The Haldane Conjecture on Spectral Gaps of Antiferromagnetic Quantum Spin Chains

The Haldane Conjecture on Spectral Gaps of Antiferromagnetic Quantum Spin Chains

OpenAI's claim

OpenAI presents a mathematical result relevant to this question in the selected manuscript. The pure spin-one antiferromagnetic Heisenberg chain is claimed to have a size-independent positive gap on even periodic rings.

Mathematical statement of the claimed result

Result 1.

For spin-one matrices satisfying $[S^a,S^b]=i\varepsilon_{abc}S^c$, $\sum_a(S^a)^2=2I$, use the even periodic ring \[ H_L=\sum_{j=0}^{L-1}\boldsymbol S_j\cdot\boldsymbol S_{j+1},\qquad\boldsymbol S_L=\boldsymbol S_0. \] Let $\gamma_L$ be the difference between its first two distinct energies. The source claims a unique ground state for even $L\ge60$, and \[ \gamma_L>\frac4{105}\log\frac{80}{79}\quad(L\ge60),\qquad \gamma_L>\frac{\log20}{784}\quad(L\ge2304), \] for even lengths. Therefore $\liminf_{L\to\infty,\ L\text{ even}}\gamma_L\ge\log20/784>0$. Simultaneously similar realizations with the same spin normalization have the same spectral conclusion.

Outline of the argument

Auxiliary twists express partition functions as spatial moments. Two purity comparisons yield a gap criterion, and two finite initializations start an iterative improvement that makes the criterion uniform in ring size.

Formal verification

Family 268 has no Lean scope page in this release. Its selected result is not presented as formalized. This site has not executed the Lean code. The scope descriptions do not certify the remaining manuscript arguments.

Status caveat

OpenAI attributes these manuscripts to work by an internal model. Its release cautions that arguments without formalization can contain mistakes. This catalogue has not independently checked the proofs; the labels report the mathematical claims attributed to OpenAI.

Sources

• OpenAI, The periodic spin-one Haldane gap, 2026-09-24; citation key OAI:The-periodic-spin-one-Haldane-gap-September-24-2026. PDF (GitHub); PDF (direct); Manuscript page and citation; Latest version; Statement source: thm:main.
• OpenAI catalogue entry.

The Haldane Conjecture on Spectral Gaps of Antiferromagnetic Quantum Spin Chains

Definitions and mathematical statement

A spin-$S$ site has Hilbert space ${\mathbb{C}}^{2S+1}$ and angular-momentum operators $S^x,S^y,S^z$. In the usual even-length periodic approximation, the antiferromagnetic chain is \[ H_L=J\sum_{j=1}^{L}\mathbf S_j\cdot\mathbf S_{j+1},\qquad J>0,\quad\mathbf S_{L+1}=\mathbf S_1. \] Let $\Delta_L$ be the energy difference between the ground energy and the next distinct energy. The proposed thermodynamic behavior is a positive gap for integer $S\ge1$, and gaplessness for $S\in\tfrac12+{\mathbb{Z}}_{\ge0}$. Periodic or infinite-chain conventions avoid confusing boundary edge modes with a bulk gap. The precise infinite-volume spectral statement is the one in the source.

Formulation and concept references: [S1].

Short English statement

Does the isotropic antiferromagnetic Heisenberg chain have a bulk energy gap exactly for integer, rather than half-integer, spin?

Research attempt

Current frontier and formulation audit.

The full accessible manuscript corresponding to the original source is arXiv:2409.09685v2. Its introduction explicitly separates the integer-spin Heisenberg conjecture from the frustration-free setting of its theorems. Its inverse-square finite-size criterion assumes frustration-freeness; Remark 1.5 also warns about periodic versus open boundaries. Those hypotheses cannot be transferred to the present Hamiltonian. [S1][E1]

The half-integer obstruction is of Lieb--Schultz--Mattis type. A precise infinite-chain theorem rules out a translation-invariant locally unique gapped ground state under the stated symmetry assumptions; it is not, for an arbitrary spin Hamiltonian, a theorem excluding symmetry-breaking gapped phases. Tasaki's review carefully distinguishes these alternatives. The source incorporated by the catalogue does not further specify which infinite-volume ground-state representation its half-integer sentence uses. The derivation below therefore proves the explicit even-periodic finite-size assertion, and does not silently identify it with every possible infinite-volume assertion. [E2]

Attack route.

Three routes were tested: an explicit twist trial state for half-integer spin; operator comparison with the spin-one AKLT parent Hamiltonian; and a continuous interpolation from that parent Hamiltonian. The first needs an orthogonality argument, not just a low trial energy. The second needs control of the actual ground energy and ground projection. The third needs a size-independent gap estimate all along the path, rather than continuity at each fixed chain length.

Attempt: a complete finite-volume calculation.

We take even $L\geq4$ and keep $J$ explicit. First we justify the ground-state input instead of assuming it. In the product basis of $S_j^z$ eigenvectors, restrict to total magnetization zero. Multiplication by \[ (-1)^{\sum_{j\ \mathrm{odd}}(S-m_j)} \] changes every nonzero off-diagonal exchange matrix element to a negative number. The configuration graph in this sector is connected: successive nearest-neighbor transfers of one unit of magnetization redistribute any configuration with the same total into any other, never exceeding the capacities $-S\leq m_j\leq S$. Equivalently, this is the connected configuration graph of a fixed number of indistinguishable units on a connected cycle, with capacity $2S$ at each site. Perron--Frobenius, applied to a sufficiently large scalar minus this matrix, gives a unique sector ground vector with strictly positive transformed coefficients.

The product of singlets on $(1,2),(3,4),\ldots,(L-1,L)$ has nonnegative transformed coefficients and nonzero overlap with this vector. Since total spin commutes with $H_L$, sector uniqueness makes the ground vector an eigenvector of total spin squared. Its overlap with a singlet makes that spin zero. Every total-spin multiplet on an even chain has integral spin and contains a zero-magnetization vector. Consequently this sector ground vector, denoted $\Omega$, is the unique global ground state and is a singlet. Translation $T$ therefore satisfies $T\Omega=e^{i\vartheta}\Omega$. These are the standard sign/positivity ingredients of the finite-chain argument; the proof above states the needed version explicitly.

Put $q=2\pi/L$ and \[ U=\exp\left(iq\sum_{j=1}^L jS_j^z\right). \] With the convention $TS_j^zT^{-1}=S_{j+1}^z$, \[ TUT^{-1}=U\exp(-iqS_{\mathrm{tot}}^z)\exp(2\pi iS_1^z). \] For half-integer $S$, the last factor is $-I$. Thus both $U\Omega$ and $U^{-1}\Omega$ have translation eigenvalue $-e^{i\vartheta}$ and are orthogonal to $\Omega$. This also covers the boundary bond, since rotation of a spin operator through $2\pi$ is the identity even when the unitary on state vectors changes sign.

Write $A_j=S_j^xS_{j+1}^x+S_j^yS_{j+1}^y$. Averaging the two twists cancels the sine/current terms exactly and gives \[ \frac12\sum_{\epsilon=\pm1} \langle U^\epsilon\Omega,(H_L-E_0)U^\epsilon\Omega\rangle =J(\cos q-1)\sum_j\langle\Omega,A_j\Omega\rangle. \] Since $\|A_j\|\leq2S^2$, at least one normalized orthogonal trial vector has energy excess at most \[ 2JS^2L(1-\cos(2\pi/L))\leq\frac{4\pi^2JS^2}{L}. \] The variational principle proves \begin{equation}\label{wsref-0-65-71-3a-33-30-30-30-36-37-33-33-3a-74-77-69-73-74} 0<\Delta_L\leq\frac{4\pi^2JS^2}{L} \qquad(S\in\tfrac12+{\mathbb{Z}}_{\geq0},\ L\geq4\text{ even}). \end{equation} The strict left inequality is finite-dimensional uniqueness, not a uniform lower bound. Equation (eq:30006733:twist) settles the gap-closing conclusion in the stated periodic approximation for every fixed half-integer $S$. It is a rederivation of an established mechanism, not a new Haldane proof. [E2]

Attempt: why a positive AKLT comparison fails.

For spin one, let $Q_j=\mathbf S_j\cdot\mathbf S_{j+1}$ and let $P_{a,j}$ project onto total bond spin $a=0,1,2$. The eigenvalues of $Q_j$ are $-2,-1,1$, respectively. Direct polynomial interpolation gives \[ P_{2,j}=\frac{Q_j^2+3Q_j+2I}{6},\qquad Q_j+2I=P_{1,j}+3P_{2,j}. \] Consequently, with $J=1$ and $H_A=\sum_jP_{2,j}$, \begin{equation}\label{wsref-0-65-71-3a-33-30-30-30-36-37-33-33-3a-63-6f-6d-70-61-72-69-73-6f-6e} H_L+2LI=3H_A+\sum_jP_{1,j}\ \geq\ 3H_A. \end{equation} The spin-one AKLT parent is a standard gapped frustration-free example within the source's framework. [E1] Nevertheless (eq:30006733:comparison) does not compare spectral gaps: it compares unshifted energies relative to different ground spaces. Already $A=\operatorname{diag}(0,1)$ and $B=\operatorname{diag}(1,1+\delta)$, $0<\delta<1$, satisfy $B\geq A$, while their gaps are $1$ and $\delta$.

In fact the shifted Heisenberg Hamiltonian is not frustration-free. The kernel of $Q_j+2I$ is the one-dimensional bond singlet. A vector that is a singlet on sites $1,2$ factors as that singlet tensor the remaining sites. Its two-site density on $2,3$ is therefore $I_{\mathcal H_2}/(2S+1)\otimes\rho_3$, which is not a pure singlet. Adjacent bond kernels have no common nonzero vector. Thus the positive extensive ground energy in (eq:30006733:comparison) is precisely what prevents the direct use of a frustration-free gap criterion.

One can interpolate using \[ H_L(t)=3H_A+t\sum_jP_{1,j},\qquad0\leq t\leq1. \] At $t=1$ this is $H_L+2LI$. At fixed $L$, eigenvalue continuity is elementary, but $\|H_L(t)-H_L(s)\|\leq L|t-s|$. Weyl's inequality therefore loses a factor $L$ when controlling the gap. A small interval of length $O(L^{-1})$ near $t=0$ is not a uniform path to $t=1$.

Here is a precise sufficient replacement. Let $P_L(t)$ be the actual ground projection and $E_L(t)$ the actual ground energy. If one could prove, with $c>0$ independent of even $L$ and $t$, that \begin{equation}\label{wsref-0-65-71-3a-33-30-30-30-36-37-33-33-3a-70-61-74-68} \langle v,(H_L(t)-E_L(t))v\rangle\geq c\|v\|^2 \quad\text{for }v\perp\operatorname{ran}P_L(t), \end{equation} then the spin-one periodic gap would follow. The obstacle is not the existence of the interpolation but the uniform estimate involving its moving ground space. This route does not even address all integer spins without further work.

Stress test and reproducibility.

The companion script constructs the local spin matrices and periodic Hamiltonians for selected small $L$ and $S$. Its eigenvalues are floating-point numerical evidence, not certified thermodynamic bounds. The identities for $P_2$ and the two-by-two gap-comparison counterexample are exact. No finite list of positive gaps is used to infer a positive limit.

The twist proof uses uniqueness to ensure that an orthogonal trial state lies above the ground energy. Without that input, orthogonality to a chosen ground vector need not exclude the rest of the ground space, especially because the catalogue defines the next distinct energy. For integer spin, $\exp(2\pi iS_1^z)=I$, so the same twist has no forced orthogonality. Finally, small periodic splittings alone are not promoted to a statement about the excitation gap above every pure infinite-chain state. [E2]

Outcome.

Outcome: Exploratory approach with unresolved gap.

The periodic half-integer gap-closing estimate is proved above with explicit constants and its ground-state hypothesis supplied. For integer spin, the attempt identifies and disproves a tempting monotonicity argument and isolates the missing size-independent moving-ground-space estimate. Neither that estimate nor the full integer-spin claim has been established. There is no claim of novelty for the half-integer branch or the AKLT facts.

Sources

• S1 Marius Lemm and Angelo Lucia, On the critical finite-size gap scaling for frustration-free Hamiltonians, Review of Mathematical Physics 37 $2025$, introduction. https://re.public.polimi.it/retrieve/2332a1fb-6d7b-4ef7-a66f-1cbb552b76f8/lemm-lucia-2025-on-the-critical-finite-size-gap-scaling-for-frustration-free-hamiltonians.pdf. Catalogue formulation source.
• E1 Marius Lemm and Angelo Lucia, On the critical finite-size gap scaling for frustration-free Hamiltonians, arXiv:2409.09685v2 $2025$, introduction, Theorem 1.4 and Remark 1.5; accessible full manuscript of [S1]. https://arxiv.org/html/2409.09685v2.
• E2 Hal Tasaki, The Lieb--Schultz--Mattis Theorem: A Topological Point of View, arXiv:2202.06243v2 $2022$, introduction, Sections 2--3, especially the finite/infinite-volume distinction and the twist-operator argument. https://arxiv.org/html/2202.06243v2.