
Imagine you have two sealed envelopes, each containing a secret message written on a grid of sticky notes. You want to transfer the message from envelope A to envelope B without ever reading it — because reading destroys it. The trick: you temporarily glue the two grids together along one edge, run a series of consistency checks on the shared border to verify the message bled through without corruption, then peel them apart. The more checks you run, the more confident you are that the transfer was clean. That's lattice surgery — and this paper just did it for real on a chip. The committed claim: this is the first experimental demonstration of fault-tolerant lattice surgery — the merge-and-split protocol that lets two surface-code logical qubits interact — on superconducting hardware. The team at ETH Zürich used 41 physical qubits to encode two distance-3 surface-code patches, merged them, ran repeated stabilizer measurement rounds to restore fault tolerance, and teleported a logical state from one patch to the other. This is not a simulation or a theoretical protocol; it is a physical experiment on a real device. Where does this sit on the ladder? The immediate predecessor is Google's 2023 demonstration of below-threshold surface-code operation on their Sycamore/Willow-class devices, plus IBM's work on distance-3 and distance-5 codes. But those experiments focused on memory — keeping a logical qubit alive — not on operations between logical qubits. Lattice surgery is the canonical method the surface-code community has converged on for performing logical gates, and demonstrating it experimentally is a distinct milestone. The distance-3 code used here is the minimum viable surface code, so the error suppression is modest, but the point is demonstrating the protocol machinery, not beating classical simulation. Architecturally, this is a stabilizer-code experiment in the topological family. Surface codes are the workhorse of the threshold-theorem approach to fault tolerance: you tile physical qubits in a 2D grid, measure local stabilizer operators (X-type and Z-type) every cycle, and use a decoder to infer and correct errors. Lattice surgery extends this by temporarily merging two code patches — expanding the stabilizer set across the boundary — to implement a logical entangling operation. The hardware is a fixed-frequency transmon chip with tunable couplers, running at ~20 mK in a dilution refrigerator. The key hardware property this leans on is the 2D nearest-neighbor connectivity that maps naturally onto the surface-code layout. The integrity picture is solid for a first demonstration. This is a real physical experiment — not a simulation — which is the gold standard for validation. The authors track logical observables step-by-step through the protocol, showing how fault tolerance degrades during the merge (when the code distance drops) and recovers with repeated stabilizer rounds afterward. They also show that optimizing the gate count and sequence duration yields further fidelity improvements, which is the kind of ablation that builds trust. The limitation is that distance-3 is too small for real error suppression; the logical error rate is not yet below the physical error rate. No pre-registration, no independent replication yet, but the device and gate sequences are described comprehensively. The milestone math is clear. Distance-3 uses ~17 data qubits per patch (41 total including ancillas and the merge region). Distance-5 — the next meaningful step — requires roughly 100+ physical qubits for two patches plus surgery region, and that's where you'd expect to see genuine error suppression from repeated stabilizer rounds. Distance-7 or higher, requiring several hundred qubits with low enough error rates, is where lattice surgery becomes a practical primitive for circuits. The timeline depends on hardware scaling: if transmon coherence and gate fidelity continue improving at their current pace, distance-5 lattice surgery is plausible within 2-3 years. The obvious experiment the authors did not run: distance-5 lattice surgery. The reason is almost certainly (a) — they ran out of qubits. A 41-qubit device is exactly sized for two distance-3 patches with a merge region. Scaling to distance-5 requires a larger chip, which is an engineering effort, not a conceptual gap. The other missing piece is a full logical circuit — using lattice surgery as a subroutine within a multi-step computation, not just a one-shot teleportation. That's likely being saved for the next paper, once the hardware supports it.
The analogy
Imagine transferring a secret message between two grids of sticky notes by temporarily gluing them together along one edge, running consistency checks on the shared border until you're confident the message bled through cleanly, then peeling them apart. That's lattice surgery: a temporary merge of two error-corrected code patches, stabilizer checks to restore integrity, then a split that completes the logical operation.
Sticky-note grids
Glue two grids along an edge, run consistency checks on the border, peel apart once the message has transferred
Lattice surgery
Merge two surface-code patches by extending stabilizers across the boundary, run repeated measurement rounds to restore fault tolerance, then split
Honest assessment
This is a genuine milestone — the first time anyone has performed lattice surgery on real quantum hardware, not just simulated it. The distance-3 code is too small for practical error suppression, so this proves the protocol works, not that it's ready for use. The real payoff comes when a future group repeats this at distance 5 or higher and shows that surgery rounds actually suppress logical errors below physical error rates. If you work in quantum error correction or superconducting hardware, this paper matters today. If you're tracking quantum computing from the outside, file it as 'the plumbing works' — necessary but not yet sufficient.
What Is CLAIMS?
77
/ 100
BedrockNews applies a structured analytical framework called CLAIMS to academic papers and technical research releases.
How CLAIMS worksClaim
4/5
Ladder
3/5
Architecture
4/5
Integrity
4/5
Milestone
4/5
Successor
4/5
What is this paper actually committing to?
Genuine first: no prior experiment has performed lattice surgery on real hardware. Distance-3 is minimal, but the protocol demonstration is the point, not the error rate.
The field fight
Can the surface-code-plus-lattice-surgery stack scale from toy demonstrations to useful fault-tolerant circuits?
Side A
Surface codes are too resource-hungry; alternative codes (color codes, qLDPC) will win at scale
Held by: Color-code advocates, qLDPC theorists
Side B
Surface codes with lattice surgery are the most hardware-compatible path to fault tolerance on 2D chips
Held by: Google, IBM, this paper (ETH Zürich)
Milestone tracker
Unlocks: Lattice surgery with real error suppression — the first step toward practical fault-tolerant logical circuits
Current
3
code distance
this paper (41 qubits)
Next
5
code distance
~100+ qubits, measurable error suppression
~2-3 years if transmon hardware scaling continues
Verified Facts (6)
41 physical qubits used in the experiment
Two distance-three logical qubits realized in a superconducting circuit
Lattice surgery used to teleport logical states between the two logical qubits
Repeated execution of stabilizer measurements after merging restores fault tolerance and yields increased fidelity
Optimizing the protocol in terms of gate count and sequence duration yields further improvements in fidelity
16 pages, 10 figures
Practical Implications (4)
Validates lattice surgery as a physically realizable primitive, not just a theoretical construct — de-risks the dominant surface-code roadmap
The gate-count and sequence-duration optimizations provide a concrete recipe for other groups building surface-code devices
Comprehensive device description lowers the barrier for independent replication on other superconducting platforms
Step-by-step observable characterization establishes a methodology for benchmarking future lattice surgery experiments at higher distances
More latest news
Read all