When the Measurement Schedule Moves the Qubit: Floquet Majorana XYZ Codes

A quantum memory is usually described as a place to keep information still. A new Floquet-code proposal asks a subtler question: what if the physical description of that information must keep changing, and the measurement schedule decides whether it moves at all?

In Floquet Majorana XYZ Codes with Tunable Logical Dynamics, posted on September 9, 2026, Xinyu Sun and Hong Yao of Tsinghua University construct a family of periodically measured quantum codes in which the same encoded degree of freedom can have static or partly time-dependent logical operators. Their control knob is not a new material or a stronger laser. It is the sequence of measurements.[1][2]

This is a theoretical preprint, not a demonstrated quantum computer, an energy-saving benchmark, or a quantum heat engine. Its interest for Floquet research is nevertheless concrete: it connects time-programmed information protection to both local qubit descriptions and a direct microscopic Majorana construction, while exposing important limits on what those constructions prove.[2]

The new idea is not simply to repeat error checks. It is to use the choice of repeating checks to determine whether a logical operator can remain fixed.

Why a quantum memory needs a schedule

Quantum error correction spreads a logical qubit—the information we actually want to preserve—across physical degrees of freedom. Rather than directly reading that information, a controller measures collective properties that help reveal errors. In a stabilizer description, these checks constrain the allowed state without specifying the encoded message.

Floquet codes make time part of this architecture. Instead of keeping one fixed set of mutually compatible constraints, they cycle through measurement families that need not all be compatible with one another. The set of constraints available at a particular round can change, and the physical operators representing the protected information may change with it.[2]

The foundational honeycomb construction by Matthew Hastings and Jeongwan Haah made this distinction especially vivid. Viewed as a static subsystem code, their construction has no logical qubits; with the prescribed measurement pattern, it supports dynamically generated logical qubits and can function as a fault-tolerant memory.[3] The schedule therefore contributes something that a static inventory of checks misses.

The new ingredient: a memory with two kinds of logical behaviour

Sun and Yao start from the Majorana XYZ subsystem code and formulate an effective-qubit version on a periodic triangular lattice. The lattice has linear size L, with L² physical qubits. Their three-step protocol successively measures families of four-qubit parallelogram operators of X, Y and Z type.[2]

X, Y and Z label different quantum observables. For a non-specialist, the useful picture is three different kinds of collective question asked in a fixed rhythm. Within each family, the questions are compatible; switching families changes which earlier constraints survive. The paper works out that measurement algebra rather than assuming that more measurements always mean more protection.

After the initial transient, the three-step protocol has three distinct cases:[2]

• Odd lattice size: one logical qubit admits time-independent logical Pauli representatives.
• Even size of the form L = 4n + 2: two logical qubits admit time-independent representatives.
• Size of the form L = 4n: one logical qubit is static, while the second is partly dynamical: one of its logical Pauli operators can remain fixed, but its conjugate must change through the cycle.

Here, “static” concerns the physical operators used to represent logical information, not a promise that a real device suffers no noise. “Partly dynamical” also does not mean that the stored message is deliberately scrambled. Different physical operators can represent the same logical observable at different stages of the measurement cycle.[2]

A changing representation is not automatically a computation

A useful analogy is tracking a location on a map whose coordinate grid changes between frames. The location can remain meaningful even though its coordinates must be updated. In this code, however, the need to update is not merely an awkward choice of notation. For the relevant lattice-size class, the authors show that no time-independent conjugate representative exists for the second logical operator across all three measurement families.[2]

That distinction keeps the headline honest. The result establishes a constraint on logical representations and their evolution. It does not establish a universal set of computational gates, nor show that this particular changing representation is more useful than a static one.

Remove one measurement family, and the logical picture changes

The most striking comparison comes from shortening the cycle. For even L, the authors find that a two-step protocol already reaches a structure encoding two logical qubits. Repeating those two measurement families allows both logical qubits to have static representatives, including the size class that was partly dynamical under the three-step protocol.[2]

The reason is algebraic. With all three families present, the combined measurement group contains an operator that prevents the second logical qubit from having a fixed conjugate. Removing the third family removes that particular obstruction.[2] The underlying lattice alone therefore does not determine whether the logical description must move.

There is an important qualification: this is not a blanket instruction to delete checks from quantum memories. The paper treats different lattice parities separately. For odd L, the third measurement family is needed to reach the stated one-logical-qubit three-step code.[2] A simplification that works for one global lattice structure need not work for another.

For designers, the lesson is precise rather than magical: measurement schedules are genuine architectural variables. Their consequences must be calculated alongside geometry, boundaries and logical bookkeeping.

Why the direct Majorana connection matters

A mathematical qubit description is not yet a hardware recipe. Operations that look local in one description can become extended strings in another. The paper addresses that problem by mapping the construction to microscopic Majorana modes and examining what happens at the boundary.[2]

Majorana modes provide a fermionic description in which parity measurements are natural building blocks. The novelty claimed by Sun and Yao is a Floquet-code family admitting both a local qubit representation and a direct microscopic Majorana realization. They qualify that priority claim with “to our knowledge”; it should not be expanded into a claim that this is the first Majorana quantum code or the first Floquet code.[1][2]

The boundary choice is essential. On a fully periodic geometry, some operators crossing an ordering seam acquire nonlocal strings. Opening one direction produces a cylindrical geometry in which suitable protocols can use local Majorana parity checks.[2] This is a theoretical construction, not evidence that a laboratory has fabricated and operated the proposed array.

This specificity is useful. An engineering team needs to know which degrees of freedom a check touches, which boundaries are permitted, and whether missing constraints can be recovered from previous rounds. Saying only “Majorana-compatible” would hide the actual implementation problem.

The distance result—and the more demanding test still ahead

The authors establish an instantaneous code distance d = L for the steady two- and three-step protocols under the stated conditions.[2] Code distance characterizes the smallest-weight nontrivial logical error in the code description. It is an important structural measure, but it is not a complete forecast of an operating memory's reliability.

The manuscript explicitly distinguishes this instantaneous distance from the spacetime fault distance of the full measurement protocol, which it does not address.[2] A physical machine can experience wrong readouts, errors during measurement, and correlated faults across rounds. Those failure paths involve time as well as space.

The authors consequently leave realistic fault-tolerance analysis, including decoding thresholds, for future work.[2] No threshold, logical lifetime or energy advantage for this new family should be borrowed from a different Floquet code just because the names sound related.

What industry-linked honeycomb work teaches us

Earlier work by Craig Gidney and colleagues on a fault-tolerant honeycomb memory provides a useful methodological comparison. Their 2021 study used Monte Carlo sampling and a decoder to estimate performance under different noise models. It found that the relative performance of honeycomb and surface-code memories depended strongly on whether the hardware model used controlled-NOT circuits or native two-body measurements.[4]

The lesson is not that the new Majorana XYZ construction inherits those results. It does not. Rather, a promising code must eventually be judged against its intended native operations and error processes. Algebra identifies a candidate; realistic simulation and experiments determine whether the candidate protects information economically.

Where is the quantum-energy connection?

For an energy research hub, the responsible connection is indirect. Reliable quantum processors could support demanding simulations of driven materials and open quantum systems. Their protection schemes also require physical control, measurement and classical processing. A new way to organize those operations may therefore deserve investigation as part of the infrastructure supporting quantum research.

But this preprint does not calculate that infrastructure's energy consumption. A two-step cycle is not automatically cheaper than a three-step cycle: individual checks may differ in duration, hardware cost and error rate. A protocol that needs more repetitions or more difficult readout could erase an apparent saving from having fewer named stages.

A shorter measurement schedule is an opportunity to investigate resource costs, not evidence that those costs have already fallen.

For the same reason, there is no beyond-Carnot claim here. The paper is about encoded information and measurement algebra, not converting heat into work. Its relevance to Floquet.ca lies in using periodic structure as a controllable resource while keeping the resulting information-theory claims separate from thermodynamic performance.

What would make the next paper decisive?

Our assessment is that three follow-up tests would move this result closer to engineering:

1. A complete noisy measurement model. Include faulty parity outcomes, data errors and the correlations induced by the proposed physical operations.
2. A decoder and memory benchmark. Track changing logical representatives and compare failure rates across schedules at matched physical resources.
3. A hardware-level resource account. Measure or model readout time, control requirements and energy consumption per reliably preserved logical qubit over a defined storage interval.

Those are proposed evaluation criteria, not results reported in the September preprint. They would test whether the newly identified freedom in logical dynamics is practically useful, merely convenient, or costly under realistic conditions.

The immediate advance is already worth understanding without exaggeration. Sun and Yao show that the static or partly dynamical character of encoded information can depend on both global lattice relations and the periodic measurement schedule.[2] Floquet engineering here does not create energy or promise effortless computation. It changes what it means to keep quantum information under control—and gives researchers a sharply defined new architecture to test.