Open problems for the coherent-state ansatz of the Holstein polaron: smooth crossover, delocalized phonons at large momentum, and extensions to bipolarons and non-Bravais lattices

The Holstein polaron - a single lattice electron coupled locally to dispersionless optical phonons - is a paradigm of electron-phonon physics. Walsh, Boettcher, and Marsiglio (arXiv:2603.07290; J. Phys.: Condens. Matter 2026) introduced two extremely cheap approximations: a coherent-state ansatz (CSA), which superposes five families of coherent phonon clouds surrounding the electron, and a restricted Hilbert space (RHS) exact diagonalization built from the same five families. Both are accurate at strong coupling and surprisingly good at weak coupling, but they leave three linked problems open: (i) the CSA exhibits a spurious discontinuous jump in the effective mass at a critical coupling for small phonon frequencies ( $\omega_E/t \lesssim 0.2$ ), where the exact solution crosses over smoothly; (ii) the five-family truncation fails where phonons delocalize from the electron (small coupling, large momentum); and (iii) extensions to the two-electron bipolaron and to non-Bravais (frustrated) lattices are proposed but undeveloped. Two documented limitations motivate these problems: (iv) exact Bonča-Trugman benchmarks converge very slowly in the delocalized-phonon regime because the ground state contains a phonon spatially separated from the electron; and (v) the accuracy of CSA/RHS energies and masses is characterized only through spot numerical benchmarks. The problem is to construct a compact, systematically improvable, quantitatively controlled variational framework of the CSA/RHS type that is uniformly valid across coupling, phonon frequency, and momentum in one and two dimensions, and transferable to bipolarons and complex lattice geometries.

Background

The Holstein model is the archetypal lattice model of local electron-phonon coupling: a single electron hops between nearest-neighbor sites of a $d$ -dimensional hypercubic lattice with amplitude $t$ while coupling linearly to a dispersionless optical phonon mode of frequency $\omega_E$ at each site. The coupling forms a polaron, an electron dressed by a cloud of virtual phonons, whose ground-state energy, dispersion $E(\mathbf{k})$ , and effective mass $m^*$ (the curvature of $E(\mathbf{k})$ at $\mathbf{k}=0$ , conventionally quoted relative to the bare band mass $m$ ) interpolate between a weakly dressed, mobile carrier at weak coupling and a heavy, quasi-localized small polaron at strong coupling. The standard parametrization uses the dimensionless coupling $\lambda = g^2\omega_E/(2dt)$ , with $g$ the dimensionless electron-phonon coupling constant: roughly, $\lambda<1$ is weak coupling, $\lambda>1$ strong coupling, and the crossover lies near $\lambda\approx 1$ . Analyticity results of Gerlach and Löwen (1991) establish that infinite-lattice polaron systems admit no genuine phase transitions: ground-state observables vary smoothly with coupling, so the very sharp crossover seen in some approximations is still a smooth crossover, not a transition. Numerically exact results come mainly from the Bonča-Trugman method (Bonča, Trugman, and Batistić, 1999): an exact-diagonalization scheme in which a basis within a fixed momentum sector is generated by repeatedly applying the off-diagonal (hopping) part of the Hamiltonian to seed states and truncating by basis size and phonon number. It is exact where it converges but expensive at strong coupling (many phonons in the ground state; the average phonon number in the strong-coupling limit is $g^2 = 2d\lambda/(\omega_E/t)$ ) and slow-converging at small $\omega_E/t$ , small $\lambda$ , and large momentum. Walsh, Boettcher, and Marsiglio (2026) introduced two approximations designed for exactly this strong-coupling, many-phonon regime. Observing that at strong coupling the ground-state probability concentrates in a handful of phonon-configuration families (phonons on the electron site, on nearest-neighbor sites, on next-nearest sites, and two further families), they built: (1) the coherent-state ansatz (CSA), a variational wavefunction that superposes five Bloch-summed families, each constrained to be a phonon coherent state - a displaced-oscillator state, i.e. a superposition of all phonon occupation numbers on the relevant sites with Poisson weights, parametrized by a single displacement amplitude $\alpha_\mu$ per family - so the whole calculation reduces to diagonalizing a small matrix with five variational parameters; and (2) the restricted Hilbert space (RHS) approximation, exact diagonalization within the individual phonon Fock states (occupation-number states of the phonon modes) spanned by the same five families, which relaxes the coherent-state constraint at slightly higher cost. The CSA is exact in both limits $\lambda\to 0$ (free electron) and $\lambda\to\infty$ (the strong-coupling limit in which the electron localizes with a Poisson cloud on its own site), and both methods are accurate at strong coupling at negligible computational cost, making two-dimensional and arbitrary-lattice applications essentially free. Three limitations motivate the open problems. First, at small phonon frequencies ( $\omega_E/t\lesssim 0.2$ ) the CSA effective mass exhibits a discontinuous jump at a critical coupling $\lambda_c$ (decreasing with decreasing $\omega_E$ in 1D, tending to a finite value in 2D): the coherent-state constraint forces the ansatz to choose between a weak-coupling form ( $\alpha_\mu\approx 0$ ) and a strong-coupling form ( $\alpha_\mu\sim g$ ), whereas the exact ground state interpolates smoothly between the two; the RHS, being a larger calculation, does reproduce a smooth crossover (in 2D an extremely abrupt but continuous one). Second, at small $\lambda$ and large $\mathbf{k}$ - momenta near the Brillouin-zone boundary, where the exact dispersion crosses the flat free-phonon level - the exact ground state contains a single phonon spatially separated from a nearly free electron (a $\mathbf{k}=0$ electron plus one free phonon carrying the momentum); the five local families cannot represent such delocalized phonons, the RHS fails quantitatively, and even exact Bonča-Trugman calculations converge slowly there because the separated phonon is reached only at high order in the iterative basis generation, even when the seeds are the strong-coupling states. Third, the accuracy of the two methods has been quantified only through spot comparisons against Bonča-Trugman benchmarks at selected parameter points. Compact variational polaron descriptions with coherent phonons have a long prior history: the Merrifield and Toyozawa variational wavefunctions (compact trial states in which a delocalized electron carries a coherent phonon cloud) and the global-local method (a variational band theory combining such delocalized clouds with localized, site-attached ones; Zhao, Brown, and Lindenberg, 1997; Brown, Lindenberg, and Zhao, 1997), the delocalized Davydov D1 ansatz (a translationally invariant version of the Davydov electron-plus-coherent-phonon trial states; Sun, Duan, and Zhao, 2013), and the momentum-average (MA) hierarchy of Berciu (2006), a diagrammatic Green's-function approximation with a closed-form self-energy that is smooth across the crossover and asymptotically exact at strong coupling. What distinguishes the CSA/RHS is the organization of the ansatz around a handful of real-space families around the electron, which is what makes applications to higher dimensions and arbitrary lattices computationally free; the open questions concern exactly whether that structure can be made uniformly valid and controlled. On the two-electron side, the bipolaron problem (two electrons with explicit Coulomb repulsion plus phonon-mediated attraction) has been studied with exact methods in 1D at intermediate and large $\omega_E/t$ (Bonča, Katrašnik, and Trugman, 2000), but not with compact coherent-state ansatze at small $\omega_E/t$ , and Holstein polarons on non-Bravais lattices - lattices whose crystallographic unit cell contains more than one site, such as honeycomb, kagome, and pyrochlore - have recently become of interest in geometrically frustrated systems where exact methods are expensive.

Problem Statement

Walsh, Boettcher, and Marsiglio pose the problem of turning their five-family coherent-state framework into a uniformly valid, quantitatively controlled variational description of the Holstein polaron. The common scientific objective uniting the requested results is: construct a variational treatment of the single Holstein polaron that retains the computational cheapness and real-space transparency of the CSA/RHS while being accurate at every point of the parameter space $(\lambda, \omega_E/t, \mathbf{k})$ in one and two dimensions, with quantified accuracy, and demonstrably transferable to the harder problems (bipolaron, non-Bravais lattices) where exact methods are expensive. Concretely, the source asks for: (1) Smooth crossover - a variational family as compact and cheap as the CSA, ideally with a comparably small number of free parameters, whose ground-state observables, in particular the effective mass $m^*(\lambda)$ , interpolate smoothly through the weak-to-strong-coupling crossover at small phonon frequencies ( $\omega_E/t \lesssim 0.2$ ) in 1D and 2D, removing the spurious first-order jump the CSA produces at the critical coupling $\lambda_c$ . (2) Delocalized phonons - a minimal systematic extension of the five-family truncation (which additional families, under what truncation criterion) that captures the regime where phonons delocalize from the electron - notably small coupling $\lambda$ and momenta $\mathbf{k}$ near the Brillouin-zone boundary, where the exact ground state contains a free phonon spatially separated from the electron - without losing the efficiency that makes the method feasible in higher dimensions. (3) Extensions - the adaptation of the CSA/RHS to the two-electron bipolaron problem with explicit electron-electron repulsion at small $\omega_E/t$ (determining under what conditions the phonon-mediated attraction binds a mobile pair) and to Holstein polarons on non-Bravais or frustrated lattices such as honeycomb, kagome, and pyrochlore: which families of phonon configurations and which variational parameters or seeding strategies are needed for the methods to remain accurate and tractable. Two supporting challenges arise from limitations the paper documents but does not explicitly pose as open questions: obtaining converged exact dispersion references in the hard regime - Bonča-Trugman basis generation, even seeded with the strong-coupling states, converges very slowly at small $\omega_E/t$ , small $\lambda$ , and large $\mathbf{k}$ , so converged references in exactly the regime that tests items (1) and (2) are currently hard to obtain - and a quantitative characterization of how close CSA/RHS-type ground-state energies and effective masses are to exact values as a function of $(\lambda, \omega_E/t, \mathbf{k})$ , beyond the spot numerical examples presented. Progress on the items can be made separately and in any order: resolving any single item - including either sub-part of item (3), the bipolaron or the non-Bravais lattices - is a legitimate partial result, while a complete resolution addresses items (1)-(3) together. A result counts as resolving an item when it shows a clear, quantified improvement over the CSA/RHS in the regime that item targets, measured against the exact Bonča-Trugman benchmarks available there. The two supporting challenges are advisory context, not additional deliverables: they identify benchmarking and accuracy-mapping difficulties that support items (1)-(3) but are not themselves required to resolve them.

References

Connor M. Walsh, Igor Boettcher, Frank Marsiglio, "Coherent-state ansatz for the Holstein polaron in one and two dimensions", arXiv:2603.07290 (2026), doi:10.48550/arXiv.2603.07290, https://arxiv.org/abs/2603.07290; published in Journal of Physics: Condensed Matter (2026), doi:10.1088/1361-648x/ae9dd1, https://doi.org/10.1088/1361-648x/ae9dd1
J. Bonča, S. A. Trugman, I. Batistić, "Holstein polaron", Physical Review B 60, 1633 (1999), doi:10.1103/PhysRevB.60.1633, https://doi.org/10.1103/PhysRevB.60.1633
Mona Berciu, "Green's Function of a Dressed Particle", Physical Review Letters 97, 036402 (2006), doi:10.1103/PhysRevLett.97.036402, https://doi.org/10.1103/PhysRevLett.97.036402
J. Bonča, T. Katrašnik, S. A. Trugman, "Mobile Bipolaron", Physical Review Letters 84, 3153 (2000), doi:10.1103/PhysRevLett.84.3153, https://doi.org/10.1103/PhysRevLett.84.3153
B. Gerlach, H. Löwen, "Analytical properties of polaron systems or: Do polaronic phase transitions exist or not?", Reviews of Modern Physics 63, 63 (1991), doi:10.1103/RevModPhys.63.63, https://doi.org/10.1103/RevModPhys.63.63
Jin Sun, Liwei Duan, Yang Zhao, "Delocalized Davydov D1 Ansatz for the Holstein polaron", Journal of Chemical Physics 138, 174116 (2013), doi:10.1063/1.4802961, https://doi.org/10.1063/1.4802961
David W. Brown, Katja Lindenberg, Yang Zhao, "Variational energy band theory for polarons: Mapping polaron structure with the global-local method", Journal of Chemical Physics 107, 3179 (1997), doi:10.1063/1.474668, https://doi.org/10.1063/1.474668
Yang Zhao, David W. Brown, Katja Lindenberg, "Variational energy band theory for polarons: Mapping polaron structure with the Toyozawa method", Journal of Chemical Physics 107, 3159 (1997), doi:10.1063/1.474667, https://doi.org/10.1063/1.474667
Jeet Shannigrahi, Janez Bonča, Mona Berciu, "Universal thermal breakdown of polaron coherence in one, two, and three dimensions", arXiv:2608.18529 (2026), https://arxiv.org/abs/2608.18529
Moritz K. A. Baumgarten, Hamlin Wu, Tong Jiang, Joonho Lee, "A Scalable Translationally Invariant Variational Theory of Ab Initio Polarons", arXiv:2605.05675 (2026), https://arxiv.org/abs/2605.05675