A room of funhouse mirrors: you stand in the center, and the echo of your voice comes back from an unexpected direction, amplified and distorted. That is how spin waves in a thin magnetic film, excited by an antenna, reflect off defects and irregularities, creating a tangled pattern of fields. A microscope with a diamond needle sensitive to magnetic fields tries to make out the original wave, but instead of it sees the sum of echoes from hundreds of scatterers.

It used to be thought that a map taken with such a microscope directly shows which waves exist in the film and what their wavelengths are. However, near the surface, where the probe senses the total field of all waves, this picture breaks down. The wave from the antenna and the waves scattered by defects interfere, and a false signal with zero wave vector appears in the spectrum, as if in the room of funhouse mirrors you hear a hum that matches no source.

Physicists measured these signals in a film of yttrium iron garnet (YIG) 140 nanometers thick, a material with record-low losses. They varied the distance between the probe and the sample from 200 to 50 nanometers, controlling how short the waves seen by the sensor could be. At a distance of 200 nm, the spectrum showed a single pair of peaks, as theory predicted. But at 50 nm, a "butterfly" appeared: a complex two-dimensional map where the signal was anomalously enhanced near zero wave vector. The model explained this by a cross term between the primary wave and the scattered waves: when they add up, their difference vector can turn to zero, even though both waves carry finite momentum.

This effect is comparable to how, in a room of funhouse mirrors, echoes from different walls add up into a hum that matches none of the original sounds. The scattered waves turned out to be not just a nuisance but the dominant contribution: in a geometry where the waves run perpendicular to the magnetization, their field exceeded the direct field of the antenna. In another geometry, where excitation is inefficient, the false signals disappeared, which confirmed their nature. Additional experiments showed that even a roughness of the antenna edge of 1% of its width (100 nanometers) generates noticeable scattering.

So the room of funhouse mirrors teaches us not to trust first impressions. For designing magnonic circuits, where spin waves are supposed to carry information, lithographic defects are unavoidable and must be taken into account at the design stage rather than written off as noise. The new method makes it possible to detect hidden scattering and to correct antennas so as to minimize parasitic modes. The next step is to separate the contributions of film defects and antenna roughness by comparing samples with different densities of growth pits, and to test the prediction that the cross term dominates for bulk waves when grounded transducers are used.