A dilution refrigerator hanging from its frame in a quiet lab looks deceptively still. It isn't. The mixing chamber plate at the bottom of the stack is sitting at sub-15 mK, the qubit chip is bonded to that plate, and every footstep on the floor above, every truck on the road outside, every compressor cycle in the building's plant room is a microscopic earthquake propagating into your transmon. Get vibration isolation wrong and you don't just lose coherence — you spend months chasing phantom T2 problems that aren't qubit physics at all, they're civil engineering. So when we started designing the Ireland Quantum 100 facility in Co. Tipperary, the building came before the chip.
Why vibration matters more for superconducting qubits than people expect
Transmon qubits are pieces of patterned aluminium or niobium on a silicon or sapphire substrate, coupled to coplanar waveguide resonators. The qubit frequency is set by a Josephson junction and a shunt capacitance. Mechanical vibration affects this in several distinct ways, and they don't all show up as the same symptom on a Ramsey trace.
First, there's direct microphonic coupling: vibrations modulate the geometry of the qubit chip and its package, which shifts capacitance, which shifts frequency. Second, vibration shakes the wiring — the coax lines running from room temperature down to the mixing chamber plate. Triboelectric noise on those lines becomes charge noise on the qubit. Third, the dilution refrigerator's pulse tube cooler is itself a vibration source; its 1.4 Hz cycle and harmonics ride straight down the cold stages unless you do something about it. Fourth, the magnetic shielding around the qubit package — typically mu-metal and superconducting cans — moves in the Earth's residual field if you let it, and a moving shield is a fluctuating field is a flux noise contribution.
The practical consequence: T1 and T2 numbers from the same chip can vary by tens of percent between a quiet day and a noisy day, with no change to control electronics or fridge state. That variability is fatal if you're trying to characterise gate fidelity, and worse if you're trying to run a surface-code experiment where every physical error budget assumption depends on stable underlying coherence.
Site selection: the building decision is a physics decision
Quantum facility design starts with the ground. Co. Tipperary, away from urban rail and heavy industry, gives us a relatively quiet seismic baseline. But "rural" is not synonymous with "quiet" — agricultural machinery, wind loading on tall structures, and nearby road traffic all generate low-frequency content in the 1–50 Hz band, which is precisely where dilution refrigerator pulse tubes and structural building modes live.
The relevant measurements before you commit to a site are ambient floor velocity spectra, taken over at least a full week to catch diurnal and weekly variation, plus a separate measurement during whatever the worst expected local activity is. You want the ambient floor to sit below the VC-E vibration criterion in the bands that matter for your fridge geometry, and ideally approaching VC-G in the qubit-relevant bands. If the ground itself fails that, no amount of platform isolation downstream will recover it cleanly — passive isolators attenuate, they don't subtract noise floors.
The other site decision is mass. A heavy building on a thick slab on competent subsoil behaves very differently from a light steel-frame structure. For a quantum building you want concrete, you want it thick, and you want it isolated from the rest of the structure that humans walk around in.
The slab-within-a-slab approach
The standard solution, and the one we're using, is an inertia block: a separate concrete mass for the cryostat that sits on its own foundation, mechanically decoupled from the surrounding floor slab. The principle is simple — make the resonant frequency of the mass-spring system formed by the inertia block and its isolators well below the lowest frequency you care about, and the block stops transmitting vibration above that corner.
The implementation details are where it gets interesting:
- The inertia block needs enough mass that the cryostat plus all its plumbing is a small perturbation. Rule of thumb: the block should be at least ten times the mass it carries.
- The isolation gap around the block is real — it's a physical air gap, sometimes filled with compressible material, that prevents structure-borne paths back into the building slab.
- Anything that crosses the gap — gas lines, cables, the helium return, the support structure for the fridge frame — has to either cross flexibly or not cross at all. Rigid bridges defeat the entire scheme.
- The isolators themselves are typically passive: pneumatic mounts or coil springs sized to give a system resonance around 1 Hz or below. Active isolation has its place but adds complexity, power, and failure modes you don't want under a fridge that takes three days to warm up and three more to cool back down.
Passive isolation, properly designed, is boring and reliable. That's exactly what you want underneath a machine that's supposed to run for years.
Decoupling the pulse tube
The dilution refrigerator's pulse tube cooler is the loudest mechanical element in the system, and it's bolted to the same fridge it's meant to keep quiet. There are a few approaches, and most modern quantum-grade fridges combine them.
The first is mechanical decoupling of the pulse tube head from the cold stages using flexible thermal links — copper braid or specifically designed soft links that conduct heat well but transmit vibration poorly. The second is mounting the pulse tube on its own bellows or flexible coupling so its 1.4 Hz reciprocation doesn't drive the fridge frame directly. The third, which fights the problem at the source, is suspending the entire cold stack from above with soft springs so it behaves as a low-frequency pendulum decoupled from both the pulse tube head and the room.
None of this is novel — it's standard practice in the field — but the integration with the building isolation matters. You don't want your inertia block tuned to 1 Hz and your fridge suspension also tuned to 1 Hz; coupled resonances at the same frequency make things worse, not better. The fridge vendor's isolation and the building's isolation need to be designed together, with the resonances spread across the band so each stage does its job.
The boring infrastructure that quietly ruins coherence
Once the cryostat sits on its inertia block and the pulse tube is decoupled, the next failure mode is everything else in the room. HVAC ducting carries fan vibration through the building. Compressors for the pulse tube and for the gas handling system are themselves significant vibration sources and need their own isolation, ideally in a separate plant room with flexible line connections. UPS systems, transformers, and chilled water pumps all radiate. Magnetic shielding inside the fridge has to be supported in a way that doesn't turn it into a vibrating drum.
And then there's the human factor. People walking, doors closing, lifts moving, deliveries arriving at the loading bay. Quantum facility design has to account for the fact that the building will be occupied — researchers and engineers need access to the control room, the wiring, the electronics rack — without their presence corrupting the experiment. The standard answer is a layered layout: cryostat hall, then a buffer zone, then the human-occupied control space, with vibration-sensitive zones on the most isolated foundation and human-occupied zones on a separate, more forgiving slab. The same logic applies to the EM environment, which deserves its own treatment — see the electromagnetic shielding strategy for how that ties together with the mechanical layout.
Measuring what you've built
The temptation when you've spent serious money on isolation is to assume it works. It doesn't, until you've measured it. The commissioning sequence we're planning runs in stages: ambient ground measurements before construction, post-construction floor and inertia block measurements with the building empty, then again with the cryostat installed but warm, then with the cryostat cold and the pulse tube running, and finally with control electronics powered and the room occupied. Each stage gives you a different transfer function, and any unexpected peak is something you'd rather find before first-light than during a fidelity benchmark.
The instrumentation is unglamorous: triaxial seismometers and geophones in the relevant bands, with continuous logging so you can correlate vibration events to qubit-level observations later. Once qubits are running, frequency noise spectroscopy on the qubit itself is the ultimate vibration sensor — anything mechanical that couples to the chip will show up as a peak in the qubit's frequency spectrum, and you can chase it back to its source.
Where to start this week
If you're planning your own quantum building, or evaluating a site for one, the single most useful thing you can do this week is put a seismometer on the floor where you think the cryostat will sit and leave it there for seven days. Not an hour, not a day — a full week, including overnight and weekend. The data will tell you more about whether the site is viable than any architectural drawing. Everything else — the inertia block, the pulse tube decoupling, the layered building layout — is engineering you can specify and buy. The ground itself is the one variable you can't change after you've poured concrete. For more on how this fits into the wider sovereign-compute build in Tipperary, see the Ireland Quantum 100 programme overview.