Entropy-Driven Structural Phase Transition in Nb$_3$Cl$_8$ via Density Functional Theory and an Effective Model

Abstract: As a prototypical flat-band cluster Mott insulator on an effective triangular lattice, Nb$_3$Cl$_8$ is a potential candidate for hosting a quantum spin liquid (QSL) state. Nevertheless, a first-order structural phase transition around 90K transforms the high-temperature paramagnetic $α$ phase into the low-temperature nonmagnetic $β$ phase, suppressing the candidate QSL regime of the $α$ phase. To clarify the microscopic origin of this transition, we combine first-principles calculations with an extended Hubbard model to construct a unified free-energy framework. This framework reveals that the transition is jointly driven by phonon and spin entropy: the $α$ phase is stabilized by softer phonons and larger paramagnetic spin entropy, whereas the $β$ phase is favored by interlayer dimerization, which hardens the phonons and quenches the spin entropy through singlet formation. Furthermore, by evaluating the pressure-dependent generalized enthalpy, we provide a thermodynamic explanation for the suppression of the transition under c-axis uniaxial pressure, where stabilizing the $α$ phase may allow the candidate QSL regime of the $α$ phase to be explored at low temperatures.

• The paper combines DFT, phonon calculations, and correlated spin models to show that phonon and spin entropy stabilize the high-temperature α phase, while interlayer singlet formation favors the low-temperature β phase.
• The model predicts a transition near 113 K, an interlayer exchange of 70.34 meV in β-Nb₃Cl₈, and an entropy change consistent with measurements, explaining the observed first-order transition near 90 K.
• The paper predicts that c-axis uniaxial pressure above approximately 2.6 GPa suppresses the transition by favoring the more compressible α phase, potentially exposing its frustrated triangular-lattice magnetic state and candidate quantum spin liquid behavior.

Overview

Nb$_3$Cl$_8$ is a layered van der Waals cluster Mott insulator in which Nb–Nb bonding trimerizes niobium atoms into $[\mathrm{Nb}_3]^{8+}$ clusters arranged on a triangular sublattice. Each cluster carries one unpaired electron in a half-filled, nearly flat $2a_1$ molecular-orbital band, making the material a candidate host for a quantum spin liquid (QSL) on an effective triangular lattice. Experimentally, however, this candidate regime is cut short: below roughly 90 K, bulk Nb$_3$Cl$_8$ undergoes a first-order structural transition from the high-temperature paramagnetic $\alpha$ phase ($P\bar{3}m1$) to a nonmagnetic $\beta$ phase, accompanied by a sharp drop in magnetic susceptibility. The paper by Zhu et al. (2607.00599) addresses the long-standing question of what thermodynamically drives this transition and why it can be suppressed under c-axis uniaxial pressure.

The central result is that the transition is jointly driven by phonon entropy and spin entropy. The $\alpha$ phase is stabilized at high temperature by softer phonons and larger paramagnetic spin entropy; the $_8$0 phase is favored at low temperature by interlayer dimerization, which hardens phonons and quenches spin degrees of freedom through singlet formation. The authors construct a unified Helmholtz free-energy framework combining DFT total energies, harmonic phonon free energies from DFPT, and the finite-temperature free energy of an extended Hubbard model reduced to an effective Heisenberg model. This framework reproduces a transition temperature of approximately 113 K—semiquantitatively consistent with experiment—and predicts full suppression of the transition above a critical uniaxial pressure of about 2.6 GPa.

Methodological framework

The total free energy is assembled as

$_8$1

where the Hartree-level double-counting correction $_8$2 removes the low-energy $_8$3 contribution already treated explicitly by the correlated model. Because standard non-spin-polarized DFT cannot describe Mott physics in the half-filled $_8$4 manifold, that subspace is handled via an extended Hubbard model built on Wannierized cluster-centered orbitals, with screened interactions obtained from constrained RPA (288 bands included for convergence). The model is then mapped to a Heisenberg Hamiltonian with $_8$5, since the transition temperature lies far below the charge-excitation scale.

The resulting parameters quantify the essential physics of the two phases:

Phase	$_8$6 (meV)	$_8$7 (meV)	$_8$8 (meV)	$_8$9 (meV)
$[\mathrm{Nb}_3]^{8+}$0	−17.3	1.12	2.12	1394.8
$[\mathrm{Nb}_3]^{8+}$1	−136.7	70.34	1.58	1445.5

The interlayer hopping in the $[\mathrm{Nb}_3]^{8+}$2 phase is enhanced by nearly an order of magnitude relative to the $[\mathrm{Nb}_3]^{8+}$3 phase, producing an interlayer exchange of 70.34 meV that drives singlet formation on interlayer dimers. For the $[\mathrm{Nb}_3]^{8+}$4 phase, spin thermodynamics was computed via tenth-order linked-cluster high-temperature series expansion (HTSE) with Padé continuation; for the $[\mathrm{Nb}_3]^{8+}$5 phase, a second-order expansion around the dimer-singlet limit was used, supplemented at low temperature by an isolated-dimer model.

A methodological caveat deserves emphasis: the DFT-level energy difference between phases (about 100 meV per unit cell favoring $[\mathrm{Nb}_3]^{8+}$6) cannot be interpreted as the true correlated splitting, because Kohn-Sham theory artificially delocalizes the half-filled $[\mathrm{Nb}_3]^{8+}$7 electrons. The double-counting scheme corrects this by subtracting the Wannier-subspace hopping energy and the AMF-equivalent on-site Hartree term, isolating the suppressed kinetic scale of the Mott state. Bond-order-dependent Fock terms are omitted from the DC correction, an approximation the authors argue is justified by their small magnitude relative to the Hartree term.

Structural mechanism and energy barrier

The dominant structural change across the transition is an interlayer slip: the uniform staggered $[\mathrm{Nb}_3]^{8+}$8 stacking of the $[\mathrm{Nb}_3]^{8+}$9 phase transforms into the alternating $2a_1$0 sequence of the $2a_1$1 phase, which alternates between weak (staggered) and strong (eclipsed) interlayer couplings and thereby promotes c-axis dimerization. Among competing structural assignments for the $2a_1$2 phase ($2a_1$3, $2a_1$4, $2a_1$5), the authors find that the lower-symmetry distortions are energetically marginal—the relaxed $2a_1$6 and $2a_1$7 structures lie only 4 meV and 2 meV per unit cell below $2a_1$8, respectively—and adopt the highest-symmetry $2a_1$9 structure as representative. This choice implicitly disfavors the charge-disproportionation scenario ($_3$0), which relies on bond-length asymmetries absent in the higher-symmetry models.

Along a linear interpolation path between the relaxed endpoint structures, the DFT total energy exhibits a barrier of roughly 300 meV per two-site unit cell, consistent with the thermal hysteresis observed experimentally. The authors correctly note that this value is path-dependent and should not be taken as the exact kinetic barrier.

Entropy balance and transition temperature

The phonon spectra reveal systematic softening in the $_3$1 phase, traceable to its weaker interlayer coupling and larger volume. The layer-breathing mode shifts from 1.729 THz ($_3$2) to 1.826 THz ($_3$3), and the Raman-active $_3$4 mode near 5.3 THz moves from 5.228 THz to 5.354 THz across the transition, in agreement with Raman and terahertz measurements. Softer modes lower the zero-point energy and raise the vibrational entropy of the $_3$5 phase; although the internal-energy difference has a nontrivial temperature dependence, the entropic term dominates and drives the $_3$6-phase phonon free energy down more rapidly with increasing temperature.

Spin thermodynamics supplies the second entropy channel. In the $_3$7 phase, interlayer exchanges comparable to the intralayer scale leave the system paramagnetic near the transition, retaining spin entropy approaching the high-temperature limit of $_3$8 per two-site unit cell. In the $_3$9 phase, the large $_8$0 locks spins into interlayer singlets, quenching this entropy. Combining all contributions, the total free-energy crossing occurs at $_8$1 K within the present approximations—on the same scale as the experimental transition near 90 K, though not a precise prediction.

Two independent checks support the picture. First, specific-heat measurements yield an entropy change of approximately 85% of $_8$2 per unit cell, consistent with the calculated spin-entropy loss once impurity moments and remnant $_8$3-phase contributions are accounted for. Second, recent HREELS measurements showing a quasi-2D linearly dispersing exciton in the $_8$4 phase evolving into split quasi-3D excitonic bands in the $_8$5 phase directly corroborate the substantial reconstruction of interlayer coupling that underlies the dimerization mechanism.

Several approximations bound the quantitative accuracy of $_8$6: phonons are treated harmonically at zero-temperature structures without anharmonic renormalization or explicit electron-phonon coupling beyond static relaxation; the HTSE Padé continuation is reliable near $_8$7 but fails below roughly 20 K, where it cannot resolve any putative QSL state; itinerant electronic entropy is neglected (justified by both phases being insulating); and residual double-counting ambiguities in the correlated total energy remain.

Pressure response

To address the experimental observation that c-axis uniaxial pressure suppresses the transition, the authors construct a $_8$8 generalized enthalpy including the stress-work term $_8$9. The mechanism follows from differential compressibility: the $\alpha$0 phase, with its stronger interlayer bonding, is less compressible along c than the $\alpha$1 phase, so under equal applied pressure the $\alpha$2-phase enthalpy decreases more rapidly. The calculated enthalpy crossing yields a critical pressure $\alpha$3 GPa for complete suppression, while at 1.8 GPa—comparable to the experimental range—the transition is only partially suppressed, consistent with magnetic-susceptibility measurements. The authors flag that finite-temperature phonon and spin entropies under pressure are excluded, so this threshold is semiquantitative. By symmetry of argument, c-axis tension should stabilize the $\alpha$4 phase and raise the transition temperature.

The practical implication stated in the same context is direct: stabilizing the $\alpha$5 phase at low temperature via uniaxial pressure opens an experimental route to the candidate QSL regime that the structural transition otherwise masks.

Limitations and open questions

The paper is explicit about its boundaries. The harmonic treatment of phonons omits anharmonic effects of low-frequency interlayer modes, which may be significant given the slip-dominated character of the transition. The interpolation-path barrier is not the true kinetic barrier. The HTSE description of the $\alpha$6 phase cannot access low enough temperatures to characterize any QSL state, so the framework says nothing about whether suppressing the transition actually realizes one. The critical pressure neglects pressure-induced entropy renormalization. Finally, the structural assignment of the $\alpha$7 phase remains unsettled among $\alpha$8, $\alpha$9, and $P\bar{3}m1$0; the energetic differences among these candidates (2–4 meV per unit cell) are small enough that subtle distortions could still matter for fine spectroscopic details even if they do not affect the thermodynamic balance.

An open question the work leaves unresolved is the microscopic fate of the $P\bar{3}m1$1 phase under sufficient uniaxial pressure: whether the frustrated triangular-lattice magnetism indeed develops QSL behavior, or instead orders at temperatures accessible once the structural instability is removed.

Conclusion

This work establishes a unified thermodynamic account of the first-order structural transition in Nb$P\bar{3}m1$2Cl$P\bar{3}m1$3, resolving the competition between an entropy-stabilized high-temperature $P\bar{3}m1$4 phase and an energy-stabilized, dimerized $P\bar{3}m1$5 phase through a free-energy framework spanning DFT, harmonic lattice dynamics, and correlated spin models. The joint phonon-plus-spin entropy mechanism, the semiquantitative reproduction of the transition temperature and entropy change, and the prediction of pressure-induced suppression together provide a coherent explanation of existing experiments and a concrete thermodynamic handle—c-axis uniaxial pressure—for accessing the low-temperature physics of the $P\bar{3}m1$6 phase. Beyond this material, the framework offers a template for analyzing structural transitions in cluster Mott systems where lattice, charge, and spin degrees of freedom are comparably intertwined.