WorksPhysics and Astronomy
Ruddlesden-Popper chalcogenides push the limit of mechanical stiffness and glass-like thermal conductivity in single crystals
Abstract
Insulating materials featuring ultralow thermal conductivity for diverse applications also require robust mechanical properties. Conventional thinking, however, which correlates strong bonding with high atomic-vibration-mediated heat conduction, led to diverse weakly bonded materials that feature ultralow thermal conductivity and low elastic moduli. One must, therefore, search for strongly-bonded single crystals in which heat transport is impeded by other means. Here, we report intrinsic, glass-like, ultralow thermal conductivity and ultrahigh elastic-modulus/thermal-conductivity ratio in single-crystalline Ruddlesden-Popper Ban+1ZrnS3n+1, n = 2,3, which are derivatives of BaZrS3. Their key features are strong anharmonicity and intra-unit-cell rock-salt blocks. The latter produce strongly bonded intrinsic superlattices, impeding heat conduction by broadband reduction of phonon velocities and mean free paths and concomitant strong phonon localization. The present study initiates a paradigm of mechanically stiff phonon glasses.
Jordan A. Hachtel
Center for Nanophase Materials Sciences, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37830, USA
Jayakanth Ravichandran
Mork Family Department of Chemical Engineering and Materials Science, University of Southern California, Los Angeles, California 90089, USA
Email: [email protected]
Sokrates T. Pantelides
Department of Physics and Astronomy, Vanderbilt University, Nashville, Tennessee 37235, USA
Department of Electrical and Computer Engineering, Vanderbilt University, Nashville, Tennessee 37235, USA
Email: [email protected]
Patrick E. Hopkins
Department of Mechanical and Aerospace Engineering, University of Virginia, Charlottesville, Virginia 22904, USA
Department of Materials Science and Engineering, University of Virginia, Charlottesville, Virginia 22904, USA
Department of Physics, University of Virginia, Charlottesville, Virginia 22904, USA
Email: [email protected]
Md Shafkat Bin Hoque, Eric R. Hoglund, and Boyang Zhao contributed equally to this work.
Jayakanth Ravichandran, Sokrates T. Pantelides, and Patrick E. Hopkins are corresponding authors.
Introduction
Designing materials with ultralow thermal conductivity () without reducing their density and degrading their mechanical properties typically evades century-old theories on microscopic heat conduction. In non-metallic crystals, the primary modes of thermal transport are lattice vibrations, namely phonons. Dating back to theories originally pioneered by Peierls, Leibfreid and Schlomann, Einstein, and Debye, the thermal transport of atomic vibrations are directly related to the interatomic bond strengths.[1, 2, 3, 4, 5, 6, 7, 8, 9] Hence, "ultralow" thermal conductivity in solids often comes at the cost of “weak” interatomic-bond strength and “soft” elastic modulus (), limiting their mechanical performance.[10, 11, 12, 13, 14] In recent years, a family of halide perovskite-structure crystalline solids has been found to feature ultralow thermal conductivities, of order 0.2-0.3 W m K, but their weak van der Waals (vdW) bonding leads to very low values of elastic and shear moduli.[15, 16, 17, 18, 19] Breaking this Pareto-normality in the design of crystals to create ultrahigh materials must, therefore, involve different strategies than those traditionally proposed to limit phonon transport. Such decoupling would have major implications in many technological applications, such as thermal barrier coatings,[20] thermoelectrics,[21, 22] and memory devices.[23, 24]
To overcome the barriers of achieving ultrahigh materials and create new directions for achieving single crystals with ultralow thermal conductivity, one seeks to inhibit the propagation of phonons in a “hard” lattice. In perfect crystals, phonons travel with intrinsic group velocities and are inhibited by phonon-phonon scattering caused by anharmonic effects. Historically, chalcogenides (S, Se, and Te) have been investigated for ultralow thermal conductivity because of strong anharmonicity (e.g., PbSe, BiTe, SnSe).[25, 26, 27, 28, 29] However, many of these materials tend to crystallize in layered structures with vdW interactions, leading to poor mechanical properties across the layers due to ease of shearing. Thus, covalently-bonded layered chalcogenides present a potential pathway to low thermal conductivity and high elastic modulus in the direction across the layers. We, therefore, present Ruddlesden-Popper (RP) sulfides[30, 31] as potential candidates for this scenario. Unlike some RP halide perovskites that are bonded by weak vdW forces,[16, 19] the RP sulfides hold both the promise of high anharmonicity and strong bonding in a layered, superlattice-like structure with intrinsic interfaces.
Herein, we report on the ultralow thermal conductivity of RP phases , and 3 of barium zirconium sulfide () single crystals, enabled by strong anharmonicity and a large fraction of localized and low-velocity vibrational modes throughout the entire vibrational spectrum, thus achieving broadband attenuation of thermal transport. We reveal the origin of the ultralow thermal conductivities in these RP derivatives of the chalcogenide perovskite using a combination of experiments and first-principles- and machine learning-driven computational approaches. Unlike previously studied vdW layered crystals, including the RP halides, [16, 19] the strong Ba-S and Zr-S chalcogenide bonds across the rock-salt layers and the layers bring elastic-moduli values of the RP phases nearly an order of magnitude higher than other ultralow-thermal-conductivity inorganic crystals. As a result, the RP phases of these single crystals exhibit record setting values of for any single-crystalline material discovered to date, while maintaining glass-like, ultralow thermal conductivities. Such a useful combination of thermal and mechanical properties makes the materials highly important in the fields of thermal barrier coatings and thermoelectric materials.[33]
Results and discussion
To investigate the impact of sub-unit cell structures on thermal conductivity, we consider perovskite and two RP phases, and . The perovskite shown in Figure 1(a) consists of tilted octahedra (Figure 1(c)) and polyhedra (Figure 1(d)). The RP-phase shown in Figure 1(b) contains two perovskite sections (red brackets) that are separated by rock-salt packed layers (Figure 1(e)). differs from by adding one more BaS and atomic layer to each perovskite section. Figure 1(f) shows an iDPC image of and Figure 1(g,Figure 1h) show the layered periodic stacking of perovskite layers in the two RP phases with the enlarged regions of interest, emphasizing one of two rock-salt layers that are present in a single unit cell. The elemental maps generated by electron-energy-loss spectroscopy (EELS) and the corresponding atomic-number-contrast (Z-contrast) image shown in Figure 1(i-Figure 1l) show the high degree of chemical ordering in each sublattice and the change in local symmetry at the rock-salt layers.

Figure 1. Structure of perovskite and Ruddlesden-Popper . Ball-and-stick model of (a) Pnma perovskite and (b) I4mmm Ruddlesden-Popper showing grey octahedra and green BaS bonds. Red markers in (b) indicate the perovskite blocks of the RP phase separated by the rock-salt blocks. Blue dotted lines between BaS atomic planes indicate the midplane of the rock-salt blocks. Ball-and-stick model of a (c) Zr octahedra, (d) undistorted Ba polyhedral, and (e) rock-salt building block resulting from layering in . Integrated differential phase contrast images of the (f) , (g) , (h) crystals. Enlarged regions from the cyan annotations are shown below each image. In the enlargements, two perovskite unit cells are annotated with Ba (green), Zr (grey) and S (yellow) circles. (i) Z-contrast image and (j) BaZr composite image from a STEM-EELS spectrum image. Intensity maps of (k) Ba-M, (l) Zr-M background-subtracted edges. (m,n) Section views along (200) and (110) of the valence electron density of the RP-phase calculated by DFT, respectively. Red marks and blue dashed lines help to correlate atomic structure to that of panel (b). (o) A line profile of the valence electron density along the white dashed line in (m). Blue bars illustrate the rock-salt-block regions.
To understand the bonding in the RP phases, density functional theory (DFT) calculations were performed on to calculate the valence electron density, as shown in Figure 1(m,Figure 1n). A line profile along the dashed white line in Figure 1(m) is shown in Figure 1(o) for a more quantitative evaluation. The valence electron density within the rock-salt regions is non-zero and comparable to those inside the perovskite blocks, which suggests similar intra- and inter-perovskite-block bonding strength. In other words, the bonding in the rock-salt regions, namely across the “gaps” highlighted in Figure 1(b), is not of the weak, vdW type. The overall strong bonding is also reflected in the calculated elastic moduli, which have comparable values along the cross-plane and in-plane directions. The elastic-moduli values are three to four times higher than those observed in other ultralow-thermal-conductivity halide perovskites, such as, , [17] , [18] the RP-phase . [16] and several metal halide perovskites (see Supplemental Table S3). The presence of strong intra- and inter-perovskite-block bonding strength in the sulfide RP phases described here mitigates the role of bond strength in the observed ultralow thermal conductivity.
To understand phonon transport through the structures, we measured the cross-plane (-axis) thermal conductivity using time-domain thermoreflectance (TDTR) from 100 to 400 K, as shown in Figure 2(a). The thermal conductivities exhibit several unusual features for single-crystalline materials. First, the thermal conductivity of crystalline increases from 100 to 250 K, then remains relatively temperature independent. Such trend is observed in amorphous materials and disordered crystals, but it is rare in single crystals.[7] The weak or negligible temperature dependence cannot be explained by prior first-principles three-phonon scattering calculations [34] (the purple dashed curve in Figure 2a). In contrast, our DFT-based machine-learning-interatomic-potential (MLIP)-driven molecular-dynamics (MD) simulation results (orange open dots in Figure 2a) show good overall agreement with experimentally measured data, with only small differences above 200 K. This agreement demonstrates the accuracy of MLIP-MD (or MLMD) simulations, which implicitly capture all the atomic-vibration contributions to thermal conductivity, including the effects of three-phonon scattering, four-phonon scattering, finite-temperature phonon renormalization, temperature correction to force constant, and diffuson contributions.[35, 36] This agreement also confirms that the glass-like thermal conductivity trend of single-crystalline is not from extrinsic defects, but rather mechanisms intrinsic to the crystal. However, the exact mechanism behind the temperature trend in the thermal conductivity of single-crystalline remains an open question. Supporting Figure S9 shows that perovskites have been reported to exhibit both glass- and crystalline-like thermal conductivity trends. Isolating the mechanisms that underlie the thermal conductivities observed in this class of materials is beyond the scope of the present study, as we focus on the RP phases.

Figure 2. Thermal conductivities – experimental and simulated. (a) TDTR-measured cross-plane thermal conductivity of and its RP derivatives. c- and a- denote crystalline and amorphous , respectively. For comparison, we also include the three-phonon and MLMD predicted thermal conductivity of and RP phases. The three-phonon prediction is adopted from Osei-Agyemang et al. [34] The minimum limit and diffuson limit refer to Cahill et al. [4] and Agne et al.’s [6] theoretical models. (b) Thermal conductivity distribution of and RP phases as a function of heavy ion-irradiation doses. The thermal conductivity of the crystalline is lower compared to panel (a) due to the presence of nano-domains (see Supporting Information for details). (c) Anisotropic thermal conductivity of measured by TDTR.
The RP phases and possess ultralow thermal conductivities, i.e., and , respectively, over a relatively large temperature range. These values are times lower than those of single-crystalline , in agreement with the Agne et al.’s diffuson limit,[6] and lower than Cahill et al.’s glass limit. [4] These two limits are two of the most commonly used theoretical models to predict the lowest possible thermal conductivity of a crystalline material. Additionally, the thermal conductivities of (green diamonds) and (blue triangles) show glass-like temperature trends, comparable to that of amorphous (a-) (black circles). The MLMD thermal conductivity simulations of show quantitative agreement with the experimental data. The cross-plane thermal conductivities of the sulfide RP phases are even lower than those of vdW layered materials, except when thin film samples and interlayer rotations are involved. [10, 37] On the other hand, the sulfide RP phases feature far superior mechanical properties of all vdW layered materials.
Defects in materials have also been known to lead to ultralow thermal conductivity and glass-like temperature trends.[7] To rule out the influence of defect scattering on our experimental data, we irradiated the BaZrS and BaZrS crystals with high energy gold ions. The measured cross-plane thermal conductivities of the heavily ion-irradiated crystals as a function of ion dose are shown in Figure 2(b). The thermal conductivity of BaZrS exhibits a sigmoidal reduction, typically characteristic of irradiated crystalline materials.[38, 39] At low doses, irradiation introduces low concentrations of clustered point defects and vacancies. The overall crystal structure remains relatively unchanged, whereby the thermal conductivity is nearly constant. At high doses, point-defect concentrations increase and damaged regions overlap, which gradually decreases the thermal conductivity to that of an amorphous solid. These observations, together with the agreement of the simulations of the thermal conductivity of c-BaZrS with the experimental data, also prove the high quality of our unirradiated BaZrS single-crystals. Large concentrations of defects in the unirradiated crystals would have caused the thermal conductivity to drop below the shown value.
Compared to BaZrS, a completely different thermal conductivity trend is observed vs. ion dose in the RP phases; the thermal conductivity of the BaZrS crystals remains nearly constant regardless of gold ion dose. TEM micrographs show that the layering of BaZrS crystals remains uninterrupted throughout the range of doses, although high doses can introduce amorphous pockets (see Supporting Information). In vdW layered materials with interlayer rotation, ion irradiation can lead to increased thermal conductivity due to increases in interatomic bonding.[10] However, due to the already strong bonding of the RP phases, no such trend is observed in the present study. The resistance of the RP crystals to irradiation damage makes them a highly suitable thermal barrier coating in deep space applications in radiation environments.
To gain insight into the role of anisotropy in ultralow thermal conductivity of the RP phases, we measured the thermal conductivity of BaZrS along the in-plane direction (perpendicular to -axis) at room temperature as shown in Figure 2(c). The in-plane thermal conductivity is 1.06 0.14 W m K which is 2.5 times higher than the cross-plane thermal conductivity (0.45 0.07 W m K). Noteworthy, the MLMD-simulated in-plane and cross-plane thermal conductivities are 1.02 0.17 and 0.53 0.02 W m K, respectively, showing excellent quantitative agreement with the experimental anisotropy. The major structural difference between the in-plane and cross-plane directions is the periodic rock-salt and perovskite layers, relative to continuous layers in-plane. This result suggests that, despite the strong bonding, the rock-salt layers are causing the ultralow and anisotropic thermal conductivity in the RP phases.
To understand how the introduction of rock-salt layering leads to the ultralow thermal conductivities in the RP phases compared to that of crystalline BaZrS, we consider three factors that may contribute to ultralow thermal conductivity: 1) anharmonic scattering, 2) decreased phonon group velocities (), and 3) phonon localization within unit cells.[40]
To assess the role of anharmonic scattering, we calculated the spectral energy density (SED) of BaZrS and BaZrS based on MLMD, as shown in Figure 3(a,Figure 3b). Compared to a typical semiconductor (e.g., silicon, gallium nitride), the SEDs of both BaZrS and BaZrS show much more blurred and broadened linewidths, indicating strong anharmonicity and large phonon scattering rates.[40, 41] These features could be a critical reason for the ultralow thermal conductivity. The large overlapping of branches due to the anharmonic broadening also indicates that the interband tunneling (i.e., diffusions) should be significant, based on the Wigner formalism.[42, 43] This feature could be a reason for the glass-like thermal conductivity exhibited by the experimental data. Recent calculations using the Wigner formalism show that diffusion contributes 30% to thermal conductivity of BaZrS. We expect the diffusion contribution to be even larger in BaZrS since the broadening of SED is more significant.

Figure 3. Evaluation of factors that influence thermal conductivity. Spectral energy density calculated from MLMD for BaZrS and (b) BaZrS. (c) of the BaZrS and RP structure along the cross-plane direction as a function of frequency with the dashed black line serving as a reference for air. (d) Cumulative number of phonon modes having up to a given group velocity as a function of frequency for the crystals along cross- () and in- () plane directions. (e) Participation ratio (PR) as a function of frequency for BaZrS and (b) BaZrS. (f) Cumulative number of phonon modes having up to a given PR as a function of frequency.
Group velocity () is another important parameter to understand the diffusivity of vibrational modes in a material.[13] The group velocities along the cross-plane direction () of BaZrS and BaZrS are compared in Figure 3(c). The latter shows ultralow group velocities compared to the former at nearly all frequencies, except for very few optical modes, whose fraction is very small. To quantify the number of modes having ultralow group velocities, we plot the cumulative number of phonon modes as a function of group velocity in Figure 3(d). We find that 80% of phonon modes of BaZrS have a lower than the sound speed of air, which is astonishing. In comparison, the component shows a larger group velocity, which is similar to that of BaZrS. This feature would indicate that the in-plane thermal conductivity of BaZrS is similar to that of BaZrS but the cross-plane thermal conductivity would be much lower, which is consistent with both the experimental observations and the simulations. Thus, the ultralow phonon velocities, induced by the presence of periodic building blocks in unit cells, is another key contributing factor to the ultralow thermal conductivities.
The third possible factor contributing to ultralow thermal conductivity is phonon localization. For that, we calculate the participation ratio (PR) of the vibrational modes, which represents the spatial localization of phonon waves within a unit cell. Localized vibrational modes are usually defined as having a participation ratio lower than 0.1.[44] As shown in Figure 3(e,Figure 3f), the participation ratio of is significantly lower than that of across all frequencies. It is noteworthy that some low-frequency modes ( THz) in the RP phase have a participation ratio that is smaller than 0.1, which is comparable to the localization expected for locons in amorphous materials.[44, 45] This result provides evidence that the presence of rock-salt building block layers in the RP phases causes a significant number of vibrational modes to become highly localized.[18]
To further show the localization of the vibrational modes, we estimate the average mean free path of phonons in in Supporting Figure S10. Assuming the diffuson thermal conductivity is zero, the average mean free path of phonons is estimated as 1 nm, to match with experimental thermal conductivity. Since diffuson contribution is nonzero, the actual mean free path of phonons should be smaller than 1 nm, the inter-gap thickness. These considerations indicate that the phonons are localized inside the rock-salt layers of by the gaps, being consistent with the participation ratio results.
Figure S10. Comparison of phonon thermal conductivity assuming an average mean free path of 1 nm with experimental data.
In summary, the presence of intra-unit-cell rock-salt blocks in the sulfide RP phases derived from and the corresponding selenide and telluride RP phases effectively produces strongly bonded, intrinsic superlattices with different periodicities and interfacial regions that largely reduce phonon velocities and mean free paths, inducing strong localization. Combined with the strong anharmonicity that is known to be intrinsic to chalcogenides,[25, 26, 27, 28, 29] the chalcogenide RP phases are a class of single-crystalline materials that can achieve broadband restriction of thermal transport, leading to ultralow thermal conductivities, while sustaining high elastic moduli and hence high ratios.
In Figure 4, we compare the of and its RP derivatives with a wide range of single crystals. The materials shown here range from soft, insulating crystals (e.g., ) to stiff, conductive crystals (e.g., diamond). The elastic moduli of and RP phases are significantly higher than those of other ultralow-thermal-conductivity single crystals (e.g., superatoms, metal halide perovskites, and layered perovskites) and are surpassed only by oxides and some semiconductors. Despite such strong bonding, the RP phases possess an ultralow thermal conductivity. As a result, the ratio of the single crystal is the highest reported to date. Though some polycrystalline materials have larger ratios,[20, 46] the record ratio shown in Figure 4 among single-crystalline solids is very significant in its own right as not only it eliminates any extrinsic spurious influences in values, but also allows us to study the fundamental mechanisms behind the “mechanically stiff phonon glass” paradigm. Additionally, the ratio of the RP phase is 3 times higher than that of . Such difference between two phases of the same material is unprecedented in literature.

Figure 4. Elastic modulus/thermal conductivity () ratio for a wide range of single-crystalline materials at room temperature. The crystals are grouped into superatoms, metal halide perovskites, semiconductors, oxides, layered materials, and and its RP derivatives. The thermal conductivity and elastic modulus data of the single crystals are provided in the Supporting Information.
As shown in Figure 4, oxides generally possess a high elastic modulus. By replacing oxygen with sulfur, we reduced thermal conductivity significantly with moderate reduction in the elastic modulus. Introduction of strongly bonded periodic interfaces can further reduce the thermal conductivity without sacrificing stiffness proportionately. The sulfides studied here sit in an ideal regime of relatively high elastic modulus and low thermal conductivity thereby opening a paradigm for finding high ratio materials. For example, RP phases of are also likely to exhibit a high ratio. Moreover, the study of chalcogenide perovskites has gained momentum only recently. We anticipate the discovery and synthesis of a broad range of layered phases such as Aurivillius phases, Dion Jacobson phases in the near future.[47, 48] These layered materials will present additional opportunities to test and further the paradigm of “mechanically stiff phonon glass”.
Conclusion
The RP phases of is found to possess ultralow thermal conductivity and ultrahigh modulus-to-thermal conductivity ratio. We find that the rock-salt layers separating the perovskite sections of the RP structure lead to highly anisotropic thermal conductivity, with the cross-plane reaching values comparable to the amorphous solid despite similar and highly strong bonding across the full unit cell. Together with simulations, our results provide evidence that the rock-salt layers in the single crystal RP phases lead to ultralow phonon velocities, ultrashort phonon mean free paths, and strong localization within rock-salt layers, leading to ultralow, glass-like thermal conductivity. Our study provides a detailed overview of the mechanisms needed to achieve ultralow thermal conductivity in a non-vdW, strongly bonded, layered material.
Acknowledgement
We thank Kevin Ye and Rafael Jaramillo from Department of Materials Science and Engineering of Massachusetts Institute of Technology for their help with the sample preparation. Md.S.B.H., E.R.H., E.A.S., J.A.T., S.M., K.A. and P.E.H. appreciate support from the Office of Naval Research Grant No. N0014-23-1-2630. D.-L.B. and S.T.P. acknowledge support by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, Materials Science and Technology Division Grant No. DE-FG02-09ER46554 and by the McMinn Endowment at Vanderbilt University. Computations were performed at the National Energy Research Scientific Computer Center (a U.S. Department of Energy Office of Science User Facility located at Lawrence Berkeley National Laboratory, operated under contract no. DE-AC02-05CH11231. B.Z., M.S., and J.R. acknowledge support from an ARO MURI program (W911NF21-1-0327), an ARO grant (W911NF-19-1-0137), and National Science Foundation (DMR-2122071). H.Z. and T.F. acknowledge support from National Science Foundation (NSF) (award number: CBET 2212830). H.Z. and T.F. used the computational resource of Bridges-2 at Pittsburgh Supercomputing Center through allocation PHY220002 from the Advanced Cyber infrastructure Coordination Ecosystem: Services & Support (ACCESS) program, which is supported by National Science Foundation grants #2138259, #2138286, #2138307, #2137603, and #2138296, National Energy Research Scientific Computing Center, a DOE Office of Science User Facility supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231 using NERSC award BES-ERCAP0022132, and Center for High Performance Computing (CHPC) at the University of Utah. A.G. and S.T. acknowledge support from the Office of Naval Research Grant No. N00014-21-1-2622 and National Science Foundation (NSF Award No. 2119365). This work was performed, in part, at the Center for Integrated Nanotechnologies, an Office of Science User Facility operated for the U.S. Department of Energy (DOE) Office of Science. Sandia National Laboratories is a multi-mission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International, Inc., for the U.S. DOE’s National Nuclear Security Administration under contract DE-NA-0003525. The views expressed in the article do not necessarily represent the views of the U.S. DOE or the United States Government.
Supporting Information
Growth details of and its RP derivatives
The and single crystals are synthesized using the flux method, details of which can be found in previous publications.[49] 1 g of powder (Alfa Aesar, 99.998%) is grounded and mixed with 0.5 g of stoichiometric mixtures of precursor powders (, , and ), and loaded into a quartz tube. For , the tube is heated to 1050 °C at a rate of 1.6 °C/min, held at 1050 °C for 100 h, cooled to 800 °C at a rate of 0.1 °C/min, and then to room temperature in an uncontrolled manner by shutting off the furnace. For , the tube is heated to 1050 °C at a rate of 0.3 °C/min, held at 1050 °C for 40 h, cooled to 400 °C at a rate of 1 °C/min, and then to room temperature in an uncontrolled manner. is formed as a secondary phase twined with some large crystals. X-ray diffraction along 001 orientation[31] is used to characterize their existence, so layers are exfoliated to access them. The obtained samples are washed repeatedly with deionized water and isopropyl alcohol to remove excess flux before drying in airflow. The crystal dimensions are on the order of . Amorphous thin films are grown from a phase pure target in a background gas mixture of hydrogen sulfide and argon (Ar-) on substrate. The dense, polycrystalline target is prepared by sintering powders by high pressure torsion method.[50]
Time-domain thermoreflectance (TDTR)
We use a two-tint time-domain thermoreflectance (TDTR) setup to measure the thermal conductivity of the crystalline and amorphous , , and specimens.[51, 52] In our TDTR setup, a Ti:sapphire oscillator (80 MHz, nm central wavelength, and nm full width at half maximum) emanates subpicosecond laser pulses that are split into a high-power pump and a low-power probe beam. The pump beam is modulated at a frequency of 8.4 MHz by an electro-optic modulator (EOM) to create oscillatory heating events at the sample surface. The probe beam is then directed through a mechanical delay stage to detect the temporal change in thermoreflectivity which is related to the surface temperature change. Using a lock-in amplifier and a balanced photodetector, the probe beam measures the temperature decay up to 5.5 ns. The TDTR data are analyzed by fitting a cylindrically symmetric, multilayer thermal model to the ratio of in-phase to out-of-phase signal () from the RF lock-in amplifier.[53, 54, 55, 56] Figure S1 shows the best-fit thermal model to the TDTR data for the , , and single crystals. For all the measurements, the pump and probe beams are coaxially focused to and 11 diameters, respectively on the sample surface. Prior to the measurements, all the samples are coated with an nm aluminum film via electron beam evaporation for optothermal transduction.[57] Our TDTR setup is calibrated with a sapphire and a high purity fused silica wafer.
Figure S1. (a) Best-fit thermal model to the TDTR data for BaZrS, BaZrS, and BaZrS crystals at room temperature. (b) Sensitivity of TDTR measurements to the ratio of in-phase to out-of-phase signal () for BaZrS at room temperature. Here, , , , and represent volumetric heat capacity, thermal conductivity, thermal boundary conductance, and thickness, respectively.
In TDTR, we fit for the thermal conductivity of the specimen and thermal boundary conductance between the aluminum transducer and specimen. Figure S1 shows the TDTR sensitivity calculations for the crystal. The sensitivity analysis is performed following the methodology published in prior literature.[58, 59, 60] As exhibited here, TDTR measurements have low sensitivity to the thermal boundary conductance and high sensitivity to the thermal conductivity. TDTR measurements also have high sensitivity to the volumetric heat capacity of the aluminum transducer and crystals.
Steady-state temperature rise during TDTR measurements
Since the RP phases possess ultralow thermal conductivity, steady-state temperature rise from pulsed laser heating can be a major concern, particularly at low temperatures. Therefore, at each temperature, we calculate the steady-state temperature rise for the precise aluminum/sample geometry from the numerical solution to the cylindrical heat equation.[61] For each sample, we optimize the laser power to obtain good signal to noise ratio and minimal steady-state temperature rise. The maximum steady-state temperature rise observed in our study is 11 K. This degree of steady-state temperature rise is not expected to change the properties of aluminum and samples significantly to cause any error.[62]
Uncertainty analysis
We calculate the uncertainty () of TDTR measurements using the following equation[63, 64, 65]
here the first term represents the standard deviation of multiple measurements across different spots. The second term represents the uncertainty in determining the absolute value of the phase from the lock-in amplifier. , , and stand for the ratio of in-phase to out-of-phase signal, uncertainty of the phase, and sensitivity of to in the thermal model, respectively. The third term represents the propagation of uncertainty from individual input parameters to the thermal model. and are uncertainty of parameter and sensitivity of to in the thermal model, respectively.
We conduct the TDTR measurements at 8.4 MHz modulation frequency. At such high modulation frequency, the signal noise and is quite small.[58] Therefore, the main sources of uncertainty in the TDTR measurements are the standard deviation and input parameters.
The four input parameters to the thermal model are: volumetric heat capacity of aluminum and samples, aluminum transducer thickness, and aluminum thermal conductivity. The volumetric heat capacities of aluminum and samples are adopted from literature[66] and density functional theory (DFT) calculations, respectively. The thickness and thermal conductivity of the aluminum transducer are measured by picosecond acoustics[67] and four-point probe using the Wiedemann–Franz law[68], respectively.
Figure S2-(Figure S2) shows the contour uncertainty analysis for various input parameters. The dashed lines represent the 95% confidence interval for the thermal conductivity corresponding to a specific value of the input parameter. This confidence interval confirms the uncertainty bounds of thermal conductivity for the parameters. Details of the contour uncertainty analysis can be found in previous publications.[69]

Figure S2. Uncertainty contour analysis of TDTR measurements showing a combination of BaZrS₃ thermal conductivity and (a) volumetric heat capacity of aluminum (2.42 MJ m K), (b) volumetric heat capacity of BaZrS₃ (1.505 MJ m K), (c) aluminum thermal conductivity (180 W m K), and (d) aluminum thickness (77 nm).
Steady-state thermoreflectance (SSTR)
Along with TDTR, we employ another optical pump-probe technique named steady-state thermoreflectance (SSTR) to measure the thermal conductivity. Unlike TDTR, SSTR technique is insensitive to the volumetric heat capacity of aluminum and samples. We employ this technique to validate the volumetric heat capacity assumed for amorphous film. In SSTR setup, a continuous wave pump laser (532 nm wavelength) is modulated at 100 Hz frequency to create steady-state temperature rise at the sample surface. The reflectivity change and temperature rise are detected by a continuous wave probe laser (786 nm wavelength). By changing the pump laser power, the heat flux deposited on the sample surface and corresponding steady-state temperature rise can be changed. By correlating the pump laser power and temperature rise via Fourier’s law, the thermal conductivity of any material can be derived. More details regarding the SSTR setup can be found in previous publications.[63, 70, 71] We use pump and probe diameters of for the measurements.
Volumetric heat capacity of , , and
The volumetric heat capacity of crystalline and are calculated using DFT. Within the framework of DFT, the projector augmented wave method (PAW)[72] as implemented in the Vienna ab initio simulation package (VASP)[73] code is employed for relaxing the cell parameter and atomic positions of all perovskite unit cells. The PBEsol[74] functional for solids is utilized for relaxing the unit cell and atomic positions. The Methfessel-Paxton[75] smearing scheme with the gamma parameter is set to 0.1 eV and an energy cut-off of 500 eV is used for the planewaves expansion. A Monkhorst-Pack[76] special grid sampling of the k-points is used for integration in the Brillouin zone for structural optimization. For resolution of the Kohn-Sham equations, the self-consistent field procedure is considered by setting energy changes for each cycle at eV as the convergence criterion between two successive iterations. The atomic forces are calculated by minimizing the total forces until the energy convergence is less than eV. The interatomic force constants (IFCs) are computed by utilizing density functional perturbation theory with a supercell including up to the seventh nearest neighbor interactions. The PHONOPY[77] code is used to obtain the harmonic displacements while considering all neighboring interactions. For the anharmonic interactions, a supercell with up to the seventh nearest neighbor interactions is employed with the THIRDOORDER.PY code.[78, 79]
Figure S3 shows the volumetric heat capacity of crystalline and . We assume that the crystalline and amorphous possess the same volumetric heat capacity. To ensure the validity of this assumption, we measure the room-temperature thermal conductivity of a nm amorphous film grown on silicon substrate via both TDTR and SSTR techniques. The TDTR technique is sensitive to heat capacity whereas the SSTR is insensitive to it. The measured thermal conductivity is in excellent agreement between the two techniques, thus proving the accuracy of the volumetric heat capacity assumed for amorphous in TDTR. We further assume that the and possess the same volumetric heat capacity. A 10% uncertainty is used with all the heat capacity values in TDTR measurements.

Figure S3. Volumetric heat capacity of crystalline BaZrS3 and Ba3Zr2S7 calculated via DFT.
Ion irradiation of the BaZrS3 and Ba4Zr3S10 crystals
The BaZrS3 and Ba4Zr3S10 single crystals are irradiated with gold (Au) ions at an energy of 2.8 MeV using a 6 MV tandem Van de Graaff accelerator. The ion implantation depths are calculated via SRIM simulations for an ion energy of 2.8 MeV. Details of the SRIM simulations for determining the stopping range of ions can be found in previous publications.[39, 80] The implantation depths of the Au ions are greater than 450 nm. This length scale is much larger than the thermal penetration depth of TDTR measurements.[81] Therefore, the thermally probed region and the measured thermal conductivity are of the defected region pre-end-of-range.[38, 82]
For the irradiation, six crystals of each material are carbon-taped onto a silicon substrate and loaded into the implant chamber. Afterwards, the chamber is pumped down to a pressure on the order of torr. Due to the small size of the crystals () relative to the incident ion beam, spatial uniformity is achieved during the implantation. Nominal fluences ranging from to are applied to each material.
TEM images of irradiated crystals are shown in Figure S4. As exhibited here, when the ion dose is , no visible damage is observed in the crystal. However, as the ion dose is increased to , amorphous pockets are introduced in the system. For both ion doses, the layerings of remain uninterrupted. This verifies the strong bonding across the rock-salt layers in the RP phases and makes them suitable for deep space applications in radiation environments.
Figure S4. TEM images of crystals corresponding to (a) and (b) ion doses.
Presence of nano-domains in BaZrS₃ crystals
One of the biggest requirements for any thermoreflectance measurement is optically smooth sample surface. To obtain such surface quality, we mechanically polish the samples. Polishing deforms the crystals and creates nano-scale domains in as exhibited in Figure S5. The presence of such nano-domains can obfuscate the origin of thermal behaviors in . Therefore, as an alternate to polishing, we use cleaving on several crystals. Although cleaving does not deform the material and create nano-domains, obtaining large quantities of optically smooth crystals can be highly challenging via cleaving. As a result, we use polished crystals for many of our measurements. The thermal conductivity shown in Figure 2(a) corresponds to a cleaved crystal, whereas the ones shown in Figure 2(c) belong to polished crystals. The cleaved and polished crystals have thermal conductivities of and , respectively, at room temperature.
Figure S5. TEM images of domains in (a) BaZrS and (b) BaZrS crystals. (c) Thermal conductivity of cleaved and polished BaZrS crystals as a function of temperature.
Polishing also deforms the RP crystals and creates domains. However, due to the high stiffness of the RP phases, the domain sizes are greater than 500 nm as shown in Figure S5. At such length scales, the domains are not expected to impact the thermal conductivity of the RP phases.[34] As a result, we use mechanically polished RP crystals for all of our measurements.
The thermal conductivity of the cleaved and polished BaZrS crystals are shown in Figure S5 as a function of temperature. Due to phonon scattering at the domain boundary, the thermal conductivity of the polished crystal is lower than the cleaved one throughout the temperature range.
Machine learning molecular dynamics (MLMD) simulation details
To construct the training database of machine learning potential, first principles calculations of both BaZrS and BaZrS are performed by using VASP[83] with the PAW[84] method based on the DFT. Local density approximation (LDA)[85] is chosen as the exchange-correlation functional. The plane-wave energy cutoff is selected as 500 eV. The primitive cell is first relaxed with the energy convergence threshold of eV and the force convergence threshold of eV/Å between atoms. For BaZrS, the calculational primitive cell contains 20 atoms following the Pnmb symmetry, where the lattice constants are 7.03 Å, 9.86 Å, and 6.89 Å along , , and directions, respectively. The k-mesh is set to . For BaZrS, the calculational primitive cell contains 24 atoms following the I4mmm symmetry, where the lattice constants are 4.92 Å, 4.92 Å, and 25.24 Å along , , and directions, respectively. The k-mesh is set to . After the relaxation, the VASP + PHONOPY combination is utilized to calculate the phonon properties of BaZrS and BaZrS. Ab initio molecular dynamics (AIMD) simulations are performed to sample the potential energy surface. For BaZrS and BaZrS, supercells are adopted containing 160 and 192 atoms, respectively, in the simulation domains. The time step of AIMD is set to 5 fs and a total step of 8000 is run for each individual AIMD. Independent AIMDs at 100 K, 200 K, 300 K, 400 K, and 500 K with Canonical ensemble (NVT) are performed. During the AIMD process, energies, forces, and stresses are recorded together with corresponding atomic configurations to form the training database for machine learning potential.
The potential energy surface is approximated by moment tensor potential (MTP) trained by Machine-Learning Interatomic Potentials (MLIP) package.[86] MTP represents the energy of an atomic configuration cfg () as a sum of contributions of local atomic environments of each atom, denoted by , shown as
MTP with different levels are provided and a higher level means more fitting parameters, potential higher accuracy and possible issues of overfitting. Given a training database containing K configurations (k=1, 2, . . . , K) with their corresponding DFT calculated energies, forces, and stresses, the training process of MTP is minimizing the following expression:
where is the number of atoms in the kth configuration, , , and are non-negative weights. More details about MTP and the training process can be found in Novikov et al.[86] In this work, MTP with the level of 22 is adopted to fit the potential energy surface and the training step is set to be 20,000.
Once the MTP is developed, equilibrium molecular dynamics (EMD) is performed via LAMMPS[87] package to get the lattice thermal conductivity, . A supercell of conventional cell is adopted for BaZrS and for BaZrS, respectively, which should be large enough to avoid size effect.[88] The time step of EMD is 5 fs, and periodic boundary conditions are implemented in all three directions. The simulation is carried out first by 2 ns NVT, followed by a 2 ns NVE to fully relax the lattice, and then another NVE of 4 ns, during which the heat current is recorded. Based on the Kubo-formula, [89, 90] along a certain direction is proportional to the integral of autocorrelation of heat current:
where is Boltzmann constant, is the volume of the simulation domain, is temperature, and is time. In LAMMPS, the heat current vector is defined as
where is the total energy of the th atom, is the velocity vector of the th atom, is the force interaction between the th and th atom, and represents the position vector between the th and th atom. The final is averaged over 8 independent EMD simulations at each temperature with different initial velocities. Quantum correction is taken into account due to the classical specific heat used in LAMMPS is much higher than the real specific heat at low temperature. [91]
Spectral energy density (SED) calculations
To gain insights into the intrinsic mechanisms that dictate the thermal transport in BaZrS and its RP phase BaZrS structures, we calculate the phonon mode specific properties by performing spectral energy density (SED) calculations in MD simulations, which is given as [40, 92]
where is the total simulation time, is the cartesian direction, is a unit cell, is the number of unit cells in the crystal, is the atom label in a given unit cell, is the atomic number in the unit cell, is the mass of atom in the unit cell, denotes the velocity along the direction at time , and is the equilibrium position of each unit cell.
To ensure a high resolution in our SED calculations, we create the computational domain for both BaZrS and its RP phase BaZrS structures with (representing the total number of unit cells) and perform SED calculations for 100 q-points. Initially, we equilibrate our supercell structure using the Nosé-Hoover thermostat and barostat [93] for 1 ns with a time step of 0.5 fs where the number of particles, pressure, and temperature of the system are held constant at ambient pressure. After the NPT integration, we further equilibrate our structures in the NVT ensemble, keeping the volume and temperature constant for an additional 1 ns. Finally, for the data collection for our SED calculations, we record the velocities and positions of each atom in the microcanonical ensemble (or NVE ensemble) for a total of 1.5 ns.
Elastic modulus measurements and derivation of sound speed
We measure the elastic modulus of the BaZrS crystals using a nanoindenter (MTS XP) based on the standard continuous stiffness measurement (CSM) method.[94] Before the measurements, the nanoindenter is calibrated with a silica standard. Details of the calibration procedure can be found in previous publications.[13, 95] The elastic modulus of BaZrS is determined to be 84.2 5.4 GPa. This value is in agreement with literature.[96]
The elastic modulus of BaZrS is measured by a Bruker Hysitron TI 950 Triboindenter. The indenter is calibrated with a single crystal quartz sample with known modulus and hardness (69.6 GPa and 9.25 GPa, respectively). The measured elastic modulus of BaZrS is 69 4.5 GPa. The indentation curves exhibit pop-in features, which are sudden displacement excursions; a common feature observed in single crystals.[97]
Assuming isotropic elastic properties, the sound speed components of BaZrS can be derived using the following equations[98]
where , , , , , and represent longitudinal sound speed, transverse sound speed, sound velocity, elastic modulus, density, and Poisson’s ratio, respectively. We take the density and Poisson’s ratio to be 4.39 g cm and 0.28 0.05, respectively.[31] The derived sound speed components of BaZrS are tabulated Table S1. The error bars incorporate the uncertainty of the elastic modulus and Poisson’s ratio.
| Sound speed components | Velocity (m s) |
| 4952 410 | |
| 2737 112 | |
| 3475 143 |
Table S1. Longitudinal sound speed, transverse sound speed, and sound velocity of the BaZrS crystals.
To further verify the longitudinal sound speed, we use picosecond acoustics on a 416 nm crystalline BaZrS film. Details of our picosecond acoustics metrology and data reduction can be found in previous publications.[98, 99] Picosecond acoustics reveal the longitudinal sound speed of BaZrS to be 4818 512 m s, in excellent agreement with the elastic modulus derived value. The error bar of the picosecond acoustic measurement incorporates the standard deviation and uncertainty of the film thickness (44 nm).
Elastic modulus calculations
To determine the stress-strain relationship for our and its RP phase structures, we use MLMD simulations to calculate the elastic moduli. We deform the simulation cell in the uniaxial direction using a strain rate of . Concurrently, we maintain zero pressure at the boundaries in the remaining two lateral directions. This process is performed under the NPT ensemble, where the system’s number of particles, pressure, and temperature are kept constant. The strain is computed at intervals of 0.5 fs using the formula , where represents the current length and denotes the initial length of the domain in the direction of the applied strain. Simultaneously, the stress of the entire structure is recorded every 50 fs to enable the generation of a stress-strain relationship for the structures.
For , we calculate a Young’s modulus value of 85 GPa in the cross-plane direction. For , we calculate Young’s modulus values of 64.5 GPa and 64.3 GPa in the in-plane directions, and 81.3 GPa in the cross-plane direction.
Minimum limit and diffuson limit
Cahill et al.’s minimum limit model states that the lower bound of thermal conductivity can be achieved when the phonon mean free path reduces to half of its wavelength.[7] This model assumes that the heat is transported through a solid via random walks between localized oscillators. Cahill et al.’s minimum limit model is also known as the amorphous limit model and can be expressed as the following[4]
where , , , , , and represent thermal conductivity prediction, Boltzmann constant, number density, Planck constant, Debye temperature, and temperature, respectively.
Cahill et al.’s model has successfully predicted the thermal conductivity of many disordered solids and amorphous materials. However, several recent studies have showed that the thermal conductivity of a few materials can fall below this limit.[10, 11, 12, 23] Therefore, to better estimate the theoretical lower bound of thermal conductivity, Agne et al.[6] predicted a model based on the diffuson-mediated thermal transport. Agne et al.’s diffuson model assumes that the heat is transported through a material via diffusons (i.e., non-propagating, delocalized vibrational modes). According to this diffuson model, the thermal conductivity () can be expressed as
The minimum limit and diffuson limit shown in Figure 2(a) correspond to BaZrS and are calculated using the properties mentioned in supporting information section S11.
Anisotropic minimum thermal conductivity limit
As RP structured crystals has certain anisotropy, we use the anisotropic thermal conductivity limit proposed by Chen and Dames[100] for . According to this model, the minimum thermal conductivity along the cross-plane direction () of a layered material can be expressed as:
Where , , , and represent Boltzmann constant, Planck constant, -axis sound velocity, -plane sound velocity, Debye temperature, and temperature, respectively.
For , we derive and ( represents longitudinal polarizations) from the elastic constants obtained via MD. The derived values are 4288 and 3807 m s, respectively. Using these values, the anisotropic minimum thermal conductivity model of Figure S6 is obtained. As elastic constants have an uncertainty of and we assume [23], the anisotropic model of Figure S6 has an uncertainty of .
Figure S6. Anisotropic minimum thermal conductivity model[100] and experimental measurements of .
The anisotropic version of the minimum thermal conductivity model is still higher than the experimental values of . This can be attributed to the comparable bond strength across the rock-salt layers and layers. As a result, the sound velocities are also comparable between -axis and -plane. In other anisotropic materials, such as , the bonding and sound velocities are very different between -axis and -plane.[100]
Due to comparable sound velocities between -axis and -plane, the predictions of anisotropic thermal conductivity model are close to the predictions of isotropic thermal conductivity model.
Machine learning potentials (MLPs)
In order to perform MLMD simulations over a wide temperature range, we develope two machine learning potentials (MLPs) for each of and , respectively. One potential is for relatively low-temperature (LT) MLMDs and the other is for relatively high-temperature (HT) MLMDs. In particular, the LT-MLP of is used to run 100 K, 150 K, 200 K, and 250 K MLMDs, while the HT-MLP of is used to run 300 K, 350 K, and 400 K MLMDs. The LT-MLP of is for 100 K and 150 K MLMDs, while HT-MLP of is for 200 K, 250 K, 300 K, 350 K, and 400 K MLMDs.
| RMS absolute difference | MLP () | MLP () | ||
| Train | Test | Train | Test | |
| Energy per atom (meV) | 0.348 | 0.354 | 0.293 | 0.311 |
| Atomic force (eV/Å) | 0.054 | 0.054 | 0.047 | 0.047 |
Table S2. The root-mean-square (RMS) errors of energies per atom and atomic forces for MLPs across both the training and testing datasets.

Figure S7. Discrepancies between the energy/atom and atomic forces obtained using the machine-learned force field (MLIP) and those obtained using DFT over the testing datasets for BaZrS and BaZrS.
Figure S8. Calculated relative energies as a function of bi-axial strain between DFT and MLIP for BaZrS and BaZrS. The results between DFT and MLIP show very good agreement.
Literature data of elastic modulus/thermal conductivity () ratio
The elastic modulus/thermal conductivity () ratios of a wide range of single crystalline materials are listed in Table S3.
| Group | Material | (GPa) | (GPa m K W) |
| Superatom | CoS | 4 | 18.18[27] |
| Superatom | CoSe | 2.3 | 12.78[27] |
| Superatom | CoTe | 0.62 | 4.77[27] |
| Superatom | [CoSe][C] | 8.1 | 32.4[27] |
| Superatom | [CoTe][C] | 1.5 | 9.38[27] |
| Semiconductor | Diamond (I/ IIa/ IIb) | 1144.81 | 1.27/0.49/0.84[101, 102, 103, 104] |
| Semiconductor | cBN | 909 | 1.03[105, 106, 107] |
| Semiconductor | AlN | 374 | 1.17[108, 109] |
| Semiconductor | GaN | 295 | 1.17[110, 111] |
| Semiconductor | Si | 165.82 | 1.11[101, 102, 103, 104] |
| Semiconductor | Ge | 135.4 | 2.25[101, 102, 103, 104] |
| Semiconductor | AgSbTe | 49.49 | 72.78[112, 113] |
| Semiconductor | PbTe | 67.23 | 28.01[112, 114] |
| Semiconductor | InAs | 79.7 | 2.95[115, 116] |
| Semiconductor | PbSe | 65.2 | 24.7[117] |
| Semiconductor | PbS | 70.2 | 25[117] |
| Metal halide perovskite | MAPbCl (cubic) | 23 | 31.5[15] |
| Metal halide perovskite | MAPbBr (cubic) | 17.8 | 34.9[15] |
| Metal halide perovskite | MAPbI (tetragonal) | 12 | 35.3[15] |
| Metal halide perovskite | CsPbBr (Orthorhombic) | 13.5 | 29.35[15] |
| Metal halide perovskite | FAPbBr (Orthorhombic) | 10.2 | 20.82[15] |
Table S3. The elastic modulus/thermal conductivity () ratios of different groups of single crystalline materials at room temperature.
| Group | Material | E (GPa) | (GPa m K W) |
| Oxide | 310 | 5.96 [13, 118] | |
| Oxide | 345 | 10.15 [13] | |
| Oxide | 260.85 | 23.71 [119, 120, 121] | |
| Oxide | 181 | 42.1 [122] | |
| Oxide | 175 | 92.11 [122] | |
| Oxide | -stabilized (YSZ) | 210 | 98.6 [122] |
| Oxide | 175 | 5.15 [123, 124] | |
| Oxide | J14 ESO | 152 | 51.53 [13] |
| Oxide | J34 ESO | 180.8 | 125.56 [13] |
| Oxide | J36 ESO | 229.9 | 143.69 [13] |
| Layered materials | 17.4 | 87 [17] | |
| Layered materials | * | 20.6 | 55.7 [16] |
| Layered materials | 167.3 | 107.2 [10, 125] | |
| 84.2 | 54.3 | ||
| 69 | 153.3 |
Table S3 (PDF p. 29). The elastic modulus/thermal conductivity () ratios of different groups of single crystalline materials at room temperature (cont.).
Temperature-dependent thermal conductivity of perovskites

Figure S9. Temperature-dependent thermal conductivity of BaZrS, SrTiO[126], NaNbO[127], CsPbI[128], CsSnI[129], BaTiS[62], and CsBiNbO[130]. The solid and hollow symbols represent measurements taken in this study and literature values, respectively.
Phonon mean free path of
Heat-current autocorrelation functions
Figure S11. The calculated heat-current autocorrelation functions for at various considered temperatures. Recall that we performed eight independent MLMDs at each temperature. For simplicity, each panel in the figure represents one MLMD simulation case out of these eight independent cases. Blue, orange, and green colors represent x, y, and z-direction results, respectively.
Figure S12. The calculated heat-current autocorrelation functions for Ba₃Zr₂S₇ at various considered temperatures. Recall that we performed eight independent MLMDs at each temperature. For simplicity, each panel in the figure represents one MLMD simulation case out of these eight independent cases. Blue, orange, and green colors represent x, y, and z-direction results, respectively.
Calculated thermal conductivities vs correlation time
Figure S13. The calculated thermal conductivities of BaZrS3 as a function of correlation time at various considered temperatures. Only the z-components of the thermal conductivity are plotted. Within each panel, the colorful curves represent the calculated results for all eight cases at each temperature. The black curves depict the average of the eight colorful curves, while the red curves are obtained by averaging every 200 data points from the black curve.
Figure S14. The calculated thermal conductivities of as a function of correlation time at various considered temperatures. Only the -components of the thermal conductivity are plotted. Within each panel, the colorful curves represent the calculated results for all eight cases at each temperature. The black curves depict the average of the eight colorful curves, while the red curves are obtained by averaging every 200 data points from the black curve.
DFT-calculated phonon dispersion
Figure S15. DFT-calculated phonon dispersion of BaZrS and BaZrS. Note that they are in reasonable agreement with the SED results.
Eigen-displacements of phonon modes

Figure S16. Eigen-displacements of three prototype phonon modes for Ba₃Zr₂S₇, illustrating that the inter- and intra-cell rock-salt blocks contribute to phonon localizations. The gray bars highlight the rock-salt-like interfaces.
References
- [1]Kittel, C. Interpretation of the thermal conductivity of glasses. Physical Review 1949, 75, 972.DOI
- [2]Ioffe, A.; Regel, A. Progress in semiconductors; 1960; pp 237–291.
- [3]Slack, G. A. The thermal conductivity of nonmetallic crystals. Solid state physics 1979, 34, 1–71.DOI
- [4]Cahill, D. G.; Watson, S. K.; Pohl, R. O. Lower limit to the thermal conductivity of disordered crystals. Physical Review B 1992, 46, 6131.pubmed.ncbi.nlm.nih.gov/10002297
- [5]Clarke, D. R. Materials selection guidelines for low thermal conductivity thermal barrier coatings. Surface and Coatings Technology 2003, 163, 67–74.DOI
- [6]Agne, M. T.; Hanus, R.; Snyder, G. J. Minimum thermal conductivity in the context of diffusion-mediated thermal transport. Energy & Environmental Science 2018, 11, 609–616.
- [7]Beekman, M.; Cahill, D. G. Inorganic crystals with glass-like and ultralow thermal conductivities. Crystal Research and Technology 2017, 52, 1700114.DOI
- [8]Mukhopadhyay, S.; Parker, D. S.; Sales, B. C.; Puretzky, A. A.; McGuire, M. A.; Lindsay, L. Two-channel model for ultralow thermal conductivity of crystalline Tl3VSe4. Science 2018, 360, 1455–1458.DOI
- [9]Kim, S. E.; Cahill, D. G. Pushing low thermal conductivity to the limit. Science 2021, 373, 963–964.DOI
- [10]Chiritescu, C.; Cahill, D. G.; Nguyen, N.; Johnson, D.; Bodapati, A.; Keblinski, P.; Zschack, P. Ultralow thermal conductivity in disordered, layered WSe2 crystals. Science 2007, 315, 351–353.pubmed.ncbi.nlm.nih.gov/17170252
- [11]Duda, J. C.; Hopkins, P. E.; Shen, Y.; Gupta, M. C. Exceptionally low thermal conductivities of films of the fullerene derivative PCBM. Physical review letters 2013, 110, 015902.pubmed.ncbi.nlm.nih.gov/23383810
- [12]Wang, X.; Liman, C. D.; Treat, N. D.; Chabinyc, M. L.; Cahill, D. G. Ultralow thermal conductivity of fullerene derivatives. Physical Review B 2013, 88, 075310.DOI
- [13]Braun, J. L.; Rost, C. M.; Lim, M.; Giri, A.; Olson, D. H.; Kotsonis, G. N.; Stan, G.; Brenner, D. W.; Maria, J.-P.; Hopkins, P. E. Charge-induced disorder controls the thermal conductivity of entropy-stabilized oxides. Advanced materials 2018, 30, 1805004.DOI
- [14]Qian, X.; Zhou, J.; Chen, G. Phonon-engineered extreme thermal conductivity materials. Nature Materials 2021, 20, 1188–1202.DOI
- [15]Elbaz, G. A.; Ong, W.-L.; Doud, E. A.; Kim, P.; Paley, D. W.; Roy, X.; Malen, J. A. Phonon speed, not scattering, differentiates thermal transport in lead halide perovskites. Nano letters 2017, 17, 5734–5739.DOI
- [16]Acharyya, P.; Ghosh, T.; Pal, K.; Kundu, K.; Singh Rana, K.; Pandey, J.; Soni, A.; Waghmare, U. V.; Biswas, K. Intrinsically ultralow thermal conductivity in Ruddlesden–Popper 2D perovskite Cs2PbI2Cl2: localized anharmonic vibrations and dynamic octahedral distortions. Journal of the American Chemical Society 2020, 142, 15995–15603.DOI
- [17]Acharyya, P.; Ghosh, T.; Pal, K.; Rana, K. S.; Dutta, M.; Swain, D.; Etter, M.; Soni, A.; Waghmare, U. V.; Biswas, K. Glassy thermal conductivity in Cs3Bi2I6Cl3 single crystal. Nature communications 2022, 13, 5053.DOI
- [18]Acharyya, P.; Pal, K.; Ahad, A.; Sarkar, D.; Rana, K. S.; Dutta, M.; Soni, A.; Waghmare, U. V.; Biswas, K. Extended Antibonding States and Phonon Localization Induce Ultralow Thermal Conductivity in Low Dimensional Metal Halide. Advanced Functional Materials 2023, 2304607.DOI
- [19]Christodoulides, A. D.; Guo, P.; Dai, L.; Hoffman, J. M.; Li, X.; Zuo, X.; Rosenmann, D.; Brumberg, A.; Kanatzidis, M. G.; Schaller, R. D.; others Signatures of coherent phonon transport in ultralow thermal conductivity two-dimensional Ruddlesden–Popper phase perovskites. ACS nano 2021, 15, 4165–4172.DOI
- [20]Yang, J.; Qian, X.; Pan, W.; Yang, R.; Li, Z.; Han, Y.; Zhao, M.; Huang, M.; Wan, C. Diffused lattice vibration and ultralow thermal conductivity in the binary Ln–Nb–O oxide system. Advanced materials 2019, 31, 1808222.DOI
- [21]Snyder, G. J.; Toberer, E. S. Materials for sustainable energy: a collection of peer-reviewed research and review articles from Nature Publishing Group; World Scientific, 2011; pp 101–110.
- [22]Ravichandran, J.; Yadav, A. K.; Cheaito, R.; Rossen, P. B.; Soukiassian, A.; Suresha, S.; Duda, J. C.; Foley, B. M.; Lee, C.-H.; Zhu, Y.; others Crossover from incoherent to coherent phonon scattering in epitaxial oxide superlattices. Nature materials 2014, 13, 168–172.DOI
- [23]Aryana, K.; Stewart, D. A.; Gaskins, J. T.; Nag, J.; Read, J. C.; Olson, D. H.; Grobis, M. K.; Hopkins, P. E. Tuning network topology and vibrational mode localization to achieve ultralow thermal conductivity in amorphous chalcogenides. Nature communications 2021, 12, 1–9.DOI
- [24]Gibson, Q. D.; Zhao, T.; Daniels, L. M.; Walker, H. C.; Daou, R.; Hébert, S.; Zanella, M.; Dyer, M. S.; Claridge, J. B.; Slater, B.; others Low thermal conductivity in a modular inorganic material with bonding anisotropy and mismatch. Science 2021, 373, 1017–1022.DOI
- [25]Chiritescu, C.; Mortensen, C.; Cahill, D. G.; Johnson, D.; Zschack, P. Lower limit to the lattice thermal conductivity of nanostructured Bi2Te3-based materials. Journal of Applied Physics 2009, 106.DOI
- [26]Manley, M. E.; Hellman, O.; Shulumba, N.; May, A. F.; Stonaha, P. J.; Lynn, J. W.; Garlea, V. O.; Alatas, A.; Hermann, R. P.; Budai, J. D.; others Intrinsic anharmonic localization in thermoelectric PbSe. Nature communications 2019, 10, 1928.DOI
- [27]Ong, W.-L.; O’Brien, E. S.; Dougherty, P. S.; Paley, D. W.; Fred Higgs III, C.; McGaughey, A. J.; Malen, J. A.; Roy, X. Orientational order controls crystalline and amorphous thermal transport in superatomic crystals. Nature materials 2017, 16, 83–88.DOI
- [28]Zhao, L.-D.; Lo, S.-H.; Zhang, Y.; Sun, H.; Tan, G.; Uher, C.; Wolverton, C.; Dravid, V. P.; Kanatzidis, M. G. Ultralow thermal conductivity and high thermoelectric figure of merit in SnSe crystals. nature 2014, 508, 373–377.DOI
- [29]Su, L.; Wang, D.; Wang, S.; Qin, B.; Wang, Y.; Qin, Y.; Jin, Y.; Chang, C.; Zhao, L.-D. High thermoelectric performance realized through manipulating layered phonon-electron decoupling. Science 2022, 375, 1385–1389.DOI
- [30]Niu, S.; Sarkar, D.; Williams, K.; Zhou, Y.; Li, Y.; Bianco, E.; Huyan, H.; Cronin, S. B.; McConney, M. E.; Haiges, R.; others Optimal bandgap in a 2D Ruddlesden–Popper perovskite chalcogenide for single-junction solar cells. Chemistry of Materials 2018, 30, 4882–4886.DOI
- [31]Niu, S.; Zhao, B.; Ye, K.; Bianco, E.; Zhou, J.; McConney, M. E.; Settens, C.; Haiges, R.; Jaramillo, R.; Ravichandran, J. Crystal growth and structural analysis of perovskite chalcogenide BaZrS3 and Ruddlesden–Popper phase Ba3Zr2S7. Journal of Materials Research 2019, 34, 3819–3826.DOI
- [33]Toberer, E. S.; Baranowski, L. L.; Dames, C. Advances in thermal conductivity. Annual Review of Materials Research 2012, 42, 179–209.DOI
- [34]Osei-Agyemang, E.; Balasubramanian, G. Understanding the extremely poor Lattice thermal transport in chalcogenide perovskite BaZrS3. ACS Applied Energy Materials 2019, 3, 1139–1144.DOI
- [35]Tiwari, J.; Feng, T. Accurate prediction of thermal conductivity of Al 2 O 3 at ultrahigh temperatures. Physical Review B 2024, 109, 075201.
- [36]Kulnitskiy, B.; Blank, V.; Kuznetsov, M.; Nosukhin, S.; Terentiev, S. The effect of boron on the structure and lattice parameters of diamond single crystals. Diamond and Related Materials 2024, 144, 111015.DOI
- [37]Kim, S. E.; Mujid, F.; Rai, A.; Eriksson, F.; Suh, J.; Poddar, P.; Ray, A.; Park, C.; Fransson, E.; Zhong, Y.; others Extremely anisotropic van der Waals thermal conductors. Nature 2021, 597, 660–665.DOI
- [38]Scott, E. A.; Hattar, K.; Braun, J. L.; Rost, C. M.; Gaskins, J. T.; Bai, T.; Wang, Y.; Ganski, C.; Goorsky, M.; Hopkins, P. E. Orders of magnitude reduction in the thermal conductivity of polycrystalline diamond through carbon, nitrogen, and oxygen ion implantation. Carbon 2020, 157, 97–105.DOI
- [39]Scott, E. A.; Hattar, K.; Lang, E. J.; Aryana, K.; Gaskins, J. T.; Hopkins, P. E. Reductions in the thermal conductivity of irradiated silicon governed by displacement damage. Physical Review B 2021, 104, 134306.DOI
- [40]Thomas, J. A.; Turney, J. E.; Iutzi, R. M.; Amon, C. H.; McGaughey, A. J. Predicting phonon dispersion relations and lifetimes from the spectral energy density. Physical Review B 2010, 81, 081411.DOI
- [41]de Koker, N. Thermal conductivity of MgO periclase from equilibrium first principles molecular dynamics. Physical review letters 2009, 103, 125902.pubmed.ncbi.nlm.nih.gov/19792447
- [42]Simoncelli, M.; Marzari, N.; Mauri, F. Unified theory of thermal transport in crystals and glasses. Nature Physics 2019, 15, 809–813.DOI
- [43]Simoncelli, M.; Mauri, F.; Marzari, N. Thermal conductivity of glasses: first-principles theory and applications. npj Computational Materials 2023, 9, 106.DOI
- [44]Seyf, H. R.; Henry, A. A method for distinguishing between propagons, diffusons, and locons. Journal of Applied Physics 2016, 120, 025101.DOI
- [45]Lv, W.; Henry, A. Non-negligible contributions to thermal conductivity from localized modes in amorphous silicon dioxide. Scientific reports 2016, 6, 35720.DOI
- [46]Guo, Y.; Feng, S.; Yang, Y.; Zheng, R.; Zhang, Y.; Fu, J.; Wang, H.; Li, J. High-entropy titanate pyro chlore as newly low-thermal conductivity ceramics. Journal of the European Ceramic Society 2022, 42, 6614–6623.DOI
- [47]Kendall, K. R.; Navas, C.; Thomas, J. K.; zur Loye, H.-C. Recent developments in oxide ion conductors: Aurivillius phases. Chemistry of materials 1996, 8, 642–649.DOI
- [48]Benedek, N. A. Origin of Ferroelectricity in a Family of Polar Oxides: The Dion- Jacobson Phases. Inorganic Chemistry 2014, 53, 3769–3777.DOI
- [49]Ye, K.; Koocher, N. Z.; Filippone, S.; Niu, S.; Zhao, B.; Yeung, M.; Bone, S.; Robinson, A. J.; Vora, P.; Schleife, A.; others Low-energy electronic structure of perovskite and Ruddlesden- Popper semiconductors in the Ba-Zr-S system probed by bond-selective polarized x-ray absorption spectroscopy, infrared reflectivity, and Raman scattering. Physical Review B 2022, 105, 195203.DOI
- [50]Surendran, M.; Chen, H.; Zhao, B.; Thind, A. S.; Singh, S.; Orvis, T.; Zhao, H.; Han, J.-K.; Htoon, H.; Kawasaki, M.; others Epitaxial thin films of a chalcogenide perovskite. Chemistry of Materials 2021, 33, 7457–7464.DOI
- [51]Olson, D. H.; Avincola, V. A.; Parker, C. G.; Braun, J. L.; Gaskins, J. T.; Tomko, J. A.; Opila, E. J.; Hopkins, P. E. Anisotropic thermal conductivity tensor of β-Y2Si2O7 for orientational control of heat flow on micrometer scales. Acta Materialia 2020, 189, 299–305.DOI
- [52]Olson, D. H.; Gaskins, J. T.; Tomko, J. A.; Opila, E. J.; Golden, R. A.; Harrington, G. J.; Chamberlain, A. L.; Hopkins, P. E. Local thermal conductivity measurements to determine the fraction of α-cristobalite in thermally grown oxides for aerospace applications. Scripta Materialia 2020, 177, 214–217.DOI
- [53]Cahill, D. G.; Goodson, K.; Majumdar, A. Thermometry and thermal transport in micro/nanoscale solid-state devices and structures. J. Heat Transfer 2002, 124, 223–241.DOI
- [54]Cahill, D. G. Analysis of heat flow in layered structures for time-domain thermoreflectance. Review of scientific instruments 2004, 75, 5119–5122.DOI
- [55]Schmidt, A. J.; Chen, X.; Chen, G. Pulse accumulation, radial heat conduction, and anisotropic thermal conductivity in pump-probe transient thermoreflectance. Review of Scientific Instruments 2008, 79, 114902.DOI
- [56]Hopkins, P. E.; Serrano, J. R.; Phinney, L. M.; Kearney, S. P.; Grasser, T. W.; Harris, C. T. Criteria for cross-plane dominated thermal transport in multilayer thin film systems during modulated laser heating. Journal of Heat Transfer 2010, 132.DOI
- [57]DeCoster, M. E.; Chen, X.; Zhang, K.; Rost, C. M.; Hoglund, E. R.; Howe, J. M.; Beechem, T. E.; Baumgart, H.; Hopkins, P. E. Thermal conductivity and phonon scattering processes of ALD grown PbTe–PbSe thermoelectric thin films. Advanced Functional Materials 2019, 29, 1904073.DOI
- [58]Koh, Y. K.; Singer, S. L.; Kim, W.; Zide, J. M.; Lu, H.; Cahill, D. G.; Majumdar, A.; Gossard, A. C. Comparison of the 3 ω method and time-domain thermoreflectance for measurements of the cross-plane thermal conductivity of epitaxial semiconductors. Journal of Applied Physics 2009, 105, 054303.DOI
- [59]Jiang, P.; Huang, B.; Koh, Y. K. Accurate measurements of cross-plane thermal conductivity of thin films by dual-frequency time-domain thermoreflectance (TDTR). Review of Scientific Instruments 2016, 87, 075101.DOI
- [60]Giri, A.; Niemelä, J.-P.; Tynell, T.; Gaskins, J. T.; Donovan, B. F.; Karppinen, M.; Hopkins, P. E. Heat-transport mechanisms in molecular building blocks of inorganic/organic hybrid superlattices. Physical Review B 2016, 93, 115310.DOI
- [61]Braun, J. L.; Szwejkowski, C. J.; Giri, A.; Hopkins, P. E. On the steady-state temperature rise during laser heating of multilayer thin films in optical pump–probe techniques. Journal of Heat Transfer 2018, 140.DOI
- [62]Sun, B.; Niu, S.; Hermann, R. P.; Moon, J.; Shulumba, N.; Page, K.; Zhao, B.; Thind, A. S.; Mahalingam, K.; Milam-Guerrero, J.; others High frequency atomic tunneling yields ultralow and glass-like thermal conductivity in chalcogenide single crystals. Nature Communications 2020, 11, 6039.DOI
- [63]Braun, J. L.; Olson, D. H.; Gaskins, J. T.; Hopkins, P. E. A steady-state thermoreflectance method to measure thermal conductivity. Review of Scientific Instruments 2019, 90, 024905.DOI
- [64]Wei, C.; Zheng, X.; Cahill, D. G.; Zhao, J.-C. Invited Article: Micron resolution spatially resolved measurement of heat capacity using dual-frequency time-domain thermoreflectance. Review of Scientific Instruments 2013, 84.DOI
- [65]Yang, J.; Ziade, E.; Schmidt, A. J. Uncertainty analysis of thermoreflectance measurements. Review of scientific instruments 2016, 87.DOI
- [66]Touloukian, Y.; Boyco, E. Specific Heat: Metallic Solids volume 4 of Thermophysical Properties of Matter Eds YS Touloukian. CY Ho (New York: IFI/Plenum) 1970,
- [67]O’Hara, K.; Hu, X.; Cahill, D. G. Characterization of nanostructured metal films by picosecond acoustics and interferometry. Journal of Applied Physics 2001, 90, 4852–4858.DOI
- [68]Park, J.-H.; Choi, Y.-S.; Shim, H. C.; Ahn, J.-P.; Lee, J.-C. Direct Measurement of Electric Resistivity of Solid Electrolyte Interface Using 4-Point-Probe Technique. Korean Journal of Metals and Materials 2018, 57, 67–70.DOI
- [69]Koh, Y. R.; Shi, J.; Wang, B.; Hu, R.; Ahmad, H.; Kerdsongpanya, S.; Milosevic, E.; Doolittle, W. A.; Gall, D.; Tian, Z.; others Thermal boundary conductance across epitaxial metal/sapphire interfaces. Physical Review B 2020, 102, 205304.DOI
- [70]Hoque, M. S. B.; Koh, Y. R.; Braun, J. L.; Mamun, A.; Liu, Z.; Huynh, K.; Liao, M. E.; Hussain, K.; Cheng, Z.; Hoglund, E. R.; others High in-plane thermal conductivity of aluminum nitride thin films. ACS nano 2021, 15, 9588–9599.DOI
- [71]Hoque, M. S. B.; Koh, Y. R.; Aryana, K.; Hoglund, E. R.; Braun, J. L.; Olson, D. H.; Gaskins, J. T.; Ahmad, H.; Elahi, M. M. M.; Hite, J. K.; others Thermal conductivity measurements of subsurface buried substrates by steady-state thermoreflectance. Review of Scientific Instruments 2021, 92, 064906.DOI
- [72]Kresse, G.; Joubert, D. From ultrasoft pseudopotentials to the projector augmented-wave method. Physical review b 1999, 59, 1758.DOI
- [73]Hafner, J. Ab-initio simulations of materials using VASP: Density-functional theory and beyond. Journal of computational chemistry 2008, 29, 2044–2078.DOI
- [74]Perdew, J. P.; Ruzsinszky, A.; Csonka, G. I.; Vydrov, O. A.; Scuseria, G. E.; Constantin, L. A.; Zhou, X.; Burke, K. Restoring the density-gradient expansion for exchange in solids and surfaces. Physical review letters 2008, 100, 136406.DOI
- [75]Methfessel, M.; Paxton, A. High-precision sampling for Brillouin-zone integration in metals. Physical Review B 1989, 40, 3616.pubmed.ncbi.nlm.nih.gov/9992329
- [76]Monkhorst, H. J.; Pack, J. D. Special points for Brillouin-zone integrations. Physical review B 1976, 13, 5188.
- [77]Togo, A.; Tanaka, I. First principles phonon calculations in materials science. Scripta Materialia 2015, 108, 1–5.arxiv.org/abs/1506.08498
- [78]Carrete, J.; Vermeersch, B.; Katre, A.; van Roekeghem, A.; Wang, T.; Madsen, G. K.; Mingo, N. almaBTE: A solver of the space–time dependent Boltzmann transport equation for phonons in structured materials. Computer Physics Communications 2017, 220, 351–362.DOI
- [79]Li, W.; Mingo, N.; Lindsay, L.; Broido, D. A.; Stewart, D. A.; Katcho, N. A. Thermal conductivity of diamond nanowires from first principles. Physical Review B 2012, 85, 195436.DOI
- [80]Scott, E. A.; King, S. W.; Jarenwattananon, N. N.; Lanford, W. A.; Li, H.; Rhodes, J.; Hopkins, P. E. Thermal conductivity enhancement in ion-irradiated hydrogenated amorphous carbon films. Nano Letters 2021, 21, 3935–3940.DOI
- [81]Koh, Y. K.; Cahill, D. G. Frequency dependence of the thermal conductivity of semiconductor alloys. Physical Review B 2007, 76, 075207.DOI
- [82]Scott, E. A.; Braun, J. L.; Hattar, K.; Sugar, J. D.; Gaskins, J. T.; Goorsky, M.; King, S. W.; Hopkins, P. E. Probing thermal conductivity of subsurface, amorphous layers in irradiated diamond. Journal of Applied Physics 2021, 129, 055307.DOI
- [83]Kresse, G.; Furthmüller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Physical review B 1996, 54, 11169.pubmed.ncbi.nlm.nih.gov/9984901
- [84]Blöchl, P. E. Projector augmented-wave method. Physical review B 1994, 50, 17953.pubmed.ncbi.nlm.nih.gov/9976227
- [85]Perdew, J. P.; Zunger, A. Self-interaction correction to density-functional approximations for many-electron systems. Physical Review B 1981, 23, 5048.DOI
- [86]Novikov, I. S.; Gubaev, K.; Podryabinkin, E. V.; Shapeev, A. V. The MLIP package: moment tensor potentials with MPI and active learning. Machine Learning: Science and Technology 2020, 2, 025002.arxiv.org/abs/2007.08555
- [87]Kryuchkov, N. P.; Yurchenko, S. O.; Fomin, Y. D.; Tsiok, E. N.; Ryzhov, V. N. Complex crystalline structures in a two-dimensional core-softened system. Soft Matter 2018, 14, 2152–2162.arxiv.org/abs/1712.04707
- [88]Wang, Z.; Safarkhani, S.; Lin, G.; Ruan, X. Uncertainty quantification of thermal conductivities from equilibrium molecular dynamics simulations. International Journal of Heat and Mass Transfer 2017, 112, 267–278.arxiv.org/abs/1612.05107
- [89]Green, M. S. Markoff random processes and the statistical mechanics of time-dependent phenomena. II. Irreversible processes in fluids. The Journal of chemical physics 1954, 22, 398–413.DOI
- [90]Kubo, R. Statistical-mechanical theory of irreversible processes. I. General theory and simple applications to magnetic and conduction problems. Journal of the physical society of Japan 1957, 12, 570–586.DOI
- [91]Turney, J.; McGaughey, A.; Amon, C. Assessing the applicability of quantum corrections to classical thermal conductivity predictions. Physical Review B 2009, 79, 224305.DOI
- [92]Thomas, J. A.; Turney, J. E.; Lutzi, R. M.; Amon, C. H.; McGaughey, A. J. Erratum: Predicting phonon dispersion relations and lifetimes from the spectral energy density [Phys. Rev. B 81, 081411 (R)(2010)]. Physical Review B 2015, 91, 239905.
- [93]Hoover, W. G. Canonical dynamics: Equilibrium phase-space distributions. Physical review A 1985, 31, 1695.pubmed.ncbi.nlm.nih.gov/9895674
- [94]Oliver, W. C.; Pharr, G. M. An improved technique for determining hardness and elastic modulus using load and displacement sensing indentation experiments. Journal of materials resarch 1992, 7, 1564–1583.DOI
- [95]Ding, Z.; Ridley, M.; Deijkers, J.; Liu, N.; Hoque, M. S. B.; Gaskins, J.; Zebarjadi, M.; Hopkins, P. E.; Wadley, H.; Opila, E.; others The thermal and mechanical properties of hafnium orthosilicate: experiments and first-principles calculations. Materialia 2020, 12, 100793.arxiv.org/abs/2003.07546
- [96]Zhang, P.; Chen, B.; Zhu, W.; Wang, C.; Zhang, W.; Li, Y.; Liu, W. First-principles prediction of structural, mechanical and thermal properties of perovskite BaZrS3. The European Physical Journal B 2020, 93, 1–7.DOI
- [97]Xia, Y.; Gao, Y.; Pharr, G. M.; Bei, H. Single versus successive pop-in modes in nanoindentation tests of single crystals. Journal of Materials Resarch 2016, 31, 2065–2075.DOI
- [98]Braun, J. L.; King, S. W.; Giri, A.; Gaskins, J. T.; Sato, M.; Fujiseki, T.; Fujiwara, H.; Hopkins, P. E. Breaking network connectivity leads to ultralow thermal conductivities in fully dense amorphous solids. Applied Physics Letters 2016, 109, 191905.DOI
- [99]Hoque, M. S. B.; Brummel, I. A.; Hoglund, E. R.; Dionne, C. J.; Aryana, K.; Tomko, J. A.; Gaskins, J. T.; Hirt, D.; Smith, S. W.; Beechem, T.; others Interface-independent sound speed and thermal conductivity of atomic-layer-deposition-grown amorphous AlN/Al2 O 3 multilayers with varying oxygen composition. Physical Review Materials 2023, 7, 025401.
- [100]Chen, Z.; Dames, C. An anisotropic model for the minimum thermal conductivity. Applied Physics Letters 2015, 107.
- [101]Mason, W. Properties of transducer materials. American Institute of Physics Handbook 1972, 3–118.
- [102]Mason, W. Acoustic properties of solids. American Institute of physics handbook 1972, 3–98.
- [103]Gray, D. E. American Institute of Physics (AIP). Handbook. New York: McGraw-Hill 1972,
- [104]Ho, C. Y.; Powell, R. W.; Liley, P. E. Thermal conductivity of the elements. Journal of Physical and Chemical Reference Data 1972, 1, 279–421.DOI
- [105]Grimsditch, M.; Zouboulis, E.; Polian, A. Elastic constants of boron nitride. Journal of applied physics 1994, 76, 832–834.DOI
- [106]Lehmann, G.; Hess, P.; Weissmantel, S.; Reisse, G.; Scheible, P.; Lunk, A. Young’s modulus and density of nanocrystalline cubic boron nitride films determined by dispersion of surface acoustic waves. Applied Physics A 2002, 74, 41–45.DOI
- [107]Chen, K.; Song, B.; Ravichandran, N. K.; Zheng, Q.; Chen, X.; Lee, H.; Sun, H.; Li, S.; Udalamatta Gamage, G. A. G.; Tian, F.; others Ultrahigh thermal conductivity in isotope-enriched cubic boron nitride. Science 2020, 367, 555–559.DOI
- [108]Yonenaga, I.; Shima, T.; Sluiter, M. H. Nano-indentation hardness and elastic moduli of bulk single-crystal AlN. Japanese journal of applied physics 2002, 41, 4620.DOI
- [109]Cheng, Z.; Koh, Y. R.; Mamun, A.; Shi, J.; Bai, T.; Huynh, K.; Yates, L.; Liu, Z.; Li, R.; Lee, E.; others Experimental observation of high intrinsic thermal conductivity of AlN. Physical Review Materials 2020, 4, 044602.DOI
- [110]Nowak, R.; Pessa, M.; Suganuma, M.; Leszczynski, M.; Grzegory, I.; Porowski, S.; Yoshida, F. Elastic and plastic properties of GaN determined by nano-indentation of bulk crystal. Applied physics letters 1999, 75, 2070–2072.DOI
- [111]Shibata, H.; Waseda, Y.; Ohta, H.; Kiyomi, K.; Shimoyama, K.; Fujito, K.; Nagaoka, H.; Kagamitani, Y.; Simura, R.; Fukuda, T. High thermal conductivity of gallium nitride (GaN) crystals grown by HVPE process. Materials Transactions 2007, 48, 2782–2786.
- [112]Morelli, D.; Jovovic, V.; Heremans, J. Intrinsically minimal thermal conductivity in cubic I- V - VI 2 semiconductors. Physical review letters 2008, 101, 035901.pubmed.ncbi.nlm.nih.gov/18764265
- [113]Berri, S.; Maouche, D.; Medkour, Y. Ab initio study of the structural, electronic and elastic properties of AgSbTe2, AgSbSe2, Pr3AlC, Ce3AlC, Ce3AlN, La3AlC and La3AlN compounds. Physica B: Condensed Matter 2012, 407, 3320–3327.DOI
- [114]Miller, A.; Saunders, G.; Yogurtcu, Y. Pressure dependences of the elastic constants of PbTe, SnTe and Ge0. 08Sn0. 92Te. Journal of Physics C: Solid State Physics 1981, 14, 1569.DOI
- [115]Gerlich, D. Elastic constants of single-crystal indium arsenide. Journal of Applied Physics 1963, 34, 2915–2915.DOI
- [116]Maycock, P. Thermal conductivity of silicon, germanium, III–V compounds and III–V alloys. Solid-state electronics 1967, 10, 161–168.
- [117]Xiao, Y.; Chang, C.; Pei, Y.; Wu, D.; Peng, K.; Zhou, X.; Gong, S.; He, J.; Zhang, Y.; Zeng, Z.; others Origin of low thermal conductivity in SnSe. Physical Review B 2016, 94, 125203.DOI
- [118]Chung, D.-H. Elastic moduli of single crystal and polycrystalline MgO. Philosophical Magazine 1963, 8, 833–841.DOI
- [119]Bell, R.; Rupprecht, G. Elastic constants of strontium titanate. Physical Review 1963, 129, 90.DOI
- [120]Yu, C.; Scullin, M. L.; Huijben, M.; Ramesh, R.; Majumdar, A. Thermal conductivity reduction in oxygen-deficient strontium titanates. Applied physics letters 2008, 92, 191911.DOI
- [121]Oh, D.-W.; Ravichandran, J.; Liang, C.-W.; Siemons, W.; Jalan, B.; Brooks, C. M.; Huijben, M.; Schlom, D. G.; Stemmer, S.; Martin, L. W.; others Thermal conductivity as a metric for the crystalline quality of SrTiO 3 epitaxial layers. Applied physics letters 2011, 98, 221904.DOI
- [122]Vassen, R.; Cao, X.; Tietz, F.; Basu, D.; Stöver, D. Zirconates as new materials for thermal barrier coatings. Journal of the American Ceramic Society 2000, 83, 2023–2028.
- [123]Lewis, F.; Saunders, N. The thermal conductivity of NiO and CoO at the Neel temperature. Journal of Physics C: Solid State Physics 1973, 6, 2525.DOI
- [124]Jifang, W.; Fisher, E.; Manghnznmi, M. Elastic constants of nickel oxide. Chinese Physics Letters 1991, 8, 153.DOI
- [125]Zhang, R.; Koutsos, V.; Cheung, R. Elastic properties of suspended multilayer WSe2. Applied Physics Letters 2016, 108.DOI
- [126]Dawley, N. M.; Pek, E. K.; Lee, C. H.; Ragasa, E. J.; Xiong, X.; Lee, K.; Phillpot, S. R.; Chernatynskiy, A. V.; Cahill, D. G.; Schlom, D. G. Thermal Conductivity of the n= 1-5 and 10 Members of the (SrTiO3)nSrO Ruddlesden-Popper Superlattices. 2021.DOI
- [127]Tachibana, M.; Kolodiazhnyi, T.; Takayama-Muromachi, E. Thermal conductivity of perovskite ferroelectrics. Applied Physics Letters 2008, 93, 092902.DOI
- [128]Lee, W.; Li, H.; Wong, A. B.; Zhang, D.; Lai, M.; Yu, Y.; Kong, Q.; Lin, E.; Urban, J. J.; Grossman, J. C.; others Ultralow thermal conductivity in all-inorganic halide perovskites. Proceedings of the National Academy of Sciences 2017, 114, 8693–8697.DOI
- [129]Bhui, A.; Ghosh, T.; Pal, K.; Singh Rana, K.; Kundu, K.; Soni, A.; Biswas, K. Intrinsically Low Thermal Conductivity in the n-Type Vacancy-Ordered Double Perovskite Cs2SnI6: Octahedral Rotation and Anharmonic Rattling. Chemistry of Materials 2022, 34, 3301–3310.DOI
- [130]Cahill, D. G.; Melville, A.; Schlom, D. G.; Zurbuchen, M. A. Low thermal conductivity of CsBiNb2O7 epitaxial layers. Applied Physics Letters 2010, 96, 121903.DOI












