Scientists Film Ultrafast Spin Currents at Nanoscale with Terahertz Nanoscopy

Physicists have long dreamed of watching electrons and their spins move through materials on the scales where the fastest processes in nature actually unfold — femtoseconds in time and nanometers in space. Those two scales are not arbitrary: they match the dimensions of the electronic and spintronic devices envisioned to operate at terahertz rates, and they coincide with the native ranges that govern how electronic systems relax their energy, linear momentum, and angular momentum. Now, a team of researchers from Germany, China, the United States, and France reports in Light: Science & Applications that it has captured a movie of ultrafast, spin-driven charge currents with exactly this combination of resolutions, using a technique the authors call terahertz emission nanoscopy, or TEN. The work resolves a long-standing puzzle about what near-field probes are actually seeing when they image currents flowing parallel to a surface, and it opens a route toward fully vectorial, table-top imaging of nanoscale spin and charge dynamics.

The starting point for the study is a well-established measurement tool known as terahertz-emission spectroscopy. In this technique, a femtosecond optical pulse strikes a material and triggers a burst of electrical current; that transient current, in turn, radiates a pulse of terahertz electromagnetic waves into the far field. By analyzing the amplitude and polarization of the emitted terahertz radiation, researchers can reconstruct the magnitude and direction of the underlying currents, making the method a genuinely vectorial probe of ultrafast carrier motion. The catch is diffraction. Like any far-field optical approach, terahertz-emission spectroscopy cannot resolve spatial features smaller than roughly the wavelength of light, limiting its spatial resolution to the micrometer range and blinding it to local current distributions within or near the excitation spot.

Near-field scanning probes offer a way around this limit. In terahertz-emission nanoscopy, a sharp metallic tip of a scattering-type near-field optical microscope acts as a terahertz nano-antenna, confining the interaction to a region set by the tip apex — ideally tens of nanometers. The approach has already yielded striking local insights into ultrafast out-of-plane photocurrents in two-dimensional materials, semiconductors, and metallic nanostructures. But a major obstacle remained: many of the most technologically important ultrafast phenomena involve currents flowing in the plane of a thin film, coupled to the flow of electron spin. Because near-field tips are primarily sensitive to electric fields pointing perpendicular to the sample surface, it has been unclear whether — and how — terahertz-emission nanoscopy could capture these in-plane, spin-driven charge dynamics at all.

To attack the problem, the team built a clever model system: a spintronic terahertz emitter sputtered directly onto the tip of a single-mode optical fiber. The emitter is a metallic trilayer, with a 1.8-nanometer-thick ferromagnetic CoFeB layer sandwiched between 2-nanometer films of tungsten and platinum, two heavy metals whose spin-Hall angles are large but opposite in sign. When a femtosecond laser pulse travels down the fiber and hits the stack, it triggers ultrafast demagnetization of the ferromagnet and injects a burst of spin current into the adjacent heavy metals. The inverse spin Hall effect then converts those spin currents into transverse, in-plane charge currents — and because tungsten and platinum have opposite spin-Hall angles, the two charge currents add constructively, radiating a strong terahertz pulse. Mounting this fiber-coupled emitter on a three-axis piezo stage beneath the stationary tip of a near-field microscope, and raster-scanning it laterally, allowed the researchers to map the terahertz dynamics in real space.

The fiber coupling is more than a packaging convenience; it is the key to making the images quantitatively trustworthy. In conventional setups, moving the tip changes the setup response function with respect to the terahertz detector, while moving the sample changes the optical pump-path length, the excitation profile, and the excitation position. In the new design, the single-mode fiber guarantees a scan-position-independent Gaussian excitation profile with sub-wavelength dimensions, and the flexible fiber maintains a constant pump-probe delay and an invariant detection response throughout the scan. Because the optical path length barely changes as the fiber bends, the temporal delay axis is identical for every pixel of the image, so the spatiotemporal evolution of the terahertz signal can be filmed directly without post-hoc corrections of relative timing.

The resulting movies delivered a surprise. The near-field terahertz signal did not peak where the pump intensity — and hence the current density — was largest, at the center of the Gaussian excitation spot on the fiber core. Instead, the signal traced a counterintuitive dipolar pattern: two lobes of opposite sign centered on the fiber core, with the dipole axis oriented perpendicular to the in-plane magnetization of the ferromagnetic layer. The researchers interpret this pattern using the analytical solution for the electric field surrounding a point-like in-plane dipole embedded in a thin metallic film. The out-of-plane field component decays as one over the distance squared, multiplied by the cosine of the angle relative to the dipole axis — a two-lobe structure with a central zero line, exactly what the experiment shows. Because the charge-current dipole generated by inverse spin Hall conversion is expected to lie orthogonally to the magnetization, the observed orientation confirms that the images are indeed the out-of-plane signature of the spin-driven in-plane charge currents.

This mechanism also explains a persistent ambiguity in the field. Previous terahertz-emission nanoscopy studies had managed to detect signals from spintronic emitters despite the probe’s supposed lack of in-plane sensitivity. The new analysis suggests that in those free-space excitation experiments, the laser spot was likely positioned off-center relative to the tip, allowing the tip to pick up the finite out-of-plane fields at the edges of the in-plane current distribution. In the frequency domain, the team measured the spectral amplitude and phase at a center frequency of about one terahertz, consistent with the 70-femtosecond pump pulse duration. The spectral-amplitude map at one terahertz reproduced the dipolar profile, while the spectral phase showed an abrupt, nanoscale pi phase jump across the dipole’s zero line, plus a constant radial gradient — the frequency-domain fingerprint of a propagation delay along the surface, tentatively attributed to a surface plasmon polariton moving along the air-metal interface.

To put the interpretation on rigorous footing, the team turned to finite-element simulations of the full experimental geometry, including the fiber facet, the 6-nanometer metallic film, and the scanning probe, driven by a Gaussian-shaped in-plane current density. The agreement with experiment is remarkably close: simulated and measured spectral-amplitude maps at one terahertz match for scan step sizes of one micrometer and 300 nanometers, and center line scans reproduce the measured peak positions and the dip shape, with a width of roughly 400 nanometers. The simulations also show that the experiment sits near an optimum: the 10.5-micrometer mode-field diameter of the single-mode fiber maximizes the out-of-plane field generation, while the film conductivity of about one million siemens per meter lies in a regime where the signal amplitude has saturated, making the measurements robust against moderate variations in film quality. Linear fits to the phase gradient yield a subluminal phase velocity of about sixty percent of the speed of light, in line with known surface plasmon-polariton modes.

The implications stretch well beyond one device. Spintronic terahertz emitters are themselves candidate components for future terahertz data-processing technology, and understanding their current distributions at the nanoscale is a step toward engineering them more precisely. More broadly, the demonstration shows that terahertz-emission nanoscopy can serve as a fully vectorial probe, combining its established sensitivity to out-of-plane currents with newly explained access to in-plane, spin-driven charge transport. The authors foresee applications across heterogeneous material systems and nanostructures, including magnetic nanodomains in emerging altermagnets and two-dimensional magnets, as well as photocurrent heterogeneities in a wide range of systems on their native spatiotemporal scales. From ultrafast spin transport in atomically thin materials to spin-to-charge conversion at metal interfaces, the fastest and smallest currents in modern physics may soon be routinely filmed — one femtosecond and a few nanometers at a time.

Subject of Research: Nanoscale spatiotemporal imaging of ultrafast spintronic charge currents using terahertz emission nanoscopy

Article Title: Spatiotemporal terahertz emission nanoscopy of spintronic photocurrent

Article References: Spatiotemporal terahertz emission nanoscopy of spintronic photocurrent. (n.d.). https://doi.org/10.1038/s41377-026-02468-2

Image Credits: AI Generated

DOI: 10.1038/s41377-026-02468-2

Keywords: terahertz emission nanoscopy, spintronics, spintronic terahertz emitter, near-field microscopy, inverse spin Hall effect, ultrafast spin currents, femtosecond dynamics, nanoscale imaging, surface plasmon polariton, CoFeB, spin-to-charge conversion, SNOM