AWS Ocelot
AWS Center for Quantum Computing
AWS Ocelot
Register profile — the machine's variant of its architecture
qubit type: dissipatively stabilised cat qubits (two-photon dissipation via buffer modes), best operating point alpha^2 ~ 1.5
gate mechanism: noise-biased cat-transmon CX with chi-matching; CX cycle 3 us, QEC cycle 2.8 us
connectivity: 1-D repetition-code chain, d=3 and d=5
control: room-temperature control
control placement: room-temp
readout: multiplexed dispersive readout of the transmon ancillas
gate evidence: a benchmark paper with its protocol
decoder mode: offline — syndromes decoded after the run
codes: cat_repetition, two-legged-cat
flags: no distance scaling shown
Technologies by layer — the machine's cell per layer
| 1 Qubit carrier | Bosonic cavity mode · primary ▣ Fig. 1a-c (storage modes S1-S5, buffers B1-B5); §II ✅ 5 cat data modes in CPW resonators (5.24–5.41 GHz), stabilised by two-photon dissipation via 5 buffer modes; storage T1 >60 µs, T2 >80 µs |
| 2 Encoding | Cat-code encoding (biased noise) · primary ▣ Fig. 1e (cat Wigner functions), Fig. 1f (bit-/phase-flip times) ✅ two-legged cat, |α|² swept 1→4, best ≈1.5; at |α|²=2 bit-flip >1 ms, phase-flip 27–33 µs, noise bias >30 (Feb 2025) |
| 3 Gate mechanism | Ancilla-mediated bosonic gates (cat CX, dual-rail CZ, beamsplitter) · primary ▣ Fig. 2 (noise-biased CX gate); §III ✅ noise-biased cat–transmon CX, ancilla in |g⟩+|f⟩ with χ-matching; CX cycle 3 µs; per-cycle bit-flip 3.5(4)e-3, phase-flip 9.6(4)e-2 |
| 4 Connectivity / transport | Static nearest-neighbour lattice · primary ▣ Fig. 1a-b,d (5 cats + 4 ancillas, tunable couplers C_ij); §II ✅ 1-D repetition-code chain of 5 cats + 4 transmon ancillas: d=3 (two 3-qubit subsections) and d=5 (full array); no 2-D code |
| 5 Control | Room-temperature electronics + per-qubit coax/flex · primary ▤ App. B (Control lines and fridge setup) 🔎 room-temperature control; ~19 microwave modes (buffers, readout, Purcell and bandpass filters, tunable couplers) on a 9-qubit device |
| 6 Readout | Dispersive microwave readout (+TWPA, Purcell) · primary ▤ ✅ multiplexed dispersive readout of the 4 fixed-frequency transmon ancillas through Purcell-filtered readout resonators |
| 7 Code | Bosonic concatenation (repetition-cat, LDPC-cat, GKP+qLDPC) · primary ▣ Fig. 3 (phase-flip correction circuit), Fig. 5 (logical memory performance); §IV-VI ✅ two-legged cat ⊗ repetition code; logical error/cycle d=3 1.75(2)%, d=5 1.65(3)% at |α|²≈1.5 ⇒ Λ≈1.06 (Feb 2025) |
| 8 Decoder | MWPM / Sparse Blossom (+ correlated matching) · primary ▤ App. H (Repetition Code Data Analysis) 🔎 1-D matching decoder run offline; no real-time decoder claimed, no decode latency published vs 2.8 µs QEC cycle — none — nothing in this layer ▤ App. H (Repetition Code Data Analysis) 🔎 no real-time decoder: 1-D matching run offline only |
| 9 Interconnect | Multi-die packaging in one module (flip-chip, bump-bonded control die) · primary ▤ App. A (Device fabrication) ✅ two dies flip-chip bonded with indium bumps in one module: Al die 16×14 mm² (junctions, buffers, transmons) + Ta die 20×20 mm² (storage modes) |
| 10 Manufacturing | Superconducting-qubit lithography (Al/AlOx/Al junctions, Nb wiring, 300 mm) · primary ▤ App. A (Device fabrication) ✅ separately fabricated Al die (nonlinear elements) and Ta die (high-Q storage) on high-resistivity Si, decoupling Ta growth from junction processing |
Standard records
2Q error under full-width parallel operation → 0.096 · typical (CX phase-flip per cycle; bit-flip 3.5e-3, bias >25)
Feb 2025 · source
idle / memory error per round → 0.0175 · typical (d=3; d=5 gives 1.65e-2, so Lambda ~1.06)
the only published two-distance bosonic-concatenated result · Feb 2025 · source
T1 relaxation → 1.0 ms · best-of-line (cat bit-flip time at photon number 2)
Feb 2025 · source
T2 coherence (echo/DD) → 27 µs · typical (phase-flip time, 27-33 us)
Feb 2025 · source
evidence: ✅ 8 / 11
Pictures
Every picture carries the story of what it shows and its credit line.