacts. Several factors could account for the discrepancy. At higher reactant concentration conditions, the oxidation reaction over nanoparticles may become highly anisotropic due to kinetic factors. The increased concentration of Fe3+ ions can enhance reaction rates at specific sites on the nanoparticle surface, leading to non-uniform oxidation30. Conversely, at low reactant concentrations, reaction kinetics is completely diffusion-controlled, ensuring isotropy in particle reaction dynamics through radial mass transfer in the liquid phase.

We considered the possibility that the observed fluctuations were due to rolling behavior of the nanoparticles induced by force imbalances from vigorous anisotropic oxidation31,32,33. Rolling may exacerbate blinking frequency or amplitude fluctuations in the scattering intensity. To rule out this factor, we conducted supplementary experiments on silver nanocubes (Ag NCs) and silver nanowires (Ag NWs), which have a larger contact area with the substrate and thus exhibit stronger adhesion, minimizing rolling during reactions34,35. However, we still observed drastic changes in scattering patterns and intensity curves in the oxidation experiments of Ag NCs and Ag NWs (Supplementary Fig. 9, Fig. 10), suggesting that the signals observed mainly originate from the reaction itself rather than particle motion.

The electron transfer kinetics at the solid-liquid interface are closely tied to the reactivity of nanoparticles. Sharp corners or defects often exhibit enhanced reactivity36,37, while oxide patches on the surface of Ag nanoparticles can significantly hinder local oxidation reactions38. Such heterogeneously distributed reactivity could promote a transition of from isotropic to anisotropic reaction behavior. In our system, this corresponds to the precipitation of AgCl on the surface of Ag NSs. To eliminate the interference from AgCl, we repeated the oxidation experiment with 10 mM Fe(NO3)3 solution. The results showed that most nanoparticles maintained stable scattering patterns, with intensity curves declining monotonically, similar to the behavior observed in 10 μM FeCl3 solutions (Fig. 4a, Supplementary Movie 4). The occasional non-monotonic trends in intensity curves were likely due to the presence of natural oxides on the Ag NSs, which can lead to anisotropic oxidation38. However, the proportion of these particles was small (~30%), and the blinking events were occasional (1–2 times). In contrast, when Ag NSs reacted in a 50 μM FeCl3 solution, we observed high-frequency blinking of scattering patterns and sharp fluctuations in scattering intensity (Fig. 4b, Supplementary Movie 5). Given that the reaction time scales for both Fe(NO3)3 and FeCl3 experiments were comparable, we conclude that the precipitation of AgCl, rather than the spatial distribution of nanoparticle reactivity, primarily drives the strong anisotropy observed in high-concentration FeCl3 solutions. The presence of Cl⁻ ions leads to the formation of AgCl precipitates on the nanoparticle surface, disrupting the symmetry and causing anisotropic oxidation.

Fig. 4: Concentration-dependent reaction kinetics and morphological transformations of Ag NSs.
Fig. 4: Concentration-dependent reaction kinetics and morphological transformations of Ag NSs.
Full size image

Statistical normalized scattering intensity variations of Ag NSs over time when oxidating in 10 mM Fe(NO3)3 solution (a) and in 50 µM FeCl3 solution (b). The thick lines in the middle corresponds to the median intensities, and the top and bottom thin lines correspond to the first and third quartiles of the intensities. When FeCl3 concentration is low, Ag+ do not precipitate on the surface of Ag NPs (c). However, as the FeCl3 concentration increases, AgCl starts to precipitate and adhere to particle surfaces (e), altering the mass transfer dynamics around the particles and intensifying the reaction’s anisotropy (d). f Apparent reaction rate profile as a function of FeCl3 concentration. The thick line in the middle corresponds to the median apparent reaction rate, and the top and bottom two thin lines correspond to the first and third quartiles of the apparent reaction rate. g Mean blinking times curve of the particles with blinking behavior during the oxidation process versus with the concentration of FeCl3 solutions. h Proportion curve of particles with blinking behavior during the oxidation process versus with the concentration of FeCl3 solutions. To extract the critical concentration and better study the transformation of reaction kinetics, the experimental data in (g, h) are fitted with a sigmoid function. Based on the fitting results, three concentration intervals with distinct reaction kinetics were identified: I, isotropic dissolution stage; II, transition stage; III, anisotropic oxidation stage. i SEM image of Ag NSs after reacting in 1 mM FeCl3 solution. Ag NSs reaction mechanisms when reacting in low (j) and high (k) concentration FeCl3 solutions.

To deconvolute the respective effects of Fe3+ and Cl- on Ag NSs oxidation, we performed decouple experiments (Supplementary Fig. 11). The results confirm that varying the concentration of Fe3+, which acts as the oxidant, directly modulates the reaction rate (Supplementary Fig. 11a, b). Concurrently, the addition of Cl- further accelerates the dissolution kinetics by facilitating the formation of reaction products with low solubility (Supplementary Fig. 11a, b). Replacing Fe3+ with dissolved oxygen, a weaker oxidant, resulted in a slower oxidation rate (Supplementary Fig. 12).

Unveiling reaction anisotropy with sub-particle resolution

The formation of structural intermediates, such as AgCl nucleation sites, at the nanoparticle surface requires a critical degree of supersaturation39,40,41. In our experiment, when the FeCl3 concentration was below 30 μM, supersaturation was insufficient to promote AgCl nucleation on the surface of Ag NSs (Fig. 4c, Supplementary Section 7). As FeCl3 concentration increases, the probability of nucleation rose sharply, and the resulting AgCl precipitates significantly influenced the reaction dynamics (Fig. 4d, e). To investigate this transition, we conducted concentration gradient experiments. The oxidation rate of Ag NSs increased approximately linearly with FeCl3 concentration up to around 100 µM, after which the rate of increase noticeably slowed, indicating a plateau (Fig. 4f). Statistical analysis of the mean blinking times (MBT) and the proportion of blinking particles (PBP) among all Ag NSs revealed three distinct concentration ranges (Fig. 4g, h, statistical method detailed in Supplementary Fig. 14): (I) At low concentrations (up to 30 μM), approximately 20% particles exhibited occasional blinking (1-2 events), independent of FeCl3 concentration, while the reaction rate increased linearly with FeCl3 concentration; (II) As FeCl3 concentration increased, a growing number of particles exhibited blinking, while the reaction rate increased at a slower rate; (III) At high FeCl3 concentrations, all particles exhibited frequent blinking, and both reaction rate and blinking behaviors gradually stabilized.

During stage I, low FeCl3 concentrations result in a slow dissolution rate of Ag+, preventing the accumulation of AgCl nucleation. The dissolution of Ag occurred uniformly across the nanoparticle surface, as indicated by the isotropic scattering patterns observed throughout the reaction (Fig. 4c, d, Fig. 4j). In our system, the reactivity differences across various crystal planes of Ag NSs are negligible, further contributing to the isotropic nature of the dissolution process. Occasional blinking events, which were independent of FeCl3 concentration, were likely due to inherent surface defects, such as unevenly distributed oxide patches, which influence localized dissolution. As FeCl3 concentration increases, MBT and PBP show a sigmoidal concentration-dependent response. Fitting the transition behavior of MBT and PBP with sigmoid functions yields critical transition concentrations of 70 µM and 40 µM, respectively. The PBP curve exhibits a significantly steeper gradient than the MBT curve, highlighting the substantial influence of AgCl precipitation on the reaction dynamics. The initial monotonic decrease in scattering intensity across all concentrations (Fig. 3c, light blue block; Supplementary Fig. 15) corresponded to an induction period before nucleation, during which the supersaturation near the particle surface gradually increased. Nucleation rate is highly sensitive to supersaturation40,41, with small increases in FeCl3 concentration extremely accelerating the nucleation rate, sharply enhancing the probability of AgCl precipitation and contributing to the steep transition observed in the PBP curve.

Stage II: As the FeCl3 concentration increased further, localized AgCl nucleation, inferred through changes in scattering pattern symmetry and intensity, disrupted the isotropic nature of the reaction, introducing significant asymmetry. This behavior marked the onset of anisotropic growth, where the formation of AgCl intermediates at discrete sites broke the uniformity of the reaction fronts observed at lower concentrations. Unlike the traditional nano-Kirkendall effect, where voids form due to imbalanced ion diffusion through an oxide layer and reaction fronts propagate along specific energy-favorable crystal faces42,43, the anisotropic growth observed here was driven by the random nucleation of AgCl, leaving isolated regions of high reactivity. As AgCl nuclei grew, they consumed Ag+ and Cl- ions, thereby limiting further nucleation and creating isolated regions of high reactivity40. This behavior was further corroborated by ex situ scanning electron microscopy (SEM) characterization, which revealed the presence of multiple AgCl aggregates rather than a continuous AgCl shell or single particle (Fig. 4i). The competitive growth of AgCl clusters altered the propagation of the reaction front, resulting in the complex anisotropies captured by RESM.

During Stage III, at high FeCl3 concentrations, the oxidation process was dominated by repetitive cycles of AgCl nucleation and growth, where these nucleation sites acted as structural intermediates that induced periodic changes in the scattering patterns. Large AgCl clusters formed, covering substantial portions of the nanoparticle surface and hindering further local oxidation reactions. Uncovered regions continued to dissolve, periodically accumulating supersaturation and triggering subsequent cycles of nucleation. This repetitive nucleation-growth behavior, manifested as rapid blinking and sharp fluctuations in scattering intensity, demonstrated that the system had entered a stable yet dynamically evolving state, where the balance between Ag⁺ consumption and release dictated the overall reaction kinetics (Fig. 4k). In contrast to the conventional oxidation of metal nanoparticles, where the oxide layer forms uniformly on the particle surface (Supplementary Fig. 16), the precipitation of AgCl in FeCl3 solutions results in spatially heterogeneous reactivity, with separate AgCl clusters forming in solution rather than directly on the Ag surface. Similarly, the much higher solubility product (Ksp) of Ag2SO4 compared to AgCl prevents precipitate formation in Fe2(SO4)3 solution. The single-particle scattering signals exhibited a monotonic decrease without significant fluctuations (Supplementary Fig. 17), indicating complete dissolution without forming stable asymmetric precipitate. This distinction highlights the fundamentally different oxidation pathways in liquid-phase environments and highlights the versatility of RESM in capturing these complex sub-particle dynamics. Thus, the ability to dynamically map these processes offers valuable insights into the interplay between reaction conditions, intermediate formation, and nanoparticle reactivity.

To further demonstrate the versatility of RESM, we applied this technique to study the intercalation reaction of Prussian blue (PB) nanoparticles, a promising electrode material for large-scale energy storage due to its excellent chemical properties and low cost44. By periodically switching the applied potential between 0.2 V and −0.2 V, we observed periodic changes in the scattering intensity of these nanoparticles, indicating the reversible transformation between PB and Prussian white (PW) (Supplementary Fig. 18a). During each unidirectional cycle (PB  →  PW or PW  →  PB), the scattering intensity changed monotonically, and the orientation of the scattering pattern remained largely unchanged (Supplementary Fig. 18a, b). This suggests that the PB/PW intercalation-deintercalation process is nearly isotropic due to the highly symmetrical structure of the PB nanoparticles. However, when analyzing the intensity evolution of different subregions within the scattering patterns, we found variations in synchronicity at the sub-particle scale (Supplementary Fig. 18c). Notably, the response of these subregions varies among different particles, implying that such heterogeneity originates from intrinsic particle properties rather than imaging artifacts (Supplementary Fig. 18c). Such heterogeneity may stem from spatial variations in the microstructure of the catalyst45. Furthermore, poor contact between the particles and the substrate could also contribute to localized differences in intercalation kinetics46. By employing an enhanced PSF, RESM effectively amplifies sub-particle-scale differences, highlighting subtle variations that are undetectable by conventional optical imaging techniques.

Discussion

We have developed an optical technique, RESM, that is capable of real-time and label-free imaging of solution chemical reactions at the sub-particle scale, overcoming the limitations of conventional optical microscopy. By dynamically encoding sub-particle structural information into distinctive scattering patterns via PSF engineering, RESM enables the visualization of complex reaction dynamics at the sub-particle level and the identification of structural intermediate with high spatiotemporal resolution and without the need for additional labeling. The capability of RESM to track subtle structural transformations was demonstrated using the oxidation of Ag NSs as a model system, revealing concentration-dependent reaction pathways that transition from isotropic dissolution to anisotropic growth of AgCl clusters. The detailed analysis of scattering pattern evolution suggests the involvement of reaction intermediates in localized nucleation events and sub-particle anisotropy, revealing changes in reaction dynamics and transformation pathways that were previously undetectable with traditional optical imaging. Furthermore, by applying RESM to PB nanoparticles, we captured anisotropic intercalation dynamics of PB nanoparticles within highly symmetric materials. These findings highlight the ability of RESM to capture reaction heterogeneity and sub-particle structural changes, offering new insights into the role of reaction intermediates in governing nanoparticle morphology and reaction kinetics. Importantly, the designability of the PSF in RESM offers high flexibility. By customizing the PSF according to specific objectives and highlighting signals of interest, RESM can enhance imaging sensitivity, making it a versatile methodology with broad applicability in fields ranging from catalysis and electrochemistry to nanomaterial synthesis. By pushing the limits of optical detection and analysis capabilities, this work contributes to the broader goal of developing advanced imaging techniques that bridge the gap between structural characterization and functional performance in nanomaterials. As optical imaging continues to evolve, RESM offers a promising platform for elucidating fundamental processes in photonics and catalysis, and could support for future studies in high-sensitivity optical imaging of dynamic chemical and material systems.

Methods

Setup of RESM

The RESM was built based on an inverted microscope configuration. A 660 nm laser (OBIS LX, Coherent) was employed for microscopic imaging experiments. An oil-immersion objective lens (60X, N.A. = 1.49, Nikon) was utilized for illumination and collection of scattered light. Adjustment of the illumination angle was achieved by focusing the incident light to different positions on the BFP of the objective with a scanning galvanometer (GVS012, Thorlabs, USA). Reflected light was effectively blocked using an iris at the conjugate plane of the BFP, thus achieving dark-field effect, while the remaining scattered light was collected and imaged by a camera (Pike F-032B, Allied Vision Technologies, Germany) at the sampling rate of 10 fps (Temporal resolution: 100 ms) or 100 fps (Temporal resolution: 10 ms), depending on the rate of reactions. Each RESM image is formed by synchronizing the camera exposure with the azimuthal scan so that the illumination completes an integer number of full rotations during a single exposure. Images obtained from the setup were further magnified by a factor of 1.8, resulting in a final magnification of 108. For specific optical pathways and detailed descriptions of the setup, please refer to the Supplementary Information (see Supplementary Information Section 1 for details).

Microfluidic Experiments

PDMS microfluidic chips with a straight micro flow channel of 0.5 mm in diameter were purchased from Anhui Chixin Biotechnology Co., Ltd (Fuyang, China). 24 mm × 60 mm coverslips (Citotest Scientific Co., Ltd, Nanjing, China) were used as the substrate to deposit Ag nanoparticles. Before the experiment, both the chip and the coverslip were sonicated with deionized water for several minutes, then rinsed three times with deionized water and dried with nitrogen gas. After drying, the chip and coverslip were attached together by draining the air between them. Two Tygon tubes (1.52 mm outer diameter, 0.51 mm inner diameter) were inserted into the inlet and outlet (1.3 mm diameter) of the flow channel. The Tygon tube connected to the outlet was connected to a syringe, while the tube connected to the inlet was inserted into a centrifuge tube. An equal volume of KNO3 solution (0.1 M) was added to the solution of ultrasonically dispersed Ag nanoparticles and mixed rapidly to destabilize the particles and make them adhesion to the glass substrate easier, and subsequently added to the centrifuge tube. The syringe was operated continuously and slowly to ensure a constant flow of fresh solution into the channel. The process was monitored by a camera in real time, and the solution was switched to deionized water when the density of silver nanoparticles in the field of view was moderate. The syringe was pumped repeatedly to flush away loosely adhered nanoparticles from the substrate. Before the formal experiment, the solution was switched to a specific concentration of FeCl3 solution or Fe(NO3)3 solution. The Tygon tube was pulled out of the syringe and placed at the appropriate height. Subsequently, the oxidant solution flows spontaneously under gravity and capillary action to trigger the reaction. The flow rate of the solution is predetermined by adjusting the height of the Tygon tube, ensuring the concentration of the oxidant remains constant within the flow channel (Supplementary Fig. 4).

Characterization

TEM (TEM, H7650, Hitachi Co., Japan) was used to characterize the morphology of the nanoparticles. And the SEM images of Ag NSs and Ag NCs used in the experiments were in Supplementary Fig. 19.

Numerical model for calculating the RESM image of nanoparticles

To calculate the RESM pattern of nanoparticles, we first used a commercial finite element simulation software (COMSOL 6.0) to calculate the near field of the scattering light from nanoparticles on a glass substrate excited by light illuminated at a particular azimuth (\(\psi\)). The calculated near field was then transformed to far field by an open-source program (RETOP)47. The far field result served as input for calculating the complex electric field distribution on the camera with vectorial diffraction integrals. We take the square of the amplitude of the complex electric field as the result of the pattern acquired by the camera. The previous calculation is repeated by changing \(\psi\) in steps of \(\frac{2\pi }{N}\), resulting in N patterns. Finally, these N pictures are summed to serve as the RESM pattern for this particle. The larger N is, the closer the obtained pattern is to the true result. N = 36 was chosen for all simulations in this work. See Supplementary Information Section 13 for details.

Data processing

All data were processed using custom-written programs in MATLAB software (2023a, MathWorks). The position of nanoparticles was determined with an open-source localization algorithm48. Due to the usage of a highly coherent light source, there exists significant speckle noise within the imaging field of view, which has been smoothed out by the azimuth-rotation strategy, providing a slowly varying background. To eliminate signal fluctuations caused by environmental changes, all intensity data were corrected by means of background subtraction. The regions of interest (ROI) for particles and their respective backgrounds were determined as

$${RO}{I}_{{particle}}=\left\{(x,y)\left|\sqrt{{\left(x-{x}_{c}\right)}^{2}+{\left(y-{y}_{c}\right)}^{2}}\right.\le 10\right\}$$
(4)
$${RO}{I}_{{background}}=\left\{(x,y)\left|40 < \sqrt{{\left(x-{x}_{c}\right)}^{2}+{\left(y-{y}_{c}\right)}^{2}}\right.\le 50\right\}$$
(5)

where \(({x}_{c},{y}_{c})\) is the coordinates of the nanoparticle with the unit being pixel. The mean intensities of these two ROIs were used as the signal and background, respectively. Considering the slow-changing nature of the background, the selection of the above two ROIs is reasonable.