betaFeedback

WorksPhysics and Astronomy

OpenBTE: a Solver for ab-initio Phonon Transport in Multidimensional Structures

Abstract

Controlling heat flow at the nanoscales is pivotal to several applications, including thermal energy harvesting and heat management. However, engineering nanostructures is challenging because phonon-boundary interaction, not contemplated by Fourier's law, must be taken into account. Nondiffusive models, such as the Boltzmann transport equation (BTE), have been successfully employed to capture size effects in complex structures; however, their widespread has been hindered by the limited offer of open-source solvers in this space. We fill this void by introducing OpenBTE, an efficient solver for the steady-state phonon BTE in multidimensional structures. This tool is interfaced to first-principles calculations, thus it unlocks the calculations of thermal-related properties with no fitting-parameters. As an example, we employ OpenBTE to compute the temperature and flux maps, as well as the mode-resolved, effective thermal conductivity of Si membranes with infinite and finite thickness. By unlocking fast nanoscale heat transport simulations, OpenBTE may help accelerate the design of nanomaterials for thermal energy applications.

Program Summary

Program Title: OpenBTE

Developer’s repository link: https://github.com/romanodev/openbte.git

Licensing provisions: MIT

Programming language: Python

Nature of problem: Ab-initio calculation of thermal transport properties, including temperature and heat flux maps, as well as the effective thermal conductivity of multidimensional nanostructures.

Solution method: We implement the space-dependent, ab-initio Boltzmann transport equation (BTE) for phonons. The first-principles calculation of phonon group velocities, frequencies and scattering times, is delegated to external packages to which OpenBTE is interfaced. The current implementation focuses on a flavor of the BTE that interpolates phonon populations onto their vectorial mean-free-paths, allowing for fast and accurate thermal transport simulations.

Additional comments: Vectorization via NUMPY and parallelization (including with shared memory) with MPI4Py.

Introduction

The increasing accuracy of first-principles calculations of heat transport, along with the growing availability of computing resources, has enabled the systematic prediction of lattice thermal conductivity in semiconductors and two-dimensional (2D) materials [1, 2, 3]. If key theoretical developments have solidified our understanding of phonon dynamics at the nanoscales, the fast progress of this field has been unarguably aided by the release of open-source packages. For example, ShengBTE [4] implements the iterative solution of the Boltzmann transport equation (BTE) [5], with force constants computed by first-principles [6, 7]. The tool can be primarily used for either bulk materials or nanowires [8]. Later iterations include AlmaBTE [9], which, among several improvements, features a Monte Carlo solver for superlattices, and FourPhonons [10, 11], a tool that, as the name suggests, can handle four-phonon scattering. Another notable package is Phono3Py [12], where the BTE is solved both within the relaxation time approximation (RTA) and including the full scattering operator [13]. Albeit not strictly required, these tools are mostly used in tandem with super-cell approaches. When a density functional perturbation theory (DFPT) is available for a given material, it is often convenient to compute directly the Fourier-transformed force constants, without resorting to supercells; a prominent package in this space is d3q [14], implemented on top of QUANTUM ESPRESSO [15]. This software also includes the possibility of simulating thin films [16] and coherence effects [17]. In a similar effort, the recently released Kaldo [18] combines the Green-Kubo model and the BTE to take into account material disorder [19].

While the dissemination of these software manifests the increasing excitement around phonon related applications, their focus is primarily on bulk materials or simple geometries. In fact, most phonon dynamics simulations in complex geometries are currently performed by MonteCarlo solvers [20, 21], with a recent implementation provided by MCBTE [22]. Thus, an open-source tool that solves space-dependent heat transport deterministically, i.e. in the same spirit as in bulk calculations, is still pending. We fill this gap by introducing OpenBTE, a solver for the first-principles, multidimensional phonon BTE. Key capabilities include the calculation of the temperature and heat flux maps, as well as the mode-resolve effective thermal conductivity. Ab-initio harmonic and anharmonic properties are delegated to external software. While flexible and straightforward to expand, OpenBTE currently implements the anisotropic-mean-free-path BTE (aMFP-BTE) [23], a method based on the interpolation of the phonon populations onto the vectorial MFP space. This method has shown a 50x speed up with respect to the mode-resolve BTE for thick Si membranes, while not compromising on the accuracy. Furthermore, it scales with constant time with respect to the number of phonon branches, therefore enabling fast simulations of size effects in complex-unit-cell materials, such as bismuth telluride.

The paper is organized as follows. First, we review the model and the assumptions therein, then we provide details on the implemented algorithm and the software’s main modules. An example of a porous Si membrane will follow, and final remarks conclude the paper. We also provides technical implementation details in the appendices. Enabling fast simulations of heat transport in nanostructures, OpenBTE sets out to accelerate the development of thermal-related applications, such as thermal management and thermoelectrics. OpenBTE is released under the GPL-3.0 license and currently hosted on GitHub [24].

Model

The steady-state, linearized phonon BTE in the temperature formulation reads [23]

−Cμvμ⋅∇ΔTμ(r)=∑νWμνΔTν(r),(1)-C_\mu\mathbf{v}_\mu\cdot\nabla\Delta T_\mu(\mathbf{r}) = \sum_\nu W_{\mu\nu}\Delta T_\nu(\mathbf{r}), \tag*{(1)}

where μ\mu, ν\nu collectively indicate phonon branch and wave vector, running up to NpN_p and NqN_q, respectively. The unknowns ΔTμ(r)\Delta T_\mu(\mathbf{r}) are the phonon pseudo-temperatures (referred to hereafter simply as temperatures), connected to the deviational non-equilibrium phonon population via ΔTμ(r)=Δnμ(r)ℏωμCμ−1\Delta T_\mu(\mathbf{r}) = \Delta n_\mu(\mathbf{r)\hbar\omega_\mu C_\mu^{-1}}. The mode-specific heat capacity is Cμ=kB(ημsinh⁡ημ)2C_\mu= k_B \left(\frac{\eta_\mu}{\sinh\eta_\mu}\right)^2, with ημ=ℏωμ/kB/T/2\eta_\mu= \hbar\omega_\mu/k_B/T/2. The terms ℏωμ\hbar\omega_\mu and vμ\mathbf{v}_\mu are the phonon frequencies and the group velocities, respectively. In this formulation, the thermal flux is given by J(r)=(NqV)−1∑μΔTμ(r)Sμ\mathbf{J}(\mathbf{r}) = (N_q\mathcal{V})^{-1}\sum_\mu\Delta T_\mu(\mathbf{r})\mathbf{S}_\mu, where V\mathcal{V} is the volume of the unit cell, and Sμ=Cμvμ\mathbf{S}_\mu= C_\mu\mathbf{v}_\mu. The term WμνW_{\mu\nu} is the energy form of the scattering operator [25]. While we plan to make available the implementation of Eq. (1), currently only the relaxation-time-approximation (RTA) is supported; within the RTA, the RHS of Eq. (1) becomes Cμ/τμ[ΔTμ(r)−ΔT(r)]C_\mu/\tau_\mu\left[\Delta T_\mu(\mathbf{r})-\Delta T(\mathbf{r})\right], leading to

−vμ⋅∇ΔTμ(r)=ΔTμ(r)−ΔT(r)τμ,(2)-\mathbf{v}_\mu\cdot\nabla\Delta T_\mu(\mathbf{r}) = \frac{\Delta T_\mu(\mathbf{r})-\Delta T(\mathbf{r})}{\tau_\mu}, \tag*{(2)}

where τμ\tau_\mu is the relaxation time, and ΔT(r)\Delta T(\mathbf{r}) is the pseudo lattice temperature; this term is evaluated so that energy conservation is satisfied, i.e. ∇⋅J(r)=0\nabla\cdot\mathbf{J}(\mathbf{r})=0, yielding ΔT(r)=[∑νCν/τν]−1∑μCμ/τμΔTμ(r)\Delta T(\mathbf{r}) = \left[\sum_\nu C_\nu/\tau_\nu\right]^{-1}\sum_\mu C_\mu/\tau_\mu\Delta T_\mu(\mathbf{r}) [26].

The simulation domain is a cuboid of size LxL_x, LyL_y and LzL_z with periodic boundary conditions applied throughout by default. A difference of temperature ΔText\Delta T_{\mathrm{ext}} is also enforced along a chosen Cartesian axis. At the walls of the pores, energy conservation leads to ∑νΔTνSμ⋅n^=0\sum_\nu\Delta T_\nu\mathbf{S}_\mu\cdot\hat{\mathbf{n}}=0, which can be split into incoming and outgoing flux, i.e. ∑μΔTμReLu⁡(Sμ⋅n^)=−∑μΔTμReLu⁡(−Sμ⋅n^)\sum_\mu\Delta T_\mu\operatorname{ReLu}(\mathbf{S}_\mu\cdot\hat{\mathbf{n}})=-\sum_\mu\Delta T_\mu\operatorname{ReLu}(-\mathbf{S}_\mu\cdot\hat{\mathbf{n}}); note that we define ReLu⁡(x)=xΘ(x)\operatorname{ReLu}(x)=x\Theta(x), with Θ(x)\Theta(x) being the Heaviside function. The boundary condition is imposed to phonons leaving the surface; generally, we have ΔTμ−=∑νRμνΔTν+\Delta T_\mu^-=\sum_\nu R_{\mu\nu}\Delta T_\nu^+, where −- (++) are for outgoing (incoming) phonons, and RμνR_{\mu\nu} is a mode-resolved bouncing matrix. However, currently only a simplified model is available, with which we assume that phonons thermalize to a single temperature [20]; this choice corresponds to Rμν=[∑kReLu⁡(Sk⋅n^)]−1ReLu⁡(Sμ⋅n^)R_{\mu\nu}=\left[\sum_k\operatorname{ReLu}(\mathbf{S}_k\cdot\hat{\mathbf{n}})\right]^{-1}\operatorname{ReLu}(\mathbf{S}_\mu\cdot\hat{\mathbf{n}}). Lastly, the effective thermal conductivity, assuming we have imposed the temperature gradient along xx, reads

κeff=−LxΔText∑μ∫AdSAJμ⋅n^.(3)\kappa^{\mathrm{eff}}=-\frac{L_x}{\Delta T_{\mathrm{ext}}}\sum_\mu\int_A\frac{dS}{A}\mathbf{J}_\mu\cdot\hat{\mathbf{n}}. \tag*{(3)}

Equation 2, when solved iteratively, requires inverting as many linear systems as the number of phonons modes, an endeavour that may easily become prohibitive. To ameliorate the computational effort, we instead implement the recently introduced anisotropic-MFP-BTE (aMFP-BTE), a model based on interpolating the phonon temperatures onto the vectorial MFP space [23]. In practice, instead of solving Eq. (2) NpNqN_pN_q times at each iteration, we solve it for a set of vectorial MFPs, Fml\mathbf{F}_{ml}, located on a spherical grid. The indices mm and ll label the MFP Λm\Lambda_m (up to NΛN_\Lambda) and phonon directions s^l\hat{\mathbf{s}}_l (up to NΩN_\Omega), respectively. As NΛNΩ≪NpNqN_{\Lambda}N_{\Omega} \ll N_{p}N_{q}, while not compromising on accuracy, this scheme results in a much faster runtime with respect to the case with mode-resolved resolution, i.e. no interpolation.

Implementation

As derived in Appendix A, the discretization of aMFP-BTE model in real and momentum space yields the following iterative linear system

AmlΔTml(n)=∑m′l′Smlm′l′ΔTm′l′(n−1)+Pml,(4)\mathbf{A}_{ml}\Delta\mathbf{T}_{ml}^{(n)} = \sum_{m'l'} \mathbf{S}_{ml}^{m'l'}\Delta\mathbf{T}_{m'l'}^{(n-1)} + \mathbf{P}_{ml}, \tag*{(4)}

where Aml\mathbf{A}_{ml} is the stiffness matrix for a given MFP and direction, Smlm′l′\mathbf{S}_{ml}^{m'l'} is associated to the lattice temperature and adiabatic boundary conditions, and Pml\mathbf{P}_{ml} represents the perturbation. Once Eq. (4) converges, the effective thermal conductivity is computed as κeff=κbal/2+∑mlKmlT⋅ΔTml\kappa_{\mathrm{eff}} = \kappa_{\mathrm{bal}}/2 + \sum_{ml} \mathbf{K}_{ml}^{T} \cdot\Delta\mathbf{T}_{ml}, where κbal\kappa_{\mathrm{bal}} is the ballistic thermal conductivity, and Kml\mathbf{K}_{ml} is associated to the thermal flux.

The first guess to Eq. (4) is given by the standard heat conduction equation. While it is straightforward to discretize the diffusive equation in orthogonal grids, it becomes convoluted for unstructured meshes. In OpenBTE, we use the approach described in [27], where the non-orthogonal contribution of the flux is treated explicitly, i.e. it depends on previous solutions. The resulting system, analogously to Eq. (4), is an iterative linear system,

AΔT(n)=SΔT(n−1)+P,(5)\mathbf{A}\Delta\mathbf{T}^{(n)} = \mathbf{S}\Delta\mathbf{T}^{(n-1)} + \mathbf{P}, \tag*{(5)}

where S\mathbf{S} arises from the non-orthogonality of the mesh. Note that Smlm′l′\mathbf{S}_{ml}^{m'l'} in Eq. (4) and S\mathbf{S} in Eq. (5) are not related. The same applies with the terms sharing the same basename in these two equations. The derivation of Eq. (5) is given in Appendix B.

Equation 4 entails, for each iteration nn, the inversion of NΛNΩN_{\Lambda}N_{\Omega} matrices, whose size is the number of volumes in the computational domain. To enhance the computational efficiency, three distinct strategies have been put in place. First, by noting that the stiffness matrix does not depend on nn, we are able to compute LU factorizations for the first iteration, and reuse them until convergence. Another major speed up is achieved by vectorizing the assembly of the stiffness matrices, while also exploiting their sparsity. Lastly, for each nn, Eq. (4) consists of a set of independent problems, thus it can be trivially parallelized [28]; to this end, we assign a set of phonon directions to each core (Ωc\Omega_c), while the indices mm run serially. The implementations of these three methods rely on SCIPY, NUMPY [29] and MPI4PY [30], respectively.

When vectorization and parallelization are both used, we may save memory by noting that, as shown in Appendix § 12, the stiffness matrices have the form Aml=I+ΛmAl\mathbf{A}_{ml} = \mathbf{I} + \Lambda_m\mathbf{A}_l; thus vectorization can be done only for Al\mathbf{A}_l, while Aml\mathbf{A}_{ml} is assembled on the fly right before LU factorization. Similar arguments hold for Pml\mathbf{P}_{ml} and Smlm′l′\mathbf{S}_{ml}^{m'l'}. The overall methodology is summerized in Algorithm 1, where, for readability, only Aml\mathbf{A}_{ml} has been expanded upon.

Algorithm 1 Workflow for thermal conductivity calculations. The term Ωc\Omega_c is the set of solid angles assigned to core cc.

  1. ΔTml(0)←\Delta\mathbf{T}_{ml}^{(0)} \leftarrow Fourier’s law.

  2. while error >10−3> 10^{-3} do

  3. κeff,(n+1)=κbal/2\kappa^{\mathrm{eff},(n+1)} = \kappa_{\mathrm{bal}}/2

  4. for ll in {Ωc}\{\Omega_c\} do

  5. for m=1:NΛm = 1:N_\Lambda do

  6. Bml=∑m′l′Smlm′l′ΔTm′l′(n−1)+Pml\mathbf{B}_{ml} = \sum_{m'l'} \mathbf{S}_{ml}^{m'l'} \Delta\mathbf{T}_{m'l'}^{(n-1)} + \mathbf{P}_{ml}

  7. ΔTml(n)←(I+ΛmAl)ΔTml(n)=B\Delta\mathbf{T}_{ml}^{(n)} \leftarrow(\mathbf{I} + \Lambda_m\mathbf{A}_l)\Delta\mathbf{T}_{ml}^{(n)} = \mathbf{B}

  8. κ+eff,(n+1)=KmlT⋅ΔTml\kappa^{\mathrm{eff},(n+1)}_+ = \mathbf{K}_{ml}^{T} \cdot\Delta\mathbf{T}_{ml}

  9. end for

  10. end for

  11. error←∣(κeff,(n+1)−κeff,(n))∣/κeff,(n+1)\mathrm{error} \leftarrow\lvert(\kappa^{\mathrm{eff},(n+1)} - \kappa^{\mathrm{eff},(n)})\rvert/\kappa^{\mathrm{eff},(n+1)}

  12. n←n+1n \leftarrow n + 1

  13. end while

Workflow

The easiest way to run OpenBTE is to use it as a module in Python. A sketch for the workflow is illustrated in Figure 1. The core module is Solver, which runs both the heat conduction equation as a first guess to Eq. (4) and all subsequent iterations, outlined in Algorithm 1. Before invoking it, though, it is necessary to define the material’s internal boundaries using the Geometry module, and bulk-related data via the Material module. The latter is the interface to external packages. Once the simulation is finalized, results can be plotted using the Plot module. Note that OpenBTE follows a pure functional approach, where data is simply stored as dictionaries. Each module is expanded upon in the next sections.

Image not extracted

Figure 1. The software architecture of OpenBTE. The key components are the Material, Geometry, Solver and Plot modules. Currently, the code is interfaced to AlmaBTE and Phono3Py, which computes first-principles data. Mesh generation is delegated to GMSH, and output maps can be visualized with Paraview.

Material

The Material module converts mode-resolved bulk data, CμC_{\mu}, τμ\tau_{\mu} and vμ\mathbf{v}_{\mu}, into the spherically-resolved quantities Gml\mathbf{G}_{ml} and CmlC_{ml}. The input files, rta.npz, are created by postprocessing data obtained by external packages. Currently, OpenBTE provides interfaces to AlmaBTE [9] and Phono3Py [12]. A small set of precomputed rta.npz is also provided. The file created by Material is stored into material.npz and ready to be used by Solver. Depending on the geometry, this module has two models:

  • rta3D: this model must be used for geometries with lz>0l_z > 0. In this case, the angular interpolation of Fμ\mathbf{F}_{\mu} is performed in both polar and azimuthal angles.

  • rta2DSym: when lzl_z is not specified, the actual simulation domain is 2D, and this model must be used. Note that when the base material is 3D, using this option is equivalent to considering an infinite thickness. In fact, in this case, the mode-resolved Fμ\mathbf{F}_{\mu} are mapped into Ftk=Λtcos⁡(ϕk)\mathbf{F}_{tk} = \Lambda_t \cos(\phi_k), where Λt\Lambda_t implicitly includes the MFP and the azimuthal angle, and ϕk\phi_k is the polar angle on the xx-yy plane [23].

The model as well as the discretization grid can be chosen with

from openbte import Material
Material(filename='rta_Si_300',model='rta2DSym',n_phi=48)

where, by default, the material file is taken from the database, whose entries are updated online. In this case, we choose to discretize the polar angle into 48 slices.

Geometry module

The Geometry module allows the user to build an arbitrarily patterned nanomaterial. It interfaces with GMSH [31] to create an unstructured Delaney mesh, and creates the file geometry.npz. While we plan to handle user-provided meshes, OpenBTE currently supports 2D membranes with finite and infinite thickness. The size of the unit cell is specified by lx and ly, while the thickness is defined by lz. The direction of the applied gradient is indicated with direction (default is x). Periodic boundary conditions are applied along all directions; however, adiabatic surfaces can be imposed along specific directions with Periodic=[True,False,True], where each component declares whether a surface is periodic. Currently, there are two geometry models, as expanded upon in the next sections.

Lattice model

This model aids the creation of 2D periodic porous materials with pores having either predefined shapes or custom ones. In analogy to crystals, a porous material is defined as a set of unit vectors, which in our case are a1=lxx^\mathbf{a}_1=l_x\hat{\mathbf{x}} and a2=lyy^\mathbf{a}_2=l_y\hat{\mathbf{y}}, and a base. The latter is a list of xx-yy pairs, each of them being one pore. Here is an example with a single pore in the origin

from openbte import Geometry
Geometry(model='lattice',base=[[0,0]],shape='circle')

In the example above, we chose a simple shape, a circle; however, it is possible also to supply a user-defined shape using shape=custom. In such a case, the keyword shape_functions must be provided, optionally along with shape_options. For example, as shown in Figure 2a-Figure 2b and Figure 2c, we can define a structure having two pores, each of them described by the same custom shape.

Image not extracted

Figure 2. a) An example of structure using the keyword-value pair shape=custom for the lattice model of the Geometry module. b) the unit cell comprising two pores at designed positions. Note that when a pore crosses a boundary it gets repeated. c) singe-pore view of material. The coordinates of the shape are defined via the shape_function option. d) An example of custom shape obtained by carving out the base material with user-defined polygons.

Custom model

Using the option model=custom, it is possible to define an arbitrary patterning by providing a list of polygons, with the periodicity being automatically guaranteed. This option significantly widens up the number of possible geometries at the expense of more input by the user. In spirit of subtractive manufacturing, each polygon can be seen as a region where the material is carved out. An example of custom model is depicted in Figure 2-d.

Plot Module

Simulations results can be plotted via the Plot module. Currently, the following features are implemented:

  • vtu: the temperature and flux maps are stored in a vtu file and ready to be opened by Paraview [32].

  • maps: the maps are shown as web-ready, interactive structures.

  • kappa_mode: this model computes the mode-resolved thermal conductivity from Eq A.16, and stores in a file.

  • line: this option, which currently is implemented only for 2D systems, allows for line plots, using linear interpolation.

Thermal transport in porous Si

In some cases, it is convenient to use Fourier’s law to estimate macroscopic geometric effects. In this first part, we thus aim to assess the reliability of the diffusive solver, implemented in OpenBTE. To this end, we consider a membrane with circular pores and infinite thickness, for which κfourier/κbulk=r\kappa_{\mathrm{fourier}}/\kappa_{\mathrm{bulk}} = r can be evaluated analytically. In fact, in this case, the thermal transport suppression is given by the Eucken-Maxwell model r=(1−ϕ)/(1+ϕ)r = (1 - \phi)/(1 + \phi) [33], ϕ\phi being the porosity. In Figure 3a, we compare this model with the results from OpenBTE for different porosities, obtaining excellent agreement (<1%< 1\%).

Image not extracted

Figure 3. a) Thermal transport reduction when using Fourier’s law, for regular circular pores and for a user-defined shape; the Eucken-Maxwell model is also plotted. In the inset, the magnitude of the thermal map for the user-defined structure. The porosity is 0.1 and L=50L = 50 nm. b) a cut of the temperature maps, computed from both the BTE and Fourier’s law, along the xx-axis. c) The effective thermal conductivity, κeff\kappa^{\mathrm{eff}}, of the user-defined structure with thickness t=10t = 10 nm, and for different porosities. In the inset, the case with porosities = 0.1 and 0.17 are shown. d) The mode-resolved thermal conductivity, κμeff\kappa_{\mu}^{\mathrm{eff}}, versus the MFP, ∣Fμ,x∣\lvert F_{\mu,x}\rvert, for the cases t=10t = 10 nm and t=∞t = \infty.

We note that by using the option only_fourier=True it is possible to run only the Fourier’s model. Next, we compute rr for a more complicated structure, comprising two concatenated sub-lattices of pores with different user-provided shapes; this structure is generated using the shape=custom option from the lattice model, a two-pore base, and different options for each pore. As shown in Figure 3a, we obtain a much larger heat transport suppression due to the longest path heat must travel through the material.

Let us now compute κeff\kappa^{\mathrm{eff}} using the BTE, encoded in Eq. (4). For the chosen porosity of 0.1 and L=50L = 50nm, we obtain 17.4 Wm−1^{-1}K−1^{-1}, well below the corresponding macroscopic value 99.4 Wm−1^{-1}K−1^{-1}. In Figure 3b, we plot a cut of the temperature maps, both for the BTE and Fourier cases, along the xx-axis. As expected, they have different trends, the BTE one being flatter due to ballistic transport. When a finite thickness is added, phonon undergo stronger suppression; in fact, for thickness t=10t=10 nm and porosity 0.1, we obtain κeff=8.9\kappa^{\mathrm{eff}} = 8.9 W m−1^{-1}K−1^{-1}, roughly a half of the value with t=∞t = \infty. In Figure 3c, κeff\kappa^{\mathrm{eff}} for different porosities is plotted. From the insets, we note that, as the porosity becomes larger, the heat flux peaks around the phonon bottleneck, e.g. the areas between the pores. Lastly, the mode-resolved thermal conductivities for t=10t = 10 nm and ∞\infty are plotted in Figure 3d; analogously to [23], we can appreciate the stronger suppression for large MFPs for the case with finite thickness.

Conclusion

We have presented OpenBTE, a software for computing deterministically space-dependent heat transport in 2D and 3D structures. The efficient and modern implementation of the underlying algorithms enable accurate simulations both in the cloud and on common laptops, thus democratizing this type of simulations. Future work includes going beyond the RTA of the BTE and adding transient dynamics. GPU support and differentiability are also on the roadmap. Furthermore, we plan to develop a hybrid Fourier/BTE solver, which has the potential to dramatically reduce runtimes for the cases with 3D structures. Filling an important gap in the current offer of publicly available phonon transport solvers, OpenBTE sets out to guide experiments on nanoscale heat transport and facilitate high-throughput prediction of thermal properties of nanostructures.

Acknowledgments

Research was partially supported by the Solid-State SolarThermal Energy Conversion Center (S3TEC), an Energy Frontier Research Center funded by the U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences (BES), under Award No. DESC0001. The author thanks all the early users, who helped improving OpenBTE.

The aMFP-BTE

OpenBTE employs the upwind finite-volume discretization of Eq. (2), where the flux outgoing from volume ii contributes to the diagonal of the stiffness matrix at row ii [27]. A sketch depicting incoming and outgoing flux is illustrated in Fig. A.4a. The discretized mode-resolved BTE reads [23]

∑jμ′[δcc′δμμ′+Aijμ′+Bμμ′jδij+Dμ′δij]ΔTjμ′=Piμ,(6)\sum_{j\mu'} \left[\delta_{cc'}\delta_{\mu\mu'} + A_{ij\mu'} + B_{\mu\mu'j}\delta_{ij} + D_{\mu'}\delta_{ij}\right]\Delta T_{j\mu'} = P_{i\mu}, \tag*{(6)}

where

Aijμ=∑kReLu⁡(Fμ⋅Rik)δij−ReLu⁡(−Fμ⋅Rik)δjk,(7)A_{ij\mu} = \sum_k \operatorname{ReLu}(\mathbf{F}_{\mu}\cdot\mathbf{R}_{ik})\delta_{ij} - \operatorname{ReLu}(-\mathbf{F}_{\mu}\cdot\mathbf{R}_{ik})\delta_{jk}, \tag*{(7)}
Bμμ′j=−∑sReLu⁡(−Fμ⋅n^s)gsiReLu⁡(Sν⋅n^s)∑kReLu⁡(Sk⋅n^s),(8)B_{\mu\mu'j} = -\sum_s \operatorname{ReLu}(-\mathbf{F}_{\mu}\cdot\hat{\mathbf{n}}_s)g_{si}\frac{\operatorname{ReLu}(\mathbf{S}_{\nu}\cdot\hat{\mathbf{n}}_s)}{\sum_k\operatorname{ReLu}(\mathbf{S}_k\cdot\hat{\mathbf{n}}_s)}, \tag*{(8)}

arise flux discretization and adiabatic boundary conditions, respectively. The term Dμ=Cμ/τμ[∑kCk/τk]−1D_{\mu}=C_{\mu}/\tau_{\mu}\left[\sum_k C_k/\tau_k\right]^{-1} contributes to the lattice pseudo-temperature; lastly, the perturbation is

Piμ=−∑jReLu⁡(−Fμ⋅RijP)ΔText.(9)P_{i\mu}=-\sum_j\operatorname{ReLu}(-\mathbf{F}_{\mu}\cdot\mathbf{R}^{P}_{ij})\Delta T_{\mathrm{ext}}. \tag*{(9)}

The terms Rij\mathbf{R}_{ij}, RijP\mathbf{R}^{P}_{ij} and gscg_{sc} define the mesh, and are given by

Rij={1VcAijn^ij,if i and j are neighbors0,otherwise,(10)\mathbf{R}_{ij}= \begin{cases} \frac{1}{V_c}A_{ij}\hat{\mathbf{n}}_{ij}, & \text{if } i \text{ and } j \text{ are neighbors} \\ 0, & \text{otherwise}, \end{cases} \tag*{(10)}
RijP={1ViAijn^ij,if i and j are periodic to each other0,otherwise.(11)\mathbf{R}^{P}_{ij}= \begin{cases} \frac{1}{V_i}A_{ij}\hat{\mathbf{n}}_{ij}, & \text{if } i \text{ and } j \text{ are periodic to each other} \\ 0, & \text{otherwise}. \end{cases} \tag*{(11)}
gsi={1ViAs,if s is a side of i0,otherwise.(12)g_{si}= \begin{cases} \frac{1}{V_i}A_s, & \text{if } s \text{ is a side of } i \\ 0, & \text{otherwise}. \end{cases} \tag*{(12)}

where AijA_{ij} is the area of the surface between the volumes ii and jj, and n^ij\hat{\mathbf{n}}_{ij} is its normal, pointing toward the volume jj. The term ViV_i is the volume of the element ii. We solve A.1 iteratively,

∑j(δij+Aijμ)ΔTjμ(n)=Piμ+∑μ′(Bμμ′i+Dμ′)ΔTiμ′(n−1).(13)\sum_j \left(\delta_{ij}+A_{ij}\mu\right)\Delta T_{j\mu}^{(n)}=P_{i\mu}+\sum_{\mu'}\left(B_{\mu\mu'i}+D_{\mu'}\right)\Delta T_{i\mu'}^{(n-1)}. \tag*{(13)}

Equation A.8 requires solving as many linear systems as the number of phonons modes. To overcome this issue, OpenBTE implements the aMFP-BTE [23], where the phonon temperatures are interpolated over the vectorial MFPs, Fμ=∑mlcmlμFml\mathbf{F}_{\mu}=\sum_{ml}c_{ml}^{\mu}\mathbf{F}_{ml}; the chosen Fml=Λms^l\mathbf{F}_{ml}=\Lambda_m\hat{\mathbf{s}}_l are located uniformly on a spherical grid. The labels mm and ll refer to MFP and solid angle, respectively. Analogously, the temperatures are ΔTμ=∑mlcmlμΔTml\Delta T_{\mu}=\sum_{ml}c_{ml}^{\mu}\Delta T_{ml}. In the aMFP-BTE formulation, Eq. A.8 becomes

∑j(δij+ΛmAijl)ΔTjml(n)=ΛmPil+∑m′l′(ΛmBlm′l′i+Dm′l′)ΔTim′l′(n−1),(14)\sum_j\left(\delta_{ij}+\Lambda_m A_{ijl}\right)\Delta T_{jml}^{(n)}=\Lambda_m P_{il}+\sum_{m'l'}\left(\Lambda_m B_{lm'l'i}+D_{m'l'}\right)\Delta T_{im'l'}^{(n-1)}, \tag*{(14)}

where

Aij′l=∑kReLu⁡(s^l⋅Rik)δij−ReLu⁡(−s^l⋅Rik)δjk,(15)A_{ij'l}=\sum_k\operatorname{ReLu}(\hat{\mathbf{s}}_l\cdot\mathbf{R}_{ik})\delta_{ij}-\operatorname{ReLu}(-\hat{\mathbf{s}}_l\cdot\mathbf{R}_{ik})\delta_{jk}, \tag*{(15)}
Blm′l′i=−∑sReLu⁡(−s^l⋅n^s)gsiReLu⁡(Sm′l′⋅n^s)∑m′′l′′ReLu⁡(Sm′′l′′⋅n^s).(16)B_{lm'l'i}=-\sum_s\operatorname{ReLu}(-\hat{\mathbf{s}}_l\cdot\hat{\mathbf{n}}_s)g_{si}\frac{\operatorname{ReLu}(\mathbf{S}_{m'l'}\cdot\hat{\mathbf{n}}_s)}{\sum_{m''l''}\operatorname{ReLu}(\mathbf{S}_{m''l''}\cdot\hat{\mathbf{n}}_s)}. \tag*{(16)}

The term PilP_{il} is the perturbation, given by

Pil=−∑jReLu⁡(−s^l⋅RijP)ΔText.(17)P_{il}=-\sum_j\operatorname{ReLu}(-\hat{\mathbf{s}}_l\cdot\mathbf{R}_{ij}^{P})\Delta T_{\mathrm{ext}}. \tag*{(17)}

The quantities Dml=∑μcmlμDμD_{ml}=\sum_{\mu}c_{ml}^{\mu}D_{\mu} and Sml=∑μcmlμSμS_{ml}=\sum_{\mu}c_{ml}^{\mu}S_{\mu} contributes to the lattice temperature and thermal flux, respectively. Finally, after defining Sim′l′m′l′=ΛmBlm′l′i+Dm′l′S_{im'l'}^{m'l'}=\Lambda_mB_{lm'l'i}+D_{m'l'} in Eq. A.9, we obtain Eq. 4. To compute the effective thermal conductivity from Eq. 3, we need to integrate over the hot contact and the momentum space. According to the upwind scheme, the volume from which we pick the temperature depends on whether the flux is outgoing from or incoming to such volume; accordingly, Eq. 3 translates into

κeff=−LAΔText∑ijml[ΔTcmlReLu⁡(Sml⋅RijK)−(ΔTiml+ΔText)ReLu⁡(−Sml⋅RijK)]=κbal2+∑mlKmlTΔTml(18)\begin{aligned} \kappa_{\mathrm{eff}}&=-\frac{L}{A\Delta T_{\mathrm{ext}}}\sum_{ijml}\left[\Delta T_{cml}\operatorname{ReLu}(\mathbf{S}_{ml}\cdot\mathbf{R}_{ij}^{K})\right.\\ &\left.\qquad-(\Delta T_{iml}+\Delta T_{\mathrm{ext}})\operatorname{ReLu}(-\mathbf{S}_{ml}\cdot\mathbf{R}_{ij}^{K})\right]\\ &=\frac{\kappa_{\mathrm{bal}}}{2}+\sum_{ml}\mathbf{K}_{ml}^{T}\Delta\mathbf{T}_{ml} \tag*{(18)} \end{aligned}
Image not extracted

Figure A.4. a) A scheme for two adjacent elements. In this case, the vectorial MFP Fμ\mathbf{F}_{\mu} points inward the element ii, thus it contributes to the off-diagonal part of the stiffness matrix of the BTE model. b) Local reference system, e^ξ−e^η\hat{e}_{\xi}-\hat{e}_{\eta}, used to derive the non-orthogonal contribution to the heat flux for the heat-conduction equation.

where κbal=L∑mlGml⋅n^ext\kappa_{\mathrm{bal}}=L\sum_{ml}\mathbf{G}_{ml}\cdot\hat{n}_{\mathrm{ext}} is the ballistic thermal conductivity, n^ext\hat{n}_{\mathrm{ext}} being the direction of the applied temperature gradient; the matrices Kml\mathbf{K}_{ml} are

Kiml=LAΔText∑jSml⋅RijK,(19)K_{i}^{ml}=\frac{L}{A\Delta T_{\mathrm{ext}}}\sum_{j}\mathbf{S}_{ml}\cdot\mathbf{R}_{ij}^{K}, \tag*{(19)}

where we use ReLu⁡(x)−ReLu⁡(−x)=x\operatorname{ReLu}(x)-\operatorname{ReLu}(-x)=x, and

RijK={Aijn^ij,if i and j share a hot contact’s side0,otherwise.(20)\mathbf{R}_{ij}^{K}= \begin{cases} A_{ij}\hat{\mathbf{n}}_{ij}, & \text{if }i\text{ and }j\text{ share a hot contact's side}\\ 0, & \text{otherwise.} \end{cases} \tag*{(20)}

Lastly, it is also possible to compute the “mode-resolved” thermal conductivity, given by

κμeff=κbal2+∑mlKμT,mlcμmlΔTml,(21)\kappa_{\mu}^{\mathrm{eff}}=\frac{\kappa_{\mathrm{bal}}}{2}+\sum_{ml}K_{\mu}^{T,ml}c_{\mu}^{ml}\Delta T_{ml}, \tag*{(21)}

where cμmlc_{\mu}^{ml} maps Fml\mathbf{F}_{ml} to Fμ\mathbf{F}_{\mu}, and

Kiμ=LAΔText∑jSμ⋅RijK.(22)K_{i\mu}=\frac{L}{A\Delta T_{\mathrm{ext}}}\sum_{j}\mathbf{S}_{\mu}\cdot\mathbf{R}_{ij}^{K}. \tag*{(22)}

Heat conduction equation

The first guess to Eq. (4) is given by the standard diffusive equation

∇κ∇ΔT(r)=0,\nabla\kappa\nabla\Delta T(\mathbf{r})=0,

with κ\kappa being the bulk thermal conductivity tensor. In a finite-volume fashion, we integrate over the control volume Ωc\Omega_c

∫∂Ωc∑jκij∂ΔT(r)∂xjn^i=0,(23)\int_{\partial\Omega_c} \sum_j \kappa_{ij}\frac{\partial\Delta T(\mathbf{r})}{\partial x_j}\hat{n}_i = 0, \tag*{(23)}

where we used Gauss’ law. Eq. B.1 is then transformed into a sum of integrals over individual faces of the volume, i.e.

∑jgij⋅Sij=0(24)\sum_j \mathbf{g}_{ij}\cdot\mathbf{S}_{ij}=0 \tag*{(24)}

where Sij=∑βκαβRij,α\mathbf{S}_{ij}=\sum_\beta\kappa_{\alpha\beta}R_{ij,\alpha} and gij\mathbf{g}_{ij} is the gradient evaluated at the face between the volume ii and jj (which hereafter will be referred to as ss). For orthogonal grids, flux discretization simply leads to ∇T=(ΔTj−ΔTi)dij∣dij∣−2\nabla T=(\Delta T_j-\Delta T_i)\mathbf{d}_{ij}|\mathbf{d}_{ij}|^{-2}, where dij\mathbf{d}_{ij} is the distance between the centroids of the two adjacent volumes. However, for unstructured mesh, a non-orthogonal contribution must be added. In OpenBTE, we follow the approach described by Murthy [27], described as follows. With no loss of generality, we consider a 2D case. As shown in Fig. A.4b, we introduce a reference system based on the vectors e^η\hat{\mathbf{e}}_\eta and e^ξ\hat{\mathbf{e}}_\xi. The former is aligned with dij\mathbf{d}_{ij} and latter is parallel to the face. In this new system, the gradient transforms as

∂ΔT∂x=J(x,e)∂ΔT∂e,(25)\frac{\partial\Delta T}{\partial x}=\mathcal{J}(\mathbf{x},\mathbf{e})\frac{\partial\Delta T}{\partial e}, \tag*{(25)}

with J\mathcal{J} being the Jacobian. Here we will not elaborate on its structure but will only provide the final formula for the temperature gradient [27]

gij=n^ijn^ij⋅e^ξ[∂ΔT∂eξ∣s−e^ξ⋅e^η∂ΔT∂eη∣s].(26)\mathbf{g}_{ij}=\frac{\hat{\mathbf{n}}_{ij}}{\hat{\mathbf{n}}_{ij}\cdot\hat{\mathbf{e}}_\xi}\left[\left.\frac{\partial\Delta T}{\partial e_\xi}\right|_s-\hat{\mathbf{e}}_\xi\cdot\hat{\mathbf{e}}_\eta\left.\frac{\partial\Delta T}{\partial e_\eta}\right|_s\right]. \tag*{(26)}

The gradient along eξ\mathbf{e}_\xi at the face ss is simply

∂ΔT∂eξ∣s=ΔTj−ΔTi−RijP∣dij∣,(27)\left.\frac{\partial\Delta T}{\partial e_\xi}\right|_s=\frac{\Delta T_j-\Delta T_i-R^P_{ij}}{|\mathbf{d}_{ij}|}, \tag*{(27)}

where

RijP={−ΔText,if i∈hot and j∈cold and they share a sideΔText,if i∈cold and j∈hot and they share a side0otherwise(28)R^P_{ij}= \begin{cases} -\Delta T_{\mathrm{ext}}, & \text{if } i\in\mathrm{hot}\text{ and }j\in\mathrm{cold}\text{ and they share a side}\\ \Delta T_{\mathrm{ext}}, & \text{if } i\in\mathrm{cold}\text{ and }j\in\mathrm{hot}\text{ and they share a side}\\ 0 & \text{otherwise} \end{cases} \tag*{(28)}

Using Eq. B.5, we rewrite Eq. B.4 as

gij=n^ijn^ij⋅dij[ΔTj−ΔTi−RijP]+Fij,(29)\mathbf{g}_{ij} = \frac{\hat{\mathbf{n}}_{ij}}{\hat{\mathbf{n}}_{ij} \cdot\mathbf{d}_{ij}} \left[\Delta T_j - \Delta T_i - R^{P}_{ij}\right] + \mathbf{F}_{ij}, \tag*{(29)}

where Fij\mathbf{F}_{ij} is called “secondary flux.” In contrast, Eq. B.5 is termed “primary flux.” The secondary flux is treated explicitly, i.e. using precomputed temperatures. To avoid unambiguities in the 3D case, we compute F\mathbf{F} as the difference between the total flux and the primary flux, i.e.

Fij=gij−n^ijn^ij⋅dij(gij⋅dij).(30)\mathbf{F}_{ij} = \mathbf{g}_{ij} - \frac{\hat{\mathbf{n}}_{ij}}{\hat{\mathbf{n}}_{ij} \cdot\mathbf{d}_{ij}} \left(\mathbf{g}_{ij} \cdot\mathbf{d}_{ij}\right). \tag*{(30)}

Combining Eqs. B.8–B.7, and explicitly indicating the iteration index, we have

gij(n)=gij(n−1)+n^ijn^ij⋅dij[(ΔTj(n)−ΔTi(n))−gij(n−1)⋅dij].(31)\mathbf{g}_{ij}^{(n)} = \mathbf{g}_{ij}^{(n-1)} + \frac{\hat{\mathbf{n}}_{ij}}{\hat{\mathbf{n}}_{ij} \cdot\mathbf{d}_{ij}} \left[\left(\Delta T_j^{(n)} - \Delta T_i^{(n)}\right) - \mathbf{g}_{ij}^{(n-1)} \cdot\mathbf{d}_{ij}\right]. \tag*{(31)}

In light of these results, Eq. B.2 translates into the following linear system

AΔT=B+P,(32)\mathbf{A}\Delta\mathbf{T} = \mathbf{B} + \mathbf{P}, \tag*{(32)}

with stiffness matrix and right hand side

Aij=Sij⋅nijn^ij⋅dij−δij∑kSik⋅nikn^ij⋅dik,A_{ij} = \mathbf{S}_{ij} \cdot\frac{\mathbf{n}_{ij}}{\hat{\mathbf{n}}_{ij} \cdot\mathbf{d}_{ij}} - \delta_{ij} \sum_k \mathbf{S}_{ik} \cdot\frac{\mathbf{n}_{ik}}{\hat{\mathbf{n}}_{ij} \cdot\mathbf{d}_{ik}},
Bi=∑jgij⋅(dij(n^ij⋅Sij)n^ij⋅dij−Sij).(33)B_i = \sum_j \mathbf{g}_{ij} \cdot\left(\frac{\mathbf{d}_{ij}\left(\hat{\mathbf{n}}_{ij} \cdot\mathbf{S}_{ij}\right)}{\hat{\mathbf{n}}_{ij} \cdot\mathbf{d}_{ij}} - \mathbf{S}_{ij}\right). \tag*{(33)}

The perturbation term is given by

Pi=∑jSijPnijn^ij⋅dij,(34)P_i = \sum_j S^{P}_{ij} \frac{\mathbf{n}_{ij}}{\hat{\mathbf{n}}_{ij} \cdot\mathbf{d}_{ij}}, \tag*{(34)}

where SijP=SijRijPS^{P}_{ij} = S_{ij}R^{P}_{ij}. The gradients gij\mathbf{g}_{ij}, appearing in Eq. B.11, depend on precomputed temperatures. Their calculation if performed by interpolating the gradients from the elements ii and jj, i.e.

gij=ωijgi+(1−ωij)gj,(35)\mathbf{g}_{ij} = \omega_{ij}\mathbf{g}_i + (1-\omega_{ij})\mathbf{g}_j, \tag*{(35)}

where

gi=∂ΔT∂x∣i,(36)\mathbf{g}_i = \left.\frac{\partial\Delta T}{\partial\mathbf{x}}\right|_i, \tag*{(36)}

and ωij\omega_{ij} is the distance between the centroid of the element ii and the intersection between dij\mathbf{d}_{ij} and the face ss. The gradient at the element’s centroid is evaluated assuming local linear variation of ΔT\Delta T around the element ii, resulting in as many linear equations as the neighbours of ii,

ΔTi+gi⋅dij=ΔTj+RijP,(37)\Delta T_i + \mathbf{g}_i \cdot\mathbf{d}_{ij} = \Delta T_j + R_{ij}^{P}, \tag*{(37)}

which, in matrix form, become

Migi=Qi.(38)\mathbf{M}_i\mathbf{g}_i = \mathbf{Q}_i. \tag*{(38)}

Equation B.16 is a linear system to be solved for each element. Since Mi\mathbf{M}_i has more rows (number of neighbors) that columns (dimensions), we solve it using the least square method, i.e.

gi=(MiTMi)−1MiTQi.(39)\mathbf{g}_i = \left(\mathbf{M}_i^{T}\mathbf{M}_i\right)^{-1}\mathbf{M}_i^{T}\mathbf{Q}_i. \tag*{(39)}

Combining Eqs. B.17-B.11 and-B.10, we may formalize the heat conduction equation in unstructured grids as the following iterative linear system

AΔT(n)=SΔT(n−1)+P,(40)\mathbf{A}\Delta\mathbf{T}^{(n)} = \mathbf{S}\Delta\mathbf{T}^{(n-1)} + \mathbf{P}, \tag*{(40)}

as reported in the main text. Lastly, after Eq. B.18 is solved, κfourier\kappa_{\mathrm{fourier}} is computed by

κfourier=−LΔTextA∑s∈hotgij⋅Sij.(41)\kappa_{\mathrm{fourier}} = -\frac{L}{\Delta T_{\mathrm{ext}} A}\sum_{s \in\mathrm{hot}} \mathbf{g}_{ij}\cdot\mathbf{S}_{ij}. \tag*{(41)}

References

  1. [1]L. Lindsay, A. Katre, A. Cepellotti, N. Mingo, Perspective on ab initio phonon thermal transport, Journal of Applied Physics 126 (5) (2019) 050902. URL https://www.osti.gov/pages/biblio/1550742
  2. [2]L. Lindsay, C. Hua, X. Ruan, S. Lee, Survey of ab initio phonon thermal transport, Materials Today Physics 7 (2018) 106–120. URL https://www.sciencedirect.com/science/article/abs/pii/S254252931830141X
  3. [3]N. Mingo, D. Stewart, D. Broido, L. Lindsay, W. Li, Ab initio thermal transport, Length-scale dependent phonon interactions (2014) 137–173. URL https://link.springer.com/chapter/10.1007/978-1-4614-8651-0_5
  4. [4]W. Li, J. Carrete, N. A. Katcho, N. Mingo, ShengBTE: A solver of the Boltzmann transport equation for phonons, Comput. Phys. Commun. 185 (6) (2014) 1747–1758. URL http://www.sciencedirect.com/science/article/pii/S0010465514000484
  5. [5]M. Omini, A. Sparavigna, An iterative approach to the phonon boltzmann equation in the theory of thermal conductivity, Physica B: Condensed Matter 212 (2) (1995) 101–112. URL https://www.sciencedirect.com/science/article/abs/pii/0921452695000163
  6. [6]D. A. Broido, M. Malorny, G. Birner, N. Mingo, D. Stewart, Intrinsic lattice thermal conductivity of semiconductors from first principles, Applied Physics Letters 91 (23) (2007) 231922.
  7. [7]K. Esfarjani, G. Chen, H. T. Stokes, Heat transport in silicon from first-principles calculations, Phys. Rev. B 84 (8) (2011) 085204. URL https://journals.aps.org/prb/abstract/10.1103/PhysRevB.84.085204DOI
  8. [8]W. Li, N. Mingo, L. Lindsay, D. A. Broido, D. A. Stewart, N. A. Katcho, Thermal conductivity of diamond nanowires from first principles, Physical Review B 85 (19) (2012) 195436. URL https://journals.aps.org/prb/abstract/10.1103/PhysRevB.85.195436
  9. [9]J. Carrete, B. Vermeersch, A. Katre, A. van Roekegehem, T. Wang, G. K. Madsen, N. Mingo, almaBTE : A solver of the space-time dependent Boltzmann transport equation for phonons in structured materials, Comput. Phys. Commun. 220 (2017) 351–362. doi:10.1016/j.cpc.2017.06.023.DOI
  10. [10]Z. Han, X. Yang, W. Li, T. Feng, X. Ruan, Fourphonon: An extension module to shengbte for computing four-phonon scattering rates and thermal conductivity, arXiv preprint arXiv:2104.04895 (2021). URL https://journals.aps.org/prb/abstract/10.1103/PhysRevB.93.045202
  11. [11]T. Feng, X. Ruan, Quantum mechanical prediction of four-phonon scattering rates and reduced thermal conductivity of solids, Physical Review B 93 (4) (2016) 04502.
  12. [12]A. Togo, L. Chaput, I. Tanaka, Distributions of phonon lifetimes in brillouin zones, Phys. Rev. B 91 (2015) 094306. doi:10.1103/PhysRevB.91.094306.DOI
  13. [13]L. Chaput, Direct solution to the linearized phonon boltzmann equation, Physical Review Letters 110 (26) (6 2013). doi:10.1103/PhysRevLett.110.265506.DOI
  14. [14]L. Paulatto, F. Mauri, M. Lazzeri, Anharmonic properties from a generalized third-order ab initio approach: Theory and applications to graphite and graphene, Physical Review B 87 (21) (2013) 214303. URL https://journals.aps.org/prb/abstract/10.1103/PhysRevB.87.214303
  15. [15]P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, et al., Quantum espresso: a modular and open-source software project for quantum simulations of materials, Journal of physics: Condensed matter 21 (39) (2009) 395502. URL https://iopscience.iop.org/article/10.1088/0953-8984/21/39/395502/metaDOI
  16. [16]L. Paulatto, D. Fournier, M. Marangolo, M. Eddrief, P. Atkinson, M. Calandra, Thermal conductivity of bi 2 se 3 from bulk to thin films: Theory and experiment, Physical Review B 101 (20) (2020) 205419. URL https://journals.aps.org/prb/abstract/10.1103/PhysRevB.101.205419DOI
  17. [17]M. Simoncelli, N. Marzari, F. Mauri, Unified theory of thermal transport in crystals and glasses, Nature Physics 15 (8) (2019) 809–813. URL https://www.nature.com/articles/s41567-019-0520-x G. Barbalinardo, Z. Chen, N. W. Lundgren, D. Donadio, Efficient anharmonic lattice dynamics calculations of thermal transport in crystalline and disordered solids, Journal of Applied Physics 128 (13) (2020) 135104. URL https://aip.scitation.org/doi/full/10.1063/5.0020443
  18. [19]L. Isaeva, G. Barbalinardo, D. Donadio, S. Baroni, Modeling heat transport in crystals and glasses from a unified lattice-dynamical approach, Nature communications 10 (1) (2019) 1–6. URL https://www.nature.com/articles/s41467-019-11572-4DOI
  19. [20]C. D. Landon, N. G. Hadjiconstantinou, Deviational simulation of phonon transport in graphene ribbons with ab initio scattering, J. Appl. Phys. 116 (16) (2014) 163502. URL https://aip.scitation.org/doi/full/10.1063/1.4898090DOI
  20. [21]J.-P. M. Péraud, N. G. Hadjiconstantinou, Efficient simulation of multidimensional phonon transport using energy-based variance-reduced monte carlo formulations, Physical Review B 84 (20) (2011) 205331. URL https://journals.aps.org/prb/abstract/10.1103/PhysRevB.84.205331DOI
  21. [22]A. Pathak, A. Pawnday, A. P. Roy, A. J. Aref, G. F. Dargush, D. Bansal, Mcbte: A variance-reduced monte carlo solution of the linearized boltzmann transport equation for phonons, Computer Physics Communications 265 (2021) 108003.DOI
  22. [23]G. Romano, Efficient calculations of the mode-resolvedab-initio thermal conductivity innanostructures, arXiv preprint arXiv:2002.08940 (2020). URL https://arxiv.org/abs/2002.08940
  23. [24]G. Romano, Openbte, https://github.com/romanodev/OpenBTE (2013).
  24. [25]G. Romano, Phonon transport in patterned two-dimensional materials from first principles, arXiv preprint arXiv:2002.08940 (2020). URL https://arxiv.org/abs/2002.08940
  25. [26]C. Hua, A. J. Minnich, Transport regimes in quasiballistic heat conduction, Physical Review B 89 (9) (2014) 94302. URL https://journals.aps.org/prb/abstract/10.1103/PhysRevB.89.094302
  26. [27]J. Y. Murthy, S. Mathur, Numerical methods in heat, mass, and momentum transfer, School of Mechanical Engineering Purdue University (2002). URL https://engineering.purdue.edu/~scalo/menu/teaching/me608/ME608_Notes_Murthy.pdf
  27. [28]S. A. Ali, G. Kollu, S. Mazumder, P. Sadayappan, A. Mittal, Large-scale parallel computation of the phonon boltzmann transport equation, International journal of thermal sciences 86 (2014) 341–351. URL https://www.sciencedirect.com/science/article/pii/S12900729140022233
  28. [29]C. R. Harris, K. J. Millman, S. J. van der Walt, R. Gommers, P. Virtanen, D. Cournapeau, E. Wieser, J. Taylor, S. Berg, N. J. Smith, et al., Array programming with numpy, Nature 585 (7825) (2020) 357–362.DOI
  29. [30]L. D. Dalcin, R. R. Paz, P. A. Kler, A. Cosimo, Parallel distributed computing using python, Adv. Water Resour. 34 (9) (2011) 1124–1139. URL https://www.sciencedirect.com/science/article/pii/S0309170811000777
  30. [31]C. Geuzaine, J.-F. Remacle, Gmsh: A 3-D finite element mesh generator with built-in pre-and post-processing facilities, Int. J. Numer. Meth. Eng. 79 (11) (2009) 1309–1331. URL http://onlinelibrary.wiley.com/doi/10.1002/nme.2579/fullDOI
  31. [32]J. Ahrens, B. Geveci, C. Law, Paraview: An end-user tool for large data visualization, The visualization handbook 717 (8) (2005). URL https://www.paraview.org/DOI
  32. [33]D. Hasselman, L. F. Johnson, Effective thermal conductivity of composites with interfacial thermal barrier resistance, J. Compos. Mater. 21 (6) (1987) 508–515. URL https://aip.scitation.org/doi/full/10.1063/1.4945776

Paper details

vlnc:W_9sxcrrwq
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

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