Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Finite Gluing of the Three-Dimensional Ising Free Energy

Yoshiki Ueoka

Published Sep 2, 2026• 0 citations
Publisher

Abstract

We construct a single reduced free-energy generator for the simple-cubic three-dimensional Ising model by finite gluing of complementary local relations. With the convention $\Phi=+\log Z/V$, the final object is $\Phi_3(K,H,L)=\Phi_0(K)+\Delta\Phi_H(K,H)+\Phi_{\rm FSS}(K,H,L),$ from which $M=\partial_H\Phi_3$, $\chi=\partial_H^2\Phi_3$, and $U_K=\partial_K\Phi_3$ are generated. The construction covers the high- and low-temperature regimes, the critical region, finite magnetic field, strong-field saturation, periodic-film finite size, and simultaneous finite $H$ and finite $L$. The materials and operations are separated explicitly. A diagnostic dimensional stage uses $d=1,2,4$ anchors in $\mu_c=-\log\tanh K_c$ to transport critical coordinates and positive film amplitudes toward $d=3$. The practical 3D generator then protects established 3D values of $K_c$, critical exponents, selected amplitude ratios, and scalar zero-field critical data. The zero-field term joins truncated simple-cubic high- and low-temperature series through a singular critical patch with $C^2$ matching. The inner magnetic increment joins an $O(K^2)$ high-temperature field expansion, a minimal 3D parametric critical free energy fixed only by the basic 3D anchors, and a low-temperature/cluster continuation using fixed quintic switches in $K$. A recovered $z^4$ strong-field expansion is integrated to an absolute free energy. The final inner-to-strong crossover is performed by constructing a monotone beta-CDF magnetization whose endpoint values and integral reproduce the two endpoint free energies, and then integrating that derivative; thus convexity is enforced at the free-energy level rather than by clipping observables. The periodic-film excess is built from dimensionally weighted 2D and leading 4D film surfaces in common correlation-length coordinates, with an $H$-independent residual stabilizer that preserves the $H=0$ axis. A previously reconstructed full 3D parametric equation-of-state benchmark and the full 3D finite-size scaling function are withheld from construction. The pure dimensional stage gives $K_c(3)=0.22301549$ and sub-percent estimates of $\beta$, $\delta$, and $\nu$. After 3D calibration, the final implementation passes 29/29 stated structural and numerical checks. In the withheld low-field window $|K-K_c|\le0.01$, $H\le0.1$, the magnetization has RMS relative error $1.21\%$ and maximum error $4.29\%$. The result is not an exact lattice solution; it is a finite, auditable reconstruction in which the input anchors, switch intervals, matching conditions, numerical tolerances, and validation data are kept distinct.

DOI
10.5281/zenodo.22261618

Related papers

Finite Gluing of the Two-Dimensional Ising Free Energy

Yoshiki Ueoka

We test the Gluing Method on the square-lattice two-dimensional Ising model by reconstructing global free-energy generators from complementary local relations. The emphasis is kept explicit throughout: the inputs used as \emph{gluing material} are separated from quantities obtained \emph{after gluing} and from data reserved for validation. At zero magnetic field, Kramers--Wannier duality is absorbed into the coordinate $\tau=\log\sinh 2K$, so that the reduced free energy decomposes as $\Phi(\tau)=\tau/2+G(|\tau|)$. The gluing materials are then only a critical logarithmic local expansion and a low-order high-temperature outer expansion; the low-temperature sector is generated by duality rather than fitted independently. Positive flat weights preserve the prescribed critical asymptotics to every finite algebraic order. With one crossover parameter fixed from three scalar anchors, the resulting bulk free energy agrees with the exact Onsager benchmark at a maximum relative error of $1.915\times10^{-5}$ on the tested interval. Internal energy and specific heat are derived from the same glued generator. Zero-field finite-size corrections are organized separately as one-variable scaling functions multiplying $L^{-2},L^{-4},\ldots$. At nonzero field in the thermodynamic limit, the gluing materials are the established high-temperature, critical, and low-temperature local expansions of the universal Ising magnetic scaling free energy $\Phi_{\rm mag}(\eta)$, $\eta=m/|h|^{8/15}$. The low-temperature input contains the $|h|$ cusp associated with spontaneous symmetry breaking. Mapping the glued scaling free energy back to the square lattice through nonlinear scaling fields yields a single generator for the magnetic free-energy increment, magnetization, and susceptibility. The construction reproduces high-temperature linear response, the critical isotherm $M\propto |H|^{1/15}$, and the two one-sided limits $M(H\to0^\pm)=\pm M_0$ without separately fitting magnetization branches. Finally, finite size and finite field are combined in the scaling variables $x_t=Lm$ and $x_h=L^{15/8}h$. The two protected one-dimensional axes, a central mixed-response expansion, a low-temperature spectral-rounding relation, and a massive outer description are glued into a two-variable scaling surface $\mathcal Y_{\rm glue}(x_t,x_h)$. The resulting lattice free energy closes in the compact form $\Phi_L^{glue}(K,H)=\Phi_{lat}^{glue}(K,H)+\frac{\mathcal{Y}_{glue}(Lm,L^{15/8}h)- Y_{bulk}(Lm,L^{15/8}h)}{L^2}+O(L^{-4})$ where $m=m(r,H)$ and $h=h(r,H)$ are the nonlinear scaling fields and $Y_{\rm bulk}$ is the thermodynamic singular scaling surface expressed in the same variables. Thus all finite-size/finite-field information is carried by the excess $\mathcal Y_{\rm glue}-Y_{\rm bulk}$, while the thermodynamic lattice free energy remains the previously constructed $\Phi_{\rm lat}^{\rm glue}$. On eleven transfer-matrix points withheld from the construction, the blind free-energy test gives RMS absolute error $3.90\times10^{-4}$ and maximum absolute error $9.55\times10^{-4}$ in $\mathcal Y$. Differentiation of the same free energy produces finite-size magnetization and susceptibility; magnetization is especially stable on the low-temperature side, while the second derivative is more sensitive in crossover regions.

•Sep 2, 2026•0 citations

Free energy in a magnetic field and the universal scaling equation of state for the three-dimensional Ising model

P. Butera, M. Pernici

We have substantially extended the high-temperature and low-magnetic-field (and the related low-temperature and high-magnetic-field) bivariate expansions of the free energy for the conventional three-dimensional Ising model and for a variety of other spin systems generally assumed to belong to the same critical universality class. In particular, we have also derived the analogous expansions for the Ising models with spin $s=1,3/2,\dots{}$ and for the lattice Euclidean scalar-field theory with quartic self-interaction, on the simple-cubic and the body-centered-cubic lattices. Our bivariate high-temperature expansions, which extend through 24th order, enable us to compute, through the same order, all higher derivatives of the free energy with respect to the field, namely, all higher susceptibilities. These data make more accurate checks possible, in critical conditions, both of the scaling and the universality properties with respect to the lattice and the interaction structure and also help to improve an approximate parametric representation of the critical equation of state for the three-dimensional Ising model universality class.

arXiv:1012.5004•Physical Review B•Feb 25, 2011•17 citations

25th-order high-temperature expansion results for three-dimensional Ising-like systems on the simple-cubic lattice

Massimo Campostrini B, Andrea Pelissetto C, Paolo Rossi +1 more

25th-order high-temperature series are computed for a general nearest-neighbor three-dimensional Ising model with arbitrary potential on the simple cubic lattice. In particular, we consider three improved potentials characterized by suppressed leading scaling corrections. Critical exponents are extracted from high-temperature series specialized to improved potentials, obtaining $\ensuremath{\gamma}=1.2373(2),$ $\ensuremath{\nu}=0.63012(16),$ $\ensuremath{\alpha}=0.1096(5),$ $\ensuremath{\eta}=0.03639(15),$ $\ensuremath{\beta}=0.32653(10),$ and $\ensuremath{\delta}=4.7893(8).$ Moreover, biased analyses of the 25th-order series of the standard Ising model provide the estimate $\ensuremath{\Delta}=0.52(3)$ for the exponent associated with the leading scaling corrections. By the same technique, we study the small-magnetization expansion of the Helmholtz free energy. The results are then applied to the construction of parametric representations of the critical equation of state, using a systematic approach based on a global stationarity condition. Accurate estimates of several universal amplitude ratios are also presented.

arXiv:cond-mat/0201180•Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics•Jun 27, 2002•262 citations

Three-dimensional randomly dilute Ising model: Monte Carlo results

Pasquale Calabrese, Víctor Martín-Mayor, Andrea Pelissetto C +1 more

We perform a high-statistics simulation of the three-dimensional randomly dilute Ising model on cubic lattices L3 with L< or =256. We choose a particular value of the density, x=0.8, for which the leading scaling corrections are suppressed. We determine the critical exponents, obtaining nu=0.683(3), eta=0.035(2), beta=0.3535(17), and alpha=-0.049(9), in agreement with previous numerical simulations. We also estimate numerically the fixed-point values of the four-point zero-momentum couplings that are used in field-theoretical fixed-dimension studies. Although these results somewhat differ from those obtained using perturbative field theory, the field-theoretical estimates of the critical exponents do not change significantly if the Monte Carlo result for the fixed point is used. Finally, we determine the six-point zero-momentum couplings, relevant for the small-magnetization expansion of the equation of state, and the invariant amplitude ratio R(+)(xi) that expresses the universality of the free-energy density per correlation volume. We find R(+)(xi)=0.2885(15).

arXiv:cond-mat/0306272•Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics•Sep 30, 2003•54 citations

Critical finite-size-scaling amplitudes of a fully anisotropic three-dimensional Ising model

M. A. Yurishchev

A fully anisotropic simple-cubic Ising lattice in the geometry of periodic cylinders n\ifmmode\times\else\texttimes\fi{}n\ifmmode\times\else\texttimes\fi{}\ensuremath{\infty} is investigated by the transfer-matrix finite-size-scaling method. In addition to the previously obtained critical amplitudes of the inverse correlation lengths and singular part of the free energy [M. A. Yurishchev, Phys. Rev. B 50, 13 533 (1994)], the amplitudes of the usual (``linear'') and nonlinear susceptibilities and the amplitude of the second derivative of the spin-spin inverse correlation length with respect to the external field are calculated. The behavior of critical amplitude combinations (which, in accordance with the Privman-Fisher equations, do not contain in their composition the nonuniversal metric coefficients and geometry prefactor) are studied as a function of the interaction anisotropy parameters. A universality domain for the amplitude ratios is found in the quasi-one-dimensional regime of interactions in the system. In the case of a fully isotropic three-dimensional Ising model for which the high precision values of the critical coupling and critical-point free energy are available, improved estimates are obtained for the following four universal quantities: (1) the amplitude of spin-spin inverse correlation length, (2) the amplitude of singular part of the free energy, (3) the ratio of the amplitude of a second derivative of the spin-spin inverse correlation length with respect to the external field to the usual susceptibility amplitude, and (4) the ratio of the nonlinear susceptibility amplitude to the square of the linear susceptibility amplitude (i.e., for the finite-size counterpart of the four-point renormalized coupling constant).

arXiv:cond-mat/9612219•Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics•Apr 1, 1997•14 citations

On Ising model in magnetic field on the lattice

Raghav G. Jha

We study the free energy of the classical square-lattice Ising model in a uniform real magnetic field $h$, for which no exact solution is known. We first conjecture a closed-form expression built from the one-dimensional Ising chain and Onsager's hypergeometric solution. Comparing with high-precision tensor renormalization group (TRG) data, we find that it deviates by up to $2.6\%$, has no spontaneous magnetization, and gives an effective critical-isotherm exponent $δ\to 1$. However, using the Lee--Yang circle theorem, we show that the exact free energy is a superposition of one-dimensional Ising chains in the same field, weighted by a temperature-dependent (normalized, but not always positive) distribution of squared Lee--Yang gaps, of which our first expression is a single-chain approximation. Exact results (Onsager, Yang, the strong-field expansion and the Lee--Yang solution at $βh = iπ/2$) become sum rules for this distribution. We propose a simple form for it, fixed by Onsager's free energy and the exact strong-field limit without any fit to numerical data. This \emph{improved} conjecture agrees with TRG to $0.15\%$ for all $β$ and $h$, reproduces Yang's spontaneous magnetization exactly, has a positive specific heat, and yields the critical isotherm $M\propto h^{1/δ}$ with $δ= 15.2$ and amplitude $1.055$, compared with the exact $δ=15$ and $1.059$. A refined form, which builds in additional input from Ising field theory (including the exact critical isotherm, $δ=15$) and a few constants adjusted to transfer-matrix and TRG data, matches the TRG free energy to $4\times 10^{-5}$.

arXiv:2504.18744•Apr 25, 2025•0 citations
We construct a single reduced free-energy generator for the simple-cubic three-dimensional Ising model by finite gluing of complementary local relations. With the convention Φ=+log⁡Z/V\Phi=+\log Z/VΦ=+logZ/V, the final object is Φ3(K,H,L)=Φ0(K)+ΔΦH(K,H)+ΦFSS(K,H,L),\Phi_3(K,H,L)=\Phi_0(K)+\Delta\Phi_H(K,H)+\Phi_{\rm FSS}(K,H,L),Φ3​(K,H,L)=Φ0​(K)+ΔΦH​(K,H)+ΦFSS​(K,H,L), from which M=∂HΦ3M=\partial_H\Phi_3M=∂H​Φ3​, χ=∂H2Φ3\chi=\partial_H^2\Phi_3χ=∂H2​Φ3​, and UK=∂KΦ3U_K=\partial_K\Phi_3UK​=∂K​Φ3​ are generated. The construction covers the high- and low-temperature regimes, the critical region, finite magnetic field, strong-field saturation, periodic-film finite size, and simultaneous finite HHH and finite LLL. The materials and operations are separated explicitly. A diagnostic dimensional stage uses d=1,2,4d=1,2,4d=1,2,4 anchors in μc=−log⁡tanh⁡Kc\mu_c=-\log\tanh K_cμc​=−logtanhKc​ to transport critical coordinates and positive film amplitudes toward d=3d=3d=3. The practical 3D generator then protects established 3D values of KcK_cKc​, critical exponents, selected amplitude ratios, and scalar zero-field critical data. The zero-field term joins truncated simple-cubic high- and low-temperature series through a singular critical patch with C2C^2C2 matching. The inner magnetic increment joins an O(K2)O(K^2)O(K2) high-temperature field expansion, a minimal 3D parametric critical free energy fixed only by the basic 3D anchors, and a low-temperature/cluster continuation using fixed quintic switches in KKK. A recovered z4z^4z4 strong-field expansion is integrated to an absolute free energy. The final inner-to-strong crossover is performed by constructing a monotone beta-CDF magnetization whose endpoint values and integral reproduce the two endpoint free energies, and then integrating that derivative; thus convexity is enforced at the free-energy level rather than by clipping observables. The periodic-film excess is built from dimensionally weighted 2D and leading 4D film surfaces in common correlation-length coordinates, with an HHH-independent residual stabilizer that preserves the H=0H=0H=0 axis. A previously reconstructed full 3D parametric equation-of-state benchmark and the full 3D finite-size scaling function are withheld from construction. The pure dimensional stage gives Kc(3)=0.22301549K_c(3)=0.22301549Kc​(3)=0.22301549 and sub-percent estimates of β\betaβ, δ\deltaδ, and ν\nuν. After 3D calibration, the final implementation passes 29/29 stated structural and numerical checks. In the withheld low-field window ∣K−Kc∣≤0.01|K-K_c|\le0.01∣K−Kc​∣≤0.01, H≤0.1H\le0.1H≤0.1, the magnetization has RMS relative error 1.21%1.21\%1.21% and maximum error 4.29%4.29\%4.29%. The result is not an exact lattice solution; it is a finite, auditable reconstruction in which the input anchors, switch intervals, matching conditions, numerical tolerances, and validation data are kept distinct.
We test the Gluing Method on the square-lattice two-dimensional Ising model by reconstructing global free-energy generators from complementary local relations. The emphasis is kept explicit throughout: the inputs used as \emph{gluing material} are separated from quantities obtained \emph{after gluing} and from data reserved for validation. At zero magnetic field, Kramers--Wannier duality is absorbed into the coordinate τ=log⁡sinh⁡2K\tau=\log\sinh 2Kτ=logsinh2K, so that the reduced free energy decomposes as Φ(τ)=τ/2+G(∣τ∣)\Phi(\tau)=\tau/2+G(|\tau|)Φ(τ)=τ/2+G(∣τ∣). The gluing materials are then only a critical logarithmic local expansion and a low-order high-temperature outer expansion; the low-temperature sector is generated by duality rather than fitted independently. Positive flat weights preserve the prescribed critical asymptotics to every finite algebraic order. With one crossover parameter fixed from three scalar anchors, the resulting bulk free energy agrees with the exact Onsager benchmark at a maximum relative error of 1.915×10−51.915\times10^{-5}1.915×10−5 on the tested interval. Internal energy and specific heat are derived from the same glued generator. Zero-field finite-size corrections are organized separately as one-variable scaling functions multiplying L−2,L−4,…L^{-2},L^{-4},\ldotsL−2,L−4,…. At nonzero field in the thermodynamic limit, the gluing materials are the established high-temperature, critical, and low-temperature local expansions of the universal Ising magnetic scaling free energy Φmag(η)\Phi_{\rm mag}(\eta)Φmag​(η), η=m/∣h∣8/15\eta=m/|h|^{8/15}η=m/∣h∣8/15. The low-temperature input contains the ∣h∣|h|∣h∣ cusp associated with spontaneous symmetry breaking. Mapping the glued scaling free energy back to the square lattice through nonlinear scaling fields yields a single generator for the magnetic free-energy increment, magnetization, and susceptibility. The construction reproduces high-temperature linear response, the critical isotherm M∝∣H∣1/15M\propto |H|^{1/15}M∝∣H∣1/15, and the two one-sided limits M(H→0±)=±M0M(H\to0^\pm)=\pm M_0M(H→0±)=±M0​ without separately fitting magnetization branches. Finally, finite size and finite field are combined in the scaling variables xt=Lmx_t=Lmxt​=Lm and xh=L15/8hx_h=L^{15/8}hxh​=L15/8h. The two protected one-dimensional axes, a central mixed-response expansion, a low-temperature spectral-rounding relation, and a massive outer description are glued into a two-variable scaling surface Yglue(xt,xh)\mathcal Y_{\rm glue}(x_t,x_h)Yglue​(xt​,xh​). The resulting lattice free energy closes in the compact form ΦLglue(K,H)=Φlatglue(K,H)+Yglue(Lm,L15/8h)−Ybulk(Lm,L15/8h)L2+O(L−4)\Phi_L^{glue}(K,H)=\Phi_{lat}^{glue}(K,H)+\frac{\mathcal{Y}_{glue}(Lm,L^{15/8}h)- Y_{bulk}(Lm,L^{15/8}h)}{L^2}+O(L^{-4})ΦLglue​(K,H)=Φlatglue​(K,H)+L2Yglue​(Lm,L15/8h)−Ybulk​(Lm,L15/8h)​+O(L−4) where m=m(r,H)m=m(r,H)m=m(r,H) and h=h(r,H)h=h(r,H)h=h(r,H) are the nonlinear scaling fields and YbulkY_{\rm bulk}Ybulk​ is the thermodynamic singular scaling surface expressed in the same variables. Thus all finite-size/finite-field information is carried by the excess Yglue−Ybulk\mathcal Y_{\rm glue}-Y_{\rm bulk}Yglue​−Ybulk​, while the thermodynamic lattice free energy remains the previously constructed Φlatglue\Phi_{\rm lat}^{\rm glue}Φlatglue​. On eleven transfer-matrix points withheld from the construction, the blind free-energy test gives RMS absolute error 3.90×10−43.90\times10^{-4}3.90×10−4 and maximum absolute error 9.55×10−49.55\times10^{-4}9.55×10−4 in Y\mathcal YY. Differentiation of the same free energy produces finite-size magnetization and susceptibility; magnetization is especially stable on the low-temperature side, while the second derivative is more sensitive in crossover regions.
We have substantially extended the high-temperature and low-magnetic-field (and the related low-temperature and high-magnetic-field) bivariate expansions of the free energy for the conventional three-dimensional Ising model and for a variety of other spin systems generally assumed to belong to the same critical universality class. In particular, we have also derived the analogous expansions for the Ising models with spin s=1,3/2,…s=1,3/2,\dots{}s=1,3/2,… and for the lattice Euclidean scalar-field theory with quartic self-interaction, on the simple-cubic and the body-centered-cubic lattices. Our bivariate high-temperature expansions, which extend through 24th order, enable us to compute, through the same order, all higher derivatives of the free energy with respect to the field, namely, all higher susceptibilities. These data make more accurate checks possible, in critical conditions, both of the scaling and the universality properties with respect to the lattice and the interaction structure and also help to improve an approximate parametric representation of the critical equation of state for the three-dimensional Ising model universality class.
25th-order high-temperature series are computed for a general nearest-neighbor three-dimensional Ising model with arbitrary potential on the simple cubic lattice. In particular, we consider three improved potentials characterized by suppressed leading scaling corrections. Critical exponents are extracted from high-temperature series specialized to improved potentials, obtaining $\ensuremath{\gamma}=1.2373(2),$ $\ensuremath{\nu}=0.63012(16),$ $\ensuremath{\alpha}=0.1096(5),$ $\ensuremath{\eta}=0.03639(15),$ $\ensuremath{\beta}=0.32653(10),$ and $\ensuremath{\delta}=4.7893(8).$ Moreover, biased analyses of the 25th-order series of the standard Ising model provide the estimate $\ensuremath{\Delta}=0.52(3)$ for the exponent associated with the leading scaling corrections. By the same technique, we study the small-magnetization expansion of the Helmholtz free energy. The results are then applied to the construction of parametric representations of the critical equation of state, using a systematic approach based on a global stationarity condition. Accurate estimates of several universal amplitude ratios are also presented.
We perform a high-statistics simulation of the three-dimensional randomly dilute Ising model on cubic lattices L3 with L< or =256. We choose a particular value of the density, x=0.8, for which the leading scaling corrections are suppressed. We determine the critical exponents, obtaining nu=0.683(3), eta=0.035(2), beta=0.3535(17), and alpha=-0.049(9), in agreement with previous numerical simulations. We also estimate numerically the fixed-point values of the four-point zero-momentum couplings that are used in field-theoretical fixed-dimension studies. Although these results somewhat differ from those obtained using perturbative field theory, the field-theoretical estimates of the critical exponents do not change significantly if the Monte Carlo result for the fixed point is used. Finally, we determine the six-point zero-momentum couplings, relevant for the small-magnetization expansion of the equation of state, and the invariant amplitude ratio R(+)(xi) that expresses the universality of the free-energy density per correlation volume. We find R(+)(xi)=0.2885(15).
A fully anisotropic simple-cubic Ising lattice in the geometry of periodic cylinders n\ifmmode\times\else\texttimes\fi{}n\ifmmode\times\else\texttimes\fi{}\ensuremath{\infty} is investigated by the transfer-matrix finite-size-scaling method. In addition to the previously obtained critical amplitudes of the inverse correlation lengths and singular part of the free energy [M. A. Yurishchev, Phys. Rev. B 50, 13 533 (1994)], the amplitudes of the usual (``linear'') and nonlinear susceptibilities and the amplitude of the second derivative of the spin-spin inverse correlation length with respect to the external field are calculated. The behavior of critical amplitude combinations (which, in accordance with the Privman-Fisher equations, do not contain in their composition the nonuniversal metric coefficients and geometry prefactor) are studied as a function of the interaction anisotropy parameters. A universality domain for the amplitude ratios is found in the quasi-one-dimensional regime of interactions in the system. In the case of a fully isotropic three-dimensional Ising model for which the high precision values of the critical coupling and critical-point free energy are available, improved estimates are obtained for the following four universal quantities: (1) the amplitude of spin-spin inverse correlation length, (2) the amplitude of singular part of the free energy, (3) the ratio of the amplitude of a second derivative of the spin-spin inverse correlation length with respect to the external field to the usual susceptibility amplitude, and (4) the ratio of the nonlinear susceptibility amplitude to the square of the linear susceptibility amplitude (i.e., for the finite-size counterpart of the four-point renormalized coupling constant).
We study the free energy of the classical square-lattice Ising model in a uniform real magnetic field hhh, for which no exact solution is known. We first conjecture a closed-form expression built from the one-dimensional Ising chain and Onsager's hypergeometric solution. Comparing with high-precision tensor renormalization group (TRG) data, we find that it deviates by up to 2.6%2.6\%2.6%, has no spontaneous magnetization, and gives an effective critical-isotherm exponent δ→1δ\to 1δ→1. However, using the Lee--Yang circle theorem, we show that the exact free energy is a superposition of one-dimensional Ising chains in the same field, weighted by a temperature-dependent (normalized, but not always positive) distribution of squared Lee--Yang gaps, of which our first expression is a single-chain approximation. Exact results (Onsager, Yang, the strong-field expansion and the Lee--Yang solution at βh=iπ/2βh = iπ/2βh=iπ/2) become sum rules for this distribution. We propose a simple form for it, fixed by Onsager's free energy and the exact strong-field limit without any fit to numerical data. This \emph{improved} conjecture agrees with TRG to 0.15%0.15\%0.15% for all βββ and hhh, reproduces Yang's spontaneous magnetization exactly, has a positive specific heat, and yields the critical isotherm M∝h1/δM\propto h^{1/δ}M∝h1/δ with δ=15.2δ= 15.2δ=15.2 and amplitude 1.0551.0551.055, compared with the exact δ=15δ=15δ=15 and 1.0591.0591.059. A refined form, which builds in additional input from Ising field theory (including the exact critical isotherm, δ=15δ=15δ=15) and a few constants adjusted to transfer-matrix and TRG data, matches the TRG free energy to 4×10−54\times 10^{-5}4×10−5.