Quantum Hardware Platforms: Coherence vs. Gate-Error Circuit Budget

Circuit requirement

Target circuit depth 200 gates

Log scale, 1 – 1,000,000 gates. This is the 2-qubit gate count your algorithm actually needs.

Gate-error roadmap today (1×)

Slide right to see each platform with engineering-roadmap error rates reduced up to 10× (2Q infidelity divided by up to 10).

Run

Platforms (real parameter ranges)

How the budget is computed

Every platform runs the same real single-qubit relaxation model used across this site's other decoherence sims: coherence decays as C(t) = exp(−t/T2), with the physical bound T2 ≤ 2×T1. Here t is expressed in units of gates executed, t = n × t_gate, so the x-axis is genuinely comparable across wildly different hardware speeds. On top of that real decay, each applied 2-qubit gate also has a real error rate (1−fidelity), so the probability the whole circuit is still usable after n gates is:

C(n)     = exp(-n * t_gate / T2)
F(n)     = fidelity ^ n
usable(n)= C(n) * F(n)
budget:  n_half = ln(0.5) / ( ln(fidelity) - t_gate/T2 )

Superconducting (transmon) — fast gates (≈30 ns) but short coherence (T1≈100µs) and the highest gate error of the four here (≈0.5%): its budget is error-limited, not coherence-limited.
Trapped-ion — slow gates (≈10µs, >300× slower than a transmon) but T1/T2 in the second range and the best demonstrated 2Q fidelity (≈99.9%): it wins on usable depth despite being the slowest.
Neutral-atom (Rydberg) — fast Rydberg-blockade gates (≈200ns) with second-scale coherence, similar fidelity to superconducting.
Photonic — coherence is essentially never the limit (a photon barely dephases before it's measured); the real bottleneck is that linear-optical 2-qubit gates succeed probabilistically (≈11% per attempt without ancilla-assisted boosting). That's modeled as an inflated effective gate time (expected attempts to get one heralded success), not as an error — a heralded photonic gate is near-perfect (≈99.9%) when it does succeed.

Parameter values are representative real-world ranges for each hardware family, not any single vendor's datasheet — move the sliders to explore how much headroom each platform actually has.