betaFeedback

WorksPhysics and Astronomy

Mott transitions in Dirac nodal-line semimetals

Abstract

Nodal-line semimetals provide a natural setting to investigate how the geometry of a nodal manifold affects interaction-driven quantum phase transitions. Here, we study the Hubbard model on a cubic lattice supporting either nodal-point (cubic ππ-flux model) or nodal-line semimetallic states using mean-field theory, determinant quantum Monte Carlo simulations, and a weak-coupling renormalization-group (RG) analysis. In both cases, antiferromagnetic order develops only above a finite interaction strength, with the nodal-line system exhibiting a smaller critical coupling due to its enhanced low-energy density of states and markedly stronger finite-size effects. For the nodal-point model, the quantum critical behavior is compatible with (3+1)(3+1)-dimensional Gross–Neveu mean-field scaling. In contrast, the weak-coupling RG reveals that the nodal line itself is unstable against interactions: although the Hubbard interaction is perturbatively irrelevant, its second-order self-energy generates a symmetry-allowed relevant term proportional to U2cos⁡kzU^2\cos k_z, reconstructing the nodal line into isolated, strongly anisotropic nodal points. This weak-coupling reconstruction is distinct from the strong-coupling antiferromagnetic transition and highlights the role of nodal-manifold geometry in interacting semimetals.

Introduction

Topological semimetals have attracted considerable attention as systems in which the low-energy electronic structure is governed by symmetry-protected band touchings. In Dirac and Weyl semimetals, the conduction and valence bands touch at isolated points in the Brillouin zone, giving rise to quasiparticles with linear dispersion around the nodal points [1, 2]. In three dimensions, Dirac semimetals may appear either at the transition between topologically distinct insulating phases or as stable semimetallic phases protected by crystalline symmetries [3, 4]. The pointlike character of the Fermi surface leads to a density of states that vanishes as ρ(ω)∼ω2\rho(\omega) \sim\omega^2, rendering weak short-range interactions perturbatively irrelevant [5].

Despite this weak-coupling stability, three-dimensional Dirac systems may display nontrivial many-body and disorder-driven phenomena. Long-range Coulomb interactions can renormalize quasiparticle properties and, at sufficiently strong coupling, drive dynamical mass generation or excitonic instabilities [6, 7]. Short-range interactions may also destabilize the semimetal and generate scalar, pseudoscalar, or axionic insulating masses [8]. In addition, disorder can induce transitions from an incompressible Dirac semimetal to a compressible diffusive metal and, at stronger disorder, to an Anderson insulating phase [8, 9, 10, 11]. These results establish three-dimensional Dirac semimetals as a useful toy model for studying the interplay between topology, correlations, and randomness.

A particularly useful lattice realization of three-dimensional Dirac fermions is the cubic π\pi-flux model [12]. In this system, each plaquette encloses a π\pi flux, and the noninteracting spectrum hosts linearly dispersing Dirac cones. The Hubbard model on this lattice has been investigated using determinant quantum Monte Carlo simulations, which showed that the Dirac semimetal remains stable against weak onsite repulsion and undergoes a direct transition to an antiferromagnetic insulating phase only above a finite critical interaction [13]. For three-dimensional Dirac fermions, the corresponding Gross–Neveu–Yukawa theory lies at its upper critical spacetime dimension dc=4d_c = 4, such that the critical behavior is Gaussian up to logarithmic corrections [14].

Nodal-line semimetals provide a natural extension of this setting. In these systems, the conduction and valence bands touch not at isolated points, but along one-dimensional manifolds in momentum space, which may form loops, chains, or more complex structures [15, 16]. The stability of these nodal lines is generally tied to crystalline or internal symmetries, such as time-reversal, inversion, mirror, or glide symmetry [17]. When the protecting symmetry is broken, the nodal line may become fully gapped or evolve into other topological band structures. Compared with a three-dimensional Dirac cone, a nodal-line semimetal has a larger low-energy phase space. Therefore, interaction effects are expected to be enhanced relative to point-node Dirac systems, although the vanishing density of states at the Fermi level still suggests that a stable weak-coupling semimetallic regime may survive. Indeed, related studies of excitonic quantum criticality in ZrSiS suggest that nodal-line systems may also display strong quasiparticle mass renormalization near a semimetal–excitonic-insulator quantum critical point, even in regimes where a full gap is not observed experimentally [7].

The role of electronic interactions in nodal-line semimetals is therefore a natural and timely question [18, 19, 20, 21]. Field-theoretical studies based on renormalization-group and Gross–Neveu–Yukawa approaches have shown that short-range interactions can drive continuous quantum phase transitions from a nodal-line semimetal into fully gapped phases in charge, spin, and superconducting channels [22]. In these approaches, the many-body instability occurs only beyond a finite interaction strength, and the critical exponents depend on the number of order-parameter components. In particular, the dynamical critical exponent is found to be z=1z = 1 at one-loop order, while the remaining exponents acquire corrections associated with the anomalous dimension of the bosonic order parameter [22].

More broadly, studies of anisotropic and semi-Dirac semimetals emphasize that the geometry of the band touching plays a central role in determining the effects of interactions. Systems with mixed linear and quadratic dispersions interpolate between relativistic Dirac fermions and quadratic band-touching systems, leading to modified density of states scaling and distinct quantum critical behavior [23, 24, 25]. In some cases, interactions can generate additional terms in the low-energy dispersion, modify the stability of the noninteracting fixed point, or promote several competing ordered phases [19, 20, 25, 26, 27]. These results indicate that the universality class of an interaction-driven transition cannot be inferred from spatial dimensionality alone, but depends sensitively on the structure of the nodal manifold and on the dispersion normal to it.

Motivated by these developments, we construct a nodal-line semimetal model starting from the cubic π\pi-flux lattice. By adding hopping terms to the parent Dirac model, we generate a continuous nodal line along (π/2,π/2,kz)(\pi/2,\pi/2,k_z) in the Brillouin zone, as depicted in Figure 1. This construction allows us to compare two closely related systems: the original cubic π\pi-flux Dirac semimetal, and a modified nodal-line semimetal. The central questions are whether the nodal line remains stable against weak interactions, whether antiferromagnetic order appears only above a finite critical coupling, and whether the corresponding critical behavior differs from that of the three-dimensional Dirac case.

Figure 1. Noninteracting Fermi surface at half filling for (a) the cubic π\pi-flux model and (b) the corresponding nodal-line model, with nodal lines located at (±π/2,±π/2,kz)(\pm\pi/2,\pm\pi/2,k_z).

In this work, we combine mean-field theory, quantum Monte Carlo simulations, and a weak-coupling renormalization-group analysis to study the Hubbard model on both the cubic π\pi-flux lattice and the modified nodal-line system. We first benchmark our calculations in the parent Dirac semimetal by examining the antiferromagnetic critical interaction, magnetic correlations, and finite-size scaling behavior. We then apply the same analysis to the nodal-line model in order to determine the stability of the semimetallic phase. In this way, our study clarifies how the Hubbard-driven quantum phase transition is modified when isolated Dirac points are promoted to an extended nodal line. Finally, a weak-coupling RG analysis reveals a distinction between the stability of the semimetal against symmetry-breaking order and the stability of the nodal manifold itself. The paper is organized as follows. In the next section, we introduce the model and describe the methodologies employed. The results for both the nodal-point and nodal-line cases are presented in Sec. III, while our conclusions are summarized in Sec. IV.

Model and methodology

The model

We investigate interacting fermions in a three-dimensional lattice model for nodal-point and nodal-line semimetals. The

noninteracting Hamiltonian (in momentum space) reads

H0=2t∑k,σ∑τ,μ∑τ′,μ′ck,τ,μ,σ†[(τx)ττ′δμμ′cos⁡kx+(τy)ττ′(μy)μμ′cos⁡ky+(τz)τ′δμμ′χ(k)]ck,τ′,μ′,σ.(1)\mathcal{H}_0 = 2t \sum_{\mathbf{k},\sigma} \sum_{\tau,\mu} \sum_{\tau',\mu'} c_{\mathbf{k},\tau,\mu,\sigma}^{\dagger} \left[ (\tau_x)_{\tau\tau'} \delta_{\mu\mu'} \cos k_x + (\tau_y)_{\tau\tau'} (\mu_y)_{\mu\mu'} \cos k_y + (\tau_z)_{\tau'} \delta_{\mu\mu'} \chi(\mathbf{k}) \right] c_{\mathbf{k},\tau',\mu',\sigma}. \tag*{(1)}

Here τi\tau_i and μi\mu_i are Pauli matrices acting on two independent orbital degrees of freedom, and σ\sigma denotes the spin. This general Hamiltonian is useful because it allows us to interpolate between nodal-point (NP) and nodal-line (NL) regimes by choosing, respectively,

χ(k)={χNP(k)=−cos⁡kz,χNL(k)=−cos⁡kz(cos⁡2kx+cos⁡2ky).(2)\chi(\mathbf{k}) = \begin{cases} \chi_{\mathrm{NP}}(\mathbf{k}) = -\cos k_z, \\ \chi_{\mathrm{NL}}(\mathbf{k}) = -\cos k_z\left(\cos^2 k_x+\cos^2 k_y\right). \end{cases} \tag*{(2)}

As discussed below, for χ=χNP\chi=\chi_{\mathrm{NP}} the spectrum exhibits nodal/Dirac points in the Brillouin zone, as shown in Figure 1(a). By contrast, for χ=χNL\chi=\chi_{\mathrm{NL}}, the band touching extends into a nodal line along the kzk_z direction, as shown in Figure 1(b).

For convenience, we combine the two orbital labels (τ,μ)(\tau,\mu) into a single index α=A,B,C,D\alpha=A,B,C,D, according to (0,0)≡A(0,0)\equiv A, (0,1)≡B(0,1)\equiv B, (1,0)≡C(1,0)\equiv C, (1,1)≡D(1,1)\equiv D. Within this notation, Eq. (1) becomes H0=∑k,σΨkσ†h^0(k)Ψkσ\mathcal{H}_0=\sum_{\mathbf{k},\sigma}\Psi_{\mathbf{k}\sigma}^{\dagger}\hat{h}_0(\mathbf{k})\Psi_{\mathbf{k}\sigma}, where

H^0(k)=2t(χ(k)0cos⁡kx−cos⁡ky0χ(k)cos⁡kycos⁡kxcos⁡kxcos⁡ky−χ(k)0−cos⁡kycos⁡kx0−χ(k)),(3)\hat{H}_0(\mathbf{k})=2t \begin{pmatrix} \chi(\mathbf{k}) & 0 & \cos k_x & -\cos k_y \\ 0 & \chi(\mathbf{k}) & \cos k_y & \cos k_x \\ \cos k_x & \cos k_y & -\chi(\mathbf{k}) & 0 \\ -\cos k_y & \cos k_x & 0 & -\chi(\mathbf{k}) \end{pmatrix}, \tag*{(3)}

with Ψkσ=(cA,kσ,cB,kσ,cC,kσ,cD,kσ)T\Psi_{\mathbf{k}\sigma}=(c_{A,\mathbf{k}\sigma},c_{B,\mathbf{k}\sigma},c_{C,\mathbf{k}\sigma},c_{D,\mathbf{k}\sigma})^T being a four-component spinor in orbital space. Regardless of the choice of χ(k)\chi(\mathbf{k}), the corresponding energy bands are doubly degenerate and given by

E±,k=±2tcos⁡2kx+cos⁡2ky+χ2(k).(4)E_{\pm,\mathbf{k}=\pm2t\sqrt{\cos^2 k_x+\cos^2 k_y+\chi^2(\mathbf{k})}.} \tag*{(4)}

Therefore, band touchings occur when cos⁡kx=cos⁡ky=χ(k)=0\cos k_x=\cos k_y=\chi(\mathbf{k})=0. For χNL\chi_{\mathrm{NL}}, this condition is satisfied along the line (kx,ky)=(π/2,π/2)(k_x,k_y)=(\pi/2,\pi/2), with arbitrary kzk_z, while for χNP\chi_{\mathrm{NP}}, the last condition reduces to cos⁡kz=0\cos k_z=0, producing Dirac points at (kx,ky,kz)=(π/2,π/2,π/2)(k_x,k_y,k_z)=(\pi/2,\pi/2,\pi/2), as shown in Figure 1. In this representation, the noninteracting Hamiltonian reads

H^0(k)=2t[cos⁡kxΓ1+cos⁡kyΓ2+χ(k)Γ3](5)\hat{H}_0(\mathbf{k}) = 2t\left[\cos k_x \Gamma_1+\cos k_y \Gamma_2+\chi(\mathbf{k})\Gamma_3\right] \tag*{(5)}

where

Γ1=τx⊗μ0,Γ2=τy⊗μy,Γ3=τz⊗μ0,Γ4=τy⊗μx,Γ5=τy⊗μz.(6)\begin{aligned} \Gamma_1 &= \tau_x\otimes\mu_0,\qquad\Gamma_2=\tau_y\otimes\mu_y,\\ \Gamma_3 &= \tau_z\otimes\mu_0,\qquad\Gamma_4=\tau_y\otimes\mu_x,\\ \Gamma_5 &= \tau_y\otimes\mu_z. \tag*{(6)} \end{aligned}

These matrices satisfy the Clifford algebra {Γa,Γb}=2δabI4\{\Gamma_a,\Gamma_b\}=2\delta_{ab}I_4, with a,b=1,…,5a,b=1,\ldots,5.

The real-space representation of the Hamiltonian in Eqs. (1) or Eq. (5) is given in Appendix A, using the primitive vectors

a1=(2,0,0)a,a2=(0,2,0)a,a3=(0,0,1)a,(7)\mathbf{a}_1=(2,0,0)a,\qquad\mathbf{a}_2=(0,2,0)a,\qquad\mathbf{a}_3=(0,0,1)a, \tag*{(7)}

with aa being the lattice spacing. The NP limit corresponds to the three-dimensional cubic π\pi-flux Dirac model [12], as displayed in Figure 2, while the NL limit is obtained by adding extra hopping processes to this parent NP Hamiltonian. Some noninteracting properties of the NL case are discussed in Appendix B. Hereafter, we measure energies in units of the hopping amplitude tt and set the lattice constant to unity.

Figure 2. Real-space representation of the cubic π\pi-flux model. The four orbitals within the unit cell are labeled AA, BB, CC, and DD. Solid and dashed bonds denote nearest-neighbor hopping amplitudes +t+t and −t-t, respectively, arranged such that each elementary plaquette is threaded by a π\pi flux.

The mass term

Despite its simplicity, Eq. (5) does not make explicit the mass term that appears when antiferromagnetic (AFM) order emerges. From the real-space representation (see Appendix A), one can alternatively define the primitive vectors as

a~1=(1,1,0)a,a~2=(1,−1,0)a,a~3=(0,0,2)a.(8)\widetilde{\mathbf{a}}_1=(1,1,0)a,\qquad\widetilde{\mathbf{a}}_2=(1,-1,0)a,\qquad\widetilde{\mathbf{a}}_3=(0,0,2)a. \tag*{(8)}

For this choice, the orbital positions inside the unit cell are

δA=(0,0,0),δC=(1,0,0),δA′=(0,0,1),δC′=(1,0,1),(9)\delta_A=(0,0,0),\qquad\delta_C=(1,0,0),\qquad \delta_{A'}=(0,0,1),\qquad\delta_{C'}=(1,0,1), \tag*{(9)}

and a local gauge transformation crσ=qydrσc_{\mathbf{r}\sigma}=q_y d_{\mathbf{r}\sigma}, with qy=1q_y=1 for even yy and qy=iq_y=i for odd yy, is required to make the hopping amplitudes periodic.

Performing a Fourier transform, the noninteracting Hamiltonian reads

H~0=∑k,σΨ~kσ†H~0(k)Ψ~kσ,(10)\widetilde{\mathcal{H}}_0=\sum_{\mathbf{k},\sigma}\widetilde{\Psi}^{\dagger}_{\mathbf{k}\sigma}\widetilde{H}_0(\mathbf{k})\widetilde{\Psi}_{\mathbf{k}\sigma}, \tag*{(10)}

with

H~0(k)=2t[cos⁡kxΓ~1+cos⁡kyΓ~2+χ(k)Γ~3],(11)\widetilde{H}_0(\mathbf{k})=2t\left[\cos k_x\widetilde{\Gamma}_1+\cos k_y\widetilde{\Gamma}_2+\chi(\mathbf{k})\widetilde{\Gamma}_3\right], \tag*{(11)}

where Ψ~kσ=(dA,kσ,dC,kσ,dA′,kσ,dC′,kσ)T\widetilde{\Psi}_{\mathbf{k}\sigma}=(d_{A,\mathbf{k}\sigma},d_{C,\mathbf{k}\sigma},d_{A',\mathbf{k}\sigma},d_{C',\mathbf{k}\sigma})^{T} and σ\sigma labels the electron spin. Similarly, the matrices

Γ~1=μ~0⊗τ~x,Γ~2=μ~0⊗τ~y,Γ~3=μ~x⊗τ~z,Γ~4=μ~z⊗τ~z,Γ~5=μ~y⊗τ~z(12)\begin{aligned} \widetilde{\Gamma}_1&=\widetilde{\mu}_0\otimes\widetilde{\tau}_x,\qquad\widetilde{\Gamma}_2=\widetilde{\mu}_0\otimes\widetilde{\tau}_y,\\ \widetilde{\Gamma}_3&=\widetilde{\mu}_x\otimes\widetilde{\tau}_z,\qquad\widetilde{\Gamma}_4=\widetilde{\mu}_z\otimes\widetilde{\tau}_z,\\ \widetilde{\Gamma}_5&=\widetilde{\mu}_y\otimes\widetilde{\tau}_z \tag*{(12)} \end{aligned}

satisfy the Clifford algebra {Γ~a,Γ~b}=2δabI4\{\widetilde{\Gamma}_a,\widetilde{\Gamma}_b\}=2\delta_{ab}I_4. Here μ~i\widetilde{\mu}_i and τ~i\widetilde{\tau}_i are Pauli matrices acting on two independent orbital indices: μ~i\widetilde{\mu}_i distinguishes the unprimed layer (A,C)(A,C) from the primed layer (A′,C′)(A',C'), while τ~i\widetilde{\tau}_i distinguishes the AA and CC sites within each layer, with μ~0=τ~0=I2\widetilde{\mu}_0=\widetilde{\tau}_0=I_2.

Within this representation, a staggered antiferromagnetic pattern is given by a term σΔΓ~4\sigma\Delta\widetilde{\Gamma}_4, where σ=±1\sigma=\pm1, and Δ\Delta is proportional to the magnetic order parameter, with ordering wavevector Q=0\mathbf{Q}=0. Since {Γ~4,H~0(k)}=0\{\widetilde{\Gamma}_4,\widetilde{H}_0(\mathbf{k})\}=0, this contribution is a fermion mass, making explicit the coupling between the AFM order parameter and the Dirac fermions, as in Gross–Neveu systems.

Nevertheless, hereafter, we adopt the unit cell defined in Eq. (7), retaining the original hopping gauge. Although this choice makes the AFM mass term less explicit, it is particularly convenient for quantum Monte Carlo simulations. It changes the orbital grouping and the corresponding Brillouin zone without altering the underlying lattice model.

Mean-field approximation

We now include the onsite Hubbard interaction,

HU=U∑R,αnRα↑nRα↓.(13)\mathcal{H}_U=U\sum_{\mathbf{R},\alpha}n_{\mathbf{R}\alpha\uparrow}n_{\mathbf{R}\alpha\downarrow}. \tag*{(13)}

As outlined in Appendix C, within a static mean-field approximation, assuming an AFM order, one obtains the dispersion ϵ±,k,σ(Δ)=±ϵk,σ(Δ)\epsilon_{\pm,\mathbf{k},\sigma}(\Delta)=\pm\epsilon_{\mathbf{k},\sigma}(\Delta), where

ϵk,σ(Δ)=2tcos⁡2kx+cos⁡2ky+χ2(k)+Δ2,(14)\epsilon_{\mathbf{k},\sigma}(\Delta)=2t\sqrt{\cos^2 k_x+\cos^2 k_y+\chi^2(\mathbf{k})+\Delta^2}, \tag*{(14)}

with Δ=Um4t\Delta=\frac{Um}{4t}, and mm being the AFM order parameter. Here, the staggered AFM order is characterized by the ordering wave vector Q=(0,0,π)\mathbf{Q}=(0,0,\pi) in the extended Brillouin zone.

Given this, the mean-field free energy is obtained by summing over the quasiparticle levels,

F(Δ)=−8β∑k,σ=±ln⁡[1+e−βEp,k,σ(Δ)]+64Nt2UΔ2(15)F(\Delta)=-\frac{8}{\beta}\sum_{\mathbf{k},\sigma=\pm}\ln\left[1+e^{-\beta E_{\mathbf{p},\mathbf{k},\sigma}(\Delta)}\right]+64N\frac{t^2}{U}\Delta^2 \tag*{(15)}

with Δ\Delta being obtained self-consistently by minimizing F(Δ)F(\Delta). That is, by employing ∂F(Δ)/∂Δ=0\partial F(\Delta)/\partial\Delta=0 at zero temperature, one obtains

tU=14N∑k1cos⁡2kx+cos⁡2ky+χ2(k)+Δ2(16)\frac{t}{U}=\frac{1}{4N}\sum_{\mathbf{k}}\frac{1}{\sqrt{\cos^2 k_x+\cos^2 k_y+\chi^2(\mathbf{k})+\Delta^2}} \tag*{(16)}

which provides the order parameter mm. The critical interaction strength UcU_c is obtained by taking the limit Δ→0\Delta\to0,

tUc=14N∑k1cos⁡2kx+cos⁡2ky+χ2(k)(17)\frac{t}{U_c}=\frac{1}{4N}\sum_{\mathbf{k}}\frac{1}{\sqrt{\cos^2 k_x+\cos^2 k_y+\chi^2(\mathbf{k})}} \tag*{(17)}

Determinant quantum Monte Carlo method

We now turn to the determinant quantum Monte Carlo (DQMC) simulations [28, 29, 30, 31]. This method provides a numerically exact treatment of the interacting lattice problem, up to statistical errors and the controllable Trotter–Suzuki discretization error.

Considering our Hamiltonian as the sum of a quadratic kinetic term, K^≡H0\hat{K}\equiv\mathcal{H}_0, and the interaction term, P^≡HU\hat{P}\equiv\mathcal{H}_U, we perform a Trotter–Suzuki decomposition in the partition function,

Z=Tr⁡e−βH=Tr⁡[e−Δτ(K^+P^)]Lτ≈Tr⁡[e−ΔτK^e−ΔτP^]Lτ.(18)\begin{aligned} Z&=\operatorname{Tr}e^{-\beta\mathcal{H}}\\ &=\operatorname{Tr}\left[e^{-\Delta\tau(\hat{K}+\hat{P})}\right]^{L_\tau}\\ &\approx\operatorname{Tr}\left[e^{-\Delta\tau\hat{K}}e^{-\Delta\tau\hat{P}}\right]^{L_\tau}. \tag*{(18)} \end{aligned}

That is, we write the partition function as an imaginary-time path integral by discretizing the interval τ∈[0,β)\tau\in[0,\beta) into LτL_\tau slices of width Δτ\Delta\tau, with β=LτΔτ\beta=L_\tau\Delta\tau being the inverse temperature. The decomposition in Eq. (18) introduces systematic errors of order (Δτ)2(\Delta\tau)^2 in measured observables. In our simulations we use Δτ≤0.1\Delta\tau\leq0.1, for which the remaining discretization effects are smaller than, or comparable to, the statistical uncertainties.

The quartic interaction term is then decoupled by a discrete Hubbard–Stratonovich transformation [32]. For each spacetime point (Rα,τ)(\mathbf{R}_\alpha,\tau), we introduce an auxiliary Ising field sRα,τ=±1s_{\mathbf{R}_\alpha,\tau}=\pm1 through

e−ΔτU(n↑−12)(n↓−12)=12e−ΔτU/4∑s=±1eλs(n↑−n↓)(19)e^{-\Delta\tau U(n_\uparrow-\frac{1}{2})(n_\downarrow-\frac{1}{2})}=\frac{1}{2}e^{-\Delta\tau U/4}\sum_{s=\pm1}e^{\lambda s(n_\uparrow-n_\downarrow)} \tag*{(19)}

where the coupling λ\lambda is defined by cosh⁡λ=eΔτU/2\cosh\lambda=e^{\Delta\tau U/2}. After this transformation, the fermionic degrees of freedom appear only quadratically and can be traced out exactly for a fixed auxiliary-field configuration. The partition function becomes

Z=∑{s}W({s}),W({s})=det⁡M↑({s})det⁡M↓({s}),(20)Z=\sum_{\{s\}}W(\{s\}),\qquad W(\{s\})=\det M_\uparrow(\{s\})\det M_\downarrow(\{s\}), \tag*{(20)}

where MσM_\sigma is the fermionic matrix associated with spin sector σ\sigma in the background of the Hubbard–Stratonovich field. Physical observables are obtained by sampling the auxiliary fields with probability proportional to W({s})W(\{s\}).

In general, DQMC simulations of fermionic systems may suffer from the minus-sign problem when the product of determinants is not positive [33, 34]. In the present work, however, we restrict ourselves to the half-filled case, where particle-hole symmetry ensures sign-problem-free simulations for the models considered here. This allows us to access the low-temperature regime needed to analyze the onset of antiferromagnetic order and the stability of the semimetallic phase.

For each set of parameters, we perform between 10 and 30 independent DQMC runs initialized with different random seeds. In each run, we discard 0.80.8–2×1032\times10^3 sweeps for thermalization and subsequently accumulate measurements over 0.60.6–3×1033\times10^3 sweeps, depending on temperature and lattice size. These simulation parameters were chosen to ensure proper equilibration and reliable ensemble-averaged results. We also average over periodic (PBC) and antiperiodic (APBC) boundary conditions, which reduces finite-size effects for thermodynamic quantities and correlation functions.

Results

Noninteracting and mean-field cases

We first compare the NP and NL limits at the noninteracting level. Figure 3 (a) and Figure 3 (b) show the dispersion relation of Eq. (4) along high-symmetry directions for the NP and NL cases, respectively. In the nodal-point case, the spectrum exhibits a Dirac cone centered at (π/2,π/2,π/2)(\pi/2,\pi/2,\pi/2). By contrast, in the NL case, the bands touch along the kzk_z direction when kx=ky=π/2k_x=k_y=\pi/2, producing a dispersionless nodal feature as a function of kzk_z.

Figure 3. (a) Dispersion along the high-symmetry path and (c) density of states for nodal points. (b) and (d) show the same for nodal lines.

This difference in the low-energy band structure leads to distinct densities of states (DOS). As shown in Appendix B, the low-energy DOS can be obtained from the continuum expansion around the corresponding nodal manifold. For the NP case, one finds

ρNP(E)=E22π2t3(21)\rho_{\mathrm{NP}}(E)=\frac{E^2}{2\pi^2t^3} \tag*{(21)}

which shows that the DOS vanishes quadratically at the nodal point. In contrast, the NL case gives

ρNL(E)=∣E∣2πt2(22)\rho_{\mathrm{NL}}(E)=\frac{\lvert E\rvert}{2\pi t^2} \tag*{(22)}

corresponding to a DOS that vanishes linearly with energy. This behavior is confirmed by the numerical DOS shown in Figure 3 (c) and Figure 3 (d) for the NP and NL cases, respectively. The DOS analysis is useful because the different low-energy power laws provide the first indication that the two systems may respond differently once electronic interactions are included, as discussed below.

The distinction between the NP and NL cases is already visible at the mean-field level. By solving the self-consistent Eq. (16), we obtain the order parameter as a function of UU, as shown in Figure 4. Both cases exhibit a finite critical interaction for the onset of AFM order: we find Uc,NP/t≈4.392U_{c,\mathrm{NP}}/t \approx4.392 and Uc,NL/t≈3.538U_{c,\mathrm{NL}}/t \approx3.538. That is, the nodal-line system becomes unstable toward AFM order at a smaller interaction strength than the nodal-point system.

Figure 4. Mean-field AFM order parameter mm as a function of UU for the NP (purple dot-dashed line) and NL (blue dashed line) cases in the thermodynamic limit.

The behavior of the order parameter close to the transition also differs between the two cases. Assuming the thermodynamic limit for the NP case, the order parameter follows the conventional mean-field square-root behavior as a function of U−UcU-U_c, with the critical exponent β~≈1/2\widetilde{\beta} \approx1/2 (not shown). On the other hand, for the NL case, it increases approximately linearly with U−UcU-U_c, i.e., with β~≈1\widetilde{\beta} \approx1 (not shown). These differences for UcU_c and β~\widetilde{\beta} are consistent with the distinct low-energy DOS of the two systems. [35]

Considering the NL limit and expanding the dispersion around kx=ky=π/2k_x=k_y=\pi/2, with q⊥=(qx,qy)\mathbf{q}_{\perp}=(q_x,q_y) being the momenta transverse to the nodal line, one obtains

E(q⊥,kz,Δ)=±2t∣q⊥∣2+∣q⊥∣4cos⁡2(kz)+Δ2.(23)E(\mathbf{q}_{\perp},k_z,\Delta)=\pm2t\sqrt{|\mathbf{q}_{\perp}|^2+|\mathbf{q}_{\perp}|^4\cos^2(k_z)+\Delta^2}. \tag*{(23)}

Here kzk_z parametrizes the position along the nodal line. In the low-energy limit, ∣q⊥∣→0|\mathbf{q}_{\perp}|\to0, the term proportional to ∣q⊥∣4|\mathbf{q}_{\perp}|^4 is subleading, and the dispersion reduces to

E(q⊥,kz,Δ)≈±2t∣q⊥∣2+Δ2.(24)E(\mathbf{q}_{\perp},k_z,\Delta)\approx\pm2t\sqrt{|\mathbf{q}_{\perp}|^2+\Delta^2}. \tag*{(24)}

Thus, the nonanalytic dependence of the mean-field free energy on Δ\Delta is governed by the two transverse momentum directions, while the integration along the nodal line contributes only a finite prefactor. This is the same scaling structure that appears in two-dimensional Dirac systems, such as the honeycomb lattice. As a consequence, the NL case yields a mean-field magnetization exponent β~=1\widetilde{\beta}=1, in contrast to the NP case, for

which the conventional mean-field result β~=1/2\widetilde{\beta}=1/2 is obtained. Therefore, although the lattice is three-dimensional, the critical behavior of the NL semimetal is controlled by the codimension dc=2d_c=2 of the nodal manifold. That is, as the dispersion along the nodal line is subleading, we expect it to weakly modifies the leading scaling near the transition.

Quantum Monte Carlo: equal-time results

We now turn to the interacting regime using DQMC simulations. Since the model is three-dimensional, AFM long-range order may occur at finite temperature. Therefore, as a first thermodynamic probe, we analyze the energy densities,

e(T)=⟨H⟩N,(25)e(T)=\frac{\langle\mathcal{H}\rangle}{N}, \tag*{(25)}

where N=4×L3N=4\times L^3 is the total number of lattice sites, and LL is the linear size. Figures 5 (a) and 5 (d) display e(T)e(T) for the NP and NL cases, respectively. In fact, the quantity more directly related to a phase transition is the temperature derivative of the energy, namely the specific heat, cv(T)=de(T)dTc_v(T)=\frac{de(T)}{dT}. As the numerical QMC data may be noisy, it is convenient to perform a nonlinear fit of the energy density as

efit(T)=a0+∑n=1Mane−βnΔ,(26)e_{\mathrm{fit}}(T)=a_0+\sum_{n=1}^{M}a_n e^{-\beta n\Delta}, \tag*{(26)}

where MM is a cutoff typically chosen in the range 4–7, while ana_n and Δ\Delta are fitting parameters [36]. This procedure provides a smooth estimate of the energy and avoids amplifying Monte Carlo fluctuations when taking temperature derivatives.

Figures 5 (b) and 5 (e) show representative results for U/t=4U/t=4 in the NP and NL cases, respectively. The symbols correspond to the numerical derivative of the raw QMC data, while the solid lines are obtained from defit(T)/dTde_{\mathrm{fit}}(T)/dT, which agree well. Therefore, applying the same procedure to the remaining values of UU, we obtain the specific heat shown in Figure 5 (c) and Figure 5 (f) for the NP and NL cases, respectively. For small interactions, both the NP and NL cases display a single broad maximum in cv(T)c_{\mathrm{v}}(T), resembling the noninteracting behavior. As U/tU/t increases, this structure evolves into a two-peak profile. The higher-temperature peak is naturally associated with single-particle excitations, while the lower-temperature feature is associated with low-lying collective spin-wave excitations, consistent with the onset of AFM order. In both cases, this low-temperature peak becomes more evident at large interaction strength, for instance at U/t=9U/t = 9.

Figure 5. Energy density and specific heat for the NP and NL cases on a lattice with N=4×43N = 4 \times4^3 sites. Panels (a) and (d) show e(T)e(T), with symbols denoting DQMC data and lines the fits efit(T)e_{\mathrm{fit}}(T). Panels (b) and (e) compare the numerical derivative of e(T)e(T) with the derivative of efit(T)e_{\mathrm{fit}}(T) for U/t=4U/t = 4. Panels (c) and (f) show the resulting cv(T)c_{\mathrm{v}}(T) for different interaction strengths from the fit.

Although the specific heat provides a clear indication of a phase transition as UU increases, it does not give a sharp determination of the quantum critical point. Therefore, to connect these thermodynamic signatures with magnetic correlations, we analyze the real-space spin-spin correlation function

c(ri−rj)=⟨SizSjz⟩,(27)c(\mathbf{r}_i-\mathbf{r}_j)=\langle S_i^z S_j^z\rangle, \tag*{(27)}

as a function of temperature, with Siz=ni↑−ni↓S_i^z=n_{i\uparrow}-n_{i\downarrow}. We focus on correlations between equivalent orbitals, averaging over the AA-AA, BB-BB, CC-CC, and DD-DD contributions. In order to probe the longest-distance correlations available on the finite lattice, we take rmax⁡≡ri−rj=(L/2,L/2,L/2)\mathbf{r}_{\max}\equiv\mathbf{r}_i-\mathbf{r}_j=(L/2,L/2,L/2). Figure 6 shows this quantity for L=4L=4, corresponding to N=256N=256 sites, for (a) the NP and (b) the NL cases. For comparison, we introduce a small reference value at 10−310^{-3}, shown as the horizontal dashed line.

Figure 6. Long-distance same-orbital spin-spin correlation function as a function of temperature for (a) the NP case and (b) the NL case on a lattice with L=4L=4 (N=256N=256 sites). The horizontal dashed line marks an arbitrary reference value below which the correlations are considered negligible in the present finite-size analysis.

At high temperatures, the long-distance correlations are

strongly suppressed for all interaction strengths, as expected from the short-range character of the spin correlations in this regime. As the temperature is lowered, however, c(rmax⁡)c(\mathbf{r}_{\max}) increases significantly for sufficiently large U/tU/t, indicating the onset of long-range magnetic correlations. The temperature scale at which these long-distance correlations become relevant is consistent with the low-temperature peak observed in the specific heat, shown in Figure 5 (c) and (Figure 5f), which is around T/t≈0.4T/t\approx0.4 for large interactions. This supports the above interpretation that the low-temperature peak is associated with the development of a long-range AFM order.

This analysis also provides useful insights into the zero-temperature magnetic phase transition. As the temperature is lowered, the longest-distance correlations remain small for U/t≲7U/t\lesssim7 in the NP case, as shown in Figure 6 (a), even at the lowest temperatures considered. Similarly, in the NL case, shown in Figure 6 (b), the correlations are strongly suppressed for U/t≲6U/t\lesssim6 and become appreciable only at larger values of U/tU/t. Therefore, the onset of AFM ordering is expected to occur around these interaction scales, which explains the single-peak structure of the specific heat for weaker interactions.

Beyond these hints from the specific heat and c(rmax⁡)c(\mathbf{r}_{\max}), the emergence of a long-range ordered phase at low TT is determined by a finite-size scaling analysis of the spin structure factor

Figure 8. Staggered spin structure factor SsS_s for the NL case as a function of inverse temperature for different system sizes at (a) U/t=5U/t = 5, (b) U/t=6U/t = 6, and (c) U/t=6.8U/t = 6.8. Panel (d) shows the low-temperature finite-size analysis of the corresponding order parameter m(L)m(L).

Ss(q)=1N∑i,jeiq⋅(ri−rj)⟨SizSjz⟩(28)S_s(\mathbf{q}) = \frac{1}{N}\sum_{i,j}e^{i\mathbf{q}\cdot(\mathbf{r}_i-\mathbf{r}_j)}\langle S_i^z S_j^z\rangle \tag*{(28)}

where q\mathbf{q} is a wavevector of the lattice. Because the unit cell is chosen as 2×2×12 \times2 \times1, the staggered AFM order occurs at Q=(0,0,π)\mathbf{Q} = (0,0,\pi), in the extended Brillouin zone. Hereafter, we denote Ss≡Ss(Q)S_s \equiv S_s(\mathbf{Q}) as the staggered AFM structure factor. [37]

Figure Figure 7 (a) shows SsS_s as a function of U/tU/t for the NP case, using lattices with L=4,6,8L = 4, 6, 8, and 1010. Due to the presence of Lorentz symmetry, we set β=L\beta= L, in line with the low-energy theory that provides a dynamical exponent z=1z = 1. For weak interactions, SsS_s remains small and only weakly dependent on system size, consistent with the stability of the semimetallic phase. At larger values of U/tU/t, SsS_s grows rapidly with LL, signaling the development of long-range AFM correlations. The change in the behavior occurs around U/t≈7.5U/t \approx7.5, which provides a first estimate of the critical point.

Figure 7. (a) Staggered AFM structure factor SsS_s and (b) its corresponding correlation ratio RcR_c as functions of U/tU/t for different system sizes in the NP case (cubic π\pi-flux Hubbard model). The horizontal dashed line correspond to the critical point Uc/t≈7.8U_c/t \approx7.8. The insets describe their respective finite-size scaling analysis, providing ν~≈β~≈1/2\tilde{\nu} \approx\tilde{\beta} \approx1/2.

A more precise determination of UcU_c is obtained by using the correlation ratio [38, 39, 40]

Rc(L)=1−Ss(Q−δq)Ss(Q),(29)R_c(L) = 1 - \frac{S_s(\mathbf{Q}-\delta\mathbf{q})}{S_s(\mathbf{Q})}, \tag*{(29)}

where ∣δq∣=2π/L|\delta\mathbf{q}| = 2\pi/L is the smallest momentum interval allowed on a lattice with linear size LL. In the absence of long-range AFM order, the spin correlations remain short ranged and Rc→0R_c \to0 in the thermodynamic limit. By contrast, in the ordered phase, Ss(Q)S_s(\mathbf{Q}) grows more strongly than Ss(Q−δq)S_s(\mathbf{Q}-\delta\mathbf{q}), leading to Rc→1R_c \to1. At the critical point, RcR_c is a scale-invariant quantity.

Figure Figure 7 (b) shows the correlation ratio as a function of the interaction strength for the corresponding data of panel (a) [41]. A clear crossing occurs near Uc/t=7.8(1)U_c/t = 7.8(1), signaling the onset of AFM long-range order for U>UcU > U_c, and determining the value of the critical point for the NP case. This estimated critical strength is larger than the value reported in Ref. [13] for the same model. This difference may originate from details of the finite-size analysis, boundary conditions, and the range of lattice sizes considered. Since our simulations include lattices up to N=4×103N = 4 \times10^3 sites with different boundary conditions, the present result provides a benchmark for the cubic π\pi-flux Hubbard model.

We next examine the correlation-length exponent ν~\tilde{\nu} from the finite-size collapse of the correlation ratio. Since the crossing analysis gives Uc/t≈7.8U_c/t \approx7.8, we use the scaling form

Rc(L,U)=F[L1/ν~(U−Uc)Uc].(30)R_c(L,U) = \mathcal{F}\left[L^{1/\tilde{\nu}}\frac{(U-U_c)}{U_c}\right]. \tag*{(30)}

As shown in the inset of Figure 7 (b), the data collapse is consistent with ν~≈1/2\tilde{\nu} \approx1/2, in agreement with the mean-field exponent.

One may also estimate the order-parameter exponent β~\tilde{\beta} from the scaling of Ss(Q)S_s(\mathbf{Q}) [42], which obeys

Ss(L,U)=Ld−2β~/ν~F[L1/ν~(U−Uc)Uc],(31)S_s(L,U) = L^{d-2\tilde{\beta}/\tilde{\nu}}\mathcal{F}\left[L^{1/\tilde{\nu}}\frac{(U-U_c)}{U_c}\right], \tag*{(31)}

with d=3d = 3. Using UcU_c and ν~\tilde{\nu} obtained from the crossing of RcR_c, we can estimate the exponent β~\tilde{\beta} from the data collapse of Ss/Ld−2β~/ν~S_s/L^{d-2\tilde{\beta}/\tilde{\nu}}, as shown in the inset of Figure 7 (a). We obtain β~≈1/2\tilde{\beta} \approx1/2, also consistent with the expected mean-field critical behavior of the NP transition, in line with the collapse of the correlation ratio. Thus, these results provide an independent verification that the semimetal–AFM quantum phase transition in the cubic π\pi-flux model is compatible with the (3 + 1)-dimensional Gross–Neveu mean-field scaling [14].

We now turn to the NL case. In contrast to the NP limit, the nodal-line system is anisotropic and does not possess the same Dirac cone structure. Therefore, we do not impose the ground-state projection condition β=L\beta= L a priori. Although field-theoretical analyses suggest z≈1z \approx1 for nodal-line semimetals [22], here we adopt a more direct numerical procedure. We first analyze the temperature dependence of the spin structure factor and then perform the correlation ratio analysis only in a temperature regime where SsS_s has saturated within the accessible system sizes.

Figure 8 (a)–(c) shows SsS_s as a function of inverse temperature for several lattice sizes and interaction strengths. At high temperatures (small β\beta), the spin correlations are short ranged and SsS_s is weakly dependent on system size. As the temperature is lowered (larger β\beta), the magnetic correlation length increases and may become comparable to the linear system size, depending on the interaction strength. For U/t=5U/t = 5, in panel (a), the spin structure factor remains weakly dependent on LL over the accessible temperature range, which is consistent with the absence of long-range AFM order. At U/t=6U/t = 6, a weak size dependence begins to develop, while for U/t=6.8U/t = 6.8 the growth with system size becomes more pronounced, as shown in panels (b) and (c), respectively. The saturation at large β\beta indicates that the finite-size ground-state regime has been reached for the sizes considered.

To further assess whether the low-temperature state is ordered, we perform a spin-wave finite-size scaling (FSS) analysis using

Ss(Q)L3=m(L)+AL,(32)\sqrt{\frac{S_s(\mathbf{Q})}{L^3}} = m(L) + \frac{A}{L}, \tag*{(32)}

where m(L)m(L) is the AFM order parameter and AA is a fitting parameter. Figure 8 (d) shows this scaling analysis at β=20\beta= 20, where SsS_s is already saturated for the sizes studied. The extrapolation gives a vanishing order parameter for U/t=6U/t = 6, whereas a finite intercept is obtained for U/t=6.2U/t = 6.2. Although these results should be interpreted with care, as the system size is still small, they indicate that the critical interaction in the NL case lies in the interval between U/t≈6.0U/t \approx6.0 and U/t≈6.2U/t \approx6.2.

This interval for the onset of AFM order is also visible in Figure 9 (a), which shows SsS_s as a function of U/tU/t for different system sizes at fixed inverse temperature βt=20\beta t = 20. Indeed, the growth of SsS_s with LL becomes appreciable around U/t≈6U/t \approx6, consistent with the finite-size extrapolation discussed above. A more accurate estimate is obtained from the correlation ratio. Figure 9 (b) shows RcR_c as a function of U/tU/t at βt=20\beta t = 20. The curves for different lattice sizes cross near a common point, yielding our best estimate Uc/t=6.1(1)U_c/t = 6.1(1) for the AFM transition in the NL case. This value is smaller than the critical coupling obtained for the NP system, consistent with the enhanced low-energy density of states associated with the nodal line. A finite-size collapse of RcR_c can also be performed, yielding an effective exponent close to ν~≈1/2\tilde{\nu} \approx1/2. Performing a scaling of Ss(Q)S_s(\mathbf{Q}), we obtain the magnetic critical exponent β~≈1/2\tilde{\beta} \approx1/2, as shown in the insets of Figure 9.

Figure 9. (a) Staggered spin structure factor SsS_s and (b) its correlation ratio RcR_c for the NL case at fixed inverse temperature βt=20\beta t = 20. The crossing of RcR_c near Uc/t≈6.1U_c/t \approx6.1 indicates the onset of AFM order. The insets describe their respective finite-size scaling analysis, providing ν~≈β~≈1/2\tilde{\nu} \approx\tilde{\beta} \approx1/2.

At this point, it is worth emphasizing that the mean-field analysis reveals that very large system sizes are required to recover the asymptotic scaling behavior associated with the nodal-line codimension dc=2d_c = 2. Consequently, our QMC results for the NL case are expected to exhibit much stronger finite-size effects than those for the NP case, and quantitative estimates of the critical exponents should be interpreted with caution. We return to this issue below.

Quantum Monte Carlo: unequal-time results

We now complement the equal-time magnetic observables with dynamical information obtained from the unequal-time Green’s function. While the spin structure factor and the correlation ratio diagnose the onset of AFM order, the single-particle spectral function allows us to follow how the low-energy fermionic excitations evolve across the transition.

In DQMC, the quantity measured directly is the imaginary-time Green’s function, Gαα′(k,τ)G_{\alpha\alpha'}(\mathbf{k},\tau), with α\alpha and α′\alpha' denoting orbital indices. Here, we consider the orbital-averaged Green’s function

G(k,τ)=14∑αGαα(k,τ).(33)G(\mathbf{k},\tau) = \frac{1}{4}\sum_{\alpha}G_{\alpha\alpha}(\mathbf{k},\tau). \tag*{(33)}

Obtaining A(k,ω)A(\mathbf{k},\omega) from G(k,τ)G(\mathbf{k},\tau) requires an analytic continuation from imaginary time to real frequency,

G(k,τ)=∫−∞∞dω K(τ,ω)A(k,ω),(34)G(\mathbf{k},\tau) = \int_{-\infty}^{\infty} d\omega\, K(\tau,\omega)A(\mathbf{k},\omega), \tag*{(34)}

with K(τ,ω)=e−τω1+e−βωK(\tau,\omega) = \frac{e^{-\tau\omega}}{1+e^{-\beta\omega}} being the fermionic kernel. This inversion is numerically obtained by a stochastic analytic continuation (SAC) procedure [43, 44]. For both cases, we used a 4×4×84 \times4 \times8 lattice, employing both PBC and APBC to access more wavevectors. [45]

Figure 10 shows the momentum-resolved spectral function A(k,ω)A(\mathbf{k},\omega) for the NP case, at fixed inverse temperature βt=8\beta t = 8. Panels 10 (a)–10 (d) correspond to U/t=5U/t = 5, 77, 88, and 99, respectively, with the spectra being shown along the line k=(π/2,π/2,kz)\mathbf{k} = (\pi/2,\pi/2,k_z). For U/t=5U/t = 5 and U/t=7U/t = 7, the spectral weight clearly displays a Dirac-like dispersion centered around kz=π/2k_z = \pi/2, indicating that the semimetallic quasiparticles remain well defined in this interaction range. As U/tU/t is increased to 88, the low-energy spectral weight near the nodal point is suppressed and a small gap begins to develop. Upon further increasing to U/t=9U/t = 9, the gap becomes much more pronounced, showing that the system has indeed entered an insulating regime. It is important to note that the opening of a small gap around U/t≈8U/t \approx8 is consistent with the critical interaction Uc/t≈7.8U_c/t \approx7.8 obtained from the equal-time magnetic correlations.

Figure 10. Momentum-resolved spectral function A(k,ω)A(\mathbf{k},\omega) for the nodal-point (cubic π\pi-flux) case at fixed inverse temperature βt=8\beta t = 8. The spectra are shown along the momentum line k=(π/2,π/2,kz)\mathbf{k} = (\pi/2,\pi/2,k_z) for (a) U/t=5U/t = 5, (b) U/t=7U/t = 7, (c) U/t=8U/t = 8, and (d) U/t=9U/t = 9. The cyan line in panel (a) represents a linear fit to the peak positions near the nodal point, yielding an estimate vF≈1.6tav_{\mathrm{F}} \approx1.6ta.

The spectral function also allows us to estimate the Fermi velocity in the semimetallic regime. To this end, we identify the

peak positions of A(k,ω)A(\mathbf{k},\omega) for momenta close to the nodal point (π/2,π/2,π/2)(\pi/2,\pi/2,\pi/2) and fit the resulting dispersion linearly. This procedure is illustrated in Figure 10 (a), yielding vF≈1.6tav_{\mathrm{F}} \approx1.6ta. This value is close to the noninteracting result, vF=2tav_{\mathrm{F}} = 2ta, indicating that the Hubbard interaction only weakly renormalizes the Fermi velocity before the onset of the insulating phase.

We now turn to the NL case, where a similar gap opening is observed around U/t≈6U/t \approx6. Figure 11 shows A(k,ω)A(\mathbf{k},\omega) along the nodal-line direction, k=(π/2,π/2,kz)\mathbf{k} = (\pi/2,\pi/2,k_z), for (a) U/t=5U/t = 5, (b) U/t=6U/t = 6, and (c) U/t=7U/t = 7, at fixed βt=8\beta t = 8. For U/t=5U/t = 5, the low-energy spectral weight exhibits no clear distortion or appreciable momentum dependence along the nodal-line direction, consistent with the persistence of a gapless nodal-line semimetal. Increasing the interaction strength to U/t=6U/t = 6 leads to the emergence of a small gap in the spectral function, consistent with the critical interaction Uc/t=6.1(1)U_c/t = 6.1(1) estimated from the crossing of the correlation ratio. For U/t=7U/t = 7, the gap becomes more pronounced, indicating the development of an insulating phase. Thus, the spectral analysis provides further evidence of the interaction-driven semimetal-to-insulator transition in the NL case, in agreement with the equal-time magnetic correlations.

Figure 11. Momentum-resolved spectral function A(k,ω)A(\mathbf{k},\omega) for the nodal-line case along k=(π/2,π/2,kz)\mathbf{k} = (\pi/2,\pi/2,k_z) for (a) U/t=5U/t = 5, (b) U/t=6U/t = 6, and (c) U/t=7U/t = 7. Here, we fix the inverse temperature at βt=8\beta t = 8 on a 4×4×84 \times4 \times8 lattice.

Before proceeding, it is worth examining the unequal-time Green’s functions directly, particularly at k=(π/2,π/2,0)\mathbf{k}=(\pi/2,\pi/2,0) and k=(π/2,π/2,π/2)\mathbf{k}=(\pi/2,\pi/2,\pi/2), as shown in Figure 12 for βt=8\beta t=8. This comparison allows us to investigate possible differences between these two momenta without relying on SAC. Interestingly, we observe a small but noticeable difference (beyond error bars) between the two Green’s functions, with the correlations at kz=0k_z=0 being more suppressed than those at kz=π/2k_z=\pi/2 at τ=β/2\tau=\beta/2. This suggests that interactions may introduce a weak momentum dependence along the nodal line, which is not resolved in the spectral functions obtained through SAC. Therefore, although our results indicate that the nodal-line structure remains largely robust against interactions, a definitive assessment of possible small interaction-induced distortions requires complementary methods with higher spectral resolution.

Figure 12. Unequal-time Green’s functions as a function of imaginary time τ\tau for k=(π/2,π/2,0)\mathbf{k}=(\pi/2,\pi/2,0) and k=(π/2,π/2,π/2)\mathbf{k}=(\pi/2,\pi/2,\pi/2) in the NL case, at fixed U/t=5U/t=5 and βt=8\beta t=8.

Weak-coupling renormalization-group analysis

We now perform a weak-coupling renormalization-group analysis of the interacting nodal-line semimetal phase. The details of the methodology are presented in Appendix D. Close to the nodal line, the noninteracting Hamiltonian takes the long-wavelength form

H^NL(0)≃−v(qxΓ1+qyΓ2)+κq⊥2cos⁡kzΓ3,(35)\hat{H}_{\mathrm{NL}}^{(0)} \simeq-v(q_x\Gamma_1+q_y\Gamma_2)+\kappa q_{\perp}^{2}\cos k_z\Gamma_3, \tag*{(35)}

where v=κ=2tv=\kappa=2t and q⊥2=qx2+qy2q_{\perp}^{2}=q_x^{2}+q_y^{2}. To leading order in momentum the system may therefore be viewed as a continuous stack of (2+1)(2+1)-dimensional Dirac cones labeled by kzk_z. As discussed above, this dimensional reduction of the low-energy dispersion leads to the same mean-field critical exponents as those of the two-dimensional honeycomb lattice. The Wilsonian transformation integrates a tubular shell in the transverse momenta, Λ/b<q⊥<Λ\Lambda/b<q_{\perp}<\Lambda, while leaving kzk_z unscaled. Here b>1b>1 is the coarse-graining factor, and

we introduce the logarithmic RG scale ℓ=ln⁡b\ell=\ln b, such that increasing ℓ\ell corresponds to flowing toward the infrared. Accordingly, the running transverse cutoff is Λ(ℓ)=Λ0e−ℓ\Lambda(\ell)=\Lambda_0e^{-\ell} and the associated energy scale is EΛ(ℓ)=vΛ(ℓ)E_\Lambda(\ell)=v\Lambda(\ell). Tree-level power counting then gives [U]=−1[U]=-1, and [Δ]=+1[\Delta]=+1, where the latter denotes the coefficient of the symmetry-allowed quadratic operator

δH^=Δcos⁡kzΓ3.(36)\delta\hat{H}=\Delta\cos k_z\Gamma_3. \tag*{(36)}

Thus, although the short-range Hubbard interaction is irrelevant at the Gaussian nodal-line fixed point, the quadratic deformation is a relevant perturbation.

At half filling, the first-order Hartree contribution vanishes in the particle-hole-symmetric formulation. The leading momentum-dependent correction therefore arises from the second-order self-energy. In compact notation, the corresponding diagram gives

Σσ(2)(k)=U2∫qΠσˉ(0)(q)Gσ(0)(k−q),(37)\Sigma_{\sigma}^{(2)}(k)=U^2\int_q\Pi_{\bar{\sigma}}^{(0)}(q)G_{\sigma}^{(0)}(k-q), \tag*{(37)}

where k=(k,iωn)k=(\mathbf{k},i\omega_n), ∫q≡T∑Ωm∫q\int_q\equiv T\sum_{\Omega_m}\int_{\mathbf{q}}, and

Πσ(0)(q)=−∫pGσˉ(0)(p)Gσˉ(0)(p+q)(38)\Pi_{\sigma}^{(0)}(q)=-\int_p G_{\bar{\sigma}}^{(0)}(p)G_{\bar{\sigma}}^{(0)}(p+q) \tag*{(38)}

is the bare particle-hole bubble of the opposite-spin fermions. The Green’s functions in Eqs. (37) and (38) retain the orbital matrix structure of the lattice Hamiltonian; the explicit evaluation of the loop integrals is given in Appendix D.

The contribution responsible for reconstructing the nodal manifold is obtained by projecting the static self-energy onto the Γ3\Gamma_3 channel. Along the original nodal line, k=(π/2,π/2,kz)\mathbf{k}=(\pi/2,\pi/2,k_z), the leading harmonic is

Σ3(2)(kz,0)=ΔUcos⁡kz+⋯ ,(39)\Sigma_3^{(2)}(k_z,0)=\Delta_U\cos k_z+\cdots, \tag*{(39)}

with ΔU=c3U2t\Delta_U=c_3\frac{U^2}{t}, and c3≈2.91×10−3c_3\approx2.91\times10^{-3}. Thus, although the onsite Hubbard interaction is irrelevant at the Gaussian nodal-line fixed point, it generates at O(U2)O(U^2) the symmetry-allowed single-particle perturbation ΔUcos⁡kzΓ3\Delta_U\cos k_z\Gamma_3, which is relevant under the low-energy scaling. Importantly, this reconstruction occurs without accompanying weak-coupling instability toward a symmetry-broken many-body state: the Hubbard interaction itself remains perturbatively irrelevant, while its second-order self-energy reshapes the low-energy band structure.

Introducing the dimensionless Hubbard coupling uu and δ=Δ/EΛ\delta=\Delta/E_\Lambda, with EΛ=vΛE_\Lambda=v\Lambda, the corresponding weak-coupling RG equations are

dudℓ=−u+BUu2+⋯ ,(40)\frac{du}{d\ell}=-u+B_Uu^2+\cdots, \tag*{(40)}
dδdℓ=δ+CΔu2+⋯ .(41)\frac{d\delta}{d\ell}=\delta+C_\Delta u^2+\cdots. \tag*{(41)}

The first equation expresses the perturbative irrelevance of the short-range interaction, whereas the CΔu2C_\Delta u^2 term in the second equation is precisely the source term associated with the interaction-generated self-energy in Eq. (39).

For u0≪1u_0 \ll1, the nonlinear correction to the interaction flow is subleading, so that

u(ℓ)=u0e−ℓ.(42)u(\ell) = u_0e^{-\ell}. \tag*{(42)}

Starting from the microscopic nodal-line Hamiltonian, for which δ0=0\delta_0 = 0, one then obtains

δ(ℓ)=CΔu023(eℓ−e−2ℓ).(43)\delta(\ell) = \frac{C_{\Delta}u_0^2}{3}\left(e^\ell-e^{-2\ell}\right). \tag*{(43)}

The corresponding weak-coupling trajectories in the (u,δ)(u,\delta) plane are illustrated in Figure 13, for fixed q0=0.35q_0 = 0.35 and v=2tv = 2t. Hence the interaction itself flows to zero, while the relevant single-particle operator generated by it grows under coarse graining. Since the running cutoff scales as EΛ∝e−ℓE_{\Lambda} \propto e^{-\ell}, the corresponding physical amplitude remains finite in the infrared. Matching the continuum flow to the lattice self-energy gives Δphys=c3U2/t\Delta_{\mathrm{phys} = c_3U^2/t}.

Weak-coupling RG flow of the dimensionless Hubbard interaction

Figure 13. Weak-coupling RG flow of the dimensionless Hubbard interaction uu and the quadratic perturbation δ=Δ/EΛ\delta= \Delta/E_{\Lambda}. The Gaussian fixed point at (u,δ)=(0,0)(u,\delta) = (0,0) and trajectories starting from the microscopic condition δ0=0\delta_0 = 0 are shown.

This perturbation qualitatively changes the topology of the zero-energy manifold. Along the original nodal line,

H^eff(π2,π2,kz)=Δphyscos⁡kzΓ3,(44)\hat{H}_{\mathrm{eff}}\left(\frac{\pi}{2},\frac{\pi}{2},k_z\right) = \Delta_{\mathrm{phys}}\cos k_z\Gamma_3, \tag*{(44)}

so that zero energy survives only for kz=±π2k_z = \pm\frac{\pi}{2}. The weakly interacting infrared theory is consequently not a nodal-line semimetal but a strongly anisotropic nodal-point semimetal with longitudinal velocity vz≈c3U2/t≪2tv_z \approx c_3U^2/t \ll2t. However, the corresponding induced gap is small, which explains why a broad intermediate-energy regime may appear nodal-line-like in finite-size simulations.

Berry phase and Wilson-loop evolution

The interaction-driven reconstruction of the nodal manifold can also be characterized geometrically. In the noninteracting limit, a closed loop in the (qx,qy)(q_x,q_y) plane that links the nodal line encloses a Dirac singularity and carries a Berry phase π\pi modulo 2π2\pi, as expected for a band-contact line [46, 47]. In the present four-band model, however, the occupied orbital subspace is twofold degenerate. The corresponding gauge-invariant quantity is therefore the Wilson loop, whose eigenvalues may be written as eiγWe^{i\gamma_W}. In the decoupled Dirac sectors, the Wilson-loop eigenphase coincides with the corresponding Berry phase modulo 2π2\pi. Hence, in the noninteracting nodal-line limit, both occupied sectors have eigenvalue −1-1, corresponding to γW=π\gamma_W = \pi along the entire nodal line.

Once the interaction-generated term Δphyscos⁡kzΓ3\Delta_{\mathrm{phys}}\cos k_z\Gamma_3 is included, a generic fixed-kzk_z slice acquires an effective Dirac mass m(kz)=Δphyscos⁡kzm(k_z) = \Delta_{\mathrm{phys}}\cos k_z. As outlined in Appendix § 1.2.5, for a circular transverse loop of radius q0q_0, the principal Wilson-loop phase can be written as

γW(kz;q0)=π[1−∣Δphyscos⁡kz∣v2q02+Δphys2cos⁡2kz].(45)\gamma_W(k_z;q_0) = \pi\left[1 - \frac{\left|\Delta_{\mathrm{phys}}\cos k_z\right|}{\sqrt{v^2q_0^2+\Delta_{\mathrm{phys}}^2\cos^2 k_z}}\right]. \tag*{(45)}

This expression is restricted to the principal branch 0≤γW≤π0 \le\gamma_W \le\pi. In the noninteracting limit, Δphys→0\Delta_{\mathrm{phys}} \to0, one recovers γW=π\gamma_W = \pi for every kzk_z. For finite interaction, the phase departs from π\pi on the gapped fixed-kzk_z slices, whereas it remains pinned to γW=π\gamma_W = \pi at kz=±π/2k_z = \pm\pi/2, where the interaction-generated mass vanishes. In the limit of a shrinking transverse loop, q0→0q_0 \to0, one furthermore obtains γW→0\gamma_W \to0 for every gapped slice, while the value π\pi survives at the two nodal points.

Figure 14 illustrates this evolution for representative interaction strengths, using Δphys=c3U2/t\Delta_{\mathrm{phys}} = c_3U^2/t. The noninteracting system exhibits a uniform π\pi phase along the complete nodal line. Increasing UU progressively suppresses the Wilson-loop phase away from kz=±π/2k_z = \pm\pi/2, while leaving the two surviving nodal points pinned at π\pi. The evolution of γW\gamma_W therefore provides a direct geometrical signature of the interaction-driven reconstruction of the nodal line into isolated nodal points.

Figure 14. Principal Wilson-loop phase γW\gamma_W evaluated along the original nodal line as a function of kzk_z for representative interaction strengths. The interaction-induced scale is parametrized as Δphys=c3U2/t\Delta_{\mathrm{phys}} = c_3U^2/t.

Conclusions

In this work, we investigated the effects of local electronic interactions in three-dimensional semimetals with Dirac nodal-point and Dirac nodal-line Fermi surfaces. By considering two closely related lattice models, we directly examined how the dimensionality of the nodal manifold affects the stability of the semimetallic phase and the interaction-driven transition toward antiferromagnetic order. To this end, we combined mean-field calculations, determinant quantum Monte Carlo simulations, and a weak-coupling renormalization-group analysis.

Already at the noninteracting and mean-field levels, the two systems display important differences. While the density of states vanishes quadratically in energy for the nodal-point case, it vanishes linearly for the nodal-line system. As a consequence, the latter is more susceptible to electronic interactions and it is expected to develop antiferromagnetic order at a smaller critical coupling. Within mean-field theory, we obtain Uc,NP/t≈4.39U_{\mathrm{c,NP}}/t \approx4.39 and Uc,NL/t≈3.54U_{\mathrm{c,NL}}/t \approx3.54, for the NP and NL cases, respectively. In addition, the mean-field order parameter displays distinct critical behavior in the two cases. The nodal-point system exhibits the conventional mean-field behavior β~=1/2\tilde{\beta}=1/2, whereas the nodal-line system approaches β~=1\tilde{\beta}=1 in the thermodynamic limit. These features (the codimension-two low-energy scaling, including the linear density of states and the associated mean-field critical behavior) are expected to apply broadly to smooth nodal lines, in particular those with linear transverse dispersion.

The unbiased DQMC simulations confirm that both systems remain stable against weak-coupling antiferromagnetic ordering and undergo a transition to an antiferromagnetic insulating state only above a finite interaction strength. For the nodal-point model, our finite-size analysis gives Uc/t=7.8(1)U_{\mathrm{c}}/t=7.8(1), and the scaling of the correlation ratio and spin structure factor is compatible with ν~=1/2\tilde{\nu}=1/2 and β~=1/2\tilde{\beta}=1/2. Therefore, we provide an independent verification that the quantum phase transition in the cubic π\pi-flux Hubbard model is compatible with the (3+1)-dimensional Gross–Neveu mean-field scaling [14]. The single-particle spectral function further shows that the Dirac-like quasiparticles survive throughout the semimetallic regime, with only a moderate renormalization of the Fermi velocity, while a gap develops upon entering the antiferromagnetic phase.

For the nodal-line system, the transition occurs at the smaller interaction strength Uc/t=6.1(1)U_{\mathrm{c}}/t=6.1(1). This reduction of the critical coupling is consistent with the enhanced low-energy density of states associated with the extended nodal manifold. The thermodynamic and magnetic observables provide a consistent picture of the transition, with the low-temperature structure in the specific heat accompanying the development of long-range spin correlations. We also show that the finite-size scaling of the correlation ratio and spin structure factor is compatible with ν~=1/2\tilde{\nu}=1/2 and β~=1/2\tilde{\beta}=1/2, as in the NP case.

However, our QMC results are not sufficient to establish the robustness of the nodal line against interactions. Although we observe small differences in the unequal-time Green’s functions at different momenta along the nodal line, these differences cannot be reliably resolved in the spectral functions obtained through analytic continuation. Therefore, possible interaction-induced renormalizations of the low-energy band structure may remain below the momentum and energy resolution accessible in the present simulations.

We address this question by employing a renormalization-group analysis in the weak-coupling limit. Our main result is that the nodal lines are unstable against interactions, ultimately leaving isolated nodal points at kz=±π/2k_z=\pm\pi/2 along each original nodal line. However, the associated renormalization of the band structure remains weak throughout the semimetallic regime. In particular, the interaction-induced gap at k=(π/2,π/2,0)\mathbf{k}=(\pi/2,\pi/2,0) remains smaller than 0.1t0.1t, placing it at the limit of the energy resolution accessible in our QMC simulations.

Although small in energy, this reconstruction changes the dimensionality of the band-touching manifold, from nodal lines to isolated nodal points, with potentially important consequences for the critical behavior. In particular, the resulting low-energy spectrum may favor conventional three-dimensional mean-field criticality, providing a possible explanation for the exponents ν~≈1/2\tilde{\nu}\approx1/2 and β~≈1/2\tilde{\beta}\approx1/2 obtained in our QMC simulations for the NL case. Nevertheless, given the strong finite-size effects discussed above, these numerical results should be treated as rough estimates.

In summary, our results suggest that, while the codimension of the nodal manifold controls the leading magnetic critical behavior (both the critical point and the critical exponents), the symmetries of the microscopic Hamiltonian determine whether the nodal manifold itself remains stable against interactions. That is, we have highlighted the important role played by the geometry and codimension of the Fermi surface in interaction-driven quantum phase transitions. We also emphasize that the instability of the nodal manifold is not a generic property of all symmetry-protected nodal-line semimetals, but rather of nodal-line realizations for which such a perturbation is permitted. If an additional crystalline symmetry forbids this bilinear, the nodal line may remain stable throughout the symmetry-preserving weak-coupling regime.

Acknowledgments

The authors are grateful to H. Guo, Z.-Y. Meng, and R. G. Pereira for many insightful discussions on the subject. A.R.M.S., L.O.L., M.B.S.N. and N.C.C. thank the Brazilian Agencies Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), and Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro, FAPERJ. A.R.M.S. thanks the International Institute of Physics of Federal University of Rio Grande do Norte (IIP-UFRN) and Fundação de Amparo à Pesquisa do Estado do Rio Grande do Norte (Grant No. 1699/24 IIF-FINEP). M.B.S.N. acknowledges INCT – Advanced Quantum Materials, involving CNPq (Proc. 408766/2024-7), FAPESP (Proc. 2025/27091-3), and CAPES. N.C.C. acknowledges support from FAPERJ Grant No. E-26/210.592/2025, CNPq Grant No. 308130/2025-1, Serrapilheira Institute Grant No. R2502-52037, and the Alexander von Humboldt Foundation through a Humboldt Research Fellowship for Experienced Researchers. F.F.A. acknowledges financial support by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) through the Würzburg-Dresden Cluster of Excellence ctd.qmat – Complexity, Topology and Dynamics in Quantum Matter (EXC 2147, Project No. 390858490). The authors gratefully acknowledge the Gauss Centre for Supercomputing e.V. (www.gauss-centre.eu) for funding this project by providing computing time on the GCS Supercomputer SuperMUC-NG at Leibniz Supercomputing Centre (www.lrz.de).

Real-space Hamiltonian

A real-space representation is obtained by considering a cubic Bravais lattice with primitive vectors a1=a(2,0,0)\mathbf{a}_1=a(2,0,0), a2=a(0,2,0)\mathbf{a}_2=a(0,2,0), a3=a(0,0,1)\mathbf{a}_3=a(0,0,1), where aa is the lattice constant. The four orbitals inside the unit cell are located at

RA=R,RB=R+a12+a22,RC=R+a12,RD=R+a22.(46)\begin{aligned} \mathbf{R}_A &= \mathbf{R},\\ \mathbf{R}_B &= \mathbf{R}+\frac{\mathbf{a}_1}{2}+\frac{\mathbf{a}_2}{2},\\ \mathbf{R}_C &= \mathbf{R}+\frac{\mathbf{a}_1}{2},\\ \mathbf{R}_D &= \mathbf{R}+\frac{\mathbf{a}_2}{2}. \tag*{(46)} \end{aligned}

as shown in Figure 2. By performing an inverse Fourier transform [48], one finds the Hamiltonian for the nodal-point case as

H0(NP)=t∑R[cA,RA†cC,RC−cA,RA†cD,RD−cA,RA†cA,RA+a3+cC,RC†cA,RA+a1+cC,RC†cB,RB+cC,RC†cC,RC+a3+cD,RD†cB,RB+cD,RD†cD,RD+a3−cD,RD†cA,RA+a2+cB,RB†cD,RD+a1+cB,RB†cC,RC+a2−cB,RB†cB,RB+a3+H.c.].(47)\begin{aligned} \mathcal{H}_0(\mathrm{NP})=t\sum_{\mathbf{R}}\big[& c^\dagger_{A,\mathbf{R}_A}c_{C,\mathbf{R}_C} -c^\dagger_{A,\mathbf{R}_A}c_{D,\mathbf{R}_D} -c^\dagger_{A,\mathbf{R}_A}c_{A,\mathbf{R}_A+\mathbf{a}_3}\\ &+c^\dagger_{C,\mathbf{R}_C}c_{A,\mathbf{R}_A+\mathbf{a}_1} +c^\dagger_{C,\mathbf{R}_C}c_{B,\mathbf{R}_B} +c^\dagger_{C,\mathbf{R}_C}c_{C,\mathbf{R}_C+\mathbf{a}_3}\\ &+c^\dagger_{D,\mathbf{R}_D}c_{B,\mathbf{R}_B} +c^\dagger_{D,\mathbf{R}_D}c_{D,\mathbf{R}_D+\mathbf{a}_3} -c^\dagger_{D,\mathbf{R}_D}c_{A,\mathbf{R}_A+\mathbf{a}_2}\\ &+c^\dagger_{B,\mathbf{R}_B}c_{D,\mathbf{R}_D+\mathbf{a}_1} +c^\dagger_{B,\mathbf{R}_B}c_{C,\mathbf{R}_C+\mathbf{a}_2} -c^\dagger_{B,\mathbf{R}_B}c_{B,\mathbf{R}_B+\mathbf{a}_3} +\mathrm{H.c.}\big]. \tag*{(47)} \end{aligned}

Here, cα,Rα†c^\dagger_{\alpha,\mathbf{R}_\alpha} and cα,Rαc_{\alpha,\mathbf{R}_\alpha} are, respectively, the creation and annihilation operators for an electron in orbital α\alpha at position Rα\mathbf{R}_\alpha. The spin index has been omitted for simplicity. The symbol H.c. denotes the Hermitian conjugate of the preceding hopping terms. Notice that H0(NP)\mathcal{H}_0(\mathrm{NP}) corresponds to the three-dimensional π\pi-flux Hamiltonian, whose hopping pattern is shown in Figure 2.

The difference between the nodal-line and nodal-point limits is encoded in additional hopping processes. These terms are collected in H′\mathcal{H}', so that the nodal-line Hamiltonian can be written as H0(NL)=H0(NP)+H′\mathcal{H}_0(\mathrm{NL})=\mathcal{H}_0(\mathrm{NP})+\mathcal{H}', with

Noninteracting density of states

The four-band Hamiltonian in Eq. (3) has the spectrum

E±(k)=±2tcos⁡2kx+cos⁡2ky+χ2(k),(48)E_{\pm}(\mathbf{k})=\pm2t\sqrt{\cos^2 k_x+\cos^2 k_y+\chi^2(\mathbf{k})}, \tag*{(48)}

where each branch is twofold degenerate in orbital space. Including spin, each branch therefore has a total degeneracy of four.

For the nodal-point case, χNP(k)=cos⁡kz\chi_{\mathrm{NP}}(\mathbf{k})=\cos k_z, and the reduced Brillouin zone contains two inequivalent nodal points at kz=±π/2k_z=\pm\pi/2. Expanding around either point gives an isotropic linear dispersion,

E±(q)≈±v∣q∣,(49)E_{\pm}(\mathbf{q})\approx\pm v|\mathbf{q}|, \tag*{(49)}

with v=2tv=2t. The corresponding low-energy density of states is

ρNP(E)=gNP∫d3q(2π)3δ(∣E∣−v∣q∣)=gNPE22π2v3,(50)\rho_{\mathrm{NP}}(E)=g_{\mathrm{NP}}\int\frac{d^3q}{(2\pi)^3}\delta(|E|-v|\mathbf{q}|)=g_{\mathrm{NP}}\frac{E^2}{2\pi^2v^3}, \tag*{(50)}

with gNP=2×2×2=8g_{\mathrm{NP}}=2\times2\times2=8, accounting for orbital degeneracy, spin, and the two nodal points. Hence, ρNP(E)=E22π2t3\rho_{\mathrm{NP}}(E)=\frac{E^2}{2\pi^2t^3}, as presented in Eq. (21).

For the nodal-line case, χNL(k)=cos⁡kz(cos⁡2kx+cos⁡2ky)\chi_{\mathrm{NL}}(\mathbf{k})=\cos k_z(\cos^2 k_x+\cos^2 k_y). Expanding transverse to the nodal line gives

E±(q⊥,kz)=±2tq⊥2+q⊥4cos⁡2kz≈±2tq⊥.(51)E_{\pm}(\mathbf{q}_{\perp},k_z)=\pm2t\sqrt{q_{\perp}^2+q_{\perp}^4\cos^2 k_z}\approx\pm2tq_{\perp}. \tag*{(51)}

Since the nodal line extends over the full kzk_z direction, the low-energy density of states becomes

ρNL(E)=gNL∫−ππdkz2π∫d2q⊥(2π)2δ(∣E∣−2tq⊥),(52)\rho_{\mathrm{NL}}(E)=g_{\mathrm{NL}}\int_{-\pi}^{\pi}\frac{dk_z}{2\pi}\int\frac{d^2q_{\perp}}{(2\pi)^2}\delta(|E|-2tq_{\perp}), \tag*{(52)}

where gNL=2×2=4g_{\mathrm{NL}}=2\times2=4 accounts for orbital degeneracy and spin. Performing the integrals gives ρNL(E)=∣E∣2πt2\rho_{\mathrm{NL}}(E)=\frac{|E|}{2\pi t^2}, as presented in Eq. (22).

Mean-field approximation

We perform a static Hartree decoupling in the Hubbard term considering

nRα↑nRα↓≈⟨nRα↑⟩nRα↓+nRα↑⟨nRα↓⟩−⟨nRα↑⟩⟨nRα↓⟩.(53)n_{\mathbf{R}\alpha\uparrow}n_{\mathbf{R}\alpha\downarrow}\approx\langle n_{\mathbf{R}\alpha\uparrow}\rangle n_{\mathbf{R}\alpha\downarrow}+n_{\mathbf{R}\alpha\uparrow}\langle n_{\mathbf{R}\alpha\downarrow}\rangle-\langle n_{\mathbf{R}\alpha\uparrow}\rangle\langle n_{\mathbf{R}\alpha\downarrow}\rangle. \tag*{(53)}

ends_mid=1 with α=A,B,C,D\alpha=A,B,C,D labeling the four sites within the unit cell. For an arbitrary collinear magnetic state, we parametrize the local densities as

⟨nRα,σ⟩=n2+σmRα2,(54)\langle n_{R\alpha,\sigma}\rangle=\frac{n}{2}+\sigma\frac{m_{R\alpha}}{2}, \tag*{(54)}

where σ=+1\sigma=+1 for spin up and σ=−1\sigma=-1 for spin down, while nn and mm define the local average density and the local average magnetization, respectively. At half filling, the spin-independent Hartree contribution can be absorbed into the chemical potential. Then, we assume a staggered antiferromagnetic pattern characterized by the wavevector Q=(0,0,π)\mathbf{Q}=(0,0,\pi), and the local magnetization

mRA=mRB=−mRC=−mRD=mcos⁡(Q⋅R).(55)m_{R_A}=m_{R_B}=-m_{R_C}=-m_{R_D}=m\cos(\mathbf{Q}\cdot\mathbf{R}). \tag*{(55)}

By performing a Fourier transform, as the magnetic order carries momentum Q\mathbf{Q}, the mean-field term couples states at k\mathbf{k} and k+Q\mathbf{k}+\mathbf{Q}. Therefore, one should work with an enlarged spinor representation

Φkσ=(Ψkσ,Ψk+Q,σ)T.(56)\Phi_{\mathbf{k}\sigma}=(\Psi_{\mathbf{k}\sigma},\Psi_{\mathbf{k}+\mathbf{Q},\sigma})^{T}. \tag*{(56)}

The corresponding mean-field Hamiltonian can be written in block form as

H^MF,σ(k)=(H^0(k)−σUm2Λ^−σUm2Λ^H^0(k+Q)),(57)\hat{H}_{\mathrm{MF},\sigma}(\mathbf{k})= \begin{pmatrix} \hat{H}_0(\mathbf{k}) & -\sigma\frac{Um}{2}\hat{\Lambda}\\ -\sigma\frac{Um}{2}\hat{\Lambda} & \hat{H}_0(\mathbf{k}+\mathbf{Q}) \end{pmatrix}, \tag*{(57)}

in which H^0(k)\hat{H}_0(\mathbf{k}) is the noninteracting Hamiltonian defined in Eq. (3), and Λ^=diag⁡(1,1,−1,−1)\hat{\Lambda}=\operatorname{diag}(1,1,-1,-1). The overall sign of the off-diagonal term can be changed by a redefinition of the staggered order parameter and does not affect the spectrum. Defining Δ=Um4t\Delta=\frac{Um}{4t}, and diagonalizing the resulting 8×88\times8 matrix, one obtains the quasiparticle energies of Eq. (14).

The mean-field free energy is then obtained by summing over the quasiparticle levels and adding the constant contribution generated by the Hartree decoupling, which leads to Eq. (15). The order parameter is obtained by imposing the stationary condition ∂F/∂Δ=0\partial F/\partial\Delta=0. At finite temperature, the nontrivial solution satisfies

tU=14N∑ktanh⁡[βtcos⁡2kx+cos⁡2ky+χ2(k)+Δ2]cos⁡2kx+cos⁡2ky+χ2(k)+Δ2.(58)\frac{t}{U}=\frac{1}{4N}\sum_{\mathbf{k}}\frac{\tanh\left[\beta t\sqrt{\cos^{2}k_x+\cos^{2}k_y+\chi^{2}(\mathbf{k})+\Delta^{2}}\right]}{\sqrt{\cos^{2}k_x+\cos^{2}k_y+\chi^{2}(\mathbf{k})+\Delta^{2}}}. \tag*{(58)}

In the zero-temperature limit, the hyperbolic tangent approaches unity, leading to

tU=14N∑k1cos⁡2kx+cos⁡2ky+χ2(k)+Δ2.(59)\frac{t}{U}=\frac{1}{4N}\sum_{\mathbf{k}}\frac{1}{\sqrt{\cos^{2}k_x+\cos^{2}k_y+\chi^{2}(\mathbf{k})+\Delta^{2}}}. \tag*{(59)}

Weak-coupling RG and Berry phase

We provide here the details of the weak-coupling RG analysis summarized in Sec. III D. The main point is that the onsite Hubbard interaction is irrelevant at the Gaussian nodal-line fixed point, but generates at second order a symmetry-allowed relevant quadratic perturbation. Although parametrically small at weak coupling, this term changes the asymptotic low-energy structure from a nodal line to isolated nodal points.

Using Γ1=τx⊗μ0\Gamma_1=\tau_x\otimes\mu_0, Γ2=τy⊗μy\Gamma_2=\tau_y\otimes\mu_y, and Γ3=τz⊗μ0\Gamma_3=\tau_z\otimes\mu_0, the lattice nodal-line Hamiltonian can be written as

H^0(k)=2t[cos⁡kxΓ1+cos⁡kyΓ2+cos⁡kz(cos⁡2kx+cos⁡2ky)Γ3].(60)\hat{H}_0(\mathbf{k})=2t\left[\cos k_x\Gamma_1+\cos k_y\Gamma_2+\cos k_z\left(\cos^{2}k_x+\cos^{2}k_y\right)\Gamma_3\right]. \tag*{(60)}

Expanding around the nodal line, kNL=(π/2,π/2,kz)\mathbf{k}_{\mathrm{NL}}=(\pi/2,\pi/2,k_z), with kx=π/2+qxk_x=\pi/2+q_x and ky=π/2+qyk_y=\pi/2+q_y, gives

H^0≃−ν(qxΓ1+qyΓ2)+κq⊥2cos⁡kzΓ3,(61)\hat{H}_0\simeq-\nu(q_x\Gamma_1+q_y\Gamma_2)+\kappa q_{\perp}^{2}\cos k_z\Gamma_3, \tag*{(61)}

where ν=κ=2t\nu=\kappa=2t and q⊥2=qx2+qy2q_{\perp}^{2}=q_x^{2}+q_y^{2}. To leading order in q⊥q_\perp, the system may therefore be viewed as a continuous family of (2+1)(2+1)-dimensional Dirac cones parametrized by kzk_z.

The Wilsonian RG transformation proceeds by integrating out modes in the transverse momentum shell Λ/b<q⊥<Λ\Lambda/b<q_\perp<\Lambda, with b>1b>1, while the full kzk_z dependence is retained. We parametrize the coarse graining by the logarithmic running scale

ℓ=ln⁡b,Λ(ℓ)=Λ0e−ℓ,(62)\ell=\ln b,\qquad\Lambda(\ell)=\Lambda_0e^{-\ell}, \tag*{(62)}

so that increasing ℓ\ell corresponds to progressively lower energies and momenta. Equivalently, along the RG flow, ω(ℓ)=ω0e−ℓ\omega(\ell)=\omega_0e^{-\ell} and q⊥(ℓ)=q⊥,0e−ℓq_\perp(\ell)=q_{\perp,0}e^{-\ell}, whereas kzk_z is not rescaled. The corresponding running energy cutoff is EΛ(ℓ)=νΛ(ℓ)=E0e−ℓE_\Lambda(\ell)=\nu\Lambda(\ell)=E_0e^{-\ell}.

With this anisotropic scaling, the fermionic field has scaling dimension [ψ]=1[\psi]=1. Consequently, the onsite four-fermion coupling has [U]=−1[U]=-1, whereas the coefficient Δ\Delta of the symmetry-allowed operator Δcos⁡kzΓ3\Delta\cos k_z\Gamma_3 has [Δ]=+1[\Delta]=+1. The bare curvature coefficient is also irrelevant, with [κ]=−1[\kappa]=-1. Thus the Hubbard interaction is perturbatively irrelevant at the Gaussian nodal-line fixed point, while a momentum-independent Γ3\Gamma_3 term modulated by cos⁡kz\cos k_z is a relevant perturbation.

The Hubbard interaction is conveniently written in particle-hole symmetric form as

HU=U∑R,a(nRa↑−12)(nRa↓−12),(63)H_U=U\sum_{\mathbf{R},a}\left(n_{\mathbf{R}a\uparrow}-\frac{1}{2}\right)\left(n_{\mathbf{R}a\downarrow}-\frac{1}{2}\right), \tag*{(63)}

where a=A,B,C,Da=A,B,C,D labels the orbitals within the unit cell. At half filling the first-order Hartree contribution vanishes. The leading momentum-dependent correction is consequently the second-order self-energy represented by Figure 15.

Figure 15. Second-order Hubbard self-energy responsible for the leading momentum-dependent correction in the weak-coupling regime. The first-order Hartree contribution vanishes at half filling in the particle-hole-symmetric formulation. The O(U2)O(U^2) diagram generates, upon projection onto the Γ3\Gamma_3 channel, the cos⁡kzΓ3\cos k_z\Gamma_3 perturbation discussed in the text.

We denote fermionic and bosonic four-momenta by k=(k,iωn)k=(\mathbf{k},i\omega_n) and q=(q,iΩm)q=(\mathbf{q},i\Omega_m), respectively, and use the notation Jk=(T/N)∑k,ΩnJ_k=(T/N)\sum_{k,\Omega_n}. The bare propagator is the 4×44\times4 orbital matrix

G(0)(k)=[iωn−H^0(k)]−1.(64)G^{(0)}(k)=\left[i\omega_n-\hat{H}_0(\mathbf{k})\right]^{-1}. \tag*{(64)}

For the local Hubbard vertex, the opposite-spin particle-hole bubble connecting orbitals aa and bb is

Πab,σˉ(0)(q)=−∫pGab,σˉ(0)(p+q)Gba,σˉ(0)(p),(65)\Pi^{(0)}_{ab,\bar{\sigma}}(q)=-\int_p G^{(0)}_{ab,\bar{\sigma}}(p+q)G^{(0)}_{ba,\bar{\sigma}}(p), \tag*{(65)}

@page 14/17 starts_mid=1 ends_mid=1 with α=A,B,C,D\alpha=A,B,C,D labeling the four sites within the unit cell. For an arbitrary collinear magnetic state, we parametrize the local densities as

⟨nRα,σ⟩=n2+σmRα2,(66)\langle n_{R\alpha,\sigma}\rangle=\frac{n}{2}+\sigma\frac{m_{R\alpha}}{2}, \tag*{(66)}

where σ=+1\sigma=+1 for spin up and σ=−1\sigma=-1 for spin down, while nn and mm define the local average density and the local average magnetization, respectively. At half filling, the spin-independent Hartree contribution can be absorbed into the chemical potential. Then, we assume a staggered antiferromagnetic pattern characterized by the wavevector Q=(0,0,π)\mathbf{Q}=(0,0,\pi), and the local magnetization

mRA=mRB=−mRC=−mRD=mcos⁡(Q⋅R).(67)m_{R_A}=m_{R_B}=-m_{R_C}=-m_{R_D}=m\cos(\mathbf{Q}\cdot\mathbf{R}). \tag*{(67)}

By performing a Fourier transform, as the magnetic order carries momentum Q\mathbf{Q}, the mean-field term couples states at k\mathbf{k} and k+Q\mathbf{k}+\mathbf{Q}. Therefore, one should work with an enlarged spinor representation

Φkσ=(Ψkσ,Ψk+Q,σ)T.(68)\Phi_{\mathbf{k}\sigma}=(\Psi_{\mathbf{k}\sigma},\Psi_{\mathbf{k}+\mathbf{Q},\sigma})^{T}. \tag*{(68)}

The corresponding mean-field Hamiltonian can be written in block form as

H^MF,σ(k)=(H^0(k)−σUm2Λ^−σUm2Λ^H^0(k+Q)),(69)\hat{H}_{\mathrm{MF},\sigma}(\mathbf{k})= \begin{pmatrix} \hat{H}_0(\mathbf{k}) & -\sigma\frac{Um}{2}\hat{\Lambda}\\ -\sigma\frac{Um}{2}\hat{\Lambda} & \hat{H}_0(\mathbf{k}+\mathbf{Q}) \end{pmatrix}, \tag*{(69)}

in which H^0(k)\hat{H}_0(\mathbf{k}) is the noninteracting Hamiltonian defined in Eq. (3), and Λ^=diag⁡(1,1,−1,−1)\hat{\Lambda}=\operatorname{diag}(1,1,-1,-1). The overall sign of the off-diagonal term can be changed by a redefinition of the staggered order parameter and does not affect the spectrum. Defining Δ=Um4t\Delta=\frac{Um}{4t}, and diagonalizing the resulting 8×88\times8 matrix, one obtains the quasiparticle energies of Eq. (14).

The mean-field free energy is then obtained by summing over the quasiparticle levels and adding the constant contribution generated by the Hartree decoupling, which leads to Eq. (15). The order parameter is obtained by imposing the stationary condition ∂F/∂Δ=0\partial F/\partial\Delta=0. At finite temperature, the nontrivial solution satisfies

tU=14N∑ktanh⁡[βtcos⁡2kx+cos⁡2ky+χ2(k)+Δ2]cos⁡2kx+cos⁡2ky+χ2(k)+Δ2.(70)\frac{t}{U}=\frac{1}{4N}\sum_{\mathbf{k}}\frac{\tanh\left[\beta t\sqrt{\cos^{2}k_x+\cos^{2}k_y+\chi^{2}(\mathbf{k})+\Delta^{2}}\right]}{\sqrt{\cos^{2}k_x+\cos^{2}k_y+\chi^{2}(\mathbf{k})+\Delta^{2}}}. \tag*{(70)}

In the zero-temperature limit, the hyperbolic tangent approaches unity, leading to

tU=14N∑k1cos⁡2kx+cos⁡2ky+χ2(k)+Δ2.(71)\frac{t}{U}=\frac{1}{4N}\sum_{\mathbf{k}}\frac{1}{\sqrt{\cos^{2}k_x+\cos^{2}k_y+\chi^{2}(\mathbf{k})+\Delta^{2}}}. \tag*{(71)}

Weak-coupling RG and Berry phase

We provide here the details of the weak-coupling RG analysis summarized in Sec. III D. The main point is that the onsite Hubbard interaction is irrelevant at the Gaussian nodal-line fixed point, but generates at second order a symmetry-allowed relevant quadratic perturbation. Although parametrically small at weak coupling, this term changes the asymptotic low-energy structure from a nodal line to isolated nodal points.

Using Γ1=τx⊗μ0\Gamma_1=\tau_x\otimes\mu_0, Γ2=τy⊗μy\Gamma_2=\tau_y\otimes\mu_y, and Γ3=τz⊗μ0\Gamma_3=\tau_z\otimes\mu_0, the lattice nodal-line Hamiltonian can be written as

H^0(k)=2t[cos⁡kxΓ1+cos⁡kyΓ2+cos⁡kz(cos⁡2kx+cos⁡2ky)Γ3].(72)\hat{H}_0(\mathbf{k})=2t\left[\cos k_x\Gamma_1+\cos k_y\Gamma_2+\cos k_z\left(\cos^{2}k_x+\cos^{2}k_y\right)\Gamma_3\right]. \tag*{(72)}

Expanding around the nodal line, kNL=(π/2,π/2,kz)\mathbf{k}_{\mathrm{NL}}=(\pi/2,\pi/2,k_z), with kx=π/2+qxk_x=\pi/2+q_x and ky=π/2+qyk_y=\pi/2+q_y, gives

H^0≃−ν(qxΓ1+qyΓ2)+κq⊥2cos⁡kzΓ3,(73)\hat{H}_0\simeq-\nu(q_x\Gamma_1+q_y\Gamma_2)+\kappa q_{\perp}^{2}\cos k_z\Gamma_3, \tag*{(73)}

where ν=κ=2t\nu=\kappa=2t and q⊥2=qx2+qy2q_{\perp}^{2}=q_x^{2}+q_y^{2}. To leading order in q⊥q_\perp, the system may therefore be viewed as a continuous family of (2+1)(2+1)-dimensional Dirac cones parametrized by kzk_z.

The Wilsonian RG transformation proceeds by integrating out modes in the transverse momentum shell Λ/b<q⊥<Λ\Lambda/b<q_\perp<\Lambda, with b>1b>1, while the full kzk_z dependence is retained. We parametrize the coarse graining by the logarithmic running scale

ℓ=ln⁡b,Λ(ℓ)=Λ0e−ℓ,(74)\ell=\ln b,\qquad\Lambda(\ell)=\Lambda_0e^{-\ell}, \tag*{(74)}

so that increasing ℓ\ell corresponds to progressively lower energies and momenta. Equivalently, along the RG flow, ω(ℓ)=ω0e−ℓ\omega(\ell)=\omega_0e^{-\ell} and q⊥(ℓ)=q⊥,0e−ℓq_\perp(\ell)=q_{\perp,0}e^{-\ell}, whereas kzk_z is not rescaled. The corresponding running energy cutoff is EΛ(ℓ)=νΛ(ℓ)=E0e−ℓE_\Lambda(\ell)=\nu\Lambda(\ell)=E_0e^{-\ell}.

With this anisotropic scaling, the fermionic field has scaling dimension [ψ]=1[\psi]=1. Consequently, the onsite four-fermion coupling has [U]=−1[U]=-1, whereas the coefficient Δ\Delta of the symmetry-allowed operator Δcos⁡kzΓ3\Delta\cos k_z\Gamma_3 has [Δ]=+1[\Delta]=+1. The bare curvature coefficient is also irrelevant, with [κ]=−1[\kappa]=-1. Thus the Hubbard interaction is perturbatively irrelevant at the Gaussian nodal-line fixed point, while a momentum-independent Γ3\Gamma_3 term modulated by cos⁡kz\cos k_z is a relevant perturbation.

The Hubbard interaction is conveniently written in particle-hole symmetric form as

HU=U∑R,a(nRa↑−12)(nRa↓−12),(75)H_U=U\sum_{\mathbf{R},a}\left(n_{\mathbf{R}a\uparrow}-\frac{1}{2}\right)\left(n_{\mathbf{R}a\downarrow}-\frac{1}{2}\right), \tag*{(75)}

where a=A,B,C,Da=A,B,C,D labels the orbitals within the unit cell. At half filling the first-order Hartree contribution vanishes. The leading momentum-dependent correction is consequently the second-order self-energy represented by Figure 15.

We denote fermionic and bosonic four-momenta by k=(k,iωn)k=(\mathbf{k},i\omega_n) and q=(q,iΩm)q=(\mathbf{q},i\Omega_m), respectively, and use the notation Jk=(T/N)∑k,ΩnJ_k=(T/N)\sum_{k,\Omega_n}. The bare propagator is the 4×44\times4 orbital matrix

G(0)(k)=[iωn−H^0(k)]−1.(76)G^{(0)}(k)=\left[i\omega_n-\hat{H}_0(\mathbf{k})\right]^{-1}. \tag*{(76)}

For the local Hubbard vertex, the opposite-spin particle-hole bubble connecting orbitals aa and bb is

Πab,σˉ(0)(q)=−∫pGab,σˉ(0)(p+q)Gba,σˉ(0)(p),(77)\Pi^{(0)}_{ab,\bar{\sigma}}(q)=-\int_p G^{(0)}_{ab,\bar{\sigma}}(p+q)G^{(0)}_{ba,\bar{\sigma}}(p), \tag*{(77)}

where the minus sign originates from the closed fermion loop. The second-order self-energy may then be written as

Σab,σ(2)(k)=U2∫qΠaσˉ(0)(q)Gab,σ(0)(k−q).(78)\Sigma^{(2)}_{ab,\sigma}(k)=U^2\int_q\Pi^{(0)}_{a\bar{\sigma}}(q)G^{(0)}_{ab,\sigma}(k-q). \tag*{(78)}

Equivalently, inserting Eq. (D6),

Σab,σ(2)(k)=−U2∫p,qGab,σ(0)(k−q)Gab,σˉ(0)(p+q)Gba,σˉ(0)(p).(79)\Sigma^{(2)}_{ab,\sigma}(k)=-U^2\int_{p,q}G^{(0)}_{ab,\sigma}(k-q)G^{(0)}_{ab,\bar{\sigma}}(p+q)G^{(0)}_{ba,\bar{\sigma}}(p). \tag*{(79)}

This is the orbital-space form of the second-order Hubbard self-energy. The explicit orbital indices are important because the interaction is local and diagonal in the orbital basis, whereas the noninteracting propagator generally contains off-diagonal orbital components.

The component responsible for deforming the nodal manifold is obtained by projecting the static self-energy onto Γ3\Gamma_3,

Σ3(2)(k,0)=14Tr⁡[Γ3Σ(2)(k,0)].(80)\Sigma^{(2)}_3(\mathbf{k},0)=\frac{1}{4}\operatorname{Tr}\left[\Gamma_3\Sigma^{(2)}(\mathbf{k},0)\right]. \tag*{(80)}

Along the original nodal line, k=(π/2,π/2,kz)\mathbf{k}=(\pi/2,\pi/2,k_z), its dependence on kzk_z can be expanded in lattice harmonics. The leading symmetry-allowed term is the first harmonic,

Σ3(2)(π2,π2,kz;0)=A1cos⁡kz+⋯ ,(81)\Sigma^{(2)}_3\left(\frac{\pi}{2},\frac{\pi}{2},k_z;0\right)=A_1\cos k_z+\cdots, \tag*{(81)}

where the ellipsis represents higher harmonics and higher-order interaction corrections. The coefficient A1A_1 is obtained by projecting onto cos⁡kz\cos k_z,

A1=1π∫−ππdkzcos⁡kzΣ3(2)(π2,π2,kz;0).(82)A_1=\frac{1}{\pi}\int_{-\pi}^{\pi}dk_z\cos k_z\Sigma^{(2)}_3\left(\frac{\pi}{2},\frac{\pi}{2},k_z;0\right). \tag*{(82)}

Evaluating the internal momentum and Matsubara sums with the full lattice propagator gives

A1=c3U2t,c3≈2.91×10−3.(83)A_1=c_3\frac{U^2}{t},\qquad c_3\approx2.91\times10^{-3}. \tag*{(83)}

The leading interaction-generated correction to the single-particle Hamiltonian is therefore

δH^int=ΔUcos⁡kzΓ3,ΔU=c3U2t.(84)\delta\hat{H}_{\mathrm{int}}=\Delta_U\cos k_z\Gamma_3,\qquad\Delta_U=c_3\frac{U^2}{t}. \tag*{(84)}

We now incorporate this result into the Wilsonian flow. Introducing the dimensionless Hubbard coupling uu and δ=Δ/EΛ\delta=\Delta/E_\Lambda, with EΛ=vΛE_\Lambda=v\Lambda, the weak-coupling RG equations have the form

dudℓ=−u+Buu2+⋯ ,\frac{du}{d\ell}=-u+B_u u^2+\cdots,
dδdℓ=δ+CΔu2+⋯ .(85)\frac{d\delta}{d\ell}=\delta+C_\Delta u^2+\cdots. \tag*{(85)}

The first terms in these equations follow directly from tree-level power counting. The CΔu2C_\Delta u^2 contribution is the source term associated with the second-order self-energy derived above. The coefficients BuB_u and CΔC_\Delta depend on the normalization of the dimensionless couplings and on the cutoff convention, whereas the coefficient c3c_3 obtained by matching to the full lattice self-energy is a physical microscopic quantity.

In the weak-coupling regime, u0≪1u_0\ll1, the nonlinear correction to the first flow equation can be neglected, giving

u(ℓ)=u0e−ℓ.(86)u(\ell)=u_0e^{-\ell}. \tag*{(86)}

For the microscopic nodal-line Hamiltonian, δ0=0\delta_0=0. Substitution of Eq. (86) into the second flow equation yields

δ(ℓ)=CΔu023(eℓ−e−2ℓ).(87)\delta(\ell)=\frac{C_\Delta u_0^2}{3}\left(e^\ell-e^{-2\ell}\right). \tag*{(87)}

The interaction itself therefore flows to zero, whereas the relevant quadratic perturbation generated by it grows under coarse graining. This does not imply a divergent physical gap: since the running energy cutoff behaves as EΛ=E0e−ℓE_\Lambda=E_0e^{-\ell}, the product EΛδ(ℓ)E_\Lambda\delta(\ell) approaches a finite infrared value. Matching to the lattice self-energy gives Δphys=ΔU=c3U2/t\Delta_{\mathrm{phys}}=\Delta_U=c_3U^2/t.

The consequences for the nodal manifold follow immediately. Along the original nodal line the effective Hamiltonian becomes

H^eff(π2,π2,kz)=ΔUcos⁡kzΓ3,(88)\hat{H}_{\mathrm{eff}}\left(\frac{\pi}{2},\frac{\pi}{2},k_z\right)=\Delta_U\cos k_z\Gamma_3, \tag*{(88)}

with spectrum

E±(kz)=±∣ΔUcos⁡kz∣.(89)E_\pm(k_z)=\pm\left|\Delta_U\cos k_z\right|. \tag*{(89)}

Zero energy therefore survives only at kz=±π/2k_z=\pm\pi/2. Expanding around, for example, kz=π/2+qzk_z=\pi/2+q_z, for which cos⁡kz≈−qz\cos k_z\approx-q_z, gives

H^NP≈−vxqxΓ1−vyqyΓ2−vzqzΓ3,vz≈ΔU.(90)\hat{H}_{\mathrm{NP}}\approx-v_xq_x\Gamma_1-v_yq_y\Gamma_2-v_zq_z\Gamma_3,\qquad v_z\approx\Delta_U. \tag*{(90)}

Thus the weak-coupling infrared theory is a strongly anisotropic nodal-point semimetal, with vz∼U2v_z\sim U^2, while vx∼vy≈2tv_x\sim v_y\approx2t. The crossover between nodal-line-like and nodal-point-like behavior occurs at the parametrically small energy scale E×∼ΔU∼c3U2/tE_\times\sim\Delta_U\sim c_3U^2/t.

Berry phase and Wilson loop

The Berry-phase evolution provides a complementary geometrical characterization of this reconstruction. Since μy\mu_y commutes with the Hamiltonian, the four-component orbital problem decomposes into two sectors η=±1\eta=\pm1, defined by μy∣η⟩=η∣η⟩\mu_y|\eta\rangle=\eta|\eta\rangle. In either sector the low-energy Hamiltonian becomes

hη=−vqxτx−ηvqyτy+M(kz,q⊥)τz,(91)h_\eta=-v q_x\tau_x-\eta v q_y\tau_y+M(k_z,q_\perp)\tau_z, \tag*{(91)}

with

M(kz,q⊥)=cos⁡kz(ΔU+κq⊥2).(92)M(k_z,q_\perp)=\cos k_z\left(\Delta_U+\kappa q_\perp^2\right). \tag*{(92)}

For U=0U=0, the mass term vanishes as q⊥→0q_\perp\to0. A transverse loop linking the nodal line therefore surrounds a two-dimensional Dirac point for every kzk_z, and each sector carries Berry phase γη=ηπ\gamma_\eta=\eta\pi modulo 2π2\pi. Because the occupied orbital subspace is twofold degenerate, the gauge-invariant statement is that the corresponding Wilson-loop eigenvalues are both −1-1. This is the nodal-line counterpart of the π\pi Berry phase associated with an isolated Dirac point and of the band-contact-line picture discussed by Mikitik and Sharlai [46, 47].

For finite UU, consider a circular transverse loop, qx=q0cos⁡ϕq_x=q_0\cos\phi and qy=q0sin⁡ϕq_y=q_0\sin\phi. Neglecting the irrelevant curvature contribution κq02\kappa q_0^2, the effective mass on a fixed-kzk_z slice is m(kz)=ΔUcos⁡kzm(k_z)=\Delta_U\cos k_z. The d\mathbf{d} vector of the corresponding two-level problem traces a circle of fixed polar angle on the Bloch sphere. Since the Berry phase is one half of the enclosed solid angle, one obtains

γη(kz;q0)=ηπ[1−ΔUcos⁡kzv2q02+ΔU2cos⁡2kz](mod2π).(93)\gamma_\eta(k_z;q_0)=\eta\pi\left[1-\frac{\Delta_U\cos k_z}{\sqrt{v^2q_0^2+\Delta_U^2\cos^2 k_z}}\right]\quad(\mathop{\rm mod} 2\pi). \tag*{(93)}

For U=0U=0, Eq. (93) reduces to γη=ηπ\gamma_\eta=\eta\pi independently of kzk_z. For finite interaction, the mass vanishes at the surviving nodes, kz=±π/2k_z=\pm\pi/2, and the Berry phase remains exactly π\pi modulo 2π2\pi there. At a generic kzk_z, however, the corresponding two-dimensional slice is gapped, so the Berry phase of a finite loop is no longer quantized.

For displaying this evolution it is convenient to use the principal Wilson-loop phase,

γW(kz;q0)=π[1−∣ΔUcos⁡kz∣v2q02+ΔU2cos⁡2kz],(94)\gamma_W(k_z;q_0)=\pi\left[1-\frac{|\Delta_U\cos k_z|}{\sqrt{v^2q_0^2+\Delta_U^2\cos^2 k_z}}\right], \tag*{(94)}

which takes values between 00 and π\pi. In the noninteracting case γW=π\gamma_W=\pi along the entire nodal line, whereas for U≠0U\ne0 the π\pi phase remains pinned only at kz=±π/2k_z=\pm\pi/2. Moreover, in the limit q0→0q_0\to0 the Berry phase approaches zero modulo 2π2\pi for every gapped slice kz≠±π/2k_z\ne\pm\pi/2, while remaining π\pi at the two nodal points.

References

References

  1. [1]A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K. Geim, The electronic properties of graphene, Reviews of Modern Physics 81, 109 (2009).DOI
  2. [2]T. O. Wehling, A. M. Black-Schaffer, and A. V. Balatsky, Dirac materials, Advances in Physics 63, 1 (2014).DOI
  3. [3]S. M. Young, S. Zaheer, J. C. Y. Teo, C. L. Kane, E. J. Mele, and A. M. Rappe, Dirac Semimetal in Three Dimensions, Phys. Rev. Lett. 108, 140405 (2012).pubmed.ncbi.nlm.nih.gov/22540776
  4. [4]N. P. Armitage, E. J. Mele, and A. Vishwanath, Weyl and Dirac semimetals in three-dimensional solids, Reviews of Modern Physics 90, 015001 (2018).DOI
  5. [5]J. Maciejko and R. Nandkishore, Weyl semimetals with short-range interactions, Physical Review B 90, 035126 (2014).DOI
  6. [6]J. González, Phase diagram of the quantum electrodynamics of two-dimensional and three-dimensional Dirac semimetals, Phys. Rev. B 92, 125115 (2015).
  7. [7]J.-R. Wang, G.-Z. Liu, X. Wan, and C. Zhang, Quantum criticality of the excitonic insulating transition in the nodal-line semimetal ZrSiS, Phys. Rev. B 101, 245151 (2020).
  8. [8]B. Roy and S. Das Sarma, Quantum phases of interacting electrons in three-dimensional dirty Dirac semimetals, Phys. Rev. B 94, 115137 (2016).
  9. [9]J. H. Pixley, P. Goswami, and S. Das Sarma, Anderson Localization and the Quantum Phase Diagram of Three Dimensional Disordered Dirac Semimetals, Phys. Rev. Lett. 115, 076601 (2015).
  10. [10]B. Fu, W. Zhu, Q. Shi, Q. Li, J. Yang, and Z. Zhang, Accurate Determination of the Quasiparticle and Scaling Properties Surrounding the Quantum Critical Point of Disordered Three-Dimensional Dirac Semimetals, Phys. Rev. Lett. 118, 146401 (2017).DOI
  11. [11]J. P. Santos Pires, B. Amorim, A. Ferreira, İ. Adagideli, E. R. Mucciolo, and J. M. Viana Parente Lopes, Breakdown of universality in three-dimensional Dirac semimetals with random impurities, Phys. Rev. Research 3, 013183 (2021).DOI
  12. [12]P. Hosur, S. Ryu, and A. Vishwanath, Chiral topological insulators, superconductors, and other competing orders in three dimensions, Phys. Rev. B 81, 045120 (2010).
  13. [13]Y. Huang, H. Guo, J. Maciejko, R. T. Scalettar, and S. Feng, Antiferromagnetic transitions of Dirac fermions in three dimensions, Phys. Rev. B 102, 155152 (2020).
  14. [14]M. Moshe and J. Zinn-Justin, Quantum field theory in the large N limit: A review, Physics Reports 385, 69 (2003).DOI
  15. [15]C. C. Fang, H. Weng, X. Dai, and Z. Fang, Topological nodal line semimetals, Chinese Physics B 25, 117106 (2016), arXiv:1609.05414 [cond-mat.mes-hall].DOI
  16. [16]Z. Liu, L. Wang, and D.-X. Yao, Triply degenerate nodal lines in topological and nontopological metals, Phys. Rev. B 103, 205145 (2021).DOI
  17. [17]F. Liu, F. Qu, I. Žutić, and M. Malard, Vacancy-engineered nodal-line semimetals, Sci. Rep. 12, 14981 (2022).DOI
  18. [18]Y. Huh, E.-G. Moon, and Y. B. Kim, Long-range Coulomb interaction in nodal-ring semimetals, Phys. Rev. B 93, 035138 (2016).DOI
  19. [19]B. Roy, Interacting nodal-line semimetal: Proximity effect and spontaneous symmetry breaking, Phys. Rev. B 96, 041113(R (2017).DOI
  20. [20]J. Kang, J. Zou, K. Li, S.-L. Yu, and L.-B. Shao, Quasiparticle properties under interactions in Weyl and nodal line semimetals, Scientific Reports 9, 2824 (2019).DOI
  21. [21]A. R. Medeiros-Silva, M. Malard, R. G. Pereira, and T. Paiva, Electronic interactions in a vacancy-engineered honeycomb lattice: Transition from a nodal-line semimetal to a magnetic insulator, Phys. Rev. B 111, 191552 (2025).
  22. [22]G. Jose and B. Uchôa, Quantum critical scaling of gapped phases in nodal-line semimetals, Phys. Rev. B 101, 115123 (2020).DOI
  23. [23]S. Sur and B. Roy, Unifying Interacting Nodal Semimetals: A New Route to Strong Coupling, Phys. Rev. Lett. 123, 207601 (2019).DOI
  24. [24]M. D. Uryszek, F. Krüger, and E. Christou, Fermionic criticality of anisotropic nodal point semimetals away from the upper critical dimension: Exact exponents to leading order in 1/Nf, Phys. Rev. Research 2, 043265 (2020).
  25. [25]J.-R. Wang, W. Li, and C.-J. Zhang, Correlated interaction effects in the three-dimensional semi-Dirac semimetal, Phys. Rev. B 107, 155125 (2023).
  26. [26]S. Pujari, T. C. Lang, G. Murthy, and R. K. Kaul, Interaction-Induced Dirac Fermions from Quadratic Band Touching in Bilayer Graphene, Phys. Rev. Lett. 117, 086404 (2016).DOI
  27. [27]M. M. Elsayed, B. Uchôa, and V. N. Kotov, Coulomb interactions in systems of generalized semi-Dirac fermions, Phys. Rev. B 111, 165127 (2025).DOI
  28. [28]R. Blankenbecler, D. J. Scalapino, and R. L. Sugar, Monte Carlo calculations of coupled boson-fermion systems. I, Phys. Rev. D 24, 2278 (1981).DOI
  29. [29]J. E. Hirsch, Two-dimensional Hubbard model: Numerical simulation study, Phys. Rev. B 31, 4403 (1985).pubmed.ncbi.nlm.nih.gov/9936373
  30. [30]S. R. White, D. J. Scalapino, R. L. Sugar, E. Y. Loh, J. E. Gubernatis, and R. T. Scalettar, Numerical study of the two-dimensional Hubbard model, Phys. Rev. B 40, 506 (1989).pubmed.ncbi.nlm.nih.gov/9990942
  31. [31]R. R. dos Santos, Introduction to quantum Monte Carlo simulations for fermionic systems, Brazilian Journal of Physics 33, 36 (2003).
  32. [32]J. E. Hirsch, Discrete Hubbard-Stratonovich transformation for fermion lattice models, Phys. Rev. B 28, 4059 (1983).DOI
  33. [33]E. Y. Loh, J. E. Gubernatis, R. T. Scalettar, S. R. White, D. J. Scalapino, and R. L. Sugar, Sign problem in the numerical simulation of many-electron systems, Phys. Rev. B 41, 9301 (1990).pubmed.ncbi.nlm.nih.gov/9993272
  34. [34]M. Troyer and U.-J. Wiese, Computational Complexity and Fundamental Limitations for Fermionic Quantum Monte Carlo Simulations, Phys. Rev. Lett. 94, 170201 (2005).pubmed.ncbi.nlm.nih.gov/15904269
  35. [35]Here, we denote the magnetization critical exponent by β̃ to avoid confusion with the inverse temperature β = 1/kBT.
  36. [36]The cutoff M is chosen by comparing the numerical derivative of the energy data with the derivative obtained from efit(T).
  37. [37]To account for the staggered AFM pattern within the xy plane, we include an additional form factor (−1)xi+yi in the spin correlations. This phase factor, which alternates the sign between neighboring sites along the x and y directions, is used throughout but omitted from the notation above for brevity.
  38. [38]R. K. Kaul, Spin nematics, valence-bond solids, and spin liquids in SO(n) quantum spin models on the triangular lattice, Phys. Rev. Lett. 115, 157202 (2015).DOI
  39. [39]T. Sato, F. F. Assaad, and T. Grover, Quantum Monte Carlo simulation of frustrated Kondo lattice models, Phys. Rev. Lett. 120, 107201 (2018).DOI
  40. [40]Y. Otsuka, K. Seki, S. Sorella, and S. Yunoki, Dirac electrons in the square-lattice Hubbard model with a d-wave pairing field: The chiral Heisenberg universality class revisited, Phys. Rev. B 102, 235105 (2020).DOI
  41. [41]For the analysis of Rc, we chose δq = (0, 0, 2π/L).
  42. [42]Here, we denote the magnetization critical exponent by β̃ to avoid confusion with the inverse temperature β = 1/kBT.
  43. [43]A. W. Sandvik, Stochastic method for analytic continuation of quantum Monte Carlo data, Phys. Rev. B 57, 10287 (1998).DOI
  44. [44]K. S. D. Beach, Identifying the maximum entropy method as a special limit of stochastic analytic continuation, arXiv preprint cond-mat/0403055 (2004).DOI
  45. [45]It is worth noting that finite-size effects are partially mitigated by averaging results obtained with PBC and APBC. However, for momentum-resolved Green’s functions, changing the boundary conditions changes the set of allowed wavevectors, so that such an average cannot generally be performed at fixed k.
  46. [46]G. P. Mikitik and Y. V. Sharlai, Manifestation of Berry’s Phase in Metal Physics, Phys. Rev. Lett. 82, 2147 (1999).DOI
  47. [47]G. P. Mikitik and Y. V. Sharlai, Electron Energy Spectrum and the Berry Phase in a Graphite Bilayer, Phys. Rev. B 77, 113407 (2008).DOI
  48. [48]The Fourier transform of Eqs. (A2) and (A3) gives Eq. (1), with kx ≡ k·a1/k2, ky ≡ k·a2/k2, kz ≡ k·a3. Thus kx, ky ∈ [−π/2, π/2), while kz ∈ [−π, π).

Paper details

vlnc:W_u7uv5sc5
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 5

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents