← Quantum Technology AtlasTechnologies · Machines · Architectures · OrganisationsРусскийעברית

Sycamore-class 72-qubit processor

Google Quantum AI

Sycamore-class 72-qubit processor

Google Quantum AI · superconducting circuits · DEMONSTRATED (2022 (Nature Feb 2023)) · 72 physical qubits · access: lab-only

Register profile — the machine's variant of its architecture

qubit type: frequency-tunable transmon
gate mechanism: tunable coupler CZ, 34 ns; 1Q 25 ns; reset 160 ns; full QEC cycle 921 ns
connectivity: square lattice, degree 4
control: room-temperature DACs
control placement: room-temp
readout: dispersive, 500 ns measurement
gate evidence: no published figure
decoder mode: offline — syndromes decoded after the run
codes: quantum_repetition, twist_defect_surface, xzzx

Technologies by layer — the machine's cell per layer

1 Qubit carrier
Transmon · primary ▣ Fig. 1a (72-qubit Sycamore device, d=5 embedding) ✅
frequency-tunable transmon, 72 qubits; T1 mean 20 µs, T2,CPMG mean 30 µs (2022)
2 Encoding
Bare physical qubit (no encoding) · primary ◌ 🔎
bare transmon two-level subspace; no encoding beyond the physical qubit
3 Gate mechanism
Tunable-coupler CZ / iSWAP · primary ▣ Fig. 1b (stabilizer circuit, CZ gates); Fig. 1c (CZ error CDF) ✅
tunable-coupler CZ 34 ns; 1Q 25 ns; reset 160 ns; full QEC cycle 921 ns (2022)
4 Connectivity / transport
Static nearest-neighbour lattice · primary ▣ Fig. 1a (square-lattice layout); Fig. 1b (four NN CZs) ✅
degree-4 square lattice, 72 transmons + 121 tunable couplers
5 Control
Room-temperature electronics + per-qubit coax/flex · primary ▤ 🔎
room-temperature DACs
6 Readout
Fast unconditional / dissipative qubit reset · primary ▣ Fig. 1b (MR step, 921-ns cycle budget); Fig. 4a (measure+reset error) ✅
reset 160 ns as a QEC-cycle budget line item (no technology); 500 ns measurement within 921 ns cycle
Dispersive microwave readout (+TWPA, Purcell) · alternate ▣ Fig. 1b (measure and reset); Fig. 1c (Meas. error CDF) ✅
dispersive readout, 500 ns measurement (2022)
7 Code
Rotated surface code (+ yoked variants) · primary ▣ Fig. 1a (d=5 code); Fig. 3 (logical error d=5 vs d=3); Fig. 4b (repetition) ✅
rotated/XZZX surface: d=5 2.914(16)% vs d=3 3.028(23)% logical error/cycle, Λ₃/₅ ≈ 1.04; d=25 repetition floor 1.7e-6/round
8 Decoder
MWPM / Sparse Blossom (+ correlated matching) · primary ▤ §Decoding and logical error probabilities ✅
matching / tensor-network decoding, offline
Neural decoders (AlphaQubit 2, CNN, transformers) · alternate ▤ 🔎
AlphaQubit neural decoder trained on this device's d=3/d=5 data; offline
— none — nothing in this layer ▤ §Decoding and logical error probabilities 🔎
decoding offline only (matching / tensor-network, AlphaQubit); no in-loop deployment
9 Interconnect
Multi-die packaging in one module (flip-chip, bump-bonded control die) · primary ◌ 🔎
single processor module in the Sycamore flip-chip lineage; no inter-module interconnect
10 Manufacturing
Superconducting-qubit lithography (Al/AlOx/Al junctions, Nb wiring, 300 mm) · primary ◌ 🔎
Superconducting-qubit lithography (Nb/Al JJ, 300 mm) — IBM Albany 300 mm / Anderon foundry; Google Santa Barbara; Rigetti ABAA trimming (97.4% frequency-targeting success); OQC 500-qubit wafer-scale…

Standard records

correlated-burst rate → 1.70e-6 per round · typical (d=25 repetition floor; 1.6e-7 excluding one high-energy event)
2022 · source
idle / memory error per round → 0.0291 · typical (logical error per cycle at d=5)
2022 · source
reset / initialisation time → 160 ns · typical
2022 · source
evidence: ✅ 7 / 13

Pictures

Every picture carries the story of what it shows and its credit line.

Fig. 1 — Implementing surface code logical qubits (schematic of a 72-qubit Sycamore device with distance-5 surface code)
Three-panel paper figure: a drawing of the 72-qubit Sycamore superconducting chip as a square grid, with 25 gold data qubits and 24 blue measure qubits forming a distance-5 surface code and a red outline marking a smaller distance-3 patch; the repeating gate circuit along a time axis; and error-rate distributions for gates, measurement and idling. From Google Quantum AI, Nature 614, 676 (2023). Notice the measure qubits between data qubits: they repeatedly check neighbours for errors without reading the data directly. “Fig. 1 — Implementing surface code logical qubits (schematic of a 72-qubit Sycamore device with distance-5 surface code)” by Google Quantum AI (Rajeev Acharya, Igor Aleiner, Richard Allen, et al.), www.nature.com/articles/s41586-022-05434-1, CC BY 4.0 (creativecommons.org/licenses/by/4.0/)