Optical Near-field Electron Microscopy (ONEM)

• Optical Near-field Electron Microscopy (ONEM) is a measurement paradigm that encodes optical near-fields into electron signals via methods like photocathode-LEEM and PINEM.
• It incorporates diverse architectures that range from non-invasive optical excitation with photoelectron conversion to energy-resolved free-electron interactions.
• Variants use distinct physical principles such as photoemission, energy exchange, and momentum deflection to achieve subwavelength spatial resolution in complex imaging applications.

Searching arXiv for ONEM and closely related PINEM work to ground the article in current literature. Optical Near-field Electron Microscopy (ONEM) denotes a family of microscopy concepts in which optical near-fields are encoded into electron signals and read out with electron optics at nanometric or deep-subwavelength resolution. In the literature, the term covers both a proposed hybrid architecture in which visible-light near-fields are converted into photoelectron emission at a planar photocathode and imaged by low-energy electron microscopy (LEEM), so that the specimen is never exposed to the imaging electrons, and implemented free-electron methods based on photon-induced near-field electron microscopy (PINEM), in which fast electrons traverse or pass near illuminated nanostructures and the resulting energy gain, loss, or momentum transfer is used to reconstruct optical fields inside or around the structure (Marchand et al., 2021, Fishman et al., 2022).

1. Definitions and architectural variants

ONEM is not a single instrument design but a broader measurement paradigm linking optical near-fields to electron-based readout. A useful distinction is between photocathode-conversion ONEM and free-electron interaction ONEM.

Variant	Near-field transduction	Electron readout
Photocathode-LEEM ONEM	Optical near-field converted into photoelectron flux at a nearby planar photocathode	LEEM images emitted electrons
PINEM-based ONEM	Fast electrons exchange photon quanta with an evanescent optical near-field	EELS, energy-filtered TEM, or sideband analysis
Deflection-based ONEM	Optical Lorentz force imparts transverse momentum to electrons	4D STEM/U4DSTEM detector records angular redistribution
Photoemission-based near-field imaging	Local optical field enhancement drives photoemission directly from the surface	PEEM images emitted electrons

The proposed photocathode architecture was formulated as a hybrid method combining non-invasive optical excitation with high-spatial-resolution electron-optical readout, explicitly to avoid electron-beam exposure of delicate specimens such as soft matter, biological interfaces, and electrochemical systems (Marchand et al., 2021). In that scheme, ONEM is wide-field, label-free in principle, and compatible with liquid cells.

PINEM constitutes a foundational ONEM modality because the near-field is read out through the coherent inelastic interaction of a free-electron beam with an illuminated nanostructure. In this sense, ONEM extends beyond image formation in the conventional optical sense and includes spectroscopic mapping of optical near-fields through electron energy sidebands, spatially resolved energy gain, or force-induced beam deflection (Zhou et al., 2019).

This terminological breadth is important. In some usages, ONEM refers specifically to the photocathode-LEEM concept and its later experimental realization; in others, it encompasses PINEM-family methods and related electron-near-field microscopies. A common misconception is that these are interchangeable implementations. They are not: they share the objective of mapping optical near-fields with electron-based resolution, but they differ fundamentally in where the electron interacts, what observable is measured, and whether the specimen itself ever encounters the electron beam (Marchand et al., 2021, Zykov et al., 5 Nov 2025).

2. Physical principles and measured observables

In photocathode-based ONEM, the specimen is illuminated with visible light, generating a scattered near-field whose intensity varies on nanometric scales. At the photocathode plane, the relevant optical intensity is described as

$$I(\mathbf{r}) \propto \left\| \mathbf{E}_{\mathrm{tot}}(\mathbf{r},t) \right\|^2 = \left\| \mathbf{E}_{\mathrm{ref}}(\mathbf{r},t)+\mathbf{E}_{\mathrm{scat}}(\mathbf{r},t) \right\|^2.$$

A thin, low-work-function photocathode converts this near-field pattern into a spatially varying photoelectron flux via the photoelectric effect; LEEM then images that electron distribution. Because the conversion occurs in the near-field regime, the spatial resolution is set by the object–photocathode separation and near-field decay, rather than by the optical diffraction limit (Marchand et al., 2021).

In PINEM-based ONEM, a free electron passing through or near an illuminated nanostructure exchanges energy with the optical near-field in discrete photon quanta. The central coupling is to the longitudinal field component along the electron trajectory. For dielectric laser accelerator imaging, the measured acceleration-profile amplitude is

$$|g(x,y)|=\left|\int E_z(x,y,z)\,e^{-i\omega z/v_e}\,dz\right|,$$

with the phase factor selecting the synchronous Fourier component satisfying

$q=\omega/v_e.$

The measurement is therefore mode-selective: it isolates the longitudinal spatial Fourier component relevant for synchronous acceleration rather than returning a generic optical intensity image (Fishman et al., 2022).

A second important misconception follows from this. PINEM images are not direct pictures of material geometry, nor are they always direct maps of field intensity. Energy filtering selects electrons that gained energy, so the image represents a near-field-derived electron signal whose contrast can depend on mode symmetry, phase matching, membrane-mediated interference, edge scattering, or sideband overlap (Fishman et al., 2022, Meuret et al., 2023).

Deflection-based ONEM replaces energy-sideband analysis with transverse momentum metrology. In ultrafast 4D STEM, the local observable is the integrated Lorentz-force impulse

$$\Delta p_j(x,y) = \int_{-\infty}^{\infty} F_j(x,y,t)\,dt, \qquad F_j=e\left[\mathbf{E}+(\mathbf{v}\times\mathbf{B})\right]_j.$$

This makes the method directly sensitive to transverse force components and to magnetic as well as electric contributions, in contrast to conventional PINEM, which is primarily sensitive to the longitudinal electric field component sampled along the electron trajectory (Koutenský et al., 11 Feb 2025, Koutenský et al., 22 Apr 2026).

Strong-coupling theory further enlarges the ONEM landscape. In the PINEM regime, well-separated photon sidebands appear. When sidebands overlap strongly, quantum interference can produce linear particle accelerator (LPA) behavior, i.e. net spectrum shift rather than a symmetric ladder, and with pre-interaction drift can generate anomalous PINEM (APINEM), spectral focusing, and periodic bunching in energy or momentum space (Zhou et al., 2019).

3. Instrumentation and implementation strategies

The original ONEM proposal specified a thin, visible-light-responsive photocathode close to the specimen, photoelectron extraction by an electrostatic field of approximately $100\,\mathrm{kV\,cm^{-1}}$, acceleration to about $15\,\mathrm{keV}$, and aberration-corrected LEEM readout. The proposal highlighted bialkali antimonide $\mathrm{K_2CsSb}$ at $\sim 1.9\,\mathrm{eV}$, cesiated graphene at $\sim 1.3\,\mathrm{eV}$, and a graphene support at $\sim 4.4\,\mathrm{eV}$, with green illumination at $$|g(x,y)|=\left|\int E_z(x,y,z)\,e^{-i\omega z/v_e}\,dz\right|,$$0 ($$|g(x,y)|=\left|\int E_z(x,y,z)\,e^{-i\omega z/v_e}\,dz\right|,$$1) chosen such that emitted photoelectrons would not traverse back into the specimen region (Marchand et al., 2021).

A later experimental realization implemented that architecture with 405 nm illumination, cesiated graphene or cesiated chromium photocathodes, an aberration-corrected LEEM, and for liquid studies a custom UHV-compatible liquid cell. In this realized form, the electron gun of the LEEM is off during ONEM acquisition; emitted photoelectrons alone form the image (Zykov et al., 5 Nov 2025).

Continuous-wave PINEM inside nanophotonic accelerator channels was realized in a JEOL JEM-2100 Plus transmission electron microscope with a LaB$$|g(x,y)|=\left|\int E_z(x,y,z)\,e^{-i\omega z/v_e}\,dz\right|,$$2 thermionic source modified for optical coupling. Electrons at 189 keV propagated parallel to the DLA channel while a continuous-wave 1064 nm laser illuminated the structure from above, perpendicular to the electron direction. A deliberately broadened $$|g(x,y)|=\left|\int E_z(x,y,z)\,e^{-i\omega z/v_e}\,dz\right|,$$3 beam enabled channel-wide sampling for imaging, while a $$|g(x,y)|=\left|\int E_z(x,y,z)\,e^{-i\omega z/v_e}\,dz\right|,$$4 nm beam spot enabled local spectroscopy. An EELS spectrometer with a tunable slit, opened to a 7 eV window and selecting only energy-gain electrons, generated energy-filtered TEM images of the acceleration profile (Fishman et al., 2022).

Ultrafast PINEM on membrane-supported metallic nanostructures employed a customized 200 kV cold-field-emission Hitachi HF2000 ultrafast TEM, femtosecond optical excitation delivered by a parabolic mirror at $$|g(x,y)|=\left|\int E_z(x,y,z)\,e^{-i\omega z/v_e}\,dz\right|,$$5, and a Gatan PEELS 666 spectrometer to measure the position-dependent energy spectrum and extract the local coupling constant $$|g(x,y)|=\left|\int E_z(x,y,z)\,e^{-i\omega z/v_e}\,dz\right|,$$6 (Meuret et al., 2023).

Ultrafast scanning electron microscope implementations established a complementary route. One platform used Schottky-tip photoemission, synchronized infrared excitation of dielectric nanogratings, and a spectrometer-based detection chain to study resonance, dephasing, and acceleration. Another replaced the spectrometer entirely with a Timepix3 hybrid pixel detector and 4D STEM acquisition, using 20 keV, 500–520 fs electron pulses and a 21 nm focused spot to recover transverse momentum transfer without energy filtering (Kozák et al., 2018, Koutenský et al., 11 Feb 2025).

These instrument classes correspond to distinct operating regimes: wide-field photoemission imaging, energy-filtered inelastic microscopy, and force- or deflection-based momentum imaging. Their coexistence is a defining feature of ONEM as a field.

4. Representative measurements and benchmark results

A landmark PINEM-based result was the first measurement of the field distribution inside a nanophotonic accelerator channel. Two dielectric laser accelerator designs were compared: a dual-pillar structure with distributed Bragg reflector and an inverse-designed resonant structure with an enclosed channel. Both showed strong channel-localized fields, but their symmetries differed sharply. The dual-pillar device exhibited a null at the channel center, consistent with an anti-symmetric $$|g(x,y)|=\left|\int E_z(x,y,z)\,e^{-i\omega z/v_e}\,dz\right|,$$7-like mode, whereas the inverse-designed structure showed a more symmetric $$|g(x,y)|=\left|\int E_z(x,y,z)\,e^{-i\omega z/v_e}\,dz\right|,$$8-like mode with nonzero center amplitude, appropriate for acceleration. Fits of the measured profiles with symmetric and anti-symmetric channel modes yielded $$|g(x,y)|=\left|\int E_z(x,y,z)\,e^{-i\omega z/v_e}\,dz\right|,$$9 for the dual-pillar structure and $q=\omega/v_e.$0 for the inverse-designed structure. Full 3D simulations matched the dual-pillar experiment only after imposing a pillar-diameter correction of $q=\omega/v_e.$1, whereas the inverse-designed structure matched with $q=\omega/v_e.$2 and showed greater fabrication tolerance (Fishman et al., 2022).

On membrane-supported metallic apertures, PINEM revealed that the measured coupling is not determined by the nanostructure alone. Reflection and transmission at the $q=\omega/v_e.$3 membrane, edge scattering from the metallic aperture, and interference with any nanostructure inside the aperture all contribute comparably. Control data showed that when the membrane was removed, no detectable PINEM signal appeared at the aperture center if the electron was sufficiently far from the metallic edge. This established the membrane as essential for coupling in the aperture interior and demonstrated that the signal is a coherent sum, not a simple local-intensity readout (Meuret et al., 2023).

PEEM provided a further near-field-sensitive electron-imaging modality for semiconductor nanophotonics. Broken-symmetry silicon metasurfaces supporting quasi-BIC resonances were imaged after in-situ potassium deposition lowered the work function to about $q=\omega/v_e.$4, enabling two-photon photoemission in the near-infrared. Polarization-resolved hyperspectral imaging tracked resonances A–E, and edge-to-bulk analysis showed that coupling between about eight resonators or more establishes the collective excitations of the metasurface (Boehm et al., 2023).

Photoelectron rescattering at metal nanotips represented a much more localized near-field probe. Using 800 nm, $q=\omega/v_e.$5 fs pulses and a retarding-field spectrometer, rescattered-electron cutoff energies were used to extract field enhancement factors within roughly $q=\omega/v_e.$6 of the tip surface. For tungsten, $q=\omega/v_e.$7 increased from $q=\omega/v_e.$8 at $q=\omega/v_e.$9 to $$\Delta p_j(x,y) = \int_{-\infty}^{\infty} F_j(x,y,t)\,dt, \qquad F_j=e\left[\mathbf{E}+(\mathbf{v}\times\mathbf{B})\right]_j.$$0 at $$\Delta p_j(x,y) = \int_{-\infty}^{\infty} F_j(x,y,t)\,dt, \qquad F_j=e\left[\mathbf{E}+(\mathbf{v}\times\mathbf{B})\right]_j.$$1; for gold, $$\Delta p_j(x,y) = \int_{-\infty}^{\infty} F_j(x,y,t)\,dt, \qquad F_j=e\left[\mathbf{E}+(\mathbf{v}\times\mathbf{B})\right]_j.$$2 to $$\Delta p_j(x,y) = \int_{-\infty}^{\infty} F_j(x,y,t)\,dt, \qquad F_j=e\left[\mathbf{E}+(\mathbf{v}\times\mathbf{B})\right]_j.$$3 over $$\Delta p_j(x,y) = \int_{-\infty}^{\infty} F_j(x,y,t)\,dt, \qquad F_j=e\left[\mathbf{E}+(\mathbf{v}\times\mathbf{B})\right]_j.$$4 to $$\Delta p_j(x,y) = \int_{-\infty}^{\infty} F_j(x,y,t)\,dt, \qquad F_j=e\left[\mathbf{E}+(\mathbf{v}\times\mathbf{B})\right]_j.$$5. Comparison with Maxwell simulations supported the conclusion that the enhancement is governed predominantly by geometry rather than plasmonic resonance under most conditions (Thomas et al., 2012).

The realized photocathode-LEEM ONEM platform demonstrated three application classes in a single framework: polarization-dependent plasmon modes in 100 nm Ag nanocubes, live Escherichia coli imaging in liquid with orientation-resolved contrast, and real-time copper nanocluster electrodeposition from solution. In the biological case, bacteria were imaged for hours and remained alive afterward; in the electrochemical case, nucleation, growth, coarsening, and stripping were visualized at 3 Hz (Zykov et al., 5 Nov 2025).

5. Resolution, contrast, strengths, and limitations

The proposed photocathode ONEM analysis predicted strong near-interface selectivity because Michelson contrast decays approximately as $$\Delta p_j(x,y) = \int_{-\infty}^{\infty} F_j(x,y,t)\,dt, \qquad F_j=e\left[\mathbf{E}+(\mathbf{v}\times\mathbf{B})\right]_j.$$6, suppressing background from objects not very close to the interface. For a protein-sized object at $$\Delta p_j(x,y) = \int_{-\infty}^{\infty} F_j(x,y,t)\,dt, \qquad F_j=e\left[\mathbf{E}+(\mathbf{v}\times\mathbf{B})\right]_j.$$7, the simulated signal-to-background ratio was about 1.4%. In the shot-noise-limited regime, about $$\Delta p_j(x,y) = \int_{-\infty}^{\infty} F_j(x,y,t)\,dt, \qquad F_j=e\left[\mathbf{E}+(\mathbf{v}\times\mathbf{B})\right]_j.$$8 photoelectrons per spatially resolved area were estimated to be required for signal-to-noise ratio 1; assuming about 3% conversion efficiency and unity detection efficiency, this corresponded to about $$\Delta p_j(x,y) = \int_{-\infty}^{\infty} F_j(x,y,t)\,dt, \qquad F_j=e\left[\mathbf{E}+(\mathbf{v}\times\mathbf{B})\right]_j.$$9 photons for a 1 kHz frame at 5 nm resolution, or roughly $100\,\mathrm{kV\,cm^{-1}}$0 illumination intensity. Within the simulated range, the full width at half maximum increased approximately linearly with distance $100\,\mathrm{kV\,cm^{-1}}$1, with slope $100\,\mathrm{kV\,cm^{-1}}$2, supporting nanometric and plausibly sub-5 nm resolution at sufficiently small separations (Marchand et al., 2021).

The later experimental realization reported 31 nm spatial resolution and sub-second temporal resolution, with the spatial value extracted as an upper bound from an error-function fit yielding $100\,\mathrm{kV\,cm^{-1}}$3. The paper explicitly stated that ONEM resolution is not given by the optical diffraction limit but by the distance between the object and the photocathode (Zykov et al., 5 Nov 2025).

PINEM-based imaging reaches deep-subwavelength operation for a different reason: the optical wavelength can be much larger than the channel being imaged, while the electron readout is sensitive to the synchronous evanescent mode. In the DLA study, a $100\,\mathrm{kV\,cm^{-1}}$4 nm optical field was mapped inside channels only a few hundred nanometers across (Fishman et al., 2022).

Deflection-based ONEM in U4DSTEM demonstrated 21 nm spatial resolution and avoided electron spectral filtering entirely, but the reported sensitivity limit was set by detector pixel angular size at about 0.33 mrad, corresponding to a minimum detectable field amplitude of about 1 GV/m. Background from elastically scattered electrons and optical stray light were identified as additional practical constraints (Koutenský et al., 11 Feb 2025).

Across modalities, limitations are highly architecture-dependent. Photocathode ONEM requires a custom sample chamber, in-situ photocathode coating, photocathode stability against aging, poisoning, and charging, and speckle-free optical excitation (Marchand et al., 2021). PINEM on supported nanostructures requires careful electrodynamic modeling because the measured $100\,\mathrm{kV\,cm^{-1}}$5 can encode membrane effects, edge diffraction, and coherent interference rather than a direct field amplitude (Meuret et al., 2023). In accelerator structures, scanning electron microscopy geometry alone is insufficient to predict the near-field, and 2D field models miss substrate-induced and out-of-plane effects that become evident only in 3D simulations (Fishman et al., 2022).

6. Tomography, wavefunction engineering, and future directions

A major open direction is the extension from 2D projected near-field maps to full 3D reconstruction. The accelerator-field imaging work proposed a tomography strategy based on selective illumination of individual longitudinal subsections, either through integrated apertures or by scanning a focused laser spot, followed by deconvolution of the resulting $100\,\mathrm{kV\,cm^{-1}}$6-slice measurements into a 3D field map. The same work also suggested dark-field imaging to probe transverse fields such as $100\,\mathrm{kV\,cm^{-1}}$7 (Fishman et al., 2022).

A more formal tomographic program was later developed as photon-induced near-field electron tomography (PINET). There, the measured quantity is the complex PINEM coupling

$100\,\mathrm{kV\,cm^{-1}}$8

and multiple angular projections are combined through a Radon-like inversion to reconstruct the full 3D complex vector field. The reconstruction geometry populates tangent planes in $100\,\mathrm{kV\,cm^{-1}}$9-space outside an inaccessible low-$15\,\mathrm{keV}$0 sphere set by $15\,\mathrm{keV}$1. This implies that PINET is intrinsically high-pass: it can reconstruct deeply subwavelength structure but not the full low-spatial-frequency background without additional assumptions or physics-informed regularization (Shpiro et al., 28 Oct 2025).

Another branch of development treats ONEM interactions not only as imaging probes but as electron-wavefunction control mechanisms. Strong-coupling PINEM theory predicts the emergence of LPA and APINEM regimes from sideband overlap and interference, including optical spectral focusing and periodic bunching that can improve spectral resolution by about a factor of three in the example given at optimal drift length (Zhou et al., 2019). A related proposal used structured optical near-fields as a photonic aberration corrector, imprinting a lateral phase on the electron wave function to compensate spherical aberration and to generate tailored focal profiles after monochromator selection of a sideband (Konečná et al., 2020).

Deflection-based ONEM is likewise expanding toward resonant photonic structures. U4DSTEM was used to visualize the transverse component of the Lorentz force in a periodic silicon structure designed for photonic acceleration, demonstrating that such devices can support not only longitudinal acceleration and deceleration but also strong transverse electron streaking at optical frequencies. A plausible implication is that longitudinal PINEM imaging and transverse-force U4DSTEM are complementary channels for eventual full vector-field characterization (Koutenský et al., 22 Apr 2026).

In this broader view, ONEM is evolving from a proposal for non-destructive interface imaging into a heterogeneous research area spanning wide-field photocathode imaging, PINEM-based buried-field mapping, PEEM and rescattering variants, force-based 4D STEM, tomographic reconstruction, and optical control of free-electron phase space. The unifying principle remains constant: optical near-fields are not observed directly in the far field, but are transduced into electron observables that preserve nanoscale electromagnetic structure beyond the reach of conventional diffraction-limited optics.