The spin-1/2 kagome Heisenberg antiferromagnet is a central model in the study of quantum spin liquids, yet its ground-state properties remain unresolved. We perform large-scale variational Monte-Carlo calculations using neural-network quantum states (NQSs) with kagome-geometry-aware message-passing convolution on two-dimensional tori containing up to 300 spins, and achieve exceptionally low variational energies for the kagome model. The spin and dimer structure factors show no evidence of magnetic or valence-bond long-range order. Moreover, the dominant symmetry-resolved spin and dimer correlations exhibit a common scaling behavior consistent with the emergent $SO(6)$ monopole multiplet of the Dirac spin liquid. Finally, the fidelity of the optimized NQSs with Gutzwiller-projected mean-field states increasingly favors the Dirac state over the competing $Z_{2}$ candidates as the system size grows. Together, these complementary results provide evidence for a $U(1)$ Dirac spin-liquid ground state in the kagome Heisenberg antiferromagnet.
The spin-1/2 kagome Heisenberg antiferromagnet is a central model in the study of quantum spin liquids, yet its ground-state properties remain unresolved. We perform large-scale variational Monte-Carlo calculations using neural-network quantum states (NQSs) with kagome-geometry-aware message-passing convolution on two-dimensional tori containing up to 300 spins, and achieve exceptionally low variational energies for the kagome model. The spin and dimer structure factors show no evidence of magnetic or valence-bond long-range order. Moreover, the dominant symmetry-resolved spin and dimer correlations exhibit a common scaling behavior consistent with the emergent SO(6) monopole multiplet of the Dirac spin liquid. Finally, the fidelity of the optimized NQSs with Gutzwiller-projected mean-field states increasingly favors the Dirac state over the competing Z2 candidates as the system size grows. Together, these complementary results provide evidence for a U(1) Dirac spin-liquid ground state in the kagome Heisenberg antiferromagnet.
Motivated by recent density-matrix renormalization group (DMRG) calculations [Yan, Huse, and White, Science 332, 1173 (2011)], which claimed that the ground state of the nearest-neighbor spin-1/2 Heisenberg antiferromagnet on the kagome lattice geometry is a fully gapped spin liquid with numerical signatures of Z2 gauge structure, and a further theoretical work [Lu, Ran, and Lee, Phys. Rev. B 83, 224413 (2011)], which gave a classification of all Schwinger-fermion mean-field fully symmetric Z2 spin liquids on the kagome lattice, we have thoroughly studied Gutzwiller-projected fermionic wave functions by using quantum variational Monte Carlo techniques, hence implementing exactly the constraint of one fermion per site. In particular, we investigated the energetics of all Z2 candidates (gapped and gapless) that lie in the neighborhood of the energetically competitive U(1) gapless spin liquids. By using a state-of-the-art optimization method, we were able to conclusively show that the U(1) Dirac state is remarkably stable with respect to all Z2 spin liquids in its neighborhood, and in particular for opening a gap toward the so-called ${\mathbb{Z}}_{2}[0,\ensuremath{\pi}]\ensuremath{\beta}$ state, which was conjectured to describe the ground state obtained by the DMRG method. Finally, we also considered the addition of a small second nearest-neighbor exchange coupling of both antiferromagnetic and ferromagnetic type, and obtained similar results, namely, a U(1) Dirac spin-liquid ground state.
We investigate the impact of spin-phonon coupling on the S=1/2 Heisenberg model on the kagome lattice. For the pure spin model, there is increasing evidence that the low-energy properties can be correctly described by a Dirac spin liquid, in which spinons with a conical dispersion are coupled to emergent gauge fields. Within this scenario, the ground-state wave function is well approximated by a Gutzwiller-projected fermionic state [Ran, Hermele, Lee, and Wen, Phys. Rev. Lett. 98, 117205 (2007)]. However, the existence of U(1) gauge fields may naturally lead to instabilities when small perturbations are included. Phonons are ubiquitous in real materials and may play a relevant role in the determination of the actual physical properties of the kagome antiferromagnet. Therefore, we perform a step forward in this direction, including phonon degrees of freedom (at the quantum level) and applying a variational approach based upon Gutzwiller-projected fermionic Ans\"atze. Our results suggest that the Dirac spin liquid is stable for small spin-phonon couplings, while valence-bond solids are obtained at large couplings. Even though different distortions can be induced by the spin-phonon interaction, the general aspect is that the energy is lowered by maximizing the density of perfect hexagons in the dimerization pattern.
Due to strong geometric frustration and quantum fluctuation, the S=1/2 quantum Heisenberg antiferromagnet on the kagome lattice has long been considered as an ideal platform to realize a spin liquid (SL), a phase exhibiting fractionalized excitations without any symmetry breaking. A recent numerical study (Yan et al., e-print arXiv:1011.6114) of the Heisenberg S=1/2,kagome lattice model (HKLM) shows, in contrast to earlier results, that the ground state is a singlet-gapped SL with signatures of Z2 topological order. Motivated by this numerical discovery, we use the projective symmetry group to classify all 20 possible Schwinger fermion mean-field states of Z2 SLs on the kagome lattice. Among them we found only one gapped Z2 SL (which we call the ${\mathbb{Z}}_{2}[0,\ensuremath{\pi}]\ensuremath{\beta}$ state) in the neighborhood of the U(1) Dirac SL state. Since its parent state, i.e., the U(1) Dirac SL, was found [Ran et al., Phys. Rev. Lett. 98, 117205 (2007)] to be the lowest among many other candidate U(1) SLs, including the uniform resonating-valence-bond states, we propose this ${\mathbb{Z}}_{2}[0,\ensuremath{\pi}]\ensuremath{\beta}$ state to be the numerically discovered SL ground state of the HKLM.
The spin- 1 / 2 kagome antiferromagnet (AFM) is one of the most studied models in frustrated magnetism since it is a promising candidate to host exotic spin-liquid states. However, despite numerous studies using both analytical and numerical approaches, the nature of the ground state and low-energy excitations in this system remains elusive. This challenge is related to the difficulty in determining the spin gap in various calculations. We present the results of our investigation of the kagome AFM using the recently developed group equivariant convolutional neural networks—a novel machine learning technique for studying strongly frustrated models. This approach, combined with variational Monte Carlo method, introduces significant improvement of the achievable results’ accuracy for frustrated spin systems in comparison with approaches based on other neural-network architectures. Contrary to the results obtained previously with various methods, which predicted Z 2 or U ( 1 ) Dirac spin-liquid states, our results strongly indicate that the ground state of the kagome lattice antiferromagnet is a spinon pair density wave that does not break time-reversal symmetry or any of the lattice symmetries. The state appears due to the spinon Cooper pairing instability close to two Dirac points in the spinon energy spectrum, and it resembles the pair density wave state studied previously in the context of underdoped cuprate superconductors in connection with the pseudogap phase. This state has significantly lower energy than the lowest-energy states found by the S U ( 2 ) symmetric density matrix renormalization group calculations and other methods.
Evidence for a Z2 Dirac spin liquid in the generalized Shastry-Sutherland model
We present a multimethod investigation of the recently reported quantum spin liquid (QSL) phase in the spin-1/2 Heisenberg antiferromagnet on the Shastry-Sutherland lattice. A comprehensive projective symmetry group classification of fermionic mean-field Ansätze on this lattice yields 46 U(1) and 80 Z2 states. Using density-matrix renormalization group (DMRG) and exact diagonalization, we find that the Shastry-Sutherland model and the square-lattice J1-J2 Heisenberg antiferromagnet share the same QSL phase. We therefore map our Ansätze to those on the square lattice and identify the counterpart of the square-lattice Z2 Dirac QSL (Z2Azz13) in the Shastry-Sutherland system. Using state-of-the-art variational Monte Carlo with Gutzwiller-projected wavefunctions improved by Lánczos steps, we demonstrate excellent agreement in energies and correlation functions between a gapless (Dirac) Z2 spin liquid, characterized by only a few variational parameters, and results from neural quantum states and DMRG. Within a unified variational family that continuously connects the spin liquid to the adjacent Néel and plaquette valence-bond-crystal states, we locate both phase boundaries in an unbiased way from order-parameter finite-size scaling and correlation-ratio crossings, and find that both order parameters vanish continuously. Finally, the recently developed Keldysh formulation of the pseudo-fermion functional renormalization group yields a dynamical spin structure factor with features consistent with Dirac cones, providing strong independent evidence for a Dirac QSL ground state. Our identification of a d-wave pairing Z2 Dirac QSL is consistent with recently observed signatures of QSL behavior in Pr2Ga2BeO7 and makes predictions for future experiments.
We study the energy and the static spin structure factor of the ground state of the spin-1/2 quantum Heisenberg antiferromagnetic model on the kagome lattice. By the iterative application of a few Lanczos steps on accurate projected fermionic wave functions and the Green's function Monte Carlo technique, we find that a gapless (algebraic) U(1) Dirac spin liquid is competitive with previously proposed gapped (topological) Z2 spin liquids. By performing a finite-size extrapolation of the ground-state energy, we obtain an energy per site $E/J=\ensuremath{-}0.4365(2)$, which is equal, within three error bars, to the estimates given by the density-matrix renormalization group (DMRG). Our estimate is obtained for a translationally invariant system, and, therefore, does not suffer from boundary effects, like in DMRG. Moreover, on finite toric clusters at the pure variational level, our energies are lower compared to those from DMRG calculations.