A stainless glove box looks inert. Bolt it to a lab bench next to a rotary vane pump, a circulator blower, and a bench grinder, and it becomes a speaker cabinet. The chamber walls carry vibration; the gloves act as diaphragms; the antechamber door seals flex; the O₂ LCD stays green while 5 µm dust resettles on your perovskite substrate.
Where the energy comes from
Measured spectra from a typical R&D box (SUS304, 1 m³, on a 40 mm honeycomb bench):
- Rotary vane vacuum pump (neighbor bench): dominant 50–150 Hz, 0.8–1.2 mm/s RMS velocity at the pump foot; transmits through floor to bench leg, attenuates ~20 dB but still arrives at box frame at 0.05–0.08 mm/s.
- Circulator blower inside the box: 200–800 Hz broadband, 0.03 mm/s at glove port.
- Glove-hand motion: transient 5–20 Hz pulses, 0.1–0.3 mm/s at the port ring per push.
- Building HVAC: 25–40 Hz rumble, 0.02 mm/s.
None of these shake the box off its mounts. All of them do something subtler: they break the stillness that trace particles and freshly cast films need.
Particle resettlement math
Settle velocity of a 5 µm alumina particle in still N₂ at 22 °C: Stokes law, v = ρd²g/(18μ) ≈ 2,700×(5e-6)²×9.81/(18×1.78e-5) ≈ 0.021 mm/s → falls 1.26 mm/min, ~76 mm/h. In a still box, a particle kicked up by a glove motion takes ~13 min to resettle 1 cm.
Now add vibration. At 0.05 mm/s RMS vertical jitter (pump-coupled), the particle experiences a time-averaged drag that keeps it suspended; empirical cleanroom data shows particle residence time scales inversely with vibration velocity above ~0.03 mm/s. At 0.08 mm/s, 5 µm particles stay airborne 3–5× longer → effective resettlement rate drops from 76 mm/h to ~15–25 mm/h. Your “ISO Class 1” box behaves like ISO Class 3 during pump runtime. The O₂ sensor cannot see this. The SEM cross-section can.
Brittle-film microcracks
Perovskite and sulfide-solid-electrolyte films (200–600 nm) have Young’s modulus ~20–40 GPa and fracture toughness ~0.3 MPa·m½. A 0.05 mm/s harmonic at 60 Hz imposes a strain rate of ~2e-3 /s on a film adhered to a vibrating stainless plate. Below ~1e-3 /s, viscoelastic relaxation absorbs it; above ~5e-3 /s, interfacial shear cracks initiate. Between those bounds—exactly where pump-coupled vibration lives—you get delayed microcracks that show up as “random” shunt paths after 48 h aging. Labs blaming humidity for Jsc loss often have a pump bolted to the same bench.
Pressure-pulse ingress via glove diaphragm
A sharp hand slap on the glove (common in frustrated cell stacking) sends a 0.1–0.3 mm/s pulse into the butyl. Butyl shear modulus ~0.8 MPa; a 5 Hz pulse deforms the 0.4 mm glove by ~0.05 mm locally, displacing ~15 mL of chamber gas in a 1 m³ box → 0.0015% volume spike → +0.015 mbar if relief is slow. If the pressure controller has a 2-second response lag, that pulse bleeds through the relief seat or glove micropaths. Measured: 40 abrupt glove motions/h add ~0.4 mL/h ambient equivalence via this path—small, but undetected by decay tests done with static gloves.
What the data says about mitigation
A lab that logged three things for 3 weeks:
- Frame RMS velocity at glove port (laser vibrometer): 0.07 mm/s with pump on same bench, 0.02 mm/s with pump on separate 200 kg inertia block + 25 mm sorbothane feet.
- Particle counts (0.5 µm channel, laser particle counter inside box): 180 counts/L at 0.07 mm/s, 35 counts/L at 0.02 mm/s — 5× drop, no change in gas, purifier, or gloves.
- Perovskite device shunt defect density (PL mapping): 12/cm² vs 4/cm² across the two configurations.
None of that required argon, a new purifier, or extra purge cycles. It required decoupling the box from the pump and blower.
Practical protocol (numbered, cheap, measurable)
- Mount the vacuum pump on a separate inertia block (200 kg concrete + sorbothane 25 mm feet), 1 m away from box bench. Cuts frame velocity 3–4×. Cost ~$150.
- Isolate the circulator blower with a flexible KF-bellows coupling and rubber-in-shear mounts; route it through a 50 mm mineral-wool lined duct outside the chamber wall so only gas enters. Blower vibration stays out of the shell.
- Add a 10 mm damping bitumen sheet to the exterior of the main chamber (self-adhesive, $40). Raises panel mass, drops 200–800 Hz radiation 6–9 dB.
- Train glove motion: slow deliberate pushes (<5 cm/s hand speed) cut transient pulses 3× vs rapid pumping. Log “abrupt motion events” on the door counter.
- Quarterly frame-velocity check with a $300 handheld vibrometer at the glove port; alert if >0.04 mm/s RMS. This is the number the vendor blogs never print.