How do localized sources reveal hidden nonlocality in plasmonic heterodimers?

Opposite surface shifts can hide nonlocality

A plasmonic heterodimer here means two nanospheres of different plasmonic materials, with radii between 4 and 20 nm and surface-to-surface gaps between 1 and 5 nm, excited along their common axis. The paper asks whether the quantum surface response of such a pair can vanish from the measured spectrum even though each particle responds nonlocally.

The quantum correction is carried by the perpendicular Feibelman parameter d⊥, a length that measures how far the centroid of the induced charge sits from the assumed mathematical interface. Its real part is positive for electron spill-out, as in sodium, and negative for spill-in caused by enhanced screening, as in silver. Because the sign sets the direction of the spectral shift, a spill-in particle and a spill-out particle push their coupled modes in opposite directions.

The authors first test this with a toy model: two Drude spheres with the same plasma energy of 4 eV and damping of 0.1 eV, assigned constant, opposite parameters d⊥ = ±0.4 nm. In that setting the two corrections can cancel, and the resulting mesoscopic spectrum does not differ from the local-response approximation (LRA), which treats the metal surface as an abrupt boundary with no quantum correction. The cancellation is therefore a real possibility within the model, not an artifact of one material choice.

Plane waves excite both particles alike

A plane wave illuminates the whole dimer at once. Both spheres are driven by essentially the same extended field, so their contributions to the far-field spectrum are summed, and the opposing spectral shifts can offset each other. The far-field spectrum then reports the net effect rather than the response of either particle.

This is the sense in which the nonlocality is hidden. The local-response approximation and the surface-response formalism can produce nearly identical dimer spectra even when the individual particles would behave differently, because the observable is an average over both constituents. The paper's argument is that the cancellation is a property of how the structure is interrogated, not evidence that quantum surface effects are absent.

Localized sources weight the two particles unequally

The paper replaces the plane wave with two localized excitations: a point dipole placed at defined positions near the dimer, and a swift electron beam traveling along z past the structure. Both couple to the near field rather than to the whole structure uniformly, so a source placed close to one sphere drives that sphere more strongly than its partner.

This asymmetry is what makes the hidden response visible. When the dipole sits at the side of the dimer, close to one component, the surface-response total decay rate departs considerably from the LRA value and resembles the result for a dipole near that single particle. When the dipole sits at the center, it sees both spheres efficiently and the spectrum becomes richer, with the sodium resonance near 4.1 eV, silver-related resonances around 3.5 eV, and the bonding dimer plasmon at 3.3 eV all identifiable in the LRA case.

The two localized probes do not resolve the same information. Point dipoles access the complete modal landscape, while electron beams access it through higher-order modes. That difference matters for anyone choosing an experimental probe: the source determines which part of the nonlocal response becomes measurable.

Electron beams recover less than dipoles

The electron beam is an evanescent probe, and the paper finds that its coupling limits what it can reveal. Comparing LRA and surface-response results, the differences remain small regardless of where the electron passes, apart from additional broadening of higher-order coupled modes. The energy shifts themselves are described as experimentally irrelevant in this comparison.

The authors use this to argue that point dipoles may be the only one of the two localized sources that can truly disentangle the different nonlocal contributions. The near-field plots support this reading: the way the evanescent electron beam couples to the dimer restricts the modes it can address. This is a statement about the modeled probe–dimer coupling, not a general claim that electron spectroscopy cannot detect nonlocality in other geometries.

Realistic metals need a size or gap knob

The toy model uses constant, opposite Feibelman parameters. For realistic materials the authors pair sodium, treated with density-functional-theory-derived d⊥ values and a plasma energy of 5.89 eV with damping of 0.1 eV, against silver, whose d⊥ is obtained from the specular reflection model in the long-wavelength limit using experimental bulk data.

With these dispersive, material-specific parameters, the cancellation can still emerge, but only if an additional free parameter is added, such as the nanoparticle size or their separation. The paper states this explicitly in its discussion. The practical consequence is that a measured far-field spectrum matching LRA predictions does not by itself rule out quantum surface effects in a realistic heterodimer; the geometry must be specified before that inference is safe.

The authors also note a pragmatic inconsistency in their inputs: sodium parameters come from density-functional theory while silver parameters come from a hydrodynamic-level model, a mismatch justified by the scarcity of tabulated Feibelman parameters. That choice bounds how precisely the realistic-material results should be read.

The decay rate is more sensitive than the far field

Across the configurations studied, the nonlocal effects tend to cancel in the far field, while the near-field response stays sensitive to configuration. The clearest example is the decay rate of a nearby dipole, and therefore the local density of optical states it experiences.

This is the paper's most decision-relevant split. A measurement that collects far-field scattering may see a dimer that looks local, whereas an emitter placed in the gap or beside one sphere can register a substantially different decay rate under the surface-response formalism than under LRA. The authors conclude that small molecules and quantum dots can act as efficient probes of the nonclassical response, and that quantum corrections need to be included when plasmonic cavities are combined with quantum emitters.

The claim is scoped to the modeled sphere dimers and the dipole positions examined. It is a prediction of the surface-response formalism with Feibelman parameters taken from a planar interface, and the authors flag that extreme geometries might require additional geometrical corrections or a full nonlocal framework.

Choose the probe before judging the dimer

The paper's central comparison is not between materials but between interrogation methods. With plane-wave excitation, a spill-in/spill-out heterodimer can present a spectrum that matches the local-response approximation, and the quantum surface response remains hidden. With a point dipole near one constituent, the full modal landscape becomes accessible and the two particles' responses separate. With an electron beam, only higher-order modes carry a measurable difference, and the shifts are small.

For a reader deciding how to interpret or design such a measurement, the priority follows directly: specify the source and its position before concluding that a heterodimer is or is not nonlocal. In realistic sodium–silver pairs, the cancellation additionally depends on particle size or gap, so a null far-field result should be read together with the geometry. The strongest signature the paper identifies is the decay rate of an emitter close to the dimer, which stays configuration-sensitive even when the far field does not.

Sources used in this answer