This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Cross-sublattice Spin Pumping and Magnon Level Attraction in van der Waals Antiferromagnets

Roberto E. Troncoso Center for Quantum Spintronics, Department of Physics, Norwegian University of Science and Technology, NO-7491 Trondheim, Norway    Mike A. Lund Department of Engineering Sciences, University of Agder, 4879 Grimstad, Norway Center for Quantum Spintronics, Department of Physics, Norwegian University of Science and Technology, NO-7491 Trondheim, Norway    Arne Brataas Center for Quantum Spintronics, Department of Physics, Norwegian University of Science and Technology, NO-7491 Trondheim, Norway    Akashdeep Kamra Center for Quantum Spintronics, Department of Physics, Norwegian University of Science and Technology, NO-7491 Trondheim, Norway
Abstract

We theoretically study spin pumping from a layered van der Waals antiferromagnet in its canted ground state into an adjacent normal metal. We find that the resulting dc spin pumping current bears contributions along all spin directions. Our analysis allows for detecting intra- and cross-sublattice spin-mixing conductances via measuring the two in-plane spin current components. We further show that sublattice symmetry-breaking Gilbert damping can be realized via interface engineering and induces a dissipative coupling between the optical and acoustic magnon modes. This realizes magnon level attraction and exceptional points in the system. Furthermore, the dissipative coupling and cross-sublattice spin pumping contrive to produce an unconventional spin current in the out-of-plane direction. Our findings provide a route to extract the spin mixing conductance matrix and uncovers the unique opportunities, such as level attraction, offered by van der Waals antiferromagnet-normal metal hybrids.

Introduction.– The dawn of magnetic van der Waals (vdW) materials has renewed and invigorated interest in low-dimensional phenomena hosted by solid state systems Burch et al. (2018); Gibertini et al. (2019). These layered vdWs materials have been found to host various forms of magnetic order Huang et al. (2017); Deng et al. (2018); Huang et al. (2018); Song and Gabor (2018); Klein et al. (2018); Lado and Fernández-Rossier (2017) with antiferromagnets (AFs) taking a special place due to their various unique advantages Jungwirth et al. (2016); Baltz et al. (2018); Gomonay and Loktev (2014); Gomonay et al. (2018). Among these, a control over interfacial exchange coupling to an adjacent metal and canted or noncollinear ground state offers unprecedented pathways to achieve intriguing physics and applications Bender et al. (2017); Kamra and Belzig (2017); Troncoso et al. (2020); Flebus et al. (2019); Hellman et al. (2017); Johnsen et al. (2020); Rabinovich et al. (2019). The vdW magnets offer an effective control over both - interface and ground state - due to their layered structure and relatively weak interlayer antiferromagnetic exchange Huang et al. (2017); Deng et al. (2018); Wang et al. (2019).

Capitalizing on these features, magnon-magnon coupling resulting in level repulsion and hybridization has recently been observed in the vdW AF CrCl3 MacNeill et al. (2019); Kapoor et al. (2020); Sklenar and Zhang (2020). Approaching the challenge from a different direction, similar magnonic hybridization via their mutual coupling Kamra et al. (2017); Rezende et al. (2019); Yu et al. (2020) has been discovered in carefully chosen platforms including compensated ferrimagnets Liensberger et al. (2019) and synthetic AFs Sud et al. (2020); Chen et al. (2018). These investigations have, in part, been driven by the desire to control magnonic systems for quantum information applications Lachance-Quirion et al. (2019) and were preceded by the realization of strong magnon-photon coupling Huebl et al. (2013); Harder and Hu (2018). The latter, being a relatively mature field, has started to explore dissipative magnon-photon coupling Harder et al. (2018); Wang and Hu (2020) resulting in intriguing phenomena that foray into the realm of non-Hermitian physics Ashida et al. (2020) providing a powerful model platform. While not explored thus far, such phenomena can also result from dissipative magnon-magnon coupling offering various advantages over the magnon-photon platform Kamra et al. (2020). Demonstrating these for a vdW AF interfaced with a heavy normal metal (NM) forms a key contribution of this work.

Heterostructures comprising a magnetic insulator interfaced with a thin NM layer have become basic building blocks in an emerging spin-based paradigm for information transport and processing Goennenwein et al. (2015); Cornelissen et al. (2015); Hou et al. (2019); Troncoso et al. (2020); Lebrun et al. (2018); Wimmer et al. (2019); Chumak et al. (2015); Althammer (2018); Nakata et al. (2017); Brataas et al. (2020). In such structures, magnonic spin in the magnetic insulator can be interfaced with the electronic spin in NM thereby allowing their integration with conventional electronics. Furthermore, spin generated in NM allows to control, and even negate Wimmer et al. (2019), dissipation in the magnetic system via spin transfer torques Ralph and Stiles (2008). Invigorated by recent breakthroughs, especially employing the magnonic spin in AFs Lebrun et al. (2018); Vaidya et al. (2020); Li et al. (2020); Troncoso et al. (2020), an interface-engineering and control of spin transfer from AF to NM via magnonic spin pumping Tserkovnyak et al. (2002, 2005); Cheng et al. (2014); Kamra and Belzig (2017); Bender et al. (2017) assumes a central role. While coherently driven spin pumping from an AF into NM has recently been observed in its collinear ground state Vaidya et al. (2020); Li et al. (2020), a noncollinear or canted AF should enable unique and novel phenomena emerging from cross-sublattice spin pumping Kamra and Belzig (2017); Kamra et al. (2018); Liu et al. (2017). Besides providing the much needed understanding of spin transfer across the AF-NM interface, cross-sublattice pumping may also offer a direct probe into the aforementioned dissipative and non-Hermitian magnon-magnon coupling phenomena, as shown in this work.

Refer to caption
Figure 1: (a) Schematic setup for the measurement of pumped spin currents in the antiferromagnet (AF)-normal metal (NM) structure. The in-plane spin-polarization of the spin current can be detected by the measurements of inverse spin-Hall voltages VISHExV^{x}_{\text{ISHE}} and VISHEyV^{y}_{\text{ISHE}}. Possible interface microstructures are schematically depicted in (b), (c), and (d). The cross-sublattice spin mixing conductance gABg_{AB} vanishes [is finite] for the interface depicted in (c) [(b) and (d)].

In this Letter, we demonstrate heterostructures comprising vdW AFs and a heavy NM to be a unique platform for observing intriguing phenomena emerging from cross-sublattice spin pumping and dissipative magnon-magnon coupling. This niche is enabled by the layered structure of vdW AFs resulting in the possibility of AF-NM interface engineering and canted AF ground states with application of relatively small magnetic fields. We show that in such a non-collinear ground state, the AF pumps spin into the NM along all three directions on excitation via an rf magnetic field. A detection of the two in-plane spin components via inverse spin Hall effect allows to determine the complete spin mixing conductance matrix of the interface. We further find an unconventional out-of-plane spin pumping component that results from a concerted effect of cross-sublattice spin pumping and a dissipative coupling resulting from the sublattice-symmetry breaking at AF-NM interface. Furthermore, the ensuing dissipative coupling is found to result in magnon-magnon level attraction and coalescence observable via typical magnetic resonance experiments. The system thus constitutes a novel and unique platform for investigating this interplay between magnon level repulsion, attraction, and non-Hermitian physics via in-situ damping matrix engineering by, for example, spin transfer torques.

Model.– We treat the vdW material as a two-sublattice magnet described by the magnetization fields 𝑴A{\boldsymbol{M}}_{A} and 𝑴B{\boldsymbol{M}}_{B} that correspond to the sublattices AA and BB. We consider magnetic free-energy density MacNeill et al. (2019) F=μ0HE𝑴A𝑴B/Ms+μ0(MAz2+MBz2)/2μ0𝑯(𝑴A+𝑴B)F=\mu_{0}H_{E}{\boldsymbol{M}}_{A}\cdot{\boldsymbol{M}}_{B}/M_{s}+{\mu_{0}}\left(M^{2}_{Az}+M^{2}_{Bz}\right)/2-\mu_{0}{\boldsymbol{H}}\cdot\left({\boldsymbol{M}}_{A}+{\boldsymbol{M}}_{B}\right), with MsM_{s} the saturation magnetization of each sublattice. The inter-sublattice exchange coupling parametrized by HE>0H_{E}>0 favors antiferromagnetic order. In addition, an external dc magnetic field 𝑯{\boldsymbol{H}} is applied in-plane. The second term represents the easy-plane anisotropy. Gilbert damping is accounted for by the viscous Rayleigh dissipation functional Gilbert (2004); Kamra et al. (2018); Yuan et al. (2019) via the symmetric matrix ηζζ\eta_{\zeta\zeta^{\prime}}: R[𝑴˙A,𝑴˙B]=ζζV𝑑𝒓ηζζ𝑴˙ζ𝑴˙ζ/2R[\dot{\boldsymbol{M}}_{A},\dot{\boldsymbol{M}}_{B}]=\sum_{\zeta\zeta^{\prime}}\int_{V}d{\boldsymbol{r}}{\eta_{\zeta\zeta^{\prime}}}\dot{\boldsymbol{M}}_{\zeta}\cdot\dot{\boldsymbol{M}}_{\zeta^{\prime}}/2, where {ζ,ζ}={A,B}\{\zeta,\zeta^{\prime}\}=\{A,B\}. The ensuing magnetization dynamics is described by the coupled Landau-Lifshitz-Gilbert (LLG) equations,

𝒎˙ζ=μ0γ𝒎ζ×𝒉ζeff+αζζ𝒎ζ×𝒎˙ζ+𝝉ζ,\displaystyle\dot{\boldsymbol{m}}_{\zeta}=-\mu_{0}\gamma{\boldsymbol{m}}_{\zeta}\times{\boldsymbol{h}}^{\text{eff}}_{\zeta}+\alpha_{\zeta\zeta^{\prime}}{\boldsymbol{m}}_{\zeta}\times\dot{\boldsymbol{m}}_{\zeta^{\prime}}+{\boldsymbol{\tau}}_{\zeta}, (1)

in terms of the unit vectors 𝒎ζ𝑴ζ/Ms{\boldsymbol{m}}_{\zeta}\equiv{\boldsymbol{M}}_{\zeta}/M_{s}. The effective fields are given by 𝒉ζeff=(1/μ0)×F/𝑴ζ=𝑯HEσζζx𝒎ζMs(𝒎ζ𝒛^)𝒛^{\boldsymbol{h}}^{\text{eff}}_{\zeta}=(1/\mu_{0})\times\partial F/\partial{\boldsymbol{M}}_{\zeta}={\boldsymbol{H}}-H_{E}\sigma^{x}_{\zeta\zeta^{\prime}}{\boldsymbol{m}}_{\zeta^{\prime}}-M_{s}\left({\boldsymbol{m}}_{\zeta}\cdot{\hat{\boldsymbol{z}}}\right){\hat{\boldsymbol{z}}}, with σx\sigma^{x} the Pauli matrix and γ>0\gamma>0 is the gyromagnetic ratio magnitude. The Gilbert damping parameters are defined through αζζγMsηζζ\alpha_{\zeta\zeta^{\prime}}\equiv\gamma M_{s}\eta_{\zeta\zeta^{\prime}} where, in particular, αAB=αBAαod\alpha_{AB}=\alpha_{BA}\equiv\alpha_{od}. Note that sublattice asymmetry in our model is broken only by the Gilbert damping Kamra et al. (2018). It results from the AF-NM interface and spin pumping-mediated losses Tserkovnyak et al. (2002); Kamra et al. (2018). The magnetization dynamics may be excited by a time-dependent magnetic field 𝒉μ0γ𝒉RF(t){\boldsymbol{h}}\equiv\mu_{0}\gamma{\boldsymbol{h}}_{\text{RF}}(t) that produces a torque 𝝉ζ=𝒎ζ×𝒉{\boldsymbol{\tau}}_{\zeta}={\boldsymbol{m}}_{\zeta}\times{\boldsymbol{h}}.

Refer to caption
Figure 2: The frequency and field dependence of the spin-pumping current is represented by the plot of ±j(ω)\mathscr{F}^{j}_{\pm}(\omega), with jj the directions of polarization and damping parameters α¯=0.05\bar{\alpha}=0.05, αod=0.01\alpha_{od}=0.01 and Δα¯=0.005\Delta\bar{\alpha}=0.005. At panels (a), (b) and (c), we have depicted the spin current with spin-polarization along xx, yy- and zz-direction, respectively, due to coherent excitation of optical modes. Similarly, in panels (d), (e) and (f), we have plotted the polarization components of the spin current when acoustic modes are excited. At the inset of each panel we show ±j\mathscr{F}^{j}_{\pm}, evaluated at the resonant frequencies ω=ω±\omega=\omega_{\pm}. In the inset of panels (a), (b), (d) and (e), the curves correspond to different αod\alpha_{od}, while at panels (c) and (f) Δα¯\Delta\bar{\alpha} is modified. Other parameters employed at the plots were extracted for the vdW antiferromagnet CrCl3 MacNeill et al. (2019).

Magnetization dynamics and magnon modes.– We now investigate the magnon modes in the material when the two sublattice magnetizations are non-collinear in their equilibrium configuration [Fig. 1(a)]. In the presence of an in-plane external magnetic field 𝑯=H𝒚^{\boldsymbol{H}}=H\hat{\boldsymbol{y}}, the magnetic ground state becomes 𝒎ζeq=±cosϕ𝒙^+sinϕ𝒚^{\boldsymbol{m}}^{\text{eq}}_{\zeta}=\pm\cos\phi\hat{\boldsymbol{x}}+\sin\phi\hat{\boldsymbol{y}}, where the non-collinearity is captured by the finite angle ϕ\phi that satisfies sinϕ=H/2HE\sin\phi=H/2H_{E} [see Fig. 1(a)]. The magnetic ground state is invariant under a twofold rotational operation around the yy axis, 𝒞2y{\cal C}_{2y}, in combination with sublattice exchange ABA\leftrightarrow B, i.e. 𝒞2y𝒎Aeq=𝒎Beq{\cal C}_{2y}{\boldsymbol{m}}^{\text{eq}}_{A}={\boldsymbol{m}}^{\text{eq}}_{B}. Linearizing the LLG equations (1), considering 𝒎ζ=𝒎ζeq+δ𝒎ζeiωt{\boldsymbol{m}}_{\zeta}={\boldsymbol{m}}^{\text{eq}}_{\zeta}+\delta{\boldsymbol{m}}_{\zeta}e^{i\omega t}, the coupled dynamical equations become,

iωδ𝒎+\displaystyle i\omega\delta{\boldsymbol{m}}_{+} =𝒎Aeq×(𝒜+δ𝒎++iωΔα¯δ𝒎)+𝝉+,\displaystyle={\boldsymbol{m}}^{\text{eq}}_{A}\times\left({\cal A}_{+}\delta{\boldsymbol{m}}_{+}+i\omega\Delta\bar{\alpha}\delta{\boldsymbol{m}}_{-}\right)+{\boldsymbol{\tau}}_{+}, (2a)
iωδ𝒎\displaystyle i\omega\delta{\boldsymbol{m}}_{-} =𝒎Aeq×(𝒜δ𝒎+iωΔα¯δ𝒎+)+𝝉,\displaystyle={\boldsymbol{m}}^{\text{eq}}_{A}\times\left({\cal A}_{-}\delta{\boldsymbol{m}}_{-}+i\omega\Delta\bar{\alpha}\delta{\boldsymbol{m}}_{+}\right)+{\boldsymbol{\tau}}_{-}, (2b)

with the two magnetization dynamics or magnon modes described by the fields δ𝒎±=δ𝒎A±𝒞2yδ𝒎B\delta{\boldsymbol{m}}_{\pm}=\delta{\boldsymbol{m}}_{A}\pm{\cal C}_{2y}\delta{\boldsymbol{m}}_{B}, the torques 𝝉±=𝝉A±𝒞2y𝝉B{\boldsymbol{\tau}}_{\pm}={\boldsymbol{\tau}}_{A}\pm{\cal C}_{2y}{\boldsymbol{\tau}}_{B} and the operator 𝒜±=(μ0γHE+iωα¯)±(μ0γHE+iωαod)𝒞2y{\cal A}_{\pm}=\left(\mu_{0}\gamma H_{E}+i\omega\bar{\alpha}\right)\pm\left(\mu_{0}\gamma H_{E}+i\omega{\alpha}_{od}\right){\cal C}_{2y}. Furthermore, we have reformulated the Gilbert damping parameters as αAA=α¯+Δα¯\alpha_{AA}=\bar{\alpha}+\Delta\bar{\alpha} and αBB=α¯Δα¯\alpha_{BB}=\bar{\alpha}-\Delta\bar{\alpha}. Note that when sublattice symmetry is assumed, i.e., αAA=αBB\alpha_{AA}=\alpha_{BB}, the Eqs. (2a) and (2b) become decoupled since Δα¯=0\Delta\bar{\alpha}=0. In the absence of dissipation, the magnon eigenmodes are captured well by the fields δ𝒎±\delta{\boldsymbol{m}}_{\pm}, with the eigenfrequencies for the so-called optical and acoustic magnon modes being ω+=μ0γ2MsHEcos2ϕ\omega_{+}=\mu_{0}\gamma\sqrt{2M_{s}H_{E}\cos^{2}\phi} and ω=μ0γ2HE(Ms+2HE)sinϕ\omega_{-}=\mu_{0}\gamma\sqrt{2H_{E}\left(M_{s}+2H_{E}\right)}\sin\phi, respectively. The two modes can be excited selectively by a careful choice of the rf-field 𝒉{\boldsymbol{h}} MacNeill et al. (2019). In general, the excitation of (±)(\pm)-modes one at a time, which demands 𝝉=0{\boldsymbol{\tau}}_{\mp}=0 for the torque, imposes the conditions 𝒉=±C2y𝒉{\boldsymbol{h}}=\pm C_{2y}{\boldsymbol{h}}, respectively.

Spin pumping.– We now investigate spin transport across the AF-NM interface resulting from the excitation of magnetization dynamics by rf magnetic field. The dc spin pumping current injected into the adjacent NM is given by Kamra and Belzig (2017)

e𝑰s=ζζ{A,B}gζζ𝒎ζ×𝒎˙ζ,\displaystyle\frac{e}{\hbar}{\boldsymbol{I}}_{s}=\sum_{\zeta\zeta^{\prime}\in\{A,B\}}g_{\zeta\zeta^{\prime}}\langle{\boldsymbol{m}}_{\zeta}\times\dot{\boldsymbol{m}}_{\zeta^{\prime}}\rangle, (3)

where \langle\cdots\rangle stands for the time-average over the period of oscillation. The diagonal elements of the matrix gζζg_{\zeta\zeta^{\prime}} describe the intra-sublattice spin mixing conductance. An asymmetric interfacial coupling Bender et al. (2017); Kamra and Belzig (2017); Troncoso et al. (2020); Flebus et al. (2019), resulting in gAAgBBg_{AA}\neq g_{BB}, occurs when the two magnetic sublattices are incommensurately exposed to the NM (see Fig. 1(c) for an example in which gAA=0g_{AA}=0). The off-diagonal cross-sublattice conductance satisfy gAB=gBAg_{AB}=g_{BA} and are nonzero when both the sublattices are (partly) exposed to NM [Fig. 1(b) and (d)]. Such interfaces can be achieved with layered magnets Yin et al. (2018); Huang et al. (2018) (Fig. 1), but are not possible with synthetic AFs Sud et al. (2020). Our goal here is to establish the experimentally detectable spin pumping current as a direct probe of the 2×22\times 2 spin mixing conductance matrix [Eq. (3)]. In particular, we are interested in establishing unique signatures of the cross-sublattice conductances gAB=gBAg_{AB}=g_{BA} that elude a direct experimental observation thus far.

In typical experimental setups (Fig. 1), the spin pumping current is detected via inverse spin-Hall effect (ISHE) Sinova et al. (2015); Ando et al. (2011); Valenzuela and Tinkham (2006); Saitoh et al. (2006). In metals with strong spin-orbit coupling, a nonequilibrium spin current 𝒋s{\boldsymbol{j}}_{s} induces a transverse charge current 𝒋c=θSH(2e/)𝒋s×σ{\boldsymbol{j}}_{c}=\theta_{\text{SH}}(2e/\hbar){\boldsymbol{j}}_{s}\times\vec{\sigma}, where σ\vec{\sigma} denotes the spin polarization direction and θSH\theta_{\text{SH}} is the spin Hall angle Sinova et al. (2015). Under open circuit conditions, the generated charge current is countered by an induced inverse spin Hall voltage VISHEV_{\text{ISHE}} proportional to the charge, and thus spin, current. The voltage thus generated is proportional to the spin current injected into the NM [Eq. (3)] and is directly detected in experiments Mosendz et al. (2010).

In order to relate spin mixing conductance matrix elements to experimental observables, we evaluate the spin pumping current Eq. (3) in the limit αAAαBB\alpha_{AA}\approx\alpha_{BB}, i.e., Δα¯\Delta\bar{\alpha} small. In this perturbative regime, Eqs. (2a) and (2b) decouple and the system eigenmodes are the optical and acoustic magnon modes. The corresponding spin pumping currents are evaluated to be

𝐈z±=\displaystyle{\bf I}^{\pm}_{z}= (gAAgBB)±x(ω)𝒙^+gAB±z(ω)𝒛^\displaystyle\left(g_{AA}-g_{BB}\right)\mathscr{F}^{x}_{\pm}(\omega)\hat{\boldsymbol{x}}+g_{AB}\mathscr{F}^{z}_{\pm}(\omega)\hat{\boldsymbol{z}}
+(gAA+gBB±2gAB)±y(ω)𝒚^.\displaystyle\qquad\qquad+\left(g_{AA}+g_{BB}\pm 2g_{AB}\right)\mathscr{F}^{y}_{\pm}(\omega)\hat{\boldsymbol{y}}. (4)

The ±\pm index labels optical and acoustic modes, which are driven by the rf magnetic field with frequency ω\omega and amplitude h+,ϕ=2hycosϕh_{+,\phi}=2h_{y}\cos\phi and h,ϕ=2hxsinϕh_{-,\phi}=2h_{x}\sin\phi, respectively. The expression in Eq. (Cross-sublattice Spin Pumping and Magnon Level Attraction in van der Waals Antiferromagnets) represents a pure spin current that flows across the AF-NM interface (along zz axis). Its various components pertaining to directions in the spin space are proportional to the functions ±j(ω)\mathscr{F}^{j}_{\pm}(\omega), as detailed in the Supplemental Material in SM (see Eqs. (41)-(43)), which have been obtained to the first order in Δα¯\Delta\bar{\alpha}. The in-plane components ±x,y\mathscr{F}^{x,y}_{\pm}, satisfying ±y(ω)=tanϕ±x(ω)\mathscr{F}^{y}_{\pm}(\omega)=\tan\phi\mathscr{F}^{x}_{\pm}(\omega), are independent of Δα¯\Delta\bar{\alpha}. However, the out-of-plane component of the spin current ±z(ω)\propto\mathscr{F}^{z}_{\pm}(\omega) scales linearly with Δα¯\Delta\bar{\alpha}. The resulting field- and frequency-dependence of the spin current 𝐈z±{\bf I}^{\pm}_{z} components are plotted in Fig. 2 employing system parameters relevant for the vdW AF CrCl3 MacNeill et al. (2019). The various spin current components are displayed in panels (a), (b) and (c) of Fig. 2 for the optical mode, and in panels (d), (e), and (f) for the acoustic mode.

In contrast with the case of spin pumping via collinear magnets, in which a dc spin current polarized along the equilibrium order is generated Tserkovnyak et al. (2002); Cheng et al. (2014); Mosendz et al. (2010); Saitoh et al. (2006); Vaidya et al. (2020); Li et al. (2020), the canted AF under consideration pumps spin with components along all three directions [Eq. (Cross-sublattice Spin Pumping and Magnon Level Attraction in van der Waals Antiferromagnets)]. As detailed in the Supplemental Material SM , ±y,zsinϕ\mathscr{F}^{y,z}_{\pm}\propto\sin\phi implying yy and zz components of the spin pumping current vanish for ϕ=0\phi=0. Our result thus reduces to the existing understanding of collinear AFs Kamra and Belzig (2017); Troncoso et al. (2020). These additional components of the spin pumping current for the canted AF together with the existence of two independent magnon modes constitute some of the unique features and opportunities offered by this system. For example, detection of the ISHE voltage in two orthogonal directions [as depicted in Fig. 1(a)] enables determination of both in-plane spin current components. By detecting these while exciting the two magnon modes one at a time, we may determine the full spin mixing conductance matrix with the cross-sublattice term given by

gAB=gAAgBB4tanϕ(Is,y+Is,x+Is,yIs,x),\displaystyle{g_{AB}}=\frac{g_{AA}-g_{BB}}{4\tan\phi}\left(\frac{{I}^{+}_{s,y}}{{I}^{+}_{s,x}}-\frac{{I}^{-}_{s,y}}{{I}^{-}_{s,x}}\right), (5)

which assumes the condition gAAgBBg_{AA}\neq g_{BB}. This accomplishes a key goal and constitutes a main result of this paper.

Furthermore, existence of the spin current zz component [Eq. (Cross-sublattice Spin Pumping and Magnon Level Attraction in van der Waals Antiferromagnets)] that we find is unconventional and counterintuitive as the magnon spin is expected to lie in the same plane as the equilibrium sublattice magnetizations Kamra et al. (2017); Rezende et al. (2019). However, this component is nonzero only when Δα¯0\Delta\bar{\alpha}\neq 0, gAAgBBg_{AA}\neq g_{BB}, gAB0g_{AB}\neq 0, and ϕ0\phi\neq 0 implying that it results from a complex interplay of the sublattice-symmetry breaking dissipative coupling and a cross-sublattice interference effect. Such physics, especially dissipative coupling Harder et al. (2018); Wang and Hu (2020), appears to go beyond the magnon picture considered thus far and constitutes another key result of our work. While an out-of-plane spin component has not been measured in typical spin pumping experiments Saitoh et al. (2006); Sinova et al. (2015); Mosendz et al. (2010), the recent thermal drag-mediated detection of such an out-of-plane spin component Avci et al. (2020) provides one possible method for its direct observation.

Refer to caption
Figure 3: Eigenfrequencies of the coupled magnon modes are plotted as a function of the external dc magnetic field. The real (ωr\omega_{r}) and imaginary (ωi\omega_{i}) parts of the frequencies are displayed in panels (a) and (c) respectively, for α¯=0.1\bar{\alpha}=0.1 and αod=Δα¯=0.07\alpha_{od}=\Delta\bar{\alpha}=0.07. In panel (b), we zoom-in on the level crossing of panel (a). The additional curves shown in purple and red correspond to the same α¯\bar{\alpha} and Δα¯\Delta\bar{\alpha} as (a), but with αod=0\alpha_{od}=0. The level attraction is depicted in gray when Δα¯=0.07\Delta\bar{\alpha}=0.07 and α¯=αod=0\bar{\alpha}=\alpha_{od}=0. (d) Imaginary part of the magnetic susceptibility Im[χϕϕ+]\text{Im}[\chi^{+}_{\phi\phi}] vs. applied magnetic field for various values of Δα¯\Delta\bar{\alpha}. The susceptibility corresponds to the optical mode and has been evaluated at ω=ω+\omega=\omega_{+}.

Magnon level attraction.– We now discuss the magnon eigenmodes which become dissipatively coupled [see Eqs. (2a) and (2b)] due to the sublattice-symmetry breaking Gilbert damping Kamra et al. (2018), i.e. nonzero Δα¯\Delta\bar{\alpha}. While a “reactive” coupling between magnon modes has been observed in various systems Chen et al. (2018); Liensberger et al. (2019); Sud et al. (2020); MacNeill et al. (2019); Kapoor et al. (2020), dissipative coupling remains less explored and invokes non-Hermitian physics Ashida et al. (2020); Flebus et al. (2020). Solving Eqs. (2a) and (2b) without an external rf drive, we obtain the complex eigenmode frequencies ω±=ωr±+iωi±\omega_{\pm}=\omega_{r\pm}+i\omega_{i\pm}. ωr\omega_{r} and ωi\omega_{i}, respectively, capture the energy and inverse lifetime of the magnon modes and have been plotted against the external dc magnetic field in Fig. 3 (a)-(c). Due to the dissipative nature of the coupling, a magnon-magnon level attraction is observed. The grey curve in Fig. 3 (b), corresponding to α¯=αod=0\bar{\alpha}=\alpha_{od}=0 and Δα¯0\Delta\bar{\alpha}\neq 0, depicts a perfect level coalescence or mode synchronization Harder et al. (2018); Wang and Hu (2020). Such values for Gilbert damping matrix require a dc spin transfer torque drive in the NM. An undriven system however imposes constraints αAA,αBB>0\alpha_{AA},\alpha_{BB}>0, i.e. α¯>Δα¯\bar{\alpha}>\Delta\bar{\alpha} and αodαAAαBB\alpha_{od}\leq\sqrt{\alpha_{AA}\alpha_{BB}} Kamra et al. (2018). In this scenario, we find a complex interplay of repulsion and attraction between the two modes. The resulting eigenfrequencies split slightly (see Fig. 3(a) when α¯=0.1\bar{\alpha}=0.1, αod=0.07\alpha_{od}=0.07 and Δα¯=0.07\Delta\bar{\alpha}=0.07) while coalescing at a specific point, the so-called exceptional point Tserkovnyak (2020). Furthermore, Fig. 3(d) depicts the imaginary part of dynamic susceptibility, which is directly accessible in experiments Liensberger et al. (2019); MacNeill et al. (2019); Berger et al. (2018). A peak in this susceptibility provides an additional experimental signature of the level attraction when Δα¯\Delta\bar{\alpha} is sufficiently large. Thus, vdW AFs under consideration constitute a rich platform for realizing non-Hermitian physics and magnon-magnon level attraction via AF-NM interface engineering and spin transfer torques exerted on AF by the NM.

Summary.– We have theoretically uncovered unique and intriguing cross-sublattice spin pumping and magnon-magnon level attraction effects in a model canted antiferromagnet. By providing guidance to experiments in extracting the interfacial spin mixing conductance matrix and key signatures of level attraction, we hope to establish van der Waals antiferromagnets interfaced with a heavy metal layer as a fertile and convenient platform for realizing and investigating unconventional non-Hermitian physics.

Note added: During the manuscript preparation, we noticed a recent related preprint Lu et al. (2020) that studies level-repulsion and hybridization of magnonic modes in bulk symmetry-breaking synthetic antiferromagnets. It however does not discuss spin pumping or the dissipative magnon level attraction - the two key novelties of our work.

Acknowledgements.
This work was supported by the European Union’s Horizon 2020 Research and Innovation Programme under Grant No. DLV-737038 “TRANSPIRE,” and the Research Council of Norway through its Centres of Excellence funding scheme, Project No. 262633, “QuSpin”.

References

I Supplemental Material

In this Supplemental Material, we explicitly show the calculation of two-sublattice spin pumping in AF-NM structures. The result is applied when the magnetization dynamics is coherently driven by a linearly polarized ac magnetic field. Basic details of magnon-magnon hybridization are also provided.

I.1 Magnetization dynamics and magnon modes

Magnetic fluctuations of the two-sublattice magnet results from the linearization of LLG equations. Representing the fluctuations by 𝒎ζ=𝒎ζeq+δ𝒎ζeiωt{\boldsymbol{m}}_{\zeta}={\boldsymbol{m}}^{\text{eq}}_{\zeta}+\delta{\boldsymbol{m}}_{\zeta}e^{i\omega t}, with ζ={A,B}\zeta=\{A,B\}, the resulting coupled dynamical equations becomes,

iωδ𝒎+\displaystyle i\omega\delta{\boldsymbol{m}}_{+} =𝒎Aeq×[(μ0γHE+iωα¯)δ𝒎++(μ0γHE+iωαod)𝒞2yδ𝒎+]+iωΔα¯𝒎Aeq×δ𝒎+𝝉+,\displaystyle={\boldsymbol{m}}^{\text{eq}}_{A}\times\left[\left(\mu_{0}\gamma H_{E}+i\omega\bar{\alpha}\right)\delta{\boldsymbol{m}}_{+}+\left(\mu_{0}\gamma H_{E}+i\omega{\alpha}_{od}\right){\cal C}_{2y}\delta{\boldsymbol{m}}_{+}\right]+i\omega\Delta\bar{\alpha}{\boldsymbol{m}}^{\text{eq}}_{A}\times\delta{\boldsymbol{m}}_{-}+{\boldsymbol{\tau}}_{+}, (6a)
iωδ𝒎\displaystyle i\omega\delta{\boldsymbol{m}}_{-} =𝒎Aeq×[(μ0γHE+iωα¯)δ𝒎(μ0γHE+iωαod)𝒞2yδ𝒎]+iωΔα¯𝒎Aeq×δ𝒎++𝝉,\displaystyle={\boldsymbol{m}}^{\text{eq}}_{A}\times\left[\left(\mu_{0}\gamma H_{E}+i\omega\bar{\alpha}\right)\delta{\boldsymbol{m}}_{-}-\left(\mu_{0}\gamma H_{E}+i\omega{\alpha}_{od}\right){\cal C}_{2y}\delta{\boldsymbol{m}}_{-}\right]+i\omega\Delta\bar{\alpha}{\boldsymbol{m}}^{\text{eq}}_{A}\times\delta{\boldsymbol{m}}_{+}+{\boldsymbol{\tau}}_{-}, (6b)

with the torques 𝝉±=𝒎A×𝒉±{\boldsymbol{\tau}}_{\pm}={\boldsymbol{m}}_{A}\times{\boldsymbol{h}}_{\pm}. The Gilbert damping terms are represented by αAB=αBA=αod\alpha_{AB}=\alpha_{BA}=\alpha_{od}, αAA=α¯+Δα¯\alpha_{AA}=\bar{\alpha}+\Delta\bar{\alpha} and αBB=α¯Δα¯\alpha_{BB}=\bar{\alpha}-\Delta\bar{\alpha}. Note that when sublattice symmetry is restored, i.e., αAA=αBB\alpha_{AA}=\alpha_{BB}, the modes δ𝒎+\delta{\boldsymbol{m}}_{+} and δ𝒎\delta{\boldsymbol{m}}_{-} become decoupled and characterize, optical and acoustic modes respectively. In a compact form the Eqs. (6a) and (6b) read,

(𝒉+𝒉)=(𝕄+𝕋𝕋𝕄)𝕄(δ𝒎+δ𝒎)\displaystyle\left(\begin{array}[]{c}{\boldsymbol{h}}_{+}\\ {\boldsymbol{h}}_{-}\end{array}\right)=\underbrace{\left(\begin{array}[]{c|c}\mathbb{M}_{+}&\mathbb{T}\\ \hline\cr\mathbb{T}^{*}&\mathbb{M}_{-}\end{array}\right)}_{\mathbb{M}}\left(\begin{array}[]{c}\delta{\boldsymbol{m}}_{+}\\ \delta{\boldsymbol{m}}_{-}\end{array}\right) (13)

with the ac fields 𝒉+=(h+,ϕ,0)T{\boldsymbol{h}}_{+}=(h_{+,\phi},0)^{T} and 𝒉=(h,ϕ,h,θ)T{\boldsymbol{h}}_{-}=(h_{-,\phi},h_{-,\theta})^{T}, where h+,ϕ=2hycosϕh_{+,\phi}=2h_{y}\cos\phi, h,ϕ=2hxsinϕh_{-,\phi}=2h_{x}\sin\phi and h,θ=2hzh_{-,\theta}=2h_{z}. The matrices 𝕄±\mathbb{M}_{\pm} and 𝕋\mathbb{T} are defined as

𝕋\displaystyle\mathbb{T} =(iωΔα¯00iωΔα¯),\displaystyle=\left(\begin{array}[]{cc}-i\omega\Delta\bar{\alpha}&0\\ 0&i\omega\Delta\bar{\alpha}\end{array}\right), (16)
𝕄+\displaystyle\mathbb{M}_{+} =(Aiω(α¯+αodcos2ϕ)iωiωB+iω(α¯αod)),\displaystyle=\left(\begin{array}[]{cc}-A-i\omega\left(\bar{\alpha}+\alpha_{od}\cos 2\phi\right)&i\omega\\ i\omega&B+i\omega\left(\bar{\alpha}-\alpha_{od}\right)\end{array}\right), (19)
𝕄\displaystyle\mathbb{M}_{-} =(C+iω(α¯αodcos2ϕ)iωiωDiω(α¯+αod)),\displaystyle=\left(\begin{array}[]{cc}C+i\omega\left(\bar{\alpha}-\alpha_{od}\cos 2\phi\right)&-i\omega\\ -i\omega&-D-i\omega\left(\bar{\alpha}+\alpha_{od}\right)\end{array}\right), (22)

where we introduced the following constants A=2μ0γHEcos2ϕA=2\mu_{0}\gamma H_{E}\cos^{2}\phi and B=μ0γMsB=\mu_{0}\gamma M_{s}, C=2μ0γHEsin2ϕC=2\mu_{0}\gamma H_{E}\sin^{2}\phi and D=μ0γ(Ms+2HE)D=\mu_{0}\gamma\left(M_{s}+2H_{E}\right). It is worth noting that det[𝕄]=det[𝕄+]det[𝕄𝕋𝕄+1𝕋]det[𝕄+]det[𝕄]\det\left[\mathbb{M}\right]=\det\left[\mathbb{M}_{+}\right]\det\left[\mathbb{M}_{-}-\mathbb{T}\mathbb{M}^{-1}_{+}\mathbb{T}\right]\approx\det\left[\mathbb{M}_{+}\right]\det\left[\mathbb{M}_{-}\right] in the limit of small Δα¯\Delta\bar{\alpha}. Beyond this approximation, the optical and acoustic magnonic modes are no longer decoupled. Instead, these modes hybridize with Δα¯\Delta\bar{\alpha} being the dissipative coupling that generates level attraction. The magnon-magnon hybrid eigenfrequencies were numerically calculated and depicted in Fig. 3.

Assuming that Δα¯\Delta\bar{\alpha} is small, we use the standard formula for the inverse of a block matrix to obtain the fields δ𝒎±\delta{\boldsymbol{m}}_{\pm},

δ𝒎+\displaystyle\delta{\boldsymbol{m}}_{+} =𝕄+1𝒉+𝕄+1𝕋𝕄1𝒉,\displaystyle=\mathbb{M}^{-1}_{+}{\boldsymbol{h}}_{+}-\mathbb{M}^{-1}_{+}\mathbb{T}\mathbb{M}^{-1}_{-}{\boldsymbol{h}}_{-}, (23)
δ𝒎\displaystyle\delta{\boldsymbol{m}}_{-} =𝕄1𝒉+𝕄1𝕋𝕄+1𝒉+.\displaystyle=\mathbb{M}^{-1}_{-}{\boldsymbol{h}}_{-}+\mathbb{M}^{-1}_{-}\mathbb{T}\mathbb{M}^{-1}_{+}{\boldsymbol{h}}_{+}. (24)

The first term at the right-hand side is the zero order correction in Δα¯\Delta\bar{\alpha}, where 𝕄±1\mathbb{M}^{-1}_{\pm} corresponds to the symmetric dynamic susceptibility matrix. The components are defined as [𝕄±1]11=χϕϕ±\left[\mathbb{M}^{-1}_{\pm}\right]_{11}=\chi^{\pm}_{\phi\phi}, [𝕄±1]22=χθθ±\left[\mathbb{M}^{-1}_{\pm}\right]_{22}=\chi^{\pm}_{\theta\theta} and [𝕄±1]12=iχϕθ±\left[\mathbb{M}^{-1}_{\pm}\right]_{12}=i\chi^{\pm}_{\phi\theta}, where

χϕϕ+\displaystyle\chi^{+}_{\phi\phi} =B+i(α¯αod)ωω2(B+iω(α¯αod))(A+iω(α¯+αodcos2ϕ)),\displaystyle=\frac{B+i\left(\bar{\alpha}-\alpha_{od}\right)\omega}{\omega^{2}-\left(B+i\omega(\bar{\alpha}-\alpha_{od})\right)\left(A+i\omega(\bar{\alpha}+\alpha_{od}\cos 2\phi)\right)}, (25)
χϕθ+\displaystyle\chi^{+}_{\phi\theta} =ωω2(B+iω(α¯αod))(A+iω(α¯+αodcos2ϕ)),\displaystyle=\frac{-\omega}{\omega^{2}-\left(B+i\omega(\bar{\alpha}-\alpha_{od})\right)\left(A+i\omega(\bar{\alpha}+\alpha_{od}\cos 2\phi)\right)}, (26)

and

χϕϕ\displaystyle\chi^{-}_{\phi\phi} =Diω(α¯+αod)ω2(D+iω(α¯+αod))(C+iω(α¯αodcos2ϕ)),\displaystyle=\frac{-D-i\omega(\bar{\alpha}+\alpha_{od})}{\omega^{2}-\left(D+i\omega(\bar{\alpha}+\alpha_{od})\right)\left(C+i\omega(\bar{\alpha}-\alpha_{od}\cos 2\phi)\right)}, (27)
χθθ\displaystyle\chi^{-}_{\theta\theta} =C+iω(α¯αodcos2ϕ)ω2(D+iω(α¯+αod))(C+iω(α¯αodcos2ϕ)),\displaystyle=\frac{C+i\omega(\bar{\alpha}-\alpha_{od}\cos 2\phi)}{\omega^{2}-\left(D+i\omega(\bar{\alpha}+\alpha_{od})\right)\left(C+i\omega(\bar{\alpha}-\alpha_{od}\cos 2\phi)\right)}, (28)
χϕθ\displaystyle\chi^{-}_{\phi\theta} =ωω2(D+iω(α¯+αod))(C+iω(α¯αodcos2ϕ)).\displaystyle=\frac{\omega}{\omega^{2}-\left(D+i\omega(\bar{\alpha}+\alpha_{od})\right)\left(C+i\omega(\bar{\alpha}-\alpha_{od}\cos 2\phi)\right)}. (29)

The second contribution in Eqs. (23) and (24) is linear in the damping Δα¯\Delta\bar{\alpha}, with the 2×22\times 2 matrices,

𝕄+1𝕋𝕄1\displaystyle\mathbb{M}^{-1}_{+}\mathbb{T}\mathbb{M}^{-1}_{-} =ωΔαdet[𝕄+]det[𝕄](i(BD+ω2)(B+C)ω(A+D)ωi(AC+ω2)),\displaystyle=\frac{\omega\Delta\alpha}{\det\left[\mathbb{M}_{+}\right]\det\left[\mathbb{M}_{-}\right]}\left(\begin{array}[]{cc}i\left(BD+\omega^{2}\right)&\left(B+C\right)\omega\\ \left(A+D\right)\omega&-i\left(AC+\omega^{2}\right)\end{array}\right), (32)
𝕄1𝕋𝕄+1\displaystyle\mathbb{M}^{-1}_{-}\mathbb{T}\mathbb{M}^{-1}_{+} =ωΔαdet[𝕄+]det[𝕄](i(BD+ω2)(A+D)ω(B+C)ωi(AC+ω2)).\displaystyle=\frac{\omega\Delta\alpha}{\det\left[\mathbb{M}_{+}\right]\det\left[\mathbb{M}_{-}\right]}\left(\begin{array}[]{cc}i\left(BD+\omega^{2}\right)&\left(A+D\right)\omega\\ \left(B+C\right)\omega&-i\left(AC+\omega^{2}\right)\end{array}\right). (35)

I.2 Two-sublattice Spin Pumping

In this section we evaluate the spin pumping in the two-sublattice magnet. Final expressions for the injected spin currents are found in terms of the dynamical magnetic susceptibility. To start with, let us consider the spin pumping current into the normal metal given by Eq. (3). The fluctuations are represented by 𝒎ζ=𝒎ζeq+δ𝓜ζ{\boldsymbol{m}}_{\zeta}={\boldsymbol{m}}^{\text{eq}}_{\zeta}+\delta{\boldsymbol{\cal M}}_{\zeta}, with the real-valued fields defined as δ𝓜A=δA,θ𝒛^+δA,ϕ(𝒛^×𝒎Aeq)\delta{\boldsymbol{\cal M}}_{A}=\delta{\cal M}_{A,\theta}\hat{\boldsymbol{z}}+\delta{\cal M}_{A,\phi}\left(\hat{\boldsymbol{z}}\times{\boldsymbol{m}}^{\text{eq}}_{A}\right) and δ𝓜B=δB,θ𝒛^+δB,ϕ(𝒛^×𝒎Beq)\delta{\boldsymbol{\cal M}}_{B}=\delta{\cal M}_{B,\theta}\hat{\boldsymbol{z}}+\delta{\cal M}_{B,\phi}\left(\hat{\boldsymbol{z}}\times{\boldsymbol{m}}^{\text{eq}}_{B}\right). We find that the spin pumping current 𝐈s{\bf I}_{s} becomes

4π𝐈s=gAAδ𝓜A×δ𝓜˙A+gAB(δ𝓜A×δ𝓜˙B+δ𝓜B×δ𝓜˙A)+gBBδ𝓜B×δ𝓜˙B.\displaystyle\frac{4\pi}{\hbar}{\bf I}_{s}=g_{AA}\langle\delta{\boldsymbol{\cal M}}_{A}\times\delta\dot{\boldsymbol{\cal M}}_{A}\rangle+g_{AB}\left(\langle\delta{\boldsymbol{\cal M}}_{A}\times\delta\dot{\boldsymbol{\cal M}}_{B}\rangle+\langle\delta{\boldsymbol{\cal M}}_{B}\times\delta\dot{\boldsymbol{\cal M}}_{A}\rangle\right)+g_{BB}\langle\delta{\boldsymbol{\cal M}}_{B}\times\delta\dot{\boldsymbol{\cal M}}_{B}\rangle. (36)

Note that since a time-average is involved in the evaluation of previous equation, linear terms in the fluctuations do not contribute. In order to relate each term in Eq. (36) with the dynamical susceptibility, Eqs. (25)-(29), we represent δ𝓜ζ\delta{\boldsymbol{\cal M}}_{\zeta} by complex-valued fields as δζ,μ=Re[δmζ,μeiωt]\delta{{\cal M}}_{\zeta,\mu}=\text{Re}\left[\delta{m}_{\zeta,\mu}e^{i\omega t}\right], where μ={θ,ϕ}\mu=\{\theta,\phi\}. Next, we write the fields δmζ,μ\delta{m}_{\zeta,\mu} in the eigenbasis by the following relations δmA/B,θ=12(δm±,θ±δm,θ)\delta m_{A/B,\theta}=\frac{1}{2}\left(\delta m_{\pm,\theta}\pm\delta m_{\mp,\theta}\right) and δmA/B,ϕ=12(δm±,ϕ±δm,ϕ)\delta m_{A/B,\phi}=\frac{1}{2}\left(\delta m_{\pm,\phi}\pm\delta m_{\mp,\phi}\right). Therefore, we obtain

δ𝓜A×δ𝓜˙A=ω4(Im[δm+,ϕδm+,θ]+Im[δm,ϕδm,θ]+Im[δm,ϕδm+,θ]+Im[δm+,ϕδm,θ])𝒎Aeq,\displaystyle\langle\delta{\boldsymbol{\cal M}}_{A}\times\delta\dot{\boldsymbol{\cal M}}_{A}\rangle=\frac{\omega}{4}\left(\text{Im}\left[\langle\delta m_{+,\phi}\delta m^{*}_{+,\theta}\rangle\right]+\text{Im}\left[\langle\delta m_{-,\phi}\delta m^{*}_{-,\theta}\rangle\right]+\text{Im}\left[\langle\delta m_{-,\phi}\delta m^{*}_{+,\theta}\rangle\right]+\text{Im}\left[\langle\delta m_{+,\phi}\delta m^{*}_{-,\theta}\rangle\right]\right){\boldsymbol{m}}^{\text{eq}}_{A}, (37)
δ𝓜B×δ𝓜˙B=ω4(Im[δm,ϕδm,θ]+Im[δm+,ϕδm+,θ]Im[δm,ϕδm+,θ]Im[δm+,ϕδm,θ])𝒎Beq,\displaystyle\langle\delta{\boldsymbol{\cal M}}_{B}\times\delta\dot{\boldsymbol{\cal M}}_{B}\rangle=\frac{\omega}{4}\left(\text{Im}\left[\langle\delta m_{-,\phi}\delta m^{*}_{-,\theta}\rangle\right]+\text{Im}\left[\langle\delta m_{+,\phi}\delta m^{*}_{+,\theta}\rangle\right]-\text{Im}\left[\langle\delta m_{-,\phi}\delta m^{*}_{+,\theta}\rangle\right]-\text{Im}\left[\langle\delta m_{+,\phi}\delta m^{*}_{-,\theta}\rangle\right]\right){\boldsymbol{m}}^{\text{eq}}_{B}, (38)
δ𝓜A×δ𝓜˙B+δ𝓜B×δ𝓜˙A=ω2cosϕ(Im[δm+,ϕδm,θ]Im[δm,ϕδm+,θ])𝒙^\displaystyle\langle\delta{\boldsymbol{\cal M}}_{A}\times\delta\dot{\boldsymbol{\cal M}}_{B}\rangle+\langle\delta{\boldsymbol{\cal M}}_{B}\times\delta\dot{\boldsymbol{\cal M}}_{A}\rangle={\frac{\omega}{2}}\cos\phi\left(\text{Im}\left[\langle\delta m_{+,\phi}\delta m^{*}_{-,\theta}\rangle\right]-\text{Im}\left[\langle\delta m_{-,\phi}\delta m^{*}_{+,\theta}\rangle\right]\right)\hat{\boldsymbol{x}}
+ω2sinϕ(Im[δm,ϕδm,θ]Im[δm+,ϕδm+,θ])𝒚^+ω2sin(2ϕ)Im[δm+,ϕδm,ϕ]𝒛^.\displaystyle\qquad\qquad\qquad\qquad+{\frac{\omega}{2}}\sin\phi\left(\text{Im}\left[\langle\delta m_{-,\phi}\delta m^{*}_{-,\theta}\rangle\right]-\text{Im}\left[\langle\delta m_{+,\phi}\delta m^{*}_{+,\theta}\rangle\right]\right)\hat{\boldsymbol{y}}+{\frac{\omega}{2}}\sin(2\phi)\text{Im}\left[\langle\delta m_{+,\phi}\delta m^{*}_{-,\phi}\rangle\right]\hat{\boldsymbol{z}}. (39)

Below, we present examples of spin pumping current, evaluating Eqs. (37-39), due to the coherent magnonic excitation by linearly polarized external ac fields.

I.2.1 Exciting the Optical and Acoustic Modes

According to the selection rules described in the main text, excitation of the optical mode requires a field that satisfy 𝒉=𝟎{\boldsymbol{h}}_{-}={\boldsymbol{0}} and h+,ϕ0h_{+,\phi}\neq 0. On the other hand, excitation of the acoustic mode requires 𝒉+=0{\boldsymbol{h}}_{+}={0} and 𝒉0{\boldsymbol{h}}_{-}\neq 0. From Eqs. (23) and (24), we find the fields δ𝒎±\delta{\boldsymbol{m}}_{\pm} when the ac magnetic field is 𝒉+=hy𝒚^{\boldsymbol{h}}_{+}=h_{y}\hat{\boldsymbol{y}} (optical mode) and 𝒉=hx𝒙^{\boldsymbol{h}}_{-}=h_{x}\hat{\boldsymbol{x}} (acoustic mode). To determine the spin current, we evaluate each term given by the Eqs. (39)-(38). Up to linear order in the Gilbert damping difference Δα¯\Delta\bar{\alpha} , we find

e𝐈s±=ω4Im[δm±,ϕδm±,θ]\displaystyle\frac{e}{\hbar}{\bf I}^{\pm}_{s}=\frac{\omega}{4}\text{Im}\left[\langle\delta m_{\pm,\phi}\delta m^{*}_{\pm,\theta}\rangle\right] [(gAAgBB)cosϕ𝒙^+((gAA+gBB)2gAB)sinϕ𝒚^]\displaystyle\left[\left(g_{AA}-g_{BB}\right)\cos\phi\hat{\boldsymbol{x}}+\left(\left(g_{AA}+g_{BB}\right)\mp 2g_{AB}\right)\sin\phi\hat{\boldsymbol{y}}\right]
+ω2Im[δm+,ϕδm,ϕ]±gABsin(2ϕ)𝒛^.\displaystyle\qquad\qquad\qquad\qquad\qquad+\frac{\omega}{2}\text{Im}\left[\langle\delta m_{+,\phi}\delta m^{*}_{-,\phi}\rangle\right]_{\pm}g_{AB}\sin(2\phi)\hat{\boldsymbol{z}}. (40)

The functions ±j(ω)\mathscr{F}^{j}_{\pm}(\omega) introduced in the main text, are thus defined as

±x(ω)\displaystyle\mathscr{F}^{x}_{\pm}(\omega) =ω4Im[δm±,ϕδm±,θ]cosϕ,\displaystyle=\frac{\omega}{4}\text{Im}\left[\langle\delta m_{\pm,\phi}\delta m^{*}_{\pm,\theta}\rangle\right]\cos\phi, (41)
±y(ω)\displaystyle\mathscr{F}^{y}_{\pm}(\omega) =ω4Im[δm±,ϕδm±,θ]sinϕ,\displaystyle=\frac{\omega}{4}\text{Im}\left[\langle\delta m_{\pm,\phi}\delta m^{*}_{\pm,\theta}\rangle\right]\sin\phi, (42)
±z(ω)\displaystyle\mathscr{F}^{z}_{\pm}(\omega) =ω2Im[δm+,ϕδm,ϕ]±sin(2ϕ).\displaystyle=\frac{\omega}{2}\text{Im}\left[\langle\delta m_{+,\phi}\delta m^{*}_{-,\phi}\rangle\right]_{\pm}\sin(2\phi). (43)

The explicit expression for each contribution in Eq. (I.2.1) obeys

Im[δm+,ϕδm+,θ]\displaystyle\text{Im}\left[\langle\delta m_{+,\phi}\delta m^{*}_{+,\theta}\rangle\right] =ωBh+,ϕ2|det[𝕄+]|2,\displaystyle=\frac{\omega Bh^{2}_{+,\phi}}{\left|\det\left[\mathbb{M}_{+}\right]\right|^{2}}, (44)
Im[δm,ϕδm,θ]\displaystyle\text{Im}\left[\langle\delta m_{-,\phi}\delta m^{*}_{-,\theta}\rangle\right] =ωDh,ϕ2|det[𝕄]|2,\displaystyle=\frac{\omega Dh^{2}_{-,\phi}}{\left|\det\left[\mathbb{M}_{-}\right]\right|^{2}}, (45)
Im[δm+,ϕδm,ϕ]+\displaystyle\text{Im}\left[\langle\delta m_{+,\phi}\delta m^{*}_{-,\phi}\rangle\right]_{+} =ΔαωB(BD+ω2)(ω2ω2)h+,ϕ2|det[𝕄+]|2|det[𝕄]|2,\displaystyle=\Delta\alpha\frac{\omega B\left(BD+\omega^{2}\right)\left(\omega^{2}_{-}-\omega^{2}\right)h^{2}_{+,\phi}}{\left|\det\left[\mathbb{M}_{+}\right]\right|^{2}\left|\det\left[\mathbb{M}_{-}\right]\right|^{2}}, (46)
Im[δm+,ϕδm,ϕ]\displaystyle\text{Im}\left[\langle\delta m_{+,\phi}\delta m^{*}_{-,\phi}\rangle\right]_{-} =ΔαωD(BD+ω2)(ω2ω+2)h,ϕ2|det[𝕄+]|2|det[𝕄]|2.\displaystyle=\Delta\alpha\frac{\omega D\left(BD+\omega^{2}\right)\left(\omega^{2}-\omega^{2}_{+}\right)h^{2}_{-,\phi}}{\left|\det\left[\mathbb{M}_{+}\right]\right|^{2}\left|\det\left[\mathbb{M}_{-}\right]\right|^{2}}. (47)