betaFeedback

WorksPhysics and Astronomy

Magnons and excitons from Bethe-Salpeter equation and its application to graphene-based magnets

Abstract

With the recent isolation of two-dimensional semiconducting magnetic materials, the theoretical study of spin-flip collective excitations (magnons) along with spin-conserving electron-hole bound pairs (excitons) has become particularly relevant. Describing both types of quasiparticles within the same theoretical framework is possible, in principle, but the well-established many-body formalism for the calculation of excitons, based on the intertwined Bethe-Salpeter and Hedin's equations, is not satisfactory for magnons, since the usual approximations produce a spurious gapped Goldstone mode. Here, we show that excitons and magnons can be computed on the same footing under the same Bethe-Salpeter formalism, while respecting the gapless Goldstone mode for magnons. We demonstrate how this can be achieved through the evaluation of the full Bethe-Salpeter equation, beyond the Tamm-Dancoff approximation in model Hamiltonians. As an example, we apply it to ferromagnetic and antiferromagnetic graphene, faithfully represented by the extended Hubbard model. Our results clarify the roles of on-site and non-local interactions in exciton and magnon appearance and identify consistency requirements relevant to extending the approach to first-principles calculations.

Introduction

With the recent isolation of two-dimensional semiconducting magnetic materials [1, 2, 3], such as CrI3\mathrm{CrI}_3 [4], Cr2Ge2Te6\mathrm{Cr}_2\mathrm{Ge}_2\mathrm{Te}_6 [5], CrSBr\mathrm{CrSBr} [6], or PtSe2\mathrm{PtSe}_2 [7], the study of spin-flip excitations (magnons) along with spin-conserving excitations (excitons) has become a focus of increasing interest [8]. In these same semiconducting materials, as is also the case in non-magnetic ones, it has been recognized that excitons are key to understanding their subgap optical response [9, 10]. The coexistence of excitons and magnons, along with their reciprocal couplings, has raised fundamental questions that are being addressed by experimental [11, 12, 13] and theoretical [14, 15] works. The effects of the presence of excitons and magnons in a system have been discussed with respect to transport phenomena [12, 16], second-harmonic generation [17], negative refraction [18], as well as the magnetic control of exciton–polariton condensation [19] and the magnetic binding of excitons in Mott insulators [20].

Theoretical and computational studies of excitons in two-dimensional semiconductors are gaining momentum with new implementations and the development of codes such as BerkeleyGW [21], Yambo [22], EDUS [23], XATU [24, 25], Exciting [26], among others. The methodologies considered can be essentially divided into those using real-time dynamical approaches [23, 27, 28] and those based on solving the Bethe-Salpeter equation (BSE) in the frequency domain [21, 22, 24, 25, 26, 29]. Regarding the computation of magnons, theoretical works in the current literature are mainly based on Heisenberg models [30, 31] through linear spin wave theory [32] with exchange parameters obtained from density functional theory (DFT) calculations [33, 34, 35, 36]. The computation of the parameters has been standardized by codes such as TB2J [37]. On the other hand, many efforts have been reported to calculate magnons from time-dependent density functional theory (TDDFT) [38, 39, 40, 41, 42, 43, 44, 45, 47, 48, 49, 50] [46] or many-body perturbation theory [51, 52, 53, 54, 55, 56, 57]. These methods have been mostly confined to the study of standard ferromagnets, but their use has recently been extended to 2D magnets [56, 57, 58, 59].

One of the simplest and yet still experimentally relevant systems is that of hydrogenated graphene [60, 61, 62], where the controlled adsorption of hydrogen atoms creates unpaired electrons which can couple ferro or antiferromagnetically [63, 64, 65]. Recently, it has been shown that graphene can host all nonrelativistic magnetic phases (i.e., diamagnetism, ferromagnetism, antiferromagnetism, paramagnetism, altermagnetism, ferrimagnetism, and fully compensated ferrimagnetism) [66]. In this work, we are interested in the evaluation of excitons and magnons in this system using the BSE as a common methodology. By focusing on model Hamiltonians, we single out one of the reasons for a known failure of the application of the BSE formalism to magnons, the appearance of a gap in the Goldstone mode, and propose how to successfully address this problem. Our conclusions logically extend to the first-principles evaluation of magnons under the BSE, although there are other reasons for this more challenging case that are still being discussed in the literature [51].

We begin in Sec. II by providing an overview of the BSE in its various forms and approximations. In Sec. III, we narrow the discussion down to model Hamiltonians, in particular focusing on the extended Hubbard model. We provide explicit expressions for the BSE, dissecting the kernel according to the different interaction terms in the Hamiltonian. We also clarify why the (gapless) Goldstone can be obtained within the BSE formalism. Finally, in Sec. IV, we apply the previously discussed methodology to the computation of excitons and magnons in magnetic graphene (ferromagnetic and antiferromagnetic), for which the extended Hubbard Hamiltonian provides a faithful description.

different forms of the Bethe-Salpeter equation

Consider a generic Hamiltonian written as

H=HMF+(Ve−e−Ve−eMF),(1)H = H^{\mathrm{MF}} + \left(V_{\mathrm{e-e}} - V_{\mathrm{e-e}}^{\mathrm{MF}}\right), \tag*{(1)}

where the first term

HMF=H0+Ve−eMFH^{\mathrm{MF}} = H_0 + V_{\mathrm{e-e}}^{\mathrm{MF}}

is a self-consistent mean-field (MF) Hamiltonian with H0H_0 the non-interacting term (kinetic plus electron-ion interaction energy) and Ve−eMFV_{\mathrm{e-e}}^{\mathrm{MF}} the MF potential. The second term in Eq. (1) thus represents the “residual” interaction. Our main focus here is on non-relativistic systems, so spin-orbit coupling will be disregarded throughout this work. We denote the creation and annihilation operators of a Bloch electron in a state ψnkσ(r)\psi_{nk\sigma}(r) with band index nn, crystal momentum kk, spin σ\sigma, and energy ξnkσ\xi_{nk\sigma} by cnkσ†c^\dagger_{nk\sigma} and cnkσc_{nk\sigma}, respectively, where these operators diagonalize the MF Hamiltonian:

HMF=∑nkσξnkσcnkσ†cnkσ.H^{\mathrm{MF}} = \sum_{nk\sigma} \xi_{nk\sigma} c^\dagger_{nk\sigma}c_{nk\sigma}.

In this basis, the term Ve−eV_{\mathrm{e-e}} is given by

Ve−e=12∑{ni,ki}∑σ,σ′Vnk1σ,mk2σ′;m′k4σ,n′k3σ′cnk1σ†cmk2σ′†cn′k3σ′cm′k4σV_{\mathrm{e-e}} = \frac{1}{2}\sum_{\{n_i,k_i\}} \sum_{\sigma,\sigma'} V_{nk_1\sigma,mk_2\sigma';m'k_4\sigma,n'k_3\sigma'} c^\dagger_{nk_1\sigma}c^\dagger_{mk_2\sigma'}c_{n'k_3\sigma'}c_{m'k_4\sigma}

with

Vnk1σ,mk2σ′;m′k3σ,n′k4σ′=∫dr dr′ ψnk1σ∗(r)ψmk2σ′∗(r′)V(r,r′)ψm′k3σ(r)ψn′k4σ′(r′).(2)V_{nk_1\sigma,mk_2\sigma';m'k_3\sigma,n'k_4\sigma'} = \int dr\,dr'\,\psi^*_{nk_1\sigma}(r)\psi^*_{mk_2\sigma'}(r')V(r,r')\psi_{m'k_3\sigma}(r)\psi_{n'k_4\sigma'}(r'). \tag*{(2)}

In principle, the bare interaction should be considered here. In model Hamiltonians, as shown below, simplifications are made for both the non-interacting and interaction terms.

Both excitons, considered here as spin-conserving electron-hole (e-h) bound pairs, and magnons, spin-flip collective excitations, can, in principle, be evaluated solving the BSE in its various forms. The bound nature of the electron and the hole and an associated energy below the band gap is characteristic of the exciton, although the BSE can also describe unbound e-h pairs in the continuum. On the other hand, we will generally refer to a magnon when the excitation flips a single spin from majority bands to minority bands and lowers the total spin (total SzS_z component) by one unit. Furthermore, we will refer to magnon only when the character of this excitation is collective rather than particle-hole type. The latter are known as Stoner-like excitations and can also be obtained from the BSE. Spin-flip excitations from minority bands to majority bands, that is, those raising the total spin by one unit in the ferromagnetic system will be simply called dark excitons, regardless of its binding nature.

Here, we briefly summarize the different approaches and fix our terminology. Notice that it feeds on nuclear physics, quantum chemistry, and condensed matter terminologies:

  • Independent particle approximation (IPA). The ground state of the system is considered to be a single Slater determinant, ∣Φ0⟩|\Phi_0\rangle, obtained from a MF procedure, either Hartree-Fock (HF) or DFT. The excitations are approximated by single particle-hole pairs:

∣X⟩+=cckσ†cvkσ∣Φ0⟩,|X\rangle^+ = c^\dagger_{ck\sigma}c_{vk\sigma}|\Phi_0\rangle,

where c†/cc^\dagger/c create/destroy MF quasiparticles with c,vc,v being conduction and valence bands. Clearly, ∣X⟩−=cvkσ†cckσ∣Φ0⟩=0|X\rangle^- = c^\dagger_{vk\sigma}c_{ck\sigma}|\Phi_0\rangle= 0. This approximation is only valid if the two-body residual interactions are neglected in the Hamiltonian and may be unnecessarily labelled as BSE/IPA.

  • Tamm-Dancoff approximation (TDA). The ground state is also approximated by a MF single Slater determinant ∣Φ0⟩|\Phi_0\rangle, but the excitations are now collective e-h pairs:

∣X⟩+=∑c,v,kAc,vσ′σ(k)cckσ′†cvkσ∣Φ0⟩.(3)|X\rangle^+ = \sum_{c,v,k} A^{\sigma'\sigma}_{c,v}(k)c^\dagger_{ck\sigma'}c_{vk\sigma}|\Phi_0\rangle. \tag*{(3)}

Again, ∣X⟩−=∑c,v,kBc,vσ′σ(k)cvkσ†cckσ∣Φ0⟩=0|X\rangle^- = \sum_{c,v,k} B^{\sigma'\sigma}_{c,v}(k)c^\dagger_{vk\sigma}c_{ck\sigma}|\Phi_0\rangle= 0. Choosing HF as the MF approximation, one obtains the BSE/TDA equation through equation of motion (EOM), although a similar result can be obtained by applying the variational principle to the excitation or through a simplification of the linearized time-dependent HF (TDHF) equations. In practice, however, the MF ground state is preferably obtained within a DFT approximation. The quasi-particle energies resulting in the BSE/TDA equations are then substituted by the DFT ones and the HF kernel by the Hartree plus exchange correlation kernel (this is actually justified as an approximation to time-dependent DFT (TDDFT)). Alternatively, an effective screened interaction can be considered in the HF kernel, while still using the DFT quasiparticle energies[24]. If the starting point is a model Hamiltonian, these considerations do not apply and the MF plus the BSE/TDA calculations are carried out simply on the basis of the model Hamiltonian at the HF level.

  • Random-phase approximation (RPA). The ground state ∣Φ~0⟩\left|\widetilde{\Phi}_0\right\rangle is no longer a single Slater determinant, being a correlated state composed of an infinite number of configurations with even-numbered e-h pair excitations[67]. The collective excitations are now considered to be

∣X⟩=∑c,v,k(Ac,vσσ′(k)cckσ′†cvkσ−Bc,vσ′σ(k)cvkσ′†cckσ)∣Φ~0⟩.(4)|X\rangle= \sum_{c,v,\mathbf{k}} \left(A_{c,v}^{\sigma\sigma'}(\mathbf{k})c_{c\mathbf{k}\sigma'}^{\dagger}c_{v\mathbf{k}\sigma} - B_{c,v}^{\sigma'\sigma}(\mathbf{k})c_{v\mathbf{k}\sigma'}^{\dagger}c_{c\mathbf{k}\sigma}\right)\left|\widetilde{\Phi}_0\right\rangle. \tag*{(4)}

Again, the EOM leads to the BSE/RPA equation, also known as the Casida equation[68] or full BSE. The actual ground state ∣Φ~0⟩\left|\widetilde{\Phi}_0\right\rangle is, in practice, not needed, with the derivation still based on ∣Φ0⟩\left|\Phi_0\right\rangle using the quasi-boson approximation (QBA) of EOM[69]. A common approximation in solids involves discarding the Fock term to get a simplified version of the RPA which is still useful to describe plasmons. A linearized time-dependent HF or TDDFT leads to the same type of equation and similar alternatives as those in the BSE/TDA apply here as well. HF is deprecated in favour of DFT and TDDFT actually justifies the use of DFT quasiparticle energies as in the TDA case. The actual exchange-correlation kernel or, alternatively, a HF screened effective kernel can be used. Again, in the case of model Hamiltonians, the bare HF approximation is the natural choice.

  • BSE/GW approximation. This fully first-principles approach is based on Green’s function many-body perturbation theory where screening is explicitly taken into account through Hedin’s equations[29]. In its most popular form, the quasiparticle energies are replaced by GW quasiparticle energies, hence the name. In other words, the local and static DFT self-energy operator is replaced by the non-local and dynamical GW self-energy operator in the calculation of the quasiparticle energies. The direct term in the kernel is usually screened by a static dielectric matrix. Both TDA and RPA flavours can be used, with the TDA the more frequently used due to its relative computational simplicity.

The different flavours of the BSE discussed above apply equally to the discussion of e-h pairs with finite centre-of-mass (COM) momentum and to the calculation of spin-conserving (σ′=σ\sigma' = \sigma) or spin-flip (σ′=σ‾\sigma' = \overline{\sigma}) excitations.

However, it is well documented[38, 39, 40, 41, 42, 43, 44, 45, 51, 52, 53, 54, 70, 71, 72, 73, 74, 75] [46] that even the most ambitious calculations to date, namely, BSE/GW/TDA, cannot reproduce the zero-energy Goldstone mode in magnetic systems, essentially because of the inconsistency between the dynamical screening in the quasiparticle self-energies and the static screening in the interaction kernel of the BSE. Importantly, we note that fixing this issue does not solve the problem, as we discuss below.

magnons and excitons in model Hamiltonians

Without loss of generality, here we are interested in model Hamiltonians such as the single-site extended Hubbard model:

H=T+VH = T + V
T=∑i,jtciσ†cjσT = \sum_{i,j} t c_{i\sigma}^{\dagger}c_{j\sigma}
V=U∑ini↑ni↓+J∑⟨i,j⟩∑σ,σ′niσnjσ′,(5)V = U \sum_i n_{i\uparrow}n_{i\downarrow} + J \sum_{\langle i,j\rangle} \sum_{\sigma,\sigma'} n_{i\sigma}n_{j\sigma'}, \tag*{(5)}

where ciσ†c_{i\sigma}^{\dagger} (ciσc_{i\sigma}) is the creation (annihilation) operator for an electron with spin σ\sigma at site ii, tt is the hopping amplitude (only ≠0\ne0 for nearest neighbours), niσ=ciσ†ciσn_{i\sigma} = c_{i\sigma}^{\dagger}c_{i\sigma} is the number density operator, and <i,j>< i,j > represents that the summation is done over nearest-neighbouring sites with indices ii and jj, each pair counted once.

In this case, the natural choice of action to compute excitons and magnons is through BSE/TDA or BSE/RPA, starting from a reference HF ground state. The inconsistency mentioned at the end of the previous section does not appear in this case since the quasi-particle self-energies and the interaction kernel are naturally evaluated using the same interaction model. This singles out a second, frequently overlooked difficulty when it comes to magnon computation, namely, that S2S^2 is typically not a good quantum number in the MF ground state ∣Φ0⟩\left|\Phi_0\right\rangle. This also opens a gap in the Goldstone mode[76] at the BSE/TDA level. Here, we demonstrate that it is possible to fix this by using BSE/RPA on an unrestricted HF (UHF) reference ground state.

BSE/TDA

The simplest way to arrive at the BSE/TDA is to apply the variational principle to the excitation proposed in Eq. (3). This leads to an eigenvalue equation where

Hvc,v′c′(k,k′,Q)=⟨v,c,k,Q∣H∣v′,c′,k′,Q⟩=δk,k′δvv′[δcc′ϵck+Q+∑cc′V(k+Q,k+Q)]−δk,k′δcc′[δvv′ϵvk+∑vv′v(k,k)]−Dvc,v′c′(k,k′,Q)+Xvc,v′c′(k,k′,Q),(6)H_{vc,v'c'}(\mathbf{k},\mathbf{k}',\mathbf{Q}) = \langle v,c,\mathbf{k},\mathbf{Q}|H|v',c',\mathbf{k}',\mathbf{Q}\rangle = \delta_{\mathbf{k},\mathbf{k}'}\delta_{vv'}\left[\delta_{cc'}\epsilon_{c\mathbf{k}+\mathbf{Q}} + \sum_{cc'} V(\mathbf{k}+\mathbf{Q},\mathbf{k}+\mathbf{Q})\right] - \delta_{\mathbf{k},\mathbf{k}'}\delta_{cc'}\left[\delta_{vv'}\epsilon_{v\mathbf{k}} + \sum_{vv'} v(\mathbf{k},\mathbf{k})\right] - D_{vc,v'c'}(\mathbf{k},\mathbf{k}',\mathbf{Q}) + X_{vc,v'c'}(\mathbf{k},\mathbf{k}',\mathbf{Q}), \tag*{(6)}

is the Hamiltonian in the single e-h pair excitation basis ∣v,c,k,Q⟩=cck+Q†cvk∣Φ0⟩|v,c,\mathbf{k},\mathbf{Q}\rangle= c^{\dagger}_{c\mathbf{k}+\mathbf{Q}}c_{v\mathbf{k}}|\Phi_{0}\rangle with respect to the MF ground state ∣Φ0⟩|\Phi_{0}\rangle. ϵc/vk\epsilon_{c/v\mathbf{k}} stands for non-interacting energies and the quasiparticle self-energy is given by

Σnm(k,k′)=∑v′′,k′′occ(Vnk,v′′k′′;mk′,v′′k′′−Vnk,v′′k′′;v′′k′′,mk′).(7)\Sigma_{nm}(\mathbf{k},\mathbf{k}')=\sum_{\substack{v'',\mathbf{k}''\\ \mathrm{occ}}}\left(V_{n\mathbf{k},v''\mathbf{k}'';m\mathbf{k}',v''\mathbf{k}''}-V_{n\mathbf{k},v''\mathbf{k}'';v''\mathbf{k}'',m\mathbf{k}'}\right). \tag*{(7)}

where the summation runs over all the occupied states. The MF quasi-particle energies ξc/vk\xi_{c/v\mathbf{k}} result from diagonalizing

ϵc/vk+Σnm(k).\epsilon_{c/v\mathbf{k}}+\Sigma_{nm}(\mathbf{k}).

The direct term

Dvcv′c′(k,k′,Q)=Vck+Q,v′k′;c′k′+Q,vk,(8)D_{vcv'c'}(\mathbf{k},\mathbf{k}',\mathbf{Q})=V_{c\mathbf{k}+\mathbf{Q},v'\mathbf{k}';c'\mathbf{k}'+\mathbf{Q},v\mathbf{k}}, \tag*{(8)}

and the exchange term

Xvcv′c′(k,k′,Q)=Vck+Q,v′k′;vk,c′k′+Q,(9)X_{vcv'c'}(\mathbf{k},\mathbf{k}',\mathbf{Q})=V_{c\mathbf{k}+\mathbf{Q},v'\mathbf{k}';v\mathbf{k},c'\mathbf{k}'+\mathbf{Q}}, \tag*{(9)}

form the BSE kernel

Kvcv′c′(k,k′,Q)=−Dvcv′c′(k,k′,Q)+Xvcv′c′(k,k′,Q).(10)K_{vcv'c'}(\mathbf{k},\mathbf{k}',\mathbf{Q})=-D_{vcv'c'}(\mathbf{k},\mathbf{k}',\mathbf{Q})+X_{vcv'c'}(\mathbf{k},\mathbf{k}',\mathbf{Q}). \tag*{(10)}

By comparison with the direct and exchange terms in their integral formulations (see Eq. (2)) and considering the δk,k′\delta_{\mathbf{k},\mathbf{k}'} prefactor in the diagonal terms, we can identify each self-energy term as a direct or exchange self-energy and write

Σnm(k,k′)=∑v′′,k′′occ(ΣnmD(k,k′)−ΣnmX(k,k′)).(11)\Sigma_{nm}(\mathbf{k},\mathbf{k}')=\sum_{\substack{v'',\mathbf{k}''\\ \mathrm{occ}}}\left(\Sigma^{D}_{nm}(\mathbf{k},\mathbf{k}')-\Sigma^{X}_{nm}(\mathbf{k},\mathbf{k}')\right). \tag*{(11)}

Notice that, for clarity’s sake, we have suppressed the spin indices. One can always safely absorb them in the labels c/vc/v.

As an illustrative exercise, we can consider distinguishable or non-identical (NI) particles so that the non-interacting ground state corresponds to a Hartree ground state ∣Φ0NI⟩|\Phi_{0}^{\mathrm{NI}}\rangle instead of a Slater determinant. In this hypothetical case, the BSE/TDA equation acquires the form (see Appendix A)

Hvc,v′c′NI(k,k′,Q)=δk,k′δvv′[δcc′ϵck+Q+Σcc′NI(k+Q,k+Q)]−δvv′δk,k′δcc′[ϵvk+ΣvvNI(k,k)]−δvv′δkk′Dvcvc′(k,k,Q)+(1−δvv′)Xvcv′c′(k,k′,Q).(12)\begin{aligned} H^{\mathrm{NI}}_{vc,v'c'}(\mathbf{k},\mathbf{k}',\mathbf{Q})={}&\delta_{\mathbf{k},\mathbf{k}'}\delta_{vv'}\left[\delta_{cc'}\epsilon_{c\mathbf{k}+\mathbf{Q}}+\Sigma^{\mathrm{NI}}_{cc'}(\mathbf{k}+\mathbf{Q},\mathbf{k}+\mathbf{Q})\right]\\ &-\delta_{vv'}\delta_{\mathbf{k},\mathbf{k}'}\delta_{cc'}\left[\epsilon_{v\mathbf{k}}+\Sigma^{\mathrm{NI}}_{vv}(\mathbf{k},\mathbf{k})\right]\\ &-\delta_{vv'}\delta_{\mathbf{k}\mathbf{k}'}D_{vcvc'}(\mathbf{k},\mathbf{k},\mathbf{Q})\\ &+(1-\delta_{vv'})X_{vcv'c'}(\mathbf{k},\mathbf{k}',\mathbf{Q}). \tag*{(12)} \end{aligned}

where ΣnmNI(k,k′)\Sigma^{\mathrm{NI}}_{nm}(\mathbf{k},\mathbf{k}') is the quasiparticle self-energy for non-identical electrons,

ΣnmNI(k,k′)=∑v′′,k′′occ[(1−δnv′′δkk′′)(1−δmv′′δk′k′′)Vnk,v′′k′′;mk′,v′′k′′].(13)\Sigma^{\mathrm{NI}}_{nm}(\mathbf{k},\mathbf{k}')=\sum_{\substack{v'',\mathbf{k}''\\ \mathrm{occ}}}\left[(1-\delta_{nv''}\delta_{\mathbf{k}\mathbf{k}''})(1-\delta_{mv''}\delta_{\mathbf{k}'\mathbf{k}''})V_{n\mathbf{k},v''\mathbf{k}'';m\mathbf{k}',v''\mathbf{k}''}\right]. \tag*{(13)}

and it consists of Hartree integrals only. Σvv′NI\Sigma^{\mathrm{NI}}_{vv'} is non-zero only when v=v′v=v', whereas the Hartree integral for identical electrons can be non-zero even for v≠v′v\ne v'. Notice that the kernel for non-identical electrons has the direct term DD only for v=v′v=v', and not for any pair of valence bands. Dvcvc′D_{vcvc'} cancels the self energy term for the interaction of the electron added in the conduction band c/c′c/c' with the electron removed from the valence band vv in Σcc′NI\Sigma^{\mathrm{NI}}_{cc'}. So, the direct term present in Hvc,v′c′NIH^{\mathrm{NI}}_{vc,v'c'} is not the general direct term like in Eq. (6). Additionally, Σvv′NI\Sigma^{\mathrm{NI}}_{vv'} is only non-zero for v=v′v=v' and the exchange term Xvcv′c′X_{vcv'c'} for v=v′v=v' is not present. This explicitly shows that the exchange interaction, only present for identical particles, is behind the exciton formation, since it is the DD term in the kernel that binds the electron and the hole.

BSE/RPA

Including anti-resonant excitations and their coupling with resonant excitations as described in Sec. II we arrive at the BSE/RPA equation, also known as Casida’s equation [29, 77] [78]. Similarly to the BSE/TDA case, we have that the general Hamiltonian remains block-diagonal in Q\mathbf{Q}:

He−h=(Q10⋯0Q2⋯⋮⋱),(14)H^{\mathrm{e-h}}= \begin{pmatrix} \boxed{Q_{1}} & 0 & \cdots\\ 0 & \boxed{Q_{2}} & \cdots\\ \vdots& \ddots& \end{pmatrix}, \tag*{(14)}

where each Qi\boxed{Q_{i}} block [25, 29, 79, 80] is now given by

Qi=(H(r,r)(Qi)H(r,a)(Qi)−H(r,a)(Qi)†−H(a,a)(Qi)∗),(15)\boxed{Q_{i}}= \begin{pmatrix} H^{(r,r)}(Q_{i}) & H^{(r,a)}(Q_{i})\\ -H^{(r,a)}(Q_{i})^{\dagger} & -H^{(a,a)}(Q_{i})^{*} \end{pmatrix}, \tag*{(15)}

and where H(r,r)H^{(r,r)} is the resonant part of the BSE Hamiltonian as considered in the TDA (i.e.i.e., creation of e-h pairs). −H(a,a)∗-H^{(a,a)*} is the antiresonant part corresponding to the annihilation of e-h pairs and H(r,a)H^{(r,a)} involves the coupling between the two distinct sectors [25, 81]. Having already discussed

H(r,r)(Qi)≡Hvc,v′c′(k,k′,Q),(16)H^{(r,r)}(Q_{i})\equiv H_{vc,v'c'}(\mathbf{k},\mathbf{k}',\mathbf{Q}), \tag*{(16)}

in Eq. (6), we now move to the anti-resonant and coupling terms. For the anti-resonant block, we have

Hvc,v′c′(a,a)(k,k′,Q)=δkk′δvv′[δcc′ϵck+Σcc′(k,k)]−δkk′δcc′[δvv′ϵvk+Q+Σv′v(k+Q,k+Q)]−Vck,v′k′+Q;c′k′,vk+Q+Vck,v′k′+Q;vk+Q,c′k′.(17)\begin{aligned} H^{(a,a)}_{vc,v'c'}(\mathbf{k},\mathbf{k}',\mathbf{Q})={}&\delta_{\mathbf{k}\mathbf{k}'}\delta_{vv'}\left[\delta_{cc'}\epsilon_{c\mathbf{k}}+\Sigma_{cc'}(\mathbf{k},\mathbf{k})\right]\\ &-\delta_{\mathbf{k}\mathbf{k}'}\delta_{cc'}\left[\delta_{vv'}\epsilon_{v\mathbf{k}+\mathbf{Q}}+\Sigma_{v'v}(\mathbf{k}+\mathbf{Q},\mathbf{k}+\mathbf{Q})\right]\\ &-V_{c\mathbf{k},v'\mathbf{k}'+\mathbf{Q};c'\mathbf{k}',v\mathbf{k}+\mathbf{Q}}+V_{c\mathbf{k},v'\mathbf{k}'+\mathbf{Q};v\mathbf{k}+\mathbf{Q},c'\mathbf{k}'}. \tag*{(17)} \end{aligned}

For the coupling terms, following existing literature [25, 80], we have

H(r,a)(Qi)=−D~(r,a)+X~(r,a).(18)H^{(r,a)}(Q_{i})=-\widetilde{D}^{(r,a)}+\widetilde{X}^{(r,a)}. \tag*{(18)}

where the off-diagonal kernel terms are given by

D~(r,a)=∫dr dr′ ψck+Q∗(r) ψc′k′(r′) V(r,r′) ψv′k′+Q(r) ψvk(r′)=Vck+Q,c′k′;v′k′+Q,vk(19)\widetilde{D}^{(r,a)} = \int d r\,d r'\,\psi^{*}_{c k+Q}(r)\,\psi_{c' k'}(r')\,V(r,r')\,\psi_{v' k'+Q}(r)\,\psi_{v k}(r') = V_{c k+Q,c' k';v' k'+Q,v k} \tag*{(19)}

and

X~(r,a)=∫dr dr′ ψck+Q∗(r) ψc′k′(r′) V(r,r′) ψvk(r) ψv′k′+Q(r′)=Vck+Q,c′k′;vk,v′k′+Q.(20)\widetilde{X}^{(r,a)} = \int d r\,d r'\,\psi^{*}_{c k+Q}(r)\,\psi_{c' k'}(r')\,V(r,r')\,\psi_{v k}(r)\,\psi_{v' k'+Q}(r') = V_{c k+Q,c' k';v k,v' k'+Q}. \tag*{(20)}

It can also be seen that, at Q=0Q=0, H(a,a)=H(r,r)H^{(a,a)}=H^{(r,r)} and H(r,a)H^{(r,a)} is complex symmetric, so that, with A≡H(r,r)A\equiv H^{(r,r)} and B=H(r,a)B=H^{(r,a)},

Qi=(AB−B∗−A∗)(21)Q_i=\begin{pmatrix} A & B \\ -B^{*} & -A^{*} \end{pmatrix} \tag*{(21)}

the familiar Casida form [25, 29, 79, 80].

As each QiQ_i block is pseudo-Hermitian following the relation [25]

Qi†=ηQiη,(22)Q_i^{\dagger}=\eta Q_i\eta, \tag*{(22)}

where η=diag⁡(I,−I)\eta=\operatorname{diag}(I,-I), with II the identity matrix with the same dimensions as each HH block, the eigenvalues of the system are real and come in pairs ω(Q),−ω(−Q)\omega(Q),-\omega(-Q) (i.e. ±ω\pm\omega with time-reversal symmetry) if

K=(H(r,r)(Qi)H(r,a)(Qi)H(r,a)(Qi)†H(a,a)(Qi∗))=ηQi(23)K=\begin{pmatrix} H^{(r,r)}(Q_i) & H^{(r,a)}(Q_i)\\ H^{(r,a)}(Q_i)^{\dagger} & H^{(a,a)}(Q_i^{*}) \end{pmatrix} =\eta Q_i \tag*{(23)}

is positive definite, which is Thouless’ stability condition for the mean-field state [67]. If KK is positive semidefinite, its null vectors are zero-frequency modes.

Goldstone theorem and BSE/RPA

Goldstone’s theorem states that, whenever a continuous symmetry of a system is broken, a zero mode appears in the excitation spectrum of said system[82]. The proof follows from [67, 82] by considering a broken symmetry one-body operator

G=∑pqgpqcp†cq(24)G=\sum_{pq}g_{pq}c_{p}^{\dagger}c_q \tag*{(24)}

that commutes with the Hamiltonian. Separating the sum into a particle-hole part and particle-particle plus hole-hole parts, we have

G=∑cv(gcvcc†cv+gvccv†cc)⏟G1+∑cc′gcc′cc†cc′⏟+∑vv′gvv′cv†cv′⏟G2.(25)G=\underbrace{\sum_{cv}\left(g_{cv}c_c^{\dagger}c_v+g_{vc}c_v^{\dagger}c_c\right)}_{G_1}+\underbrace{\sum_{cc'}g_{cc'}c_c^{\dagger}c_{c'}}_{ }+\underbrace{\sum_{vv'}g_{vv'}c_v^{\dagger}c_{v'}}_{G_2}. \tag*{(25)}

We have now temporarily suppressed for clarity’s sake the momentum indices. Since GG commutes with the Hamiltonian, one can apply the equation of motion (EOM) formulation [83] to write

⟨[cv†cc,[H,G1]]⟩0+⟨[cv†cc,[H,G2]]⟩0=0,(26)\left\langle[c_v^{\dagger}c_c,[H,G_1]]\right\rangle_0+\left\langle[c_v^{\dagger}c_c,[H,G_2]]\right\rangle_0=0, \tag*{(26)}

where ⟨⋅⟩0\langle\cdot\rangle_0 is defined as taking the expectation value relative to the HF ground state ∣Φ0⟩\lvert\Phi_0\rangle. Focusing on the second term of Eq. (26), we apply the Jacobi identity to write

T2=⟨[[cv†cc,H],G2]⟩0+⟨[H,[cv†cc,G2]]⟩0,(27)T_2=\left\langle[[c_v^{\dagger}c_c,H],G_2]\right\rangle_0+\left\langle[H,[c_v^{\dagger}c_c,G_2]]\right\rangle_0, \tag*{(27)}

which can be further simplified knowing that

cc′†cc∣Φ0⟩=0,cv′†cv∣Φ0⟩=δvv′∣Φ0⟩,(28)c_{c'}^{\dagger}c_{c}\lvert\Phi_0\rangle=0,\qquad c_{v'}^{\dagger}c_v\lvert\Phi_0\rangle=\delta_{vv'}\lvert\Phi_0\rangle, \tag*{(28)}

meaning

⟨[[cv†cc,H],G2]⟩0=0(29)\left\langle[[c_v^{\dagger}c_c,H],G_2]\right\rangle_0=0 \tag*{(29)}

by applying G2G_2 to the left and to the right. For the commutator between cv†ccc_v^{\dagger}c_c and G2G_2, we can easily see that

[cv†cc,G2]≃[cv†cc,cc′†cc′′]+[cv†cc,cv′†cv′′]≃δcc′cv†cc′′−δvc′′cc′†cc+δcv′cv†cv′′−δvv′′cv′†cc≃δcc′cv†cc′′−δvv′′cv′†cc,(30)[c_v^{\dagger}c_c,G_2]\simeq[c_v^{\dagger}c_c,c_{c'}^{\dagger}c_{c''}]+[c_v^{\dagger}c_c,c_{v'}^{\dagger}c_{v''}] \simeq\delta_{cc'}c_v^{\dagger}c_{c''}-\delta_{vc''}c_{c'}^{\dagger}c_c+\delta_{cv'}c_v^{\dagger}c_{v''}-\delta_{vv''}c_{v'}^{\dagger}c_c \simeq\delta_{cc'}c_v^{\dagger}c_{c''}-\delta_{vv''}c_{v'}^{\dagger}c_c, \tag*{(30)}

where we have dropped the coefficients gg to focus on the operators themselves, meaning that the outer commutator then is simply

⟨[H,[cv†cc,G2]]⟩0≃⟨[H,cv†cc]⟩0≃−⟨Φ0∣cv†ccH∣Φ0⟩≃−⟨Φvc∣H∣Φ0⟩=0,(31)\left\langle[H,[c_v^{\dagger}c_c,G_2]]\right\rangle_0\simeq\left\langle[H,c_v^{\dagger}c_c]\right\rangle_0 \simeq-\langle\Phi_0|c_v^{\dagger}c_cH|\Phi_0\rangle \simeq-\langle\Phi_{vc}|H|\Phi_0\rangle=0, \tag*{(31)}

per the vanishing Hamiltonian matrix elements between the HF ground state and singly-excited determinants ∣Φvc⟩\lvert\Phi_{vc}\rangle (Brillouin’s theorem [84]). Therefore, we have shown that the G2G_2 contribution vanishes. Regarding G1G_1, we have that T1T_1 must then also be zero, meaning

⟨[cv†cc,[H,G1]]⟩0=∑c′v′(gc′v′⟨[cv†cc,[H,cc′†cv′]]⟩0+gv′c′⟨[cv†cc,[H,cv′†cc′]]⟩0)=0.(32)\left\langle[c_v^{\dagger}c_c,[H,G_1]]\right\rangle_0 =\sum_{c'v'}\left(g_{c'v'}\left\langle[c_v^{\dagger}c_c,[H,c_{c'}^{\dagger}c_{v'}]]\right\rangle_0 +g_{v'c'}\left\langle[c_v^{\dagger}c_c,[H,c_{v'}^{\dagger}c_{c'}]]\right\rangle_0\right)=0. \tag*{(32)}

The two expectation values present are, however, the EOM definitions of the resonant and coupling blocks of Eq. (21) evaluated in the UHF ground state ∣Φ0⟩|\Phi_0\rangle [83]

Acv,c′v′=⟨[cv†cc,[H,cc′†cv′]]⟩0Bcv,c′v′=−⟨[cv†cc,[H,cv′†cc′]]⟩0.(33)\begin{aligned} A_{cv,c'v'} &= \left\langle\left[c_v^\dagger c_c,\left[H,c_{c'}^\dagger c_{v'}\right]\right]\right\rangle_0 \\ B_{cv,c'v'} &= -\left\langle\left[c_v^\dagger c_c,\left[H,c_{v'}^\dagger c_{c'}\right]\right]\right\rangle_0 . \tag*{(33)} \end{aligned}

This therefore means that Eq. (32) can be written simply as

∑c′v′(gc′v′′Acv,c′v′−gv′c′′Bcv,c′v′)=0,\sum_{c'v'}\left(g_{c'v'}'A_{cv,c'v'}-g_{v'c'}'B_{cv,c'v'}\right)=0,

which is directly the zero-mode equation for K\mathbf{K} as

\begin{pmatrix} A & B \\ B^* & A^* \end{pmatrix} \begin{pmatrix} g_{c'v'} \\ -g_{v'c'} \end{pmatrix} = \begin{pmatrix} 0 \\ 0 \end{pmatrix}. \tag*{(34)}\end{pmatrix}

Rotations about the magnetization axis leave a collinear state invariant, so the broken symmetry generators are SxS^x and SyS^y [85]. Combining them into

S±=Sx±iSy=∑nmkk′σσ′⟨nkσ∣S±∣mk′σ′⟩cnkσ†cmk′σ′,(35)\begin{aligned} S^\pm&= S^x \pm iS^y \\ &= \sum_{n m k k' \sigma\sigma'} \langle n k \sigma|S^\pm|m k'\sigma'\rangle c^\dagger_{n k \sigma}c_{m k'\sigma'}, \tag*{(35)} \end{aligned}

where n,mn,m run over the spin-polarized bands of both spins and ∣nkσ⟩|nk\sigma\rangle is the single-particle UHF spin-orbital, our null-vectors from Eq. (34) will then be

ZG±=(Sc′v′±−Sv′c′±).(36)Z_G^\pm= \begin{pmatrix} S^\pm_{c'v'} \\ -S^\pm_{v'c'} \end{pmatrix}. \tag*{(36)}

Writing each UHF spin-orbital as the direct product of a spatial Bloch function and a spin component as

∣nkσ⟩=∣ψnkσ⟩⊗∣σ⟩,(37)|nk\sigma\rangle=|\psi_{nk\sigma}\rangle\otimes|\sigma\rangle, \tag*{(37)}

we have that

S±=∑nmkk′σσ′⟨ψnkσ∣ψmk′σ′⟩⟨σ∣S±∣σ′⟩cnkσ†cmk′σ′.(38)S^\pm=\sum_{n m k k' \sigma\sigma'}\langle\psi_{nk\sigma}|\psi_{mk'\sigma'}\rangle\langle\sigma|S^\pm|\sigma'\rangle c^\dagger_{nk\sigma}c_{mk'\sigma'}. \tag*{(38)}

Since S+S^+ gives the companion zero mode in the ΔSz=+1\Delta S_z=+1 sector, we work with S−S^- without loss of generality

S−=∑nmkk′σσ′⟨ψnkσ∣ψmk′σ′⟩⟨σ∣S−∣σ′⟩cnkσ†cmk′σ′=∑nmkk′σσ′⟨ψnkσ∣ψmkσ′⟩δkk′δσ↓δσ′↑cnkσ†cmk′σ′=∑nmk⟨ψnk↓∣ψmk↑⟩cnk↓†cmk↑.(39)\begin{aligned} S^-&=\sum_{n m k k' \sigma\sigma'}\langle\psi_{nk\sigma}|\psi_{mk'\sigma'}\rangle\langle\sigma|S^-|\sigma'\rangle c^\dagger_{nk\sigma}c_{mk'\sigma'} \\ &=\sum_{n m k k' \sigma\sigma'}\langle\psi_{nk\sigma}|\psi_{mk\sigma'}\rangle\delta_{kk'}\delta_{\sigma\downarrow}\delta_{\sigma'\uparrow}c^\dagger_{nk\sigma}c_{mk'\sigma'} \\ &=\sum_{n m k}\langle\psi_{nk\downarrow}|\psi_{mk\uparrow}\rangle c^\dagger_{nk\downarrow}c_{mk\uparrow}. \tag*{(39)} \end{aligned}

Therefore,

ZG−=(⟨ψck↓∣ψvk↑⟩−⟨ψvk↓∣ψck↑⟩)=(xG−yG−).(40)Z_G^-= \begin{pmatrix} \langle\psi_{ck\downarrow}|\psi_{vk\uparrow}\rangle\\ -\langle\psi_{vk\downarrow}|\psi_{ck\uparrow}\rangle \end{pmatrix} = \begin{pmatrix} x_G^- \\ y_G^- \end{pmatrix}. \tag*{(40)}

will be a null-vector of K\mathbf{K} at Q=0Q=0 [67, 86], meaning an exact zero-frequency mode of Qi\mathbf{Q}_i (analogously for S+S^+). This requires (i) a self-consistent mean field, (ii) the same interaction in the calculation of the HF ground state and in the BSE kernel (so that the AA and BB matrices used in the BSE computation match with those stemming from the EOM), and (iii) a complete particle-hole space.

Since the BSE/TDA keeps only H(r,r)H^{(r,r)}, it contains the zero mode only if H(r,r)xG−=0H^{(r,r)}x_G^-=0, which requires yG−y_G^- to be equivalent to BSE/RPA (or H(r,a)H^{(r,a)} to be the zero matrix, see Table I as to why that is not the case). However, for a UHF ground state, ∥yG−∥2\lVert y_G^-\rVert^2 is precisely its spin contamination [84, 87, 88] as we will now show. Recalling Eq. (39), the amplitudes of S+=(S−)†S^+=(S^-)^\dagger acting on the ground state are given by

S+∣Φ0⟩=∑nmk⟨ψnk↑∣ψmk↓⟩cnk↑†cmk↓∣Φ0⟩.(41)S^+|\Phi_0\rangle=\sum_{n m k}\langle\psi_{nk\uparrow}|\psi_{mk\downarrow}\rangle c^\dagger_{nk\uparrow}c_{mk\downarrow}|\Phi_0\rangle. \tag*{(41)}

Since ∣Φ0⟩|\Phi_0\rangle is an eigenstate of SzS_z and S2=S−S++Sz(Sz+1)S^2=S^-S^++S_z(S_z+1), the norm of Eq. (41), which the orthonormality of the singly excited determinants reduces to a sum of squared overlaps, is the spin contamination as defined in the literature [84, 89]

∥yG−∥2=⟨Φ0∣S−S+∣Φ0⟩=⟨Φ0∣S2−Sz(Sz+1)∣Φ0⟩=⟨S2⟩UHF−⟨Sz(Sz+1)⟩UHF=Δ⟨S2⟩UHF.(42)\begin{aligned} \lVert y_G^-\rVert^2 &= \langle\Phi_0|S^-S^+|\Phi_0\rangle\\ &=\langle\Phi_0|S^2-S_z(S_z+1)|\Phi_0\rangle\\ &=\langle S^2\rangle_{\mathrm{UHF}}-\langle S_z(S_z+1)\rangle_{\mathrm{UHF}} \\ &=\Delta\langle S^2\rangle_{\mathrm{UHF}}. \tag*{(42)} \end{aligned}

Hence, a finite spin contamination means that, in general,

Δ⟨S2⟩≠0⟺yG−≠0⟺H(r,r)xG−=−H(r,a)yG−≠0.(43)\Delta\langle S^2\rangle\ne0\Longleftrightarrow y_G^-\ne0\Longleftrightarrow H^{(r,r)}x_G^-=-H^{(r,a)}y_G^-\ne0. \tag*{(43)}

so xˉG\bar{x}_{G} is not a zero mode of the BSE/TDA and its Goldstone mode is gapped, whereas the BSE/RPA keeps the exact zero. Focusing our attention on spin-flip excitations, it can be seen that exchange type integrals are zero (see Table I). In the ΔSz=−1\Delta S_z=-1 sector, xˉG\bar{x}_{G} lives on the pairs (v↑,c↓)(v \uparrow,c \downarrow) and yˉG\bar{y}_{G} on (v↓,c↑)(v \downarrow,c \uparrow), and couple through D~(r,a)≠0\widetilde{D}^{(r,a)} \ne0 while X~(r,a)=0\widetilde{X}^{(r,a)}=0 (see Table I). This coupling is what carries yˉG\bar{y}_{G} in Eq. (40) and restores the zero-energy Goldstone mode, as we will see ahead. Remarkably, while a small change is also visible in the higher energy bound magnons, as we will see ahead, working in the BSE/RPA leads to nearly identical excitons to those obtained within the BSE/TDA for the systems studied, as visible in Figure 8 and Figure 9.

Figure 8. Excitonic (red and black) and IPA (blue) contributions to the diagonal tensor elements of the linear optical conductivity for ferromagnetic graphene. As expected, only spin-conserving transitions are optically active for both contributions. Red and black lines correspond to excitons computed via BSE/RPA and BSE/TDA, respectively.

Wave functions of the Goldstone magnon and lowest bright exciton for ferromagnetic graphene

Figure 9. Wave function of the Goldstone magnon (first and second columns, computed via BSE/TDA and BSE/RPA respectively) and the lowest bright exciton (third and fourth columns, computed via BSE/TDA and BSE/RPA respectively) of ferromagnetic graphene for COM momentum QQ (top to bottom) corresponding to the high symmetry points Γ\Gamma, MM, KK. Whenever a state is degenerate (see Figure 4), wave functions of both degenerate states were summed. Highlighted points in the Brillouin zone are Γ\Gamma (black or white, for contrast), MM (blue), K/K′K/K' (green), with then darker green points representing the sampled points for finite QQ computations.

The symplectic norm of the zero mode is fixed by the magnetization [80, 90]

ZG−†ηZG−=∥xˉG∥2−∥yˉG∥2=⟨Φ0∣[S+,S−]∣Φ0⟩=2⟨Sz⟩.(44)Z_{G}^{-\dagger}\eta Z_{G}^{-}=\lVert\bar{x}_{G}\rVert^{2}-\lVert\bar{y}_{G}\rVert^{2}=\langle\Phi_{0}|[S^{+},S^{-}]|\Phi_{0}\rangle=2\langle S_{z}\rangle. \tag*{(44)}

In a ferromagnet ⟨Sz⟩≠0\langle S_{z}\rangle\ne0, giving one Goldstone mode with finite norm and quadratic dispersion (type B) [91], and the S+S^{+} zero mode has norm −2⟨Sz⟩-2\langle S_{z}\rangle and is its −ω-\omega partner. In an antiferromagnet ⟨Sz⟩=0\langle S_{z}\rangle=0, so the zero mode has zero norm and pairs with its conjugate coordinate, giving two linear modes (type A) [92, 93]. Both cases satisfy the counting rule nA+2nB=nBG=2n_{A}+2n_{B}=n_{\mathrm{BG}}=2 [85, 94, 95], with nB=1n_{B}=1 for the ferromagnet and nA=2n_{A}=2 for the antiferromagnet.

Ferromagnetic graphene

We first consider ferromagnetic graphene [65, 66] as a scenario where both excitons and magnons can be computed. Ferromagnetism in graphene appears when hydrogen atoms are adsorbed on the same sublattice[62, 65, 66] (vacancies also do the trick [96], although lattice reconstructions may diminish the effect[97]). The unit cell we consider is obtained by doubling the lattice vectors of the standard graphene lattice, as plotted in Figure 1 and drastically increasing the on-site energy of the leftmost site in the figure to mimic the effect of the vacancy or hydrogen atom bonded to the carbon atom below. The non-interacting band structure for t=1t=1 eV is then computed, giving the expected flat band at the Fermi level[96], as shown in Figure 2(a).

Figure 1. Unit cell considered for hydrogenated graphene system, with 8 Carbon atoms per unit cell. The dashed circle around the leftmost atomic site indicates the position of the added Hydrogen atom. Lattice vectors a1/2a_1/2 correspond to doubling the lattice vectors of the standard graphene lattice.

Figure 2. Single-particle non-interacting (left panel) and unrestricted Hartree-Fock quasi-particle (right panel) band structure for ferromagnetic hydrogenated graphene system with 8 Carbon atoms per unit cell. In the right hand side plot, solid lines correspond to an UHF calculation with U=4U = 4 eV and J=1.25J = 1.25 eV, while dashed lines have J=0J = 0 eV. Quasi-particle gaps between the three distinct band pairings relevant are highlighted by the transitions in question.

Preliminary analysis

Table I shows a complete dissection of how the different interaction terms in the Hamiltonian determine which terms in the self-energy and in the kernel are present or vanish. The first thing we learn is that the Fock term in the self-energy ΣX\Sigma^{X} vanishes if only local interactions are considered. This translates, in turn, into the vanishing of the direct term in the kernel either at the TDA or the RPA levels. As this term is responsible for the binding of the exciton, one concludes that excitons cannot form for the standard Hubbard model (something similar happens in the case of distinguishable particles for any interaction model, as previously discussed). The exchange term is non-zero, but does not contribute to the binding. Remarkably, when considering spin-flip excitations, the finite and the vanishing terms swap roles, leading to the binding of magnons with the standard Hubbard model and their disappearance when considering only non-local interactions.

For the calculations, we will generally consider all the 14 bands of the system in question (the vacancy-related bands are decoupled from the low energy spectrum). Sometimes, for illustration purposes, we will take only the 4 bands closest to the Fermi level as to allow for easier visualization of both magnon and exciton states. This corresponds to two bands of opposite spins full and two bands of opposite spin empty (see right panel of Figure 2). In this plot, these three distinct transitions are highlighted, together with their respective UHF bandgaps given by

E↑↓U=1.28 eV,E↑↑U=3.84 eV,E↓↓U=2.84 eV.(45)E_{\uparrow\downarrow}^{\mathrm{U}}=1.28\,\mathrm{eV},\quad E_{\uparrow\uparrow}^{\mathrm{U}}=3.84\,\mathrm{eV},\quad E_{\downarrow\downarrow}^{\mathrm{U}}=2.84\,\mathrm{eV}. \tag*{(45)}

For this plot, we have used U=4U=4 eV and J=1.25J=1.25 eV. We have verified our previous observations by performing calculations setting UU or JJ to zero independently on top of the same ground state discussed above. Setting U=0U=0, and keeping a finite value of JJ (in this case, 1.251.25 eV), we observe that magnons are completely lost and only continuum spin-flip excitations (i.e., Stoner excitations [98, 99]) remain in that energy range, while excitons remain largely unaffected. On the contrary, setting J=0J=0 and keeping a finite value for U=4U=4 eV, excitons are lost and only continuum spin-conserving excitations remain, while magnons remain untouched. This is all shown in Figure 3.

Effects of including only on-site versus off-site on the excitation energies

Figure 3. Effects of including only on-site versus off-site on the excitation energies in the BSE/TDA formalism. First six eigenvalues included in each spin sector, separated by the vertical dashed lines as ↑↓\uparrow\downarrow, ↓↓\downarrow\downarrow, and ↑↑\uparrow\uparrow. Horizontal dashed lines correspond to the bandgap for each of these sectors.

BSE/TDA with unrestricted Hartree-Fock ground state

Although UHF is a standard MF approach for magnetic systems, it suffers from spin contamination [87, 88, 100, 101, 102] as ∣Φ0⟩|\Phi_{0}\rangle is not an eigenstate of S2S^{2} (it remains an eigenstate of SzS_{z}). Its spin contamination Δ⟨S2⟩\Delta\langle S^{2}\rangle is given by Eq. (42), as we have shown above.

For now, let us temporarily ignore this spin contamination and proceed with the BSE/TDA computations knowing that a gap in the Goldstone mode will be present. One obtains the excitation dispersion as shown in Figure 4. Besides the (gapped) Goldstone mode (black line), one clearly observes two other bound magnons degenerate at Γ\Gamma (also in black, almost flat dispersion), usually called "optical" modes [58, 71, 72, 103, 104], followed by the Stoner continuum (grey lines). Looking further up in the energy series, we reach spin-conserving excitations in the form of excitons, first for spin-down transitions (blue lines) and then for spin-up transitions (red lines). The excitonic continuum and transitions corresponding to different band pairings (for instance, the spin-down band which touches one of the highlighted bands at Γ\Gamma, recall the right panel of Figure 2) are also clearly visible, filling the space between the two excitonic sectors.

Magnon and exciton dispersion for COM momentum Q

Figure 4. Magnon (black lines) and exciton (blue and red lines) dispersion for COM momentum QQ along the K-Γ\Gamma-M-K′'-Γ\Gamma-K path in the first Brillouin zone with BSE/TDA. Black lines highlight the magnon Goldstone mode and the optical magnons, grey lines are the Stoner continuum. Solid dots at Γ\Gamma in the exciton lines denotes an optically bright state. Excitons depicted in blue/red correspond to transitions associated with the band pairings denoted as E↓↓/E↑↑E_{\downarrow\downarrow}/E_{\uparrow\uparrow} in Figure 2, respectively.

Interaction term / bound statesOn-site interactionNon-local interaction (i≠ji \ne j)Non-local interaction (i≠ji \ne j)General Hamiltonian
Hartree term in Σ\Sigma≠0\ne0≠0\ne0≠0\ne0≠0\ne0
Fock term in Σ\Sigma0000≠0\ne0≠0\ne0
DD for spin-conserving excitations0000≠0\ne0≠0\ne0
XX for spin-conserving excitations≠0\ne0≠0\ne0≠0\ne0≠0\ne0
D~\widetilde{D} for spin-conserving excitations0000≠0\ne0≠0\ne0
X~\widetilde{X} for spin-conserving excitations≠0\ne0≠0\ne0≠0\ne0≠0\ne0
DD for spin-flip excitations≠0\ne0≠0\ne000≠0\ne0
XX for spin-flip excitations00000000
D~\widetilde{D} for spin-flip excitations≠0\ne0≠0\ne000≠0\ne0
X~\widetilde{X} for spin-flip excitations00000000
ExcitonNoYesYes
MagnonYesNoYes

Table 1. Contribution of on-site and non-local interactions to different terms of the BSE/TDA and BSE/RPA matrices and to the appearance of bound states.

Regarding the optically bright/dark nature of each state, magnons are optically dark as expected while excitons can be bright or dark depending on their dipole matrix elements. This is highlighted in Figure 4, where optically bright excitons are identified with a solid dot. Furthermore, in Figure 8 we plot the linear optical conductivity of this system, clearly observing both the spin selectivity of optical transitions and the highly localized low energy peaks, characteristic of an almost flat band structure.

BSE/RPA with UHF ground state

Recalling Secs. III B and III C, we will now show that BSE/RPA fully closes the spin-contamination gap which appears in the Goldstone mode. Including the antiresonant excitations and their coupling, the Goldstone mode becomes exactly gapless, as can be seen in Figure 5. The spin contamination of the UHF reference itself is unchanged, but it no longer gaps the Goldstone mode, as discussed in Eq. (42). Furthermore, while no change is observed in the Stoner continuum, the optical magnon modes also get shifted down in energy in similar fashion to the Goldstone mode. At small QQ the Goldstone mode is quadratic with dispersion

Magnon dispersion from UHF computed via BSE/TDA or BSE/RPA

Figure 5. Magnon dispersion from UHF computed via BSE/TDA (black lines) or BSE/RPA (red dashed lines) for COM momentum QQ along the K-Γ\Gamma-M-K′'-Γ\Gamma-K path in the first Brillouin zone.

ℏω(Q)≃2.307 eVA˚ ∣Q∣2,(46)\hbar\omega(\mathbf{Q}) \simeq2.307\ \mathrm{eV\mathring{A}}\,|\mathbf{Q}|^2, \tag*{(46)}

as expected from the magnetization-based classification from Eq. (44), against a gap at Γ\Gamma of 0.14 eV for BSE/TDA. Additionally, it is clear that the energy difference is not a pure scissor cut, being instead dependent on both QQ and on which state one is discussing.

Regarding the exciton states, we have omit their dispersion to allow for a clearer observation of the exactly-zero nature of the Goldstone magnon. However, we

still compare the (optically active) excitons in both BSE forms via the linear optical conductivity, as presented in Figure 8. Consistent with previous work in the literature [80], the effect of including anti-resonant excitations and their coupling to resonant excitations in the linear optical conductivity is negligible. Regarding dark excitons with ΔSz=1\Delta S_z = 1, as mentioned previously, these are outside of the range plotted due to the ≈5\approx5 eV energy difference between the bands involved in such transitions. For another comparison between the two forms of the BSE, we show-case the excitations’ kk-resolved state weights (analogous to the wave function for BSE/TDA) in Appendix B.

Antiferromagnetic Graphene

Having shown that the Goldstone mode can be correctly computed in a ferromagnet, we will now consider the case of antiferromagnetic graphene. Considering the same unit cell as before, namely the one obtained by doubling the lattice vectors of the standard graphene lattice, as shown in Figure 1, as well as the same model interaction as presented in Eq. (5), we now perform the UHF calculation for an antiferromagnetic ground state. Even for pristine graphene, this ground state appears when the on-site interaction strength reaches U=7U = 7 eV, with band structure shown in Figure 6. For this specific case, we keep J=0J = 0.

Figure 6. Unrestricted Hartree-Fock quasi-particle band structure for an antiferromagnetic graphene system with 8 Carbon atoms per unit cell.

With ⟨Sz⟩=0\langle S_z\rangle= 0, Eq. (44) gives two zero-norm zero modes at Γ\Gamma and the BSE/RPA Goldstone magnon is linear and twofold degenerate with dispersion

ℏω(Q)≃4.42 eVA˚ ∣Q∣,(47)\hbar\omega(\mathbf{Q}) \simeq4.42\ \mathrm{eV\mathring{A}}\,|\mathbf{Q}|, \tag*{(47)}

close to Γ\Gamma against a 1.91 eV gap in the BSE/TDA as shown in Figure 7. While in the ferromagnetic case the difference in magnon dispersion between BSE/TDA and BSE/RPA might seem, at first glance, like a simple downward shift in energy, here the difference is much more drastic. Besides the much larger energy difference between the two at Γ, the curves are also qualitatively distinct, with BSE/TDA failing to capture the linear dispersion of the Goldstone magnons and presenting a quadratic-like dispersion similar to the ferromagnetic case. BSE/RPA, on the contrary, accurately captures the Goldstone magnons’ linear dispersion and the two exact zero modes at Γ.

Figure 7. Magnon dispersion from UHF computed via BSE/TDA (black lines) or BSE/RPA (red dashed lines) for COM momentum QQ along the K-Γ-M-K′-Γ-K path in the first Brillouin zone.

ticeable when one discusses antiferromagnets, where the linear Goldstone mode is completely lost in the Tamm-Dancoff approximation, instead becoming quadratic and presenting a sizeable gap.

Moving on to excitons, we have verified that the Tamm-Dancoff approximation is very accurate for these states, matching the observations of previous works that the changes in the excitonic optical response are minimal. Furthermore, the connection between the existence of both excitons and magnons and the presence of onsite and non-local interactions, as well as their respective roles in the Bethe-Salpeter kernel, has been discussed and verified.

Conclusions

In this paper, we have shown that magnons and excitons can be computed on the same footing under the Bethe-Salpeter formalism while respecting the gapless Goldstone mode for magnons. This result rests on working beyond the Tamm-Dancoff approximation through the evaluation of the full Bethe-Salpeter equation with the same interaction in the kernel as that which was used in the computation of the starting mean-field ground state. This follows directly from the equation-of-motion derivation of the Bethe-Salpeter equation and the Goldstone theorem, and we have shown that keeping the antiresonant block and its coupling to the resonant block of the Bethe-Salpeter matrix is fundamental if one wishes to preserve its zero modes.

Following this procedure, we have shown that the vestigial Goldstone gap, present throughout the current literature and usually closed by a manual scissor cut in ferromagnets, closes fully and the exact zero-modes are recovered. Additionally, we have also shown that the correction to the magnon dispersion is not purely a scissor cut, being strongly dependent on which state one is analysing, on the centre-of-mass momentum QQ, and on the magnetic phase of the material. This is especially noticeable when one discusses antiferromagnets, where the linear Goldstone mode is completely lost in the Tamm-Dancoff approximation, instead becoming quadratic and presenting a sizeable gap.

Acknowledgments

We acknowledge financial support from MCIU/AEI/10.13039/501100011033 through Grants PID2022-141712NB-C21, PCI2026-177428-1, PID2025-174732NB-100, and CEX2023-001316-M within the María de Maeztu Program for Units of Excellence in R&D. We acknowledge the Comunidad de Madrid through the project “Disruptive 2D materials” (MAD2D-CM-UAM7) within the Recovery, Transformation and Resilience Plan. We acknowledge the European Union through the M.ERA-NET Joint Call 2025 (ALTMAG project) and the NextGenerationEU programme. We also acknowledge computational resources from CCC at UAM and the Spanish RES through projects FI-2025-1-12 and FI-2026-2-0015.

Author Contributions

Simran Chourasia and Maurício F. C. Martins Quintela contributed equally to this work. J. J. Palacios conceived and supervised the work. All authors contributed equally to writing the manuscript.

Distinguishable electrons

In this appendix, we outline the derivation of the BSE/TDA considering distinguishable electrons. Recalling the definitions of the main text

H=T+V,(48)H = T + V, \tag*{(48)}
T=∑l1,l2Tl1l2cl1†cl2=∑αεαaα†aα,(49)T = \sum_{l_1,l_2} T_{l_1 l_2} c^\dagger_{l_1} c_{l_2} = \sum_{\alpha} \varepsilon_{\alpha} a^\dagger_{\alpha} a_{\alpha}, \tag*{(49)}
V=12∑l1,l2,l3,l4Vl1l2l3l4cl1†cl2†cl4cl3,(50)V = \frac{1}{2} \sum_{l_1,l_2,l_3,l_4} V_{l_1 l_2 l_3 l_4} c^\dagger_{l_1} c^\dagger_{l_2} c_{l_4} c_{l_3}, \tag*{(50)}

the ground state is assumed as ∣Φ0⟩=∏εv≤EFav†∣0⟩|\Phi_0\rangle= \prod_{\varepsilon_v \le E_F} a^\dagger_v |0\rangle. Here, subscript vv and cc represent valence and conduction band state, respectively. In the e-h excitation basis, ∣v,c⟩=ac†av∣Φ0⟩|v,c\rangle= a^\dagger_c a_v |\Phi_0\rangle, the Hamiltonian becomes

⟨v,c∣H∣v′,c′⟩=δcc′δvv′[⟨Φ0∣H∣Φ0⟩+εc−εv]−δcc′∑εα≤EF[⟨εv′,εα∣V∣εv,εα⟩−⟨εα,εv′∣V∣εv,εα⟩]+δvv′∑εα≤EF[⟨εc,εα∣V∣εc′,εα⟩−⟨εc,εα∣V∣εα,εc′⟩]−⟨εc,εv′∣V∣εc′,εv⟩+⟨εc,εv′∣V∣εv,εc′⟩.(51)\begin{aligned} \langle v,c|H|v',c'\rangle={}& \delta_{cc'}\delta_{vv'}\left[\langle\Phi_0|H|\Phi_0\rangle+\varepsilon_c-\varepsilon_v\right] \\ &-\delta_{cc'}\sum_{\varepsilon_\alpha\le E_F}\left[\langle\varepsilon_{v'},\varepsilon_\alpha|V|\varepsilon_v,\varepsilon_\alpha\rangle-\langle\varepsilon_\alpha,\varepsilon_{v'}|V|\varepsilon_v,\varepsilon_\alpha\rangle\right] \\ &+\delta_{vv'}\sum_{\varepsilon_\alpha\le E_F}\left[\langle\varepsilon_c,\varepsilon_\alpha|V|\varepsilon_{c'},\varepsilon_\alpha\rangle-\langle\varepsilon_c,\varepsilon_\alpha|V|\varepsilon_\alpha,\varepsilon_{c'}\rangle\right] \\ &-\langle\varepsilon_c,\varepsilon_{v'}|V|\varepsilon_{c'},\varepsilon_v\rangle+\langle\varepsilon_c,\varepsilon_{v'}|V|\varepsilon_v,\varepsilon_{c'}\rangle. \tag*{(51)} \end{aligned}

A system with non-identical interacting electrons will only have Hartree interaction. The Fock interaction term

appears because all the electrons are identical. Since the Fock term carries all the information of many-body correlations, it is important to know which terms of the kernel arise from it. Looking at TDA matrix elements for non-identical electrons will help us identify these terms. To study non-identical electrons, let us define the creation operator cl†(x)c^\dagger_l(x) which creates the electron with label "x" in the electronic state ll, and cl(x)c_l(x) is the corresponding annihilation operator. Now, the Hamiltonian for non-identical (NI) electrons with two-body interaction has the following non-interacting and interaction terms

HNI=TNI+VNI,(52)H^{\mathrm{NI}} = T^{\mathrm{NI}} + V^{\mathrm{NI}}, \tag*{(52)}
TNI=∑x1∑l1,l2Tl1l2cl1†(x1)cl2(x1)=∑x1∑αεαaα†(x1)aα(x1),(53)\begin{aligned} T^{\mathrm{NI}} &= \sum_{x_1}\sum_{l_1,l_2} T_{l_1 l_2} c^\dagger_{l_1}(x_1)c_{l_2}(x_1) \\ &= \sum_{x_1}\sum_{\alpha}\varepsilon_\alpha a^\dagger_\alpha(x_1)a_\alpha(x_1), \tag*{(53)} \end{aligned}
VNI=12∑x1,x2∑l1,l2,l3,l4(1−δx1x2)Vl1l2l3l4(x1,x2)cl1†(x1)cl2†(x2)cl4(x2)cl3(x1).(54)V^{\mathrm{NI}} = \frac{1}{2}\sum_{x_1,x_2}\sum_{l_1,l_2,l_3,l_4}\left(1-\delta_{x_1x_2}\right)V_{l_1l_2l_3l_4}(x_1,x_2)c^\dagger_{l_1}(x_1)c^\dagger_{l_2}(x_2)c_{l_4}(x_2)c_{l_3}(x_1). \tag*{(54)}

The N-electron ground takes the form

∣Φ0NI⟩=∣εv1(p1),εv2(p2),…,εvN(pN)⟩=av1†(p1)av2†(p2)…avN†(pN)∣0⟩,(55)|\Phi^{\mathrm{NI}}_0\rangle= |\varepsilon_{v_1}(p_1),\varepsilon_{v_2}(p_2),\ldots,\varepsilon_{v_N}(p_N)\rangle = a^\dagger_{v_1}(p_1)a^\dagger_{v_2}(p_2)\ldots a^\dagger_{v_N}(p_N)|0\rangle, \tag*{(55)}

and the Hamiltonian in the e-h excitation basis,

∣vi(pi),c(η)⟩=ac†(η)avi(pi)∣Φ0⟩,\left|v_i(p_i),c(\eta)\right\rangle=a^\dagger_c(\eta)a_{v_i}(p_i)\left|\Phi_0\right\rangle,
⟨vi(pi),c(η)|HNI|vj(pj),c′(η′)⟩=δcc′δηη′δvivjδpipj[⟨Φ0NI|HNI|Φ0NI⟩+εc−εvi]−δcc′δηη′δvivjδpipj∑r=1N(1−δvrviδprpi)⟨εvr(pr),εvi(pi)|VNI|εvr(pr),εvi(pi)⟩+δηη′δvivjδpipj∑r=1N(1−δprη)⟨εvr(pr),εc(η)|VNI|εvr(pr),εc′(η)⟩−δηη′δvivjδpipj(1−δpiη)⟨εvi(pi),εc(η)|VNI|εvi(pi),εc′(η)⟩+(1−δvjviδpjpi)δη′pjδηpi⟨εvj(pj),εc(pi)|VNI|εc′(pj),εvi(pi)⟩.(56)\begin{aligned} \left\langle v_i(p_i),c(\eta)\middle|H^{\mathrm{NI}}\middle|v_j(p_j),c'(\eta')\right\rangle ={}&\delta_{cc'}\delta_{\eta\eta'}\delta_{v_i v_j}\delta_{p_i p_j} \left[\left\langle\Phi_0^{\mathrm{NI}}\middle|H^{\mathrm{NI}}\middle|\Phi_0^{\mathrm{NI}}\right\rangle+\varepsilon_c-\varepsilon_{v_i}\right]\\ &-\delta_{cc'}\delta_{\eta\eta'}\delta_{v_i v_j}\delta_{p_i p_j} \sum_{r=1}^{N}\left(1-\delta_{v_r v_i}\delta_{p_r p_i}\right) \left\langle\varepsilon_{v_r}(p_r),\varepsilon_{v_i}(p_i)\middle|V^{\mathrm{NI}}\middle|\varepsilon_{v_r}(p_r),\varepsilon_{v_i}(p_i)\right\rangle\\ &+\delta_{\eta\eta'}\delta_{v_i v_j}\delta_{p_i p_j} \sum_{r=1}^{N}\left(1-\delta_{p_r\eta}\right) \left\langle\varepsilon_{v_r}(p_r),\varepsilon_c(\eta)\middle|V^{\mathrm{NI}}\middle|\varepsilon_{v_r}(p_r),\varepsilon_{c'}(\eta)\right\rangle\\ &-\delta_{\eta\eta'}\delta_{v_i v_j}\delta_{p_i p_j}\left(1-\delta_{p_i\eta}\right) \left\langle\varepsilon_{v_i}(p_i),\varepsilon_c(\eta)\middle|V^{\mathrm{NI}}\middle|\varepsilon_{v_i}(p_i),\varepsilon_{c'}(\eta)\right\rangle\\ &+\left(1-\delta_{v_jv_i}\delta_{p_jp_i}\right)\delta_{\eta'p_j}\delta_{\eta p_i} \left\langle\varepsilon_{v_j}(p_j),\varepsilon_c(p_i)\middle|V^{\mathrm{NI}}\middle|\varepsilon_{c'}(p_j),\varepsilon_{v_i}(p_i)\right\rangle. \tag*{(56)} \end{aligned}

To compare the above expression with Eq. (51), let us suppress the particle labels and replace vi→vv_i \to v, vj→v′v_j \to v', and ∑r=1N→∑εα≤EF\sum_{r=1}^{N} \to\sum_{\varepsilon_\alpha\le E_F}. The resulting expression is

⟨v,c|HNI|v′,c′⟩=δcc′δvv′[⟨Φ0NI|HNI|Φ0NI⟩+εc−εv]−δcc′δvv′∑εα≤EF(1−δαv)⟨εα,εv|VNI|εα,εv⟩+δvv′∑εα≤EF⟨εα,εc|VNI|εα,εc′⟩−δvv′⟨εc,εv|VNI|εc′,εv⟩+(1−δv′v)⟨εc,εv′|VNI|εv,εc′⟩,(57)\begin{aligned} \left\langle v,c\middle|H^{\mathrm{NI}}\middle|v',c'\right\rangle ={}&\delta_{cc'}\delta_{vv'}\left[\left\langle\Phi_0^{\mathrm{NI}}\middle|H^{\mathrm{NI}}\middle|\Phi_0^{\mathrm{NI}}\right\rangle+\varepsilon_c-\varepsilon_v\right]\\ &-\delta_{cc'}\delta_{vv'}\sum_{\varepsilon_\alpha\le E_F}\left(1-\delta_{\alpha v}\right)\left\langle\varepsilon_\alpha,\varepsilon_v\middle|V^{\mathrm{NI}}\middle|\varepsilon_\alpha,\varepsilon_v\right\rangle\\ &+\delta_{vv'}\sum_{\varepsilon_\alpha\le E_F}\left\langle\varepsilon_\alpha,\varepsilon_c\middle|V^{\mathrm{NI}}\middle|\varepsilon_\alpha,\varepsilon_{c'}\right\rangle\\ &-\delta_{vv'}\left\langle\varepsilon_c,\varepsilon_v\middle|V^{\mathrm{NI}}\middle|\varepsilon_{c'},\varepsilon_v\right\rangle +\left(1-\delta_{v'v}\right)\left\langle\varepsilon_c,\varepsilon_{v'}\middle|V^{\mathrm{NI}}\middle|\varepsilon_v,\varepsilon_{c'}\right\rangle, \tag*{(57)} \end{aligned}

and it takes the form of Eq. (12) for a crystal.

Wave functions

In this appendix, we provide a visual comparison between the BSE/TDA and BSE/RPA excitation kk-resolved state weight in the symplectic (η\eta) norm, which we will now designate as “wave function” for simplicity. From Eqs. (4) and (44), it reads

ρQ(k)=ψcvQ†(k)ηψcvQ(k)=∑vc(∣AcvQ(k)∣2−∣BcvQ(k)∣2),(58)\rho_Q(k)=\psi_{cv}^{Q\dagger}(k)\eta\psi_{cv}^{Q}(k) =\sum_{vc}\left(\left|A_{cv}^{Q}(k)\right|^2-\left|B_{cv}^{Q}(k)\right|^2\right), \tag*{(58)}

with ψQ(k)=(AQ(k),BQ(k))T\psi^Q(k)=\left(A^Q(k),B^Q(k)\right)^T. This quantity is defined analogously to the BSE/TDA case [24], but now taking advantage of the symplectic (η\eta) normalization of the BSE/RPA eigenstates. For B=0B=0 it reduces to the BSE/TDA weight ∑vc∣AcvQ(k)∣2\sum_{vc}\left|A_{cv}^{Q}(k)\right|^2, the probability of finding the electron-hole pair at kk.

Computing these, one observes that the two different forms of the BSE yield very similar wave functions for ferromagnetic graphene, as can be seen in Figure 9. However, some differences can still be seen, especially in the Goldstone magnon at Q=0Q=0 where the RPA wave function much more clearly showcases the expected fully flat character of a collective spin flip. The first exciton is essentially indistinguishable, matching the expectations set by the essentially identical optical conductivity between the two forms of the BSE (see Figure 8).

† [email protected], corresponding author

References

References

  1. [1]K. Hao, R. Schreiner, A. Kindseth, and A. A. High, “Optically controllable magnetism in atomically thin semiconductors,” Science Advances, vol. 8, no. 39, p. eabq7650, 2022.DOI
  2. [2]G. Shi, N. Huang, J. Qiao, X. Zhang, F. Hu, H. Hu, X. Zhang, and J. Shang, “Recent Progress in Two-Dimensional Magnetic Materials,” Nanomaterials, vol. 14, p. 1759, Jan. 2024.DOI
  3. [3]J. Phillips, J. C. G. Henriquez, J. Fernández-Rossier, and A. T. Costa, “Prediction of room-temperature two-dimensional π-electron half-metallic ferrimagnets,” 2026.
  4. [4]B. Huang, G. Clark, E. Navarro-Moratalla, D. R. Klein, R. Cheng, K. L. Seyler, D. Zhong, E. Schmidgall, M. A. McGuire, D. H. Cobden, W. Yao, D. Xiao, P. Jarillo-Herrero, and X. Xu, “Layer-dependent ferromagnetism in a van der waals crystal down to the monolayer limit,” Nature, vol. 546, no. 7657, pp. 270–273, 2017.DOI
  5. [5]C. Gong, L. Li, Z. Li, H. Ji, A. Stern, Y. Xia, T. Cao, W. Bao, C. Wang, Y. Wang, Z. Q. Qiu, R. J. Cava, S. G. Louie, J. Xia, and X. Zhang, “Discovery of intrinsic ferromagnetism in two-dimensional van der waals crystals,” Nature, vol. 546, no. 7657, pp. 265–269, 2017.
  6. [6]K. Lee, A. H. Dismukes, E. J. Telford, R. A. Wiscons, J. Wang, X. Xu, C. Nuckolls, C. R. Dean, X. Roy, and X. Zhu, “Magnetic order and symmetry in the 2d semiconductor CrSBr,” Nano Letters, vol. 21, no. 8, pp. 3511–3517, 2021.DOI
  7. [7]A. Avsar, C.-Y. Cheon, M. Pizzocchero, M. Tripathi, A. Ciarrocchi, O. V. Yazyev, and A. Kis, “Probing magnetism in atomically thin semiconducting ptse₂,” Nature Communications, vol. 11, p. 4806, Sep 2020.DOI
  8. [8]P. C. Adak, F. Dirnberger, S. Acharya, A. Kamra, X. Xu, and V. M. Menon, “Excitons in van der waals magnetic materials,” Nature Materials, Jul 2026.
  9. [9]K. F. Mak, C. Lee, J. Hone, J. Shan, and T. F. Heinz, “Atomically thin MoS₂ A new direct-gap semiconductor,” Phys. Rev. Lett., vol. 105, p. 136805, Sep 2010.
  10. [10]Y. Li, A. Chernikov, X. Zhang, A. Rigosi, H. M. Hill, A. M. van der Zande, D. A. Chenet, E.-M. Shih, J. Hone, and T. F. Heinz, “Measurement of the optical dielectric function of monolayer transition-metal dichalcogenides: MoS₂, MoSe₂, WS₂, and WSe₂,” Phys. Rev. B, vol. 90, p. 205422, Nov 2014.DOI
  11. [11]Y. Sun, F. Meng, S. Ke, K. Xu, H. Zhang, A. Soll, Z. Sofer, A. Majumdar, R. Ramesh, J. B. Neaton, J. Yao, and J. Ornstein, "Electron-magnon coupling at the interface of a "twin-twisted" antiferromagnet, 2025.
  12. [12]Z. A. Iakovlev, A. Kamra, and M. M. Glazov, “Boltzmann transport theory of magnon-exciton drag,” 2025.DOI
  13. [13]B. Datta, P. C. Adak, S. Yu, A. Valiyaparamil Dharampalan, S. J. Hall, A. Vakulenko, F. Komis-sarenko, E. Kurganov, J. Quan, W. Wang, K. Mosina, Z. Sofer, D. Pashov, M. van Schilfgaarde, S. Acharya, A. Kamra, M. Y. Sfeir, A. Alù, A. B. Khaniakev, and V. M. Menon, “Magnon-mediated exciton-exciton interaction in a van der waals antiferromagnet,” Nature Materials, vol. 24, pp. 1027–1033, Jul 2025.
  14. [14]A. Ebrahimian, F. J. García-Vidal, and J. J. Palacios, “Magnetic order tuning of excitons in the magnetic semiconductor crcl₃ through strain,” 2025.arxiv.org/abs/2509.20761
  15. [15]S. Gupta and S. S.-L. Zhang, “Roles of electron-magnon cross diffusion in unidirectional magnetoresistance of metallic magnetic bilayers,” Phys. Rev. B, vol. 113, p. 094434, Mar 2026.DOI
  16. [16]F. Dirnberger, S. Terres, Z. A. Iakovlev, K. Mosina, Z. Sofer, A. Kamra, M. M. Glazov, and A. Chernikov, “Exciton transport driven by spin excitations in an antiferromagnet,” Nature Nanotechnology, vol. 21, pp. 65–70, Jan 2026.
  17. [17]Z. Wang, X.-X. Zhang, Y. Shiomi, T.-h. Arima, N. Nagaosa, Y. Tokura, and N. Ogawa, “Exciton-magnon splitting in the van der waals antiferromagnet mnps₃ unveiled by second-harmonic generation,” Phys. Rev. Res., vol. 5, p. L042032, Nov 2023.DOI
  18. [18]J. Ma, X. Wang, Y. Gong, C. Hu, Q. Wang, K. Feng, Z. Lin, T. Ishihara, N. Fang, X. Yin, S. Zhang, Z. Chen, X. Liu, X. Cui, and X. Zhang, “Excitonic negative refraction mediated by magnetic orders,” Nature Nanotechnology, vol. 21, pp. 374–379, Mar 2026.
  19. [19]H. Zhang, N. Nilforoushan, C. Weidgans, T. Inzenhofer, M. Liebich, J. Riepl, J. Hirschmann, I. Gronwald, K. Mosina, Z. Sofer, R. Tyagi, J. Wilhelm, F. Mooshammer, F. Dirnberger, and R. Huber, “Magnetic control of an exciton-polariton condensate in a van der waals magnet,” 2026.DOI
  20. [20]O. Mehio, X. Li, H. Ning, Z. Lenarčič, Y. Han, M. Buchhold, Z. Porter, N. J. Laurita, S. D. Wilson, and D. Hsieh, “A hubbard exciton fluid in a photo-doped antiferromagnetic mott insulator,” Nature Physics, vol. 19, pp. 1876–1882, Dec 2023.DOI
  21. [21]J. Deslippe, G. Samodszide, D. A. Strubbe, M. Jain, M. L. Cohen, and S. G. Louie, “Berkeleygw: A massively parallel computer package for the calculation of the quasiparticle and optical properties of materials and nanostructures,” Computer Physics Communications, vol. 183, no. 6, pp. 1269–1289, 2012.DOI
  22. [22]D. Sangalli, A. Ferretti, H. Miranda, C. Attaccalite, I. Marri, E. Cannuccia, P. Melo, M. Marsili, F. Paleari, A. Marrazzo, G. Prandini, P. Bonfà, M. O. Atambo, F. Affinito, M. Palummo, A. Molina-Sánchez, C. Hogan, M. Gruning, D. Varsano, and A. Marini, “Many-body perturbation theory calculations using the yambo code,” Journal of Physics: Condensed Matter, vol. 31, p. 325902, may 2019.DOI
  23. [23]G. Cistaro, M. Malakhov, J. J. Esteve-Paredes, A. J. Uría-Álvarez, R. E. F. Silva, F. Martín, J. J. Palacios, and A. Picón, “Theoretical approach for electron dynamics and ultrafast spectroscopy (edus),” Journal of Chemical Theory and Computation, vol. 19, no. 1, pp. 333–348, 2023.DOI
  24. [24]A. J. Uría-Álvarez, J. J. Esteve-Paredes, M. A. García-Blázquez, and J. J. Palacios, “Efficient computation of optical excitations in two-dimensional materials with the Xatu code,” Computer Physics Communications, vol. 295, p. 109001, 2024.
  25. [25]M. A. García-Blázquez and J. J. Palacios, “First-principles excitons in periodic systems with gaussian density fitting and ewald potential functions,” Phys. Rev. Res., vol. 7, p. 013156, Feb 2025.
  26. [26]M. Rava-Moreno, A. Burchi, N. A. Dasch, N. Farahani, I. Gonzalez Oliva, A. Gulans, S. Hong, M. Hossain, H. Kleine, M. Kuban, S. Lubeck, B. Maurer, P. Pavone, F. Peschel, D. Popova-Gorelova, L. Qiao, E. Richter, S. Rigamonti, R. R. Pela, M. Schébek, K. Sinha, D. T. Speckhard, J. Stutz, S. Tillack, D. Tumakov, J. Užulis, M. Voiculescu, C. Vona, M. Yang, and C. Draxl, “An exciting approach to theoretical spectroscopy,” Advanced Science, vol. n/a, no. n/a, p. e76167, 2026.
  27. [27]S. V. B. Jensen, L. B. Madsen, A. Rubio, and N. Tangcogne-Dejean, “High-harmonic spectroscopy of strongly bound excitons in solids,” Phys. Rev. A, vol. 109, p. 063104, Jun 2024.
  28. [28]D. Tumakov and D. Popova-Gorelova, “Real-time exciton dynamics in two-dimensional materials under ultrashort laser pulses,” Phys. Rev. B, vol. 114, p. 045127, Jul 2026.
  29. [29]X. Blase, I. Duchemin, D. Jacquemin, and P.-F. Loos, “The bethe-salpeter equation formalism: From physics to chemistry,” The Journal of Physical Chemistry Letters, vol. 11, pp. 7371–7382, 08 2020.DOI
  30. [30]D. L. Esteras, A. Rybakov, A. M. Ruiz, and J. J. Baldoví, “Magnon straintronics in the 2d van der waals ferromagnet crsbr from first-principles,” Nano Letters, vol. 22, pp. 8771–8778, 09 2022.DOI
  31. [31]D. L. Esteras and J. J. Baldoví, “Strain engineering of magnetic exchange and topological magnons in chromium trihalides from first-principles,” Materials Today Electronics, vol. 6, p. 100072, 2023.DOI
  32. [32]S. Toth and B. B. Lake, “Linear spin wave theory for single-q incommensurate magnetic structures,” Journal of Physics: Condensed Matter, vol. 27, p. 166002, mar 2015.DOI
  33. [33]F. L. Durhuus, T. Skovhus, and T. Olsen, “Plane wave implementation of the magnetic force theorem for magnetic exchange constants: application to bulk fe, co and ni,” Journal of Physics: Condensed Matter, vol. 35, p. 105802, jan 2023.
  34. [34]Y.-H. Shen, Y. Liu, and M. Luo, “Electrical reversal of the magnon thermal hall coefficient in a van der waals bilayer antiferromagnet,” Phys. Rev. B, vol. 110, p. 045423, Jul 2024.DOI
  35. [35]B. Kaman, J. Lim, Y. Liu, and A. Hoffmann, “Emulating 2d materials with magnons,” Phys. Rev. X, vol. 16, p. 011034, Feb 2026.
  36. [36]M. Amini, T. V. C. Antão, L. Jing, Z. Wang, A. Karjasilta, R. Drost, S. Kezilebieke, J. L. Lado, A. O. Fumega, and P. Liljeroth, “Observation of electromagnons in a monolayer multiferroic,” 2025.arxiv.org/abs/2510.08253
  37. [37]X. He, N. Helbig, M. J. Verstraete, and E. Bousquet, “Tb2j: A python package for computing magnetic interaction parameters,” Computer Physics Communications, vol. 264, p. 107938, 2021.DOI
  38. [38]P. Buczek, Spin dynamics of complex itinerant magnets. PhD thesis, Martin-Luther-Universität Halle-Wittenberg, Halle-Wittenberg, 2009.DOI
  39. [39]S. Y. Savrasov, “Linear response calculations of spin fluctuations,” Phys. Rev. Lett., vol. 81, pp. 2570–2573, Sep 1998.
  40. [40]P. Buczek, A. Ernst, and L. M. Sandratskii, “Different dimensionality trends in the landau damping of magnons in iron, cobalt, and nickel: Time-dependent density functional study,” Phys. Rev. B, vol. 84, p. 174418, Nov 2011.DOI
  41. [41]P. Buczek, A. Ernst, P. Bruno, and L. M. Sandratskii, “Energies and lifetimes of magnons in complex ferromagnets: A first-principle study of heusler alloys,” Phys. Rev. Lett., vol. 102, p. 247206, Jun 2009.pubmed.ncbi.nlm.nih.gov/19659045
  42. [42]S. Lounis, A. T. Costa, R. B. Muniz, and D. L. Mills, “Dynamical magnetic excitations of nanostructures from first principles,” Phys. Rev. Lett., vol. 105, p. 187205, Oct 2010.DOI
  43. [43]S. Lounis, A. T. Costa, R. B. Muniz, and D. L. Mills, “Theory of local dynamical magnetic susceptibilities from the korringa-kohn-rostoker green function method,” Phys. Rev. B, vol. 83, p. 035109, Jan 2011.DOI
  44. [44]S. Lounis, M. dos Santos Dias, and B. Schweflinghaus, “Transverse dynamical magnetic susceptibilities from regular static density functional theory: Evaluation of damping and g shifts of spin excitations,” Phys. Rev. B, vol. 91, p. 104420, Mar 2015.DOI
  45. [45]M. dos Santos Dias, B. Schweflinghaus, S. Blügel, and S. Lounis, “Relativistic dynamical spin excitations of [46] B. Rousseau, A. Eiguren, and A. Bergara, “Efficient computation of magnon dispersions within time-dependent density functional theory using maximally localized wannier functions,” Phys. Rev. B, vol. 85, p. 054305, Feb 2012.
  46. [47]T. Skovhus and T. Olsen, “Magnons in antiferromagnetic bcc cr and cr₂o₃ from time-dependent density functional theory,” Phys. Rev. B, vol. 106, p. 085131, Aug 2022.DOI
  47. [48]T. Skovhus and T. Olsen, “Dynamic transverse magnetic susceptibility in the projector augmented-wave method: Application to fe, ni, and co,” Phys. Rev. B, vol. 103, p. 245110, Jun 2021.DOI
  48. [49]T. Gorni, I. Timrov, and S. Baroni, “Spin dynamics from time-dependent density functional perturbation theory,” The European Physical Journal B, vol. 91, no. 10, p. 249, 2018. arXiv:1804.06573 [cond-mat.mtrl-sci].DOI
  49. [50]N. Tancogne-Dejean, F. G. Eich, and A. Rubio, “Time-dependent magnons from first principles,” 2020. Published in J. Chem. Theory Comput. 16, 1007 (2020).
  50. [51]M. C. T. D. Muller, C. Friedrich, and S. Blügel, “Acoustic magnons in the long-wavelength limit: Investigating the goldstone violation in many-body perturbation theory,” Physical Review B, vol. 94, no. 6, p. 064433, 2016.DOI
  51. [52]K. Karlsson and F. Aryasetiawan, “A many-body approach to spin-wave excitations in itinerant magnetic systems,” Journal of Physics: Condensed Matter, vol. 12, p. 7617, aug 2000.DOI
  52. [53]T. Kotani and M. van Schilfgaarde, “Spin wave dispersion based on the quasiparticle self-consistent gw method: Nio, mno and α-mno,” Journal of Physics: Condensed Matter, vol. 20, p. 295214, jul 2008.DOI
  53. [54]E. Şaşıoğlu, A. Schindlmayr, C. Friedrich, F. Freimuth, and S. Blügel, “Wannier-function approach to spin excitations in solids,” Phys. Rev. B, vol. 81, p. 054434, Feb 2010.DOI
  54. [55]K. B. Le, A. Eschumbre-Kučukalić, H.-Y. Chen, I. Maliyov, Y. Luo, J.-J. Zhou, D. Sangalla, A. Molina-Sánchez, and M. Bernardi, “Magnon-phonon interactions from first principles,” Phys. Rev. B, vol. 112, p. L180403, Nov 2025.arxiv.org/abs/2502.05385
  55. [56]T. Olsen, “Unified treatment of magnons and excitons in monolayer cri₃ from many-body perturbation theory,” Phys. Rev. Lett., vol. 127, p. 166402, Oct 2021.DOI
  56. [57]L. Ke and M. I. Katsnelson, “Electron correlation effects on exchange interactions and spin excitations in 2d van der waals materials,” npj Computational Materials, vol. 7, no. 1, p. 4, 2021.
  57. [58]A. Eschumbre-Kučukalić, K. B. Le, A. García-Cristóbal, M. Bernardi, D. Sangalli, and A. Molina-Sánchez, “Magnons in chromium trilayers calculated with the ab initio bethe-salpeter equation,” Phys. Rev. B, vol. 112, p. 184412, Nov. 2025.
  58. [59]W. Beida, E. Şaşıoğlu, C. Friedrich, G. Bihlmayer, Y. Mokrousov, and S. Blügel, “Chiral split magnons in metallic g-wave altermagnets: insights from many-body perturbation theory,” npj Quantum Materials, vol. 10, p. 97, Sep 2025.arxiv.org/abs/2505.08103
  59. [60]O. V. Yazyev, “Emergence of magnetism in graphene materials and nanostructures,” Reports on Progress in Physics, vol. 73, p. 056501, apr 2010.DOI
  60. [61]W. Han, R. K. Kawakami, M. Gmitra, and J. Fabian, “Graphene spintronics,” Nature Nanotechnology, vol. 9, pp. 794–807, Oct 2014.DOI
  61. [62]D. Soriano, F. Muñoz Rojas, J. Fernández-Rossier, and J. J. Palacios, “Hydrogenated graphene nanoribbons for spintronics,” Phys. Rev. B, vol. 81, p. 165409, Apr 2010.DOI
  62. [63]S. M. Hollen and J. A. Gupta, “Painting magnetism on a canvas of graphene,” Science, vol. 352, no. 6284, pp. 415–416, 2016.DOI
  63. [64]S. Tjung, S. Hollen, G. Gambrel, N. Santagata, E. Johnston-Halperin, and J. Gupta, “Crystalline hydrogenation of graphene by scanning tunneling microscope tip-induced field dissociation of h2,” Carbon, vol. 124, pp. 97–104, 2017.DOI
  64. [65]H. González-Herrero, J. M. Gómez-Rodríguez, P. Mallet, M. Moaied, J. J. Palacios, C. Salgado, M. M. Ugeda, J.-Y. Veuillen, F. Yndurain, and A. I. Bihrugeha, “Atomic-scale control of graphene magnetism by using hydrogen atoms,” Science, vol. 352, no. 6284, pp. 437–441, 2016.DOI
  65. [66]B. Viña-Bausà, M. A. García-Blázquez, S. Chourasia, R. Carrasco, D. Expósito, I. Brihuega, and J. J. Palacios, “Building Unconventional Magnetic Phases on Graphene by H Atom Manipulation: From Alternating Magnets to Lieb Ferrimagnets,” Nano Letters, vol. 25, pp. 11554–11561, July 2025.DOI
  66. [67]D. Thouless, “Stability conditions and nuclear rotations in the hartree-fock theory,” Nuclear Physics, vol. 21, pp. 225–232, 1960.DOI
  67. [68]M. E. CASIDA, Time-Dependent Density Functional Response Theory for Molecules, vol. Volume 1 of Recent Advances in Computational Chemistry, pp. 155–192. WORLD SCIENTIFIC, Nov 1995. 0.DOI
  68. [69]G. Co’, “Introducing the random phase approximation theory,” Universe, vol. 9, no. 3, 2023.arxiv.org/abs/2303.05801
  69. [70]J. F. Cooke, “Neutron scattering from itinerant-electron ferromagnets,” Phys. Rev. B, vol. 7, pp. 1108–1116, Feb 1973.DOI
  70. [71]J. F. Cooke, J. A. Blackman, and T. Morgan, “New interpretation of spin-wave behavior in nickel,” Phys. Rev. Lett., vol. 54, pp. 718–721, Feb 1985.pubmed.ncbi.nlm.nih.gov/10031598
  71. [72]J. F. Cooke, J. W. Lynn, and H. L. Davis, “Calculations of the dynamic susceptibility of nickel and iron,” Phys. Rev. B, vol. 21, pp. 4118–4131, May 1980.DOI
  72. [73]P. Buczek, A. Ernst, and L. M. Sandratskii, “Interface electronic complexes and landau damping of magnons in ultrathin magnets,” Phys. Rev. Lett., vol. 106, p. 157204, Apr 2011.pubmed.ncbi.nlm.nih.gov/21568609
  73. [74]M. van Schilfgaarde, T. Kotani, and S. Faleev, “Quasiparticle self-consistent gw theory,” Phys. Rev. Lett., vol. 96, p. 226402, Jun 2006.
  74. [75]S. V. Faleev, M. van Schilfgaarde, and T. Kotani, “All-electron self-consistent gw approximation: Application to si, mno, and nio,” Phys. Rev. Lett., vol. 93, p. 126406, Sep 2004.pubmed.ncbi.nlm.nih.gov/15447292
  75. [76]Y. Cui, I. W. Bulik, C. A. Jiménez-Hoyos, T. M. Henderson, and G. E. Scuseria, “Proper and improper zero energy modes in hartree-fock theory and their relevance for symmetry breaking and restoration,” The Journal of Chemical Physics, vol. 139, p. 154107, 10 2013.DOI
  76. [77]M. E. Casida, C. Jamorski, F. Bohr, J. Guan, and D. R. Salahub, Optical Properties from Density-Functional Theory, ch. 8, pp. 145–163. ACS Publications, 1996.DOI
  77. [79]A. Drew and M. Head-Gordon, “Single-reference ab initio methods for the calculation of excited states of large molecules,” Chemical Reviews, vol. 105, no. 11, pp. 4009–4037, 2005.
  78. [80]T. Sander, E. Maggio, and G. Kresse, “Beyond the Tamm-Dancoff approximation for extended systems using exact diagonalization,” Physical Review B, vol. 92, p. 045209, July 2015.
  79. [81]A. J. Uría-Álvarez, “Optical excitations and disorder in two-dimensional topological insulators,” 2025.
  80. [82]J. Goldstone, A. Salam, and S. Weinberg, “Broken Symmetries,” Physical Review, vol. 127, pp. 965–970, Aug. 1962.
  81. [83]P. Ring and P. Schuck, The nuclear many-body problem. Theoretical and Mathematical Physics, Berlin, Germany: Springer, Mar. 2004.
  82. [84]A. Szabo and N. S. Ostlund, Modern quantum chemistry. Dover Books on Chemistry, Mineola, NY: Dover Publications, Jan. 1996.
  83. [85]H. Watanabe and H. Murayama, “Unified description of nambu-goldstone bosons without lorentz invariance,” Phys. Rev. Lett., vol. 108, p. 251602, Jun. 2012.
  84. [86]D. Thouless, “Variational states of nuclei in the random phase approximation,” Nuclear Physics, vol. 22, no. 1, pp. 78–95, 1961.DOI
  85. [87]J. S. Andrews, D. Jayatilaka, R. G. A. Bone, N. C. Handy, and R. D. Amos, “Spin contamination in single-determinant wavefunctions,” Chemical Physics Letters, vol. 183, pp. 423–431, Sept. 1991.
  86. [88]P. K. Nandi, T. Kar, and A. B. Sannigrahi, “Effect of spin contamination in UHF wavefunctions on charge density-based local quantities,” Journal of Molecular Structure: THEOCHEM, vol. 362, pp. 69–75, Jan. 1996.DOI
  87. [89]J. Baker, A. Scheiner, and J. Anzelm, “Spin contamination in density functional theory,” Chemical Physics Letters, vol. 216, no. 3, pp. 380–388, 1993.DOI
  88. [90]F. Furche, “On the density matrix based approach to time-dependent density functional response theory,” The Journal of Chemical Physics, vol. 114, pp. 5982–5992, 04 2001.DOI
  89. [91]T. Holstein and H. Primakoff, “Field dependence of the intrinsic domain magnetization of a ferromagnet,” Phys. Rev., vol. 58, pp. 1098–1113, Dec. 1940.DOI
  90. [92]P. W. Anderson, “An approximate quantum theory of the antiferromagnetic ground state,” Phys. Rev., vol. 86, pp. 694–701, Jun. 1952.
  91. [93]R. Kubo, “The spin-wave theory of antiferromagnetics,” Phys. Rev., vol. 87, pp. 568–580, Aug 1952.DOI
  92. [94]Y. Hidaka, “Counting rule for nambu-goldstone modes in nonrelativistic systems,” Phys. Rev. Lett., vol. 110, p. 091601, Feb. 2013.pubmed.ncbi.nlm.nih.gov/23496701
  93. [95]H. Watanabe, “Counting rules of nambu-goldstone modes,” Annu. Rev. Condens. Matter Phys., vol. 11, no. 1, pp. 169–187, 2020.DOI
  94. [96]J. J. Palacios, J. Fernández-Rossier, and L. Brey, “Vacancy-induced magnetism in graphene and graphene ribbons,” Phys. Rev. B, vol. 77, p. 195428, May 2008.DOI
  95. [97]J. J. Palacios and F. Ynduráin, “Critical analysis of vacancy-induced magnetism in monolayer and bilayer graphene,” Physical Review B, vol. 85, p. 245443, June 2012.DOI
  96. [98]E. C. Stoner, “Collective electron ferromagnetism,” Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences, vol. 165, pp. 372–414, 04 1938.DOI
  97. [99]A. Nag, Y. Peng, J. Li, S. Agrestini, H. C. Roberts, M. García-Fernández, A. C. Walters, Q. Wang, Q. Yin, H. Lei, Z. Yin, and K.-J. Zhou, “Correlation driven near-flat band Stoner excitations in a Kagome magnet,” Nature Communications, vol. 13, p. 7317, Nov. 2022.DOI
  98. [100]D. C. Young, “Spin Contamination,” in Computational Chemistry, pp. 227–230, Nashville, TN: John Wiley & Sons, Ltd., 2001.DOI
  99. [101]R. Rodríguez-Guzmán, K. W. Schmid, C. A. Jiménez-Hoyos, and G. E. Scuseria, “Symmetry-projected variational approach for ground and excited states of the two-dimensional Hubbard model,” Physical Review B, vol. 85, p. 245130, June 2012.DOI
  100. [102]R. Rodríguez-Guzmán, C. A. Jiménez-Hoyos, R. Schutski, and G. E. Scuseria, “Multireference symmetry-projected variational approaches for ground and excited states of the one-dimensional Hubbard model,” Physical Review B, vol. 87, p. 235129, June 2013.DOI
  101. [103]H. A. Mook and D. M. Paul, “Neutron-scattering measurement of the spin-wave spectra for nickel,” Phys. Rev. Lett., vol. 54, pp. 227–229, Jan 1985.pubmed.ncbi.nlm.nih.gov/10031446
  102. [104]S. V. Halilov, H. Eschrig, A. Y. Perlov, and P. M. Oppeneer, “Adiabatic spin dynamics from spin-density-functional theory: Application to fe, co, and ni,” Phys. Rev. B, vol. 58, pp. 293–302, Jul 1998.DOI

Paper details

vlnc:W_ak4gym5p
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

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