Exact solution of the three-dimensional Ising transfer-matrix eigenvalue problem: the quartic-form obstruction left open by Nambu's anticommuting-vector method

Nambu (1950) reformulated Onsager's transfer-matrix eigenvalue problem for the two-dimensional Ising model using $2N$ mutually anticommuting vector operators, reducing the transfer matrix to a product of rotations in a $2N$-dimensional real space and reproducing Onsager's spectrum in a few lines. He noted that in three dimensions the same construction produces exponentials of quartic forms (products of four anticommuting vectors) that rotations cannot diagonalize, and left the three-dimensional eigenvalue problem unsolved; he also set aside the supplementary parity conditions that matter for delicate quantities such as boundary tension, and pointed to other planar lattices and future applications. The parity bookkeeping and boundary tension were settled for the two-dimensional model by later work (Kaufman's spinor sectors, Yang's spontaneous magnetization, Watson 1967, Abraham–Gallavotti–Martin-Löf 1973); the honeycomb and triangular lattices were treated by Husimi and Syôzi themselves; and the application program flourished (Nambu's paper introduced the XY chain). The surviving open core is the exact solution of the three-dimensional Ising transfer-matrix eigenvalue problem — a closed-form thermodynamic free energy per site of the simple-cubic Ising model with its spectral structure — which remains open after three-quarters of a century: claimed exact solutions have repeatedly failed verification — most recently Zhang 2021, refuted by Perk 2023 — and the critical point and exponents are known only numerically.

Background

The Ising model places spins $\sigma=\pm 1$ on the sites of a lattice with nearest-neighbor coupling; its partition function encodes order-disorder transitions in magnets and alloys. Onsager (1944) solved the model on the two-dimensional square lattice by writing the partition function of an $N\times M$ strip as $\mathrm{Tr}\,T^M$, where $T$ is a transfer matrix acting on the $2^N$-dimensional space of configurations of one row of spins. The free energy per site in the thermodynamic limit is $-kT\ln\lambda_{\max}$, with $\lambda_{\max}$ the largest eigenvalue of $T$, so solving the model means solving an eigenvalue problem. Onsager's evaluation gave the first exactly solved phase transition in a many-particle system.

Kaufman (1949) rederived the result by spinor analysis: the transfer-matrix factors can be written as exponentials of bilinear forms in $2N$ mutually anticommuting operators $x_1,\dots,x_{2N}$ (a Clifford algebra, $\{x_r,x_s\}=2\delta_{rs}$, meaning $x_rx_s=-x_sx_r$ for $r\neq s$ and $x_r^2=1$). Such bilinear exponentials act as rotations in the $2N$-dimensional real vector space spanned by the $x_r$, and products of rotations are diagonalized by elementary linear algebra. The spectrum organizes into fermion-parity sectors (the eigenspaces split according to whether they contain an even or an odd number of fermionic mode excitations), bookkeeping that matters for boundary-sensitive quantities.

Nambu (1950), having completed the work independently by 1947, exposed this structure with unusual clarity. Representing Onsager's operators as $s_ns_{n+1}=ix_{2n}x_{2n+1}$ and $c_n=ix_{2n-1}x_{2n}$, he obtained Onsager's dispersion relation $\cosh 2\gamma_k=\cosh 2\alpha\cosh 2\beta-\sinh 2\alpha\sinh 2\beta\cos\phi_k$ in a few lines, treated the Kramers–Wannier 'screw' lattice (a two-dimensional lattice wound on a helix, introduced by Kramers and Wannier in 1941) even more simply, and in passing introduced the isotropic XY chain — a one-dimensional quantum model of spins whose couplings favor alignment in the $xy$-plane, later known simply as the XY model — which became a major exactly solvable model. His paper closes with remarks that define the present question. First, the supplementary conditions $S=\prod_n s_ns_{n+1}=1$ and $C=\prod_n c_n=\pm 1$ — constraints on the allowed eigenvectors equivalent to fixing the product of all the $x_r$ — were ignored in the derivation; they drop out of the bulk free energy but are essential for delicate quantities such as boundary tension, the line tension of an interface, which is extracted from the gap between the largest eigenvalues of different parity sectors. Second, no substantial physical applications had yet been made, and the method should extend to other planar lattices such as the honeycomb lattice then being studied by Husimi and Syôzi. Third — the remark that survives as an open problem — in three dimensions the same construction represents the transfer-matrix factors no longer as bilinear exponentials (rotations) but as exponentials of quartic forms, products of four anticommuting vectors such as $x_1x_3'x_4x_2''$ (primes marking vectors of adjacent planes); Nambu set the representation up but did not solve the resulting eigenvalue problem, and no rotation-type diagonalization of quartic forms is known.

The two-dimensional story completed itself. Kaufman's sector bookkeeping underlies Yang's (1952) exact spontaneous magnetization; Watson (1967) treated the boundary tension of two-dimensional Ising systems, and Abraham, Gallavotti and Martin-Löf (1973) established the surface tension of the planar Ising interface; Husimi and Syôzi (1950) themselves worked out the honeycomb and triangular lattice statistics; the XY chain grew into a pillar of exactly solvable physics (documented by Perk and Au-Yang 2009). The three-dimensional simple-cubic Ising model, by contrast, has never been solved: its critical point is known only numerically (for the isotropic model, the best current estimates give $\beta_cJ\approx 0.221654626$), its critical exponents only numerically, and every claimed exact solution to date has been refuted. Whether Nambu's quartic-form formalism — or any other method — can crack the three-dimensional eigenvalue problem is the surviving question.

Problem Statement

Determine exactly the eigenvalue problem for the transfer matrix of the Ising model on the simple cubic lattice (the three-dimensional Ising model) with nearest-neighbor couplings $J_x,J_y,J_z$ and zero magnetic field. The target is an exact closed-form expression for the thermodynamic-limit free energy per site — the limit of $-\ln\lambda_{\max}/(L_xL_y)$ (in units of $kT$), where $\lambda_{\max}$ is the largest eigenvalue of the transfer matrix acting on the $2^{L_xL_y}$-dimensional configuration space of one $L_x\times L_y$ plane of spins, as the cross-section grows — equivalently the exact free energy per site $f(K_x,K_y,K_z)$, with $K_i=J_i/kT$, valid for all temperatures (a formula, an integral, or a convergent series; a purely numerical evaluation is not a closed-form solution). The exact solution also determines the spectral structure of the transfer matrix — the decomposition into parity sectors and the controlling subdominant eigenvalues — from which the exact critical point and critical behavior follow, raising the three-dimensional model to the fully explicit status that Onsager, Kaufman, and Nambu established for the two-dimensional model. In Nambu's formulation the specific obstruction is that the three-dimensional transfer matrix, expressed through anticommuting vectors, contains exponentials of quartic forms (products of four of the $x_r$) which are not rotations and admit no known diagonalization. The components of Nambu's original remark that later literature settled (the supplementary parity conditions and the boundary tension of the two-dimensional model, the other planar lattices, and applications such as the XY chain) are context for this question, not part of what remains open.

References

1. Yôichirô Nambu, 'A Note on the Eigenvalue Problem in Crystal Statistics', Progress of Theoretical Physics 5(1), 1–13 (1950). DOI: 10.1143/ptp/5.1.1, https://doi.org/10.1143/ptp/5.1.1
2. Lars Onsager, 'Crystal Statistics. I. A Two-Dimensional Model with an Order-Disorder Transition', Physical Review 65, 117–149 (1944). DOI: 10.1103/PhysRev.65.117, https://doi.org/10.1103/PhysRev.65.117
3. H. A. Kramers and G. H. Wannier, 'Statistics of the Two-Dimensional Ferromagnet. Part I', Physical Review 60, 252–262 (1941). DOI: 10.1103/PhysRev.60.252, https://doi.org/10.1103/PhysRev.60.252
4. Bruria Kaufman, 'Crystal Statistics. II. Partition Function Evaluated by Spinor Analysis', Physical Review 76, 1232–1243 (1949). DOI: 10.1103/PhysRev.76.1232, https://doi.org/10.1103/PhysRev.76.1232
5. C. N. Yang, 'The Spontaneous Magnetization of a Two-Dimensional Ising Model', Physical Review 85, 808–816 (1952). DOI: 10.1103/PhysRev.85.808, https://doi.org/10.1103/PhysRev.85.808
6. Kodi Husimi and Itiro Syôzi, 'The Statistics of Honeycomb and Triangular Lattice. I', Progress of Theoretical Physics 5(2), 177–186 (1950). DOI: 10.1143/ptp.5.177, https://doi.org/10.1143/ptp.5.177
7. P. G. Watson, 'Boundary tension of two-dimensional Ising systems', Proceedings of the Physical Society 91, 940–946 (1967). DOI: 10.1088/0370-1328/91/4/322, https://doi.org/10.1088/0370-1328/91/4/322
8. D. B. Abraham, G. Gallavotti and A. Martin-Löf, 'Surface tension in the two-dimensional Ising model', Physica 65(1), 73–88 (1973). DOI: 10.1016/0031-8914(73)90234-6, https://doi.org/10.1016/0031-8914(73)90234-6
9. Lars Brink and Pierre Ramond, 'Nambu and the Ising Model' (2022). arXiv:2209.01122, DOI: 10.48550/arXiv.2209.01122, https://arxiv.org/abs/2209.01122
10. Jacques H. H. Perk and Helen Au-Yang, 'New Results for the Correlation Functions of the Ising Model and the Transverse Ising Chain', Journal of Statistical Physics 135, 599–619 (2009). DOI: 10.1007/s10955-009-9758-5, arXiv:0901.1931, https://arxiv.org/abs/0901.1931
11. Degang Zhang, 'Exact Solution for Three-Dimensional Ising Model', Symmetry 13(10), 1837 (2021). DOI: 10.3390/sym13101837, https://doi.org/10.3390/sym13101837
12. Jacques H. H. Perk, 'Comment on Zhang, D. Exact Solution for Three-Dimensional Ising Model. Symmetry 2021, 13, 1837', Symmetry 15(2), 374 (2023). DOI: 10.3390/sym15020374, https://doi.org/10.3390/sym15020374 (the arXiv preprint, arXiv:2202.03136, https://arxiv.org/abs/2202.03136, carries the longer title 'Comment on "Exact Solution for Three-Dimensional Ising Model" by Degang Zhang followed by Response to the Reply by Degang Zhang')
13. J. Ambjørn and G. W. Semenoff, 'Fermionized spin systems and the boson-fermion mapping in (2+1)-dimensional gauge theory', Physics Letters B 226, 107–112 (1989). DOI: 10.1016/0370-2693(89)90296-7, https://doi.org/10.1016/0370-2693(89)90296-7
14. Peter Orland, 'Exact solution of a quantum gauge magnet in 2+1 dimensions', Nuclear Physics B 372, 635–653 (1992). DOI: 10.1016/0550-3213(92)90369-M, https://doi.org/10.1016/0550-3213(92)90369-M
15. I. M. Ratner, 'Eigenvalues of the Transfer Matrix of the Three-Dimensional Ising Model in the Particular Case n = m = 2', Theoretical and Mathematical Physics 199, 909–921 (2019). DOI: 10.1134/S0040577919060102, https://doi.org/10.1134/S0040577919060102