betaFeedback

WorksPhysics and Astronomy

Planar parallel phonon Hall effect and local symmetry breaking

Abstract

Y-kapellasite [Y3Cu9(OH)19Cl8] is a frustrated antiferromagnetic insulator which remains paramagnetic down to a remarkably low Néel temperature of about 2 K. Having studied this material in the paramagnetic regime, in which phonons are the only possible heat carriers, we report the observation of a planar parallel thermal Hall effect coming unambiguously from phonons. This is an advantage over the Kitaev quantum spin liquid candidates α-RuCl3 and Na2Co2TeO6 where in principle other heat carriers can be involved [1-4]. As it happens, Y-kapellasite undergoes a structural transition attributed to the positional freezing of a hydrogen atom below about 33 K. Above this transition, the global crystal symmetry forbids the existence of a planar parallel signal - the same situation as in Na2Co2TeO6 and cuprates [3-5]. This points to the notion of a local symmetry breaking at the root of the phonon Hall effect. In this context, the advantage of Y-kapellasite over Na2Co2TeO6 (with high levels of Na disorder and stacking faults) and cuprates (with high levels of disorder coming from dopants and oxygen vacancies) is its clean structure, where the only degree of freedom available for local symmetry breaking is this hydrogen atom randomly distributed over six equivalent positions above 33 K. This provides a specific and concrete case for the general idea of local symmetry breaking leading to the phonon Hall effect in a wide range of insulators.

Rather unexpectedly, C. Strohm and colleagues discovered the phonon Hall effect (PHE) through conventional thermal Hall effect measurements in a paramagnetic dielectric garnet, in which phonons are the only possible heat carriers [6]. Given a longitudinal thermal gradient ΔTi\Delta T_i produced by the heat flux qiq_i set at one end of the sample along the direction ii, an orthogonal temperature gradient ΔTj\Delta T_j develops along the direction jj when a magnetic field BkB_k is applied fully normal to the (ij)(ij) plane along the direction kk, originating in a finite QijkQ_{ijk} Righi-Leduc tensor component: qi=QijkΔTjBkq_i = Q_{ijk}\Delta T_jB_k, see Fig. 1a. In a conventional metal, where electrons carry heat in tandem with phonons, the thermal Hall conductivity κij\kappa_{ij} naturally includes a sizeable Lorentz force-like contribution, directly related to the electrical Hall effect through the Wiedemann-Franz law. Conversely, observing a finite κij\kappa_{ij} in an insulator, where all possible heat carriers are necessarily neutral, is quite counter-intuitive.

Over the past few years, a number of theoretical and experimental studies focusing on magnetic insulators highlighted that phonons are not the only possible neutral heat carriers and that collective spin excitations such as magnons – conventional spin waves – may also generate a thermal Hall effect [7, 8, 9, 10, 11, 12, 13]. Naturally, the coupling of these magnetic quasiparticles to the magnetic field appears less cryptic. A tantalising aspect is that most studies conducted on frustrated quantum magnets reported a sizeable thermal Hall signal presented as compelling evidence for the emergence of long-sought exotic spin excitations while assuming a marginal or null PHE.

The most recent and salient examples are those of two Kitaev quantum spin liquid candidates, namely α\alpha-RuCl3_3 and Na2_2Co2_2TeO6_6, in which a startling planar parallel thermal Hall effect was detected when a magnetic field BiB_i is applied fully parallel to the heat flux [1, 2, 3, 4], see Fig. 1a. Until now, this finite QijiQ_{iji} was systematically attributed to unconventional magnetic edge states rather than phonons. In particular, regarding the highly scrutinised α\alpha-RuCl3_3, there is a heated debate to decide between the two scenarios proposed so far: chiral Majorana fermions from the gapped Kitaev quantum spin liquid [1], expected to yield a quantized temperature dependence of κij\kappa_{ij}, versus topological magnons [2], expected to yield a steeper temperature dependence, typical of bosons.

Yet, given that phonons contribute to the conventional thermal Hall effect in these two materials [14, 15], one wonders if they also contribute to the planar parallel thermal Hall effect. If so, this would cast doubt on the putative evidence for exotic spin excitations.

In our recent studies of Na2_2Co2_2TeO6_6 [4] and of the Mott insulating (antiferromagnetic) cuprate Nd2−x_{2-x}Cex_xCuO4_4 (with x=0.04x = 0.04) [5], we argued that the similar temperature dependence of the phonon-dominated longitudinal thermal conductivity and the planar parallel thermal Hall conductivity is a strong indication that phonons do contribute to the planar parallel thermal Hall effect. To support our interpretations, it is now crucial to provide an unambiguous observation of the planar parallel phonon Hall effect in a material where phonons are the only possible heat carriers.

Beyond the nature of the involved heat carriers, the symmetry requirements for the planar parallel thermal Hall effect constitute a key issue. In Na2_2Co2_2TeO6_6 for instance, it is forbidden by the global crystal symmetries of the P6322P6_3 22 space group even though finite signals were reported by two independent groups, using samples from different sources [3, 4]. There is nonetheless a striking difference in magnitude and temperature dependence between the two sets of results, which suggests that mechanisms related to the sample quality and history, e.g., defects and domains, whether structural or magnetic, are responsible for the planar parallel thermal Hall effect.

In Nd2−xCexCuO4\mathrm{Nd}_{2-x}\mathrm{Ce}_{x}\mathrm{CuO}_{4} (with x=0.04x = 0.04), it is also forbidden by the global crystal symmetries of the I4/mmmI4/mmm space group [5]. We thus wonder if local symmetry breaking, around positional disorder or extrinsic defects, is at the root of the planar parallel thermal Hall effect.

In the present study, we investigate the thermal Hall effect in clean, phase pure single crystals of Y-kapellasite [Y3Cu9(OH)19Cl8\mathrm{Y}_{3}\mathrm{Cu}_{9}(\mathrm{OH})_{19}\mathrm{Cl}_{8}], a kagome-based insulating frustrated antiferromagnet which does not display quantum spin liquid physics nor exotic spin excitations down to the lowest temperatures reached experimentally [16]. While it undergoes two structural transitions at TS1≃33T_{\mathrm{S1}} \simeq33 K and TS2≃13T_{\mathrm{S2}} \simeq13 K, it remains a simple paramagnet down to a very low Néel temperature TN≃2T_{\mathrm{N}} \simeq2 K above which phonons are the only possible heat carriers. In the high-temperature crystal structure (above TS1T_{\mathrm{S1}}) defined by the space group R3ˉR\bar{3} (148), the inter-plane hydrogen is randomly distributed over six equivalent positions, see Fig. 1b. The transition at TS2T_{\mathrm{S2}} is still to be clarified but the transition at TS1T_{\mathrm{S1}} is attributed to the positional freezing of this atom and potentially leads to a crystal structure defined by the space group P1P1 with three different twin domains. This system thus provides a unique platform in which one can investigate the PHE and correlate symmetry considerations with the temperature dependence of specific thermal transport coefficients.

We performed thermal transport measurements on three high-quality single crystals of Y-kapellasite, labelled S1, S2 and S3. We focused on five distinct configurations listed in Table 1 so as to assess the Q123Q_{123} (conventional κ12\kappa_{12}, in S1), Q213Q_{213} (conventional κ21\kappa_{21}, in S2), Q212Q_{212} (planar parallel κ21\kappa_{21}, in S2), Q211Q_{211} (planar orthogonal κ21\kappa_{21}, in S2) and Q321Q_{321} (conventional κ32\kappa_{32}, in S3) Righi-Leduc tensor components, where 1, 2 and 3 respectively denote the aa, b∗b^* and cc orthonormal lattice vectors of the R3ˉR\bar{3} structure, see Fig. 1b.

First examining the longitudinal thermal conductivities associated with the three corresponding heat flux directions, κ11\kappa_{11} (in S1), κ22\kappa_{22} (in S2) and κ33\kappa_{33} (in S3) as displayed in Figure 2, we obtain confirmation that there are no other heat carriers than phonons at all considered temperatures from 80 down to 2 K. The temperature dependence of the three κii\kappa_{ii} is similar in shape, with clear anomalies at TS1T_{\mathrm{S1}} and TS2T_{\mathrm{S2}} reflecting the two structural transitions, see Figure 2a. Actually, the three curves are almost identical up to a multiplicative factor which includes potential variations in crystalline quality and geometric factor uncertainties on top of any real anisotropy, with κ22:κ11:κ33≃1.00:1.43:5.30\kappa_{22} : \kappa_{11} : \kappa_{33} \simeq1.00 : 1.43 : 5.30 for a perfect match at 25 K, see Figure 2b. Here, the fact that κ33\kappa_{33} is about five times smaller than κ11\kappa_{11} and κ22\kappa_{22} reflects the quasi two-dimensional nature of the structure. Note that any contribution from mobile spin excitations, necessarily contained within the kagome planes and thus unable to carry heat along cc, would preclude such scaling and would be substantially affected by the magnetic field. On the contrary, κ22\kappa_{22} and κ33\kappa_{33} are found to be field independent up to 15 T while κ11\kappa_{11} displays a minute field dependence below TS1T_{\mathrm{S1}}, see Figure 2a. Most probably, the latter is related to the scattering of phonons by the paramagnetic spin fluctuations on a temperature range where short-range spin correlations (or paramagnons) start to develop [17]. Increasing the magnetic field (here applied along cc) gaps out some of these spin fluctuations, which in turn slightly enhances κ11\kappa_{11}.

Phonon longitudinal thermal conductivity in Y-kapellasite

Figure 2. Phonon longitudinal thermal conductivity in Y-kapellasite. a Temperature dependence of the three longitudinal thermal conductivities κ11\kappa_{11} (along aa, measured on sample S1), κ22\kappa_{22} (along b⋆b^{\star}, measured on sample S2) and κ33\kappa_{33} (along cc, measured on sample S3). Zero field data, in light colours, are plotted against B3=15B_3 = 15 T data, in dark colours (magnetic field applied along cc). b The zero field data are reproduced from a with a multiplicative factor applied to κ11\kappa_{11} (1.43) and κ33\kappa_{33} (5.30) for a perfect match with κ22\kappa_{22} at 25 K. The thick orange curve indicates the estimate of the low-temperature boundary scattering limit for κ22\kappa_{22}. The two vertical black dashed lines in a and b indicate the structural transitions at TS1≃33T_{S1} \simeq33 K and TS2≃13T_{S2} \simeq13 K.

It should also be pointed out that the three κii\kappa_{ii} are of remarkably modest magnitude, e.g., far below the low-temperature boundary scattering limit, and similar to the universal thermal conductivity of amorphous solids, see Figure 2b. In particular, when considering the evolution of κii\kappa_{ii} with increasing temperature, a key feature is the slow and almost linear rise observed after a plateau-like regime ending at TS1T_{\mathrm{S1}}. Such behaviour was explained at the theoretical level in terms of anharmonic interactions between phonons and fractons, which are short-scale vibrational excitations [18]. Here, it seems logical to attribute the glass-like thermal conductivity to a strong scattering of phonons by the randomly distributed inter-plane hydrogen. While the latter freezes below TS1T_{\mathrm{S1}}, κii\kappa_{ii} drops less rapidly with decreasing temperature: it remains more or less constant down to TS2T_{\mathrm{S2}}, at which it experiences a minor improvement (most noticeable in κ33\kappa_{33}) before vanishing.

Now turning to the thermal Hall effect results presented in Figure 3, we obtain small but finite Q123Q_{123}, Q213Q_{213} and Q212Q_{212} components, while the Q211Q_{211} component is found to be virtually null over the whole temperature range. As for the Q321Q_{321} component, we were not able to obtain satisfactory data owing to the poor κ33\kappa_{33} which prevented us from generating any detectable thermal Hall gradient ΔT2\Delta T_{2} given our experimental sensitivity. In other words, a clear PHE is observed in two of the conventional configurations and in the planar parallel configuration.

Conventional and planar parallel PHE in Y-kapellasite

Figure 3. Conventional and planar parallel PHE in Y-kapellasite. a Temperature dependence of the Q123Q_{123} Righi-Leduc component (measured on sample S1) plotted as κ12/B3\kappa_{12}/B_{3} for several magnetic fields from 7.5 to 17 T (applied along cc). b Temperature dependence of the four Righi-Leduc components Q123Q_{123} (conventional thermal Hall effect measured on sample S1), Q213Q_{213} (conventional thermal Hall effect measured on sample S2), Q212Q_{212} (planar parallel thermal Hall effect measured on sample S2) and Q211Q_{211} (planar orthogonal thermal Hall effect measured on sample S2), obtained in magnetic fields of 15 T and plotted as κij\kappa_{ij}. The two vertical black dashed lines in a and b indicate the structural transitions at TS1≃33T_{\mathrm{S1}} \simeq33 K and TS2≃13T_{\mathrm{S2}} \simeq13 K.

We first concentrate on the Q123Q_{123} component, which corresponds to the conventional thermal Hall conductivity κ12\kappa_{12}, see Figure 3a. It has a positive sign and a visible response to both structural transitions with a well-defined maximum position at TS1T_{\mathrm{S1}} and a fuzzier peak centred around TS2T_{\mathrm{S2}}. Within reproducibility tolerance, it scales linearly with the magnetic field up to 17 T, in stark contrast to expectations for the magnon Hall effect which declines with increasing magnetic field [9]. This weak conventional PHE translates into a thermal Hall angle κ12/κ22/B3≃7×10−5 T−1\kappa_{12}/\kappa_{22}/B_{3} \simeq7 \times10^{-5}\ \mathrm{T}^{-1} at TS1T_{\mathrm{S1}} in 15 T. It is instructive to note that κ12\kappa_{12} and κ22\kappa_{22} have a different temperature dependence. While a shape similarity between κij\kappa_{ij} and κjj\kappa_{jj} is often considered as an evidence that a single type of heat carriers (hence phonons) is involved [4, 14], we demonstrate here that this criterion is a condition that may be sufficient but by no means necessary. For instance, when considering the evolution with decreasing temperature, κ12\kappa_{12} increases down to TS1T_{\mathrm{S1}} whereas κ22\kappa_{22} decreases. Then, down to TS2T_{\mathrm{S2}}, κ12\kappa_{12} decreases more significantly than κ22\kappa_{22}, which remains more or less constant. The latter decrease reveals that κ12\kappa_{12} is not enhanced by the positional freezing of the inter-plane hydrogen at low temperatures.

We now consider the Q213Q_{213} component, which corresponds to the conventional thermal Hall conductivity κ21\kappa_{21}. According to the symmetry-adapted form of the Righi-Leduc tensor for the space groups R3ˉR\bar{3} and P1, Q213Q_{213} and Q123Q_{123} are Onsager-Casimir reciprocal, with Q213=−Q123Q_{213}=-Q_{123}, see Ext. Data Tables 1, 2. As depicted in Figure 3b, we confirm that Q213Q_{213} mirrors Q123Q_{123} with a negative sign and no significant magnitude difference. Note that this beautiful verification was carried out using two distinct samples, which underlines the reliability and reproducibility of our measurements.

Finally, we discuss the two most exciting components Q212Q_{212} and Q211Q_{211}, which correspond to the planar parallel and planar orthogonal thermal Hall conductivities κ21\kappa_{21} when the magnetic field is applied either parallel or orthogonal to the heat flux within the (ab)(ab) plane. We immediately rule out a spurious occurrence of Q212Q_{212} arising from a contamination by Q213Q_{213}. On the one hand, Q212Q_{212} is found to have the opposite sign, positive, and displays a different temperature dependence which may in turn suggest a different origin, see Figure 3b. In particular, we notice that Q212Q_{212} increases with decreasing temperature from TS1T_{\mathrm{S1}} to TS2T_{\mathrm{S2}} whereas ∣Q213∣\lvert Q_{213}\rvert decreases like Q123Q_{123}. On the other hand, we took great care to check the alignment of the magnetic field and estimate that it deviates from the (ab)(ab) plane towards cc by ±5 ∘\pm5\,^\circ at most. This corresponds to a maximal contamination of about 9% of ∣Q213∣\lvert Q_{213}\rvert, well below the observed Q212Q_{212}, with for instance 0.09×0.81≃0.07 mW K−1 m−10.09 \times0.81 \simeq0.07\ \mathrm{mW\,K^{-1}\,m^{-1}} versus 0.61 mW K−1 m−10.61\ \mathrm{mW\,K^{-1}\,m^{-1}} (more than eight times larger) at TS2T_{\mathrm{S2}}.

Detecting a finite Q212Q_{212} at temperatures above TS1T_{\mathrm{S1}} comes as a real surprise because it is forbidden in the symmetry-adapted form of the Righi-Leduc tensor for the space group R3ˉR\bar{3}, see Extended Data Tables 1. Its existence necessarily implies the occurrence of special symmetry breakings. We therefore listed all the space groups resulting from lowering the symmetries of R3ˉR\bar{3} down to P1P1, see Extended Data Fig. 1. Among all possible subgroups, only P1ˉP\bar{1} and P1P1 have symmetries compatible with a finite Q212Q_{212} (note that a finite Q211Q_{211} is also allowed), see Extended Data Tables 1, 2. Therefore, in contrast to the cases of Q123Q_{123} and Q213Q_{213}, the inter-plane hydrogen may play a crucial role in the establishment of Q212Q_{212}. We propose that a local symmetry breaking from R3ˉR\bar{3} to P1ˉP\bar{1} or P1P1 resulting from the random distribution of this atom is sufficient for the planar parallel PHE to emerge above TS1T_{\mathrm{S1}}. It is then only slightly enhanced when some if not all of the broken symmetries become global below TS1T_{\mathrm{S1}} and peaks around TS2T_{\mathrm{S2}}, see Figure 3b.

For now, it is also not clear why Q211Q_{211} remains vanishingly small but this observation, in line with other planar thermal Hall effect studies [1, 2, 4], may prove useful in subsequent theoretical developments to clarify the precise mechanisms responsible for the planar PHE. For comparison, QijiQ_{iji} and QijjQ_{ijj} are correspondingly reported to be finite and null in both α\alpha-RuCl3_3 (in agreement with the symmetries of the space group C2/mC2/m) and Na2_2Co2_2TeO6_6 (although a finite QijiQ_{iji} is there forbidden by the symmetries of the space group P6322P6_3 22).

In summary, our thermal transport measurements on Y-kapellasite yield a paradigm shift in the study of the thermal Hall effect in insulators. First, we report the first unambiguous detection of a planar parallel PHE. Our findings thus prompt a second look at the interpretations put forward in previous studies on quantum spin liquid candidates and open up a wider range of scenarios in which the phonon contribution cannot be neglected. Second, we now have a specific and concrete case for the general idea of local symmetry breaking at the root of the PHE in a wide range of insulators.

Methods

Structure and magnetic model of Y-kapellasite. Y-kapellasite is a derivative of the emblematic quantum spin liquid candidate herbertsmithite ZnCu3(OH)6Cl2\mathrm{ZnCu}_{3}(\mathrm{OH})_{6}\mathrm{Cl}_{2}, discovered as an interesting by-product of unsuccessful doping attempts when substituting divalent zinc for trivalent yttrium [19, 20]. In this material, copper spins S=1/2S = 1/2 decorate a slightly distorted kagome lattice with yttrium located close to the centre of hexagons, see Fig. 1b, hence the kapellasite denomination (Zn-kapellasite is a polymorph of herbertsmithite, the difference between the two structures being the position of zinc, either within or in between the kagome planes). Contrary to zinc, yttrium has a significantly larger ionic radius than copper, which precludes any intersite mixing and renders the system immune to the troublesome magnetic defects typical of herbertsmithite, Zn-kapellasite or Zn-barlowite. Hydroxyl groups and chlorine constitute thick diamagnetic layers separating the kagome planes and recent ab initio density functional theory combined with inelastic neutron scattering results confirmed the quasi two-dimensional nature of the magnetic lattice, demonstrating that three intra-plane antiferromagnetic Heisenberg couplings between nearest neighbours dominate over all other possible intra- or inter-plane couplings between further neighbours [16, 21]. These main three couplings J≃J▽≃140J \simeq J_{\bigtriangledown} \simeq140 K and J′≃63J' \simeq63 K result in an original anisotropic variant of the standard nearest neighbour Heisenberg model (recovered for J=J▽=J′J = J_{\bigtriangledown} = J'), breaking translational symmetry of the kagome lattice but retaining six-fold rotational symmetry around hexagons, see Fig. 1b. Owing to the considerable frustration produced by the lattice geometry and the competition between the latter three antiferromagnetic terms, the material remains paramagnetic down to TN≃2T_{\mathrm{N}} \simeq2 K, below which a coplanar long-range order with propagation vector Q=(1/3,1/3)Q = (1/3, 1/3) sets in, as predicted theoretically for the ground state [21]. This magnetic transition, resulting in a remarkably weak ordered moment of about 1/30 μB1/30\,\mu_{\mathrm{B}}, initially remained elusive when focusing on polycrystalline samples [20] until it was later demonstrated in the case of large phase-pure single crystals prepared via optimal synthesis [16]. For that matter, every improvement in the synthesis procedure triggered a thorough reappraisal of the exact stoichiometry and crystallographic structure which were eventually settled through inductively coupled plasma mass spectroscopy, gas extraction and detailed neutron diffraction measurements from 40 K down to 65 mK [16]. Y-kapellasite crystallises in a trigonal rhombohedral structure defined by the space group R3‾R\overline{3} (148), in which the inter-plane hydrogen randomly occupies six equivalent positions, thus locally breaking the global crystal symmetry, see Fig. 1b. Upon cooling, two structural transitions occurring at TS1≃33T_{\mathrm{S1}} \simeq33 K and TS2≃13T_{\mathrm{S2}} \simeq13 K were detected through specific heat, thermal expansion and 35Cl{}^{35}\mathrm{Cl} NMR measurements on single crystals while they remained unnoticed in prior studies of polycrystalline samples [20]. Strikingly, these transitions only reflect in the neutron diffraction data through a clear intensity increase for some Bragg peaks. Preserving the same space group and lattice parameters down to the lowest temperatures does not affect the refinement quality. The transition at TS1T_{\mathrm{S1}} is attributed to the positional freezing of the inter-plane hydrogen, thereby potentially leading to a global crystal symmetry breaking from R3‾R\overline{3} to P1 (1) with three different twin domains. This is compatible with the complex 35Cl{}^{35}\mathrm{Cl} quadrupolar line splitting reported below TS1T_{\mathrm{S1}}. Further terahertz magnetometry measurements revealed that these structural transitions are accompanied by the building up of short-range spin correlations (or paramagnons) although long-range magnons only emerge below TNT_{\mathrm{N}} as highlighted with inelastic neutron scattering [16, 17].

Optimal synthesis of Y-kapellasite. The crystal growth of Y-kapellasite was originally reported in Reference [19]. Subsequently, as described in Reference [17], the synthesis was improved to obtain inclusion-free, large, bulk single crystals by means of a horizontal external gradient method in thick-walled quartz ampoules with a wall thickness of 2.5–3 mm. Growth is achieved by slowly dissolving CuO in a YCl3_3-H2_2O solution and transporting it to the cold end. This is realised in a three-zone furnace with a gradient of 25 °C and a temperature of 240 °C at the hot end, over a length of 20 cm. The gradient was optimised because too low temperatures yielded a phase mixture of Y-kapellasite and clinoatacamite. The phase-pure, optically transparent single crystals have an average size of 3 × 3 × 1 mm3^3 up to 3 × 3 × 3 mm3^3 when grown over several weeks. Their orientation is facilitated by their hexagonal plaquette shape, with cc perpendicular to the hexagonal faces and aa (respectively b∗b^{*}) perpendicular (respectively parallel) to the hexagonal edge. The samples S1, S2 and S3 examined here are from the same batch as the single crystals investigated in References [16, 17]. S1 and S2 were selected among the thinnest and measured as grown, while S3 was cut along cc in one of the thickest.

Thermal transport measurements. Thermal transport measurements were performed using a standard steady-state method. A constant heat flux qiq_i is injected at one end of the sample along the direction ii while the other end is thermally sunk to a heat bath at temperature T0T_0 (either a copper or lithium fluoride block), see Fig. 1a. The heat flux is generated by applying an electric current through a strain gauge whose resistance (of about 5 kΩ) marginally depends on the temperature and magnetic field. Assuming a one-dimensional heat flow and an isotropic medium, the longitudinal thermal gradient ΔTi\Delta T_i is measured between two contacts separated by a distance ll along the direction ii, see Fig. 1a. This gradient is assessed using either two Cernox sensors (calibrated in situ against a reference Cernox) or two type-E thermocouples. The longitudinal thermal conductivity κii\kappa_{ii} is given by κii=qi/(αΔTi)\kappa_{ii}=q_i/(\alpha\Delta T_i), where α\alpha is a geometric factor determined by the cross section wtwt (ww: width, tt: thickness) divided by ll. In presence of an applied magnetic field BxB_x, with x∈{i,j,k}x \in\{i,j,k\}, the orthogonal thermal gradient ΔTj\Delta T_j is measured between two contacts separated by a distance ww along the direction jj, see Fig. 1a. This gradient is assessed using a differential type-E thermocouple and antisymmetrised between the two field polarities to remove any contamination by the longitudinal gradient: ΔTj(Bx)=[ΔTj(+Bx)−ΔTj(−Bx)]/2\Delta T_j(B_x)=[\Delta T_j(+B_x)-\Delta T_j(-B_x)]/2. The thermal Hall conductivity κij\kappa_{ij} is given by κij=lκjjΔTj/(wΔTi)\kappa_{ij}=l\kappa_{jj}\Delta T_j/(w\Delta T_i), which implies a two-step computation because κjj\kappa_{jj} and ΔTj\Delta T_j are not measured simultaneously. Error bars in the figures represent one standard deviation. In our mountings, all connections between the samples and the heat baths, temperature sensors and strain gauges consisted of gold and silver wires with a 17 to 100 μ\mum diameter attached using silver paste. The contacts had geometries (l×w×t)(l \times w \times t) 1859(60) × 2239(78) × 334(10) μ\mum3^3 on S1, 558(108) × 1040(60) × 85(4) μ\mum3^3 on S2 and 418(70) × 600(99) × 137(38) μ\mum3^3 on S3. Note that a quantitative determination of κij\kappa_{ij} requires measuring both κij\kappa_{ij} and ΔTj\Delta T_j in the same sample. Here, the longitudinal thermal conductivity κ11\kappa_{11} and the thermal Hall gradient ΔT2\Delta T_2 were measured in sample S1 while the longitudinal thermal conductivity κ22\kappa_{22} and the thermal Hall gradient ΔT1\Delta T_1 were measured in sample S2. Owing to the modest anisotropy between κ11\kappa_{11} and κ22\kappa_{22} (if any beyond potential variations in crystalline quality and geometric factor uncertainties), see Figure 2, we assumed that κ11≃κ22\kappa_{11}\simeq\kappa_{22} to compute κ21\kappa_{21} so as to compare κ12\kappa_{12} and κ21\kappa_{21} in a meaningful way.

Acknowledgements We are grateful to S. Fortier for extensive technical support and acknowledge valuable discussions with N. Gauthier and J.A. Quilliam. C.K. acknowledges funding from the Deutsche Forschungsgemeinschaft (DFG) through TRR 288-422213477 (project A03). L.T. acknowledges support from the Canadian Institute for Advanced Research (CIFAR) as a Fellow and funding from the Institut Quantique, the Natural Sciences and Engineering Research Council of Canada (NSERC, PIN 123817), the Fonds de Recherche du Québec - Nature et Technologies (FRQNT), the Canada Foundation for Innovation (CFI), and a Canada research chair. This research was undertaken thanks in part to funding from the Canada First Research Excellence Fund.

Author contributions Q.B., P.P. and L.T. conceived and led the project. P.P., K.M.Z. and C.K. grew the single crystals. Q.B., É.L., L.C., and A.V. carried out the thermal transport measurements and analysis. Q.B. wrote the manuscript with feedback from all the authors.

Competing interests The authors declare no competing interests.

Correspondence Correspondence and requests for materials should be addressed to Q.B., P.P. or L.T.

Thermal transport setup, structure and magnetic model of Y-kapellasite

Figure 1. Thermal transport setup, structure and magnetic model of Y-kapellasite. a Sketch of the thermal transport measurement setup showing the directions of the heat flux qiq_i, the longitudinal thermal gradient ΔTi\Delta T_i, the thermal Hall gradient ΔTj\Delta T_j and the applied magnetic field BxB_x, with x∈{i,j,k}x \in\{i,j,k\}. (i,j,k)(i,j,k) is an orthonormal basis. In a conventional thermal Hall effect configuration, the magnetic field is applied fully along kk. In a planar parallel thermal Hall effect configuration, the magnetic field is applied fully along ii. In a planar orthogonal thermal Hall effect configuration, the magnetic field is applied fully along jj. b Crystal structure of Y-kapellasite [Y3_3Cu9_9(OH)19_{19}Cl8_8], with space group R3ˉR\bar{3} (148), deduced from neutron diffraction measurements at 40 K (above TS1T_{\mathrm{S1}}) and viewed along the cc axis [16]. The orthonormal lattice vectors aa, b∗b^* and cc are numbered 1, 2 and 3. Note the 1/6 partial occupation of the inter-plane hydrogen positions (highlighted in pink). The three main antiferromagnetic couplings J≃J◊≃140J \simeq J_{\lozenge} \simeq140 K and J′≃63J' \simeq63 K, indicated by thick coloured bonds, define an anisotropic nearest neighbour Heisenberg model on the slightly distorted copper kagome lattice.

ConfigurationSampleHeat flux qiq_iThermal Hall gradient ΔTj\Delta T_jApplied magnetic field BxB_xRighi-Leduc tensor component QijxQ_{ijx}
1S1∥a(1)\parallel a^{(1)}∥b⋆(2)\parallel b^{\star(2)}∥c(3)\parallel c^{(3)}Q123Q_{123}
2S2∥b⋆(2)\parallel b^{\star(2)}∥a(1)\parallel a^{(1)}∥c(3)\parallel c^{(3)}Q213=−Q123Q_{213}=-Q_{123}
3S2∥b⋆(2)\parallel b^{\star(2)}∥a(1)\parallel a^{(1)}∥b⋆(2)\parallel b^{\star(2)}Q212Q_{212}
4S2∥b⋆(2)\parallel b^{\star(2)}∥a(1)\parallel a^{(1)}∥a(1)\parallel a^{(1)}Q211Q_{211}
5S3∥c(3)\parallel c^{(3)}∥b⋆(2)\parallel b^{\star(2)}∥a(1)\parallel a^{(1)}Q321Q_{321}

Table 1. Details of the five thermal transport configurations that were adopted. For each configuration, the directions of the heat flux, the thermal Hall gradient and the applied magnetic field are specified along with the corresponding Righi-Leduc tensor component. Conventional thermal Hall effects were investigated in samples S1, S2 and S3 using configurations 1, 2 and 5 respectively. Planar parallel and planar orthogonal thermal Hall effects were investigated in sample S2 using configurations 3 and 4 respectively.

Graph of maximal subgroups from $\mathrm{R}\bar{3}$ to P1

Figure 1 (PDF p. 14). Graph of maximal subgroups from R3ˉ\mathrm{R}\bar{3} (148) to P1 (1). The space groups P3ˉ\mathrm{P}\bar{3} (147), R3 (146), P32_2 (145), P31_1 (144), P3 (143), P1ˉ\mathrm{P}\bar{1} (2) and P1 are obtained by gradually breaking down the symmetries of R3ˉ\mathrm{R}\bar{3}. This graph was produced using the SUBGROUPGRAPH module of the Bilbao Crystallographic Server [22].

QijkQ_{ijk}kk
112233
1111000000
12120000Q123Q_{123}
1313Q131Q_{131}Q132Q_{132}00
21210000−Q123-Q_{123}
2222000000
2323−Q132-Q_{132}Q131Q_{131}00
3131−Q131-Q_{131}−Q132-Q_{132}00
3232Q132Q_{132}−Q131-Q_{131}00
3333000000

Table 1 (PDF p. 15). Symmetry-adapted form of the Righi-Leduc tensor QijkQ_{ijk} for the space groups R3ˉ\mathrm{R}\bar{3} (148), P3ˉ\mathrm{P}\bar{3} (147), R3\mathrm{R}3 (146), P32\mathrm{P}3_{2} (145), P31\mathrm{P}3_{1} (144) and P3\mathrm{P}3 (143). Finite Q123Q_{123} and Q213=−Q123Q_{213}=-Q_{123} components, corresponding to conventional thermal Hall effects, are allowed. Finite Q211Q_{211} and Q212Q_{212} components, corresponding to planar orthogonal and planar parallel thermal Hall effects, are forbidden. This table was produced using the TENSOR module of the Bilbao Crystallographic Server [23].

QijkQ_{ijk}kk
123
ijij
11000
12Q121Q_{121}Q122Q_{122}Q123Q_{123}
13Q131Q_{131}Q132Q_{132}Q133Q_{133}
21−Q121-Q_{121}−Q122-Q_{122}−Q123-Q_{123}
22000
23Q231Q_{231}Q232Q_{232}Q233Q_{233}
31−Q131-Q_{131}−Q132-Q_{132}−Q133-Q_{133}
32−Q231-Q_{231}−Q232-Q_{232}−Q233-Q_{233}
33000

Table Extended Data Table 2. Symmetry-adapted form of the Righi-Leduc tensor QijkQ_{ijk} for the space groups P1ˉ\mathrm{P}\bar{1} (2) and P1\mathrm{P}1 (1). Finite Q123Q_{123} and Q213=−Q123Q_{213}=-Q_{123} components, corresponding to conventional thermal Hall effects, are allowed. Finite Q211=−Q121Q_{211}=-Q_{121} and Q212=−Q122Q_{212}=-Q_{122} components, corresponding to planar orthogonal and planar parallel thermal Hall effects, are also allowed. This table was produced using the TENSOR module of the Bilbao Crystallographic Server [23].

References

  1. [1]T. Yokoi et al., Science 373, 6554, 568-572 (2021)
  2. [2]P. Czajka et al., Nat. Mater. 22, 36-41 (2023)
  3. [3]H. Takeda et al., Phys. Rev. Res. 4, L042035 (2022)
  4. [4]L. Chen et al., arXiv:2309.17231 (2023)DOI
  5. [5]L. Chen et al., arXiv:2310.07696 (2023)DOI
  6. [6]C. Strohm et al., Phys. Rev. Lett. 95, 155901 (2005)
  7. [7]A. Kitaev, Ann. Phys. 321, 1, 2-111 (2006)
  8. [8]H. Katsura et al., Phys. Rev. Lett. 104, 066403 (2010)
  9. [9]Y. Onose et al., Science 329, 5989, 297-299 (2010)
  10. [10]M. Hirschberger et al., Phys. Rev. Lett. 115, 106603 (2015)
  11. [11]Y. H. Gao and G. Chen, SciPost Phys. Core 2, 004 (2020)
  12. [12]Y. Teng et al., Phys. Rev. Res. 2, 033283 (2020)
  13. [13]L. E. Chern et al., Phys. Rev. Lett. 126, 147201 (2021)
  14. [14]É. Lefrançois et al., Phys. Rev. X 12, 021025 (2022)
  15. [15]M. Gillig et al., arXiv:2303.03067 (2023)DOI
  16. [16]D. Chatterjee et al., Phys. Rev. B 107, 125156 (2023)
  17. [17]T. Biesner et al., Adv. Quantum Technol. 5, 2200023 (2022)
  18. [18]S. Alexander et al., Phys. Rev. B 34, 2726 (1986)
  19. [19]P. Puphal et al., J. Mater. Chem. C 5, 2629-2635 (2017)
  20. [20]Q. Barthélémy et al., Phys. Rev. Mater. 3, 074401 (2019)
  21. [21]M. Hering et al., npj Comput. Mater. 8, 10 (2022)
  22. [22]S. Ivantchev et al., J. Appl. Cryst. 33, 1190-1191 (2000)
  23. [23]S.V. Gallego et al., Acta Cryst. A 75, 438-447 (2019)

Paper details

vlnc:W_ue5sz63b
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents