Gold is diamagnetic. Silver is diamagnetic. Neither should show significant conductance changes in a magnetic field. At bulk scale, theory and measurement agree: the magnetic response of noble metals is negligible.
At atomic-scale contacts — single atoms bridging two electrodes — gold shows a ~15% conductance decrease under 20-Tesla fields. Silver shows even more dramatic effects during contact formation. The theoretical prediction of “practically non-existent magnetic dependence” fails.
The through-claim: purity is a scale-dependent property. The magnetic response comes not from the gold itself but from residual oxygen molecules adsorbed near the atomic contact. In bulk, these trace impurities contribute nothing measurable — their signal drowns in the material's dominant behavior. At the atomic scale, a single adsorbed molecule near a single-atom contact generates spin-polarized currents that dominate the transport measurement.
The impurity doesn't contaminate the measurement. At this scale, it IS the measurement. What counts as “pure gold” at the macroscale is “gold plus oxygen” at the nanoscale, and the oxygen is doing most of the interesting physics.
This is not a failure of sample preparation. It's a fundamental feature of scale reduction: as the system shrinks, the surface-to-volume ratio diverges, and surface adsorbates that were negligible become the primary actors. The bulk property (diamagnetic, non-magnetic) isn't wrong — it's incomplete. It describes the material at one resolution. At another resolution, the same material has different properties because different components dominate.
Every “pure” system is pure only relative to the scale at which you're asking the question.