Irreversible dynamics imaged in space and time

Abstract

Ultrafast electron microscopy (UEM), also referred to as four-dimensional electron microscopy (4D-EM), has emerged as a transformative tool capable of recording atomic-scale dynamics with high spatiotemporal resolution. While the stroboscopic pump–probe mode has been extensively applied to reversible processes, many important phenomena are intrinsically irreversible and demand single-shot imaging. This review focuses on the principles and recent progress of single-shot 4D-EM for capturing non-repeatable events in space and time. We describe the two main single-shot modalities — single-frame and movie-mode (multi-frame) imaging — which rely on intense, ultrashort electron pulses to capture a transient following a single pump excitation. Representative applications are systematically surveyed, including laser-induced melting and rapid solidification, phase transitions, nanoparticle jumping and coalescence, redox and eutectic reaction dynamics, and liquid-phase behaviors. Key challenges, including the conflict between electron number and pulse duration imposed by space-charge effects, beam coherence, and detector limitations, are discussed, where correspondingly future perspectives and strategies are offered. The progress establishes single-shot 4D-EM as an indispensable platform for unraveling irreversible and stochastic processes across materials science, chemistry, physics, and biology.

Graphical abstract

Keywords

irreversible dynamics / single-shot imaging / ultrafast electron microscopy / 4D imaging / spatiotemporal characterization

1 Introduction

A reversible process is an ideal process in which the system and environment can be restored to exactly the same original states. Although truly reversible process is infrequent in nature, certain processes are treated close to reversible and therefore use the consequences of the corresponding reversible processes as a starting point or reference. Alternatively, an irreversible process is defined as a process in which the system and the surroundings do not return to their initial condition once the process is initiated. These processes have been extensively studied by various ex-situ/in-situ spectroscopy and microscopy techniques in the past years [1−8]. Nevertheless, ultrafast characterization techniques should be applied for the processes that occur on time scales ranging from nanoseconds to femtoseconds, owing to the typical response limitation of common detectors. Photons and electrons are two fundamental sources for characterization technologies, each of which possesses its own merits. For instance, photons have no dispersion in free space and no elementary charge that specialize in generating ultrashort pulses suitable for high temporal studies, while electrons with picometer wavelengths and strong interaction cross section are good at forming images with high spatial resolution. As such, the pump-probe scheme through these sources is normally used to achieve the high spatiotemporal resolution, with the highest level of attoseconds achieved recently [9−22].

Combining the advantages of these imaging sources, ultrafast electron microscopy, also known as 4D electron microscopy (4D-EM), is a powerful technique capable of imaging the dynamical processes in both space and time domains with high resolutions [19, 20]. This technique is based on transmission/scanning electron microscopy (TEM/SEM) that has been modified through the integration of ultrafast laser systems. It is categorized into two operation modes, namely, stroboscopic and single-shot modes. These two modes are utilized for probing the reversible and irreversible processes, respectively. For the stroboscopic mode, a sample is measured at a sufficient number of pump–probe time delays to capture the dynamics of the relevant processes. These measurements are then repeated to obtain an adequate signal-to-noise ratio (the sample should return to the same state before subsequent pump excitation and be stable throughout a potentially hours-long measurement). Over the past 20 years, 4D-EM has enabled the widespread use of its stroboscopic mode in materials science, physical, chemical and biological studies regarding the reversible processes, including phase transitions [23−25], mechanical oscillations [26−28], phonon-related dynamics [29, 30], photon-electron-matter interactions [31−35], magnetization-associated dynamics [36−39], charge transfer [40−43], biological dynamics [44−47], and technological developments [48−50].

On the other hand, the single-shot methodology has evolved progressively: it demonstrated early attempt at fast-TEM imaging employing a three-channel wave-memory system [51], then advanced with the introduction of photo-induced electron sources for nanosecond-scale imaging [52]. Subsequently, it has continued to evolve with the pump–probe strategy, which has enabled the imaging of a wide range of irreversible processes with temporal resolution reaching the nanosecond and picosecond regimes [53−55]. This technique enables the measurement (acquirement of an image at a certain time delay) of an evolving system through a single shot (exposure to a pump pulse for triggering the dynamics). Compared to the stroboscopic studies, the progress regarding the single-shot technique has been relatively slow due to the challenging issues in the technique (e.g., low signal-to-noise ratio). Since the irreversible processes prevail in nature, the development of the single-shot technique is indispensable and the technique has demonstrated successful applications in the fields of crystal growth, melting, nanoparticle motion and chemical reactions in recent years.

This review article will start from the principle of single-shot imaging in 4D-EM with a focus on the laser pump–electron probe scheme. Note that the single-shot imaging using optical detection sources [56, 57] falls out of the scope here. In the next section, we provide a comprehensive review on the applications of this technique to non-repeatable processes. After that, a summary and a perspective are provided to conclude this article.

2 Principle of single-shot imaging

The single-shot measurement encodes a transient into a single pulse pair imaged onto the detector. To achieve this aim, the laser repetition rate is normally down-converted from high rates to 1 Hz, ensuring that each image only consists of one pulse-pair per image. For the pump-probe scheme in 4D-EM (Fig. 1), there are two ways of single-shot techniques, namely, single-shot single-frame imaging and single-shot multiframe one [58]. In a typical measurement, the pump and probe pulses are spatially overlapped onto the sample. The signal measured by the probe pulse is dependent on the excitation density caused by the pump pulse. When the spatial overlap and pump intensity are consistent over the time delays, the background signals should be constant across all the measured time delays, which can be corrected by subtracting the baseline signal before time zero for reflecting the excited state dynamics of the sample. Such information is typically reported as the differential signal. Although this suggests that the signal is independent of the probe intensity, low probe intensity leads to a noisier signal. Therefore, a higher and more uniform probe intensity (typically larger than 10⁴ e⁻/pulse) is normally required to achieve a sufficient signal-to-noise ratio across the time delays, allowing the capture of suitable data for later analyses.

For the single-shot multiframe (movie-mode) imaging, it acquires a transient in a single pump pulse excitation by probing the signal over a series of time delays [Fig. 1(b)]. The technique relies on multiple intense electron pulses for imaging a laser-excited sample, with each pulse being sequentially deflected to a distinct position on the detector across a series of frames (typically 9 or 16 frames, each corresponding to a different time delay; see Fig. 1(c) for an example). Deflection is accomplished by a deflector system that generates the appropriate deflecting electric field, ensuring that each probe pulse reaches its intended detector location. Distinct from the single-shot single-frame imaging, this removes the need for imaging a transient through multiple pump-probe measurements on different samples. Further, it also reduces the inconsistency and ambiguity caused by shot-to-shot laser fluctuations and inhomogeneity from different samples. Since each acquisition is subdivided into tens of frames in the single-shot multiframe mode, the number of electrons in the probe pulses is typically larger than 10⁶ e⁻ so that an adequate signal-to-noise ratio is achieved in each frame on the detector. The single-shot multiframe mode of UEM enables the tracking of irreversible processes, and the entire dynamic process can be captured by varying the time interval of the experiment.

3 Representative examples for 4D single-shot visualization

The single-shot technology in 4D-EM has demonstrated the unique applications in the fields of melting, phase transformation, nanoparticle jumping, reaction dynamics and liquid dynamics. In this section, we review recent advancements of the irreversible processes visualized by such single-shot imaging techniques.

3.1 Melting and solidification

When a solid metal receives the irradiation from an intense laser pulse, it normally changes into a liquid (melting), followed by a rapid solidification due to the rapid cooling rates of 10⁵−10⁷ K/s during such laser-induced non-equilibrium process. The melting and solidification behaviors in metals and alloys initialized by laser have been observed by the single-shot multiframe mode of 4D-EM [59−61]. Figure 2(a) shows the laser-induced melting and resolidification in the 160 nm thick Al thin films [59]. The Al films were deposited on 50 nm thick amorphous Si₃N₄ membranes (electron transparent) of commercially available TEM grids. The experiments were conducted on different locations with varied distances, \mathrm{\Delta }x, measured from the center of the incident laser irradiation to the edge of the electron transparent window. The heating laser pulse was incident at an angle of ~45°, resulting in elliptically shaped irradiation cross-sections in the samples. The electron pulse train in the movie-mode series probed the involved transient processes, producing the nine-frame images at different time points. The diffuse elliptical shape was attributed to the increased thermal diffuse scattering associated with the laser-induced heating and melting of Al, where the intermediate-contrast regions presented the solid-to-liquid-to-solid transition sequences. Solidification rates were larger for the liquid in closer proximity to the edge of the electron transparent window than that on the opposite side of the melt pool at larger distances. In this regard, the last remaining liquid stayed at an off-center location relative to the center of the incident laser, e.g., the marker x in the frame (5) image. The solidification process completed after 10.2 μs. When the distance was larger, e.g., \mathrm{\Delta }x = 75 μm, the distortions of the melt pool shape (elliptical shape) were negligible. The last remaining liquid was nearly in the center of the laser irradiated area, and the solidification completed in a longer time, i.e., after 15.3 μs. For the locations with \mathrm{\Delta }x = 100 μm or larger, the incident laser pulse was far from the edge positions (effective heat sink) so that the thermal transport from the melt pool was essentially two-dimensional and limited to the Al layer. This gave rise to the symmetric heat extraction rates during the transformation sequence, and the solidification completed in the center of the elliptical shape irradiated area but with a much longer time, i.e., after 17.85 μs. Based on the spatiotemporal observations, the average migration velocity of the solidification interface section could be thereby estimated, which increased from ~1.35 to 2.5 m/s, with an average of ~1.9 m/s.

Figure 2(b) shows the rapid solidification behavior in the Al−4 at.%Cu thin-film alloys [60]. The initial dimensions of the melt pool were ~38 and 31 μm along the semi-major and semi-minor axes, respectively. The solid−liquid interface was sharp, and no obvious dendritic growth was observed. Directional crystal growth commenced between 5.1 and 7.65 μs, which was evidenced by the columnar grains that propagated radially inward. Such behavior became more apparent after ~10.2 μs, and was complete by ~32.75 μs. The measurement of the solid−liquid interface motion allowed for tracking of the solidification front evolution with time. It is seen that the solid−liquid interface accelerated as the solidification progressed, with the velocity along the semi-major axis increasing from ~0.67 m/s at 5.1 μs to 2.1 m/s at 30.2 μs. At the end of the solidification process (~28−32 μs), the morphological instability developed at the solid−liquid interface. The instability at the solidification front occurred at a critical velocity of ~2 m/s, which agreed with the ex-situ measurement results [62]. At this instability point, the growth transited from the inward radial direction to the oscillatory growth mode where the solidification front advanced along the directions both parallel and perpendicular to the isotherms of the melt pool. The banded microstructure was thereby formed when such oscillatory instability initiated at the critical velocity, typically stated as the point at which the solidification front velocity exceeded the diffusivity of solute in the liquid [63].

3.2 Phase transitions

A phase transition is the process in which a substance changes from one state of matter to another by absorbing or releasing energy. Such transformations take place between solid, liquid, and gaseous phases. The phase transitions can be either reversible (e.g., the monoclinic phase to the tetragonal rutile phase in VO₂ [64]) or irreversible (e.g., the amorphous phase to ordered crystalline phase in TiO₂ [65]), which have been investigated by the stroboscopic or single-shot mode of 4D-EM with high spatiotemporal resolution. In principle, the melting phenomena described in Section 3.1 are also types of phase transitions, namely solid–liquid transitions. Due to the typically low signal-to-noise ratio in single-shot imaging, phase transitions involving gas species are difficult to probe. Hence, this work demonstrates single-shot investigations specifically on irreversible solid-state phase transformations. Figures 3(a)−(d) show the phase transition of TiO₂ nanofilms with the thickness of ~88 nm from the amorphous state to the crystalline phase [65]. The sample was heated by a pump laser pulse at 355 nm (3.5 eV, exceeding the bandgap of TiO₂ by 0.3 eV) with a fluence of 120 mJ·cm⁻². The initial nanofilm presents the amorphous feature [Fig. 3(a)], whereas it becomes the crystalline state (identified by the diffraction peaks) after the single-pulse laser irradiation [Fig. 3(b)]. The transient diffraction patterns of the nanofilm before, during (1500 ns) and after the laser irradiation are displayed in Fig. 3(c). The rings in the diffraction patterns indicate that the TiO₂ nanofilm transited from the initial amorphous state to the polycrystalline phase. To quantify the crystallization process, a linear combination of the diffraction profiles before and after the transition were utilized to fit the transient-frame diffraction pattern. For example, the 1500 ns profile corresponded to the contribution from the state before irradiation (45.3% proportion) and the one in the final stage (54.7% proportion). As such, the crystallinity at 1500 ns was defined to be 0.547. The same procedure was conducted to calculate the crystallinity at different time points [Fig. 3(d)]. It exhibits two plateaus representing the degree of crystallization, which is found to vary from 0 to 0.9. The plateau period of more than 1 μs implies the presence of an intermediate structure that was the precursor of the ordered crystal state. In the two-step dynamics, the existence of the intermediate plateau was necessary for the nucleation process when the latent heat and entropy were lowered.

Another example of the phase transition in the amorphous silicon is presented in Figs. 3(e) and (f) [66]. The 532 nm laser with a fluence of ~100 mJ·cm⁻² was used to excite the samples for melting. The melting temperature of the crystalline phase is 1687 K, which is ~200 K higher than that of the amorphous phase [67, 68]. The time-dependent diffraction patterns [Fig. 3(e)], from the broad halo rings to the sharp rings) revealed the phase transformation from the initial amorphous state to the crystalline one. The diffraction rings were indexed to the (111), (022), (113) crystal planes of Si. The rings became sharper with time, which reflected the progressive growth of the crystalline phase. Using the same data analysis procedure, the crystallinity as a function of time is shown in Fig. 3(f). The plateau represents the intermediate state during the crystallization process, whereas the averaged time constant of 880 ± 140 ns suggests the characteristic time scale for the final transformation. Owing to the low thermal conductivity of amorphous silicon [67], no significant heat diffusion occurred within the irradiated volume on the microsecond time scale. Once the crystallization commenced, the heat transport became faster because of the increase in the thermal conductivity of the as-formed crystalline phase. In this way, the explosive crystallization propagated out to create the larger area of the polycrystalline silicon, allowing for the observed fast change in crystallinity. Finally, the enhanced heat transport out of the hot irradiated center resulted in a flattening of the heat profile, where the modified areas fell below the threshold for further crystallization within a few microseconds.

3.3 Jumping behavior

Different from the laser-induced melting or phase transitions in the thin-film samples, the nanoparticles (NPs) normally present the jumping phenomena when they undergo the ultrafast melting on the supporting substrates (e.g., carbon or Si₃N₄ films). The jumping processes have been closely associated to a wide range of fundamentally important phenomena, including the dewetting of liquid films, heat transfer, droplet transport and self-cleaning behavior [69−72]. Figure 4(a) shows the schematic diagram of the jumping behavior for Au NPs sandwiched between two graphene films [73]. After the illumination by a short laser pulse, the initial Au nanoprisms lose the primary outline, collapse into the contracted nanodroplets, and finally jump off the graphene surface. The time-dependent TEM images of the morphological evolution and associated jumping process of the nanoprisms are displayed in Fig. 4(b). The top-row micrographs represent the transient states of the nanoprisms imaged by only a single probe pulse, while the bottom-row ones are the ensemble dynamics by averaging more than 500 individual exposures to improve the signal-to-noise through the phase correlation-based image registration method [74]. These images reveal that after flash-melting, the nanoprism immediately began to collapse onto its center of mass. Within 2 ns, the nanoprism sides retracted inward, transiently adopting a star-shaped configuration. As the vertices continued to be pulled inward, the molten droplet gradually developed into a more triangular outline, which approached a circular shape at ~7 ns. Based on these transient observations, the motion of Au NPs and their associated parameters such as the velocity and angular distribution could be analyzed. Owing to the smallest radius of curvature and the corresponding highest Laplace pressure at the vertices, they retracted fastest with a speed of ~80 m/s, in contrast to that of ~40 m/s for the nanoprism sides. It is noteworthy that the nanoprisms did not jump straight up but instead possessed the finite in-plane velocities. After detaching from the graphene substrate, the NPs increasingly spread out without any obvious angular preference [Fig. 4(c)]. Figure 4(d) shows the time-dependent displacements of the NPs from the center of the nanoprisms. An average in-plane velocity of the detached NPs was obtained to be ~9 m/s, with some NPs reaching more than ~25 m/s. Such variations of the in-plane velocities suggest the substantial asymmetry of the dewetting process, which were distinct from the fully symmetric contraction from the NPs that jumped straight up [75, 76].

With the advancement of single-shot multiframe imaging technique, a full tracking of the NPs trajectories becomes feasible at the nanosecond-nanometer spatiotemporal resolution. Figure 4(e) shows a series of snapshots representing the full evolution picture of Au NPs upon a heating laser pulse [58]. After the pump laser pulse exceeding 36 mJ·cm⁻², the NPs began to melt, and the adjacent ones were merged into the nanospheres within ~80 ns to lower the surface energies. The merged NP-A obtained a certain velocity and almost moved straightly in the image plane. It collided with another NP-B at ~320 ns, then separated and finally stopped moving at ~480 ns. With the developed algorithm correction for frame-by-frame distortions, the entire motion trajectory of NP-A was quantitatively analyzed. The NP-A jumped away from the original position along the Y direction with a speed of ~1.2 nm/ns. From the period of 320 to 400 ns, the NP-A collided with NP-B, and its velocity was slightly decreased along the Y direction while that of the X direction was increased from 0.1 to 1 nm/ns. The dewetting/coalescence-induced jumping mechanism would account for the above observed phenomena. Upon melting the NPs with a laser pulse, the dewetting of such molten droplets occurred and the surface deformation energy transformed to the kinetic energy, leading to the detachment from the supporting substrate. When multiple NPs were close to each other, they were coalesced to reduce the surface energy. Part of the excess surface energy was converted to the NP kinetic energy, resulting in the subsequent detachment and jumping. Note that the balance between the released surface energy and the dissipative energy (e.g., adhesion-induced dissipation and viscous dissipation) would govern the jumping behavior of the NPs. For instance, when the released surface energy was fully transformed to the dissipative energy (no additional kinetic energy), the merged NPs (NP-B and NP-C) stayed on the substrate surface. It should be pointed out that other mechanisms like impulsive laser-induced stress can also induce the jumping behavior even though the laser fluence is below the threshold for melting the nanostructures [77, 78].

3.4 Reaction dynamics

In a typical reaction, many elementary processes take place on time scales spanning from femtoseconds to seconds or even longer. Offering nanometer spatial and nanosecond temporal resolution, the single-shot mode in 4D-EM enables the capability for in-situ detection of the irreversible reactions in both real and reciprocal space. Figure 5(a) shows the single-shot images of the Cu(TCNQ) crystal under the illumination of a short laser pulse with the wavelength of 671 nm [54]. The initial crystal before the laser excitation displays the needle-like morphology. The single-shot imaging experiment (single pulse excitation) was conducted to reveal the time scale needed for the redox reaction from Cu(TCNQ) to Cu nanocrystals (NCs). The morphological change occurred and were complete within 100 ns since no additional change was observed in the “after” image. It should be pointed out that no observable formation of Cu⁰ NCs was detected under the condition of one-pulse excitation, regardless of the time at which the morphological change was probed (e.g., longer time at 1 ms).

When two laser pulses were used for the excitation, the Cu(TCNQ) NCs initially melted in less than 10 μs as indicated by the emergence of the liquid-like phase and nanobubbles [Fig. 5(b)]. Note that small dark features appeared in the transient snapshot at around 100 μs, which were distinct from that observed under the one-pulse excitation. Detailed analyses were performed to identify the nature of such small dark features. As presented by the magnified image and selected area diffraction in Fig. 5(c), these dark features were determined to be the Cu⁰ NCs. Upon comparison with the phenomena under the one-pulse and two-pulse excitation, it indicates that the reaction reached maturation after the two-pulse illumination and the early morphological change was followed by a slower restructuring to the formation of Cu⁰ NCs. The reduction of Cu⁺ involved the charge transfer that might occur from the electronically excited TCNQ anion-radical to Cu⁺ because of the lower oxidation potential of the molecular excited state [79]. Following the Cu⁺ reduction, the growth of Cu⁰ NCs was then realized through the Ostwald ripening with the assistance of residual thermal energy from the heating laser pulses.

The diffraction and electron energy loss spectroscopy (EELS) techniques with the single-shot mode can also be utilized to investigate the irreversible reactions [80, 81]. Figure 5(d) demonstrates an example of exploring the redox reaction in NiO upon the excitation by a single infrared (IR) laser pulse [81]. The height of the oxygen edge relative to the Ni edge in EELS provides the degree of such reduction. It is seen that the intensity of the oxygen edge was considerably decreased between 1 μs and 4 μs. The timescale of the reaction was further verified through the single-shot diffraction experiments [Fig. 5(e)]. After exposing to the single IR laser pulse, two distinguishable (220) and (331) peaks of NiO became weak at 1 μs and were almost vanished at 4 μs. Note that the peaks of Ni, e.g., the characteristic reflections from (220) and (311), were not obvious until ~10 μs, suggesting that the initial as-formed Ni was liquid. Such diffraction profile did not evolve after 100 μs, which indicates the complete transformation to solid Ni. Figure 5(f) shows a series of single-shot images of the sample in real space. No obvious contrast from the NCs was visible for delays less than 1 μs due to the limited resolution at low magnification. The larger Ni NCs occurred at around 1.5 μs, however, no more significant changes of the image contrast and morphology were observed after 3 μs.

Upon comparison, it is obtained that the disappearance of the oxygen edge in EELS corresponds well with the timescale for the vanishing of the NiO reflections in diffraction and the gradual occurrence of the Ni NCs in the single-shot images. However, it should be pointed out that the diffraction peaks from the crystalline Ni appeared after the sample sufficiently cooled, starting with a broad peak at ~10 μs until the diffraction profile of Ni reached the final shape at 100 μs. The reaction to the formation of Ni was caused by the heating of the carbon film (from TEM grid) due to the IR absorption since the absorption of NiO in the IR of 1064 nm was very low. The temperature of the system after the laser heating pulse was around 2000−2100 K, which was between the melting temperatures of Ni (1728 K) and NiO (2257 K). Due to the initially very high temperature gradient between the carbon film and the NiO layer, heat conduction was very efficient, leading to a temperature rise of almost 2000 K in the NiO layer, triggering the subsequent reactions.

When multiple components are present, the eutectic-related reactions can occur at a temperature below the melting point of each individual component, which have been recently realized by the direct visualization through the single-shot imaging/diffraction in 4D-EM. Figure 6(a) shows the eutectic-related reaction process in the as-grown Au-GaAs nanowire (NW) [82]. The NW (wurtzite structure) was free-standing on the GaAs substrate, which was free from environmental contamination or disturbance that was convenient for statistical investigations. The initial GaAs NW before the laser irradiation has a length of 822 nm. When the NW was illuminated by a laser pulse of 5.5 mJ·cm⁻², the length was shortened to be 797 nm. The NW continued its length reduction after the excitation with more laser pulses, e.g., ~40% and ~70% reduction after 23 and 46 pulses, respectively. Further analyses on the volume change revealed that although the volume of the GaAs NW decreased significantly, the total volume presented no obvious change because of the increment of the top Au bead. Such behavior was understood according to the simulations of the temperature rise as well as the phase identification. The temperatures of the top Au and the GaAs body were calculated to be 566 and 574 K, which were lower than the melting point of either Au (1337 K) or GaAs (1511 K). Nevertheless, it still triggered the reaction to form the Au₇Ga₂ phase identified by the selected-area diffraction [82]. As the fluence was increased to 19.5 mJ·cm⁻², the effective temperature became 1196 K, which was high enough to facilitate the eutectic reactions for the formation of new AuGa and AuGa₂ phases.

The cooling dynamics (diffraction intensity as a function of time) were conducted to estimate the time scale that could guide the investigation of the transient evolution of eutectic-related process [Fig. 6(b)]. The time constant of 123 ± 12 ns was extracted from fitting the dynamics. Guided by such time scale, the single-shot imaging at that range was utilized to elucidate the eutectic-related phase reactions in the NWs. Three columns of images represent the sample states before, at specific delays (20−100 ns), and after the process ended, respectively. After a single pulse excitation with the fluence of 19.5 mJ·cm⁻², the shrinkage of the GaAs body as well as a distinct shape change of the top bead was observed at 20 ns. Note that the bead continued its morphological change even after the incident laser pulse was removed (see the comparison between the “20 ns” and the “after” images). At longer delay time of 80 ns, similar behaviors were also observed. However, when the delay time was increased to 100 ns, the length of the NW and the shape of the bead nearly remained unchanged, indicating that the thermal energy induced by the heating laser pulse was insufficient to propel additional changes after ~100 ns. Using the changes of the length and volume at specific delays together with the optical characteristics of Au and GaAs, the useful thermal parameters including the latent heat and specific heat for the newly formed phases could be thereby retrieved. The as-obtained latent heats of Au₇Ga₂ and AuGa were 8 and 21 kJ/mol, and their specific heats were 62 and 41 J/(mol·K), respectively, which were not easily accessible in bulk counterparts.

Besides the hexagonal wurtzite structure, GaAs also possesses another structure with the typical “ABC” stacking sequence, i.e., the cubic zinc-blende (ZB) one. Fig. 6(c) exhibits the eutectic-related phase reactions in the ZB GaAs-based NW [83]. Upon exposure to a single laser pulse with the fluence of 12 mJ·cm⁻², the NW demonstrated a transient increase of ~45 nm in length at 20 ns, although the NW length was shortened when the process had ended. The similar behavior also occurred for the NW at the delay time of 100 ns, despite a larger transient increase in length (~60 nm). It is noteworthy that such phenomenon was opposite to that observed in the wurtzite NWs. As the time was increased to 150 ns, no obvious increase of the NW length was observed. It suggests that the length of ZB-GaAs NW presented a transient increase after illumination of an incoming heating laser pulse and completed this transformation within ~150 ns. The distinct behaviors between the ZB and wurtzite NWs might be attributed to the difference of surface diffusion caused by the crystalline structures. As known, mass transport on the NW surface is usually related to lower activation energy, and is thereby faster than diffusion in the interior of the NW [84]. Owing to the low diffusion barrier on surface, the phase reactions might also occur on the NW surface except the Au/GaAs interface. Therefore, when the temperature exceeded the phase-reaction point, the liquid-phase species diffused toward a certain distance along the NW as a result of the fast surface transport. The finding indicates that the ZB structure promoted the surface diffusion, leading to the transient expansion and subsequent length increase of the NW.

3.5 Liquid associated dynamics

Many chemical, physical and biological processes occur in liquid environment, such as solution-based synthesis, catalysis, crystal growth, NP motion and self-assembly, and physiological activities [85−92]. Essentially, most of these processes are irreversible, which can serve as the prototypes studied by single-shot mode of 4D-EM. Figure 7(a) shows the motion behavior of the liquid lead in a ZnO nanotube under irradiation with a train of picosecond laser pulses [93]. A lead NP of ~150 nm in diameter was formed when the molten lead leaked through the imperfection region (nanochannel) on the nanotube wall. Such processes were vividly captured in the snapshots at different times. Briefly, the laser pulse heated the solid lead core and transformed it into a hot liquid, thereby triggering its rapid expansion and initiating the subsequent dynamics. At a delay time of 96 ns, a nearly spherical lead droplet erupted on the outside of the nanotube. The extrusion shrunk to a diameter of ~110 nm after 10 s due to the slow reabsorption through the nanotube. It further contracted to a diameter of ~50 nm at 110 s and finally disappeared at 210 s. The movement of the lead nanodroplet was the result of the initial rapid expansion upon the heating laser pulse and the slower contraction upon cooling. Note that some voids were formed in the nanotube during the cooling process, which was energetically favorable due to the reduction of the total surface energy.

Time-resolved diffraction experiments were performed to investigate the temperature jump and cooling rates, offering the time window for single-shot imaging of the fluid-associated flow dynamics. Figure 7(b) presents the selected-area diffraction pattern of the nanotube. Some diffraction spots from the lead core, e.g., the (200) ones, were visible except the reflections of the graphite substrate. Their intensity after the laser excitation was measured as a function of time by recording the stroboscopic diffraction patterns [Fig. 7(c)]. It is seen that the diffraction intensity initially dropped sharply and then recovered exponentially with a time constant of 226 ns. Such behavior was caused by the laser-induced temperature jump, which could be calibrated through the measurement of the diffraction intensity as a function of the sample temperature in the absence of the laser irradiation [Fig. 7(d)]. By extrapolating to high laser fluence together with taking the nanotube dimensions into account, the initial temperature of the system in Fig. 7(a) was estimated to greater than 1000 K, well above the melting point of lead. As such, the extracted parameters including the time constant and laser fluence would be utilized as the guidance for studying various droplet-associated behaviors.

Figure 7(e) presents three irreversible flow phenomena: liquid shooting, fission, and rupture. In the shooting case, a nanotube with a 60-nm-diameter open tip was heated by a laser pulse; the molten lead core expanded into a 110-nm droplet at the tip at 29 ns, which later shrank upon cooling (rightmost difference image between the “29 ns” and “after” ones). In the fission case, a 250-nm-long lead column inside a 55-nm-diameter nanotube split into two segments at 34 ns, moving apart at ~1.5 and 2 m/s, while a 40-nm extrusion appeared and erupted from the tube side. Upon cooling, the segments reconnected into a shorter 210-nm column due to material loss from eruption. Under higher-fluence multiple pulses, laser-induced pressure ruptured the tube wall at defect sites, causing the lead segments to deform at 18 ns and extrusions to explode into numerous smaller NPs, with wall rupture visible in the magnified rightmost image. These examples highlight that direct imaging of the liquid flow with a time resolution sufficient to capture the fast processes occurring far from equilibrium would provide useful insights into fluid dynamics at single-particle level.

Liquid cells, when integrated in TEM, offers versatile opportunities for visualizing nanoscale liquid-phase dynamics with high spatial resolution [94−101]. Developing the technology of combining both the 4D-EM and liquid cells has further extended the in-situ detection capabilities with high spatiotemporal resolutions. Figure 8 shows the translational and rotational dynamics of photon-activated Au NPs in water by liquid-cell 4D-EM [102, 103]. For the liquid cell preparation, the solution containing spherical Au NPs (~80 nm in diameter) was sealed between two electron-transparent silicon nitride membranes (20 nm in thickness) with a liquid thickness of ~300 nm [Fig. 8(a)]. At the long-time scale, the translational dynamics of a single photon-activated Au NP was traced by continuous electron beam imaging mode. Figure 8(b) shows a set of typical snapshots of the NP translational motion under repetitive (1 kHz) femtosecond laser pulse excitation at the fluence of 2.3 mJ·cm⁻². Upon excitation, the NP was heated up in hundreds of picoseconds due to the strong local photothermal effect through the surface plasmon–enhanced optical absorption of Au at the laser wavelength of 520 nm. The temperature was raised rapidly over the boiling point of the water molecules that evaporated into steam. Such water steam nucleated as nanobubbles on the NP surface, as seen in the snapshots. Consequently, the NP was activated to move randomly, which was driven by the rapid expansion, detachment and collapse of the nanobubbles around the particle surface. At the short-time scale, the transient translational dynamics of the Au NPs were further studied by using the single-pulse imaging mode of 4D-EM [Fig. 8(c)]. The first and second columns show the typical single-pulse images of the NP before the laser pulse and at the specific delay of 60 ns, respectively. The corresponding difference image is shown in the third column to highlight the contrast (the red and blue colors indicate the initial position and the position at the specific delay, respectively). Apparently, the displacements of the NP could reach several tens to hundreds of nanometers at the timescale of nanoseconds.

To understand the translational dynamics of photon-activated Au NPs, further analyses were conducted by extracting the diffusion constants from the mean square displacements (MSDs). The MSDs show a nearly linear time dependence (MSD\propto Dt), confirming Brownian diffusive behavior. As shown in Fig. 8(d), the diffusion constant D follows a power-law relation with laser fluence: D\propto {\left(J-J_c\right)}^{2.2}, where J_c = 1.25 mJ·cm⁻² is the threshold for water boiling and evaporation. The measured D values are four to five orders of magnitude larger than that of the conventional Brownian motion of Au NPs without photon activation [104], indicating superfast diffusion. The time-resolved displacement of the NP [Fig. 8(e)] reveals three stages: a slow increase within the first ~25 ns, a rapid rise from 25 to 60 ns, and a gradual saturation after ~80 ns. This implies that following femtosecond laser excitation, the NP gained momentum within 20–30 ns — the timescale for photoinduced steam and nanobubble generation. In the nanosecond range, the NP exhibited ballistic Brownian motion at ~6 m/s, three orders of magnitude faster than passive Brownian motion. Notably, although the short-time ballistic nature is similar in both cases, the photon-activated motion is driven to much higher speeds by the impulsive force from steam nanobubbles.

In addition to the translational motion, the Au NPs can also move in a rotational way under the impulsive laser excitation. Figure 8(f) shows the transient rotational morphologies of a gold NP dimer induced by a single femtosecond laser pulse at the fluence of 10 mJ·cm⁻² [103]. At different delay times, the dimer changed its orientation by specific angles with respect to the initial state (the red-blue arrow pairs in the first and third columns indicated the initial and final positions of the dimer). For convenient comparison, the orientations of the dimer before the laser excitation (black dashed lines), at specific delays (solid blue lines), and after the process ended (solid pink lines) are displayed in the fourth column. The rotation angles increased with the time delays, namely, 2° and 29° at 20 and 150 ns, respectively. Interestingly, the rotational behavior of the dimers could be modulated by tuning the morphological asymmetry of the dimers. The typical snapshots of a dimer (the two NPs had similar diameters of 60 and 67 nm with a ratio of 1:1.1) were recorded as a function of the femtosecond laser illumination times [Fig. 8(g)]. Upon the laser irradiation, the dimer shows distinct orientations (indicated by the red-blue arrow pairs) at different times. The rotation direction could even reverse at certain times (pink arc arrows), featuring the the manner of “random walk”. When the ratio was further increased (larger morphological asymmetry), a full picture depicting the transition from conventional diffusive rotation to a ballistic one could be realized, as detailed in the following section.

The time-dependent rotation angle of the dimer is presented in Fig. 8(h) to understand the rotational behavior, alongside theoretical modeling results. The angle shows a modest increase to 2° over the first 10−20 ns, followed by a rapid rise to 17° at 42 ns, and thereafter a slow approach to saturation. This behavior suggests that angular momentum was imparted to the dimer within ~20 ns after femtosecond laser excitation, with subsequent hindrance and damping over hundreds of nanoseconds due to environmental drag. The asymmetry of the dimer gives rise to nonuniform laser heating, leading to rapid nanobubble nucleation, expansion, and collapse near the NP surface, which can exert an impulsive force and torque that trigger the observed motion. The relevant parameters were extracted by fitting to a stochastic differential equation [103]; the reconstructed impulsive torque is plotted as the red dashed curve, showing a full width at half maximum (FWHM) of ~14 ns and a peak of 4.6 × 10³ nN·nm at ~11 ns (inset). This torque duration is consistent with the lifetime of femtosecond pulse-induced steam nanobubbles in water, as measured by optical scattering [105, 106]. In addition, the effective damping constant — originating from drag contributions of both liquid friction and substrate interaction — was determined from the fit to be 0.03 ns⁻¹.

The transition from conventional diffusive rotation to the ballistic behavior was analyzed through the statistical properties of the rotation angle \theta (t) by extracting the dimer trajectories at different starting times {\tau }_0 and ending times t. The angular displacement for each trajectory is defined as \mathrm{\Delta }\theta \left(t\right)=\theta \left(t+{\tau }_0\right)-\theta \left({\tau }_0\right). Figure 8(i) shows the mean square angular displacements (MSADs) of three kinds of dimers with different morphological asymmetries [103]. The MSADs displays the linear increment in the log−log scale plot with a power law of , where the exponent \alpha increases with the factor of the morphological asymmetry. The retrieved \alpha increases from 1.11 for the D1 dimer (ratio of 1:1.1), to 1.51 for the D2 dimer (ratio of 1:1.3), and further to 1.95 for the D3 dimer (ratio of 1:1.5). The \alpha \approx 1 for the D1 dimer indicates the conventional diffusive behavior. Nevertheless, the extracted rotational diffusion coefficient (4.68 rad²/s) of D1 is nearly four orders of magnitude greater than that of the colloidal nanorods without laser illumination [107, 108]. For D2, the exponent 1 < \alpha < 2 indicates superdiffusive rotation, implying that the impulsive torque preferentially acts in one direction. When the asymmetric factor is further increased, e.g., \alpha \approx 2 for the D3 dimer, the rotation is transited into a ballistic one, indicating that the photoinduced nanobubbles occur at a nearly identical position on the dimer and give rise to the unidirectional impulsive torque. Collectively, these liquid-cell 4D-EM observations reveal a photoinduced nanobubble propulsion mechanism, offering fundamental insight into the design and control of light-driven artificial micro/nanomotors.

4 Perspectives and conclusions

4D-EM, after nearly two decades of rapid development, has matured into a transformative tool to “direct” and “record” real-time structural movies of matter under photoexcitation with atomic-scale resolution. Yet, the vast majority of such experiments to date have relied on the stroboscopic pump–probe scheme [109], where a reversible process is accumulated over countless repetitions to build a movie. For studying irreversible or dose-sensitive events — such as material fracture, chemical reactions in solution, phase transitions under extreme conditions, or the conformational changes of a single biomolecule — this accumulation strategy becomes fundamentally unsuitable. It is here that the single-shot technologies emerge for realizing the full potential of 4D-EM to film a process as it happens, once and never again. Despite being less developed compared to the stroboscopic mode, the single-shot 4D-EM has also demonstrated unique capabilities for revealing irreversible processes, as reviewed in this article including the melting and jumping behaviors, phase transitions, reaction dynamics, and liquid-associated dynamics.

For further developing the single-shot 4D-EM, the central challenge is the uncompromising battle between electron number, pulse duration, and beam coherence. To form a high-quality diffraction pattern or real-space image in a single pulse, one must deliver enough electrons within a short-time window. Coulomb repulsion, however, severely broadens the pulse duration and destroys the spatiotemporal resolution as the electron density increases. Radio-frequency compression and Terahertz-driven pulse compression techniques have been adapted to time-focus electron bunches for achieving much higher temporal resolutions, which would be applicable to further promise the single-shot capability. In parallel, the development of ultrabright electron sources with high-coherence bunches could directly addresses the trade-off by providing higher charge densities at the cathode with lower intrinsic energy spread. Looking ahead, the single-shot technology may ultimately merge with the femtosecond/attosecond frontier. While attosecond single-shot electron pulses remain extraordinarily challenging, the combination of laser-driven nanotip sources with sophisticated bunch compression may enable one to record the response of valence electrons to an intense optical field in a single event, mapping light-induced electron density redistribution with sub-cycle resolution. In such a regime, single-shot 4D-EM would directly visualize the birth and death of transient states — virtual charge separation, correlation-driven band-gap collapse, or the instantaneous screening dynamics that follow photoexcitation — without requiring the process to repeat identically. For further extending the capability, it places necessary demands on detectors to improve the capture issue, in which direct electron detectors with single-electron sensitivity and radiation tolerance would be capable of recording an entire image or diffraction pattern from a single electron pulse with high dynamic range, low noise, and high speed. All these combined improvements would effectively realize the long-standing dream of making a “molecular movie” not just of the average atomic motion but of the genuine, non-repeating quantum trajectories of matter.

On the application frontier, 4D-EM will inevitably move toward in-situ and operando ultrafast characterization. By combining liquid/gas microfluidic cells with ultrafast pumping, it may be possible to atomically resolve ion migration, valence state changes, and defect dynamics under realistic catalytic reactions, nucleation and growth of a single NP from a liquid precursor, battery charge–discharge cycles, and the operating conditions of topological spin devices, natively along the evolving coordinates. For quantum materials, single-shot imaging of fluctuating order parameters — such as transient charge-density-wave domains or skyrmion nucleation — will reveal the true stochastic nature of nonequilibrium phase transitions. In the life sciences, the emergence of cryogenic 4D-EM holds promise for delivering near-atomic-resolution movies of light-triggered conformational changes in radiation-sensitive biological macromolecules. When performed in the single-shot mode, it could freeze a protein’s conformation at a defined instant following a laser-triggered release of a ligand, bypassing the fatal dose limit by distributing the signal over many particles but capturing each individual micrograph in a single exposure. Although the superimposed challenges of laser-induced heating and electron dose limitations should be overcome, a breakthrough along this path would profoundly transform the understanding of dynamic processes in structural biology.

In all, the pursuit of single-shot capability epitomizes the broader applications in ultrafast electron microscopy: from imaging averaged, repeatable dynamics to capturing the singular, the stochastic, and the irreversible processes. Although challenges regarding electron number, space charge, beam coherence, and sample damage remain, accompanied by a new wave of innovation in pulse compression, bright electron sources, fast detectors, and intelligent algorithms, single-shot 4D-EM is poised within the next decade to explore the frontier of non-equilibrium science in interdisciplinary fields, where the interesting phenomena happen only once. In this sense, the single-shot electron pulse is not merely a technical achievement; it is the key that unlocks the true, unrepeatable dynamics of the physical world, which is also expected to become a platform for actively designing and manipulating quantum states of matter.

References

[1] B. Dasgupta , D. Bagchi , T. Sontheimer , M. Driess , and P. W. Menezes , Dynamics in electrochemical organic oxidation reactions from in situ and operando techniques, Nat. Rev. Chem. 9(11), 766 (2025)

[2] W. L. Mao , L. Wang , Y. Ding , W. Yang , W. Liu , D. Y. Kim , W. Luo , R. Ahuja , Y. Meng , S. Sinogeikin , J. Shu , and H. Mao , Distortions and stabilization of simple-cubic calcium at high pressure and low temperature, Proc. Natl. Acad. Sci. USA 107(22), 9965 (2010)

[3] G. R. Heath and S. Scheuring , High-speed AFM height spectroscopy reveals μs-dynamics of unlabeled biomolecules, Nat. Commun. 9(1), 4983 (2018)

[4] N. Chada , K. Chattrakun , B. P. Marsh , C. Mao , P. Bariya , and G. M. King , Single-molecule observation of nucleotide induced conformational changes in basal SecA-ATP hydrolysis, Sci. Adv. 4(10), eaat8797 (2018)

[5] H. Ye , C. Wu , D. Cao , G. Zhang , Y. Sun , Y. Zhu , F. Wu , Z. Zhang , and R. Wang , Atomically resolved two-dimensional amorphous nuclei formed during MoS₂ chemical vapor deposition, Science 391(6785), 622 (2026)

[6] L. Zhong , Y. Zhang , X. Wang , T. Zhu , and S. X. Mao , Atomic-scale observation of nucleation- and growth-controlled deformation twinning in body-centered cubic nanocrystals, Nat. Commun. 15(1), 560 (2024)

[7] Y. Lee , H. Ki , D. Im , S. Eom , J. Gu , S. Lee , J. Kim , Y. Cha , K. W. Lee , S. Zerdane , M. Levantino , and H. Ihee , Cerium photocatalyst in action: Structural dynamics in the presence of substrate visualized via time-resolved X-ray liquidography, J. Am. Chem. Soc. 145(43), 23715 (2023)

[8] S. H. Lee , J. E. Avilés Acosta , D. Lee , D. M. Larson , H. Li , J. Chen , J. Lee , E. Erdem , D. U. Lee , S. J. Blair , A. Gallo , H. Zheng , A. C. Nielander , C. J. Tassone , T. F. Jaramillo , and W. S. Drisdell , Structural transformation and degradation of Cu oxide nanocatalysts during electrochemical CO₂ reduction, J. Am. Chem. Soc. 147(8), 6536 (2025)

[9] H. Zhu , K. Miyata , Y. Fu , J. Wang , P. P. Joshi , D. Niesner , K. W. Williams , S. Jin , and X. Y. Zhu , Screening in crystalline liquids protects energetic carriers in hybrid perovskites, Science 353(6306), 1409 (2016)

[10] J. Nishida , J. P. Breen , K. P. Lindquist , D. Umeyama , H. I. Karunadasa , and M. D. Fayer , Dynamically disordered lattice in a layered Pb-I-SCN perovskite thin film probed by two-dimensional infrared spectroscopy, J. Am. Chem. Soc. 140(31), 9882 (2018)

[11] C. Li , J. Yang , F. Su , J. Tan , Y. Luo , and S. Ye , Conformational disorder of organic cations tunes the charge carrier mobility in two-dimensional organic−inorganic perovskites, Nat. Commun. 11(1), 5481 (2020)

[12] C. Zhao , W. Tian , Q. Sun , Z. Yin , J. Leng , S. Wang , J. Liu , K. Wu , and S. Jin , Trap-enabled long-distance carrier transport in perovskite quantum wells, J. Am. Chem. Soc. 142(35), 15091 (2020)

[13] W. Liu , H. Yu , Y. Li , A. Hu , J. Wang , G. Lu , X. Li , H. Yang , L. Dai , S. Wang , and Q. Gong , Mapping trap dynamics in a CsPbBr₃ single-crystal microplate by ultrafast photoemission electron microscopy, Nano Lett. 21(7), 2932 (2021)

[14] S. Yue , F. Tian , X. Sui , M. Mohebinia , X. Wu , T. Tong , Z. Wang , B. Wu , Q. Zhang , Z. Ren , J. Bao , and X. Liu , High ambipolar mobility in cubic boron arsenide revealed by transient reflectivity microscopy, Science 377(6604), 433 (2022)

[15] R. Chen , Z. Ren , Y. Liang , G. Zhang , T. Dittrich , R. Liu , Y. Liu , Y. Zhao , S. Pang , H. An , C. Ni , P. Zhou , K. Han , F. Fan , and C. Li , Spatiotemporal imaging of charge transfer in photocatalyst particles, Nature 610(7931), 296 (2022)

[16] X. Lin , Y. Han , J. Zhu , and K. Wu , Room-temperature coherent optical manipulation of hole spins in solution-grown perovskite quantum dots, Nat. Nanotechnol. 18(2), 124 (2023)

[17] B. H. Li , H. Di , H. Li , J. C. Wang , W. Zeng , D. B. Cheng , C. Zhou , X. Wang , Y. Shi , J. Song , Y. Zhao , X. Yang , and Z. Ren , Unveiling the intrinsic photophysics in quasi-two-dimensional perovskites, J. Am. Chem. Soc. 146(10), 6974 (2024)

[18] Y. Ouyang , R. Wang , X. Wang , M. Xiao , and C. Zhang , Ultrafast energy transfer beyond the Förster approximation in organic photovoltaic blends with nonfullerene acceptors, Sci. Adv. 11(12), eadr5973 (2025)

[19] A. H. Zewail , Four-dimensional electron microscopy, Science 328(5975), 187 (2010)

[20] A. H. Zewail and J. M. Thomas, 4D Electron Microscopy: Imaging in Space and Time, Imperial College Press, London, 2009

[21] K. E. Priebe , C. Rathje , S. V. Yalunin , T. Hohage , A. Feist , S. Schäfer , and C. Ropers , Attosecond electron pulse trains and quantum state reconstruction in ultrafast transmission electron microscopy, Nat. Photonics 11(12), 793 (2017)

[22] D. Nabben , J. Kuttruff , L. Stolz , A. Ryabov , and P. Baum , Attosecond electron microscopy of sub-cycle optical dynamics, Nature 619(7968), 63 (2023)

[23] V. A. Lobastov , J. Weissenrieder , J. Tang , and A. H. Zewail , Ultrafast electron microscopy (UEM): Four-dimensional imaging and diffraction of nanostructures during phase transitions, Nano Lett. 7(9), 2552 (2007)

[24] R. M. van der Veen , O. Kwon , A. Tissot , A. Hauser , and A. H. Zewail , Single-nanoparticle phase transitions visualized by four-dimensional electron microscopy, Nat. Chem. 5(5), 395 (2013)

[25] K. Sun , S. S. Sun , C. H. Zhu , H. F. Tian , H. X. Yang , and J. Q. Li , Hidden CDW states and insulator-to-metal transition after a pulsed femtosecond laser excitation in layered chalcogenide 1T-TaS_2−xSe_x, Sci. Adv. 4(7), eaas9660 (2018)

[26] O. H. Kwon and A. H. Zewail, , 4D electron tomography, 328(5986), 1668 (2010)

[27] Y. J. Kim , H. Jung , S. W. Han , and O. H. Kwon , Ultrafast Electron Microscopy Visualizes Acoustic Vibrations of Plasmonic Nanorods at the Interfaces, Matter 1(2), 481 (2019)

[28] L. Tong , J. Yuan , Z. Zhang , J. Tang , and Z. Wang , Nanoscale subparticle imaging of vibrational dynamics using dark-field ultrafast transmission electron microscopy, Nat. Nanotechnol. 18(2), 145 (2023)

[29] B. Barwick , H. S. Park , O. H. Kwon , J. S. Baskin , A. H. Zewail , Oh. H. Kwon , J. S. Baskin , and A. H. Zewail, , 4D imaging of transient structures and morphologies in ultrafast electron microscopy, 322(5905), 1227 (2008)

[30] D. R. Cremons , D. A. Plemmons , and D. J. Flannigan , Femtosecond electron imaging of defect-modulated phonon dynamics, Nat. Commun. 7(1), 11230 (2016)

[31] B. Barwick , D. J. Flannigan , and A. H. Zewail , Photon-induced near-field electron microscopy, Nature 462(7275), 902 (2009)

[32] T. T. A. Lummen , R. J. Lamb , G. Berruto , T. LaGrange , L. Dal Negro , F. J. García de Abajo , D. McGrouther , B. Barwick , and F. Carbone , Imaging and controlling plasmonic interference fields at buried interfaces, Nat. Commun. 7(1), 13156 (2016)

[33] K. Wang , R. Dahan , M. Shentcis , Y. Kauffmann , A. Ben Hayun , O. Reinhardt , S. Tsesses , and I. Kaminer , Coherent interaction between free electrons and a photonic cavity, Nature 582(7810), 50 (2020)

[34] D. Hui , H. Alqattan , S. Yamada , V. Pervak , K. Yabana , and M. Th. Hassan , Attosecond electron motion control in dielectric, Nat. Photonics 16(1), 33 (2022)

[35] X. Fu , Z. Sun , S. Ji , F. Liu , M. Feng , B. K. Yoo , and Y. Zhu , Nanoscale-femtosecond imaging of evanescent surface plasmons on silver film by photon-induced near-field electron microscopy, Nano Lett. 22(5), 2009 (2022)

[36] M. Möller , J. H. Gaida , S. Schäfer , and C. Ropers , Few-Nm Tracking of Current-Driven Magnetic Vortex Orbits Using Ultrafast Lorentz Microscopy, Commun. Phys. 3(1), 36 (2020)

[37] G. Cao , S. Jiang , J. Akerman , and J. Weissenrieder , Femtosecond Laser Driven Precessing Magnetic Gratings, Nanoscale 13(6), 3746 (2021)

[38] S. R. Tauchert , M. Volkov , D. Ehberger , D. Kazenwadel , M. Evers , H. Lange , A. Donges , A. Book , W. Kreuzpaintner , U. Nowak , and P. Baum , Polarized phonons carry angular momentum in ultrafast demagnetization, Nature 602(7895), 73 (2022)

[39] C. Liu , Z. Li , X. Hu , J. Guo , Y. Hu , Y. Deng , H. Wang , S. Ji , C. Gao , F. Liu , H. Zhang , W. He , T. Guo , S. Chen , P. H. Lu , J. Wu , Y. Hu , Z. M. Liao , J. M. Liu , D. Yu , Z. Hou , R. E. Dunin-Borkowski , Y. Zhu , and X. Fu , Picosecond-scale coherent toggle switching of topological spin helicity, Nat. Nanotechnol. 21(5), 655 (2026)

[40] A. M. El-Zohry , B. S. Shaheen , V. M. Burlakov , J. Yin , M. N. Hedhili , S. Shikin , B. Ooi , O. M. Bakr , and O. F. Mohammed , Extraordinary carrier diffusion on CdTe surfaces uncovered by 4D electron microscopy, Chem 5(3), 706 (2019)

[41] T. Kim , S. Park , V. Iyer , B. Shaheen , U. Choudhry , Q. Jiang , G. Eichman , R. Gnabasik , K. Kelley , B. Lawrie , K. Zhu , and B. Liao , Mapping the pathways of photo-induced ion migration in organic-inorganic hybrid halide perovskites, Nat. Commun. 14(1), 1846 (2023)

[42] Y. Zhang , X. Chen , Y. Yu , Y. Huang , M. Qiu , F. Liu , M. Feng , C. Gao , S. Deng , and X. Fu, , A femtosecond electron-based versatile microscopy for visualizing carrier dynamics in semiconductors across spatiotemporal and energetic domains, Adv. Sci. (Weinh.) 11(31), 2400633 (2024)

[43] Y. Tian , D. Yang , Y. Ma , Z. Li , J. Li , Z. Deng , H. Tian , H. Yang , S. Sun , and J. Q. Li , Spatiotemporal visualization of photogenerated carriers on an avalanche photodiode surface using ultrafast scanning electron microscopy, Nanomaterials (Basel) 14(3), 310 (2024)

[44] D. J. Flannigan , B. Barwick , and A. H. Zewail , Biological imaging with 4D electron microscopy, Proc. Natl. Acad. Sci. USA 107(22), 9933 (2010)

[45] U. J. Lorenz and A. H. Zewail , Biomechanics of DNA structures visualized by 4D electron microscopy, Proc. Natl. Acad. Sci. USA 110(8), 2822 (2013)

[46] A. W. P. Fitzpatrick , U. J. Lorenz , G. M. Vanacore , and A. H. Zewail, , 4D cryo-electron microscopy of proteins, J. Am. Chem. Soc 135(51), 19123 (2013)

[47] M. Kaplan , B. K. Yoo , J. Tang , T. E. Karam , B. Liao , D. Majumdar , D. Baltimore , G. J. Jensen , and A. H. Zewail , Photon-induced near-field electron microscopy of eukaryotic cells, Angew. Chem. Int. Ed. 56(38), 11498 (2017)

[48] A. Feist , K. E. Echternkamp , J. Schauss , S. V. Yalunin , S. Schäfer , and C. Ropers , Quantum coherent optical phase modulation in an ultrafast transmission electron microscope, Nature 521(7551), 200 (2015)

[49] C. Kealhofer , W. Schneider , D. Ehberger , A. Ryabov , F. Krausz , and P. Baum , All-optical control and metrology of electron pulses, Science 352(6284), 429 (2016)

[50] S. Sun , X. Sun , J. Williams , and C. Y. Ruan , Development of RF-compressed high-throughput femtosecond electron microscope, Microsc. Microanal. 26(S2), 430 (2020)

[51] A. Takaoka and K. Ura , Velocity measuring system for interface motion in TEM, J. Electron Microsc. (Tokyo) 37, 117 (1988)

[52] H. Dömer and O. Bostanjoglo , High-speed transmission electron microscope, Rev. Sci. Instrum. 74(10), 4369 (2003)

[53] J. S. Kim , T. LaGrange , B. W. Reed , M. L. Taheri , M. R. Armstrong , W. E. King , N. D. Browning , and G. H. Campbell , Imaging of transient structures using nanosecond in situ TEM, Science 321(5895), 1472 (2008)

[54] S. T. Park , D. J. Flannigan , and A. H. Zewail , Irreversible chemical reactions visualized in space and time with 4D electron microscopy, J. Am. Chem. Soc. 133(6), 1730 (2011)

[55] R. K. Li and P. Musumeci , Single-shot MeV transmission electron microscopy with picosecond temporal resolution, Phys. Rev. Appl. 2(2), 024003 (2014)

[56] J. Liang and L. V. Wang , Single-shot ultrafast optical imaging, Optica 5(9), 1113 (2018)

[57] K. S. Wilson , Z. S. Walbrun , and C. Y. Wong , Single-shot transient absorption spectroscopy techniques and design principles, Spectrochim. Acta A Mol. Biomol. Spectrosc. 253, 119557 (2021)

[58] Y. Meng , Y. Zhou , X. Wang , W. Wei , Y. Hu , B. Chen , and D. Zhong , Direct nanosecond multiframe imaging of irreversible dynamics in 4D electron microscopy, Nano Lett. 24(24), 7219 (2024)

[59] K. Zweiacker , J. T. McKeown , C. Liu , T. LaGrange , B. W. Reed , G. H. Campbell , and J. M. K. Wiezorek , Determination of crystal growth rates during rapid solidification of polycrystalline aluminum by nano-scale spatio-temporal resolution in situ transmission electron microscopy, J. Appl. Phys. 120(5), 055106 (2016)

[60] J. T. McKeown , K. Zweiacker , C. Liu , D. R. Coughlin , A. J. Clarke , J. K. Baldwin , J. W. Gibbs , J. D. Roehling , S. D. Imhoff , P. J. Gibbs , D. Tourret , J. M. K. Wiezorek , and G. H. Campbell , Time-resolved in Situ measurements during rapid alloy solidification: Experimental insight for additive manufacturing, J. Miner. Met. Mater. Soc. 68(3), 985 (2016)

[61] J. T. McKeown , A. K. Kulovits , C. Liu , K. Zweiacker , B. W. Reed , T. LaGrange , J. M. K. Wiezorek , and G. H. Campbell , In situ transmission electron microscopy of crystal growth-mode transitions during rapid solidification of a hypoeutectic Al–Cu alloy, Acta Mater. 65, 56 (2014)

[62] S. C. Gill and W. Kurz , Rapidly solidified Al Cu alloys – I. Experimental determination of the microstructure selection map, Acta Metall. Mater. 41(12), 3563 (1993)

[63] M. J. Aziz and T. Kaplan , Continuous growth model for interface motion during alloy solidification, Acta Metall. 36(8), 2335 (1988)

[64] M. S. Grinolds , V. A. Lobastov , J. Weissenrieder , and A. H. Zewail , Four-dimensional ultrafast electron microscopy of phase transitions, Proc. Natl. Acad. Sci. USA 103(49), 18427 (2006)

[65] B. K. Yoo , O. H. Kwon , H. Liu , J. Tang , and A. H. Zewail , Observing in space and time the ephemeral nucleation of liquid-to-crystal phase transitions, Nat. Commun. 6(1), 8639 (2015)

[66] O. H. Kwon , B. Barwick , H. S. Park , J. S. Baskin , and A. H. Zewail, , 4D visualization of embryonic, structural crystallization by single-pulse microscopy, Proc. Natl. Acad. Sci. USA 105(25), 8519 (2008)

[67] M. G. Grimaldi , P. Baeri , and M. A. Malvezzi , Melting temperature of unrelaxed amorphous silicon, Phys. Rev. B 44(4), 1546 (1991)

[68] M. O. Thompson , G. J. Galvin , J. W. Mayer , P. S. Peercy , J. M. Poate , D. C. Jacobson , A. G. Cullis , and N. G. Chew , Melting temperature and explosive crystallization of amorphous silicon during pulsed laser irradiation, Phys. Rev. Lett. 52(26), 2360 (1984)

[69] W. Huang , L. Zhao , X. He , Y. Li , C. P. Collier , Z. Zheng , J. Liu , D. P. Briggs , and J. Cheng , Low-temperature Leidenfrost-like jumping of sessile droplets on microstructured surfaces, Nat. Phys. 20(8), 1274 (2024)

[70] C. Ma , L. Chen , L. Wang , W. Tong , C. Chu , Z. Yuan , C. Lv , and Q. Zheng , Condensation droplet sieve, Nat. Commun. 13(1), 5381 (2022)

[71] A. Habenicht , M. Olapinski , F. Burmeister , P. Leiderer , and J. Boneberg , Jumping nanodroplets, Science 309(5743), 2043 (2005)

[72] C. W. E. Lam , K. Regulagadda , M. Donati , A. Tripathy , G. C. Pal , C. S. Sharma , A. Milionis , and D. Poulikakos , Condensate droplet roaming on nanostructured superhydrophobic surfaces, Nat. Commun. 16(1), 1167 (2025)

[73] P. K. Olshin , J. M. Voss , M. Drabbels , and U. J. Lorenz , Real-time observation of jumping and spinning nanodroplets, Struct. Dyn. 7(1), 011101 (2020)

[74] B. S. Reddy and B. N. Chatterji , An FFT-based technique for translation, rotation, and scale-invariant image registration, IEEE Trans. Image Process. 5(8), 1266 (1996)

[75] M. Fuentes-Cabrera , B. H. Rhodes , M. I. Baskes , H. Terrones , J. D. Fowlkes , M. L. Simpson , and P. D. Rack , Controlling the velocity of jumping nanodroplets via their initial shape and temperature, ACS Nano 5(9), 7130 (2011)

[76] S. Afkhami and L. Kondic , Numerical simulation of ejected molten metal nanoparticles liquified by laser irradiation: Interplay of geometry and dewetting, Phys. Rev. Lett. 111(3), 034501 (2013)

[77] Y. Zhou , Y. Meng , G. Luo , B. Chen , D. Zhong , and Y. Hu , Laser-induced stress-driven nanoplate jumping visualized by ultrafast electron microscopy, ACS Nano 18, 19783 (2024)

[78] Y. Zhou , X. Wang , G. Luo , M. Qu , B. Chen , D. Zhong , and Y. Hu , Spatiotemporal visualization of nanoscale rotational dynamics by movie-mode transmission electron microscopy, ACS Nano 20(3), 2821 (2026)

[79] E. I. Kamitsos and W. M. Jr Risen , Raman studies in CuTCNQ: Resonance Raman spectral observations and calculations for TCNQ ion radicals, J. Chem. Phys. 79(12), 5808 (1983)

[80] J. S. Kim , T. LaGrange , B. W. Reed , M. L. Taheri , M. R. Armstrong , W. E. King , N. D. Browning , and G. H. Campbell , Imaging of transient structures using nanosecond in situ TEM, Science 321(5895), 1472 (2008)

[81] S. K. Sinha , A. Khammari , M. Picher , F. Roulland , N. Viart , T. LaGrange , and F. Banhart , Nanosecond electron pulses in the analytical electron microscopy of a fast irreversible chemical reaction, Nat. Commun. 10(1), 3648 (2019)

[82] B. Chen , X. Fu , J. Tang , M. Lysevych , H. H. Tan , C. Jagadish , and A. H. Zewail , Dynamics and control of gold-encapped gallium arsenide nanowires imaged by 4D electron microscopy, Proc. Natl. Acad. Sci. USA 114(49), 12876 (2017)

[83] B. Chen , X. Fu , M. Lysevych , H. H. Tan , and C. Jagadish , Four-dimensional probing of phase-reaction dynamics in Au/GaAs nanowires, Nano Lett. 19(2), 781 (2019)

[84] P. Kratzer , S. Sakong , and V. Pankoke , Catalytic role of gold nanoparticle in GaAs nanowire growth: A density functional theory study, Nano Lett. 12(2), 943 (2012)

[85] D. Ban , M. Chen , F. Guo , S. Xu , and J. Huang , Solution synthesis of single crystalline zinc nanowires, J. Am. Chem. Soc. 147(11), 9062 (2025)

[86] R. C. Forsythe , C. P. Cox , M. K. Wilsey , and A. M. Müller , Pulsed laser in liquids made nanomaterials for catalysis, Chem. Rev. 121(13), 7568 (2021)

[87] J. Dou , Z. Sun , A. A. Opalade , N. Wang , W. Fu , and F. Tao , Operando chemistry of catalyst surfaces during catalysis, Chem. Soc. Rev. 46(7), 2001 (2017)

[88] A. Schottelius , F. Mambretti , A. Kalinin , B. Beyersdorff , A. Rothkirch , C. Goy , J. Muller , N. Petridis , M. Ritzer , F. Trinter , J. M. Fernandez , T. A. Ezquerra , D. E. Galli , and R. E. Grisenti , Crystal growth rates in supercooled atomic liquid mixtures, Nat. Mater. 19(5), 512 (2020)

[89] Z. Shabeeb , N. Goyal , P. Attah Nantogmah , and V. Jamali , Learning the diffusion of nanoparticles in liquid phase TEM via physics-informed generative AI, Nat. Commun. 16(1), 6298 (2025)

[90] Y. Li , N. Khuu , E. Prince , M. Alizadehgiashi , E. Galati , O. D. Lavrentovich , and E. Kumacheva , Nanoparticle-laden droplets of liquid crystals: Interactive morphogenesis and dynamic assembly, Sci. Adv. 5(7), eaav1035 (2019)

[91] Y. Shin and C. P. Brangwynne , Liquid phase condensation in cell physiology and disease, Science 357(6357), eaaf4382 (2017)

[92] L. Wu , Y. Wang , X. Xu , Y. Liu , B. Lin , M. Zhang , J. Zhang , S. Wan , C. Yang , and W. Tan , Aptamer-based detection of circulating targets for precision medicine, Chem. Rev. 121(19), 12035 (2021)

[93] U. J. Lorenz and A. H. Zewail , Observing liquid flow in nanotubes by 4D electron microscopy, Science 344(6191), 1496 (2014)

[94] J. M. Yuk , J. Park , P. Ercius , K. Kim , D. J. Hellebusch , M. F. Crommie , J. Y. Lee , A. Zettl , and A. P. Alivisatos , High-resolution EM of colloidal nanocrystal growth using graphene liquid cells, Science 336(6077), 61 (2012)

[95] X. Ye , M. R. Jones , L. B. Frechette , Q. Chen , A. S. Powers , P. Ercius , G. Dunn , G. M. Rotskoff , S. C. Nguyen , V. P. Adiga , A. Zettl , E. Rabani , P. L. Geissler , and A. P. Alivisatos , Single-particle mapping of nonequilibrium nanocrystal transformations, Science 354(6314), 874 (2016)

[96] X. Peng , F. C. Zhu , Y. H. Jiang , J. J. Sun , L. P. Xiao , S. Zhou , K. C. Bustillo , L. H. Lin , J. Cheng , J. F. Li , H. G. Liao , S. G. Sun , and H. Zheng , Identification of a quasi-liquid phase at solid–liquid interface, Nat. Commun. 13(1), 3601 (2022)

[97] S. Zhou , J. Shi , S. Liu , G. Li , F. Pei , Y. Chen , J. Deng , Q. Zheng , J. Li , C. Zhao , I. Hwang , C. J. Sun , Y. Liu , Y. Deng , L. Huang , Y. Qiao , G. L. Xu , J. F. Chen , K. Amine , S. Sun , and H. Liao , Visualizing interfacial collective reaction behaviour of Li–S batteries, Nature 621(7977), 75 (2023)

[98] Q. Zhang , M. C. Gallant , Y. Chen , Z. Song , Y. Liu , Q. Zheng , L. Chen , K. C. Bustillo , Y. Huang , K. A. Persson , and H. Zheng , Isothermal solidification for high-entropy alloy synthesis, Nature 646(8084), 323 (2025)

[99] C. Qian , E. Stanifer , Z. Ma , L. Yao , B. Luo , C. Liu , J. Li , P. Pan , W. Pan , X. Mao , and Q. Chen , Nanoscale phonon dynamics in self-assembled nanoparticle lattices, Nat. Mater. 24(10), 1616 (2025)

[100] L. Xu , Z. Cao , Z. Liu , C. Zheng , S. Peng , Y. Lu , H. Liu , and B. Chen , Filming evolution dynamics of Hg nanodroplets mediated at solid−gas and solid−liquid interfaces by in-situ TEM, Nat. Commun. 16(1), 3684 (2025)

[101] Y. Lu , Z. Song , Z. Cao , L. Xu , H. Liu , J. Xie , and B. Chen, , Dynamic liquid-like membrane gates mass transport in inorganic nanocells, Nano Lett.

[102] X. Fu , B. Chen , J. Tang , and A. H. Zewail , Photoinduced nanobubble-driven superfast diffusion of nanoparticles imaged by 4D electron microscopy, Sci. Adv. 3(8), e1701160 (2017)

[103] X. Fu , B. Chen , J. Tang , M. Th. Hassan , and A. H. Zewail , Imaging rotational dynamics of nanoparticles in liquid by 4D electron microscopy, Science 355(6324), 494 (2017)

[104] H. Zheng , S. A. Claridge , A. M. Minor , A. P. Alivisatos , and U. Dahmen , Nanocrystal diffusion in a liquid thin film observed by in situ transmission electron microscopy, Nano Lett. 9(6), 2460 (2009)

[105] R. Lachaine , E. Boulais , and M. Meunier , From thermo- to plasma-mediated ultrafast laser-induced plasmonic nanobubbles, ACS Photonics 1(4), 331 (2014)

[106] E. Lukianova-Hleb , Y. Hu , L. Latterini , L. Tarpani , S. Lee , R. A. Drezek , J. H. Hafner , and D. O. Lapotko , Plasmonic nanobubbles as transient vapor nanobubbles generated around plasmonic nanoparticles, ACS Nano 4(4), 2109 (2010)

[107] L. Hong , S. M. Anthony , and S. Granick , Rotation in suspension of a rod-shaped colloid, Langmuir 22(17), 7128 (2006)

[108] D. Mukhija and M. J. Solomon , Translational and rotational dynamics of colloidal rods by direct visualization with confocal microscopy, J. Colloid Interface Sci. 314(1), 98 (2007)

[109] S. Ji , Z. Sun , J. Liu , J. Guo , L. Tong , Y. Deng , J. Guo , C. Liu , X. Chen , C. Gao , F. Liu , M. Feng , S. Deng , and X. Fu , Ultrafast transmission electron microscopy: Principles and applications, Front. Phys. (Beijing) 21(3), 032301 (2026)

RIGHTS & PERMISSIONS

Higher Education Press