The interfacial tension between the two ordered phases of the Ising model equals the free-energy cost of imposing antiperiodic instead of periodic boundary conditions across one lattice direction. Híjar and Sutmann (J. Stat. Mech. 2008 P07012) showed for the two-dimensional Ising model that this free-energy difference can be estimated from ensembles of non-equilibrium switching trajectories that drive the boundary coupling λ\lambda from JJ to −J-J (creating the interface) or back (destroying it), using the Jarzynski equality and the Crooks fluctuation relation, with Onsager's exact surface tension as benchmark; the Crooks crossing estimator proved the most efficient under a time-normalized cost comparison. Four linked directions remain open beyond this two-dimensional benchmark: (i) a quantitative characterization of the interface dynamics during switching (relaxation times, capillary-wave behaviour, and their dependence on temperature, lattice geometry L,RL,R, and switching rate nn); (ii) how boundary conditions transverse to the interface (free versus periodic) modify finite-size corrections and estimator convergence; (iii) whether the methodology and its efficiency hierarchy survive in three-dimensional Ising lattices, where no exact surface-tension formula exists; and (iv) whether interface-specific cluster algorithms such as the Hasenbusch-Meyer interface cluster algorithm avoid the trivial large-cluster moves arising near extreme λ\lambda values and further improve the non-equilibrium estimators. The problem of establishing the validity, convergence behaviour, and efficiency of the non-equilibrium boundary-switching methodology for Ising interfacial tension beyond the original 2D benchmark is therefore open.

Background

Below its critical temperature the Ising model has two coexisting phases of opposite magnetization separated by interfaces whose free-energy cost per unit length (per unit area in three dimensions) is the interfacial tension σ\sigma. On a finite lattice this quantity is exactly a free-energy difference between two boundary-condition ensembles: with antiperiodic boundary conditions (APBC) in one lattice direction — one boundary bond reversed in sign, i.e. a frustrated seam — an interface is forced into the system, while with periodic boundary conditions (PBC) it is not, so the interface free energy is F(β,L,R)=−β−1ln⁡[Zap/Zp]F(\beta,L,R)=-\beta^{-1}\ln[Z_{\rm ap}/Z_{\rm p}], with β\beta the inverse temperature and LL, RR the lattice extents transverse to and along the interface. In two dimensions Onsager's exact solution of the Ising model gives the thermodynamic-limit value σ=2J−β−1ln⁡coth⁡(βJ)\sigma=2J-\beta^{-1}\ln\coth(\beta J), with JJ the exchange coupling, and exact results for the finite lattice with adjustable boundary conditions are also known, which makes 2D the natural benchmark; in three dimensions no exact formula exists and σ\sigma must be obtained numerically. Equilibrium estimation of FF is difficult: thermodynamic perturbation (TP, an exponential reweighting of equilibrated samples at nearby parameter values) and thermodynamic integration require equilibrated sampling at many intermediate Hamiltonians, and interface-carrying configurations relax slowly because capillary waves — the long-wavelength fluctuations of the interface position and profile — are slow modes. Non-equilibrium fluctuation relations offer an alternative: the Jarzynski equality ⟨e−βA⟩=e−βΔF\langle e^{-\beta A}\rangle=e^{-\beta\Delta F} states that the exponential average of the work AA performed along an arbitrarily fast (but microreversibly driven) switching process equals the equilibrium free-energy difference, and the Crooks relation states that the work distributions of the forward and time-reversed processes, PF(A)P_{\rm F}(A) and PR(−A)P_{\rm R}(-A), cross exactly at A=ΔFA=\Delta F; both yield estimators from ensembles of non-equilibrium trajectories, without equilibration at intermediate parameter values. Híjar and Sutmann (2008) combined these ingredients for the 2D Ising interface: the control parameter is the coupling λ\lambda on one lattice boundary (the control boundary), switched in nn discrete steps from λ=J\lambda=J (PBC, no interface) to λ=−J\lambda=-J (APBC, interface present) and back, with thermalization sub-steps between increments performed by either single-spin-flip Metropolis or Swendsen–Wang cluster updates (an algorithm that flips clusters of aligned spins connected by occupied bonds, permitting collective moves). For 0.4407≲β≤10.4407\lesssim\beta\le 1 and lattices from 8×88\times 8 to 128×128128\times 128 they found that the estimates approach Onsager's σ\sigma (with residual deviations of up to roughly 9% at the largest size and slowest switching rate studied), that the Crooks crossing estimator is the most efficient among the Jarzynski forward, Jarzynski reverse, Crooks, and TP estimators under a cost comparison that equates total Monte Carlo update counts, that Swendsen–Wang updates reduce relaxation times and improve the estimators, and that one-sided Jarzynski estimates from reverse (interface-destroying) trajectories are systematically better than forward ones; fitting the finite-size estimates to the asymptotic exact form F≃−β−1ln⁡[(2mL/π)K1(mR)]F\simeq-\beta^{-1}\ln[(2mL/\pi)K_{1}(mR)] (K1K_{1} a modified Bessel function, mm the inverse correlation length) reproduced mm at β=0.5\beta=0.5 with visible finite-size drift. The wider setting is as follows. The 3D Ising interface tension has been measured by equilibrium and extended-ensemble methods: Mon and Jasnow (1985); the cluster-update study of Hasenbusch and Meyer (1991), which introduced the interface cluster algorithm named in the source (a cluster-update scheme designed to move an interface collectively and thereby accelerate its slow dynamics); Hasenbusch and Pinn (1997); the high-precision determination of Caselle, Hasenbusch and Panero (2007); Bittner, Nußbaumer and Janke (2009) with enhanced-sampling simulations combined with parallel tempering (exchanging configurations between simulations at nearby temperatures); and Lima, Viana Parente Lopes, Matos and Penedones (2026), who sampled the 3D domain wall with a flat-histogram method in an extended ensemble over the boundary coupling itself — the same λ\lambda variable as the source's protocol, but sampled in equilibrium rather than by non-equilibrium switching. Chatelain (2007), before the source, had already shown that a temperature-extended Jarzynski relation can estimate the 3D Ising order-order surface tension, considering both periodic and free transverse boundary conditions; Caselle, Costagliola, Nada, Panero and Toniato (2016) applied Jarzynski-based switching to interface free energies in lattice gauge theory and quantified the overlap problem between initial and final ensembles; Batygolski, Napierała-Batygolska and Lipowski (2022) transplanted the boundary-switching protocol to a nonequilibrium Ising-doped voter model. Within this landscape, the open directions identified in the introduction constitute a single program: to extend and validate the non-equilibrium boundary-switching estimator of interfacial tension beyond the original 2D benchmark, by characterizing the interface dynamics that govern its convergence, the finite-size effects induced by boundary conditions transverse to the interface, its behaviour in three dimensions, and its efficiency relative to interface-specific cluster algorithms.

Problem Statement

Establish the validity, convergence behaviour, and efficiency of the non-equilibrium boundary-switching methodology for estimating Ising interfacial tension beyond the original two-dimensional benchmark of Híjar and Sutmann (2008). The single overarching objective — a validated, efficiency-characterized non-equilibrium estimator of the interface free energy F(β,L,R)=−β−1ln⁡[Zap/Zp]F(\beta,L,R)=-\beta^{-1}\ln[Z_{\rm ap}/Z_{\rm p}] (with β\beta the inverse temperature, JJ the exchange coupling, and ZapZ_{\rm ap}, ZpZ_{\rm p} the partition functions with antiperiodic and periodic boundary conditions in the switched direction) obtained by switching the boundary coupling λ\lambda between JJ and −J-J — is supported by four linked deliverables. These directions are motivated by the source's results and closing discussion. (1) Interface dynamics during switching: quantitatively characterize how the interface evolves during the switching protocols — relaxation times after each parameter increment and the capillary-wave dynamics of the interface — and how these depend on temperature, lattice geometry (LL transverse to the interface, RR along it), and the number of switching steps nn. (2) Transverse boundary conditions and finite-size effects: determine systematically how boundary conditions in the direction(s) transverse to the interface — free versus periodic — modify the finite-size corrections to interfacial-tension estimates and the convergence of the non-equilibrium estimators. (3) Three-dimensional generalization: assess the methodology's performance in three-dimensional Ising lattices, where no exact surface-tension formula exists — in particular whether the non-equilibrium estimates agree with established equilibrium benchmark values of the 3D interface tension within controlled statistical errors, and whether the efficiency hierarchy observed in 2D (Crooks crossing most efficient, reverse one-sided Jarzynski better than forward, cluster-assisted thermalization superior to local updates) persists in d=3d=3. (4) Algorithmic efficiency: compare the non-equilibrium estimators with alternative cluster and interface-specific algorithms — for example the Hasenbusch–Meyer interface cluster algorithm — to determine whether such algorithms avoid the trivial large-cluster moves that arise near the extreme values λ=±J\lambda=\pm J (where nearly the entire lattice joins one percolating cluster, so a cluster update degenerates into a trivial global move) and whether they further improve the efficiency of non-equilibrium interfacial-tension estimates. Deliverables (1), (2) and (4) characterize the mechanisms — relaxation, finite-size corrections, and update dynamics — that govern the estimator's accuracy and cost, and jointly ground the central extrapolation in deliverable (3); all four deliverables are required for a resolution, and a resolution — whether it confirms or refutes the accuracy or the efficiency of the methodology beyond the 2D benchmark — must support its conclusions with quantified statistical errors under a stated cost convention.

References

  1. H. Híjar and G. Sutmann, 'Monte Carlo estimates of interfacial tension in the two-dimensional Ising model from non-equilibrium methods', J. Stat. Mech. (2008) P07012. DOI: 10.1088/1742-5468/2008/07/P07012, https://doi.org/10.1088/1742-5468/2008/07/P07012
  2. H. Híjar, J. Quintana-H. and G. Sutmann, 'Non-equilibrium work theorems for the two-dimensional Ising model', J. Stat. Mech. (2007) P04010. DOI: 10.1088/1742-5468/2007/04/P04010, https://doi.org/10.1088/1742-5468/2007/04/P04010
  3. C. Jarzynski, 'Nonequilibrium Equality for Free Energy Differences', Phys. Rev. Lett. 78, 2690 (1997). DOI: 10.1103/PhysRevLett.78.2690, https://doi.org/10.1103/PhysRevLett.78.2690
  4. G. E. Crooks, 'Entropy production fluctuation theorem and the nonequilibrium work relation for free energy differences', Phys. Rev. E 60, 2721 (1999). DOI: 10.1103/PhysRevE.60.2721, https://doi.org/10.1103/PhysRevE.60.2721
  5. R. H. Swendsen and J.-S. Wang, 'Nonuniversal critical dynamics in Monte Carlo simulations', Phys. Rev. Lett. 58, 86 (1987). DOI: 10.1103/PhysRevLett.58.86, https://doi.org/10.1103/PhysRevLett.58.86
  6. M. Hasenbusch and S. Meyer, 'Cluster-update acceleration of interface roughening in the 3D Ising model', Phys. Rev. Lett. 66, 530 (1991). DOI: 10.1103/PhysRevLett.66.530, https://doi.org/10.1103/PhysRevLett.66.530
  7. K. K. Mon and D. Jasnow, 'Surface tension and universality in the three-dimensional Ising model', J. Stat. Phys. 41, 273 (1985). DOI: 10.1007/BF01020613, https://doi.org/10.1007/BF01020613
  8. C. Chatelain, 'A temperature-extended Jarzynski relation: application to the numerical calculation of surface tension', J. Stat. Mech. (2007) P04011. DOI: 10.1088/1742-5468/2007/04/P04011, https://doi.org/10.1088/1742-5468/2007/04/P04011 (preprint arXiv:cond-mat/0702044, https://arxiv.org/abs/cond-mat/0702044)
  9. C. Chatelain and D. Karevski, 'Probability distributions of the work in the two-dimensional Ising model', J. Stat. Mech. (2006) P06005. DOI: 10.1088/1742-5468/2006/06/P06005, https://doi.org/10.1088/1742-5468/2006/06/P06005 (preprint arXiv:cond-mat/0602580, https://arxiv.org/abs/cond-mat/0602580)
  10. D. B. Abraham, L. F. Ko and N. M. Švrakić, 'Ising Model with Adjustable Boundary Conditions: Exact Results for Finite Lattice Mass Gaps', Phys. Rev. Lett. 61, 2393 (1988). DOI: 10.1103/PhysRevLett.61.2393, https://doi.org/10.1103/PhysRevLett.61.2393
  11. L. Onsager, 'Crystal Statistics. I. A Two-Dimensional Model with an Order-Disorder Transition', Phys. Rev. 65, 117 (1944). DOI: 10.1103/PhysRev.65.117, https://doi.org/10.1103/PhysRev.65.117
  12. M. Caselle, M. Hasenbusch and M. Panero, 'The interface free energy: comparison of accurate Monte Carlo results for the 3D Ising model with effective interface models', J. High Energy Phys. 2007(09), 117 (2007). DOI: 10.1088/1126-6708/2007/09/117, https://doi.org/10.1088/1126-6708/2007/09/117
  13. M. Hasenbusch and K. Pinn, 'The interface tension of the three-dimensional Ising model in the scaling region', Physica A 245, 366 (1997). DOI: 10.1016/S0378-4371(97)00314-2, https://doi.org/10.1016/S0378-4371(97)00314-2
  14. E. Bittner, A. Nußbaumer and W. Janke, 'Anisotropy of the interface tension of the three-dimensional Ising model', Nucl. Phys. B 820, 694 (2009). DOI: 10.1016/j.nuclphysb.2009.05.009, https://doi.org/10.1016/j.nuclphysb.2009.05.009 (preprint arXiv:0905.4569, https://arxiv.org/abs/0905.4569)
  15. D. Lima, J. M. Viana Parente Lopes, J. Matos and J. Penedones, 'Effective string theory on a torus: the 3d Ising domain wall', J. High Energy Phys. 2026(07), 267 (2026). DOI: 10.1007/JHEP07(2026)267, https://doi.org/10.1007/JHEP07(2026)267 (preprint arXiv:2510.15206, https://arxiv.org/abs/2510.15206)
  16. M. Caselle, G. Costagliola, A. Nada, M. Panero and A. Toniato, 'Jarzynski's theorem for lattice gauge theory', Phys. Rev. D 94, 034503 (2016). DOI: 10.1103/PhysRevD.94.034503, https://doi.org/10.1103/PhysRevD.94.034503 (preprint arXiv:1604.05544, https://arxiv.org/abs/1604.05544)
  17. J. Batygolski, A. Napierała-Batygolska and A. Lipowski, 'Fluctuation relations in a nonequilibrium system: Surface tension and effective temperature in an Ising-doped voter model', Phys. Rev. E 105, 064120 (2022). DOI: 10.1103/PhysRevE.105.064120, https://doi.org/10.1103/PhysRevE.105.064120
  18. A. Rajantie and D. J. Weir, 'Quantum kink and its excitations', J. High Energy Phys. 2009(04), 068 (2009). DOI: 10.1088/1126-6708/2009/04/068, https://doi.org/10.1088/1126-6708/2009/04/068 (preprint arXiv:0902.0367, https://arxiv.org/abs/0902.0367)
  19. D. J. Weir, 'Quantum mechanics of topological solitons', PhD thesis, Imperial College London (2011). DOI: 10.25560/9146, https://doi.org/10.25560/9146, https://spiral.imperial.ac.uk/handle/10044/1/9146
  20. A. Ueda, L. Burgelman, L. Tagliacozzo and L. Vanderstraeten, 'Interface roughening in the 3-D Ising model with tensor networks', arXiv:2601.07829 (2026). arXiv:2601.07829, https://arxiv.org/abs/2601.07829