Spin-Zero Manifold in Exchange-Coupled Systems

• Spin-Zero Manifold is the singlet subspace of an array of exchange-coupled spin-½ electrons where the total spin S² equals 0, preserved by SU(2)-invariant Heisenberg interactions.
• Its Hilbert space scales as approximately 2^N/(N/2)^(3/2), nearly saturating the full 2^N space and providing an exponentially large computational resource compared to modular qubit encodings.
• Exchange-only control in semiconductor quantum-dot arrays facilitates singlet initialization and leakage-free operations, enabling robust quantum simulation and advanced benchmarking techniques.

The spin-zero manifold, commonly denoted $S_{0}$, is the subspace of the Hilbert space of an exchange-coupled array of $N$ spin-$\tfrac12$ electrons on which the total spin satisfies $S^{2}=0$. In the setting of semiconductor quantum-dot arrays, this manifold is operationally significant because pairs of electrons can be readily initialized into a product state of singlets, and the Heisenberg exchange interaction preserves total spin exactly. As a result, the full spin-zero manifold is available with exchange-only control, providing a Hilbert space of approximate dimension $2^N/(N/2)^{3/2}$, asymptotically close to the $2^N$ dimension of the full spin Hilbert space, while remaining leakage-free under exchange pulses (Hoffman et al., 7 Jul 2026). In a separate geometric usage, closely related terminology appears in the study of orientable hyperbolic 4-manifolds with $w_{2}\neq 0$, which do not admit spin structures (Riolo et al., 14 Oct 2025).

1. Formal definition and symmetry structure

For $N$ spin-$\tfrac12$ electrons, with $N$ even, the total spin operator is

$N$0

with $N$1, and $N$2 has eigenvalues $N$3. The spin-zero manifold $N$4 is the subspace of the full $N$5-dimensional Hilbert space on which $N$6. Equivalently, its states transform in the trivial irreducible representation of global SU(2) (Hoffman et al., 7 Jul 2026).

The defining dynamical feature of $N$7 is its invariance under Heisenberg exchange. For a pair of sites $N$8, the exchange Hamiltonian is

$N$9

Because $\tfrac12$0 is SU(2)-invariant, $\tfrac12$1, so any sequence of exchange pulses preserves $\tfrac12$2 exactly. In this sense, exchange-only control acts natively within $\tfrac12$3 and produces no leakage out of the manifold (Hoffman et al., 7 Jul 2026).

This SU(2)-invariant characterization distinguishes the spin-zero manifold from computational subspaces obtained by encoding logical qubits into fixed blocks of spins. In the latter approach, control is organized around modular qubit units; in $\tfrac12$4, the entire singlet sector of the array is treated as the computational resource. This suggests a different scaling regime for both controllable Hilbert-space dimension and benchmarking methodology.

2. Hilbert-space dimension and combinatorial structure

If the $\tfrac12$5 spins are partitioned into $\tfrac12$6 singlet pairs at initialization, then the number of ways to recombine $\tfrac12$7 spin-$\tfrac12$8 degrees of freedom into total spin zero is the $\tfrac12$9-th Catalan number,

$S^{2}=0$0

Using Stirling’s approximation $S^{2}=0$1, one derives for large $S^{2}=0$2

$S^{2}=0$3

Since $S^{2}=0$4 and $S^{2}=0$5, it follows asymptotically that

$S^{2}=0$6

Thus $S^{2}=0$7 nearly saturates the full $S^{2}=0$8 space, losing only a polynomial $S^{2}=0$9 factor (Hoffman et al., 7 Jul 2026).

This scaling is the central quantitative reason that the spin-zero manifold is treated as a useful computational resource. The full spin Hilbert space remains exponentially larger than any fixed-width encoded-qubit architecture based on dividing the array into blocks, whereas the singlet sector differs from the full space only by a subexponential reduction. A plausible implication is that architectures restricted to exchange-only control need not accept the severe Hilbert-space compression associated with traditional block encodings if the entire $2^N/(N/2)^{3/2}$0 sector can be exploited.

The paper also introduces a convenient fusion-tree basis for $2^N/(N/2)^{3/2}$1. In that basis, states are labeled $2^N/(N/2)^{3/2}$2 by the $2^N/(N/2)^{3/2}$3 bitstring $2^N/(N/2)^{3/2}$4 of pair-spins, with $2^N/(N/2)^{3/2}$5 singlet and $2^N/(N/2)^{3/2}$6 triplet, together with a degeneracy index $2^N/(N/2)^{3/2}$7 for higher-spin couplings. This basis makes explicit both the coarse-grained measurement structure used in benchmarking and the block-diagonal action of exchange operations (Hoffman et al., 7 Jul 2026).

3. Initialization and exchange-only control

In a semiconductor quantum-dot array, one can “dump” two neighboring electrons into a single double dot under strong exchange and wait for relaxation; the two-electron ground state is the singlet

$2^N/(N/2)^{3/2}$8

Repeating this for $2^N/(N/2)^{3/2}$9 creates

$2^N$0

which lies in $2^N$1 and is invariant under global rotations (Hoffman et al., 7 Jul 2026).

Exchange pulses are written

$2^N$2

Within the spin-zero manifold, these pulses generate any $2^N$3 on the $2^N$4 manifold of dimension $2^N$5. In the fusion-tree basis, diagonal pulses $2^N$6 act as single-qudit phases on $2^N$7, while off-pair exchanges shuffle the $2^N$8 labels in block-diagonal fashion (Hoffman et al., 7 Jul 2026).

The operational consequence is that initialization, control, and readout all admit descriptions intrinsic to the singlet sector. Because exchange is the only required control primitive, the approach can avoid microwave drives or magnetic-gradient control. The paper further states that semiconducting spin qubits operate at exchange-gate fidelities $2^N$9 and clock rates comparable to superconducting qubits (Hoffman et al., 7 Jul 2026). This places the spin-zero manifold in a near-term hardware context rather than purely in an abstract coding framework.

4. Relation to modular encoded qubits

Traditional exchange-only qubit encodings subdivide $w_{2}\neq 0$0 spins into $w_{2}\neq 0$1 blocks of $w_{2}\neq 0$2 spins each, yielding one logical qubit per block and a total logical Hilbert-space dimension $w_{2}\neq 0$3. For example, $w_{2}\neq 0$4 gives $w_{2}\neq 0$5. By contrast, using the full $w_{2}\neq 0$6 manifold yields dimension $w_{2}\neq 0$7 (Hoffman et al., 7 Jul 2026).

The ratio of available states is

$w_{2}\neq 0$8

which grows exponentially in $w_{2}\neq 0$9 for any fixed $N$0. In other words, $N$1 uses almost the entire $N$2 space rather than just $N$3 (Hoffman et al., 7 Jul 2026).

This comparison is central to the computational motivation for the manifold. In block encodings, the requirement of local logical structure sharply reduces accessible state space. In the spin-zero manifold, the exchange symmetry constraint still leaves an exponentially large sector. The paper therefore frames $N$4 as a route to larger computational space in a given array than traditional exchange-only control (Hoffman et al., 7 Jul 2026). A plausible implication is that the resource theory of exchange-coupled spin arrays changes substantially when the singlet sector is treated as the native computational substrate rather than as a protected initialization point for encoded qubits.

5. Benchmarking and out-of-equilibrium diagnostics in $N$5

The paper focuses on benchmarking metrics for resource utilization by generalizing cross-entropy benchmarking, mirror benchmarking, and out-of-time-ordered correlators to the spin-zero manifold (Hoffman et al., 7 Jul 2026).

For cross-entropy benchmarking, the coarse-grained probability of measuring bitstring $N$6 after running $N$7 on $N$8 is

$N$9

where $\tfrac12$0 is the multiplicity, identified as the Riordan number of that outcome. A Haar-random state in $\tfrac12$1 has $\tfrac12$2 distributed as $\tfrac12$3, and in the large-$\tfrac12$4 limit this approaches the Erlang form

$\tfrac12$5

The linear XEB estimator for fidelity $\tfrac12$6 is

$\tfrac12$7

with $\tfrac12$8 (Hoffman et al., 7 Jul 2026).

For mirror randomized benchmarking, the fidelity is the return probability under $\tfrac12$9 followed by its inverse,

$N$0

In practice, each exchange $N$1 is inverted by $N$2, reversing the pulse sequence. Because the all-singlet outcome $N$3 is nondegenerate, with $N$4, coarse-graining does not obscure the signal (Hoffman et al., 7 Jul 2026).

For out-of-time-ordered correlators, the paper defines for integer $N$5

$N$6

where $N$7 may be a product of $N$8 projectors onto singlet or triplet on consecutive pairs, and $N$9 is a butterfly operator. Expanding $N$00 in a Weyl–Heisenberg operator basis $N$01 on the $N$02-dimensional qudit, one obtains, for $N$03,

$N$04

where $N$05. As in the qubit case, the $N$06 OTOC collects large operator-loop interferences that are classically hard to sample when $N$07 and circuit depth are large (Hoffman et al., 7 Jul 2026).

The benchmarking framework is noteworthy because it is adapted to the coarse-grained observables natural to $N$08, rather than requiring full $N$09-outcome readout. The paper states that MRB and coarse-grained XEB together provide robust, leakage-free fidelity estimates, while OTOCs in $N$10 bypass the need for fine $N$11 readout (Hoffman et al., 7 Jul 2026).

6. Quantum-simulation significance and terminological distinction

The paper identifies pathways to quantum advantage based on the fact that $N$12 almost saturates the $N$13 Hilbert space. It states that random-circuit sampling, whether via XEB or OTOCs, or coherent quantum simulation of Heisenberg models, can outstrip classical simulability at smaller $N$14 than a modular qubit layout. It also states that OTOC-type experiments in $N$15 can probe scrambling, butterfly velocity, and phase transitions, including $N$16–$N$17 chains, with Trotterized exchange gates, offering a practically near-term route to a quantum computational advantage or even supremacy (Hoffman et al., 7 Jul 2026).

In this usage, “spin-zero manifold” refers to a many-body singlet sector of a spin array rather than to a differentiable manifold in the geometric sense. A distinct but related terminology appears in geometric topology and hyperbolic geometry. One 2025 construction gives a non-compact, orientable, hyperbolic four-manifold of finite volume that does not admit any spin structure (Riolo et al., 14 Oct 2025). There the obstruction is the non-vanishing second Stiefel–Whitney class $N$18, and the authors show this by exhibiting a closed surface $N$19 with self-intersection congruent to $N$20, using the Wu formula

$N$21

(Riolo et al., 14 Oct 2025).

This geometric usage does not define a spin-zero manifold in the SU(2) singlet-sector sense. Rather, it concerns the absence of spin structures on an orientable 4-manifold. The shared terminology can therefore be misleading: in quantum information, the term denotes the $N$22 subspace preserved by exchange interactions; in differential topology, the relevant question is whether the tangent bundle lifts to a spin bundle. The available sources support treating these as separate concepts linked only by the word “spin.”