betaFeedback

WorksPhysics and Astronomy

Theory of sound attenuation in amorphous solids from nonaffine motions

Abstract

We present a theoretical derivation of acoustic phonon damping in amorphous solids based on the nonaffine response formalism for the viscoelasticity of amorphous solids. The analytical theory takes into account the nonaffine displacements in transverse waves and is able to predict both the ubiquitous low-energy diffusive damping ∼k2\sim k^{2}, as well as a novel contribution to the Rayleigh damping ∼k4\sim k^{4} at higher wavevectors and the crossover between the two regimes observed experimentally. The coefficient of the diffusive term is proportional to the microscopic viscous (Langevin-type) damping in particle motion (which arises from anharmonicity), and to the nonaffine correction to the static shear modulus, whereas the Rayleigh damping emerges in the limit of low anharmonicity, consistent with previous observations and macroscopic models. Importantly, the k4k^4 Rayleigh contribution derived here does not arise from harmonic disorder or elastic heterogeneity effects and it is the dominant mechanism for sound attenuation in amorphous solids as recently suggested by molecular simulations.

Introduction

The mechanism of wave propagation, and, especially, wave attenuation in the disordered states of matter is a central and open topic in contemporary physics, with many ramifications in other fields from plasmas to cosmology. Since the mid 20th century it is well known that wave attenuation in the acoustic regime of low frequencies and wavelengths is dominated by anharmonic processes at the lowest wavevectors (the so-called hydrodynamic regime). In this regime, the sound attenuation, or damping, scales with the wavevector as Γ∼k2\Gamma\sim k^2. This is a diffusive law which can be derived by using conservation equations, i.e. with the methods of hydrodynamic theory and effective field theory [1].

In the context of crystals, this is also known as Akhiezer damping [2], whereas in liquids is known as Brillouin linewidth. The same phenomenon is known to dominate sound attenuation in amorphous solids, such as glasses, at low kk, where it plays an important role in determining the thermal conductivity [3]. More recently, the diffusive nature of vibrational excitations in glasses has been pointed out in numerical simulations [4, 5, 6] and used as the starting point for theoretical models of the boson peak caused by anharmonicity in glasses and crystals [7, 8].

In glasses [9], upon going to higher wavevectors, a crossover from ∼k2\sim k^2 to a ∼k4\sim k^4 regime is typically observed experimentally [10], where the ∼k4\sim k^4 scaling has been interpreted as Rayleigh-type scattering from random fluctuations of some (usually macroscopic) quantity. In this sense, the heterogeneous elasticity theory (HET) has provided a derivation of this Rayleigh type damping based on the assumption of Gaussian spatial fluctuations of the shear modulus [11]. This theory, however, is entirely at the continuum level (its starting point is the elastic modulus, which is a continuum quantity, while the modulus’ dependence on microphysics, particle displacements, interactions etc is neglected), hence it does not account for the microscopic structural order/disorder [12] nor for the underlying microscopic (nonaffine) particle dynamics [13, 14, 15].

Importantly, it has been recently showed using simulations [16] that the harmonic random fluctuations described by HET are not the only mechanism behind the ubiquitous Rayleigh term ∼k4\sim k^4. Nonaffine motions, which arise from the dynamics in non-centrosymmetric environments [12], contribute as well to the Rayleigh damping and they are indeed dominant with respect to the HET contribution, even at T=0T=0 [16]. This recent observation questions the current paradigm for sound attenuation in amorphous solids and calls for a deeper theoretical understanding of the dominant nonaffine origin of it.

While the current models of wave damping in amorphous solids are invariably at the continuum level (or at best effective medium theories [17]), much progress has been done recently in the microscopic theoretical description of the viscoelastic response of glasses, through the framework known as nonaffine response formalism or nonaffine lattice dynamics [18, 19, 20, 21, 22, 23]. This theory takes into account the fact that, upon deforming a disordered solid, the particles (atoms, molecules) undergo extra displacements on top of those dictated by the external strain, and these extra displacements, which are random, are called nonaffine motions. This framework provides quantitatively accurate predictions of the viscoelastic moduli of glasses with no fitting parameters in good agreement with simulations [23]. It is by now recognized that nonaffine motions play a central role in determining the dynamics of glasses at the microscopic level.

In this work, we present a theory of acoustic wave attenuation in amorphous solids based on nonaffine motions and we analytically predict the contributions from anharmonicity to the quadratic part of the damping Γ(k)\Gamma(k). We predict the hydrodynamic diffusive damping at low kk, including the prefactor which is related to important physical parameters such as the Debye frequency, the nonaffine correction to the shear modulus, and the microscopic friction due to anharmonicity. We provide a useful analytical expression for the Rayleigh contribution Γ∼k4\Gamma\sim k^4 arising from nonaffinity and identified in the simulations of [16] as the dominant effect on sound damping in amorphous systems. Finally, we also predict the experimentally observed [10] crossover from diffusive damping to Rayleigh damping at higher kk and we give an estimate of the critical wave-length in terms of the previously mentioned physical parameters.

Preliminaries

Within standard linear response theory, the time-dependent expectation value of the stress tensor σij(t)\sigma^{ij}(t) is given by a linear convolution

⟨σij(t)⟩≃∫−∞∞χσσijkl(t−t′)γkl(t′)+O(γ2).(1)\langle\sigma^{ij}(t) \rangle\simeq\int_{-\infty}^{\infty} \chi_{\sigma\sigma}^{ijkl}(t-t')\gamma^{kl}(t')+O(\gamma^2). \tag*{(1)}

with the strain tensor γij\gamma^{ij}, whose kernel is given by the dynamic response function (sometimes also labelled two-points function, correlator or Green function) χσσijkl(t)\chi_{\sigma\sigma}^{ijkl}(t), i.e. the stress auto-correlation function [24]. Neglecting dissipative terms, the zero frequency limit of the Fourier transformed response function is simply the elastic tensor given in terms of the various elastic constants. From now on, we will focus only on the shear response (ij=xyij=xy) and therefore Latin indices will be omitted to avoid clutter. Upon Fourier transforming, we have:

σ(ω)=χσσ(ω)γ(ω),(2)\sigma(\omega)=\chi_{\sigma\sigma}(\omega)\gamma(\omega), \tag*{(2)}

which is valid in the linear regime, to leading order in the external strain γ\gamma. Importantly, in general, χσσ(ω)\chi_{\sigma\sigma}(\omega) is a complex-valued function whose real and imaginary components encode respectively the reactive and dissipative parts of the complex dynamic modulus used in viscoelasticity theory. The real and imaginary components are related by the Kramers-Kronig relations [25] (imposed by causality) and in a simple viscoelastic system, to leading order in frequency, are given by (see e.g. Ref. [26]):

χσσ(ω)=μ+iωη+O(ω2)(3)\chi_{\sigma\sigma}(\omega)=\mu+i\omega\eta+O(\omega^2) \tag*{(3)}

where μ\mu is the static shear modulus and η\eta the shear viscosity.

Let us now consider a generic operator ϕ\phi and its conjugate external field δh\delta h. We define the dynamic response function χϕϕ\chi_{\phi\phi} associated to such operator as:

δ⟨ϕ⟩(ω,k)=χϕϕ(ω,k)δh(ω,k)(4)\delta\langle\phi\rangle(\omega,k)=\chi_{\phi\phi}(\omega,k)\delta h(\omega,k) \tag*{(4)}

where linear response is assumed as well as time/space translational invariance. The response function is complex-valued, χ=χ′+iχ′′\chi=\chi'+i\chi'', and the Kramers-Kronig relation holds:

χ′(ω)=P∫−∞∞χ′′(ω′)ω′−ωdω′π.(5)\chi'(\omega)=\mathcal{P}\int_{-\infty}^{\infty}\frac{\chi''(\omega')}{\omega'-\omega}\frac{\mathrm{d}\omega'}{\pi}. \tag*{(5)}

Let us consider the case of transverse phonons and elasticity. Following standard arguments (cfr. Section 7.3.1 in Ref. [24]), we obtain:

χuTuT(k,ω)=1−ρω2+μk2−iωηk2(6)\chi_{u_{\mathrm{T}}u_{\mathrm{T}}}(k,\omega)=\frac{1}{-\rho\omega^2+\mu k^2-i\omega\eta k^2} \tag*{(6)}

which can be generalized to:

χuTuT(k,ω)=1ρ1−ω2+Ω(k)2−iωΓ(k)(7)\chi_{u_{\mathrm{T}}u_{\mathrm{T}}}(k,\omega)=\frac{1}{\rho}\frac{1}{-\omega^2+\Omega(k)^2-i\omega\Gamma(k)} \tag*{(7)}

where:

Ω(k)2=vT2k2+…,Γ(k)=Dk2+…(8)\Omega(k)^2=v_{\mathrm{T}}^2k^2+\ldots,\qquad\Gamma(k)=Dk^2+\ldots \tag*{(8)}

and vT2=μ/ρv_{\mathrm{T}}^2=\mu/\rho and D=η/ρD=\eta/\rho as expected. Let us go back to (6) and try to expand it at low frequency. We immediately obtain:

χuTuT(k,ω)=1k2μ+iηωk2μ2+…(9)\chi_{u_{\mathrm{T}}u_{\mathrm{T}}}(k,\omega)=\frac{1}{k^2\mu}+i\frac{\eta\omega}{k^2\mu^2}+\ldots \tag*{(9)}

At ω=0\omega=0, this gives the static susceptibility:

χuTuT(k,ω=0)=1k2μ,(10)\chi_{u_{\mathrm{T}}u_{\mathrm{T}}}(k,\omega=0)=\frac{1}{k^2\mu}, \tag*{(10)}

as reported in Eq.6.4.24 in Ref. [24]. Now, comparing (9) with (3), one immediately realizes that the two response functions are related via:

χσTσT(k,ω)=μ2k2χuTuT(k,ω)(11)\chi_{\sigma_{\mathrm{T}}\sigma_{\mathrm{T}}}(k,\omega)=\mu^2 k^2\chi_{u_{\mathrm{T}}u_{\mathrm{T}}}(k,\omega) \tag*{(11)}

which will be explicitly used in the computations which follow (applied to the nonaffine part of the stress and the displacement).

Nonaffine motions

In solids, each particle is displaced to a position defined by the macroscopic strain tensor F\mathbf{F}, according to

ri=F(γ)⋅ri,0.\mathbf{r}_i=\mathbf{F}(\gamma)\cdot\mathbf{r}_{i,0}.

This position is called the affine position. Due to the structural disorder, since also the nearest neighbours of particle ii are being displaced in the local force field of interaction, the net force acting on ii in the affine position is not zero, due to the absence of inversion symmetry. This force vector is denoted as Ξi\boldsymbol{\Xi}_i and is called the affine force field since it represents the forces that trigger the nonaffine displacements [19]. It can be shown that the mean squared Ξi\boldsymbol{\Xi}_i is proportional to the mean squared nonaffine displacement, where the nonaffine displacement is defined as δrNA\delta\mathbf{r}_{\mathrm{NA}} in ri(γ)=F(γ)⋅ri,0+δrNA\mathbf{r}_i(\gamma)=\mathbf{F}(\gamma)\cdot\mathbf{r}_{i,0}+\delta\mathbf{r}_{\mathrm{NA}}, which gives the final position of the particle ii in the deformed frame.

An efficient way of representing the nonaffine displacements is as follows, i.e. by defining [19]:

ri=F(γ)⋅r˚i(γ)(12)\mathbf{r}_i=\mathbf{F}(\gamma)\cdot\mathring{\mathbf{r}}_i(\gamma) \tag*{(12)}

where the new variable r˚i\mathring{\mathbf{r}}_i does the book-keeping of the nonaffine displacements in the undeformed configuration. The ring notation thus indicates that the particle or material point coordinates are measured in the undeformed frame. In this formalism, we can write the affine force vector as follows [19]:

Ξi=−∂U∂r˚i∂γ∣γ→0.(13)\boldsymbol{\Xi}_i=-\left.\frac{\partial U}{\partial\mathring{\mathbf{r}}_i\partial\gamma}\right|_{\gamma\to0}. \tag*{(13)}

All this can be summarized into an equation of motion for the displacement xi≡r˚i(t)−r˚i(0)\mathbf{x}_i\equiv\mathring{\mathbf{r}}_i(t)-\mathring{\mathbf{r}}_i(0) of a particle ii of mass mm of the following form [19]:

mx¨i+ζx˙i+Hijxj=Ξiγ,(14)m\ddot{\mathbf{x}}_i+\zeta\dot{\mathbf{x}}_i+\mathbf{H}_{ij}\mathbf{x}_j=\boldsymbol{\Xi}_i\gamma, \tag*{(14)}

with inertial, dissipative and interaction force terms, respectively, on the left hand side and the affine-force field on the right side. Here, Hij\mathbf{H}_{ij} is the Hessian matrix that will be defined shortly below. For ease of notation, we used deformed-frame coordinates. This equation of motion, Eq. (14), can be derived from a microscopic reversible particle-bath Caldeira-Leggett Hamiltonian:

H=P22M+V(P)+12∑α=1N[pα2mα+mαωα2(Xα−Fα(Q)mαωα2)2](15)H=\frac{P^2}{2M}+V(P)+\frac{1}{2}\sum_{\alpha=1}^{N}\left[\frac{p_\alpha^2}{m_\alpha}+m_\alpha\omega_\alpha^2\left(X_\alpha-\frac{F_\alpha(Q)}{m_\alpha\omega_\alpha^2}\right)^2\right] \tag*{(15)}

where the dynamic variables {P,Q}\{P,Q\} with no subscript refer to the “particle” coupled to a bath of harmonic oscillators (the other particles in the system) labeled with subscript α\alpha (for a derivation see the Supplementary material of [23] and Ref. [27]). The particle is dynamically coupled to the α=1...N\alpha=1...N bath oscillators via a bilinear coupling with coefficients cαc_\alpha contained in Fα=cαQF_\alpha=c_\alpha Q. The bi-linear coupling between the particle and the oscillators serves as a model to represent the long-range anharmonic interactions between the tagged particle and the other particles in the system. Throughout the literature, the Caldeira-Leggett Hamiltonian has also been shown to provide a mapping onto anharmonic force-fields of real molecules in liquids [28]. Even though the mapping is only partially accurate, it is nevertheless an effective approximate way of accounting for anharmonicity. Importantly, it should also be noted that if the coupling coefficients cαc_\alpha are all identically zero, then also the viscosity of the system is identically zero, since the friction ζ\zeta in Eq. (14) can be shown to vanish identically, and hence also the low-frequency viscosity η=G′′ω\eta=G''\omega will vanish altogether [19]. This is because the friction ζ(t)\zeta(t) is proportional to ∑αcα2\sum_\alpha c_\alpha^2 as shown in [29]. Also, for better tractability we restrict our analysis to the Markovian limit, in which ζ\zeta is not a function of time. Finally, in the following, we shall take units in which m=1m=1.

Applying Fourier transformation xi(t)∼exp⁡iωtx_i(t)\sim\exp i\omega t and eigenmode decomposition of the Hessian matrix Hij\mathbf{H}_{ij} with respect to the eigenfrequencies νp\nu_p, the above equation of motion leads to the linear response theory, that was developed in [19], and to the following expression analogue of (2),

σ(ω)=Xσ(ω)γ(ω)=(μA+∑pΞ^p2ω2−νp2−iωζ)γ(ω)(16)\sigma(\omega)=X_\sigma(\omega)\gamma(\omega) =\left(\mu_A+\sum_p\frac{\widehat{\Xi}_p^2}{\omega^2-\nu_p^2-i\omega\zeta}\right)\gamma(\omega) \tag*{(16)}

where ζ\zeta is the (Langevin-type) microscopic damping coefficient for particle motion, and Ξ^p\widehat{\Xi}_p is the projection of the 3N3N-dimensional vector Ξ\boldsymbol{\Xi} onto the 3N3N-dimensional eigenvector of the Hessian matrix ∣p⟩|p\rangle, i.e. in Dirac’s bracket notation, Ξ^p=⟨Ξ∣p⟩\widehat{\Xi}_p=\langle\Xi|p\rangle. Furthermore, μA\mu_A is the affine shear elastic modulus, and νp\nu_p denotes the pp-th eigenfrequency of the solid, i.e. νp2=λp\nu_p^2=\lambda_p, where λp\lambda_p is the pp-th eigenvalue associated with eigenvector ∣p⟩|p\rangle. We also omitted prefactors with dimension of volume and assumed that particles masses are all equal to one. The Hessian matrix is defined as

Hij=∂U∂r˚i∂r˚j∣γ→0=∂U∂ri∂rj∣r→r0(17)\mathbf{H}_{ij}=\left.\frac{\partial U}{\partial\mathring{\mathbf{r}}_i\partial\mathring{\mathbf{r}}_j}\right|_{\gamma\to0} =\left.\frac{\partial U}{\partial\mathbf{r}_i\partial\mathbf{r}_j}\right|_{\mathbf{r}\to\mathbf{r}_0} \tag*{(17)}

since r˚(γ)∣γ→0=r0\mathring{\mathbf{r}}(\gamma)|_{\gamma\to0}=\mathbf{r}_0. It is a 3N×3N3N\times3N matrix with p=1,…,3Np=1,\ldots,3N eigenvalues λp\lambda_p and associated eigenvectors ∣p⟩|p\rangle. In (16) it is clear the existence of two contributions, one coming from affine displacements and encoded in the affine (infinite-frequency) shear modulus μA\mu_A, and a second term in bracket which arises from nonaffine motions and is controlled by the quantity Ξ^p2\widehat{\Xi}_p^2, i.e. the square of the affine force field in the basis of the eigenvectors of the Hessian. Also, the real part of this second term in bracket represents the (negative) nonaffine contribution to the shear modulus, and is often denoted as μNA\mu_{\mathrm{NA}} (or, with alternative notation, GNAG_{\mathrm{NA}}), defined such that

μ=μA−μNA(18)\mu=\mu_A-\mu_{\mathrm{NA}} \tag*{(18)}

is the total shear modulus containing both affine and nonaffine contributions. Importantly, the nonaffine term has potentially a dissipative (damping) component, as we are going to show next. More precisely, the imaginary part of the second term in the bracket of (16) is non-zero and it gives a finite contribution to the sound attenuation constant coming entirely from nonaffine motion.

Hence, we can identify a susceptibility associated with nonaffine motions, as follows:

χNA(ω)=∑pΞ^p2ω2−νp2−iωζ.(19)\chi_{NA}(\omega)=\sum_p\frac{\hat{\Xi}_p^2}{\omega^2-\nu_p^2-i\omega\zeta}. \tag*{(19)}

where from now on the σ\sigma labels will be omitted. Nonaffinity is not necessarily tied to anharmonicity. On the contrary, the atomic-scale friction ζ\zeta and the emergent viscosity associated to it vanish if the Caldeira-Leggett coupling coefficients cαc_\alpha of Fα=cαQF_\alpha=c_\alpha Q in Eq. (15), which mimic the long-range anharmonic interactions in the single-particle description, are all identically zero. As we will see later, the Rayleigh term in the damping does not depend on this coefficient (and therefore on anharmonicity), consistently with previous approaches from harmonic disorder/heterogeneous elasticity theory (HET) [30]. Nevertheless, and fundamentally, the induced diffusive (∼k2\sim k^2) contribution in the damping would not be present in the absence of anharmonicity, thus explaining the reason why in the HET harmonic-disorder approaches [11, 31] such a term must be included forcefully “by hand”.

We can now transform the discrete sum over eigenstates in (19) into a continuous integral over frequency, by introducing the vibrational density of states (VDOS) of the solid [19]. We thus obtain [32]:

χNA(ω)=3ρ∫0νDg(ν) ξ(ν)ω2−ν2−iωζ dν(20)\chi_{NA}(\omega)=3\rho\int_0^{\nu_D}\frac{g(\nu)\,\xi(\nu)}{\omega^2-\nu^2-i\omega\zeta}\,d\nu \tag*{(20)}

where ρ=N/V\rho=N/V is the particle density, νD\nu_D is the Debye frequency and, following [19], we have defined

ξ(ν)=⟨Ξ^p2⟩νp∈[ν+dν](21)\xi(\nu)=\left\langle\hat{\Xi}_p^2\right\rangle_{\nu_p\in[\nu+d\nu]} \tag*{(21)}

where the average is performed for all the projections of Ξ\boldsymbol{\Xi} on eigenvectors ∣p⟩\lvert p\rangle with eigenfrequency νp∈[ν+dν]\nu_p\in[\nu+d\nu]. Note that in previous literature ξ(ν)\xi(\nu) was denoted as Γ(ν)\Gamma(\nu) [19, 21], not to be confused with the sound attenuation constant discussed in this work. Note that in (20) ω\omega refers to the frequency of the external “mechanical” source (in linear response language) x(t)∼exp⁡iωtx(t)\sim\exp i\omega t, whereas ν\nu refers to the microscopic, internal eigenfrequencies of the solid that arise from diagonalization of the dynamical matrix Hij\mathbf{H}_{ij} (cfr. Eq. 42 in [19]).

In Ref. [21] it was shown that, for amorphous solids in dd space dimensions,

⟨Ξ∣p⟩⟨p∣Ξ⟩=dκR02λp∑αBα,xyxy(22)\langle\boldsymbol{\Xi}|p\rangle\langle p|\boldsymbol{\Xi}\rangle=d\kappa R_0^2\lambda_p\sum_\alpha B_{\alpha,xyxy} \tag*{(22)}

where α=x,y,z\alpha=x,y,z and Bα,xyxyB_{\alpha,xyxy} are coefficients that originate from angular averaging over bond orientation vectors. The scaling with λp\lambda_p has been verified in numerical simulations of different disordered and glassy systems in [23, 32]. Note that μNA=1V∑p⟨Ξ∣p⟩⟨p∣Ξ⟩λp\mu_{NA}=\frac{1}{V}\sum_p\frac{\langle\Xi|p\rangle\langle p|\Xi\rangle}{\lambda_p} [19, 21].

The above equation therefore implies that ξ(ν)∝λp∝ν2\xi(\nu)\propto\lambda_p\propto\nu^2 and consequently

χNA(ω)=c∫0νDν4ω2−ν2−iωζ dν(23)\chi_{NA}(\omega)=c\int_0^{\nu_D}\frac{\nu^4}{\omega^2-\nu^2-i\omega\zeta}\,d\nu \tag*{(23)}

where we used the Debye VDOS, g(ν)∼ν2g(\nu)\sim\nu^2, and hid all the pre-factors into an overall parameter cc with units Pa×s3\mathrm{Pa}\times\mathrm{s}^3 whose physical interpretation will be clearer in the following. The scaling ξ(ν)g(ν)∼ν4\xi(\nu)g(\nu)\sim\nu^4 has been successfully verified in the simulations of [16].

Sound propagation

Now that we have found the stress-stress autocorrelation function, we can use it to obtain the transverse displacement auto-correlation function χuTuT(ω,k)\chi_{u_Tu_T}(\omega,k) which in the rest of the manuscript will be indicated as C(ω,k)C(\omega,k), and that can be written as [24, 33, 34, 35]:

C(ω,k)=1m1−z2+E(k)2−Σ(ω,k)(24)C(\omega,k)=\frac{1}{m}\frac{1}{-z^2+E(k)^2-\Sigma(\omega,k)} \tag*{(24)}

where z≡ω+iϵz\equiv\omega+i\epsilon is the complex-valued frequency, E(k)E(k) is the bare phonon energy E(k)2=v2k2E(k)^2=v^2k^2 (with vv the Born speed of transverse sound in this case), Σ(ω,k)\Sigma(\omega,k) the self-energy and mm the mass density. Whether in the numerator one has 11 or 2z2z depends on the arbitrary choice of normalization of the bosonic operators in the Hamiltonian [36], here we follow the convention of Ziman [33]. By setting the self-energy to zero, and solving for the poles of the Green function in (24), we obtain ω=±vk\omega=\pm vk, the dispersion relation of the acoustic phonons in which the effects of nonaffinity are neglected. Again, explicit indices are not shown and we consider only the transverse phonons correlation function which corresponds to the shear stress response discussed before. Upon decomposing the self-energy into its real and imaginary parts,

Σ(ω,k)=Σ′(ω,k)+iΣ′′(ω,k),(25)\Sigma(\omega,k)=\Sigma'(\omega,k)+i\Sigma''(\omega,k), \tag*{(25)}

then the dispersion relation of the sound mode can be obtained as a solution of the following equation:

ω2=v2k2−Σ′(ω,k)−iΣ′′(ω,k)(26)\omega^2=v^2k^2-\Sigma'(\omega,k)-i\Sigma''(\omega,k) \tag*{(26)}

which must be compared with the general form:

ω2=Ω(k)2−iωΓ(k).(27)\omega^2=\Omega(k)^2-i\omega\Gamma(k). \tag*{(27)}

where Γ(k)\Gamma(k) is the sound attenuation [24]. A simple comparison implies

Ω(k)2=v2k2−Σ′(vk,k),Γ(k)=1vkΣ′′(vk,k),(28)\Omega(k)^2=v^2k^2-\Sigma'(vk,k),\qquad\Gamma(k)=\frac{1}{vk}\Sigma''(vk,k), \tag*{(28)}

where we have substituted ω→vk\omega\to vk assuming the limit of small wave-vector.

In other words, the expressions in (28) are valid only in the region where Re⁡ω=vk\operatorname{Re}\omega= vk. Equation (28) is in perfect agreement with the discussion in [16] and will serve as starting point to derive the sound damping based on the nonaffine formalism in the next sections.

The correction to the shear modulus and the nonaffine dynamic response function

Finally, as a consistency check of our analysis, we can study the nonaffine correction to the sound speed given by:

v′2=v2−Σ′(vk,k)k2=μA−μNAm,(29)v^{\prime2} = v^2 - \frac{\Sigma^{\prime}(vk,k)}{k^2} = \frac{\mu_A-\mu_{NA}}{m}, \tag*{(29)}

and check that it correctly reaches a constant value in the limit k→0k \to0. For that to be true, we need that:

lim⁡k→0Σ′(vk,k)∼k2.(30)\lim_{k \to0}\Sigma^{\prime}(vk,k) \sim k^2. \tag*{(30)}

Now:

Re⁡C(vk,k)=−1mΣ′(vk,k)Σ′(vk,k)2+Σ′′(vk,k)2(31)\operatorname{Re} C(vk,k) = -\frac{1}{m}\frac{\Sigma^{\prime}(vk,k)}{\Sigma^{\prime}(vk,k)^2+\Sigma^{\prime\prime}(vk,k)^2} \tag*{(31)}

and then:

Re⁡χNA(vk,k)∼−k2Σ′(vk,k)Σ′(vk,k)2+Σ′′(vk,k)2(32)\operatorname{Re}\chi_{NA}(vk,k) \sim-\frac{k^2\Sigma^{\prime}(vk,k)}{\Sigma^{\prime}(vk,k)^2+\Sigma^{\prime\prime}(vk,k)^2} \tag*{(32)}

where all the dimensionful parameters are omitted since not relevant for the purpose of this analysis. From (23), we see that in the limit of zero frequency (and therefore equivalently momentum), the real part of the nonaffine dynamic response function goes to a negative constant, Re⁡χNA(vk,k)∼−β\operatorname{Re}\chi_{NA}(vk,k) \sim-\beta, with β>0\beta> 0. Therefore, we have that:

lim⁡k→0k2Σ′(vk,k)Σ′(vk,k)2+Σ′′(vk,k)2∼β>0(33)\lim_{k \to0}\frac{k^2\Sigma^{\prime}(vk,k)}{\Sigma^{\prime}(vk,k)^2+\Sigma^{\prime\prime}(vk,k)^2} \sim\beta> 0 \tag*{(33)}

Now, we know that Σ′′∼k3\Sigma^{\prime\prime} \sim k^3, in the regime of small frequency/momentum, meaning that:

lim⁡k→0k2Σ′(vk,k)Σ′(vk,k)2+k6∼β>0(34)\lim_{k \to0}\frac{k^2\Sigma^{\prime}(vk,k)}{\Sigma^{\prime}(vk,k)^2+k^6} \sim\beta> 0 \tag*{(34)}

and therefore, expanding at small wave-vector, we get Σ′∼k2/β>0\Sigma^{\prime} \sim k^2/\beta> 0 which indeed gives a constant, and importantly a negative correction to the speed of sound and the static shear modulus (see (29)), in agreement with previous numerical calculations [16, 23, 32] and theoretical analysis [19, 20, 21].

The physical meaning of the c parameter

The previous analysis is helpful to re-write the dimensionful parameter cc appearing in the main text in terms of more transparent physical quantities. By restoring all the factors in the above formulas, one obtains that at low energy:

Re⁡χNA(vk,k)=−μNA2k2m1Σ′(vk,k).(35)\operatorname{Re}\chi_{NA}(vk,k) = -\frac{\mu_{NA}^2 k^2}{m}\frac{1}{\Sigma^{\prime}(vk,k)}. \tag*{(35)}

Moreover, by using (29), we can rewrite the real part of the self-energy as:

Σ′(vk,k)=μNAmk2+…(36)\Sigma^{\prime}(vk,k) = \frac{\mu_{NA}}{m}k^2+\ldots \tag*{(36)}

(36) together with (35) imply:

Re⁡χNA(vk,k)=−μNA(37)\operatorname{Re}\chi_{NA}(vk,k)=-\mu_{NA} \tag*{(37)}

which allows us to fix cc as:

c=3μNAvD3.(38)c=\frac{3\mu_{NA}}{v_D^3}. \tag*{(38)}

The result above will be useful to express the sound attenuation constant in terms of fundamental physical quantities.

Sound attenuation

By taking the imaginary part of the auto-correlation function in (24), we obtain:

Im⁡C(vk,k)=1mΣ′′(vk,k)Σ′′(vk,k)2+Σ′(vk,k)2(39)\operatorname{Im} C(vk,k)=\frac{1}{m}\frac{\Sigma^{\prime\prime}(vk,k)}{\Sigma^{\prime\prime}(vk,k)^2+\Sigma^{\prime}(vk,k)^2} \tag*{(39)}

where we also used the low-energy expression ω=vk\omega= vk. As demonstrated in [24] and in (11), a simple relation between the stress auto-correlation function χσσ\chi_{\sigma\sigma} and the displacements auto-correlation function χuu\chi_{uu} exists. Restricting it to the nonaffine components, that yields:

χσσNA(ω,k)=μNA2k2χuuNA(ω,k)=μNA2k2CNA(ω,k).(40)\chi_{\sigma\sigma}^{NA}(\omega,k)=\mu_{NA}^2 k^2\chi_{uu}^{NA}(\omega,k)=\mu_{NA}^2 k^2 C_{NA}(\omega,k). \tag*{(40)}

This (recall that the affine part of χ(ω)\chi(\omega) is independent of ω\omega [19]) finally implies:

Im⁡χNA(vk,k)=μNA2k2mΣ′′(vk,k)Σ′′(vk,k)2+Σ′(vk,k)2.(41)\operatorname{Im}\chi_{NA}(vk,k)=\frac{\mu_{NA}^2 k^2}{m}\frac{\Sigma^{\prime\prime}(vk,k)}{\Sigma^{\prime\prime}(vk,k)^2+\Sigma^{\prime}(vk,k)^2}. \tag*{(41)}

Now, using (28) together with (18), one can immediately verify that the real part of the self-energy at low momentum is given by Σ′(vk,k)=μNA/mk2\Sigma^{\prime}(vk,k)=\mu_{NA}/mk^2. Hence, using Σ′′=vkΓ\Sigma^{\prime\prime}=vk\Gamma, we get:

Im⁡χNA(vk,k)=Γ(k)kμNA2mvk2μNA2+Γ(k)2m2v2.(42)\operatorname{Im}\chi_{NA}(vk,k)=\frac{\Gamma(k)k\mu_{NA}^2mv}{k^2\mu_{NA}^2+\Gamma(k)^2m^2v^2}. \tag*{(42)}

Now, let us assume that Γ\Gamma scales faster than kk (as observed in all experimental and simulations results), therefore at low kk we have:

Im⁡χNA(vk,k)=mvΓ(k)k.(43)\operatorname{Im}\chi_{\mathrm{NA}}(vk,k)=mv\frac{\Gamma(k)}{k}. \tag*{(43)}

This clearly implies that a linear in kk term in the imaginary part of the self-energy will produce a diffusive ∼k2\sim k^2 in the damping, while a term k3k^3 will produce the Rayleigh damping ∼k4\sim k^4, as we are going now to verify by direct evaluation of the nonaffine theory developed above. We now recall (20) and perform the integral analytically, obtaining:

χNA(ω)=c∫0νDν4ω2−ν2−iωζ dν=−cνD33−cνDω(ω−iζ)+cω3/2(ω−iζ)3/2tanh⁡−1(νDω(ω−iζ)).(44)\chi_{\mathrm{NA}}(\omega)=c\int_{0}^{\nu_D}\frac{\nu^4}{\omega^2-\nu^2-i\omega\zeta}\,d\nu =-\frac{c\nu_D^3}{3}-c\nu_D\omega(\omega-i\zeta)+c\omega^{3/2}(\omega-i\zeta)^{3/2}\tanh^{-1}\left(\frac{\nu_D}{\sqrt{\omega(\omega-i\zeta)}}\right). \tag*{(44)}

Let us first discuss the behaviour at the lowest order in frequency. In that limit, the above integral becomes:

χNA(ω)=−cνD33+iζcνDω+…(45)\chi_{\mathrm{NA}}(\omega)=-\frac{c\nu_D^3}{3}+i\zeta c\nu_D\omega+\ldots \tag*{(45)}

implying that:

Γ(k)=cζνDmk2.(46)\Gamma(k)=c\frac{\zeta\nu_D}{m}k^2. \tag*{(46)}

This result can be simplified further by using (38), leading to:

Γ(k)=3ζμNAmνD2k2.(47)\Gamma(k)=\frac{3\zeta\mu_{\mathrm{NA}}}{m\nu_D^2}k^2. \tag*{(47)}

This is an important result: it shows that the coefficient for diffusive sound damping is proportional to the microscopic Langevin friction ζ\zeta, which in turn is related to anharmonicity, and it is also proportional to μNA\mu_{\mathrm{NA}}, meaning that in a perfect centrosymmetric crystal at zero temperature (hence with μNA=0\mu_{\mathrm{NA}}=0), the sound damping is identically zero. This is consistent with the expectation that sound damping is present only in anharmonic crystals with defects and/or thermal fluctuations (which also cause nonaffinity [37, 38, 39]), and in amorphous solids.

Despite the dependence of the leading quadratic term in Eq. (47) on the frictional parameter ζ\zeta might appear surprising, it is not. Indeed, it is well known that the diffusive damping in solids is proportional to the viscosity η\eta (see for example [24]), Γ∼ηk2\Gamma\sim\eta k^2. The latter grows with the internal friction parameter ζ\zeta leading to Γ∼ζk2\Gamma\sim\zeta k^2 as derived in Eq. (47). From a physical perspective, it is expected that a larger sound attenuation corresponds to larger internal friction, as mathematically derived in Eq. (47).

Eq. (47).

Let us emphasize, that the presence of a diffusive attenuation constant Γ∼Dk2\Gamma\sim Dk^2 for the collective excitations does not indicate dissipation of energy and/or momentum which are perfectly conserved quantities in the full system. This term is “dissipative” in the sense that it breaks time-reversal invariance (reversibility) and it is connected with entropy production. This is consistent with the discussion that the Reader can find in [40].

Hence, we find a contribution from nonaffine motions to the ubiquitous diffusive damping, which has been observed in countless experimental and simulation studies of glasses, liquids and supercooled liquids [5, 41, 42]. Importantly, the microscopic theory explains that the nonaffine contribution to the diffusion coefficient DD in Γdiff=Dk2\Gamma_{\mathrm{diff}}=Dk^2 is proportional to the microscopic damping coefficient for particle motion, D∝ζD\propto\zeta, which clarifies the close connection of the diffusive damping with anharmonicity of particle motion. The Langevin-type damping ζ\zeta can be shown, through particle-bath models of the Caldeira-Leggett type [43], to arise from long-range interparticle interactions, which in condensed matter states are due to the long-range anharmonic part of the interparticle potential [23, 28]. The most general derivation yields a non-Markovian friction coefficient ζ(t)\zeta(t) which, however, reduces to a Markovian, time-independent coefficient ζ\zeta, just as the one used here, when the coupling coefficients between each particle and the other oscillators are all the same [43, 44]. This suffices for the sake of our analysis and the more complex non-Markovian case is left for future studies.

In general, the expansion of the expression in (44) is complicated and not particularly illuminating, nevertheless it becomes rather simple in the limit of small microscopic friction, ζ≪1\zeta\ll1. The imaginary part of the tanh⁡−1\tanh^{-1} in the regime of ω≫ζ\omega\gg\zeta and ω≪νD\omega\ll\nu_D (small friction and low frequency) is approximately constant and equal to −π/2-\pi/2, leading to:

Im⁡χNA(ω)=cνDωζ+cπ2ω3+…(48)\operatorname{Im}\chi_{\mathrm{NA}}(\omega)=c\nu_D\omega\zeta+c\frac{\pi}{2}\omega^3+\ldots \tag*{(48)}

This result implies a smooth crossover between a diffusive damping at low wave-vector, Γ(k)∼k2\Gamma(k)\sim k^2, and a Rayleigh one, Γ(k)∼k4\Gamma(k)\sim k^4, at larger wave-vector:

Γ(k)=Γ2k2+Γ4k4.(49)\Gamma(k)=\Gamma_2k^2+\Gamma_4k^4. \tag*{(49)}

Importantly, the weight of the diffusive term at low wave-vector is controlled by the size of the microscopic Langevin friction ζ\zeta:

Γ2Γ4=2πνDζv2(50)\frac{\Gamma_2}{\Gamma_4}=\frac{2}{\pi}\frac{\nu_D\zeta}{v^2} \tag*{(50)}

and it vanishes in absence of microscopic friction, ζ=0\zeta=0. The behaviour of the imaginary part of the nonaffine stress correlation function is shown in Figure 1 for different strengths of the microscopic damping ζ\zeta. The full function displays a smooth crossover from a diffusive regime (Im⁡χ∼ω\operatorname{Im}\chi\sim\omega) to a Rayleigh regime (Im⁡χ∼ω3\operatorname{Im}\chi\sim\omega^3), whose location is controlled by the value of ζ\zeta. The larger the ζ\zeta, the more extended (and therefore more important) the diffusive regime. At very low values of the microscopic damping, the function is well approximated by (48) and the diffusive regime is pushed to very low frequency, whereas the Rayleigh one becomes more predominant.

Figure 1. The imaginary part of the nonaffine stress autocorrelation function as a function of the frequency ω\omega. The different colors represent different values of the microscopic damping for particle motion, ζ\zeta. The dashed line guides the eye towards the Rayleigh damping scaling. The different regimes of sound damping Γ(k)\Gamma(k) predicted by the theory (via Eq. (42)) are indicated. The dimensionful parameter cc is set to unity.

For completeness, the low frequency behaviour of the nonaffine stress autocorrelation function is shown in Figure 2 for different values of the damping ζ\zeta. Interestingly, the correction to the shear modulus at finite (but low) frequency decreases with the damping ζ\zeta, while the sound attenuation constant increases with it.

Figure 2. Real and imaginary parts of the nonaffine stress autocorrelation function at low frequency. Different colors represent different values of the microscopic damping ζ\zeta. The dimensionful coefficient cc is set to unity.

From our analytic expression, we can also derive, in the small damping limit ζ≪1\zeta\ll1, the contribution from nonaffinity to the Rayleigh ∼k4\sim k^4 term in the sound attenuation, the prefactor of which is given by:

Γ4=3π2v2μNAmνD3(51)\Gamma_4 = \frac{3\pi}{2}\frac{v^2\mu_{\mathrm{NA}}}{m\nu_D^3} \tag*{(51)}

where μNA\mu_{\mathrm{NA}} is the nonaffine part of the shear modulus and mm is the mass density. Interestingly, this contribution does not vanish in the limit ζ→0\zeta\to0 and it is therefore present even at T=0T = 0, as directly showed in the simulations of Ref. [16].

Finally, in the limit of small microscopic friction ζ\zeta, our theory provides an analytic estimate of the crossover point k∗k_* between the quadratic and quartic scaling in (49) given by:

k∗2=2νDζπv2.(52)k_*^2 = \frac{2\nu_D\zeta}{\pi v^2}. \tag*{(52)}

As evident from this expression, upon increasing the temperature, the microscopic friction parameter ζ\zeta grows and the quadratic diffusive regime extends up to larger values of the wave-vector kk. This behaviour is apparent in Figure 1.

Discussion

In summary, we presented an analytical theory of sound attenuation in amorphous solids starting from single particle motion and taking into account the inherently nonaffine dynamics. We derived in closed form the contributions from nonaffinity to both the hydrodynamic diffusive term ∼Γ2k2\sim\Gamma_2 k^2 (47) and the Rayleigh term ∼Γ4k4\sim\Gamma_4 k^4 (51), with the prefactors Γ2,Γ4\Gamma_2,\Gamma_4 expressed in explicit form as functions of important physical parameters. The Rayleigh attenuation due to nonaffine motions, (51), survives in the limit of zero temperature, or equivalently zero microscopic friction ζ=0\zeta= 0, and it has been recently shown via simulations to be the dominant process of sound attenuation in amorphous solids [16], while the harmonic disorder/HET contribution to the same scaling plays a comparatively very minor role. Furthermore, our theory analytically predicts the crossover from diffusive to Rayleigh damping at larger wavenumber (52), which was trivially obtained before only from ad-hoc combinations of continuum HET and anharmonic theory, and neglecting the microscopic nonaffine dynamics [31]. Crucially, Refs. [11, 31] are theories limited to the continuum level because their input are the fluctuations of the elastic moduli, whereas the micro-physics of the elastic moduli (i.e. how they are related to atomic-level displacements, bonding and atomic-level structure) is completely neglected. Here, instead we have did consider the microphysics of the moduli, i.e. their full eigenmode decomposition into microscopic atomic vibrations, and, crucially the atomic nonaffine displacements which prove key to demonstrate the origin of sound attenuation at the microscopic level.

The analytical prediction of Rayleigh damping from nonaffinity is fully supported by recent numerical data [16], which also prove its fundamental importance and dominance, with respect to the effects of harmonic disorder/HET, in the determination of sound attenuation in amorphous solids. Additionally, our main result in (47), namely the low-kk sound attenuation constant being quadratic in the frequency (or wave-vector) and increasing with the microscopic damping ζ\zeta, is confirmed by numerical simulations in [45] where the internal friction is parametrized in terms of the inelasticity.

Acknowledgments

We thank Giorgio Frangi, Bingyu Cui and Dmitry Parshin for reading an early version of the manuscript and for providing useful comments. We thank Lijin Wang, Grzegorz Szamel and Elijah Flenner for correspondence and useful comments which helped us to improve our manuscript. A.Z. acknowledges financial support from US Army Research Office, contract nr. W911NF-19-2-0055. M.B. acknowledges the support of the Shanghai Municipal Science and Technology Major Project (Grant No.2019SHZDZX01).

DATA AVAILABILITY

The data that supports the findings of this study are available within the article.

References

  1. [1]P. C. Martin, O. Parodi, and P. S. Pershan, “Unified hydrodynamic theory for crystals, liquid crystals, and normal fluids,” Phys. Rev. A 6, 2401–2420 (1972).DOI
  2. [2]A. I. Akhiezer, “On the absorption of sound in solids,” J. Phys. (Moscow) 1, 277 (1939).
  3. [3]J. M. Ziman, Electrons and Phonons: The Theory of Transport Phenomena in Solids, Oxford Classic Texts in the Physical Sciences (Oxford University Press, Oxford, 2001) p. 568.
  4. [4]Philip B. Allen, Joseph L. Feldman, Jaroslav Fabian, and Frederick Wooten, “Diffusons, locons and propagons: Character of atomie vibrations in amorphous si,” Philosophical Magazine B 79, 1715–1731 (1999).DOI
  5. [5]Hiroshi Shintani and Hajime Tanaka, “Universal link between the boson peak and transverse phonons in glass,” Nature Materials 7, 870–877 (2008).DOI
  6. [6]Y. M. Beltukov, V. I. Kozub, and D. A. Parshin, “Ioffe-regel criterion and diffusion of vibrations in random lattices,” Phys. Rev. B 87, 134203 (2013).DOI
  7. [7]Matteo Baggioli and Alessio Zaccone, “Universal origin of boson peak vibrational anomalies in ordered crystals and in amorphous materials,” Phys. Rev. Lett. 122, 145501 (2019).DOI
  8. [8]M. Baggioli and A. Zaccone, “Unified theory of vibrational spectra in hard amorphous materials,” Phys. Rev. Research 2, 013267 (2020).DOI
  9. [9]U. Buchenau, G. D’Angelo, G. Carini, X. Liu, and M. A. Ramos, “Sound absorption in glasses,” arXiv e-prints , arXiv:2012.10139 (2020), arXiv:2012.10139.arxiv.org/abs/2012.10139
  10. [10]C. Masciovecchio, G. Baldi, S. Caponi, L. Comez, S. Di Fonzo, D. Fioretto, A. Fontana, A. Gessini, S. C. Santucci, F. Sette, G. Viliani, P. Vilmercati, and G. Ruocco, “Evidence for a crossover in the frequency dependence of the acoustic attenuation in vitreous silica,” Phys. Rev. Lett. 87, 035501 (2006).pubmed.ncbi.nlm.nih.gov/16907508
  11. [11]W Schirmacher, “Thermal conductivity of glassy materials and the “boson peak”,” Europhysics Letters (EPL) 73, 892–898 (2006).DOI
  12. [12]R. Milkus and A. Zaccone, “Local inversion-symmetry breaking controls the boson peak in glasses and crystals,” Phys. Rev. B 93, 094204 (2016).
  13. [13]Christiane Caroli and Anaël Lemaître, “Fluctuating elasticity fails to capture anomalous sound scattering in amorphous solids,” Phys. Rev. Lett. 123, 055501 (2019).DOI
  14. [14]Christiane Caroli and Anaël Lemaître, “Key role of retardation and non-locality in sound propagation in amorphous solids as evidenced by a projection formalism,” The Journal of Chemical Physics 153, 144502 (2020), https://doi.org/10.1063/5.0019964.
  15. [15]Guan-Rong Huang, Bin Wu, Yangyang Wang, and Wei-Ren Chen, “Characterization of microscopic deformation through two-point spatial correlation functions,” Phys. Rev. E 97, 012605 (2018).DOI
  16. [16]Grzegorz Szamel and Elijah Flenner, “Sound attenuation in low temperature amorphous solids is primarily determined by non-affine displacements,” arXiv preprint arXiv:2107.14254 (2021).
  17. [17]Eric DeGiuli, Adrien Laversanne-Finot, Gustavo Düring, Edan Lerner, and Matthieu Wyart, “Effects of coordination and pressure on sound attenuation, boson peak and elasticity in amorphous solids,” Soft Matter 10, 5628–5644 (2014).arxiv.org/abs/1401.6563
  18. [18]B. A. DiDonna and T. C. Lubensky, “Nonaffine correlations in random elastic media,” Phys. Rev. E 72, 066619 (2005).DOI
  19. [19]A. Lemaître and C. Maloney, “Sum rules for the quasistatic and visco-elastic response of disordered solids at zero temperature,” Journal of Statistical Physics 123, 415 (2006).
  20. [20]C. E. Maloney and A. Lemaitre, “Amorphous systems in athermal, quasistatic shear,” Phys. Rev. E 74, 016118 (2006).pubmed.ncbi.nlm.nih.gov/16907162
  21. [21]A. Zaccone and E. Scossa-Romano, “Approximate analytical description of the nonaffine response of amorphous solids,” Phys. Rev. B 83, 184205 (2011).DOI
  22. [22]Shibu Saw, Sneha Abraham, and Peter Harrowell, “Non-affine displacements and the nonlinear response of a strained amorphous solid,” Phys. Rev. E 94, 022606 (2016).DOI
  23. [23]Vladimir V. Palyulin, Christopher Ness, Rico Milkus, Robert M. Elder, Timothy W. Sirk, and Alessio Zaccone, “Parameter-free predictions of the viscoelastic response of glassy polymers from non-affine lattice dynamics,” Soft Matter 14, 8475–8482 (2018).DOI
  24. [24]P. M. Chaikin and T. C. Lubensky, Principles of Condensed Matter Physics (Cambridge University Press, 1995).
  25. [25]John S. Toll, “Causality and the dispersion relation: Logical foundations,” Phys. Rev. 104, 1760–1770 (1956).DOI
  26. [26]Allen C Pipkin, Lectures on viscoelasticity theory, Vol. 7 (Springer Science & Business Media, 2012).
  27. [27]Bingyu Cui, Jie Yang, Jichao Qiao, Minqiang Jiang, Lanhong Dai, Yun-Jiang Wang, and Alessio Zaccone, “Atomic theory of viscoelastic response and memory effects in metallic glasses,” Phys. Rev. B 96, 094203 (2017).DOI
  28. [28]Alessandro Rognoni, Riccardo Conte, and Michele Ceotto, “Caldeira–Leggett model vs ab initio potential: A vibrational spectroscopy test of water solvation,” The Journal of Chemical Physics 154, 094106 (2021), https://doi.org/10.1063/5.0040494.
  29. [29]Robert Zwanzig, Nonequilibrium statistical mechanics (Oxford university press, 2001).DOI
  30. [30]Carl Ganter and Walter Schirmacher, “Rayleigh scattering, long-time tails, and the harmonic spectrum of topologically disordered systems,” Phys. Rev. B 82, 094205 (2010).DOI
  31. [31]Constantin Tomaras, Bernhard Schmid, and Walter Schirmacher, “Anharmonic elasticity theory for sound attenuation in disordered solids with fluctuating elastic constants,” Phys. Rev. B 81, 104206 (2010).DOI
  32. [32]Rico Milkus and Alessio Zaccone, “Atomic-scale origin of dynamic viscoelastic response and creep in disordered solids,” Phys. Rev. E 95, 023001 (2017).DOI
  33. [33]J. M. Ziman, Elements of Advanced Quantum Theory (Cambridge University Press, Cambridge, 1969).
  34. [34]Giacomo Baldi, Valentina M. Giordano, and Giulio Monaco, “Elastic anomalies at terahertz frequencies and excess density of vibrational states in silica glass,” Phys. Rev. B 83, 174203 (2011).DOI
  35. [35]Yong Xu, Jian-Sheng Wang, Wenhui Duan, Bing-Lin Gu, and Baowen Li, “Nonequilibrium green’s function method for phonon-phonon interactions and ballistic-diffusive thermal transport,” Phys. Rev. B 78, 224303 (2008).DOI
  36. [36]Daniel I. Khomskii, Basic Aspects of the Quantum Theory of Solids (Cambridge University Press, Cambridge, 2010).
  37. [37]Tamoghna Das, Surajit Sengupta, and Madan Rao, “Nonaffine heterogeneities and droplet fluctuations in an equilibrium crystalline solid,” Phys. Rev. E 82, 041115 (2010).pubmed.ncbi.nlm.nih.gov/21230246
  38. [38]Saswati Ganguly, Surajit Sengupta, Peter Sollich, and Madan Rao, “Nonaffine displacements in crystalline solids in the harmonic limit,” Phys. Rev. E 87, 042801 (2013).pubmed.ncbi.nlm.nih.gov/23679467
  39. [39]D.R. Squire, A.C. Holt, and W.G. Hoover, “Isothermal elastic constants for argon, theory and monte carlo calculations,” Physica 42, 388–397 (1969).DOI
  40. [40]V. V. Brazhkin and K. Trachenko, “Collective excitations and thermodynamics of disordered state: New insights into an old problem,” The Journal of Physical Chemistry B 118, 11417–11427 (2014), PMID: 25180672.
  41. [41]B. Ruffle, G. Guimbretiere, E. Courtens, R. Vacher, and G. Monaco, “Glass-specific behavior in the damping of acoustilike vibrations,” Phys. Rev. Lett. 96, 045502 (2006).pubmed.ncbi.nlm.nih.gov/16486840
  42. [42]J.P. Hansen and I.R. McDonald, Theory of Simple Liquids (Elsevier Science, 2006).
  43. [43]Robert Zwanzig, “Nonlinear generalized langevin equations,” Journal of Statistical Physics 9, 215–220 (1973).DOI
  44. [44]R. Zwanzig and R. D. Mountain, “High-frequency elastic moduli of simple fluids,” The Journal of Chemical Physics 43, 4464–4471 (1965).DOI
  45. [45]Kuniyasu Saitoh and Hideyuki Mizuno, “Sound damping in soft particle packings: the interplay between configurational disorder and inelasticity,” Soft Matter 17, 4204–4212 (2021).DOI

Paper details

vlnc:W_62rmfhnw
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