parameter variability

Parameter variability is the plain fact that no two devices come out of fabrication exactly alike. Build a wafer full of qubits with the same mask and the same recipe, and they still differ: each junction's resistance lands a little off, each qubit's frequency sits a little above or below where you aimed, and even their coherence times scatter. It is the quantum-chip version of something every factory knows — make a thousand of anything and you get a distribution, not a single value. The trouble is that quantum processors are far less forgiving of that spread than ordinary chips, so a few percent of variation that nobody would notice in a logic gate can quietly wreck a whole qubit lattice.

Where does the spread come from? Mostly from how thin and delicate the critical features are. A Josephson junction's behavior is set by an oxide barrier only a couple of nanometers thick, so a variation of a single atomic layer across the wafer changes its resistance, and resistance maps directly onto frequency. Lithography edges wander by a few nanometers, oxide grows slightly unevenly, films vary in thickness from center to edge, and stray two-level-system defects land in different spots on different qubits. None of these is a mistake; they are the irreducible noise of building something at the atomic scale. The result is a cloud of as-fabricated parameters scattered around the design target, and the width of that cloud is what engineers fight to shrink.

This variability is the root cause sitting underneath frequency crowding and poor yield: you cannot place a clean grid of qubit frequencies if fabrication hands you a random scatter instead. Two lines of attack help. Tighter process control — steadier oxidation, better lithography, more uniform films — narrows the spread at the source. Post-fabrication trimming, like aging or laser-annealing individual junctions, nudges stray devices back toward target afterward. Both genuinely help, but neither has eliminated the problem: across a large wafer the spread is still wide enough that hitting every qubit's target remains an open, actively worked challenge rather than a solved one.