Abstract
The search for new elementary particles is one of the most basic pursuits in physics, spanning from subatomic physics to quantum materials. Magnons are the ubiquitous elementary quasiparticle to describe the excitations of fully-ordered magnetic systems. But other possibilities exist, including fractional and multipolar excitations. Here, we demonstrate that strong quantum interactions exist between three flavors of elementary quasiparticles in the uniaxial spin-one magnet FeI2. Using neutron scattering in an applied magnetic field, we observe spontaneous decay between conventional and heavy magnons and the recombination of these quasiparticles into a super-heavy bound-state. Akin to other contemporary problems in quantum materials, the microscopic origin for unusual physics in FeI2 is the quasi-flat nature of excitation bands and the presence of Kitaev anisotropic magnetic exchange interactions.
Similar content being viewed by others
Introduction
The concept of quasiparticles is central to understand and predict the properties of condensed matter. For example, the quantization of collective atomic vibrations and spin precessions in long-range ordered solids1 leads to the familiar concepts of phonons and magnons. When motion is harmonic, these bosonic excitations are free2 and manifest in spectroscopic measurements as bands with well-defined energy-momentum dispersion. Interactions between phonons underpin many basic phenomena ranging from the anharmonic behavior of crystals and the lattice conductivity of thermoelectrics3 to the rich excitation spectrum of liquid 4He4. In magnetism, interactions between magnons5 can yield finite lifetimes by spontaneous (non-thermal) decay into multi-magnon states6, resulting in incoherent excitation bands. Magnon decay is reminiscent of elementary particle decay, a ubiquitous quantum phenomenon of the Standard Model of subatomic physics. Although magnon instabilities are expected for a broad class of models7,8,9, their experimental observation is rare and so far limited to a handful of quantum paramagnets10,11 and non-collinear spin systems12,13,14,15,16. As the search for quantum spin-liquids and their fractional excitations intensifies17, achieving a quantitative understanding of magnon interactions is a pressing issue18,19.
Unlike the Standard Model, where all elementary particles emerge from a single vacuum, a rich quasiparticle landscape20 arises from distinct vacua (ground states) in the innumerable magnetic solids. In this context, it is surprising that decay instabilities have only been investigated in detail when magnon quasiparticles and their decay products carry the same fundamental quantum spin number: ΔSz = ± 1 about the local quantization axis imposed by the underlying magnetic order. A counter-intuitive framework to realize strong magnon interactions is a system with large spin (S ≥ 1) and strong single-ion and magnetic exchange anisotropies. We illustrate this concept in Fig. 1 using a ferromagnetic spin array. A S = 1/2 system (Fig. 1a) only admits single magnons (SMs) as elementary excitations, which carry a dipolar quantum number ∣ΔSz∣ = 1. Composite excitations of multiple (free or bound) SMs are possible, but they are not elementary; understanding their interactions often requires a non-perturbative treatment. To date, the vast majority of magnon decay studies have focused on the interactions between SMs and their multi-particle states. For a S = 1 system (Fig. 1b), the enlarged local Hilbert space yields a second type of on-site excitation where the same spin is flipped twice, from Sz = + 1 to − 1. This excitation is known as “single-ion bound-state” (SIBS) and becomes a distinct elementary quasiparticle, with quantum number ∣ΔSz∣ = 2, for strongly uniaxial systems21,22. Naturally, composite excitations formed by multiple SMs and/or SIBSs are also possible, for instance four-magnon bound-states (4MBS) comprising two SIBS bound by short-range ferromagnetic exchange interactions (see Fig. 1b-right). SIBS and their composite excitations are fundamentally different from SMs as they carry a multipolar quantum number ∣ΔSz∣ = 2, 3, 4, … and form quasi-flat bands that are in principle invisible to spectroscopic tools restricted by the dipole selection rule. An opportunity to observe these excitations, however, stems from their possible hybridization with conventional quasiparticles such as phonons or (single) magnons. The former mechanism is realized in UO223, while the latter was recently uncovered and understood for FeI224, which is the subject of this work.
FeI2 is a quasi-2D Van der Waals material that comprises perfect ab-plane triangular layers of Fe2+ ions surrounded by I− ligands (see “Methods” for a detailed description). At low temperature, the Fe2+ ions bear effective Seff = 1 magnetic moments with an easy-axis anisotropy D ≈ 2 meV25 along the crystallographic c axis. Magnetic exchange interactions are frustrated26 with a ferromagnetic nearest-neighbor coupling J1 ≈ − 0.1D competing with weaker antiferromagnetic further-neighbor exchanges within and between the triangular planes24 (see Supplementary Fig. 1). This competition stabilizes a striped antiferromagnetic (AF) order below TN = 9.5 K26,27, with rows of ferromagnetically aligned spins arranged in ↑↑↓↓ domains within the triangular layers, each with four magnetic sublattices (see Fig. 1e and Supplementary Fig. 2). At T = 1.8 K, the AF phase is stable up to a magnetic field of μ0H1 = 4.8 T before evolving into a complex sequence of ferrimagnetic phases below magnetic saturation at μ0Hsat = 12.5 T26. Early neutron spectroscopy experiments28 in the AF phase of FeI2 elucidated that SIBS excitations, previously identified by infrared spectroscopy29, form quasi-flat bands that lie below SM branches throughout the Brillouin zone. Recent quantitative studies24,30 have demonstrated that off-diagonal components of the nearest-neighbor exchange interaction are responsible for the high degree of hybridization between overlapping dipolar (SM) and multipolar (SIBS, 4MBS, …) excitations. These interactions can be parameterized using the spin-non-conserving terms \({S}_{i}^{z}{S}_{j}^{+}\) and \({S}_{i}^{+}{S}_{j}^{+}\) of energy scale J z± ≈ 1.1J1 and J ±± ≈ 0.7J1, respectively, or using an extended Kitaev-Heisenberg model31. In this context, all observed magnetic excitations in FeI2 are hybrid between dipolar and multipolar quasiparticles; for simplicity, we will refer to them according to the character of their dominant quasiparticle.
Results
The delicate balance of microscopic interactions and anisotropies in FeI2 leads to a unique situation where distinct elementary quasiparticles, and their bound-states, overlap in momentum-energy space and strongly hybridize. This renormalizes their dispersion curves producing light (wide band) and heavy (narrow-band) quasiparticles (Fig. 1c). Given the regime of strong quantum interactions in FeI2, it is natural to wonder if spontaneous decays are also possible. The most straightforward mechanism is through the spin-non-conserving exchange interactions because these activate cubic decay processes (Fig. 1d). For example, the \({J}^{\pm \pm }{S}_{i}^{+}{S}_{j}^{+}\) term connects initial and final states whose quantum spin numbers differ by two. But two additional conditions are necessary to observe spontaneous decay. First, the six decay vertices of Fig. 1d must connect initial one-quasiparticle states to final two-quasiparticle states with a non-zero matrix element. As two (resp. three) combinations of quasiparticles exist for the initial (resp. final) states, this opens up many distinct decay channels. Second, decays must obey the conservation of total energy and crystal momentum. The fulfillment of these kinematic conditions depends on details of the excitation spectra amenable to external control, for instance, with a magnetic field. Kinematic decay conditions are not met in FeI2 in the absence of a magnetic field as all multi-particle continua are just above the dominant excitation branches. However, the relative Zeeman shift between initial and final states with different quantum spin numbers can, as we will observe below, overcome this discrepancy and activate decay processes for an adequate magnetic field range (Fig. 1f).
To search for quasiparticle decay in FeI2, we apply a magnetic field perpendicular to the triangular planes to tune the relative position of magnetic excitations within the AF phase (Fig. 1f) and examine the resulting momentum- and energy-resolved response using inelastic neutron scattering (see Methods). A slight misalignment between the magnetic field direction and the c-axis of our high-quality multi-gram crystal selects a single magnetic domain (see Supplementary Fig. 3, which dramatically simplifies interpretation of our results. In Fig. 2, we compare neutron-scattering data for μ0H = 0, 1, 2, 3 and 4 T with SU(3)-generalized linear spin-wave (GLSW) calculations32 for the exchange interactions of ref. 24 and g = 3.8(5) (see Table 1). For μ0H ≤ 2 T, the number, dispersion, intensity, linewidth, and field-dependence of all the observed modes are in excellent agreement with GLSW predictions for all measured momenta (Fig. 2a, see also Supplementary Figs. 4–5 for more cuts). These spectra reflect the magnetic field evolution of eight modes: a SM and a SIBS for each of the four magnetic sublattices. Half of the modes have weak intensity for the momenta shown in Fig. 2. Excitations of the spin-down ferromagnetic stripes, corresponding to ΔSz = + 1 and + 2, experience a negative Zeeman shift ΔEZeeman = − gμBμ0HΔSz, and vice-verse for excitations of the spin-up stripes. The enhanced splitting of the ΔSz = ± 2 magnon bound-states enables their unambiguous spectroscopic identification33,34, see arrows on Fig. 2a, b.
While all excitation branches are sharp below μ0H ≤ 3 T, a striking deviation from GLSW predictions is observed at μ0H = 4 T where the single-ion bound-state (SIBS) broadens considerably in the middle of the Brillouin zone, see yellow box in Fig. 2b. The line cut in Fig. 2c confirms the significant energy broadening of the SIBS peak at 3.76(1) meV with a full-width at half maximum (FWHM) of 0.38(1) meV, and reveals an anomalous energy width of 0.57(1) meV for the proximate single-magnon excitation at 4.41(3) meV, see Supplementary Table 1 for fit results. In line cuts for other momenta and fields, all branches appear resolution-limited with a FWHM of ≈ 0.20 meV. We tentatively ascribe these characteristic features to the activation of decay processes for both the SM and SIBS quasiparticles.
Further evidence for strong magnon interactions in FeI2 comes from the observation of four-magnon (4MBS) and six-magnon bound-states (6MBS) in magneto-optics30. These higher-order exchange bound-states are stabilized by the narrow-band of the system and the presence of ferromagnetic interactions at short distances in a given stripe of the underlying magnetic structure (Fig. 1e). 4MBS excitations are clearly visible in our neutron-scattering data at μ0H = 3 T as weak and non-dispersing modes, unaccounted for by GLSW. Of particular interest is the 2.5 meV mode indicated by a white arrow in Fig. 2b: it lies below the SM branch observed at 3.0 meV in Fig. 2c but predicted by GLSW at 2.8 meV, indicative of mode repulsion (see Supplementary Fig. 4 for more examples of this behavior). At μ0H = 4 T, the 4MBS excitation moves down in energy but the shift of the SM peak persists. In spite of their hexadecapolar nature (ΔSz = 4), 4MBS excitations are detected in our experiment because of their strong hybridization to dipolar fluctuations30: given their impact on the SM branch they must be treated on equal footing as a distinct quasiparticle flavor.
To explain the anomalous mode broadening uncovered by our experiments in finite magnetic field, we refine our previous SU(3)-generalized spin-wave theory using a perturbative expansion that accounts for quasiparticle interactions at the one-loop level. To capture the hybridization, energy renormalization, and decay rate of the SM and SIBS excitations, it is sufficient to retain cubic interaction vertices that couple the one- and two-quasiparticle sectors, i.e., we drop the negligible contribution from quartic vertices, see Methods for full details. The essential results of these non-linear calculations (GNLSW) are shown for μ0H = 3 T and 4 T in Fig. 3a, b. The gray and colored regions indicate the continua of allowed energies and momenta for each possible combination of two unbound SM or SIBS quasiparticles. Decays are kinematically allowed where a given excitation branch overlaps with one or several of these shaded regions, with the larger decay rates (red shading in Fig. 3a, b central panels) originating from the colored continua.
For μ0H = 4 T, Fig. 3b, our GNLSW calculations predict large decay rates for the top of the E4 and E6 bands (see band labeling in Fig. 3). This yields a broadened neutron-scattering response highlighted by the dashed yellow box in Fig. 3b, in excellent agreement with our experimental observations. While the hybrid character of all the bands is fully accounted for in our quantitative decay rate calculations, it is instructive to focus on their dominant character at a given wave-vector to elucidate their decay mechanism. The broadening of band E4 around 3.8 meV stems from the emission of a ΔSz = + 2 SIBS on branch E2 by a ΔSz = + 1 SM that correspondingly looses energy and momentum. The broadening observed around 4.4 meV for band E6, corresponds to a ΔSz = − 2 SIBS decaying into two SM excitations: one at the bottom of the E2 band with ΔSz = + 1 and one at a different wave-vector of the E6 band where the ΔSz = − 1 character dominates. These decay processes correspond to the spontaneous creation and annihilation of a single-ion bound-state through a net change of two units of angular momentum, implying that the relevant interaction vertices are mediated by the anisotropic, spin-non-conserving term \({J}^{\pm \pm }{S}_{i}^{+}{S}_{j}^{+}\), see vertices with a gray background in Fig. 1d. Although this mechanism produces a finite lifetime for all excitation branches in the AF phase of FeI2, the broadening only becomes visible in experiments when the decay rate exceeds the FWHM instrumental energy-resolution of around 0.2 meV, see Fig. 3d, Supplementary Fig. 6, and Supplementary Table 1. Thus spontaneous decays are only visible in our experiment in a narrow field range around μ0H = 4 T, due to the narrow bandwidth of the lowest-energy branch essential to the decay processes.
Surprisingly, a qualitatively different phenomenon occurs for μ0H = 3 T, Fig. 3a. While our GNSLW calculations predict strong decay for excitations at the bottom of the E4 band, no visible broadening is observed in the experimental results of Fig. 2b, c. Instead, the putative unstable branch lies proximate to the 4MBS excitation discussed previously. Including this composite quasiparticle in our GNLSW calculations is impractical as it requires to sum ladder diagrams to infinite order in a perturbative loop expansion. We avoid this problem by performing an exact diagonalization (ED) of the SU(3) spin-wave Hamiltonian at quartic order on a finite lattice. Truncating the Hilbert space to only include up to two (free or bound) composite quasiparticles allows to reach adequate system sizes (see Methods). The quartic term is essential to form a 4MBS from the continuum of two free SIBSs. Results without and with quartic vertices, Fig. 3c, explain the strong suppression of decays observed in our experiments as stemming from the finite probability of decay products to form a 4MBS instead of propagating independently in the system. At the microscopic level, this non-perturbative phenomenon, which we observe and understand for the first time, comes from the unique interplay between heavy SIBS quasiparticles, attractive (ferromagnetic) interaction at short distances, and spin-non-conserving terms.
Discussion
Our neutron-scattering experiments on FeI2 reveal a rich and field-tunable quantum many-body physics phenomenology that is quantitatively explained by our theory. We observe three distinct flavors of quasiparticles: light dipolar SM fluctuations, heavy quadrupolar SIBS quasiparticles, and super-heavy hexadecapolar 4MBS excitations (Fig. 1b) stabilized by attractive short-range interactions. These quasiparticles mix, decay and pair onto each other in a way reminiscent of high-energy particle physics. Our observations of spontaneous emission of a heavy quasiparticle by a magnon, the decay of the former into two free magnons, and the suppression of decay channels by the non-perturbative recombination of decay products into super-heavy bound-states, are observed for the first time in the realm of condensed-matter systems. Within magnetism, our work challenges the conventional view that compounds with large spin and large uniaxial anisotropy behave classically. In fact, in FeI2, this combination produces unique quantum magnon dynamics brought to light by spin-non-conserving off-diagonal exchange interactions. As such interactions are also essential to stablize quantum spin-liquids and their fractionalized excitations in Kitaev magnets18, our work considerably broadens the range of quasiparticle phenomena expected in these spin-orbit coupled quantum magnets. The theoretical tools we have developed to understand FeI2 apply to many other materials16 and may be used in the future to sharpen our general understanding of large-spin magnets.
Methods
Crystal growth
Starting materials of Iron (≥99.98% purity from Alfa Aesar) and Iodine (≥99.99% purity from Alfa Aesar) were sealed in evacuated quartz tubes. The first synthesis step consists in a chemical vapor transport growth using a tube furnace (School of Physics, Georgia Tech) with the hot end at 570 °C and the cold end at room temperature35 forming a collection of mm-size FeI2 crystals. After grinding into fine powders in a glove box with water content ≤2 ppm, the crystal structure was checked using a PANAnalytical Empyrean Cu-Kα diffractometer (Materials Characterization Facility, Georgia Tech) with samples loaded in an air-tight domed holder in the glove box. This confirmed the expected crystal structure and was consistent with results reported in ref. 24. Around 10 grams of the polycrystalline sample was sealed in a quartz tube under vacuum. The ampule was then placed in a graphite crucible attached to a rotator in a Ultrahigh Temperature Midscale Induction Bridgman/CZ Furnace (PARADIM facility, Johns Hopkins University). The crucible was passed through a hot zone of ≈ 600 °C with rotating speed 20 RPM/min and lowering rate 10 mm/hr. The growth yielded a large boule from which a 4.53 g high-quality crystal was extracted with clear c-axis facet. The resulting crystal was mounted on an aluminum sample holder sealed in a Helium-filled glove box. The holder was designed to keep the sample from moisture and oxygen contamination and had a small enough diameter to fit in the ⊘32 mm diameter of a cryomagnet. The mosaic of the crystal was checked with neutrons to be around ≤3°.
Thermo-magnetic properties of FeI2
FeI2 crystallizes in the space-group \(P\bar{3}m1\) with lattice parameters a = 4.05 Å and c = 6.75 Å at T = 300 K36. FeI2 comprises triangular layers of Fe2+ ions with magnetic interactions mediated by direct exchanges and super-exchanges through the I− ligands above and below the triangular plane. The combination of crystal-field and spin-orbit coupling effects on the Fe2+ ions leads to effective S = 1 magnetic moments with an easy-axis anisotropy along the c-axis and several transitions to higher-energy multiplets above 25 meV37. In zero magnetic field, FeI2 displays a long-range magnetic order below TN = 9.5 K26,27 through a first order transition with no apparent lattice distortion27,38. The magnetic structure is described by a propagation vector kAF = (0, 1/4, 1/4) and the phase referred to as the “AF" phase given the absence of net magnetization. Within the triangular plane, the system forms a up-up-down-down stripe order shown in Supplementary Fig. 2.
Three types of magnetic domains are typically stabilized in zero magnetic field, related by 120° rotations24 with propagation vectors \({{{{{{\boldsymbol{k}}}}}}}_{{{{{{\rm{AF}}}}}}}^{(1)}=(0,1/4,1/4),\, {{{{{{\boldsymbol{k}}}}}}}_{{{{{{\rm{AF}}}}}}}^{(2)}=(-1/4,0,1/4)\) and \({{{{{{\boldsymbol{k}}}}}}}_{{{{{{\rm{AF}}}}}}}^{(3)}=(1/4,-1/4,1/4)\). When a magnetic field is applied along the crystallographic c-axis, a series of meta-magnetic transitions were observed in bulk magnetization measurements39,40. Associated magnetic structures were investigated using neutron diffraction26 below the saturation magnetic field of μ0Hs ≈ 12.5 T39. Below T ≈ 2 K, the first magnetic transition occurs above μ0H1 ≥ 4.5 T. The results presented here are restricted to magnetic fields μ0H ≤ 4 T and temperatures T ≤ 2 K, such that the underlying magnetic structure for FeI2 is AF. As explained in the main text and shown in Supplementary Fig. 3, this was checked by taking elastic cuts through the neutron-scattering data, which also revealed that a predominantly single-domain magnetic state, corresponding to \({{{{{{\boldsymbol{k}}}}}}}_{{{{{{\rm{AF}}}}}}}^{(1)}\), was stabilized in the sample.
Neutron-scattering measurements
Inelastic neutron-scattering experiments were performed on the HYSPEC spectrometer at the Spallation Neutron Source (SNS), Oak Ridge National Laboratory (ORNL), USA41. The sample was mounted on a stick inserted in a \({\mu }_{0}{H}_{\max }=8\) T vertical-field self-shielded superconducting magnet reaching a base temperature around T = 1.8 K. The sample was rotated around its c-axis over a range of 360° degrees in steps of 1° degree allowing a complete mapping of excitations in the scattering plane. The narrow out-of-plane coverage of ± 7° degrees of the magnet restricts the momentum transfer in the out-of-plane direction. All measurements were performed in unpolarized mode with an incoming neutron energy of Ei = 9 meV and Fermi choppers speed of 420 Hz yielding an elastic full-width at half-maximum energy-resolution on the sample of 0.20 meV. The center detector bank was positioned at a 2θ angle of − 35°. Five magnetic field configurations were used corresponding to μ0H = 0, 1, 2, 3, and 4 T. When theoretical calculations are compared to experiments, the computed dynamical structure factors take into account all relevant experimental effects including magnetic form factor and neutron dipole factor.
Data analysis
Data was reduced and analyzed in MANTID42 on the SNS analysis cluster at ORNL. Symmetry operations that preserve the single-domain magnetic structure were applied to the data to increase statistics. Throughout the manuscript, the scattering intensity is measured as a function of energy transfer E and momentum transfer Q = ha* + kb* + lc* ≡ (h, k, l) where a*, b* and c* are the primitive vectors of the triangular-lattice reciprocal space and (h, k, l) are Miller indices in reciprocal lattice units. The usual convention that a* and b* make an 60° angle is used, see Supplementary Fig. 1.
Model Hamiltonian
All theoretical calculations were performed using the zero-field exchange Hamiltonian obtained in ref. 24 including an uniaxial single-ion anisotropy \(-D{\sum }_{i}{({S}_{i}^{z})}^{2}\) and exchange interactions up to third-neighbors in plane (J1 to J3) and out-of-plane (\({J}_{0}^{{\prime} }\) to \({J}_{2}^{{\prime} }\)) as defined on Supplementary Figure 1. For nearest-neighbor bonds, all symmetry allowed diagonal and off-diagonal exchange interactions are taken into account, which yields the Hamiltonian
where \({\gamma }_{ij}={e}^{i{\theta }_{ij}}\) are bond-dependent phase factors with \({\theta }_{ij}={\theta }_{ji}=0,+\frac{2}{3}, - \frac{2}{3}\) depending on the direction of the bond of the triangular lattice31.
For further-neighbor bonds, only diagonal anisotropy is considered, which yields the Hamiltonian
for bonds \({J}_{2},{J}_{3},{J}_{0}^{{\prime} },{J}_{1}^{{\prime} }\) and \({J}_{2}^{{\prime} }\).
In this work, we adopt the representative values of exchanges interactions for FeI2 obtained in ref. 24 by joint fits to the zero magnetic-field energy-integrated data in the paramagnetic phase and the energy-resolved data in the magnetically ordered phase.
Although we will not use this notation in the present manuscript, we note that, alternatively, the nearest-neighbor exchange matrix for FeI2 can be recast as an extended Kitaev-Heisenberg (K–J) model31
with J1 = − 0.41 meV, K1 = 0.53 meV, Γ1 = − 0.02 meV, and \({\Gamma }_{1}^{{\prime} }=0.01\) meV.
Finally, a Zeeman term \({{{{{{\mathcal{H}}}}}}}_{{{{{{\rm{Zeeman}}}}}}}=-g{\mu }_{{{{{{\rm{B}}}}}}}{\mu }_{0}HS\) is included to account for the effect of magnetic field. The g-factor is obtained from GLSW fitting to the neutron-scattering data at low fields (μ0H ≤ 2T) with all exchange and single-ion parameters fixed, yielding a value of 3.8(5).
Loop expansion
In zero field, FeI2 has been successfully modeled by a SU(3) GLSW theory24. To explain the field-induced effects studied in this work, one must go beyond the GLSW and consider the interaction between the quasiparticles. Explicitly, we perform a systematic perturbation theory that corresponds to an expansion in the parameter 1/M, where M is the total number of SU(N) bosons per site. Note that this expansion coincides with the well-known 1/S expansion for the particular case N = 2 (M = 2S for N = 2). As will be shown below, the order of a given Feynman diagram of the expansion is determined by the number of independent loops, i.e., of closed lines of SU(N) boson propagators.
To count the order of a given Feynman diagram, it is convenient to rescale the SU(N) boson operator by a factor \(1/\sqrt{M}\), namely \({\beta }_{i,m}={\beta }_{i,m}^{{\prime} }/\sqrt{M}\) (see the next section for explicit definition), m = 1, 2, …, N. Consequently, M becomes an overall prefactor of the rescaled Hamiltonian, \(H={H}^{{\prime} }/M\). Since the original interaction vertices V(n)(n ≥ 3), the coefficients of a triple product of the boson operators, scale as M2−n/2, all vertices of the rescaled Hamiltonian \({H}^{{\prime} }\) becomes of order M, while the propagator of a boson is still of order 1/M. Therefore, the order p of a particular one-particle irreducible diagram constructed by V vertices and I internal lines is V − I (note that the frequency ω is of order M0). Since the number of loops is L = I − V + 1, we obtain that the power p = 1 − L is only determined by the number of loops in a particular Feynman diagram.
Cubic vertex and self-energy
FeI2 is described by an effective S = 1 spin model,
where \({\hat{S}}_{i}^{\mu },\mu=x,y,z\) is the spin-1 operator and the single-ion anisotropy term is proportional to the (zz) component of quadrupolar moment \({{{{{{\mathcal{Q}}}}}}}_{i}^{\mu \nu }=({\hat{S}}_{i}^{\mu }{\hat{S}}_{i}^{\nu }+{\hat{S}}_{i}^{\nu }{\hat{S}}_{i}^{\mu })/2-2/3\) (symmetric traceless components of \({\hat{{{{{{\boldsymbol{S}}}}}}}}_{i}\otimes {\hat{{{{{{\boldsymbol{S}}}}}}}}_{i}\)). The spin-exchange tensor \({{{{{{\mathcal{J}}}}}}}_{ij}^{\mu \nu }\) is described in the main text.
To describe the magnetically ordered phase, it is convenient to work in the local reference frame defined by the SU(3) rotation
where Ui ∈ SU(3). The magnetic order corresponds to a macroscopic occupation of the βi,+1 boson \(\langle {\beta }_{i,+1}\rangle=\langle {\beta }_{i,+1}^{{{\dagger}} }\rangle \simeq \sqrt{M}\) and M = 1 for the case under consideration. Because of the strong single-ion anisotropy, we can safely assume that \({\sum }_{m\ne 1}\langle {\beta }_{i,m}^{{{\dagger}} }{\beta }_{i,m}\rangle \ll M\). This assumption justifies the 1/M expansion that was discussed in the previous section:
The corresponding semi-classical expansion of the dipolar and quadrupolar operators are
where
and \({\tilde{L}}^{\mu }(i)={U}_{i}^{{{\dagger}} }{L}^{\mu }{U}_{i},{\tilde{O}}^{zz}(i)={U}_{i}^{{{\dagger}} }{({L}^{z})}^{2}{U}_{i}\) with the matrices Lμ are the generators of the SO(3) group. The variables defined in Eq. (6) depend only on the sublattice index because of the translational symmetry of the magnetic structure.
By using the expansions (4) and (5), we obtain a generalized semi-classical expansion of the spin Hamiltonian
where \({{{{{{\mathcal{E}}}}}}}^{(0)}\) and \({{{{{{\mathcal{H}}}}}}}^{(2)}\) have been computed explicitly before24. We note that \({{{{{{\mathcal{E}}}}}}}^{(0)}\propto {M}^{2}\) and \({{{{{{\mathcal{H}}}}}}}^{(2)}\propto M\) according to the series expansion of the spin and quadrupole operators in Eqs. (4) and (5). Here, we will focus on the cubic term, \({{{{{{\mathcal{H}}}}}}}^{(3)}\), which is of \({{{{{\mathcal{O}}}}}}(\sqrt{M})\). The one-loop contributions from the quartic term, \({{{{{{\mathcal{H}}}}}}}^{(4)}\), correspond to a simple renormalization of the single-mode dispersion relation (real part of the self-energy), that is obtained by expressing \({{{{{{\mathcal{H}}}}}}}^{(4)}\) in normal ordering. The corresponding one-loop Feynman diagrams involving quartic vertexes that contribute to the single-particle self-energy are: where the lines represent the propagators of the original bosons (before performing the Bogoliubov transformation). We note here that we have neglected these quartic contributions to the self-energy because they turn out to be very small (10−3J, where J represents the energy scale of the dominant exchange interaction) due the the large single-ion anisotropy.
The cubic interaction term is:
where \({V}_{1}^{m}(i,\, j)=-\frac{1}{2}{{{{{{\mathcal{S}}}}}}}_{c}^{\mu }(i){{{{{{\mathcal{J}}}}}}}_{ij}^{\mu \nu }{{{{{{\mathcal{S}}}}}}}_{1m}^{\nu }(\, \, j)\), \({V}_{2}^{lmn}(i,\, j)={{{{{{\mathcal{S}}}}}}}_{1l}^{\mu }(i){{{{{{\mathcal{J}}}}}}}_{ij}^{\mu \nu }{{{{{{\mathcal{S}}}}}}}_{mn}^{\nu }(j)\). Translational invariance implies that \({V}_{1}^{m}(i,\, j)\equiv {V}_{1}^{m}({\alpha }_{i},{{{{{{\boldsymbol{\delta }}}}}}}_{\langle ij\rangle })\) and \({V}_{2}^{lmn}(i,\, j)\equiv {V}_{2}^{lmn}({\alpha }_{i},{{{{{{\boldsymbol{\delta }}}}}}}_{\langle ij\rangle })\) are functions of sublattice and bond. Here, αi is the sublattice index of site i, while δ〈ij〉 the vector that connects sites i and j. After performing the Fourier transform
where (α, r) denotes the lattice site with coordinate r that belongs to sublattice α and Nuc the total number of the magnetic unit cells, we obtain
with
where \(1\le a\le 3,{\sum }_{{\langle ij\rangle }^{{\prime} }}\) sums over translationally inequivalent bonds. The first term of Eq. (11) includes the off-site or bond contributions to the cubic vertex that arise from the exchange interactions. The corresponding vertex function is
where δij = rj − ri and \({\bar{{{{{{\boldsymbol{\delta }}}}}}}}_{ij}\equiv -{{{{{{\boldsymbol{\delta }}}}}}}_{ij}\) refer to the bond vectors. The second term of Eq. (11) includes the on-site contributions to the cubic vertex that arise from the single-ion anisotropy and the Zeeman term. The corresponding vertex function is
The quasiparticle modes are obtained by performing a Bogoliubov transformation
that diagonalizes the linear spin-wave Hamiltonian \({{{{{{\mathcal{H}}}}}}}^{(2)}({{{{{\boldsymbol{q}}}}}})\). Note that this transformation is redundant for ± q, implying that
The triple product of bosonic operators in Eq. (10) becomes
After putting the γ operators in normal ordering and ignoring the linear terms that arise from this process, which is justified because of the strong easy-axis anisotropy, we obtain
where
and
The final form of the cubic interaction is obtained after symmetrization of the vertex:
where
and P the permutation operator.
To compare with the inelastic neutron-scattering data, we compute the dynamical spin structure factor at zero temperature, \({S}_{\mu \nu }({{{{{\bf{q}}}}}},\omega )=2\Theta (\omega ){\chi }_{\mu \nu }^{{\prime\prime} }({{{{{\bf{q}}}}}},\omega )\), where Θ(ω) is the Heaviside step function and \({\chi }_{\mu \nu }^{{\prime\prime} }({{{{{\bf{q}}}}}},\omega )\) is the imaginary part of the dynamical spin susceptibility
where \({S}_{\alpha,{{{{{\boldsymbol{q}}}}}}}^{\mu }={N}_{uc}^{-1/2}{\sum }_{{{{{{\boldsymbol{r}}}}}}}{e}^{-i{{{{{\boldsymbol{q}}}}}}\cdot {{{{{\boldsymbol{r}}}}}}}{S}_{\alpha,{{{{{\boldsymbol{r}}}}}}}^{\mu }\). χμν(q, ω) was evaluated at zero magnetic field before24 at the linear level, i. e., without including the effect of the interaction term \({{{{{{\mathcal{H}}}}}}}^{(3)}\) in Eq. (7). We note that the longitudinal channel of the spin structure factor has only contributions from the two-magnon continuum, which are negligibly small for FeI2 because of the strong single-ion anisotropy. We will then focus on the transverse response and on the effects produced by the interactions between quasiparticles. The key observation is that the kinematic conditions for magnon decay become satisfied for certain ranges of magnetic field values, giving rise to an intrinsic broadening or finite lifetime of the corresponding quasiparticle.
To leading order in 1/M, the dynamical spin susceptibility is given by
where the 2 × 2 block matrix \({{{{{\mathcal{G}}}}}}({{{{{\boldsymbol{q}}}}}},\omega )\) is the interacting single-particle Green’s function determined by the Dyson equation
The non-interacting single-particle Green’s function \({{{{{{\mathcal{G}}}}}}}_{0}({{{{{\boldsymbol{q}}}}}},\omega )\) is given by
where
and I8×8 is the 8 × 8 identity matrix. Given that \({{{{{{\mathcal{H}}}}}}}^{(2)}\) is of order M and the energy scale of interest is ω ∝ (M)1, we obtain that \({{{{{{\mathcal{G}}}}}}}_{0}\) is of order M−1, as it was mentioned in the previous section. The single-particle self-energy, Σ(q, ω), is given by the two one-loop diagrams: that correspond to the self-energy corrections:
and
where εn,k is the linear spin-wave dispersion. Since we are working to the leading order in 1/M, it is enough to just consider diagonal elements where the initial and the final boson belong to the same single-particle band. For each band, the self-energy is evaluated in the on-shell approximation: \(\omega={\varepsilon }_{{n}_{1},{{{{{\boldsymbol{q}}}}}}}\).
Diagonalization
To study the non-perturbative effect, e.g., the avoided-decay of excitations observed at μ0H = 3T, we performed an exact diagonalization (ED) study in the truncated subspace \({{{{{{\mathcal{S}}}}}}}_{1,2}\) with number of quasiparticles ≤2 on a finite lattice of 5 × 5 × 5 unit cells (500 spins). As the Hilbert space dimension becomes prohibitively large for ED if we include states with three quasiparticles, our calculation only accounts for 1-, 2-, and 4-magnon excitations. The excluded states have only perturbative effects because of their higher-energy scales compared to that of a single quasiparticle.
The subspace \({{{{{{\mathcal{S}}}}}}}_{1,2}\) is spanned by the basis \(\{|\left.i\right\rangle,|\left.i \le j \right\rangle \}\) with \(|\left.i\right\rangle={\gamma }_{i}^{{{\dagger}} }|\left.{{\emptyset}}\right\rangle \) and \(|\left.i\le j\right\rangle={\zeta }_{ij}{\gamma }_{i}^{{{\dagger}} }{\gamma }_{j}^{{{\dagger}} }|\left.{{\emptyset}}\right\rangle \), where i stands for the dictionary index of \(({n}_{i},{{{{{{\boldsymbol{q}}}}}}}_{i}),\left|{{\emptyset}}\right\rangle \) refers to the vacuum of the γ-quasiparticles and ζi≠j = 1 and \({\zeta }_{i=j}=1/\sqrt{2!}\) are normalization factors. Introducing the projector \({{{{{{\mathcal{P}}}}}}}_{1,2}\) to the subspace \({{{{{{\mathcal{S}}}}}}}_{1,2}\), the restricted Hamiltonian is obtained by the projection
with matrix elements
Here, the function \({U}_{{n}_{i},{n}_{j},{n}_{k},{n}_{l}}({{{{{{\boldsymbol{q}}}}}}}_{i},{{{{{{\boldsymbol{q}}}}}}}_{j},{{{{{{\boldsymbol{q}}}}}}}_{k},{{{{{{\boldsymbol{q}}}}}}}_{l})\) accounts for the interaction between γ-quasiparticles,
where G is the reciprocal lattice vector. The diagonalization of \({{{{{{\mathcal{P}}}}}}}_{1,2}{{{{{\mathcal{H}}}}}}{{{{{{\mathcal{P}}}}}}}_{1,2}\) is done for a fixed center of mass momentum. It reveals that the spectrum includes four energy levels below the two-particle continuum with a strong 4-magnon character, which are identified as the 4-magnon bound states.
The spectral weights carried by these excitations are revealed by computing the dynamical spin structure factor within the subspace \({{{{{{\mathcal{S}}}}}}}_{1,2}\):
where N = 4Nuc is the total number of lattice sites, and
is the Fourier transform of the spin operators on the sublattice α. The evaluation of this correlation function is carried out by using the continued-fraction method43 based on the Lanczos algorithm44. The lattice size employed here (5 × 5 × 5 unit cells) is large enough to capture the 4-magnon bound states because their linear size is of the order of one lattice space owing to the very large effective mass of the two-magnon bound sates.
Data availability
The raw experimental data are stored on ORNL’s Neutron-Scattering Division computers. The reduced experimental data and the theory dataa are available from the corresponding authors under simple requests.
References
Bloch, F. Zur theorie des ferromagnetismus. Zeitschrift für Physik 61, 206–219 (1930).
Holstein, T. & Primakoff, H. Field dependence of the intrinsic domain magnetization of a ferromagnet. Phys. Rev. 58, 1098–1113 (1940).
Delaire, O. et al. Giant anharmonic phonon scattering in PbTe. Nat. Mater. 10, 614–619 (2011).
Glyde, H. R. Excitations in Liquid and Solid Helium (Oxford University Press, 1994).
Manousakis, E. The spin-1/2 heisenberg antiferromagnet on a square lattice and its application to the cuprous oxides. Rev. Mod. Phys. 63, 1–62 (1991).
Zhitomirsky, M. E. & Chernyshev, A. L. Colloquium: spontaneous magnon decays. Rev. Mod. Phys. 85, 219–242 (2013).
Zhitomirsky, M. E. & Chernyshev, A. L. Instability of antiferromagnetic magnons in strong fields. Phys. Rev. Lett. 82, 4536–4539 (1999).
Chernyshev, A. L. & Zhitomirsky, M. E. Spin waves in a triangular lattice antiferromagnet: decays, spectrum renormalization, and singularities. Phys. Rev. B 79, 144416 (2009).
Mook, A., Klinovaja, J. & Loss, D. Quantum damping of skyrmion crystal eigenmodes due to spontaneous quasiparticle decay. Phys. Rev. Res. 2, 033491 (2020).
Stone, M. B., Zaliznyak, I. A., Hong, T., Broholm, C. L. & Reich, D. H. Quasiparticle breakdown in a quantum spin liquid. Nature 440, 187–190 (2006).
Plumb, K. W. et al. Quasiparticle-continuum level repulsion in a quantum magnet. Nat. Phys. 12, 224–229 (2016).
Oh, J. et al. Spontaneous decays of magneto-elastic excitations in non-collinear antiferromagnet (Y,Lu)MnO3. Nat. Commun. 7, 13146 (2016).
Hong, T. et al. Field induced spontaneous quasiparticle decay and renormalization of quasiparticle dispersion in a quantum antiferromagnet. Nat. Commun. 8, 15148 (2017).
Thompson, J. D. et al. Quasiparticle breakdown and spin hamiltonian of the frustrated quantum pyrochlore Yb2Ti2o7 in a magnetic field. Phys. Rev. Lett. 119, 057203 (2017).
Park, P. et al. Momentum-dependent magnon lifetime in the metallic noncollinear triangular antiferromagnet CrB2. Phys. Rev. Lett. 125, 027202 (2020).
Do, S.-H. et al. Decay and renormalization of a higgs amplitude mode in a quasi-two-dimensional antiferromagnet Nat. Commun. 12, 5331 (2021).
Broholm, C. et al. Quantum spin liquids. Science 367, 6475 (2020).
Winter, S. M. et al. Breakdown of magnons in a strongly spin-orbital coupled magnet. Nat. Commun. 8, 1152 (2017).
Verresen, R., Moessner, R. & Pollmann, F. Avoided quasiparticle decay from strong quantum interactions. Nat. Phys. 15, 750–753 (2019).
Powell, B. J. Emergent particles and gauge fields in quantum matter. Contemp. Phys. 61, 1–36 (2020).
Silberglitt, R. & Torrance Jr, J. B. Effect of single-ion anisotropy on two-spin-wave bound state in a heisenberg ferromagnet. Phys. Rev. B 2, 772 (1970).
Oguchi, T. Theory of two-magnon bound states in the heisenberg ferro-and antiferromagnet. J. Phys. Soc. Jpn 31, 394–402 (1971).
Caciuffo, R. et al. Multipolar, magnetic, and vibrational lattice dynamics in the low-temperature phase of uranium dioxide. Phys. Rev. B 84, 104409 (2011).
Bai, X. et al. Hybridized quadrupolar excitations in the spin-anisotropic frustrated magnet FeI2. Nat. Phys. 17, 467–472 (2021).
Fujita, T., Ito, A. & Ôno, K. The mössbauer study of the ferrous ion in FeI2. J. Phys. Soc. Jpn 21, 1734–1736 (1966).
Wiedenmann, A. et al. A neutron scattering investigation of the magnetic phase diagram of fei2. J. Magn. Magn. Mater. 74, 7–21 (1988).
Gelard, J., Fert, A., Meriel, P. & Allain, Y. Magnetic structure of fei2 by neutron diffraction experiments. Solid State Commun. 14, 187–189 (1974).
Petitgrand, D., Hennion, B. & Escribe, C. Neutron inelastic scattering from magnetic excitations of FeI2. J. Magn. Magn. Mater. 14, 275–276 (1979).
Fert, A. et al. Excitation of two spin deviations by far infrared absorption in FeI2. Solid State Commun. 26, 693–696 (1978).
Legros, A. et al. Observation of 4-and 6-magnon bound-states in the spin-anisotropic frustrated antiferromagnet FeI2. Phys. Rev. Lett. 127, 267201(2020).
Maksimov, P. A., Zhu, Z., White, S. R. & Chernyshev, A. L. Anisotropic-exchange magnets on a triangular lattice: Spin waves, accidental degeneracies, and dual spin liquids. Phys. Rev. X 9, 021017 (2019).
Muniz, R. A., Kato, Y. & Batista, C. D. Generalized spin-wave theory: Application to the bilinear–biquadratic model. Prog. Theor. Exp. Phys. 2014, 083I01 (2014).
Fert, A. et al. Excitation of two spin deviations by far infrared absorption in FeI2. Solid State Commun. 26, 693–696 (1978).
Petitgrand, D., Brun, A. & Meyer, P. Magnetic field dependence of spin waves and two magnon bound states in FeI2. J. Magn. Magn. Mater. 15, 381–382 (1980).
Coleman, C. & Yamada, E. Optimization of the vapor reaction growth of single crystal FeI2. J. Cryst. Growth 132, 129–133 (1993).
Bertrand, Y., Fert, A. & Gelard, J. Susceptibilité magnétique des halogénures ferreux FeCl2, FeBr2, FeI2. J. Phys. 35, 385–391 (1974).
Lockwood, D., Mischler, G. & Zwick, A. Raman scattering from magnons, electronic excitations and phonons in antiferromagnetic FeI2. J. Phys. 6, 6515 (1994).
Trooster, J. & de Valk, W. Spin ordering in FeBr2 and FeI2. evidence for first order phase transition in FeI2. Hyperfine Interact. 4, 457–459 (1978).
Fert, A., Gelard, J. & Carrara, P. Phase transitions of FeI2 in high magnetic field parallel to the spin direction, static field up to 150 koe, pulsed field up to 250 koe. Solid State Commun. 13, 1219–1223 (1973).
Katsumata, K. et al. Phase transition of a triangular lattice ising antiferromagnet FeI2. Phys. Rev. B 82, 104402 (2010).
Stone, M. B. et al. A comparison of four direct geometry time-of-flight spectrometers at the spallation neutron source. Rev. Sci. Instrum. 85, 045113 (2014).
Arnold, O. et al. Mantid–data analysis and visualization package for neutron scattering and μSR experiments. Nucl. Instrum. Method. Phys. Res. Sect. A: Accel. Spectrom. Detect. Assoc. Equip. 764, 156–166 (2014).
Gagliano, E. & Balseiro, C. Dynamical properties of quantum many-body systems at zero temperature. Phys. Rev. Lett. 59, 2999 (1987).
Lanczos, C. An iteration method for the solution of the eigenvalue problem of linear differential and integral operators. J. Res. Natl Bur. Stand. 45, 255–282 (1950).
Acknowledgements
We thank Tyrel McQueen for his help with crystal growth at PARADIM. The work of X.B., Z.L.D., and M.M. at Georgia Tech was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, Materials Sciences and Engineering Division under award DE-SC-0018660. The work of H.Z. at the Oak Ridge National Laboratory was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, Materials Sciences and Engineering Division. The work of S.-S.Z. and C.D.B. at the University of Tennessee was supported by the Lincoln Chair of Excellence in Physics. Growth of FeI2 crystals was supported by the National Science Foundation’s PARADIM (Platform for the Accelerated Realization, Analysis, and Discovery of Interface Materials) under Cooperative Agreement No. DMR-1539918. Some of this work were performed in part at the Materials Characterization Facility at Georgia Tech that is jointly supported by the GT Institute for Materials and the Institute for Electronics and Nanotechnology, which is a member of the National Nanotechnology Coordinated Infrastructure supported by the National Science Foundation under Grant No. ECCS-2025462. This research used resources at the High Flux Isotope Reactor and Spallation Neutron Source, a DOE Office of Science User Facility operated by the Oak Ridge National Laboratory.
Author information
Authors and Affiliations
Contributions
X.B., M.M., and C.D.B. conceived the project. Z.L.D., X.B., and W.A.P grew the sample at the PARADIM facility. X.B., V.O.G., and M.M. performed the neutron-scattering measurements. X.B. analyzed the neutron-scattering data. S.-S.Z., H.Z., and C.D.B. carried out the theoretical calculations. X.B., S.-S.Z., M.M., and C.D.B. wrote the manuscript with input from all authors.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Bai, X., Zhang, SS., Zhang, H. et al. Instabilities of heavy magnons in an anisotropic magnet. Nat Commun 14, 4199 (2023). https://doi.org/10.1038/s41467-023-39940-1
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467-023-39940-1
This article is cited by
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.