Attomicroscopy: Atom & Electron Dynamics

• Attomicroscopy is a technique that integrates attosecond timing with atomic-scale resolution to capture ultrafast electron and atomic dynamics.
• It leverages multiple modalities, such as UEM, STM, and fluorescence, to overcome conventional temporal and spatial resolution limits.
• The method uses optical gating and model-based inversion to directly access and quantify dynamic processes like charge migration and lattice interactions.

Attomicroscopy is a term used in several adjacent literatures for microscopy that targets the natural scales of electrons and atoms. In ultrafast electron microscopy and diffraction, it denotes transmission-electron-microscopy or related platforms with isolated optically gated electron probes short enough to resolve sub-femtosecond electron dynamics in real time and space (Hassan, 2018). In other contexts, the same term has been applied to attosecond, angstrom-scale diffraction of bound-electron motion in solids, to atomic-scale scanning tunnelling microscopy driven by lightwave fields, and to atomic-scale localization schemes that exploit lattice discreteness or nonlinear optical response rather than attosecond timing (Yuan et al., 2024). The common thread is the attempt to access either electron motion, bond-scale current flow, atomic wavefunctions, or atomic coordinates at spatial scales from nanometers to ångströms and, in the attosecond lineage, temporal scales from tens of femtoseconds to below one femtosecond (Hassan et al., 2017).

1. Terminology and conceptual scope

The term does not denote a single instrument architecture. Instead, the literature uses it for several experimentally and conceptually distinct microscopy programs.

Usage of the term	Primary observable	Representative realization
Attosecond electron microscopy	Time-resolved diffraction or imaging of electron dynamics	Optically gated UEM/TEM (Hassan, 2018)
Diffraction attomicroscopy in solids	Bragg-intensity modulation from bound-electron motion	Graphene current imaging (Yuan et al., 2024)
Lightwave-driven STM attomicroscopy	CEP-dependent tunnelling current	Atomic-resolution STM with single-cycle NIR drive (Maier et al., 14 Jul 2025)
Fluorescence attomicroscopy	Lattice-registered emitter localization	DIGIT in diamond (Duan et al., 6 Apr 2025)
Deterministic cold-atom attomicroscopy	Intra-site atomic density and dynamics	Nonlinear optical pumping in Cs lattices (McDonald et al., 2018)
Atomic-resolution spectroscopic imaging/tomography	Atomic coordinates or elemental maps	STEM-EELS and grid-free tomography (Xin et al., 2011)

This diversity has methodological consequences. In the attosecond-electron lineage, attomicroscopy emphasizes temporal reach: isolated electron probes, optical gating, and pump–probe synchronization are the central innovations. In fluorescence, cold-atom, and tomographic usages, the emphasis is instead on beating diffraction or grid limits by exploiting a discrete lattice prior, a nonlinear optical transfer function, or a sparse point-source model (Duan et al., 6 Apr 2025). This suggests that the term functions less as a single modality label than as a claim to direct access to atomic- or electron-scale observables.

2. From femtosecond UEM to isolated attosecond electron probes

The foundational attomicroscopy program emerged from limitations of femtosecond ultrafast electron diffraction and microscopy. Conventional UED/UEM reached atomic motion but remained limited by electron pulse duration, pump–probe synchronization jitter, space-charge broadening, and velocity dispersion, motivating a transition from few-hundred-femtosecond probes toward optically gated ultrashort electron packets (Hassan, 2018). A standard resolution budget was written as

$$\Delta t_{\text{sys}} \approx \sqrt{\Delta t_{\text{e}}^2 + \Delta t_{\text{sync}}^2 + \Delta t_{\text{det}}^2},$$

with optical gating introduced precisely to reduce the probe-duration and synchronization terms.

The key enabling mechanism was optical gating through a PINEM-type interaction at the specimen. In the 2017 demonstration of high-temporal-resolution electron microscopy, isolated $\sim 30$ fs electron pulses at $200$ keV were generated by selecting electrons coupled to a visible gate pulse at a gold nanoparticle, yielding a 16-fold enhancement over the typical $\sim 500$ fs UEM pulse duration (Hassan et al., 2017). The gated electrons constituted about $8\%$ of the total counts, remained at $<1$ electron per pulse on average, and were sufficient for practical imaging; a plasmonic field image of a silver nanowire was acquired in $30$ s (Hassan et al., 2017). Because gating occurred at the specimen plane after acceleration, propagation-induced dispersion and post-gate space-charge broadening were strongly suppressed.

The 2018 review formalized this trajectory as “Attomicroscopy: from Femtosecond to Attosecond Electron Microscopy” and argued that if the gate duration satisfies $\tau_{\text{gate}} \ll \tau_{\text{e,in}}$, then the output pulse duration approaches the gate width, $\tau_{\text{e,out}} \simeq \tau_{\text{gate}}$ (Hassan, 2018). It further connected attosecond gating to multi-octave optical pulse synthesis and to broadband near-field structures required for sub-femtosecond PINEM coupling. In that framework, attomicroscopy was explicitly defined as the extension of UED/UEM to attosecond temporal resolution with nanometer spatial resolution, directed at charge migration, plasmonic oscillations, electron–electron scattering, and electron–phonon coupling (Hassan, 2018).

A first TEM realization followed in 2023. There, a polarization-gated optical gating pulse with a linearly polarized half-cycle of FWHM $\approx 625$ as gated the probe at an aluminum grid above multilayer graphene, producing an isolated attosecond electron pulse inside a TEM (Hui et al., 2023). The experiment used a few-cycle $\sim 30$0 nm optical pump of duration $\sim 30$1 fs and field strength $\sim 30$2 V/nm, scanned delay in $\sim 30$3 as steps, and extracted synchronous oscillations of first-, second-, and third-order Bragg intensities (Hui et al., 2023). Control experiments were essential: blocking the optical gating pulse, lowering the pump field below about $\sim 30$4 V/nm, or removing the aluminum grid removed the modulations (Hui et al., 2023). The authors attributed approximately $\sim 30$5, $\sim 30$6, and $\sim 30$7 of the measured modulation in the first, second, and third Bragg orders, respectively, to intraband currents (Hui et al., 2023).

3. Diffraction attomicroscopy of bound-electron motion in graphene

A decisive conceptual broadening occurred when attomicroscopy was formulated not merely as a way to shorten electron probes, but as an inversion framework for bound-electron motion in solids. In graphene, attomicroscopy was defined as an attosecond, angstrom-scale imaging concept that uses ultrafast, gated electron diffraction to visualize and control bound-electron motion in real space and real time, without relying on ionization into the continuum (Yuan et al., 2024). The measured observable is the transient diffraction intensity of Bragg peaks, linked to the electronic density matrix in reciprocal space and to transition densities in real space. In the notation of that work,

$\sim 30$8

while the reconstructed real-space density and current are

$\sim 30$9

and

$200$0

Experimentally, a few-cycle NIR pump centered at about $200$1 nm and compressed to about $200$2 fs drove single-crystal multilayer graphene, while a collinear, optically gated, sub-femtosecond electron probe at $200$3 keV recorded diffraction snapshots as a function of pump–probe delay (Yuan et al., 2024). Time-delay steps of $200$4 attoseconds resolved sub-cycle dynamics of the NIR driver, whose optical period was about $200$5 fs (Yuan et al., 2024). The pump polarization was aligned along a C–C bond, and analysis focused on first-order Bragg peaks (Yuan et al., 2024).

The dynamical interpretation combined Bloch acceleration,

$200$6

with semiconductor Bloch equations for the reciprocal-space density matrix (Yuan et al., 2024). When $200$7 along $200$8, the electron momentum distribution displaced toward $200$9 and the current flowed along $\sim 500$0; when $\sim 500$1, the momentum distribution was symmetric and the current vanished; and when $\sim 500$2, the displacement and current reversed (Yuan et al., 2024). The reconstructed bond-directed current alternated every half-cycle, whereas the bound-electron density between the carbon atoms oscillated at $\sim 500$3, consistent with current being odd in momentum and density even in momentum (Yuan et al., 2024).

Control of current direction and amplitude was demonstrated explicitly. Inserting a $\sim 500$4m-thick SiO$\sim 500$5 plate shifted the CEP by approximately $\sim 500$6 with retrieved $\sim 500$7, flipping the current direction at the same delay $\sim 500$8 fs (Yuan et al., 2024). Increasing the pump field from $\sim 500$9 to $8\%$0 V/nm produced an approximately linear increase in the peak modulation of Bragg intensities, with the relative maximum intensity change increasing by about $8\%$1–$8\%$2 over that range (Yuan et al., 2024). Damage occurred above $8\%$3 V/nm (Yuan et al., 2024). The authors also noted that, under the applied fields of roughly $8\%$4–$8\%$5 V/nm, the diamagnetic term dominated the reconstructed current (Yuan et al., 2024).

This graphene work is methodologically important because the “image” is not a direct camera-space picture. It is a reconstructed map derived from transient Bragg intensities, tight-binding transition densities, and semiconductor Bloch dynamics. A common misunderstanding is therefore that attomicroscopy always produces model-free movies of electrons; in this implementation, the real-space current and density maps are quantitative only through the inversion model that links diffraction modulation to $8\%$6, $8\%$7, and $8\%$8 (Yuan et al., 2024).

4. Scanning-probe attomicroscopy and lightwave-driven charge transfer

A second experimental branch replaced free-electron diffraction by a localized tunnel junction. In 2025, attomicroscopy in scanning tunnelling microscopy was realized by steering tunnelling with phase-controlled, single-cycle near-infrared waveforms while preserving the angstrom-scale locality of the STM barrier (Maier et al., 14 Jul 2025). The setup used two spectrally non-overlapping components centered at $8\%$9 THz and $<1$0 THz whose coherent superposition yielded a single-cycle field with intensity-envelope FWHM $<1$1 fs at delay $<1$2 (Maier et al., 14 Jul 2025). CEP was modulated at about $<1$3 kHz, repetition rate was $<1$4 MHz, pulse energies ranged from $<1$5 to $<1$6 pJ, and comparison with TD-DFT implied local peak junction fields up to $<1$7 V/nm at the highest pulse energy (Maier et al., 14 Jul 2025).

The measured observable was the CEP-dependent component $<1$8 of the tunnelling current, extracted by lock-in demodulation at the CEP-modulation frequency (Maier et al., 14 Jul 2025). This was crucial because thermally induced currents could otherwise dominate: deliberate power-modulation tests showed that thermal current modulation scaled linearly with modulation depth, whereas during attomicroscopy measurements the demodulated power at the CEP frequency was only $<1$9 nW, corresponding to a relative fluctuation of $30$0 and remaining invariant with delay (Maier et al., 14 Jul 2025). The authors therefore argued that the observed sub-cycle signals were electronic rather than thermal in origin.

The interpretation combined barrier-modulated tunnelling with strong-field excitation. The adiabaticity parameter was written as

$30$1

and for the experimental conditions lay near unity, placing the system in an intermediate regime where multiphoton-assisted excitation and field-driven barrier tilting both contribute (Maier et al., 14 Jul 2025). In steady-state STM, the current obeys

$30$2

and the experiment showed how the effective decay length changed under optical driving (Maier et al., 14 Jul 2025).

The principal result was that TD-DFT predicted isolated tunnelling bursts shorter than $30$3 fs, with current bursts of FWHM $30$4 as (Maier et al., 14 Jul 2025). The peak current lagged the field maximum by about $30$5 fs and persisted up to about $30$6 fs, in contrast to quasi-static THz-STM behavior (Maier et al., 14 Jul 2025). At $30$7 pJ, $30$8 decayed by one order of magnitude within $30$9 Å; at $\tau_{\text{gate}} \ll \tau_{\text{e,in}}$0 pJ, an exponential fit gave $\tau_{\text{gate}} \ll \tau_{\text{e,in}}$1 Å$\tau_{\text{gate}} \ll \tau_{\text{e,in}}$2, corresponding to a current $\tau_{\text{gate}} \ll \tau_{\text{e,in}}$3 decay length of about $\tau_{\text{gate}} \ll \tau_{\text{e,in}}$4 Å (Maier et al., 14 Jul 2025). Constant-height imaging over a single Cu adatom on Ag(100) showed atomic-scale contrast in both the DC current and $\tau_{\text{gate}} \ll \tau_{\text{e,in}}$5, establishing atomic spatial resolution under attosecond lightwave drive (Maier et al., 14 Jul 2025).

A theoretical extension pushed this logic toward biomolecular quantum chemistry. The proposed “Q-attomicroscope” was an attosecond scanning tunnelling microscope in which a polarization-gated half-cycle pulse created a usable tunnelling window of about $\tau_{\text{gate}} \ll \tau_{\text{e,in}}$6 as, while a DUV pump from Attosecond Light-Field Synthesizer 2.0 delivered a $\tau_{\text{gate}} \ll \tau_{\text{e,in}}$7 fs, $\tau_{\text{gate}} \ll \tau_{\text{e,in}}$8J pulse centered at $\tau_{\text{gate}} \ll \tau_{\text{e,in}}$9 eV (Golubev et al., 27 Dec 2025). Within that proposal, attosecond-resolved ab initio simulations of thymine–adenine and cytosine–guanine predicted coherent charge-migration periods of about $\tau_{\text{e,out}} \simeq \tau_{\text{gate}}$0 fs and $\tau_{\text{e,out}} \simeq \tau_{\text{gate}}$1 fs, respectively, with target observables formulated through time-dependent density, current, dipole, and tunnelling-current contrasts (Golubev et al., 27 Dec 2025). This remains a proposal rather than an experimental attomicroscopy result.

5. Atomic-scale meanings beyond attosecond electron probes

The term has also been used for atomic-scale microscopy that does not rely on attosecond timing. In fluorescence microscopy, attomicroscopy was defined as far-field localization of optically active defects at the scale of the atomic lattice by explicitly incorporating the discrete periodic structure of the host crystal into the estimator (Duan et al., 6 Apr 2025). The central innovation was DIGIT, in which a continuous maximum-likelihood localization with uncertainty $\tau_{\text{e,out}} \simeq \tau_{\text{gate}}$2 is converted into a posterior over discrete lattice sites. In one dimension,

$\tau_{\text{e,out}} \simeq \tau_{\text{gate}}$3

and the wrong-site probability scales approximately as

$\tau_{\text{e,out}} \simeq \tau_{\text{gate}}$4

Because $\tau_{\text{e,out}} \simeq \tau_{\text{gate}}$5 in standard SMLM, the paper argued that the discrete estimator undergoes an effective exponential collapse once the continuous precision approaches the lattice constant (Duan et al., 6 Apr 2025). In silicon-vacancy centers in diamond, continuous localization reached $\tau_{\text{e,out}} \simeq \tau_{\text{gate}}$6 Å after integrating $\tau_{\text{e,out}} \simeq \tau_{\text{gate}}$7 frames per emitter, while DIGIT yielded an average localization deviation $\tau_{\text{e,out}} \simeq \tau_{\text{gate}}$8 Å, compared with $\tau_{\text{e,out}} \simeq \tau_{\text{gate}}$9 Å for conventional SMLM on the same data (Duan et al., 6 Apr 2025). Widefield PLE plus DIGIT localized more than $\approx 625$0 single emitters across about $\approx 625$1 clusters in a $\approx 625$2m field of view (Duan et al., 6 Apr 2025).

In cold-atom physics, attomicroscopy denoted deterministic nanometer-scale imaging of an atomic wavefunction within a single lattice site (McDonald et al., 2018). Ultracold $\approx 625$3Cs atoms in a one-dimensional optical lattice were interrogated with a resonant optical-pumping standing wave that was dark only in a nanometer-wide window near each node. The effective PSF width was controlled by nonlinear saturation, with

$\approx 625$4

and a measured PSF FWHM of $\approx 625$5 nm was reported, together with localization precision below $\approx 625$6 nm and a pumping pulse length of $\approx 625$7s (McDonald et al., 2018). The same work introduced a moiré variant with magnification

$\approx 625$8

reaching $\approx 625$9 at the smallest detuning and thereby converting $\sim 30$00 nm features into $\sim 30$01 mm stripes (McDonald et al., 2018).

A further usage appeared in atomic-resolution spectroscopic imaging of nanocatalyst ensembles. There, “attomicroscopy” described aberration-corrected STEM-EELS with atomic resolution and statistically meaningful throughput rather than attosecond timing (Xin et al., 2011). A fifth-order aberration-corrected Nion UltraSTEM at $\sim 30$02 kV, coupled to an efficient EELS spectrometer, delivered a roughly thousandfold increase in spectroscopic data-collection speed, a $\sim 30$03–$\sim 30$04 Å probe, dwell times of $\sim 30$05–$\sim 30$06 s per spectrum, and map sizes from $\sim 30$07 pixels in about $\sim 30$08 min to $\sim 30$09 pixels in about $\sim 30$10 min (Xin et al., 2011). More than $\sim 30$11 EELS spectra were acquired across the study, enabling ensemble-resolved shell-thickness and degradation statistics in Pt–Co fuel-cell catalysts (Xin et al., 2011).

Relatedly, atomic super-resolution tomography framed attomicroscopy as recovery of atomic coordinates from very few electron projections by combining a continuous point-source model with sparsity and physically motivated interaction potentials (Ganguly et al., 2020). The forward model used the Radon transform of a measure convolved with a Gaussian PSF, and the reconstruction objective added a Lennard–Jones pair potential,

$\sim 30$12

together with a minimum-separation constraint (Ganguly et al., 2020). On synthetic interstitial, vacancy, and edge-dislocation examples, the grid-free ADCG-with-energy method consistently gave the lowest mean Euclidean position errors among the tested methods (Ganguly et al., 2020).

6. Methodological limits, misconceptions, and future directions

Across its usages, attomicroscopy is limited less by a single fundamental barrier than by a shifting set of bottlenecks specific to each platform. In optically gated UEM/TEM, the recurrent constraints are counts per pulse, saturation broadening of the gate, timing jitter, and the need for broadband nanostructures or gating media that preserve sub-femtosecond response (Hassan et al., 2017). The 2018 and 2023 attosecond-electron papers emphasized low charge per pulse, specimen-plane gating, and optical phase locking as the central strategies for suppressing space charge, dispersion, and RF-linked timing drift (Hassan, 2018). In diffraction attomicroscopy of graphene, the dominant limitations are inversion complexity, omission of core-electron scattering in the diffraction model, temporal resolution set by the optical gating window and timing jitter, and material damage above about $\sim 30$13 V/nm (Yuan et al., 2024). In lightwave-driven STM, thermal-load stabilization and lock-in isolation of CEP-sensitive current are indispensable because thermal artifacts can exceed the desired ultrafast signal if average power is not stabilized (Maier et al., 14 Jul 2025).

A useful corrective to common overstatement is that possession of an “attomicroscope” does not guarantee an attosecond measurement. In the cable-bacteria study, the platform was described as capable of attosecond resolution, but the experiment itself explicitly did not use optical gating or attosecond methods; instead it performed picosecond pump–probe EELS on biological conductive nanofibers (Zhao et al., 2 Oct 2025). That work reported a $\sim 30$14-electron peak at $\sim 30$15 eV, a bulk plasmon-like resonance at $\sim 30$16 eV, a static conductivity enhancement of about $\sim 30$17 inferred from $\sim 30$18, and picosecond decay and recovery attributed to electron–phonon coupling (Zhao et al., 2 Oct 2025). The distinction matters because the biological extension was a proof of scope for the toolset rather than a demonstration of attosecond-resolved electron motion.

The semantic breadth of the term also creates a second misconception: attomicroscopy is not synonymous with any microscope that attains atomic spatial resolution. The fluorescence, cold-atom, STEM-EELS, and tomography literatures use the term for different reasons: discrete lattice priors, nonlinear optical transfer functions, ensemble atomic-resolution spectroscopic mapping, or grid-free coordinate recovery (Duan et al., 6 Apr 2025). These approaches share an aspiration to direct atomic-scale inference, but they do not share a common temporal philosophy or a common measurement equation.

Several future directions are already specified in the literature. Diffraction attomicroscopy points toward energy-filtered attomicroscopy, synthesized light transients with arbitrary waveforms, cryo-attomicroscopy, and extension to other two-dimensional materials and molecular crystals (Yuan et al., 2024). TEM-based attomicroscopy points toward broader integration with diffraction, real-space imaging, holography, phase contrast, and energy-filtered TEM, together with higher gating efficiency and more stable optical phase control (Hui et al., 2023). STM-based attomicroscopy points toward chemically specific charge-transfer imaging in molecules, defects, and quantum materials, and ultimately toward atom-scale petahertz electronics (Maier et al., 14 Jul 2025). The Q-attomicroscope proposal extends this prospect to attosecond imaging of biomolecular charge migration in DNA (Golubev et al., 27 Dec 2025). Taken together, these trajectories indicate that attomicroscopy is becoming a family of high-information, strongly model-coupled microscopies whose defining ambition is direct access to electronic or atomic-scale dynamics on the scales at which they actually occur.