Attosecond and Ultrafast Frontiers

Status: high-order harmonic generation, isolated and train-form attosecond pulses, femtosecond pump–probe spectroscopy, attosecond streaking, RABBITT interferometry, ultrafast diffraction, and strong-field ionization are established. Quantitative charge-migration measurements, coupled electron–nuclear reconstruction, liquid and solid high-harmonic spectroscopy, attosecond X-ray methods, and quantum-light-driven strong-field physics are active. A universal tunnelling time, model-independent “movies” of many-electron wavefunctions, routine attochemical control of reaction products, and a settled record for the shortest measured pulse are not established.

Last reviewed: 26 July 2026. Pulse durations, photon fluxes, fitted delays, source repetition rates, and reconstruction algorithms are date-sensitive. Temporal resolution, delay precision, and the duration assigned by an inverse retrieval are different quantities. A time-dependent spectrum is not a direct photograph of a wavefunction.

Core Question

How can electronic and nuclear motion be reconstructed and controlled on its natural timescale without confusing the probe, the inverse model, and the underlying dynamics?

An ultrafast experiment is not a camera with a fast shutter. It is a controlled interaction followed by an inference:

$$\begin{aligned} \text{source waveform} &\longrightarrow \text{state preparation} \longrightarrow \text{field-driven evolution} \\ &\longrightarrow \text{probe interaction} \longrightarrow \text{detector record} \longrightarrow \text{reconstruction}. \end{aligned}$$

Every arrow can carry phase, selection rules, propagation effects, timing jitter, correlations, and model assumptions. Shorter pulses help, but they do not make the inverse problem disappear.

The frontier contains five linked questions:

1. Which electronic degrees of freedom can be timed or reconstructed with attosecond sensitivity?
2. How can high-harmonic sources become brighter, more stable, more controllable, and spectrally broader without losing a trustworthy temporal characterization?
3. Can electronic coherence and nuclear rearrangement be observed together in molecules rather than inferred from separate experiments?
4. Which pump–probe observables identify a mechanism uniquely, and which are compatible with several models?
5. Can strong fields steer a desired quantum pathway while retaining quantitative control over ionization, damage, propagation, and many-body response?

The most important methodological question is therefore:

What observable was measured, what forward model connects it to the stated dynamics, and which alternative models were excluded?

Why It Matters

Electrons and nuclei set different but coupled clocks

An electronic superposition with energy splitting $\Delta E$ acquires a relative phase on a timescale

$$T=\frac{2\pi\hbar}{|\Delta E|}.$$

An electron-volt splitting corresponds to a few femtoseconds; larger valence–core or continuum energy differences support subfemtosecond structure. Nuclear vibration, torsion, bond extension, and passage through a conical intersection usually unfold over several to hundreds of femtoseconds.

These clocks are coupled. Electronic coherence can launch nuclear motion. Nuclear geometry changes electronic energy gaps and nonadiabatic couplings. Ionization entangles the residual ion with the outgoing electron. Solvation and phonons dephase selected coherences. A useful experiment must decide which subsystem is observed and which is traced out.

Ultrafast observables expose mechanisms hidden in final products

Two mechanisms can produce the same final ion yield, absorption spectrum, or photoproduct distribution while differing in:

• when population leaves and enters electronic states;
• whether a wave packet crosses one or several nonadiabatic regions;
• whether charge motion is coherent or population-like;
• which continuum trajectory returns in high-harmonic generation;
• whether an apparent delay arises in emission, propagation, or detection; and
• whether a solvent changes a pathway or merely erases its coherence.

Time-, angle-, polarization-, and coincidence-resolved data can distinguish such mechanisms. The gain is explanatory, not merely cinematographic.

Attosecond sources connect spectroscopy to X-ray structure

Tabletop high-harmonic generation supplies coherent extreme-ultraviolet and soft-X-ray radiation synchronized to an optical waveform. Free-electron lasers supply much larger pulse energies and extend attosecond operation toward hard X-rays. Together they enable:

• element- and orbital-selective transient absorption;
• core-level photoelectron timing;
• nonlinear X-ray spectroscopy;
• electron and nuclear wave-packet control;
• nanoscale phase-resolved microscopy; and
• diffraction with electronic and structural sensitivity.

The source trade space includes pulse energy, duration, contrast, bandwidth, photon energy, repetition rate, coherence, timing jitter, and sample damage. There is no universally best source.

Strong-field physics is both a source mechanism and a laboratory

The intense field that generates high harmonics also drives nonperturbative ionization, rescattering, above-threshold ionization, double ionization, and lightwave currents. This dual role is powerful and dangerous. The electron wave packet is simultaneously:

• a source of short-wavelength radiation;
• a probe that returns to its parent;
• a participant in the dynamics under study; and
• an object reconstructed through a strong-field model.

The frontier advances when these roles are calibrated separately enough to support a specific claim.

Canonical Boundary

This page is the canonical home for dated frontier comparisons, unresolved interpretive questions, and current experimental milestones in attosecond and ultrafast science.

It does not repeat the canonical foundations:

• Ultrafast Spectroscopy Overview owns pulse bandwidth, density-operator preparation, pump–probe response, instrument functions, coherence and dephasing, transient absorption, streaking, RABBITT, and introductory molecular dynamics.
• Photoelectron Spectroscopy owns ionization channels, binding-energy conventions, Dyson orbitals, angular distributions, and continuum scattering.
• Mode Locking owns ultrashort laser formation, chirp, carrier-envelope phase, and pulse diagnostics.
• Gauge Choices in Light–Matter Physics owns exact gauge equivalence and gauge errors caused by truncation.
• Nonadiabatic Coupling and Conical Intersections own the molecular Hamiltonian and geometric mechanisms behind coupled electron–nuclear motion.

This page summarizes those results only far enough to evaluate frontier claims. A detailed derivation retains one canonical home.

Minimal Background

Pulse duration imposes bandwidth, not complete resolution

For a field envelope with rms widths, Fourier uncertainty gives

$$\Delta t\,\Delta\omega\geq\frac{1}{2}.$$

For a transform-limited Gaussian intensity pulse, the full-width-at-half- maximum relation is

$$\Delta\nu\,\Delta t\simeq0.441.$$

Short duration therefore requires broad coherent bandwidth. Spectral phase can stretch the pulse while leaving its spectrum unchanged. Satellite pulses and spatiotemporal coupling can make one scalar duration inadequate.

The atomic unit of time is

$$t_{\rm a}=\frac{\hbar}{E_h} \simeq24.19\ {\rm as}.$$

It is a natural unit derived from the Hartree energy, not literally the classical orbital period of a hydrogen electron. Crossing this numerical scale is a source-metrology milestone, not a new dynamical law.

An experiment’s effective resolution also includes pump duration, probe duration, timing jitter, geometry, and response:

$$\sigma_{\rm IRF}^2 \simeq \sigma_{\rm pump}^2 +\sigma_{\rm probe}^2 +\sigma_{\rm jitter}^2 +\sigma_{\rm geom}^2.$$

A $50\ {\rm as}$ probe combined with a $10\ {\rm fs}$ pump does not produce a $50\ {\rm as}$ preparation–measurement resolution.

Timing precision can be smaller than the response width

A measured trace is often a convolution:

$$S_{\rm meas}(\tau) = \int_{-\infty}^{\infty} R(t)\,G(\tau-t)\,dt +\epsilon(\tau),$$

where $R$ is the material response, $G$ the instrument response, and $\epsilon$ measurement noise. With high signal-to-noise ratio and a well-constrained model, a fitted delay can have uncertainty smaller than the width of $G$. This is parameter-estimation precision, not direct temporal resolution of arbitrary structure.

Sub-response timing claims are credible when:

• the forward model is identifiable;
• the delay axis is independently calibrated;
• pulse-shape and time-zero uncertainty are propagated;
• residuals and model alternatives are tested;
• correlations among fitted parameters are reported; and
• synthetic-data recovery is demonstrated.

The three-step model organizes gas-phase high harmonics

The semiclassical recollision picture separates high-order harmonic generation into:

1. field-induced ionization;
2. continuum acceleration; and
3. recombination with emission of a high-energy photon.

For a monochromatic field of amplitude $E_0$ and angular frequency $\omega$, the ponderomotive energy is

$$U_p=\frac{e^2E_0^2}{4m_e\omega^2}.$$

The familiar single-active-electron cutoff is

$$\hbar\Omega_{\rm cut} \simeq I_p+3.17U_p.$$

The Keldysh parameter

$$\gamma_K = \frac{\omega\sqrt{2m_e I_p}}{eE_0} = \sqrt{\frac{I_p}{2U_p}}$$

organizes multiphoton-like $(\gamma_K\gg1)$ and tunnelling-like $(\gamma_K\ll1)$ regimes. The crossover near unity is not a sharp phase boundary.

These formulas are a map, not a complete theory. Coulomb focusing, depletion, multielectron response, resonances, molecular orientation, propagation, and macroscopic phase matching can all matter.

Microscopic emission and macroscopic buildup are separate

The detected harmonic field is a coherent sum over the generation medium:

$$E_q^{\rm out} \propto \int d^3r\, P_q(\mathbf r) e^{i\Phi_q(\mathbf r)}.$$

In a one-dimensional uniform approximation,

$$I_q(L) \propto L^2 \operatorname{sinc}^2 \left( \frac{\Delta k_qL}{2} \right),$$

with coherence length

$$L_c=\frac{\pi}{|\Delta k_q|}.$$

Neutral and plasma dispersion, geometric phase, dipole phase, absorption, focusing, and ionization all contribute to $\Delta k_q$. A change in the measured harmonic spectrum can therefore arise from single-emitter dynamics, propagation, or both.

The emitted attosecond waveform also carries an energy-dependent emission time, or attochirp. Compressing a broad spectrum requires controlling or compensating this spectral phase, not merely filtering the continuum.

Streaking is a calibrated momentum mapping

In a simplified strong-field approximation, an electron released near $t_i$ acquires a final momentum related to the vector potential:

$$\mathbf p_f \simeq \mathbf p_0 -q\,\mathbf A(t_i).$$

The sign and canonical-versus-kinetic momentum convention must be stated. Real retrievals include Coulomb interaction, dipole matrix elements, pulse chirp, detector response, and volume averaging.

An attosecond streaking spectrogram is therefore a two-dimensional inverse problem:

$$S(E,\tau) = \left| \int dt\, E_{\rm XUV}(t-\tau) d(t) e^{i\Phi_{\rm cont}(t,E)} \right|^2.$$

Different retrieval algorithms may impose different continuum approximations, regularization, or pulse parameterizations. Agreement among algorithms and independent diagnostics is stronger than one low residual.

RABBITT measures a relative two-photon phase

In reconstruction of attosecond beating by interference of two-photon transitions, neighboring harmonics and absorption or emission of one infrared photon reach the same sideband. A schematic sideband signal is

$$S_{2q}(\tau) = A_{2q} +B_{2q} \cos \left[ 2\omega\tau -\Delta\phi_{\rm XUV} -\Delta\phi_{\rm at} \right].$$

The oscillation phase contains both the harmonic spectral phase and an atomic or molecular two-photon phase. Converting that phase to a delay requires a reference or theory for the measurement-induced continuum–continuum contribution.

Streaking and RABBITT are complementary. Neither reads an emission time without a model.

A photoemission delay is generally composite

A useful decomposition is

$$\tau_{\rm meas} = \tau_{\rm W} +\tau_{\rm cc} +\tau_{\rm corr} +\tau_{\rm prop} +\tau_{\rm ref},$$

where the terms may represent a Wigner-like scattering delay, continuum–continuum coupling, many-electron correlation, propagation or transport, and reference-channel contributions. The exact decomposition depends on the protocol.

Near resonances, thresholds, and multichannel openings, delays can become large and strongly energy dependent. A measured delay is not necessarily a classical dwell time between two spatial points.

Strong-field “trajectories” are saddle-point histories

In the strong-field approximation, ionization and return times arise from saddle points of an action:

$$S(\mathbf p,t,t') = \int_{t'}^t \left[ \frac{(\mathbf p+e\mathbf A(\tau))^2}{2m_e} +I_p \right]d\tau.$$

The saddle times are often complex. Short and long quantum trajectories are coherent contributions associated with different excursion histories, not classical particles whose paths were directly watched. Coulomb-corrected semiclassical methods, time-dependent Schrödinger calculations, and trajectory-resolved observables test where the picture is reliable.

Ionization creates a bipartite state

After broadband molecular photoionization, a minimal state is

$$|\Psi\rangle = \sum_j c_j |I_j\rangle |e_j\rangle,$$

where $|I_j\rangle$ labels an ionic channel and $|e_j\rangle$ its correlated photoelectron wave packet. Tracing out the electron gives ionic coherence

$$\rho_{jk}^{(I)} = c_jc_k^* \langle e_k|e_j\rangle.$$

If the electron wave packets are distinguishable, their overlap is small and ionic coherence is suppressed. An attosecond pulse can prepare a broad superposition while simultaneously entangling the subsystem whose coherence one hoped to observe.

This is central to charge migration. A time-dependent hole density inferred for the ion is not independent of the unobserved electron, nuclear wave packet, molecular orientation, and ionization matrix elements.

Molecular “movies” combine complementary observables

Transient absorption is sensitive to electronic populations, coherences, and core-level transitions. Photoelectrons are sensitive to ionization channels and continuum scattering. Coulomb explosion maps a final fragment momentum distribution through a strong probe. X-ray or electron diffraction measures scattering from pair correlations and electronic density.

For diffraction, a schematic signal is

$$I(\mathbf q,t) \propto \left\langle \Psi(t) \middle| \hat\rho^\dagger(\mathbf q) \hat\rho(\mathbf q) \middle| \Psi(t) \right\rangle.$$

Recovering real-space structure requires finite-$q$ inversion, orientation averaging, background subtraction, and often a structural prior. Resolution claimed after super-resolution or learned inversion must be distinguished from the instrument’s native diffraction limit.

An evidence ladder for ultrafast claims

1. Source evidence: spectrum, spectral phase, spatial mode, contrast, energy, carrier-envelope phase, and shot-to-shot stability.
2. Timing evidence: delay calibration, time zero, jitter, and instrument response.
3. Observable evidence: raw spectra, momenta, coincidences, absorption, emission, or diffraction with detector calibration.
4. Forward-model evidence: gauge, continuum treatment, propagation, electronic structure, nuclear dynamics, and instrument convolution.
5. Reconstruction evidence: identifiability, uncertainty, algorithm dependence, and synthetic-data recovery.
6. Mechanism evidence: alternatives excluded by additional observables or interventions.
7. Control evidence: a prepared change in pathway, state, or product with a causal control comparison.

Calling a trace a “movie” does not move it up this ladder.

What Is Known

High-order harmonics are coherent attosecond sources

Established. Gas-phase high-harmonic generation produces phase-related odd harmonics and attosecond pulse trains. Few-cycle and polarization, ionization, amplitude, or phase-matching gates can isolate one dominant emission burst. Streaking and interferometric methods retrieve temporal structure and spectral phase.

The three-step model correctly organizes the plateau, cutoff, short and long trajectory families, and wavelength scaling in its domain. Quantitative source prediction requires Maxwell propagation and a more complete single-emitter response.

In 2026, filamentation-assisted transient phase matching in semi-infinite gas cells produced high-contrast isolated pulses using post-compressed ytterbium-based drivers. Reported examples included $200\ {\rm as}$ at $65\ {\rm eV}$ in argon, $69\ {\rm as}$ at $100\ {\rm eV}$ in neon, and $65\ {\rm as}$ at $135\ {\rm eV}$ in helium. This establishes a robust source route under the reported conditions; it does not make every filament a single-pulse gate.

Attosecond X-ray free-electron-laser operation is established

Established source result; active spectroscopy frontier. In 2024, cascaded superradiant operation produced isolated soft-X-ray pulses with median energy above $100\ \mu{\rm J}$ and measured peak power up to $1.1\ {\rm TW}$. A separate European XFEL result generated approximately $200\ {\rm as}$, $9\ {\rm keV}$ hard-X-ray pulses with mean energy around $180\ \mu{\rm J}$ and demonstrated megahertz intra-train operation.

These are source milestones. Pulse duration at a free-electron laser can vary shot to shot, and synchronizing an external pump adds timing uncertainty. The scientific frontier is not source existence but pulse-resolved, damage-aware experiments with independently characterized pump and probe.

Photoionization delays contain real scattering information

Established. Relative photoemission delays in atoms, molecules, and solids depend on energy, channel, angle, resonances, correlation, and the measurement field. They are reproducible observables when the reference and retrieval are specified.

In 2024, attosecond soft-X-ray free-electron-laser measurements accessed core-level molecular ionization near the oxygen K edge of nitric oxide. Reported delays reached roughly $700\ {\rm as}$ near threshold and showed strong spectral modulation associated with shape resonances, Auger–Meitner emission, and multielectron scattering. The result is not a $700\ {\rm as}$ classical escape flight through the molecule.

Molecular charge redistribution can be resolved

Established in selected systems; active generalization. Attosecond transient absorption, high-harmonic spectroscopy, and momentum-resolved photoionization have followed electronic coherences and charge localization in small molecules.

In 2025, attosecond transient absorption and quantum-chemical modelling of $\mathrm{CF_3I^+}$ separated:

• a vibrational rearrangement time of $9.38\pm0.21\ {\rm fs}$;
• population-transfer times near $2.3$–$2.4\ {\rm fs}$; and
• a $1.46\pm0.41\ {\rm fs}$ delay between hole disappearance on fluorine and appearance on iodine.

The reported instrument-response width was $\sigma_{\rm Inst}=1.00\pm0.05\ {\rm fs}$. The sub-response delay uncertainty was obtained through a constrained forward model and element-sensitive data, not by resolving an arbitrary $0.4\ {\rm fs}$ waveform feature.

Ion–photoelectron entanglement is experimentally relevant

Established in a controlled molecular experiment. In 2026, phase-locked pairs of isolated attosecond pulses and a few-cycle near-infrared pulse were used to ionize and dissociate hydrogen molecules. Varying two delays changed the electronic coherence observed in $\mathrm{H_2^+}$ by controlling ion–photoelectron entanglement.

This result directly supports the reduced-coherence factor $\langle e_k|e_j\rangle$. It also sharpens the boundary of a charge-migration claim: the residual ion cannot generally be treated as a pure isolated electronic wave packet. Coincident detection with isolated attosecond pulses remains technically demanding.

Ultrafast diffraction can constrain transient structure

Established with model-assisted inversion. Femtosecond X-ray and electron diffraction have tracked bond extension, ring opening, dissociation, phase transitions, and coherent phonons. They measure ensemble-averaged scattering, not coordinates of one molecule.

In 2025, mega-electron-volt ultrafast electron diffraction combined with a convex super-resolution inversion was applied to 1,3-cyclohexadiene ring opening. The reconstruction reported transient carbon–carbon distance differences below $0.4\ \text{\AA}$ and an approximately $30\ {\rm fs}$ nuclear-wave-packet traversal between two conical-intersection regions. The sub-ångström statement describes the model-assisted reconstruction and its tested uncertainty, not the native real-space point-spread function alone.

Attosecond chiroptical observables are accessible

Established first-generation capability. In 2025, characterized circularly polarized attosecond pulses, a weak near-infrared field, and electron–ion coincidence detection controlled photoelectron circular dichroism in methyloxirane. Co-rotating fields nearly doubled and could reverse the dichroism relative to single-photon ionization. The experiment reported angle-dependent chiral delay asymmetries, including contributions from continuum–continuum transitions.

This establishes phase-sensitive attosecond chiroptical control in the tested molecule. It does not establish enantioselective chemistry or a universal mapping from delay to static molecular handedness.

Light-field-driven solid response is measurable

Established in selected materials. Solid-state high-harmonic generation, attosecond transient absorption and reflection, and electro-optic sampling probe interband polarization, intraband currents, excitons, screening, and virtual excitation.

In 2025, broadband attosecond transient reflection in diamond showed that virtual interband transitions materially affected the timing and adiabaticity of a response that could otherwise be described as intraband dominated. That is evidence against an overly simple band-acceleration account, not by itself a functional petahertz switch.

Phase-sensitive electron microscopy has entered the subcycle regime

Established proof of principle. Free-electron homodyne detection uses a phase-controlled reference interaction as a local oscillator for weak sample-induced modulation of an electron wavefunction. A 2024 experiment imaged plasmonic near fields with few-nanometre spatial and subcycle temporal resolution.

The phrase attosecond electron microscopy in this context refers to phase-resolved sampling of a periodic optical response. It does not mean that an isolated attosecond electron packet directly photographed arbitrary nonrepeating atomic structure.

What Is Actively Developing

Source brightness, contrast, and repetition rate

Active. Shortest duration is only one source axis. Experiments increasingly optimize

$$\left( \Delta t, \Phi_\gamma, \hbar\Omega, f_{\rm rep}, \text{contrast}, \text{stability} \right),$$

where $\Phi_\gamma$ is usable photon flux and $f_{\rm rep}$ repetition rate. High repetition helps coincidence and weak-signal measurements. High pulse energy helps nonlinear X-ray spectroscopy. Broad bandwidth supports short pulses but complicates optics, detector calibration, and sample selectivity.

Ytterbium drivers, mid-infrared sources, optical parametric chirped-pulse amplifiers, enhancement cavities, plasma mirrors, and free-electron lasers occupy different regions of this space.

XUV-pump–XUV-probe experiments

Active. Many attosecond measurements use an extreme-ultraviolet pulse and an intense infrared field. The infrared probe can dress, ionize, or reshape the very dynamics being timed. Two independently controlled XUV or X-ray pulses could separate excitation and readout more cleanly.

The difficulty is severe: splitting broadband XUV light loses flux, delay control must be attosecond stable, and nonlinear cross sections are small. High-flux tabletop sources and attosecond free-electron lasers are making such experiments more plausible, but routine perturbative XUV-pump–XUV-probe molecular spectroscopy is not yet established.

Simultaneous electronic and nuclear reconstruction

Active. A complete molecular account seeks a time-dependent joint state

$$\Psi(\mathbf r_1,\ldots,\mathbf r_N;\mathbf R,t),$$

or at least experimentally constrained electronic and nuclear reduced densities. No single observable supplies it.

Promising combinations include:

• transient absorption plus fragment coincidence;
• photoelectrons plus ion momenta;
• time-resolved X-ray photoelectron spectroscopy plus Coulomb explosion;
• diffraction plus electronic-state spectroscopy; and
• core-level probes at several elemental edges.

Joint analysis should propagate correlations among datasets rather than fit each one to a separate preferred pathway.

Coincidence-resolved attosecond molecular science

Active. COLTRIMS and velocity-map imaging can correlate electron and ion momenta, molecular orientation, fragmentation channel, and delay. The price is low count rate, detector dead time, false coincidences, and a severe data volume.

The 2026 entanglement experiment makes a specific next step compelling: coincidently measuring the ionic coherence and the associated electron under isolated-attosecond-pulse preparation. Such a measurement would test which path information suppresses or restores a chosen subsystem coherence.

High-harmonic generation in liquids

Active. Liquids offer native chemical environments, solvation structure, and condensed-phase density, but also strong absorption, scattering, replenishment, surface deformation, and propagation complexity.

In 2026, liquid high-harmonic measurements spatially separated short and long trajectory contributions and used a phase-controlled two-colour field to retrieve opposite energy–time correlations for the two families. Semiclassical recollision simulations reproduced the observations. This establishes trajectory-resolved attosecond mapping in that experiment; it does not prove that every liquid can be treated as independent gas-like recollisions.

The next tests should vary liquid thickness, composition, mean free path, orientation, absorption, and focusing while preserving phase calibration.

Strong-field physics driven by quantum light

Active and emerging. In 2026, bright squeezed vacuum with $300\ {\rm nJ}$ average pulse energy produced photoelectron momentum peaks comparable to a $7.1\ \mu{\rm J}$ coherent pulse in sodium tunnelling ionization. The more-than-20-fold “quantum boost” referred to this matched nonlinear photoelectron observable and was tuned through field correlations at fixed average energy.

The result establishes that photon statistics can control a nonperturbative atomic response. It does not imply a universal 20-fold energy-conversion gain, nor does it yet demonstrate quantum-light-driven high-harmonic sources with superior end-to-end attosecond performance. Propagation, multimode quantum statistics, depletion, and light–electron entanglement are active theory problems.

Attosecond spectroscopy of solids and nanostructures

Active. The central inverse problem is to separate:

• interband polarization;
• intraband acceleration;
• real and virtual carrier populations;
• excitons and screening;
• Berry connections and topology;
• scattering and dephasing;
• surface and bulk response; and
• propagation through a finite sample.

Polarization, crystal orientation, two-colour control, harmonic phase, transient reflection, and momentum-resolved photoemission provide complementary constraints. A fitted band trajectory is not unique unless these channels and gauge-consistent observables agree.

From observing charge migration to controlling chemistry

Active; chemical control remains unestablished in general. A coherent hole density can move before substantial nuclear rearrangement. To call this attochemistry, one must show more than oscillation:

1. prepare a reproducible electronic coherence;
2. intervene at a controlled phase or delay;
3. alter a later nuclear pathway or product;
4. exclude heating, field-ionization, and intensity-selection alternatives; and
5. report control contrast and yield.

Electronic motion may be too rapidly entangled or dephased to remain a separable control knob. That limitation is a physical result, not a failure of the field.

What Is Debated or Uncertain

Is there a unique tunnelling time?

Controversial. Quantum tunnelling does not supply one universally agreed time operator. Attoclock offsets depend on the Coulomb potential, nonadiabatic initial momentum, pulse ellipticity, focal averaging, depletion, multielectron response, and the chosen operational definition.

A 2025 phase-resolved attoclock extrapolated circular-polarization deflection angles from carrier-envelope-phase-dependent elliptical measurements and concluded that ionization-potential sensitivity dominated while tunnelling delay was a minor effect. This narrows one protocol’s interpretation; it does not prove that every proposed tunnelling-time definition is zero.

Claims should name the measured angle, calibration, forward model, and time definition rather than report “the tunnelling time.”

Is a photoemission delay an emission time?

Active. A spectral phase derivative can be a Wigner-like group delay, but the measured quantity also includes the probe transition and reference. Near resonances, wavepacket reshaping makes a single classical departure time especially misleading.

Relative delays are usually more robust than absolute ones. Agreement across streaking, RABBITT, channel-resolved theory, and angular data is stronger than agreement at one energy.

How short was the shortest pulse?

Controversial as of July 2026. A 2025 Ultrafast Science article reanalyzed a 2017 krypton streaking dataset and reported an isolated soft-X-ray pulse duration of

$$19.2^{+4.6}_{-2.1}\ {\rm as}.$$

A published 2026 comment challenged the assignment, noting that the same dataset had previously supported substantially longer estimates, including an approximately $332\ {\rm as}$ upper limit and a later $165\ {\rm as}$ reanalysis. The dispute concerns continuum modelling, retrieval assumptions, and what the spectrogram identifies.

This page therefore does not adopt $19.2\ {\rm as}$ as a settled record. It treats the episode as a valuable metrology lesson:

• publish raw data and retrieval code;
• test several forward models;
• report algorithm and prior sensitivity;
• validate on synthetic traces with known pulses;
• compare independent diagnostics; and
• distinguish precision within a model from model uncertainty.

A separate 2025 preprint reported $25\pm2\ {\rm as}$ pulses with a different source and angle-resolved helium streaking retrieval. Until peer review and independent validation mature, it should remain labelled a preprint claim.

Does a reconstructed movie uniquely identify a wavefunction?

Usually no. Density and current observables do not generally determine a many-body wavefunction uniquely. Finite detector acceptance, orientation averaging, and unobserved entanglement further reduce information.

A defensible “movie” states what is reconstructed:

• a one-particle density;
• a hole density conditional on an ionization model;
• a pair distribution;
• a set of structural coordinates;
• a reduced density matrix; or
• an ensemble of trajectories consistent with the data.

It also gives uncertainty and alternative reconstructions.

Can high-harmonic spectra perform orbital tomography?

Conditional. Harmonic amplitude and phase contain recombination dipoles, continuum propagation, ionization amplitudes, orientation, multielectron response, and macroscopic propagation. Under controlled assumptions, tomographic inversions can recover useful orbital or charge information.

The inversion is not generally unique. Plane-wave recombination, single-active- electron structure, and negligible propagation are approximations whose errors can resemble the feature being reconstructed.

Does subcycle response imply petahertz electronics?

Speculative at the systems level. Reversible polarization and current can follow an optical field within a cycle. A useful electronic device additionally requires:

• reproducible state preparation;
• input and output encoding;
• gain or switching contrast;
• fan-out;
• low error and heating;
• endurance;
• integration; and
• an energy and throughput advantage.

Observing a subcycle current or reflectivity modulation is a physical milestone, not yet a processor benchmark.

Can strong fields be noninvasive probes?

Context dependent. Strong-field probes can be exquisitely sensitive, but they ionize, dress, accelerate, and sometimes fragment the target. A nonperturbative probe is not disqualified; it requires a forward model and an intensity scan that show which inferred feature belongs to the unprobed dynamics and which is induced by readout.

Key Experimental Platforms

Platform	Native advantage	Frontier observable	Dominant caveat
Tabletop gas HHG	intrinsic optical synchronization and coherent XUV	isolated pulses, streaking, transient absorption	low conversion, propagation, attochirp
Enhancement-cavity HHG	high repetition and average flux	coincidence and precision spectroscopy	intracavity stability, target loading
Mid-infrared and OPCPA drivers	extended cutoff and waveform control	water-window harmonics, long trajectories	lower repetition, wavelength scaling of yield
X-ray free-electron lasers	high pulse energy and hard-X-ray reach	core timing, nonlinear X-ray response, diffraction	shot variation, timing jitter, facility access
COLTRIMS and reaction microscopes	electron–ion coincidence and molecular frame	channel-resolved delays and entanglement	low rates, false coincidences, detector acceptance
Velocity-map imaging	efficient momentum projection	angular streaking and photoelectron asymmetry	inversion symmetry assumptions, projection
Transient absorption and reflection	element- and band-sensitive phase response	charge transfer, core shifts, solid response	overlapping pathways and line-shape inversion
Ultrafast electron diffraction	strong scattering and compact geometry	bond and pair-distance evolution	space charge, ensemble averaging, inversion priors
Time-resolved X-ray diffraction	element and momentum selectivity	structural dynamics and phase transitions	source jitter, radiation damage, weak signals
Free-electron microscopy	nanometre spatial resolution	phase-resolved near fields and polarizations	periodic reference, electron coherence
Liquid jets and flat sheets	native solvated chemistry	liquid HHG and ultrafast reaction dynamics	absorption, interfaces, thickness fluctuations
Solid and metasurface HHG	compactness and band sensitivity	interband/intraband dynamics, resonant enhancement	damage, propagation, local-field calibration
Quantum-light strong-field sources	tunable photon statistics	correlation-controlled ionization and harmonics	multimode statistics, low peak intensity, calibration

Reporting vectors

For an attosecond source, report

$$\mathcal S = \left( \Delta t, \Delta E, \Phi_\gamma, \hbar\Omega, f_{\rm rep}, C_{\rm sat}, \sigma_{\rm CEP} \right),$$

where $C_{\rm sat}$ characterizes satellite contrast and $\sigma_{\rm CEP}$ carrier-envelope-phase stability.

For a pump–probe result, report

$$\mathcal T = \left( \sigma_{\rm IRF}, \sigma_{\rm jitter}, \delta\tau_{\rm cal}, \delta\tau_{\rm fit}, \text{time zero}, \text{model} \right).$$

For a reconstruction, report

$$\mathcal R = \left( \text{raw observable}, \text{forward model}, \text{prior}, \text{resolution}, \text{uncertainty}, \text{alternatives} \right).$$

Record values without these denominators are poor comparison data.

Key Theoretical Tools

Time-dependent Schrödinger equations

Numerically solving the time-dependent Schrödinger equation is the reference for one- and few-electron model systems:

$$i\hbar\frac{\partial}{\partial t} |\Psi(t)\rangle = H(t)|\Psi(t)\rangle.$$

Validation requires convergence in spatial extent, grid spacing, angular momentum, absorber position, time step, and gauge-consistent observables. Ionization spectra are especially sensitive to boundaries and projection methods.

Strong-field and Coulomb-corrected approximations

The strong-field approximation neglects the ionic potential during continuum propagation and organizes amplitudes by saddle points. It is efficient and interpretable. It is least reliable when Coulomb focusing, low-energy structures, frustrated tunnelling, resonances, or multielectron response dominate.

Eikonal, Coulomb-quantum-orbit, analytical $R$-matrix, and trajectory-based methods restore selected Coulomb effects. Their agreement should be tested against exact model calculations where possible.

Many-electron dynamics

Time-dependent density-functional theory, time-dependent configuration interaction, multiconfiguration time-dependent Hartree–Fock, coupled-cluster response, nonequilibrium Green functions, and reduced-density-matrix methods cover different correlation and scale regimes.

No method is automatically “ab initio enough” for an attosecond claim. Important checks include:

• active-space and basis convergence;
• exchange–correlation functional sensitivity;
• core and continuum treatment;
• channel-resolved observables;
• gauge agreement after convergence; and
• comparison with stationary scattering data.

Maxwell propagation and phase matching

Single-emitter dipoles must be propagated through the gas, liquid, solid, or waveguide. Unidirectional pulse propagation, paraxial and nonparaxial Maxwell solvers, plasma dispersion, absorption, and thermal or hydrodynamic target models may be required.

Source retrieval and sample response can become coupled: a medium that generates the pulse also reshapes it. Simulating only the microscopic dipole cannot validate a far-field waveform.

Nonadiabatic molecular dynamics

Wavepacket propagation, multiconfiguration time-dependent Hartree, exact factorization, surface hopping, Ehrenfest dynamics, ab initio multiple spawning, and Gaussian basis methods trade quantum coherence against system size.

A trajectory ensemble can reproduce populations while missing geometric phase or nuclear interference. A full wavepacket can be accurate on an undersized active space. Agreement with several observables is more valuable than one preferred dynamics label.

Scattering and diffraction forward models

Photoelectron and diffraction observables require continuum states, transition dipoles, elastic and inelastic scattering, detector acceptance, orientation distributions, and instrument response. Plane-wave and independent-atom models are useful baselines, not universal truths.

For super-resolution inversion, use:

• explicit regularization;
• held-out or synthetic validation;
• uncertainty under structural priors;
• native-resolution comparisons; and
• blind tests on simulated mixtures.

Bayesian and ensemble inference

Ultrafast reconstruction is naturally probabilistic:

$$p(\theta|D) \propto p(D|\theta)p(\theta),$$

where $D$ is detector data and $\theta$ includes pulse, timing, and dynamics parameters. Posterior correlations expose whether a delay can trade against chirp, time zero, or line-shape width.

Bootstrap, profile likelihood, Bayesian sampling, and ensembles of retrieval algorithms answer different uncertainty questions. A covariance matrix from one local optimum does not capture model discrepancy.

Gauge and observable consistency

Length, velocity, and acceleration forms can converge at different rates. Strong-field calculations should state the gauge, basis transformation, observable definition, and convergence. A discrepancy between gauges after a small truncation diagnoses the approximation, not new physics.

The canonical discussion is Gauge Choices in Light–Matter Physics.

Reproducible pulse retrieval

A mature attosecond source report should archive:

• raw spectrograms and detector calibration;
• preprocessing and background subtraction;
• retrieval code and version;
• random seeds and initial guesses;
• continuum matrix elements or approximations;
• regularization and stopping rules;
• ensembles of accepted reconstructions; and
• synthetic and experimental cross-checks.

Pulse metrology is part of the scientific result, not a preliminary service step.

Representative Papers and Reviews

Foundations and broad reviews

1. P. B. Corkum, “Plasma perspective on strong-field multiphoton ionization,” Physical Review Letters 71, 1994–1997 (1993).
2. M. Lewenstein, P. Balcou, M. Y. Ivanov, A. L’Huillier, and P. B. Corkum, “Theory of high-harmonic generation by low-frequency laser fields,” Physical Review A 49, 2117–2132 (1994).
3. F. Krausz and M. Ivanov, “Attosecond physics,” Reviews of Modern Physics 81, 163–234 (2009).
4. P. B. Corkum and F. Krausz, “Attosecond science,” Nature Physics 3, 381–387 (2007).
5. R. Pazourek, S. Nagele, and J. Burgdörfer, “Attosecond chronoscopy of photoemission,” Reviews of Modern Physics 87, 765–802 (2015).
6. F. Calegari et al., “Advances in attosecond science,” Journal of Physics B 49, 062001 (2016).
7. M. Nisoli, P. Decleva, F. Calegari, A. Palacios, and F. Martín, “Attosecond electron dynamics in molecules,” Chemical Reviews 117, 10760–10825 (2017).
8. S. Ghimire and D. A. Reis, “High-harmonic generation from solids,” Nature Physics 15, 10–16 (2019).

Pulse generation and metrology

9. P. M. Paul et al., “Observation of a train of attosecond pulses from high harmonic generation,” Science 292, 1689–1692 (2001).
10. M. Hentschel et al., “Attosecond metrology,” Nature 414, 509–513 (2001).
11. E. Goulielmakis et al., “Single-cycle nonlinear optics,” Science 320, 1614–1617 (2008).
12. P. Franz et al., “Terawatt-scale attosecond X-ray pulses from a cascaded superradiant free-electron laser,” Nature Photonics 18, 698–703 (2024).
13. J. Yan et al., “Terawatt-attosecond hard X-ray free-electron laser at high repetition rate,” Nature Photonics 18, 1293–1298 (2024).
14. Y.-E. Chien et al., “Filamentation-assisted isolated attosecond pulse generation,” Nature Communications 17, 3501 (2026).
15. F. Ardana-Lamas, S. L. Cousin, J. Lignieres, and J. Biegert, “Brilliant source of 19.2-attosecond soft X-ray pulses below the atomic unit of time,” Ultrafast Science 5, 0128 (2025).
16. M. Han et al., “Comment on ‘Brilliant Source of 19.2-Attosecond Soft X-ray Pulses below the Atomic Unit of Time’,” Ultrafast Science 6 (2026).

Electron timing and molecular dynamics

17. M. Schultze et al., “Delay in photoemission,” Science 328, 1658–1662 (2010).
18. T. Driver et al., “Attosecond delays in X-ray molecular ionization,” Nature 632, 762–767 (2024).
19. E. Orunesajo et al., “Phase-resolved attoclock,” Physical Review Letters 134, 203201 (2025).
20. D. T. Matselyukh et al., “Attosecond spectroscopy of molecular charge transfer uncovers a 1.5-fs delay in population transfer,” Nature Communications 16, 7211 (2025).
21. L.-M. Koll et al., “Entanglement and electronic coherence in attosecond molecular photoionization,” Nature 652, 82–88 (2026).
22. M. Han et al., “Attosecond control and measurement of chiral photoionization dynamics,” Nature 645, 95–100 (2025).

Structural, condensed-phase, and quantum-light frontiers

23. M. Chergui and E. Collet, “Photoinduced structural dynamics of molecular systems mapped by time-resolved X-ray methods,” Chemical Reviews 117, 11025–11065 (2017).
24. H. Jiang et al., “Super-resolution femtosecond electron diffraction reveals electronic and nuclear dynamics at conical intersections,” Nature Communications 16, 6703 (2025).
25. J. H. Gaida et al., “Attosecond electron microscopy by free-electron homodyne detection,” Nature Photonics 18, 509–515 (2024).
26. G. L. Dolso et al., “Attosecond virtual charge dynamics in dielectrics,” Nature Photonics 19, 999–1005 (2025).
27. W. Tao et al., “Quantum trajectory separation and attosecond mapping in liquid high-harmonic generation,” Physical Review Letters 136, 213201 (2026).
28. Z. Jiang et al., “Nonlinear atomic tunnelling boosted by bright squeezed vacuum,” Nature 654, 356–360 (2026).

The references above support representative capabilities and interpretive boundaries. A record claim should be checked against its raw-data status, retrieval method, subsequent comments, corrections, and independent measurements.

Common Misconceptions

“An attosecond pulse directly photographs an electron”

No. The pulse initiates or probes a transition, and a detector records photons, electrons, ions, or scattering. A forward model reconstructs a reduced observable or state property.

“Pulse duration equals experimental resolution”

No. Pump duration, probe duration, jitter, crossing geometry, material response, detector bandwidth, and retrieval all contribute.

“A fitted delay smaller than the instrument response is impossible”

Not necessarily. Parameter-estimation precision can be smaller than a response width when the model is identifiable and signal-to-noise ratio is high. It does not imply arbitrary sub-response temporal imaging.

“The atomic unit of time is one hydrogen orbit”

No. $t_{\rm a}=\hbar/E_h$. The Bohr-model orbital period is $2\pi t_{\rm a}$. The atomic unit is a natural scaling quantity, not a boundary between classical and quantum dynamics.

“High-harmonic cutoff determines the pulse duration”

No. Bandwidth supports a short transform limit. Spectral phase, attochirp, trajectory interference, phase matching, filtering, and satellites determine the actual waveform.

“Short and long quantum trajectories are directly observed paths”

No. They are coherent saddle-point contributions with distinct excursion times and phase-matching behavior. Trajectory-resolved observables test that description but do not turn complex saddle times into literal tracks.

“A photoemission delay is the time an electron needs to leave”

Not generally. Scattering phase, continuum–continuum coupling, correlation, transport, resonances, and the reference channel contribute.

“A time-dependent hole density is the complete molecular state”

No. It is a reduced or model-conditioned quantity. The photoelectron, nuclear motion, orientation, and environment may be entangled with the ion.

“Super-resolution diffraction beats the instrument without assumptions”

No. It trades prior structure, sparsity, or model constraints for improved parameter localization. Native information content and prior sensitivity must be reported.

“Quantum-light enhancement is a universal efficiency gain”

No. The 2026 squeezed-vacuum result compared a specific nonlinear photoelectron momentum observable at different average pulse energies. Other yields, modes, propagation conditions, and applications require separate benchmarks.

“Observing charge migration demonstrates attochemistry”

No. Attochemistry requires a controlled intervention that changes a later nuclear pathway or product with alternatives excluded.

Exercises

Exercise 1: Bandwidth of a Gaussian attosecond pulse

Estimate the minimum intensity FWHM energy bandwidth of a transform-limited Gaussian pulse with $\Delta t=100\ {\rm as}$. Use

$$\Delta\nu\Delta t=0.441, \qquad h=4.136\times10^{-15}\ {\rm eV\,s}.$$

Solution

The frequency bandwidth is

$$\Delta\nu = \frac{0.441}{100\times10^{-18}\ {\rm s}} = 4.41\times10^{15}\ {\rm Hz}.$$

Thus

$$\Delta E = h\Delta\nu \simeq (4.136\times10^{-15})(4.41\times10^{15}) \simeq 18.2\ {\rm eV}.$$

This is a transform-limit estimate. Residual spectral phase increases the pulse duration for the same bandwidth, while filtering can introduce satellites.

Exercise 2: Keldysh parameter and harmonic cutoff

An $800\ {\rm nm}$ field has intensity $I=1.0\times10^{14}\ {\rm W\,cm^{-2}}$ and ionizes argon with $I_p=15.76\ {\rm eV}$. Use

$$U_p[{\rm eV}] \simeq 9.337\times10^{-14} I[{\rm W\,cm^{-2}}] \lambda^2[\mu{\rm m}^2].$$

Calculate $U_p$, $\gamma_K$, and the semiclassical cutoff energy.

Solution

With $\lambda=0.800\ \mu{\rm m}$,

$$U_p \simeq 9.337\times10^{-14} (1.0\times10^{14}) (0.800)^2 \simeq 5.98\ {\rm eV}.$$

Therefore

$$\gamma_K = \sqrt{\frac{15.76}{2(5.98)}} \simeq 1.15.$$

The system lies near the multiphoton–tunnelling crossover rather than deep in either limit. The cutoff estimate is

$$E_{\rm cut} \simeq 15.76+3.17(5.98) \simeq 34.7\ {\rm eV}.$$

Macroscopic phase matching and the actual single-atom spectrum determine whether this cutoff is observed cleanly.

Exercise 3: Instrument response

A Gaussian pump has rms duration $5.0\ {\rm fs}$, an XUV probe has rms duration $0.10\ {\rm fs}$, and timing jitter has rms width $0.30\ {\rm fs}$. Neglect geometry. Find $\sigma_{\rm IRF}$.

Solution

The independent Gaussian variances add:

$$\sigma_{\rm IRF} = \sqrt{ (5.0)^2+(0.10)^2+(0.30)^2 }\ {\rm fs} \simeq 5.01\ {\rm fs}.$$

The attosecond probe does not make the full experiment attosecond resolved; the pump dominates. The short probe can still provide energy or core-level selectivity and precise timing of a constrained response.

Exercise 4: Convert a RABBITT phase to delay

A sideband phase changes by $\Delta\phi=0.60\ {\rm rad}$ after reference and atomic phases are accounted for. The dressing wavelength is $800\ {\rm nm}$, with optical period $T=2.67\ {\rm fs}$. Use

$$\Delta\tau=\frac{\Delta\phi}{2\omega} = \frac{\Delta\phi\,T}{4\pi}.$$

Solution

$$\Delta\tau = \frac{(0.60)(2.67\ {\rm fs})}{4\pi} \simeq 0.127\ {\rm fs} = 127\ {\rm as}.$$

Without removing the harmonic and continuum–continuum phases, the same number would be a measured interferometric phase delay, not automatically an intrinsic emission delay.

Exercise 5: Entanglement limits ionic coherence

An ion–electron state is

$$|\Psi\rangle = \frac{1}{\sqrt2} \left( |I_1\rangle|e_1\rangle +|I_2\rangle|e_2\rangle \right).$$

If $\langle e_2|e_1\rangle=0.30e^{i\chi}$, find the magnitude of the off-diagonal ionic density-matrix element. Compare with the case of identical electron wave packets.

Solution

Tracing over the electron gives

$$\rho_{12}^{(I)} = \frac{1}{2} \langle e_2|e_1\rangle.$$

Therefore

$$\left|\rho_{12}^{(I)}\right| = 0.15.$$

If the electron wave packets were identical, their overlap magnitude would be one and $|\rho_{12}^{(I)}|=0.5$, the maximum for equal populations. The distinguishable electron carries partial which-channel information and reduces ionic coherence by a factor of $0.30$.

Exercise 6: Phase-matching coherence length

For one harmonic, the phase mismatch is $|\Delta k|=20\ {\rm m^{-1}}$. Find the coherence length. What happens to the simple buildup when the medium length is $L=2L_c$?

Solution

$$L_c = \frac{\pi}{20} \simeq 0.157\ {\rm m}.$$

At $L=2L_c$, one has $\Delta kL/2=\pi$, so the idealized uniform-medium factor

$$\operatorname{sinc}^2 \left( \frac{\Delta kL}{2} \right)$$

vanishes. Radiation produced in successive coherence lengths cancels in this oversimplified model. Real focused media have varying mismatch, absorption, and intensity, so exact zero is not expected; the calculation shows why longer is not automatically brighter.

Exercise 7: Resolution versus fitted precision

An experiment has a $1.0\ {\rm fs}$ rms instrument response and reports a relative delay $1.46\pm0.41\ {\rm fs}$ from a constrained global fit. Is the uncertainty inconsistent with the response width? What evidence is needed to trust the result?

Solution

No. The response width describes blurring of arbitrary temporal structure. The uncertainty describes precision in estimating one model parameter from many noisy data points. A line centre can likewise be estimated more precisely than its linewidth.

Trust requires an independently calibrated delay axis and instrument response, an identifiable forward model, uncertainty propagation for time zero and pulse shape, parameter-correlation analysis, residual tests, synthetic-data recovery, and comparison with plausible alternative models. The result should be stated as a model-supported relative delay, not a direct $0.41\ {\rm fs}$ shutter resolution.

Exercise 8: Rewrite an ultrafast overclaim

Rewrite:

A sub-ångström, femtosecond molecular movie directly watched a wave packet cross two conical intersections.

Solution

A defensible version is:

Time-resolved mega-electron-volt electron-diffraction data from an ensemble of photoexcited 1,3-cyclohexadiene molecules were analysed with a constrained super-resolution real-space inversion. The resulting model-assisted structural distributions resolved carbon–carbon distance differences below $0.4\ \text{\AA}$ and supported an approximately $30\ {\rm fs}$ traversal between two conical-intersection regions. The claim depends on the scattering model, structural basis, inversion, and reported uncertainty; it is not a direct measurement of one molecule’s wavefunction.

This names the raw observable, ensemble, reconstruction, result, and model boundary.

Connections to Other Volumes

Approximation, Scattering, and Semiclassics supplies WKB, stationary phase, scattering phases, semiclassical trajectories, and approximation-error logic. These tools underlie tunnelling, recollision, and photoemission-delay models.

Composite Systems and Entanglement supplies reduced density matrices and entanglement theory needed to interpret ion–photoelectron and electron–nuclear states.

Measurement and Open Quantum Systems supplies dephasing, process tensors, measurement backaction, inverse-problem cautions, and environment-induced loss of electronic coherence.

Many-Body and Statistical Physics supplies nonequilibrium response, Green functions, band dynamics, correlation, and collective excitation for ultrafast solids and liquids.

Mathematical Toolkit supplies Fourier analysis, stationary phase, ill-conditioned inversion, probability, and uncertainty propagation.

Molecular Control Frontiers owns shaped-field steering, adiabatic passage, pathway interference, and product-level coherent control. This page owns the dated attosecond and ultrafast evidence used to motivate those control questions.

What Changed This Year

Ion–photoelectron entanglement became a controlled observable

New in 2026. A phase-locked pair of isolated attosecond pulses and a few-cycle near-infrared pulse controlled the electronic coherence observed in dissociating $\mathrm{H_2^+}$ by changing its entanglement with the photoelectron. This converts a familiar theoretical caveat into an experimental control parameter.

Liquid high harmonics acquired trajectory-resolved timing

New in 2026. Spatially separated short and long liquid-HHG contributions showed opposite retrieved attochirps under a two-colour field and agreed with semiclassical recollision calculations. Generalization across liquids and scattering regimes remains open.

Quantum statistics entered isolated-atom tunnelling

New in 2026. Bright squeezed vacuum produced a matched nonlinear photoelectron momentum response with more than 20 times less average pulse energy than a coherent comparison in sodium. The demonstrated “boost” is observable- and protocol-specific, not a universal strong-field efficiency.

Filamentation became an isolated-pulse gate

New in 2026. Nonlinear propagation in semi-infinite gas cells self-compressed and spatially cleaned post-compressed ytterbium pulses while transient phase matching isolated attosecond emission. Reported helium pulses reached $65\ {\rm as}$ at $135\ {\rm eV}$.

The shortest-pulse record became a reconstruction dispute

Active and controversial in 2026. The published $19.2\ {\rm as}$ reanalysis of older streaking data received a published methodological challenge. Until continuum modelling, retrieval identifiability, and independent data converge, the duration should remain a disputed result rather than a settled record.

The 2025 molecular milestones remain the current baseline

Attosecond transient absorption separated a fitted charge-transfer delay from vibrational travel and population-transfer times in $\mathrm{CF_3I^+}$. Circular attosecond fields enabled chiral photoionization control. Super-resolution electron diffraction constrained conical-intersection structures in cyclohexadiene. Attosecond transient reflection exposed virtual interband contributions in diamond.

The next review should ask whether these capabilities have crossed from selected demonstrations to transferable methodology. The strongest evidence would combine:

$$\text{independent source metrology} +\text{multiple observables} +\text{open forward models} +\text{held-out prediction}.$$