Mode-coupling formulation of heat transport in anharmonic materials

arXiv:2411.14949 · cond-mat.mtrl-sci · Submitted 2024-11-22 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.

Kai: We're looking at the paper: "Mode-coupling formulation of heat transport in anharmonic materials".

Mira: Here's the gist of the paper. Mode-coupling theory provides a formal justification for using temperature-dependent phonons to study heat transport in anharmonic crystals because it derives a theory based on the mode-coupling theory of anharmonic lattice dynamics and accounts for both diagonal and off-diagonal contributions with and without collective effects.

Kai: Theoretical Framework The paper derives a theory of heat transport in anharmonic crystals using the mode-coupling theory of anharmonic lattice dynamics starting from the Green-Kubo formula to develop a thermal conductivity tensor based on the system’s dynamical susceptibility or spectral function.

Mira: This formulation results in a thermal conductivity tensor that includes anharmonicity non-perturbatively, with both collective and coherent contributions.

Kai: The derivation begins by considering the crystalline system within the framework of the Born-Oppenheimer approximation, where ion dynamics are described by a Hamiltonian.

Title and authors: Kai: Let's talk about the title and who wrote this. "Mode-coupling formulation of heat transport in anharmonic materials". It immediately tells us they are focusing on how phonon interactions at a level beyond just simple harmonic vibrations are affecting how heat flows through the material.

Mira: They’re bringing in mode-coupling theory, which is a specific mathematical tool from lattice dynamics, to justify using temperature dependent phonons for studying heat transport in anharmonic crystals >

Lev: It’s interesting because standard perturbation theory often fails when you try to go beyond that simple harmonic approximation, and this paper claims this approach offers a formal justification for that combination >

Kai: So what does that mean in practice? It means they have a theoretical backbone to use when you're dealing with materials where the temperature changes the force constants themselves >

Mira: Right. The authors are aiming to derive a theory of heat transport in anharmonic crystals starting from the Green-Kubo formula, which is fundamental for connecting fluctuations to transport properties >

Lev: They’re using that formula to develop a thermal conductivity tensor based on the system's dynamical susceptibility or spectral function, which is how you get your information about the material's response >

Kai: So it’s not just a single number anymore; they are getting this whole tensor structure out of this dynamic response.

Mira: And they are aiming to construct a theory founded on phonon correlation functions, which is the core idea behind mode-coupling theory >

The paper's summary: Kai: To summarize, the paper outlines how they use mode-coupling theory of anharmonic lattice dynamics to derive a heat current operator consistent with that underlying theory >

Mira: They then partition this operator into a diagonal component and an off-diagonal contribution, which is what allows them to separate the thermal conductivity into kappa equals kappa d plus kappa nd >

Lev: So, the main result here is this split. It’s not just one number for heat transport; it's two different ways to look at it >

Kai: And they show that they can calculate the diagonal part, kappa d, using what they call the single mode approximation >

Mira: Which gives them a specific expression involving the product of spectral functions and group velocities and scattering matrix elements >

Lev: That expression is important because it connects the simplified picture to the more complex reality of multiple modes interacting at once >

Kai: So, in essence, they are showing that you can get a simplified result while still being rigorous about how those individual vibrations fit into the bigger picture >

The paper's improvements: Mira: Now let’s talk about the improvements they suggest for their methodology. They focus on making it more practical and efficient, specifically by reducing computational cost >

Lev: They mention a lot of work to increase efficiency here, like using linear algebra formulations of the scattering matrix elements and symmetry reduction techniques for irreducible triplets and quartets >

Kai: That sounds like they are trying to tackle that massive calculation load by simplifying the math needed for those high order interactions >

Mira: Plus, they’re implementing a Monte Carlo integration scheme to reduce computational cost by decoupling the integrations between thermal conductivity calculations and scattering processes >

Lev: That decoupling is what lets them achieve full convergence for complex unit cells and higher order anharmonic coupling, which is something that would be impossible otherwise in a standard setup >

Kai: So, they’re showing that these numerical tricks are essential to making this theory usable for real, complicated systems rather than just theoretical exercises >

Conclusion: Mira: So to wrap up the whole paper on "Mode-coupling formulation of heat transport in anharmonic materials", it provides a unified framework encompassing single mode, collective and off-diagonal contributions that cover a large range of systems from simple to complex crystals with low or high anharmonicity >

Kai: It confirms that this approach is valid for predicting transport properties across diverse material regimes based on the input system's anharmonicity measure >

Lev: From an engineering side, it shows us how to build simulations that can handle things like third and fourth order interactions efficiently without completely blowing up the computation time if you need to run real hardware >

Mira: The main caveat is that they noted that the Markovian limit is an approximation, where full frequency dependent phonon spectral functions are replaced by Lorentzians >

Lev: And another point is that they also neglected a part of the heat current operator involving the random force and memory kernel, which they suggest we need to quantify as anharmonicity increases >

Kai: So it’s a powerful tool for calculating transport properties, but with the understanding that we need to be careful about those approximations when pushing into extreme regimes >

Aloés Castellano, Ⅎ P. Alvarinhas Batista, Olle Hellman, Matthieu J. Verstraete

Nanomat group, Q-MAT research center and European Theoretical Spectroscopy Facility, Université de Liège · Department of Molecular Chemistry and Materials Science, Weizmann Institute of Science · ITP, Physics Department, Utrecht University

cond-mat.mtrl-sci

Submitted: 2024-11-22

Updated: 2025-01-30

Journal ref: Physical Review B 111, 094306 (2025)

DOI: 10.1103/PhysRevB.111.094306

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 80/100

The gist: Mode-coupling theory provides a formal justification for using temperature-dependent phonons to study heat transport in anharmonic crystals because it derives a theory based on the mode-coupling…

Key concepts

Mode-Coupling Theory
This formalism uses linear response theory to calculate correlated atomic motion beyond simple perturbation theory. It employs a Mori-Zwanzig projection scheme to describe mass-weighted displacement correlations, allowing for the inclusion of collective effects in lattice dynamics.
Heat Current Operator Partitioning
The heat current operator is split into a diagonal component (Jd) and an off-diagonal component (Jnd). This separation allows the thermal conductivity ($Ω) to be calculated as $Ω ≈ Ω d + Ω nd, where the diagonal part is calculated using a single mode approximation.
Temperature-Dependent Phonon Renormalization
Unlike standard methods that rely on Taylor expansions of the potential energy surface, this approach incorporates temperature-dependent phonon renormalization. This focuses directly on atom dynamics rather than the potential surface, ensuring accuracy even when anharmonicity is high and displacements are significant.
Computational Efficiency Strategies
To handle complex scattering processes efficiently, the paper introduces numerical acceleration techniques. These include linear algebra formulations for scattering matrix elements, symmetry reduction for triplets and quartets, and Monte-Carlo integration to decouple thermal conductivity calculations from scattering processes.

Terminology

Summary

Mode-coupling theory provides a formal justification for using temperature-dependent phonons to study heat transport in anharmonic crystals because it derives a theory based on the mode-coupling theory of anharmonic lattice dynamics and accounts for both diagonal and off-diagonal contributions with and without collective effects. p. 1

Theoretical Framework

The paper derives a theory of heat transport in anharmonic crystals using the mode-coupling theory of anharmonic lattice dynamics starting from the Green-Kubo formula to develop a thermal conductivity tensor based on the system’s dynamical susceptibility or spectral function. This formulation results in a thermal conductivity tensor that includes anharmonicity non-perturbatively, with both collective and coherent contributions. p. 1 The derivation begins by considering the crystalline system within the framework of the Born-Oppenheimer approximation, where ion dynamics are described by a Hamiltonian. p. 1

Mode-Coupling Formalism

The mode-coupling theory of anharmonic lattice dynamics uses linear response theory to calculate correlated motion beyond standard perturbation theory. This formalism aims to describe the mass-weighted displacement-displacement KCF. p. 1 The derivation involves a Mori-Zwanzig projection scheme, which introduces projection operators P and Q to project dynamical variables onto single displacements and the rest of the full dynamical variable space, respectively. p. 2 This leads to a generalized Langevin equation for the equation of motion of the displacement correlation function. p. 2

Heat Current Operators

The heat current operator is derived using the Green-Kubo formula, and in an electrically insulating solid without diffusion, it is expressed as J (t) = X i ⟨Ri⟩ E˙i(t). This operator can be partitioned into a diagonal component J d(t) and an off-diagonal contribution J nd(t), allowing the thermal conductivity to be separated into κ ≈κd + κnd. p. 4 The diagonal part is calculated using the single mode approximation, yielding κ d,SMA = 1 / V X qs v s q ⊗ v s q cs(q) 2ΓMs(q). p. 5

Computational Efficiency

The paper presents numerical strategies to increase the efficiency of thermal conductivity calculations, including a linear algebra formulation of the scattering matrix elements and symmetry reduction techniques for irreducible triplets and quartets. A Monte-Carlo integration scheme is implemented to reduce computational cost by decoupling the integrations between thermal conductivity and scattering processes. p. 1 This decoupling allows for significant acceleration, especially when considering fourth-order interactions in complex systems. p. 8

Material Applications

The formalism is applied to several systems spanning different regimes of anharmonicity and thermal conductivity mechanisms to demonstrate its universality. Examples include silicon, Li3ClO, and γ-AgI, where the results show good agreement with both experimental data and theoretical results from the literature. p. 7 The work highlights that fourth-order scattering becomes increasingly significant with rising temperatures in some systems. p. 8

Limitations

The authors stress two main approximations made in their work, namely the use of the Markovian limit, where full frequency-dependent phonon spectral functions are replaced by Lorentzians. The other main approximation is the neglect of a part of the heat current operator involving the random force and memory kernel. p. 4 The work provides a starting point to quantify these neglected terms, which are expected to be important as anharmonicity increases.

Conclusion

The final result for the κ tensor includes single mode, collective and off-diagonal contributions, making it valid for a large range of systems from simple to complex crystals with low or high anharmonicity. The work demonstrates the validity of the mode-coupling theory of anharmonic lattice dynamics to compute transport properties and the efficiency of its implementation. p. 1 The approach offers a unified framework encompassing collective, single-mode, and off-diagonal contributions to heat transport by phonons. p. 5

Acknowledgments

The authors gratefully thank Florian Knoop for reviewing the implementation and for suggestions to improve the manuscript. The authors acknowledge the Fonds de la Recherche Scientifique (FRS-FNRS Belgium) and Fonds Wetenschappelijk Onderzoek (FWO Belgium) for funding ARC project DREAMS. p. 0 MJV acknowledges funding by the Dutch Gravitation program “Materials for the Quantum Age” (QuMat, reg number 024.005.006) financed by the Dutch Ministry of Education, Culture and Science (OCW). p. 0

Appendix A

The first important relation is that between the Kubo correlation function and the generalized susceptibility, which stems from the fluctuation-dissipation theorem. p. 2 Other important quantum correlation functions are defined as G(t) = ⟨AB(t)⟩ θ(−t). p. 1

Appendix B

The decoupling of four-point Kubo correlation functions of the form (AB, C(t), D(t)) can be formally derived using a semiclassical approximation that neglects some of the coupling between the decoupled two-point function’s correlation. The resulting equation for Ys1s2(q1, q2, t) leads to the final result Ye (0) = Ke (0)−1Y. p. 4 p. 4

Appendix C

The Markovian approximation is founded on the assumption that the bath follows a dynamic on a much slower time scale than the dynamical variable, translating into taking the infinite time limit of the convolution in eq.(C2). This leads to the Markovian scattering rate ΓM s(q) = Γs(q, omegas(q)). p. 4

Appendix D

The scattering matrix of the main text is derived by computing the time integral of the four-point correlation function and is given by eq.(D10).

Appendix E

The central quantity in the phonon Boltzmann equation (PBE) is the phonon occupation ns(q, t). The PBE reduces to a matrix equation that allows solving for ρs(q) by inverting the scattering matrix. p. 5

Appendix F

For degenerate modes, a direct calculation with this formula is ill defined and we use instead degenerate perturbation theory. The diagonal component of the generalized group velocities for each of the degenerate modes are computed as the average of the eigenvalues λi of h. p. 4

Appendix G

The MLIP were constructed by fitting on ab initio data computed with DFT using the Abinit package. The agreement between DFT and MLIP energy, forces and stress for all the MLIP used in this work is excellent. p. 4

Appendix H

The approach-to-equilibrium molecular dynamics (AEMD) method directly applies Fourier’s law to estimate thermal conductivity. The results are extrapolated to the bulk limit using equation (H3). p. 4

Appendix I

The paper is organized as follows, in section II, after introducing the mode-coupling theory of anharmonic lattice dynamics, we derive a heat current operator which is consistent with the theory. Section IV presents our implementation of the theory in the TDEP package, focusing on approaches to reduce the computational cost. p. 1 Finally, in section V, we apply our formalism to several materials before concluding in section VI. p. 1

Section II

We consider a crystalline system within the framework of the Born-Oppenheimer approximation, where the system dynamics is described by the Hamiltonian H = X i P2 i 2Mi + V (R) (1). To incorporate quantum effects in the ionic motion, we utilize the quantum Liouvillian formalism. p. 1

Section III

Our final formulation of the thermal conductivity tensor is expressed as κ = κ d + κ nd, with κ d given by Eq. (48) and κ nd by Eq. (53). This result bears some resemblance to previous derivations based on Hardy’s formulation of the heat current operator. p. 5

Section IV

The generalized IFC are fit using linear least-squares on the forces, incorporating the symmetry reduction described in reference [20]. A key point of the implementation is the successive fitting of the IFC, meaning that each order is fit on the residual forces from the previous order. p. 7

Section V

The key components for computing the thermal conductivity tensor are the generalized IFC, which must be derived from MD, or PIMD for systems with significant nuclear quantum effects. The MLIP employed in this work uses the Moment Tensor Potential framework. p. 1

Section VI

In recent years, the limitations of the harmonic approximation have become more and more apparent, producing a surge in the use of temperature dependent phonon theories. The work provides a detailed derivation of the theory of thermal conductivity in the framework of the mode-coupling theory of anharmonic lattice dynamics. p. 0

Section VII

Due to the extreme computational cost incurred in the computation of scattering processes, we present numerical strategies to increase the efficiency of thermal conductivity calculations. These acceleration methods include reducing the cost of each scattering process through a linear algebra formulation and by reducing the number of processes to be computed through symmetry reduction and Monte-Carlo integration schemes. p. 1

Section VIII

The paper applies its formalism to systems going from low to high anharmonicity, demonstrating the validity of the mode-coupling theory of anharmonic lattice dynamics to compute transport properties. The results show good agreement with both experimental and theoretical results. p. 0

Section IX

The paper concludes by noting that the Markovian limit is an approximation, and the neglect of a part of the heat current operator involving the random force and memory kernel are important aspects to quantify. The work provides a starting point to quantify these terms. p. 0

If the definition of a classical correlation function is unambiguous, there exists an infinite number of way to define a quantum correlation function. The first important relation is that between the Kubo correlation function and the generalized susceptibility, which stems from the fluctuation-dissipation theorem.

The decoupling of four-point Kubo correlation functions can be formally derived using a semiclassical approximation that neglects some of the coupling between the decoupled two-point function’s correlation. The resulting equation for Ys1s2(q1, q2, t) leads to the final result Ye (0) = Ke (0)−1Y.

The PBE reduces to a matrix equation that allows solving for ρs(q) by inverting the scattering matrix.

Page 3

Fourier’s law asserts that heat transport is characterized by the thermal conductivity which is an intrinsic material property relating the temperature gradient and heat current.

Page 5

The main difficulty in employing eq.(19) or (21) lies in the a priori unknown expression of the memory kernel. p. 2

Page 6

We will begin with the diagonal part of κ, κ d = 1 / 4V kBT2 X q1q2 X s1s2 Ωs1(q1)Ωs2(q2)v s1 q1 ⊗ v s2 q2 Z ∞0 dt As1(q1)Bs1(q1), As2(q2, t)Bs2(q2, t).

Page 7

A key distinction of our approach is its inherent inclusion of temperature-dependent phonon renormalization, setting it apart from standard formalisms that rely on a Taylor expansion of the Born-Oppenheimer potential energy surface.

Page 8

In contrast, the mode-coupling theory focuses directly on the atom dynamics, rather than on the underlying potential energy surface.

Page 9

The diagonal contribution in the single mode approximation is recovered when removing the restriction of s1 and s2 being different.

Page 10

The most time-consuming part of computing thermal conductivity is the calculation of the scattering matrix elements, which are needed for a large number of triplets or quartets of q-points and modes.

Page 12

Silicon is typically considered to exhibit low anharmonicity, and perturbation theory has been shown to accurately reproduce its transport properties, at least below room temperature.

Page 55

Compared to the approach in involving group velocity differences [37, 46, 48, 49], our approach respect the symmetries of the scattering matrix and allow to keep it symmetric positive definite.

Page 60

The diagonal contribution in the single mode approximation is recovered as κ d,SMA = 1 / V X qs v s q ⊗ v s q cs(q) 2ΓMs(q).

Page 65

This iterative approach is analogous to Omini’s solution to the phonon Boltzmann equation.

Page 70

Our findings demonstrate good agreement with both experimental and theoretical results from the literature.

Page 73

Recent studies [70] and also reproduced with our AEMD simulations, with which the mode-coupling theory agrees very well.

Page 80

Recent studies [70] and also reproduced with our AEMD simulations, with which the mode-coupling theory agrees very well.

Page 83

The pseudodojo: Training and grading a 85 element optimized norm-conserving pseudopotential table.

Page 86

The results are displayed in Fig. 7, our findings demonstrate good agreement with both experimental [69] and theoretical results from the literature [37, 70]. p. 7

Page 88

Our results show some discrepancy with previous results from the literature, both experimental [71] and theoretical [72, 73].

Page 90

The second system we studied is Li3ClO, an antiperovskite with a rich Lithium composition that makes it a candidate future generation electrolyte in solid-state batteries.

Page 87

Li3ClO can be considered as a system with medium anharmonicity, as can be attested by its σanh(T) going from 0.23 to 0.37 when increasing the temperature from 200 to 550 K. p. 0

For this system, we use a 24×24×24 full grid alongside an 8 × 8 × 8 Monte-Carlo grid for third-order scattering.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Mode coupling formulation of heat transport in anharmonic materials allows AI systems to predict thermal conductivity in complex materials by deriving a theoretically robust framework for predicting heat transport in complex materials. This enables AI to model heat flow through anharmonic crystals using a theory that includes both the diagonal and off-diagonal contributions of the heat current, with and without collective effects.

  2. Implementation strategies for TDEP package, including Monte Carlo integration schemes, allow AI systems to compute phonon scattering due to third- and fourth-order interactions achieving a substantial reduction in computational cost which enables full convergence of such calculations for the first time. This means AI can accurately calculate thermal transport even in complex unit cells with higher order anharmonic coupling.

  3. The framework allows for the computation of thermal conductivity tensors that include single mode, collective and off-diagonal contributions, making it valid for a large range of systems, from simple to complex crystals with low or high anharmonicity. This capability lets AI predict heat transport across diverse material regimes based on the input system's anharmonicity measure.

  4. The use of machine learning interatomic potentials as surrogates for ab initio Born-Oppenheimer surfaces enables AI systems to perform molecular dynamics simulations efficiently, as the MLIP is trained using a self-consistent approach and yields an optimal MLIP according to the Kullback-Leibler divergence. This allows for accurate calculation of generalized interatomic force constants required for thermal conductivity.

  5. The application of AEMD, which uses Fourier's law and estimates temperature evolution via equation H1, enables AI systems to compute thermal conductivity by relating it to the decay time τ where κ(L) = LCv / 4π2S1τ (H2). This allows for the prediction of thermal conductivity in a transient conduction regime.

  6. The framework can address limitations of perturbative theory by incorporating temperature-dependent phonon renormalization, setting it apart from standard formalisms that rely on a Taylor expansion, because the mode-coupling theory focuses directly on the atom dynamics, rather than on the underlying potential energy surface. This ensures accuracy even when assumptions like small displacements fail at elevated temperatures.

Related papers