Metal near the measurement cell causes several problems.
It heats up due to eddy currents induced by the spin dressing coil; overwhelming the dilution refrigerator and leading to unacceptable changes to the Helium-3 density through the heat flush effect.
It produces Johnson noise in the SQUIDs, reducing their sensitivity.
It can get activated by scattered neutrons, leading to γ-ray induced scintillation light that can be mistaken for neutron/Helium-3 capture events.
Nevertheless, some metal is required to produce the electric field required for the experiment. We use ion implantation to embed a thin (~200 nm) copper/germanium layer into an acrylic substrate. A silicon dioxide coating (similar to a scratch resistant coating on eyeglass lenses) underneath the metal layer improves robustness to solvents and thermal cycling.