Odd or Even? A New Topological Test for Strongly Driven Qubits

A strongly driven qubit can return to its starting conditions carrying a hidden sign change. A September 2026 theory paper explains when that happens, why odd-photon and even-photon resonances behave differently, and how an experiment could read the difference.

The result is not a new heat engine or a battery performance record. It is a sharper diagnostic for the control landscape underneath such devices. If periodic driving is to become an engineering tool, researchers need to distinguish a genuinely protected crossing from an ordinary near-miss—and measure that distinction without the control pulse itself burying the signal.

In Topological charges and parity selection at Floquet quasienergy degeneracies, submitted on September 1, Sigmund Kohler and David Guéry-Odelin address precisely this problem. They derive a topological classification for two families of crossings in a driven two-level system, verify it numerically, and propose an interference-based measurement. This article covers the arXiv preprint; the linked record does not list a journal publication.^[1]

The central insight: two resonances can both have an exactly vanishing gap, yet only one may leave a topological sign change when the controls trace a loop around it.^[2]

Start with a qubit, then add two clocks

A two-level quantum system is the basic mathematical model behind a qubit. Without driving, its energy levels provide a relatively simple map. Turn on a repeating external field, however, and the useful description changes: researchers track Floquet modes and their quasienergies, which describe evolution under a repeating drive rather than the energy levels of a frozen snapshot.^[2]

The new paper adds a second, slower rhythm. While the field keeps oscillating, its control parameters change gradually around a closed path. The rapid clock creates the Floquet states; the slow clock carries one of those states around the control landscape. “Adiabatic” here means following a Floquet mode sufficiently slowly, not simply following an instantaneous energy level.^[2]

That distinction matters under strong driving. A convenient weak-drive approximation can reduce a resonantly driven qubit to an effective static spin, but the authors study crossings outside the regime where that shortcut reliably captures the problem. Their geometric phase is calculated using the complete periodic Floquet modes, including their motion within each drive cycle.^[2]

The crossing is a landmark, not an energy source

A degeneracy occurs when two quasienergies coincide. In a plot against two control parameters, a generic conical crossing resembles two cone tips touching. Moving away from the tip separates the branches; circling the tip can reveal a geometric phase that is invisible from a single starting-and-ending setting.^[2]

The relevant “space” is the space of laboratory controls, not necessarily the physical surface of a material. This is not another claim of electricity flowing along the edge of a topological insulator. Here, topology classifies how a driven quantum state behaves when the experimenter changes parameters around a degeneracy.^[2]

There is a recent foundation for this result. In May 2026, Kohler and Jesús Casado-Pascual published Hidden time-nonlocal Floquet symmetries in Quantum. They showed how exact crossings can arise when detuning matches an integer multiple of the drive's energy quantum. A hidden parity distinguishes the modes and allows quasienergies of different parity to cross.^[3]

The September paper asks the next question: do those crossings carry a measurable topological charge? Its answer is yes for genuine conical intersections in the model. A small loop enclosing one such crossing acquires the nontrivial Floquet–Berry phase. The derivation reduces the local problem to an effective spin in a two-dimensional field without requiring a high-frequency approximation.^[2]

Why the answer is a sign

The model's time-reversal symmetry allows the Floquet modes to be represented as real in the expanded mathematical space used for periodic motion, known as Sambe space. After a closed adiabatic loop, a real normalized state can return with its original sign or the opposite sign. This restricts the geometric phase to two possibilities, modulo a full phase cycle.^[2]

Physicists call this a Z₂-valued charge. It records a binary distinction rather than a continuously variable response. In the setting studied, enclosing an odd number of nontrivial crossings produces the sign change; enclosing an even number gives a trivial total phase. Oppositely oriented conical crossings have the same Z₂ charge because this diagnostic retains the winding only modulo two.^[2]

“Protected” therefore needs careful interpretation. It does not mean that every pulse error, environmental disturbance, or change of Hamiltonian becomes harmless. The classification assumes the relevant symmetry and a suitable loop that does not pass through a degeneracy. The actual motion also has to remain sufficiently adiabatic for the theoretical phase to be recovered.^[2]

The surprise: odd-photon and even-photon resonances differ

The authors identify another family of degeneracies at zero driving amplitude. They occur when the undriven level splitting matches an integer number, m, of drive-energy quanta. These are multiphoton resonances in the Floquet description, even though the crossing itself sits where the drive amplitude vanishes.^[2]

Near such a point, the effective coupling appears at order m in the signed drive amplitude A. Along the resonance line, corrected for the drive-induced Stark shift, the minimum quasienergy gap opens proportionally to |A|^m. For higher-order resonances, the touching is therefore not an ordinary cone in both control directions.^[2]

The key is what happens when A changes sign. The leading coupling changes sign for odd m but not for even m. Following a loop on both sides of zero amplitude makes the effective field wind around its origin in the odd case, while the even case is topologically trivial. The paper obtains a nontrivial charge for odd-photon resonances and a trivial charge for even-photon resonances under its generic assumptions.^[2]

A sign change that the laboratory must actually control

A negative amplitude is not a negative amount of light or microwave power. It refers to the sign of a real drive quadrature relative to a fixed laboratory clock. The authors explicitly treat the amplitude as signed: the experiment must carry that quadrature smoothly through zero, rather than quietly identify both signs as the same control setting.^[2]

This detail makes the proposed loop physically meaningful. Without a stable phase reference, a mathematical diagram that crosses from positive to negative amplitude could conceal an ambiguity in the controls. Here, the phase-referenced control path is part of what defines the observable charge.^[2]

How to see a phase that ordinary measurements hide

Identifying the charge on paper is only the first obstacle. During an experiment, a dynamical phase accumulates alongside the geometric phase and can dominate it. A standard echo approach is not automatically a solution: the authors explain that doubling a π geometric phase can make it indistinguishable from zero, erasing the very distinction being sought.^[2]

They instead propose a chirality-based protocol. A symmetry of the Floquet Hamiltonian relates modes with opposite quasienergies. Two symmetry-related halves of a control loop, combined with carefully timed branch-exchange pulses, cancel the dynamical contributions while preserving the geometric phase of the closed path.^[2]

The pulse timing is tied to the repeating drive. This is not simply a static-spin echo transplanted into a driven device: the protocol exploits the periodic structure of the Floquet modes. Simulated evolution supports the cancellation, while also revealing failures when the loop approaches small gaps too quickly.^[2]

Why the readout needs an extra reference level

A global sign of an isolated state is not directly visible. Comparing the two driven levels is also insufficient in the proposed scheme, because their opposite geometric phases yield a relative phase that loses the binary distinction. The authors therefore introduce an ancilla: an additional, separately controlled reference level.^[2]

A Ramsey-style sequence prepares interference between the driven state and the reference. After the control loop, another pulse converts their relative phase into measurable populations. With the reference phase properly calibrated, the two topological outcomes give distinct population signals. If that phase is unknown but stable, the authors describe a frequency-scan method in which an inversion of interference fringes identifies the charge.^[2]

The paper points to superconducting hardware with a usable third level and cold atoms with an internal reference state as possible implementation platforms. These are proposals, not reported demonstrations of this new protocol. The evidence presented is analytical derivation and numerical simulation.^[2]

What this contributes to quantum energy research

For an energy-focused reader, the immediate value is diagnostic rather than generative. The study does not calculate a battery's stored work, demonstrate refrigeration, or measure conversion efficiency. Its contribution is a way to classify and potentially measure special structures in a strongly driven quantum system.^[2]

Our engineering interpretation is that such diagnostics could help researchers check whether a control model describes the intended device. A predicted exact crossing and a tiny avoided crossing may look similar in limited-resolution spectroscopy. The authors specifically propose the persistence of the quantized phase as a loop shrinks as a diagnostic of an exact degeneracy.^[2]

The connection to heat and dissipation remains a research question. The paper suggests that dissipative response near the non-conical degeneracies should be sensitive to the odd-even selection rule, building on earlier work around conical crossings. It does not establish a resulting thermodynamic advantage. Testing that extension would connect the new classification more directly to energy exchange with an environment.^[2]

The tests that would make the result useful

A convincing next experiment should do more than display an attractive interference pattern. It should distinguish the predicted topological signal from calibration drift and ordinary dynamical phase. Based on the paper's limitations, the following checks would be especially informative:^[2]

• Compare odd and even resonances. The selection rule is the distinctive prediction; observing one isolated phase signal would not test its full content.
• Change the loop without crossing the degeneracy. This tests whether the result reflects the enclosed landmark rather than a particular waveform.
• Vary the traversal time. Small gaps can make adiabatic following difficult, especially at higher-order resonances.
• Characterize the reference phase and symmetry errors. Both the ancilla readout and the quantization rely on controlled conditions.
• Measure the cost separately. A successful topology measurement would still leave energy overhead and device usefulness to be established.

The paper's numerical geometric-phase calculation and its simulated measurement protocol also test different things. The first computes overlaps between Floquet modes around a discretized loop. The second asks whether finite-time evolution actually follows the desired state. A clean mathematical invariant does not, by itself, guarantee a clean laboratory return.^[2]

A better map before a better machine

The September result makes a precise addition to Floquet engineering: exact quasienergy crossings can carry a binary geometric signature, non-conical multiphoton crossings obey an odd-even rule, and a tailored interference protocol could expose the difference.^[1][2]

For quantum energy research, that is a useful kind of progress to recognize without overselling. Before a periodic drive can reliably charge, cool, or route energy, researchers must understand which states it creates and how their control paths behave. This work improves that map. Whether its new landmarks lead to better energy devices is the next question—not a result already in hand.

Sources and research status

Primary study: September 2026 arXiv preprint, with analytical and numerical results and a proposed experiment. The May 2026 paper is its peer-reviewed symmetry foundation. Sources checked September 16, 2026.

1. Topological charges and parity selection at Floquet quasienergy degeneracies — record
2. Kohler and Guéry-Odelin (2026) — full paper
3. Kohler and Casado-Pascual: Hidden time-nonlocal Floquet symmetries, Quantum 10, 2112 (2026)