Abstract
It has become common knowledge that phonons can generate thermal Hall effect in a wide variety of materials, although the underlying mechanism is still controversial. We study longitudinal κxx and transverse κxy thermal conductivity in Pr2Ir2O7, which is a metallic analog of spin ice. Despite the presence of mobile charge carriers, we find that both κxx and κxy are dominated by phonons. A T/H scaling of κxx unambiguously reveals that longitudinal heat current is substantially impeded by resonant scattering of phonons on paramagnetic spins. Upon cooling, the resonant scattering is strongly affected by a development of spin ice correlation and κxx deviates from the scaling in an anisotropic way with respect to field directions. Strikingly, a set of the κxx and κxy data clearly shows that κxy correlates with κxx in its response to magnetic field including a success of the T/H scaling and its failure at low temperature. This remarkable correlation provides solid evidence that an indispensable role is played by spin-phonon scattering not only for hindering the longitudinal heat conduction, but also for generating the transverse response.
Similar content being viewed by others
Introduction
When the heat current carried by electrons is subject to a magnetic field applied normal to the current, a trajectory of electrons is curved by the Lorentz force and a transverse temperature gradient is developed in the direction both perpendicular to the heat current and the magnetic field. This phenomenon dubbed the thermal Hall effect has been believed to be restricted to materials in which there are mobile charge carriers. However, it is shown that even if the carriers of heat are neutral, the thermal Hall effect arises in several materials including magnetic insulators1,2, multiferroics3, spin liquid candidates4,5,6,7,8,9,10,11, Mott insulators12,13,14, and nonmagnetic insulator15, providing new insight on heat transport in solids. In some of the preceding materials, phonons are identified as the heat carriers responsible for the thermal Hall effect1,2,3,5,9,11,12,13,14,15. Despite the growing number of reports presenting the phonon Hall effect, little is known about the microscopic mechanism16,17,18,19,20,21,22,23,24.
To address this issue, we carried out measurements of thermal conductivity tensors in Pr2Ir2O7, which is a kind of ‘treasure trove’ of attractive physical properties including Kondo effect in a frustrated magnet25, topological Hall effect26,27, spin ice state in a metal27, quantum criticality28, and Luttinger semimetal with a quadratic band touching29. Among them, the relevant features to this study are the absence of a long-range magnetic order down to the lowest temperature measured and semimetallicity with low carrier density. The former prevents contamination of magnon contribution in the heat transport coefficients. While the metallicity enables precise estimation of electron contribution via the Wiedemann–Franz law, the low density of electron carriers leaves room for detection of the thermal Hall effect by chargeless carriers at the same time.
In this work, we show that longitudinal thermal conductivity κxx, in which phonon contribution by far dominates electron contribution, is largely degraded by spin–phonon scattering as low as that of amorphous silica. κxx is further lowered by the magnetic field due to resonant scattering between phonons and paramagnetic spins as evidenced by a T/H scaling. Upon cooling, magnetic fluctuations arising from spin ice correlation affect the resonant scattering by rendering another source of local level splitting through an exchange field and leads to a deviation from the T/H scaling. Despite the presence of mobile electrons, we detected finite thermal Hall conductivity κxy mostly generated by phonons. Unexpectedly, we find striking similarities between κxx and κxy in their response to a magnetic field, but importantly κxy behaves oppositely to κxx. This observation explicitly indicates that a single mechanism drives both longitudinal and transverse thermal response, and spin–phonon coupling which affects the mean-free path of phonons has a skew scattering component.
Results
Low longitudinal thermal conductivity in Pr2Ir2O7
Figure 1 shows the temperature dependence of longitudinal thermal conductivity κxx of Pr2Ir2O7 measured under zero field by applying the heat current Q parallel to the (001) plane (Q∥(001)). The data are shown together with those of insulating pyrochlore magnets, Yb2Ti2O730, Y2Ti2O731, Dy2Ti2O731, and Tb2Ti2O732 where heat conduction is dominated by phonons. As shown in the inset of Fig. 1a, electronic contribution L0σxxT to κxx estimated using the Wiedemann–Franz (WF) law is more than one order of magnitude smaller than κxx for both Q∥(001) and Q∥(111), which indicates that heat is predominantly transported by phonons. Interestingly, the magnitude of κxx is extremely small and approaches that of amorphous silica33. Moreover, the so-called phonon peak, which is characteristic of phononic thermal conductivity in insulating crystalline solids, is absent. It is shown that the structural disorder has a negligible effect on the Raman phonon spectra in the sample from the same source34. We thus stress that the low κxx is not due to phonon scattering by the random disorder. Since a position of the phonon peak is scaled by the Debye temperature, which is typically around 300–400 K for pyrochlore oxides35, κxx for Yb2Ti2O7, Y2Ti2O7, and Dy2Ti2O7 has the peaks at similar temperature (~10 K). At high temperatures exceeding the peak, the magnitude of thermal conductivity is set by the rate of collisions between thermally excited phonons whose number is also scaled by the Debye temperature. Thus, it is quite reasonable that κxx for Yb2Ti2O7 and Dy2Ti2O7 are close to each other at high temperatures.
By contrast, in Pr2Ir2O7 the phonon peak is absent and κxx is smaller than those of Y2Ti2O7 and Dy2Ti2O7 by a factor of five even at high temperatures, although our samples are crystalline solids and the Debye temperature of 400 K36 is similar to the other pyrochlore oxides. This suggests the presence of additional scatterers of phonons except for other phonons and disorders. Notably, thermal conductivity is also small and the phonon peak is absent in spin liquid candidate Tb2Ti2O7 (Fig. 1a) where these striking features are attributed to strong phonon scattering by magnetic fluctuations32. As we see below, spin–phonon scattering is a leading mechanism of the low phonon thermal conductivity in Pr2Ir2O7. An intrinsic scattering of phonon by mobile electrons may be an additional thermal impedance of the heat flow.
Resonant phonon scattering and H/T scaling
In Fig. 1b, c, magneto-thermal conductivity {κxx(H)−κxx(0)}/κxx(0) are shown for H∥[111] and H∥[001], respectively. For both directions, κxx first decreases with field and takes a minimum. On warming, a position of minimum defined as \({H}_{\min }\) shifts to higher fields. (See the inset of Fig. 1c for systematic change of \({H}_{\min }\) with temperature for H∥[001].) As shown in the inset of Fig. 1b, \({H}_{\min }\) increases linearly with temperature, \({H}_{\min } \sim T\), regardless of the field directions. Such a behavior has been observed in various paramagnets and is attributed to resonant scattering between phonons and paramagnetic spins37. We note that this \({H}_{\min } \sim T\) behavior is also discernible for a different sample of Pr2Ir2O7 with larger electrical conductivity38, indicating that our observation is an intrinsic property of heat conduction by phonons in this system. The resonance can occur in the presence of a strong spin–phonon coupling when the two-level spin systems split by the Zeeman energy absorb phonon and subsequently emit another phonon of the same energy in an unrelated direction. This spin–flip process effectively scatters phonons. The scattering becomes the largest when the Zeeman splitting ΔE ~ 2MμBH (M is magnetization) is equal to phonon energy whose spectrum has a broad maximum at ~4kBT. This causes the minimum in κxx(H) at \({H}_{\min } \sim 2{k}_{{{{{{{{\rm{B}}}}}}}}}T/M{\mu }_{{{{{{\rm{B}}}}}}}\) with \({H}_{\min }\) proportional to T. Therefore, field-induced change in the longitudinal thermal conductivity measured at various temperatures is expected to be scaled as a function of H/T with a minimum at \({H}_{\min }/T \sim 2{k}_{{{{{{{{\rm{B}}}}}}}}}/M{\mu }_{{{{{{\rm{B}}}}}}}\). Such scaling is demonstrated in Fig. 2a, b where Δκxx(H) = κxx(H)−κxx(0) normalized by its minimum value \({{\Delta }}{\kappa }_{xx}(H)/{\kappa }_{xx}^{{{{{{\rm{min}}}}}}}\) is plotted against H/T for H∥[111] and H∥[001], respectively. Remarkably, all data fall onto the same curve except for H/T > 1 and present minimum at H/T ~ 1. This result unambiguously indicates that κxx(H) is controlled by the resonant phonon scattering in the region of H < T. For the free Pr3+ ion, the magnetization is expected to be M = gJ J = 3.2, where gJ and J represent the Land\(\acute{{{{{{{{\rm{e}}}}}}}}}\)’s g factor and the total angular moment, which gives \({H}_{\min }/T \sim 2{k}_{{{{{{{{\rm{B}}}}}}}}}/3.2{\mu }_{B} \sim 0.93\), in good agreement with our observations. By closer looking at the data, however, one notices that the minimum position is slightly different with respect to the field directions: \({H}_{\min }/T \sim 1.25\) and 0.75 for H∥[111] and H∥[001], respectively. We will come back to this point later.
Anisotropic deviation from H/T scaling and spin ice correlation
Let us turn our attention to the high field regions. With increasing field, {κxx(H)−κxx(0)}/κxx(0) becomes positive (Fig. 1b, c) and the H/T scaling becomes failed (Fig. 2a, b). Concomitantly, we resolved a clear anisotropy in {κxx(H)−κxx(0)}/κxx(0), especially at low temperatures: while {κxx(H)−κxx(0)}/κxx(0) for H∥[111] increases with a concave curvature, the one for H∥[001] increases with a convex curvature and shows a tendency to saturate at low temperatures. Since the resonant scattering between phonons and paramagnetic spins is responsible for the negative magneto-thermal conductivity, the observed anisotropic recovery of κxx(H) implies the magnitude of resonant scattering is substantially influenced by spin correlation.
In magnetic materials, magnetic fluctuations yield strong scattering on phonons and significantly suppress phononic heat conduction39. An application of a magnetic field, however, weakens magnetic fluctuations and leads to a striking enhancement of phonon thermal conductivity30,32,40. The observed response of κxx to magnetic fields can be understood based on this line of thought. In particular, anisotropic magneto-thermal conductivity explicitly indicates a vital role of phonon scattering by fluctuating spins with spin ice correlation. In spin ice state41, the spin system fluctuates between the energetically equivalent “2-in, 2-out” configurations within the ground state manifold. This gives rise to strong magnetic fluctuations. The macroscopic degeneracy is lifted by the external magnetic field in an anisotropic way42,43. Magnetic field along the [001] direction steeply lifts the ground state degeneracy and suppresses the magnetic fluctuations because the stable spin configuration is uniquely determined as one of the six equivalent “2-in, 2-out” configurations by the field. For H∥[111], “3-in, 1-out/1-in, 3-out” configuration is energetically favored in high field limit. However, due to a smaller Zeeman energy gain for the spins on the Kagome plane with the “3-in, 1-out/1-in, 3-out” configuration than the “2-in, 2-out” configuration, the system remains in spin ice manifold and preserves the strong magnetic fluctuations up to higher field44,45.
This anisotropic suppression of magnetic fluctuations brings about positive and anisotropic magneto-thermal conductivity. For H∥[001], the steep suppression of the magnetic fluctuations yields the rapid rise of {κxx(H)−κxx(0)}/κxx(0) (Fig. 1c). Once the polarized state with the “2-in, 2-out” configuration is stabilized by the fields and fluctuations are totally suppressed, κxx gets saturated to a value which is purely dominated by phonons. Namely, the resonant scattering does not work anymore, since the spins are fully polarized in a saturation field, and the number of phonons carrying sufficient energy to flip the spin is exponentially suppressed. Meanwhile, the persistence of the magnetic fluctuations for H∥[111] yields the slower rise of {κxx(H)−κxx(0)}/κxx(0) (Fig. 1b).
Figure 1d shows temperature dependence of {κxx(H)−κxx(0)}/κxx(0) measured at 9 T for H∥[111] and H∥[001]. On cooling, {κxx(H)−κxx(0)}/κxx(0) changes a sign from negative to positive around 4 and 7 K for H∥[111] and H∥[001], respectively. As mentioned above, the resonant phonon scattering is strongly influenced by spin correlation through a local exchange field. However, by applying a large magnetic field, or equivalently at low temperatures, the local spin splitting becomes mainly determined by an external magnetic field. This crossover causes a gradual change from negative to positive magneto-thermal conductivity. In that sense, the sign-change temperature can be regarded as a lower bound of onset temperature below which the spin-ice correlation sets in. Notably, this temperature roughly coincides with a resistivity minimum (see the inset of Fig. 1d) which is another consequence of the spin-ice correlation while in this case, the correlated spins interact with conduction electrons46.
Now, let us discuss the implication of the spin-ice correlations to the anisotropy in \({H}_{\min }\). Under the spin ice state, the Zeeman splitting ΔE ~ 2MμBH of the ground state doublet is anisotropic with respect to the field directions due to anisotropy in magnetization M25. Accordingly, given the relation of \({H}_{\min } \sim 2{k}_{{{{{{{{\rm{B}}}}}}}}}T/M{\mu }_{{{{{{\rm{B}}}}}}}\), \({H}_{\min }\) is anisotropic and its anisotropic ratio between [111] and [001] directions is expected to be held a relation of \({H}_{\min }^{{{{{{{{\rm{[111]}}}}}}}}}/{H}_{\min }^{{{{{{{{\rm{[001]}}}}}}}}} \sim {M}_{{{{{{{{\rm{[001]}}}}}}}}}/{M}_{{{{{{{{\rm{[111]}}}}}}}}}\). This means that at a given temperature the larger Zeeman splitting ΔE due to the larger M satisfies the condition of resonance at the lower field. In fact, the anisotropic ratio of \({H}_{\min }\), \(({H}_{\min }^{{{{{{{{\rm{[111]}}}}}}}}}/T)/({H}_{\min }^{{{{{{{{\rm{[001]}}}}}}}}}/T)=1.25/0.75 \sim\) 1.67 extracted from Fig. 2a, b, is in good agreement with magnetization anisotropy expected for the “2-in, 2-out” configuration, \({M}_{{{{{{{{\rm{[001]}}}}}}}}}/{M}_{{{{{{{{\rm{[111]}}}}}}}}}=\{{g}_{J}\,J(1/\sqrt{3})\}/\{{g}_{J}\,J(1+1/3\times 1)/4\} \sim\) 1.73.
Our argument that the spin–phonon scattering controls the evolution of κxx(H) is supported by a theoretical calculation. We model the interaction between the Pr doublets by a simple spin-ice-type Ising model. What is characteristic of this compound is the spin–phonon interaction: we assume a linear transverse coupling between the Pr doublets and acoustic phonons. The Pr3+ ion takes f 2 configurations in Pr2Ir2O7, and its single-ion ground state is described as a non-Kramers doublet taking Eg representation in a D3d symmetric local crystal field47,48. In this case, the local transverse component of the doublet has quadrupole nature, which enables the linear coupling to lattice deformations or phonons. Even within this simple model, we can qualitatively reproduce main experimental features of magneto-thermal transport as shown in Fig. 1e; the initial negative magneto-thermal conductivity, the presence of minimum, and the positive increase with the convex curvature at the low temperature and a high field region.
Here we note the role of magnetic excitations in thermal transport. In spin ice, magnetic monopoles are excited above the temperature of the order of exchange coupling. Indeed, the dynamics and transport of monopoles are widely discussed both experimentally and theoretically. However, in Pr2Ir2O7, we can safely ignore their contribution. Since the magnetic field applied along the [001] direction is unfavorable for the spin ice state, the number of monopoles decays faster for H∥[001] than H∥[111]. Therefore, if heat is carried by the monopoles, κxx(H) is expected to fall rapidly for H∥[001] than H∥[111]. This is indeed observed in Yb2Ti2O730, and they raised this behavior as the major evidence for monopole transport. However, in our system, κxx(H) falls rapidly for H∥[111]. This observation clearly shows that monopoles do not play a major role in the thermal transport in the measured temperature range. To raise one more evidence for the irrelevancy of monopoles, the reduction of κxx(H) with the field is observed up to as high as 80 K (Fig. 1d), which is much higher than the characteristic temperature of spin ice correlations.
Thermal Hall effect by phonons
Having established the dominant role of spin–phonon scattering in the longitudinal thermal conductivity, let us focus on the thermal Hall effect. Temperature dependence of thermal Hall conductivity divided by temperature κxy/T measured under magnetic field of 9 T for H∥[111] and H∥[001] are shown in Fig. 3a, b, respectively. In the same figures, we also show the electronic contribution L0σxy (left axis) and κxx/T (right axis). Surprisingly, a sign of L0σxy is opposite to κxy/T in the whole measured temperature range for H∥[111] and T > 4 K for H∥[001], indicating that the thermal Hall effect is mostly governed by carriers except for electrons. Given the negligible contribution of monopoles in κxx, they are not responsible for the Hall response either. Thus, phonons are most probably the unique heat carriers that can cause thermal Hall effect in this paramagnet. Moreover, both κxy/T and κxx/T peak around 20–30 K where phonons dominate the longitudinal thermal conductivity because electron contribution accounts for only L0σxyT/κxx ~ 0.6% of the total κxx (see the inset of Fig. 1a). Such a coincidence of peaks in κxx and κxy has been observed in several insulating solids and regarded as a clue to identify the thermal Hall signal generated by phonons15. This result further supports the conjecture that thermal Hall current is carried by phonons in Pr2Ir2O7. We note a ratio κxy/κxx ≃ 0.4−0.8 × 10−3 around the peak is comparable to that found in materials where phonons have been argued to cause the Hall effect2,11.
At temperatures above the peak, the magnitude of κxy/T is comparable with that of Tb2Ti2O74 and smaller than the unexpectedly large thermal Hall conductivity of SrTiO315 and La2Cu4O12 by a factor of 10 (the inset of Fig. 3b). Below the peak, κxy/T steeply decreases faster than κxx/T. κxy/T for H∥[001] seems to approach the value expected from the WF law followed by a sign change around 4 K, showing that phonons cease to contribute to the Hall response at low temperatures.
In Fig. 3c, d, we show magnetic field dependence of thermal Hall conductivity κxy(H) (triangles) together with electron contribution L0σxyT(H) (dotted lines) estimated by using the WF law for H∥[111] and H∥[001], respectively. Again, a sign of L0σxyT(H) is opposite to κxy. By subtracting L0σxyT(H) from κxy(H), we evaluated thermal Hall conductivity generated by phonons as \({\kappa }_{xy}^{ph}={\kappa }_{xy}-{L}_{0}{\sigma }_{xy}T\) (circles). As seen from the figures, at 20 K κxy(H) increases linearly with H and there is negligible electron contribution in both directions. On cooling, κxy(H) becomes non-monotonic. Namely, κxy(H) shows a peak and subsequently decreases with the field. Moreover, a fraction of the electron contribution to κxy slightly increases, which is maximized up to ∣L0σxyT∣/κxy ~ 28% at 2.9 K and H = 4 T∥[111]. Since the peak remains in \({\kappa }_{xy}^{{{{{{\rm{ph}}}}}}}(H)\) even after the subtraction of electron contribution, phonons are responsible for the non-monotonic behavior. By further decreasing temperature, an anisotropic field response emerges at high fields. κxy(H) for H∥[001] is considerably suppressed above its peak field and approaches a value expected from the WF law within an experimental error (the inset of Fig. 3d), consistent with what we saw in the temperature variation of κxy/T (Fig. 3b). By contrast, the suppression is weak for H∥[111] and \({\kappa }_{xy}^{{{{{{\rm{ph}}}}}}}(H)\) remains positive up to 9 T.
Correlation between κ xx and κ xy
One of the most striking findings of this work is a correlation between field-induced change in \({\kappa }_{xy}^{{{{{{\rm{ph}}}}}}}\) and κxx, which are displayed in the upper and lower panels of Fig. 4, respectively. In each panel, we compare two data taken at the (nearly) same temperature for H∥[111] (open circles) and for H∥[001] (closed circles). In Fig. 4, there are several things of interest. (i) The maximum and the minimum appear at nearly the same field in \({\kappa }_{xy}^{{{{{{\rm{ph}}}}}}}(H)\) and Δκxx(H), respectively, and the extreme positions shift to a lower field with decreasing temperature. (ii) Above 7 K, the relationship in the magnitude of Δκxx(H) between the two directions is the same as \({\kappa }_{xy}^{{{{{{\rm{ph}}}}}}}(H)\): the large negative magneto-thermal conductivity is accompanied by the large thermal Hall signal for H∥[111], and vice versa for H∥[001]. (iii) Below 5.1 K, steeper Δκxx(H) rises above its minimum field, stronger the suppression of \({\kappa }_{xy}^{{{{{{\rm{ph}}}}}}}(H)\) becomes above its maximum field. Upon cooling, this correlation becomes more significant for H∥[001].
From the observation (i), the striking resemblance between \({\kappa }_{xy}^{{{{{{\rm{ph}}}}}}}(H)\) and Δκxx(H) led us to expect that the H/T scaling is also valid for \({\kappa }_{xy}^{{{{{{\rm{ph}}}}}}}(H)\). As demonstrated in Fig. 2c, d, the data of \({\kappa }_{xy}^{{{{{{\rm{ph}}}}}}}(H)\) divided by its maximum value indeed collapse on a single curve for both directions for H/T < 1 as in the case of κxx(H) (Fig. 2a, b). Moreover, observation (ii) indicates that the strong paramagnetic scattering of phonons that gives rise to the negative magneto-thermal conductivity is an ingredient to enhance the phonon thermal Hall effect. These results provide compelling evidence that a prominent role is played by resonant phonon scattering not only in degrading longitudinal phonon heat conduction but also in generating the transverse signal.
We note that the scaling of \({\kappa }_{xy}^{{{{{{\rm{ph}}}}}}}\) also becomes failed for H/T > 1 (Fig. 2c, d), indicating that the paramagnetic scattering no longer plays a major role in this regime. Instead, from the remarkable correlation in the observation (iii), it is quite natural to identify another source of asymmetric scattering of phonons as magnetic fluctuations. Whereas the survival of magnetic fluctuations along the [111] direction yields the sizable \({\kappa }_{xy}^{{{{{{\rm{ph}}}}}}}\) even after the paramagnetic scattering dies out, the strong suppression of magnetic fluctuations along the [001] direction results in the substantial decrease of κxy towards the value purely dominated by electrons. Thus, it is concluded that whatever the spin state is (whether spins are paramagnetic or correlated) when phonons interact with spins, they are asymmetrically scattered and produce the thermal Hall signal.
To attempt to clarify the intriguing thermal Hall phenomena in Pr2Ir2O7, one should seriously take into account the following two facts. First, the evolution of κxx with the field can be thoroughly explained by the way spins scatter phonons. This indicates an intrinsic coupling of phonons to the magnetic environment. Second, there is a manifest correlation between κxx and κxy in their evolution within the field. These two facts impose constraints on possible scenarios that the longitudinal and transverse thermal responses should be understood in a unified way in terms of an intrinsic coupling of phonons to spins with a skew component and make a possibility of the extrinsic origin like the skew scattering of phonons by superstoichiometric rare-earth ions20, oxygen vacancies23, and dynamical defects24 unlikely.
Methods
Samples
Single crystals of Pr2Ir2O7 were grown by a flux method49. We used two different single crystals for the thermal transport measurements under a magnetic field applied parallel to the [111] and [001] directions. The [111] (H∥[111]) and [001] (H∥[001]) samples have plate-like shape with dimensions of 1.7(width) × 2.1 (length) mm2 in the (111) plane and 1.1(width) × 2.1(length) mm2 in the (001) plane, respectively. The thicknesses of the samples are about 0.5 mm.
Thermal transport measurements
Longitudinal thermal conductivity κxx and thermal Hall conductivity κxy were measured by the standard steady-state method in a high vacuum. The heat flow Q was injected in the (111) and (001) planes for the [111] and [001] samples, respectively, by heating a chip resistor attached to one end of the sample. The other end of the sample was attached to an insulating LiF plate, which was used as a cold thermal bath. The longitudinal ΔTx and transverse ΔTy temperature differences were determined by Cernox thermometers. The thermometers and the heater were connected by gold wires (∅ = 25 μm) and heat-cured silver paint (Dupont 6838) to the sample. The contact resistances were 10 mΩ. To remove the longitudinal response from the raw data due to misalignment of the contacts, we anti-symmetrized it as ΔTy(H) = {ΔTy(+H)−ΔTy(−H)}/2. κxx and κxy were obtained from the longitudinal thermal resistivity, wxx = (ΔTx/Q)(wt/l), and the thermal Hall resistivity, wxy = (ΔTy/Q)t, as \({\kappa }_{xx}={w}_{xx}/({w}_{xx}^{2}+{w}_{xy}^{2})\) and \({\kappa }_{xy}=-{w}_{xy}/({w}_{xx}^{2}+{w}_{xy}^{2})\). Here, l, w, and t are lengths between the contacts, width, and thickness of the samples, respectively. The electrical (Hall) resistivity measurements were done by using the same contacts and gold wires. κxx and κxy were checked to be independent of the thermal gradient by changing ΔTx/T in the range of 1–20%. Since ΔTy is tiny, which is as small as 0.1 mK, at low temperatures and the scattering of the data is large, the measurements were repeated several times and the data is averaged. Error bars in the main figures represent one standard deviation.
Computational
Here, we summarize a theoretical formulation to calculate the longitudinal thermal conductivity of acoustic phonons, as shown in Fig. 1e. We assume two kinds of scattering centers, non-magnetic impurities and localized Pr moments. The former gives a scattering rate weakly dependent on the energy of phonons, which results in the normal ∝ T3 behavior of phonon thermal conductivity in the low-temperature limit. The latter scattering process is characteristic of this system, in particular, the non-Kramers nature of Pr moments. It was pointed out that the transverse components of Pr doublets behave as magnetic quadrupoles rather than dipoles in the Pr pyrochlore oxides47,48. Consequently, the lattice deformation couples to the transverse components of Pr doublets, or conversely, the acoustic phonons are scattered inelastically through the flip of Pr doublets.
Combining these two types of scattering processes, the thermal conductivity can be concisely written as
where \({\kappa }_{0}\equiv \frac{2{\pi }^{2}\tau {T}^{3}}{15c}\) is the normal phonon thermal conductivity. Δj is the splitting of Pr doublet at site j due to the “local effective field”, i.e. the combined effects of the external magnetic field and the exchange interaction with surrounding doublets. δ is the variance of the local effective field, which is essential to the resonant spin–phonon scattering and is usually attributed to the randomness in the system.
In the present analysis, we adopt the nearest-neighbor spin ice model to describe the thermal fluctuation of Pr doublets,
Here the first term is the nearest-neighbor interaction between Pr doublets, σj = ±1. The second term describes the site-dependent Zeeman interaction with the external magnetic field, h. dj stands for the easy axis of the Pr doublet at site j. We assume the case of [001] field direction, and conducted the Monte Carlo simulation for N = 16 × 16 × 16 × 4 = 16,384 doublets and, made 10,000 samplings for the effective field, \({{{\Delta }}}_{j}\equiv (\frac{2}{\sqrt{3}}h-2J{\sum }_{j^{\prime} }{\sigma }_{j^{\prime} }){\sigma }_{j}\). From the thermal average, we obtain the thermal conductivity through Eq. (1).
Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.
References
Strohm, C., Rikken, G. L. J. A. & Wyder, P. Phenomenological evidence for the phonon Hall effect. Phys. Rev. Lett. 95, 155901 (2005).
Chen, L., Boulanger, M.-E., Wang, Z.-C., Tafti, F. & Taillefer, L. Large phonon thermal Hall conductivity in a simple antiferromagnetic insulator. Preprint at http://arxiv.org/abs/2110.13277 (2021).
Ideue, T., Kurumaji, T., Ishiwata, S. & Tokura, Y. Giant thermal Hall effect in multiferroics. Nat. Mater. 16, 797–802 (2017).
Hirschberger, M., Krizan, J. W., Cava, R. J. & Ong, N. P. Large thermal Hall conductivity of neutral spin excitations in a frustrated quantum magnet. Science 348, 106–109 (2015).
Sugii, K. et al. Thermal Hall effect in a phonon-glass Ba3CuSb2O9. Phys. Rev. Lett. 118, 145902 (2017).
Kasahara, Y. et al. Unusual thermal Hall effect in a Kitaev spin liquid candidate α-RuCl3. Phys. Rev. Lett. 120, 217205 (2018).
Kasahara, Y. et al. Majorana quantization and half-integer thermal quantum Hall effect in a Kitaev spin liquid. Nature 559, 227–231 (2018).
Hentrich, R. et al. Large thermal Hall effect in α-RuCl3: evidence for heat transport by Kitaev–Heisenberg paramagnons. Phys. Rev. B 99, 085136 (2019).
Akazawa, M. et al. Thermal Hall effects of spins and phonons in Kagome antiferromagnet Cd-Kapellasite. Phys. Rev. X 10, 041059 (2020).
Yokoi, T. et al. Half-integer quantized anomalous thermal Hall effect in the Kitaev material candidate α-RuCl3. Science 373, 568–572 (2021).
Lefrançois, É. et al. Evidence of a phonon Hall effect in the Kitaev spin liquid candidate α-RuCl3. Phys. Rev. X 12, 021025 (2022).
Grissonnanche, G. et al. Giant thermal Hall conductivity in the pseudogap phase of cuprate superconductors. Nature 571, 376–380 (2019).
Grissonnanche, G. et al. Chiral phonons in the pseudogap phase of cuprates. Nat. Phys. 16, 1108–1111 (2020).
Boulanger, M.-E. et al. Thermal Hall conductivity in the cuprate Mott insulators Nd2CuO4 and Sr2CuO2Cl2. Nat. Commun. 11, 5325 (2020).
Li, X., Fauqué, B., Zhu, Z. & Behnia, K. Phonon thermal Hall effect in strontium titanate. Phys. Rev. Lett. 124, 105901 (2020).
Sheng, L., Sheng, D. N. & Ting, C. S. Theory of the phonon Hall effect in paramagnetic dielectrics. Phys. Rev. Lett. 96, 155901 (2006).
Wang, J.-S. & Zhang, L. Phonon Hall thermal conductivity from the Green–Kubo formula. Phys. Rev. B 80, 012301 (2009).
Zhang, L., Ren, J., Wang, J.-S. & Li, B. Topological nature of the phonon Hall effect. Phys. Rev. Lett. 105, 225901 (2010).
Qin, T., Zhou, J. & Shi, J. Berry curvature and the phonon Hall effect. Phys. Rev. B 86, 104305 (2012).
Mori, M., Spencer-Smith, A., Sushkov, O. P. & Maekawa, S. Origin of the phonon Hall effect in rare-earth garnets. Phys. Rev. Lett. 113, 265901 (2014).
Saito, T., Misaki, K., Ishizuka, H. & Nagaosa, N. Berry phase of phonons and thermal Hall effect in nonmagnetic insulators. Phys. Rev. Lett. 123, 255901 (2019).
Yang, Y.-F., Zhang, G.-M. & Zhang, F.-C. Universal behavior of the thermal Hall conductivity. Phys. Rev. Lett. 124, 186602 (2020).
Flebus, B. & MacDonald, A. H. Charged defects and phonon Hall effects in ionic crystals. Phys. Rev. B 105, L220301 (2022).
Sun, X.-Q., Chen, J.-Y. & Kivelson, S. A. Large extrinsic phonon thermal Hall effect from resonant scattering. Preprint at http://arxiv.org/abs/2109.12117 (2021).
Nakatsuji, S. et al. Metallic spin-liquid behavior of the geometrically frustrated Kondo lattice Pr2Ir2O7. Phys. Rev. Lett. 96, 087204 (2006).
Machida, Y. et al. Unconventional anomalous Hall effect enhanced by a noncoplanar spin texture in the frustrated Kondo lattice Pr2Ir2O7. Phys. Rev. Lett. 98, 057203 (2007).
Machida, Y., Nakatsuji, S., Onoda, S., Tayama, T. & Sakakibara, T. Time-reversal symmetry breaking and spontaneous Hall effect without magnetic dipole order. Nature 463, 210–213 (2010).
Tokiwa, Y., Ishikawa, J. J., Nakatsuji, S. & Gegenwart, P. Quantum criticality in a metallic spin liquid. Nat. Mater. 13, 356–359 (2014).
Kondo, T. et al. Quadratic Fermi node in a 3D strongly correlated semimetal. Nat. Commun. 6, 10042 (2015).
Tokiwa, Y. et al. Possible observation of highly itinerant quantum magnetic monopoles in the frustrated pyrochlore Yb2Ti2O7. Nat. Commun. 7, 10807 (2016).
Kolland, G., Valldor, M., Hiertz, M., Frielingsdorf, J. & Lorenz, T. Anisotropic heat transport via monopoles in the spin-ice compound Dy2Ti2O7. Phys. Rev. B 88, 054406 (2013).
Li, Q. J. et al. Phonon-glass-like behavior of magnetic origin in single-crystal Tb2Ti2O7. Phys. Rev. B 87, 214408 (2013).
Keller, R. C. & Pohl, R. O. Thermal conductivity and specific heat of noncrystalline solids. Phys. Rev. B 4, 2029 (1971).
Xu, Y. et al. Phonon spectrum of Pr2Zr2O7 and Pr2Ir2O7 as evidence of coupling of the lattice with electronic and magnetic degrees of freedom. Phys. Rev. B 105, 075137 (2022).
Subramanian, M. A., Aravamudan, G. & Rao, G. V. S. Oxide pyrochlores—a review. Prog. Solid State Chem. 15, 55–143 (1983).
Ghosh, B. & Mukhopadhyay, S. Unified description of resistivity and thermopower of Pr2Ir2O7: possible influence of crystal field excitation in a Kondo lattice. Phys. Rev. B 103, 165135 (2021).
Berman, R. Thermal Conduction in Solids (Clarendon Press, Oxford, 1976).
Ni, J. M. et al. Giant isotropic magneto-thermal conductivity of metallic spin liquid candidate Pr2Ir2O7 with quantum criticality. Nat. Commun. 12, 307 (2021).
Sharma, P. A. et al. Thermal conductivity of geometrically frustrated, ferroelectric YMnO3: extraordinary spin–phonon interactions. Phys. Rev. Lett. 93, 177202 (2004).
Wang, X. M. et al. Large magnetothermal conductivity of HoMnO3 single crystals and its relation to the magnetic-field-induced transitions of magnetic structure. Phys. Rev. B 82, 094405 (2010).
Udagawa, M. & Jaubert, L. (eds) Spin Ice. Springer Series in Solid-State Sciences (Springer, Cham, 2021).
Fukazawa, H., Melko, R. G., Higashinaka, R., Maeno, Y. & Gingras, M. J. P. Magnetic anisotropy of the spin-ice compound Dy2Ti2O7. Phys. Rev. B 65, 054410 (2002).
Hiroi, Z., Matsuhira, K., Takagi, S., Tayama, T. & Sakakibara, T. Specific heat of Kagomé ice in the pyrochlore oxide Dy2Ti2O7. J. Phys. Soc. Jpn. 72, 411–418 (2003).
Udagawa, M., Ogata, M. & Hiroi, Z. Exact result of ground-state entropy for Ising pyrochlore magnets under a magnetic field along [111] axis. J. Phys. Soc. Jpn. 71, 2365–2368 (2002).
Moessner, R. & Sondhi, S. L. Theory of the [111] magnetization plateau in spin ice. Phys. Rev. B 68, 064411 (2003).
Udagawa, M., Ishizuka, H. & Motome, Y. Non-Kondo mechanism for resistivity minimum in spin ice conduction systems. Phys. Rev. Lett. 108, 066406 (2012).
Onoda, S. & Tanaka, Y. Quantum fluctuations in the effective pseudospin-\(\frac{1}{2}\) model for magnetic pyrochlore oxides. Phys. Rev. B 83, 094411 (2011).
Rau, J. G. & Gingras, M. J. P. Frustrated quantum rare-earth pyrochlores. Annu. Rev. Condens. Mater. 10, 357–386 (2019).
Millican, J. N. et al. Crystal growth and structure of R2Ir2O7 (R = Pr, Eu) using molten KF. Mater. Res. Bull. 42, 928–934 (2007).
Acknowledgements
We acknowledge discussions with Kamran Behnia. This work was partially supported by Grants-in-Aid for Scientific Research (19H01840 and 20H05655), and JST-CREST (JPMJCR18T3), Japan Science and Technology Agency. The work at the Institute for Quantum Matter, an Energy Frontier Research Center was funded by DOE, Office of Science, Basic Energy Sciences under Award # DE-SC0019331.
Author information
Authors and Affiliations
Contributions
Y.M. conceived and designed the study. T.U. and Y.M. performed the transport measurements and analyzed the data. T.O. and S.N. synthesized the single crystal samples. M.U. performed the theoretical calculations. Y.M. and M.U. wrote the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks Zhaoming Tian, Yi-feng Yang and the other, anonymous, reviewer for their contribution to the peer review of this work.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
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
Uehara, T., Ohtsuki, T., Udagawa, M. et al. Phonon thermal Hall effect in a metallic spin ice. Nat Commun 13, 4604 (2022). https://doi.org/10.1038/s41467-022-32375-0
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467-022-32375-0
This article is cited by
-
The phonon thermal Hall angle in black phosphorus
Nature Communications (2023)
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.