betaFeedback

WorksPhysics and Astronomy

Gauge Invariance of the Thermal Conductivity in the Quantum Regime

Abstract

The widely used linear-response (LR) theory of thermal conduction in the quantum regime rests on the yet unproven assumption, that the thermal conductivity is invariant with respect to the gauge of the energy density of the system. This assumption manifests itself clearly in, e.g., Hardy's formulation of the heat-flux operator [Phys. Rev. 132, 168 (1963)]. In this work, we rigorously prove this assumption. Our proof, being valid for the nuclear and electronic subsystems, not only puts the state-of-the-art theory on solid grounds, but also enables going beyond the scope of the widely-used Boltzmann transport equation (BTE) within the LR framework for many-body systems.

Linear-response (LR) theory is one of the most important techniques of theoretical physics to calculate materials properties [1, 2] such as optical spectra [3, 4], electron-phonon coupling strength [5, 6], magnetic susceptibility [7] and thermal conductivity [8, 9, 10, 11]. The thermal conductivity of a material determines the amount of heat that flows through it in presence of a temperature gradient ∇T\nabla T. It is an important property for applications, such as thermographic defect detection [12], thermal-barrier coatings [13], and for thermoelectrics, that can tap waste heat to generate electricity [14].

In crystals, heat is carried by electrons and lattice vibrations (phonons). To calculate the lattice contribution, the state-of-the-art methodologies are the Green-Kubo (GK) [8, 15] and the Boltzmann transport equation (BTE) [16, 17, 18]. Despite recent advancements in the application of GK to real materials [19], using it as a general framework for calculating heat transport is still a formidable task [20]. The BTE, therefore, remains the workhorse for computing the lattice thermal conductivity. Similarly, the electronic contribution to the thermal conductivity can be calculated using the Kubo-Greenwood approach [21] or the Boltzmann transport equation [22] where due to the computational complexity of the former, likewise the BTE remains the most often used technique.

The assumption underlying the BTE is that the heat-carrying particles, i.e., phonons or electrons, act as in a gas, in which they do not scatter very often and behave almost like classical particles [17]. This enables one to treat them (almost) independently and assign them a defined long lifetime which leads to the relaxation-time approximation (RTA) [22]. Having lifetime and dispersion, one can calculate the respective contribution to the thermal conductivity of a material [23]. The BTE gives very good results for many materials [19, 24, 25, 26, 27, 28]. However, this method becomes questionable if the scattering rates of the particles are large [29], as, for instance, phonons in materials with low lattice thermal conductivity.

The BTE can be derived as an approximation to Kubo’s formula [8, 30] (Eq. (1) below) for electronic as well as lattice thermal transport [2, 31]. Therefore, it should be possible to derive correction terms that may answer the question when the BTE is applicable to a material, and if not, how to improve it. However, this approach faces a severe obstacle: For thermal transport, a temperature gradient is required. Therefore, to formulate the LR, small subsystems in local equilibrium are considered [32], which gives rise to the notion of a local energy density [2]. For interacting systems, this local energy density is, however, ambiguous: Whenever two or more particles interact, portions of the interaction energy can be arbitrarily assigned to any of them, or even to other points in space [10]. As a consequence, its definition involves the choice of a gauge. This is the case, for instance, for the interacting electrons and atomic nuclei responsible for the thermal conduction.

Even though it has been argued long ago that the LR approach is applicable also in the presence of a thermal disturbance (∇T≠0\nabla T \ne0) [8, 30], it remained unclear, whether the result is invariant with respect to the specific definition, i.e., the gauge, of the local energy density. Only for the classical case a proof has been given a few years ago [33]. In contrast, however, the many-body system of interacting particles that form any material, is a quantum problem. This is where our work comes in.

In this Letter, we prove the gauge invariance of the thermal conductivity in the quantum regime. In analogy to the classical counterpart [33], here we derive a quantum Einstein relation, which relates the thermal conductivity to the equal-time autocorrelation of a quantum energy-displacement operator. This enables us to leverage a generalized version of the Cauchy-Schwarz inequality to show that, for any two gauges, the thermal conductivity tensors must be identical. We then use Hardy’s formulation of the heat flux operator [10] to analyze the thermodynamic limit and illustrate the impact of the gauge. Finally, we show how the gauge invariance is fulfilled in a simple case of two-body interactions. This example motivates an alternative proof that is given at the end.

The lattice thermal conductivity is given as [2, 8, 32]

κ^ij(ω)=ΩT∫0βdλ∫0∞dt ei(ω+iη)t⟨S^iS^j(t+iℏλ)⟩(1)\hat{\kappa}^{ij}(\omega)=\frac{\Omega}{T}\int_{0}^{\beta}d\lambda\int_{0}^{\infty}dt\,e^{i(\omega+i\eta)t}\langle\hat{S}^{i}\hat{S}^{j}(t+i\hbar\lambda)\rangle \tag*{(1)}

where Ω\Omega is the system volume, TT the temperature, β=(kBT)−1\beta=(k_{\mathrm{B}}T)^{-1}, and kBk_{\mathrm{B}} the Boltzmann constant. η\eta is a positive infinitesimal. The operators are written in the Heisenberg picture. Angle brackets denote canonical thermal averages: ⟨O^⟩=(Tr⁡ρ^)−1Tr⁡(ρ^O^)\langle\hat{O}\rangle=(\operatorname{Tr}\hat{\rho})^{-1}\operatorname{Tr}(\hat{\rho}\hat{O}) with ρ^=exp⁡(−βH^)\hat{\rho}=\exp(-\beta\hat{H}) the density matrix, and H^\hat{H} the system hamiltonian in equilibrium. S^i\hat{S}^{i} are the cartesian components of the heat flux operator S^\hat{\mathbf{S}}. It can be obtained from the volume average of a local heat-flux operator s^(x)\hat{\mathbf{s}}(x) which fulfills the continuity equation [10]

1iℏ[h^(x),H^]=−∇⋅s^(x),(2)\frac{1}{i\hbar}[\hat{h}(x),\hat{H}]=-\nabla\cdot\hat{\mathbf{s}}(x), \tag*{(2)}

where h^(x)\hat{h}(x) is a hamiltonian density that is only constrained by the fact that its volume integral is H^\hat{H}. There remains a gauge freedom for h^(x)\hat{h}(x), since the interaction energy can be located at any of the involved particles or spatial coordinates. Multiplying Eq. (2) with xx and integrating over space, using partial integration, leads to

1iℏ[P^,H^]=−1Ω∫Ωdx x∇⋅s^(x)=S^.(3)\frac{1}{i\hbar}[\hat{P},\hat{H}]=-\frac{1}{\Omega}\int_{\Omega}dx\,x\nabla\cdot\hat{\mathbf{s}}(x)=\hat{\mathbf{S}}. \tag*{(3)}

Here,

P^=1Ω∫Ωdx xh^(x)(4)\hat{P}=\frac{1}{\Omega}\int_{\Omega}dx\,x\hat{h}(x) \tag*{(4)}

is the (gauge-dependent) energy polarization per unit volume. The surface term in Eq. (3) vanishes since the integrand xs^(x)x\hat{\mathbf{s}}(x) is zero outside the system [10]. From Eq. (4), it is clear that different choices of the gauge in h^(x)\hat{h}(x) lead in general to different definitions of S^\hat{\mathbf{S}}. While it is clear, that a physical observable is indeed independent of arbitrary mathematical gauge choices, the question remains whether the result of LR theory is also independent. In what follows, we rigorously prove that the LR result for κ^ij\hat{\kappa}^{ij} is indeed independent of the gauge of h^(x)\hat{h}(x).

We define the quantum energy displacement

D^i(t)=(ΩkB2)1/2∫−β/2β/2dλ∫0tdt′ S^i(t′−iℏλ).(5)\hat{D}^{i}(t)=\left(\frac{\Omega k_{\mathrm{B}}}{2}\right)^{1/2}\int_{-\beta/2}^{\beta/2}d\lambda\int_{0}^{t}dt'\,\hat{S}^{i}(t'-i\hbar\lambda). \tag*{(5)}

Now, we relate the thermal conductivity κ^ij\hat{\kappa}^{ij} to Re⁡⟨D^i(t)D^j(t)⟩\operatorname{Re}\langle\hat{D}^{i}(t)\hat{D}^{j}(t)\rangle. With the aid of [34]

∫cc+ydx1dx2 f(x1−x2)=∫−yydγ (y−∣γ∣)f(γ),(6)\int_{c}^{c+y}dx_{1}dx_{2}\,f(x_{1}-x_{2})=\int_{-y}^{y}d\gamma\,(y-|\gamma|)f(\gamma), \tag*{(6)}

where cc is some constant, we find

⟨D^i(t)D^j(t)⟩=ΩkB2∫−ββdϕ (β−∣ϕ∣)∫−ttdσ (t−∣σ∣)⟨S^i(σ−iℏϕ)S^j⟩.(7)\langle\hat{D}^{i}(t)\hat{D}^{j}(t)\rangle= \frac{\Omega k_{\mathrm{B}}}{2}\int_{-\beta}^{\beta}d\phi\,(\beta-|\phi|)\int_{-t}^{t}d\sigma\,(t-|\sigma|)\langle\hat{S}^{i}(\sigma-i\hbar\phi)\hat{S}^{j}\rangle. \tag*{(7)}

Applying the relation

lim⁡t→∞∫−ttdσ(1−∣σ∣t)exp⁡(iEσℏ)=2πℏδ(E)(8)\lim_{t\to\infty}\int_{-t}^{t}d\sigma\left(1-\frac{|\sigma|}{t}\right)\exp\left(i\frac{E\sigma}{\hbar}\right)=2\pi\hbar\delta(E) \tag*{(8)}

to Eq. (7) in the eigenfunction representation of H^\hat{H} and taking the limit t→∞t\to\infty, we arrive at

lim⁡t→∞1t⟨D^i(t)D^j(t)⟩=πℏΩkBβ2Tr⁡(ρ^)∑γ,ξe−βEγδ(Eγξ)S^γ,ξiS^ξ,γj,(9)\lim_{t\to\infty}\frac{1}{t}\langle\hat{D}^{i}(t)\hat{D}^{j}(t)\rangle= \frac{\pi\hbar\Omega k_{\mathrm{B}}\beta^{2}}{\operatorname{Tr}(\hat{\rho})} \sum_{\gamma,\xi}e^{-\beta E_{\gamma}}\delta(E_{\gamma\xi})\hat{S}^{i}_{\gamma,\xi}\hat{S}^{j}_{\xi,\gamma}, \tag*{(9)}

where S^γ,ξi\hat{S}^{i}_{\gamma,\xi} is a matrix element of the heat-flux operator between H^\hat{H}’s eigenstates ∣ξ⟩|\xi\rangle and ∣γ⟩|\gamma\rangle, Eξ/γE_{\xi/\gamma} are the corresponding eigenvalues, and Eγξ=Eγ−EξE_{\gamma\xi}=E_{\gamma}-E_{\xi}. This limit is identical to the static thermal conductivity κ^ij\hat{\kappa}^{ij}, as obtained from the real part of Eq. (1) in the ω→0\omega\to0, then η→0\eta\to0 limit. Therefore

lim⁡t→∞1tRe⁡⟨D^i(t)D^j(t)⟩=κ^ij.(10)\lim_{t\to\infty}\frac{1}{t}\operatorname{Re}\langle\hat{D}^{i}(t)\hat{D}^{j}(t)\rangle=\hat{\kappa}^{ij}. \tag*{(10)}

This is the quantum Einstein relation that we wanted to prove.

Now, only one more step is required: Assume there are two different heat fluxes, S^1i\hat{S}^{i}_{1} and S^2i\hat{S}^{i}_{2} (corresponding to two different gauges), causing two different conductivities, κ^1ii\hat{\kappa}^{ii}_{1} and κ^2ii\hat{\kappa}^{ii}_{2}, as obtained from Eq. (10) with D^1i\hat{D}^{i}_{1} and D^2i\hat{D}^{i}_{2}, respectively. We define Ψ^i=D^2i−D^1i\hat{\Psi}^{i}=\hat{D}^{i}_{2}-\hat{D}^{i}_{1}. Then

κ^12ii=lim⁡t→∞1tRe⁡⟨(D^1i(t)+D^2i(t))2⟩=κ^1ii+κ^2ii+lim⁡t→∞1tRe⁡⟨D^1i(t)D^2i(t)+D^2i(t)D^1i(t)⟩=2(κ^1ii+κ^2ii)−lim⁡t→∞1tRe⁡⟨Ψ^i(t)Ψ^i(t)⟩=2(κ^1ii+κ^2ii).(11)\begin{aligned} \hat{\kappa}^{ii}_{12}=\lim_{t\to\infty}\frac{1}{t}\operatorname{Re}\langle(\hat{D}^{i}_{1}(t)+\hat{D}^{i}_{2}(t))^{2}\rangle\\ &=\hat{\kappa}^{ii}_{1}+\hat{\kappa}^{ii}_{2}+\lim_{t\to\infty}\frac{1}{t}\operatorname{Re}\langle\hat{D}^{i}_{1}(t)\hat{D}^{i}_{2}(t)+\hat{D}^{i}_{2}(t)\hat{D}^{i}_{1}(t)\rangle\\ &=2\left(\hat{\kappa}^{ii}_{1}+\hat{\kappa}^{ii}_{2}\right)-\lim_{t\to\infty}\frac{1}{t}\operatorname{Re}\langle\hat{\Psi}^{i}(t)\hat{\Psi}^{i}(t)\rangle\\ &=2\left(\hat{\kappa}^{ii}_{1}+\hat{\kappa}^{ii}_{2}\right). \tag*{(11)} \end{aligned}

The last step is possible, because

lim⁡t→∞1t⟨Ψ^i(t)Ψ^i(t)⟩=0.(12)\lim_{t\to\infty}\frac{1}{t}\langle\hat{\Psi}^{i}(t)\hat{\Psi}^{i}(t)\rangle=0. \tag*{(12)}

We will elaborate on this point further below. Comparing the second with the fourth line, we find

lim⁡t→∞1t⟨D^1iD^2i+D^2iD^1i⟩=κ^1ii+κ^2ii.(13)\lim_{t\to\infty}\frac{1}{t}\langle\hat{D}^{i}_{1}\hat{D}^{i}_{2}+\hat{D}^{i}_{2}\hat{D}^{i}_{1}\rangle=\hat{\kappa}^{ii}_{1}+\hat{\kappa}^{ii}_{2}. \tag*{(13)}

Using the Cauchy-Schwarz inequality for hermitian operators A^\hat{A} and B^\hat{B}

4⟨A^2⟩⟨B^2⟩≥⟨A^B^+B^A^⟩2,(14)4\langle\hat{A}^{2}\rangle\langle\hat{B}^{2}\rangle\ge\langle\hat{A}\hat{B}+\hat{B}\hat{A}\rangle^{2}, \tag*{(14)}

we arrive at:

2κ^1iiκ^2ii≥κ^1ii+κ^2ii2\sqrt{\hat{\kappa}^{ii}_{1}\hat{\kappa}^{ii}_{2}}\ge\hat{\kappa}^{ii}_{1}+\hat{\kappa}^{ii}_{2}
⇒0≥(κ^1ii−κ^2ii)2.(15)\Rightarrow0\ge\left(\hat{\kappa}^{ii}_{1}-\hat{\kappa}^{ii}_{2}\right)^{2}. \tag*{(15)}

The invariance of the tensor κij\kappa^{ij} follows immediately from Eq. (15) which is true for all directions, not just the cartesian axes. Since the tensors κ1\kappa_1 and κ2\kappa_2 are real and symmetric [35], their difference θ=κ1−κ2\theta= \kappa_1 - \kappa_2 is also symmetric and has an eigenbasis. The representation of the tensor θ\theta in the coordinate system defined by this eigenbasis must be exactly zero, according to Eq. (15). As it vanishes in this basis, it vanishes in general. Therefore, κ1=κ2\kappa_1 = \kappa_2.

An important step in the above derivation is Eq. (12). We will now prove it. Replacing D^i\hat{D}^{i} with Ψ^i\hat{\Psi}^{i} in Eq. (7), one finds

lim⁡t→∞1t⟨Ψ^i(t)Ψ^i(t)⟩=πΩkBβ2ℏTr⁡(ρ^)∑γ,ξe−βEγδ(Eγξ)∣ΔP^γ,ξi∣2Eγξ2.(16)\lim_{t \to\infty} \frac{1}{t} \langle\hat{\Psi}^{i}(t)\hat{\Psi}^{i}(t) \rangle= \frac{\pi\Omega k_{\mathrm{B}}\beta^2}{\hbar\operatorname{Tr}(\hat{\rho})} \sum_{\gamma,\xi} e^{-\beta E_{\gamma}} \delta(E_{\gamma\xi}) \lvert\Delta\hat{P}^{i}_{\gamma,\xi} \rvert^2 \mathcal{E}_{\gamma\xi}^2 . \tag*{(16)}

Here, ΔP^γ,ξi\Delta\hat{P}^{i}_{\gamma,\xi} are matrix elements of the ii component of ΔP^=P^2−P^1\Delta\hat{P} = \hat{P}_2 - \hat{P}_1, i.e. the change in energy polarization P^\hat{P} due to the change of gauge. This quantity must be bounded for physical gauges. A way to understand this is to consider the relocation of a part of the energy δv^(x)\delta\hat{v}(\boldsymbol{x}) of a subsystem AA of volume ΩA\Omega_A by a constant vector c\boldsymbol{c}, such that the hamiltonian density is changed to h^′(x)=h^(x)+Δh^(x)\hat{h}'(\boldsymbol{x}) = \hat{h}(\boldsymbol{x}) + \Delta\hat{h}(\boldsymbol{x}),

Δh^(x)=ϑA(x−c)δv^(x−c)−ϑA(x)δv^(x),(17)\Delta\hat{h}(\boldsymbol{x}) = \vartheta_A(\boldsymbol{x} - \boldsymbol{c}) \delta\hat{v}(\boldsymbol{x} - \boldsymbol{c}) - \vartheta_A(\boldsymbol{x}) \delta\hat{v}(\boldsymbol{x}), \tag*{(17)}

where ϑA(x)\vartheta_A(\boldsymbol{x}) is an indicator function that is one in the volume ΩA\Omega_A and otherwise zero. Then, the change of polarization is given by

ΔP^=∫dx xΔh^(x)=cΩA∫ΩAdx δv^(x).(18)\Delta\hat{P} = \int dx\, \boldsymbol{x}\Delta\hat{h}(\boldsymbol{x}) = \frac{\boldsymbol{c}}{\Omega_A} \int_{\Omega_A} dx\, \delta\hat{v}(\boldsymbol{x}). \tag*{(18)}

Under the premises, that the relocation c\boldsymbol{c} of the energy should not be macroscopically large [32] and that the total energy of a subsystem should remain finite, ΔP^\Delta\hat{P} must be bounded. As ΔP^\Delta\hat{P} is bounded, Eq. (16) equals to zero. This confirms the validity of Eq. (12) and the proof is complete.

Before passing to an example, we discuss a few aspects concerning the thermodynamic limit and the heat flux operator. In the thermodynamic limit (Ω→∞\Omega\to\infty), the energy polarization as given by Eq. (4) is ill-defined, due to the presence of x\boldsymbol{x} in the integral. Thus, in this limit, no meaningful expression for κij\kappa^{ij} in terms of P^\hat{P} can be written, as would, e.g., be obtained by replacing S^\hat{S} in Eq. (9) by Eq. (3):

κij=πΩkBβ2ℏTr⁡(ρ^)∑γ,ξe−βEγδ(Eγξ)Eγξ2Re⁡(P^γ,ξiP^ξ,γj).(19)\kappa^{ij} = \frac{\pi\Omega k_{\mathrm{B}}\beta^2}{\hbar\operatorname{Tr}(\hat{\rho})} \sum_{\gamma,\xi} e^{-\beta E_{\gamma}} \delta(E_{\gamma\xi}) \mathcal{E}_{\gamma\xi}^2 \operatorname{Re}\left(\hat{P}^{i}_{\gamma,\xi}\hat{P}^{j}_{\xi,\gamma}\right). \tag*{(19)}

One may argue that a realistic system will never be of infinite size, and employ the above equation. In such a case, though, the term δ(Eγξ)Eγξ2\delta(E_{\gamma\xi})\mathcal{E}_{\gamma\xi}^2 would seemingly lead to zero conductivity. However, for a finite system the interaction with the environment must be taken into consideration. This interaction, which enables the transfer of energy through the material, leads to a broadening of δ(Eγξ)\delta(E_{\gamma\xi}) and thereby to a finite thermal conductivity.

In contrast, the analogous expression in terms of the heat flux operator, namely

κij=πℏΩkBβ2Tr⁡(ρ^)∑γ,ξe−βEγδ(Eγξ)Re⁡(S^γ,ξiS^ξ,γj),(20)\kappa^{ij} = \frac{\pi\hbar\Omega k_{\mathrm{B}}\beta^2}{\operatorname{Tr}(\hat{\rho})} \sum_{\gamma,\xi} e^{-\beta E_{\gamma}} \delta(E_{\gamma\xi}) \operatorname{Re}\left(\hat{S}^{i}_{\gamma,\xi}\hat{S}^{j}_{\xi,\gamma}\right), \tag*{(20)}

is well defined also in the thermodynamic limit. This can be understood from the general expression for S^\hat{S} as was found by Hardy [10]

S^=12Ω∑lp^lMlH^l−i2Ωℏ∑l,kr^lk[T^l,V^k]+h.c.(21)\hat{S} = \frac{1}{2\Omega} \sum_l \frac{\hat{p}_l}{M_l}\hat{H}_l - \frac{i}{2\Omega\hbar} \sum_{l,k} \hat{r}_{lk}\left[\hat{T}_l,\hat{V}_k\right] + \mathrm{h.c.} \tag*{(21)}

Here, H^l=T^l+V^l\hat{H}_l = \hat{T}_l + \hat{V}_l, T^l=p^l2/2Ml\hat{T}_l = \hat{p}_l^2/2M_l, r^lk=r^l−r^k\hat{r}_{lk} = \hat{r}_l - \hat{r}_k, p^l\hat{p}_l and r^l\hat{r}_l are the momentum and position operators of particle ll, MlM_l is its mass, and V^l\hat{V}_l its potential energy. As will be shown in the example below, the definition of the latter depends on the choice of the gauge, the only constraint being that the sum of the V^l\hat{V}_l must equal the potential-energy operator of the system, V^\hat{V}. The first term of Eq. (21) can be interpreted as the transport of heat carried by the particle flux, and is dominant if the particles form a gas, as in the case of electrons [10, 21, 32]. The second term can be interpreted as the rate of work that particle kk does on particle ll [32], and dominates the lattice transport. Similar (approximate) operators for the thermal current, can be found in literature, e.g., [30, 36].

The main reason why S^\hat{S}, as given by Eq. (21), can be employed in Eq. (20) in the thermodynamic limit, relies on the fact that it only contains position differences, r^lk\hat{r}_{lk}. This is in contrast to P^\hat{P}, which contains absolute positions, x\boldsymbol{x}. Moreover, analogous premises as given after Eq. (18) apply here: (i) No gauge choice should lead to divergent local energies H^l\hat{H}_l, and (ii) interactions should not have a macroscopic range, i.e., the energy cannot be relocated over macroscopic distances. More precisely, as the particle separation dd increases, the term r^lk[T^l,V^k]\hat{r}_{lk}[\hat{T}_l,\hat{V}_k] must vanish faster than 1/d31/d^3. This is the case, for instance, for the nuclei in non-polar insulators and most metals, for which the interatomic force constants decay as d−5d^{-5} or faster [37]. For polar insulators and semiconductors, the interatomic force constants decay like d−3d^{-3} but due to the total charge neutrality of the system, they oscillate, leading to a bounded heat-flux operator. This is more clear in reciprocal space [10] as the heat-flux operator is well-defined also in presence of LO-TO splitting [37].

We exemplify our proof with a system that has only two-body interactions between the particles. An analogous example was considered in Ref. [38] for the classical case. For this system, the local energy density h^(x)\hat{h}(x) can be written as

h^(x)=12∑l[δ(x−r^l)H^l(Γ)+H^l(Γ)δ(x−r^l)](22)\hat{h}(x) = \frac{1}{2}\sum_l \left[\delta(x-\hat{r}_l)\hat{H}_l(\Gamma)+\hat{H}_l(\Gamma)\delta(x-\hat{r}_l)\right] \tag*{(22)}

where H^l(Γ)\hat{H}_l(\Gamma) is the local energy of particle ll, given as H^l(Γ)=T^l+V^l(Γ)\hat{H}_l(\Gamma)=\hat{T}_l+\hat{V}_l(\Gamma), with

V^l(Γ)=12∑k≠lv^lk(1+Γlk).(23)\hat{V}_l(\Gamma)=\frac{1}{2}\sum_{k\ne l}\hat{v}_{lk}(1+\Gamma_{lk}). \tag*{(23)}

Here, v^lk\hat{v}_{lk} is the interaction operator between particles ll and kk that depends on their positions. The assumptions after Eq. (21) regarding the interaction range apply to v^lk\hat{v}_{lk} as well. Γ\Gamma is an arbitrary antisymmetric matrix, thus ∑lV^l(Γ)=V^\sum_l\hat{V}_l(\Gamma)=\hat{V}, independently of Γ\Gamma, as v^lk=v^kl\hat{v}_{lk}=\hat{v}_{kl}. In other words, Γ\Gamma parametrizes the gauge choice. For instance, if Γlk=sgn⁡(l−k)\Gamma_{lk}=\operatorname{sgn}(l-k), all the interaction energy between particles ll and kk is assigned to the particle with index min⁡(l,k)\min(l,k). By plugging Eq. (23) into Eq. (21), we obtain the change in the heat-flux operator due to the application of a gauge Γ\Gamma, i.e., ΔS^(Γ)=S^(Γ)−S^(Γ=0)\Delta\hat{S}(\Gamma)=\hat{S}(\Gamma)-\hat{S}(\Gamma=0). Then, using ΔS^(Γ)=dΔP^(Γ)/dt\Delta\hat{S}(\Gamma)=d\Delta\hat{P}(\Gamma)/dt (cf. Eq. (3)), we get the corresponding change in energy polarization (up to a time-independent constant),

ΔP^(Γ)=12Ω∑l,k>lr^lkv^lkΓlk+h.c.\Delta\hat{P}(\Gamma)=\frac{1}{2\Omega}\sum_{l,k>l}\hat{r}_{lk}\hat{v}_{lk}\Gamma_{lk}+\mathrm{h.c.}

Here, the spatial relocation of the energy is limited by the range of the interaction v^lk\hat{v}_{lk}, which plays a role analogous to the vector cc in Eq. (18). This result completes the example, since the energy polarization difference ΔP^(Γ)\Delta\hat{P}(\Gamma) is bounded, and, therefore, Eq. (12) is fulfilled (see Eq. (16)).

From this example, we note that, interestingly, the difference in κij\kappa^{ij} for two different definitions of S^\hat{S} contains at least one ΔS^\Delta\hat{S} term, as can be deduced from Eq. (20). Therefore, by using ΔS^=1iℏ[ΔP^,H^]\Delta\hat{S}=\frac{1}{i\hbar}[\Delta\hat{P},\hat{H}], the gauge related change in κij\kappa^{ij} will be

−iπΩkBβ2Tr⁡(ρ^)∑γ,ξRe⁡(ΔP^γξiS^ξγj)e−βEγδ(Eγξ)Eγξ=0.(24)-\frac{i\pi\Omega k_{\mathrm{B}}\beta^2}{\operatorname{Tr}(\hat{\rho})}\sum_{\gamma,\xi}\operatorname{Re}\left(\Delta\hat{P}^{i}_{\gamma\xi}\hat{S}^{j}_{\xi\gamma}\right)e^{-\beta E_\gamma}\delta(E_{\gamma\xi})E_{\gamma\xi}=0. \tag*{(24)}

which is zero because xδ(x)=0x\delta(x)=0 and both ΔP^\Delta\hat{P} and S^\hat{S} are well defined in the thermodynamic limit. This may be considered as a condensed version of our proof above.

To conclude, we have shown that the thermal conductivity is invariant of the specific choice of gauge in the energy density. This guarantees that one obtains its unique, true value in the realm of linear-response theory. With this, derivations of correction terms to the Boltzmann transport equation for thermal conductivity become possible, as any gauge of the energy density can be used. The correction terms, as proposed for example in Ref. [39], will enable one to check for which materials the Boltzmann transport equation can be used and for which it needs to be improved. This has been a longstanding, important issue for applied materials, such as thermoelectrics. Last but not least, we emphasize that, following our procedure, the gauge invariance for other quantities related to the heat-flux like the Seebeck coefficient, may be derived.

Acknowledgements

Work supported by the Studienstiftung des Deutschen Volkes. The authors thank Keith Gilmore, Nakib H. Protik and Pasquale Pavone for fruitful discussions.

References

  1. [1]R. Kubo, Statistical-mechanical theory of irreversible processes. i. general theory and simple applications to magnetic and conduction problems, Journal of the Physical Society of Japan 12, 570 (1957).DOI
  2. [2]G. D. Mahan, Many-particle physics (Springer Science & Business Media, 2013).
  3. [3]N. Wiser, Dielectric constant with local field effects included, Physical Review 129, 62 (1963).DOI
  4. [4]S. L. Adler, Quantum theory of the dielectric constant in real solids, Physical Review 126, 413 (1962).DOI
  5. [5]F. Giustino, Electron-phonon interactions from first principles, Rev. Mod. Phys. 89, 015003 (2017).DOI
  6. [6]S. Baroni, S. De Gironcoli, A. Dal Corso, and P. Giannozzi, Phonons and related crystal properties from density-functional perturbation theory, Rev. Mod. Phys. 73, 515 (2001).DOI
  7. [7]I. Lifshitz and A. Kosevich, Theory of magnetic susceptibility in metals at low temperatures, Sov. Phys. JETP 2, 636 (1956).
  8. [8]R. Kubo, M. Yokota, and S. Nakajima, Statistical-mechanical theory of irreversible processes. ii. response to thermal disturbance, J. Phys. Soc. Jpn. 12, 1203 (1957).DOI
  9. [9]S. de Groot and P. Mazur, Non-equilibrium thermodynamics (1962).DOI
  10. [10]R. J. Hardy, Energy-flux operator for a lattice, Phys. Rev. 132, 168 (1963).DOI
  11. [11]W. Nolting and W. D. Brewer, Fundamentals of many-body physics (Springer, 2008).DOI
  12. [12]D. P. Almond and S. G. Pickering, An analytical study of the pulsed thermography defect detection limit, Journal of applied physics 111, 093510 (2012).DOI
  13. [13]R. Valben, M. O. Jarligo, T. Steinke, D. E. Mack, and D. Stöwer, Overview of advanced thermal barrier coatings, Surface and Coatings Technology 205, 938 (2010).
  14. [14]G. J. Snyder and E. S. Toberer, Complex thermoelectric materials, in materials for sustainable energy: a collection of peer-reviewed research and review articles from Nature Publishing Group (World Scientific, 2011) pp. 101–110.
  15. [15]B. Qiu and X. Ruan, Molecular dynamics simulations of lattice thermal conductivity of bismuth telluride using two-body interatomic potentials, Physical Review B 80, 165203 (2009).
  16. [16]B. R. Nag, Electron transport in compound semiconductors, Springer series in solid-state sciences ; 11 (Springer-Verlag, Berlin, 1980).DOI
  17. [17]A. Altland and B. Simons, Condensed Matter Field Theory (Cambridge University Press, 2010).
  18. [18]D. A. Broido, M. Malorny, G. Birner, N. Mingo, and D. Stewart, Intrinsic lattice thermal conductivity of semiconductors from first principles, Applied Physics Letters 91, 231922 (2007).DOI
  19. [19]C. Carbogno, R. Ramprasad, and M. Scheffler, Ab initio green-kubo approach for the thermal conductivity of solids, Phys. Rev. Lett. 118, 175901 (2017).DOI
  20. [20]J. Chen, G. Zhang, and B. Li, How to improve the accuracy of equilibrium molecular dynamics for computation of thermal conductivity?, Phys. Lett. A 374, 2392 (2010).DOI
  21. [21]B. Holst, M. French, and R. Redmer, Electronic transport coefficients from ab initio simulations and application to dense liquid hydrogen, Phys. Rev. B 83, 235120 (2011).DOI
  22. [22]J. M. Ziman, Electrons and phonons: the theory of transport phenomena in solids (Oxford university press, 2001).
  23. [23]G. Grimmall, Thermophysical Properties of Materials, edited by E. Wohlfarth (Elsevier Science Publisher BV, 1986).
  24. [24]Y. Chen, D. Li, J. R. Lukes, and A. Majumdar, Monte carlo simulation of silicon nanowire thermal conductivity, J. Heat Transfer 127, 1129 (2005).DOI
  25. [25]D. Lacroix, K. Joulain, D. Terris, and D. Lemonnier, Monte carlo simulation of phonon confinement in silicon nanostructures: Application to the determination of the thermal conductivity of silicon nanowires, Appl. Phys. Lett. 89, 103104 (2006).DOI
  26. [26]F. Zhou, W. Nielson, Y. Xia, V. Ozoliņš, et al., Lattice anharmonicity and thermal conductivity from compressive sensing of first-principles calculations, Phys. Rev. Lett. 113, 185501 (2014).DOI
  27. [27]T. Tadano and S. Tsuneyuki, Self-consistent phonon calculations of lattice dynamical properties in cubic strio 3 with first-principles anharmonic force constants, Phys. Rev. B 92, 054301 (2015).
  28. [28]N. H. Protik and D. A. Broido, Coupled transport of phonons and carriers in semiconductors: A case study of n-doped gaas, Physical Review B 101, 075202 (2020).DOI
  29. [29]D. Pines, Theory of Quantum Liquids: Normal Fermi Liquids (CRC Press, 2018).
  30. [30]J. Luttinger, Theory of thermal transport coefficients, Physical Review 135, A1505 (1964).
  31. [31]W. C. Schieve and R. L. Peterson, Correlation function calculation of thermal conductivity, Phys. Rev. 126, 1458 (1962).DOI
  32. [32]P. B. Allen and J. L. Feldman, Thermal conductivity of disordered harmonic solids, Phys. Rev. B 48, 12581 (1993).pubmed.ncbi.nlm.nih.gov/10007626
  33. [33]A. Marcolongo, P. Umari, and S. Baroni, Microscopic theory and quantum simulation of atomic heat transport, Nat. Phys. 12, 80 (2016).DOI
  34. [34]E. Helfand, Transport coefficients from dissipation in a canonical ensemble, Phys. Rev. 119, 1 (1960).DOI
  35. [35]S. R. De Groot, Thermodynamics of irreversible processes, Vol. 242 (North-Holland Amsterdam, 1951).
  36. [36]V. Zlatić and R. Monnier, Modern Theory of Thermoelectricity (Oxford University Press, Oxford, 2014).DOI
  37. [37]X. Gonze and C. Lee, Dynamical matrices, born effective charges, dielectric permittivity tensors, and interatomic force constants from density-functional perturbation theory, Phys. Rev. B 55, 10355 (1997).DOI
  38. [38]S. Baroni, R. Bertossa, L. Ercole, F. Grasselli, and A. Marcolongo, Heat transport in insulators from ab initio green-kubo theory, in Handbook of Materials Modeling (Springer International Publishing, 2020) pp. 809–844.
  39. [39]L. Sham, Temperature propagation in anharmonic solids, Physical Review 163, 401 (1967).DOI

Paper details

vlnc:W_nnhbhmn4
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