Spin-phonon interaction in yttrium iron garnet

Spin-phonon interaction is an important channel for spin and energy relaxation in magnetic insulators. Understanding this interaction is critical for developing magnetic insulator-based spintronic devices. Quantifying this interaction in yttrium iron garnet (YIG), one of the most extensively investigated magnetic insulators, remains challenging because of the large number of atoms in a unit cell. Here, we report temperature-dependent and polarization-resolved Raman measurements in a YIG bulk crystal. We first classify the phonon modes based on their symmetry. We then develop a modified mean-field theory and define a symmetry-adapted parameter to quantify spin-phonon interaction in a phononmode specific way for the first time in YIG. Based on this improved mean-field theory, we discover a positive correlation between the spin-phonon interaction strength and the phonon frequency. Introduction Magnetic insulators are of considerable interest in spintronics due to their minimal spin damping [1–3]. This low damping originates in part from the absence of low-energy electronic excitations, leaving the spins to interact primarily with other spins (magnons) and the lattice (phonons). Beyond their role in spin excitation damping, interactions between the magnons and phonons play a crucial role in developing devices based on thermally driven spin transport [4–6], spin pumping through hybrid spin-lattice excitations[7], and magnon cavity quantum electrodynamics [8, 9]. Of various magnetic insulators explored for spintronic devices, yttrium iron garnet (YIG): Y3Fe5O12 is the most widely investigated due to its remarkably low spin damping and its high transition temperature of 560 K [10, 11]. However, due to its massive unit cell (160 atoms as an inset of Fig.1a), extracting the spin-phonon interaction (SPI) of YIG from ab initio studies is remarkably difficult. The SPI in YIG has been investigated through different types of experiments. Brillouin light scattering and spin Seebeck transport measurements of YIG have examined the interactions of magnons and phonons through quasiparticle hybridization [12–14]. Other studies have touched upon the SPI by measuring the magnon-phonon energy relaxation length and time [4, 15]. However, no study provides a direct and quantitative measurement of the strength of the SPI in YIG in a phonon-mode specific way. 2 Without knowing the SPI strength, it is difficult to develop accurate models of spin relaxation in YIG or compare YIG to other magnetic insulators for device development. Here we report Raman spectroscopy studies of optical phonons in a YIG bulk crystal. By analyzing their symmetry properties and temperature-dependent phonon frequency shift, we investigate if SPI changes systematically for each phonon mode. We determine that the complex unit cell precludes a direct correlation between symmetry or frequency of a phonon mode with the conventional λ-model of the SPI strength [16–18]. By developing a mean-field model and defining a new parameter to describe SPI strength, we observe a correlation between this mean-field SPI parameter and phonon frequency. These results provide crucial information and advance the understanding of how magnons and phonons interact in YIG. Experiment YIG (Y3Fe5O12) is an insulating ferrimagnet (FiM) with Curie temperature TC = 570 K and easy axis in the [111] direction [19, 20]. YIG crystals exhibit symmetries described by cubic space group Ia3%d (No. 230) and point group Oh at the Γ point [21–23]. Inversion symmetry present in Oh implies that the phonon modes show mutually-exclusive infrared and Raman activity. The possible Raman irreducible representations in Oh are either T2g, Eg, or A1g. The crystal structure is composed of Y atoms occupying the 24c Wyckoff sites, Fe ions in the 16a and 24d positions, and O atoms in the 96h sites. The conventional unit cell has eight formula units, with 24 Y ions, 40 Fe ions, and 96 O ions for a total of 160 atoms. Raman measurements were taken with a 532 nm laser incident on a YIG single crystal with [111] oriented along the surface normal. The scattered light was collected in a backscattering geometry and directed to a diffraction grating-based spectrometer. The observed optical phonon modes in the Raman spectra agree with previous measurements of YIG [24, 25]. Low-temperature measurements from 8.8 K to 313.65 K were performed in a closed-loop cryostat, and high-temperature measurements from 313.65 K to 631.95 K were performed with a ceramic heater. Between each temperature, the sample was allowed to equilibrate for 15 minutes or longer. A saturating magnetic field was applied in the sample plane for all measurements. Due to constraints of the experiment systems, low-temperature measurements used a 300 mT saturating field, and the high-temperature measurements used a 50 mT saturating field. Since both fields were above the saturating field, this difference did not noticeably affect the magnetic ordering of YIG or the Raman spectra. Raman spectra were collected with a fixed s-polarization incident on the sample. Fig.1a shows the pand s-polarizations components of the scattered light for the sample at low temperature (8.8 K). Phonon modes of different symmetries scatter light with different polarizations. Fig. 1(b) shows the intensity of the Raman signal from the scattered light as it passed through a linear polarizer, with the polarization axis rotated in steps of 20° from 0° to 180°. Based on the results, the phonons are categorized with their respective irreducible representations: T2g, Eg, or A1g. The temperature dependence of the phonon frequencies was determined by fitting with a Lorentzian function and extracting the central frequencies. We plot the measured Raman spectra for one T2g mode at three different representative temperatures 8.8 K, 313.65 K, and 632 K in Fig. 2 (a), (b), and (c). At low temperatures (e.g. 8.8 K), the low thermal population of the phonons reduces the Raman intensity. In contrast, the phonon modes exhibit a broader linewidth at high temperatures due to increased phonon-phonon and phonon-magnon scattering, which lowers the peak intensity. Consequently, the temperature-dependent frequency was only measurable for a subset of the observed phonons. The temperature dependence of the peak frequencies for the two modes is shown in Fig. 2d and 2e. The temperature dependence of peak frequencies of all the measurable phonon modes can be found in Supplementary Information.