2 agrees with those of the phonon theory of liquid [7–10]. V x ( ) Compression of particle

II. MICROSCOPIC DYNAMICS IN LIQUIDS

In this section, we begin by reviewing in detail the qualitative picture of the microscopic dynamics of parti- cles in liquids explained in Refs. [7, 8, 10, 12]. Following this picture, we clarify the relationship between the dy- namics of particles in liquids and the dynamical modes x retained in the continuum mechanical equations, that is, the elastic and hydrodynamic equations. Difusion of particle Potential barrier In liquids, a particle experiences many-body interac- tions between neighboring particles due to the van der Waals force or the Coulomb force [1] and the mean free path is the same order of the particle spacing in solids [22], which results in the complicated dynamical corre- lations in liquids. According to Refs. [7–10, 12], under such an environment, a tagged particle undergoes the solid-like oscillatory motion in the effective potential bar- riers formed by the surrounding particles. However, the FIG. 1. A schematic pictures of the microscopic dynamics tagged particle is not permanently trapped in the poten- of particles in liquids. In the shorter time regime than τF, tial, and undertakes the hydrodynamic diffusive motion a tagged (green) particle oscillates in the potential barrier by which the tagged particle escapes from the potential formed by the surrounding (blue) particles. In the longer time regime than τ , on the other hand, the tagged particle moves well on average with a characteristic time scale τ . In F F to one of the neighboring local minima through the diffusive addition, the distances between local minima of the po- motion. In both of these time regimes, a density fluctuation tential also fluctuate as a consequence of the density fluc- of surrounding particles occurs. This mode is intuitively un- tuation of the surrounding particles. These motions are derstood as a change in the distances of local minima of the illustrated in Fig.1. The above picture is based on a sin- potential. gle characteristic relaxation time for liquids, which leads to viscoelastic model of liquids. We can identify τF as the Maxwell relaxation time τM = η/G∞, where η is the vis- cosity and G∞ is the high-frequency shear modulus. The mechanism of the elastic lattice vibration in the Debye dynamical property of liquids is explained by the com- model, which produces two transverse and one longitudi- bination of these modes, and the value of τF decides the nal modes of phonons. Therefore, the dynamics of parti- relative weight of each of these two contributions, which cles clearly obeys the elastic wave equations as is a function of temperature. In order to make our dis-  2  cussion clearer, we summarize the dynamics of particles 1 ∂ 2 2 2 − ∇ ~um = 0, (1) in each frequency regime separated by the characteristic cs ∂t frequency ωF = 1/τF: where cs is the sound speed, m = L or T, L and T being 1. In the higher frequency regime ω > ωF, which we the longitudinal and transverse modes. In the hydrody- refer to as the elastic regime, the tagged particle os- namic regime, it is important to microscopically under- cillates in the potential formed by the surrounding stand the dynamical modes retained in the hydrodynamic particles. equations. To simplify the problem, we adopt the ideal fluid. In hydrodynamics, the equation of motion of the 2. In the lower frequency ω < ωF, which we refer to ideal fluid is expressed by the Euler equation as the hydrodynamic regime, the tagged particle escapes from the potential well via the hydrody- ∂ 1 ~v + (~v · ∇)~v = − ∇p, (2) namic diffusive motion, and the surrounding parti- ∂t ρ cles show the density fluctuation as collective mo- tions. and auxiliary conditions such as the equation of continu- ity of density Now we propose the quasiparticle description of these motions. The result presented in the following is basi- ∂ ρ + ~v · ∇ρ + ρ∇ · ~v = 0, (3) cally same as the phonon theory of liquid [8–10], but dif- ∂t ferent from their theory especially in the interpretation of the hydrodynamic regime. In the elastic regime, the where ~v is the fluid velocity, ρ is the fluid density and p motion of particles are correctly captured by the same is the pressure. If we assume an isentropic flow, we can 3 utilize a simple thermodynamical relation among the de- 3.2 viations of enthalpy h per a mass, pressure p and density Experiment ρ from their equilibrium values [20, 23] as 3.0 Theory 1 c2 s 2.8

∆h = ∆p = ∆ρ, (4) B ρ ρ Nk / V In the following, the variables stand for the deviations C 2.6 from their equilibrium values. By using Eq. (4) and linearizing Eqs. (2) and (3) with respect to ~v and h, we 2.4 finally obtain the linearized Euler equation as 2.2 ∂ 250 300 350 400 450 500 550 600 h + c2∇ · ~v = 0, (5) ∂t s T (K) ∂ ~v + ∇h = 0. (6) ∂t FIG. 2. Heat capacity of liquid mercury per atom as a func- tion of temperature. The red curve is the experimental data In Eqs. (5) and (6), there exist two dynamical indepen- in Ref. [25]. The blue curve is calculated by using Eq.(10). dent modes: the one is the enthalpy field h and the other The set of parameters used here is based on [7]. The devia- is the velocity potential Φ, which is defined by ~v = ∇Φ tion the experimental data from the theoretical prediction is [24]. The vorticity is now conserved through Eq. (6). within the experimental error shown in Ref. [25]. The enthalpy field reflects the density fluctuation via Eq. (4), which corresponds to the fluctuation of distances of local minima of the potential. The velocity potential, on the other hand, represents the diffusive flow of particles we obtain from the partition function the total energy as in and out of the local region, which corresponds to the  αT  diffusion motion among the local minima of potential. hHˆ i =NkBT 1 + These quasiparticles can be associated with a gauge field 2 ! in U(1) gauge theory in the next section.  ω 3 × 3D(x ) − F D(x ) , (10) In the above discussion, we clarified the number of D ω F modes in each regime: three modes in the elastic regime D and two modes in the hydrodynamic regime. This fact enables us to immediately guess a form of the partition where D(x) is the Debye function function in each regime: 3 Z x y3dy D(x) ≡ 3 y , (11) 1. In the elastic regime, x 0 e − 1

 3 N is the number of particles in the system and α is

e Y −βhω¯ ~ ~ the thermal expansion coefficient. The derivation of Eq. Ztotal =  e k  for k s.t. ωF < ω~ < ωD. (7) k (10) is explained in Supplemental Material. This ex- ~k pression depends on the ratio ωF/ωD, which reduces to ωF/ωD ≈ 0 at the melting temperature, and also reduces 2. In the hydrodynamic regime, to ωF/ωD ≈ 1 at the critical temperature by taking into account the temperature dependence of the Frenkel fre-  2 quency [7–10]. The heat capacity evaluated from Eq. Y h −βhω¯ ~k ~ Ztotal =  e  for k s.t. 0 < ω~k < ωF. (8) (10) quantitatively agree with the experimental date as ~k shown in Fig. (2). Although the total energy obtained from our discussion is the same as in the phonon theory Here, we define thath ¯ is the reduced Planck constant, of liquid [7–10], the key idea in our approach is that the β = k T is the inverse temperature, T is the tempera- heat capacity reflects the change in the number of quasi- B particles in each regime, which reproduces the proper ture, k is the Boltzmann constant, ~k is the wave number, B behavior that interpolates between the number of modes ω is the Debye frequency, and ¯hω =hc ¯ |~k| is the en- D ~k s at the melting point and at that of the critical point. ~ ergy of the mode specified by k. By taking into account The decrease in heat capacity from 3NkB to 2NkB when the temperature dependence of the Frenkel frequency ωF, lowering the temperature from the melting point to the which is related to that of viscosity through the relation critical point can be associated with the reduction of the number of quasiparticles from 3N to 2N modes. By un- G∞ derstanding the dynamical modes via quasiparticles, we ωF = ωM = , (9) η can derive the total energy much easier than the phonon 4 theory of liquid [7–10]. We can also appropriately incor- This gauge-fixing condition corresponds to the conser- porate the contribution from hydrodynamic diffusion ne- vation of the vorticity in the hydrodynamics. We give glected in their theory [7–10]. This treatment, however, a detailed explanation for this gauge-fixing condition in is not based on the single Hamiltonian or Lagrangian, Supplemental Material. We also define ~v⊥ as which a physical system is expected to fundamentally  2  possess. In the following section, we present a possi- ∂ 1 ∂ 2 ~v⊥ ≡ ac − ∇ A,~ (19) ble way to construct a unified single Lagrangian which ∂t c2 ∂t2 reproduces the number of quasiparticles at the melting temperature and at the critical point, and to interpolate and we finally obtain the following equations: between them. ∂ h + c2∇ · ~v = 0, (20) ∂t ∂ III. LAGRANGIAN THEORY FOR LIQUIDS ~v + ∇h = ~0, (21) ∂t

For the purpose mentioned in the last section, we where ~v ≡ ~vk +~v⊥. These equations are obviously equiva- demonstrate in this section that the hydrodynamic Euler lent to Eqs. (5) and (6) by replacing c with cs. Therefore, equation is equivalent to U(1) gauge theory with a spe- we have establish the equivalence between hydrodynam- cific gauge-fixing condition. We define a hydrodynamic ics and U(1) gauge theory with a specific gauge-fixing Lagrangian based on this equivalence. This Lagrangian condition. From this equivalence, we can immediately constitutes the basis of our theory for heat capacity of define the hydrodynamic Lagrangian by Eq. (12) with liquids. In addition, we add to the hydrodynamic La- Eq. (17). This hydrodynamic Lagrangian, however, can- grangian a neutral scalar field to incorporate the elastic not be applied to evaluate the thermodynamical property contribution explained in the last section. of liquids because the elastic behavior of liquids in high We start by considering the dynamics of the electro- frequency regime, which was explained in the last section, magnetic fields through the U(1) gauge Lagrangian is not incorporated yet. To recover the contribution from the phonons in high frequency regime, we add the neu- 1 µν LA = − Fµν F , (12) tral scalar field φ with its mass mF, and we propose a 4 Lagrangian for liquids as follows: where F = ∂ A − ∂ A . In the present paper, we de- µν µ ν ν µ 2 2 µ ν µ 0 ~ 1 µ mFcs 2 fine x = (ct, ~x), xµ = ηµν x ,A = (A , A) and adopt a L = LA + ∂µφ∂ φ − φ , (22) µν ν 2 2 notation η = ηµν = diag(1, −1, −1, −1),Aµ = ηµν A , 2¯h where Greek indices run over space-time coordinates and 2 where mF ≡ ¯hωF/cs . The equations of motion derived the repeated greek index is summed over the space-time from this Lagrangian are expressed as coordinates 0, 1, 2, 3. The Euler-Lagrange equation for this Lagrangian leads to the Maxwell’s equations in vac- µν µ 2 2 2
∂µF = 0, ∂µ∂ + mFcs /¯h φ = 0. (23) uum 0 i 1 ∂ By adopting the Coulomb gauge, A = 0 and ∂iA = 0, ∇ · E~ = 0, ∇ × B~ − E~ = 0, (13) c ∂t the equations of motion are rewritten as where  1 ∂2  − ∇2 Ai = 0, (24) c2 ∂t2 ~ 1 ∂ ~ 0 ~ ~ s E = − A − ∇A , B = ∇ × A, (14)  2 2 2  c ∂t 1 ∂ 2 mFcs 2 2 − ∇ + 2 φ = 0. (25) and c is the speed of light. By combining Eqs. (13) and cs ∂t ¯h (14), we obtain In the high frequency regime, that is, the high energy ∂ regime m c2  ¯hω, we can neglect the mass term of the h + c2∇ · ~v = 0, (15) F s ∂t k scalar field φ. Here, we define  2  ∂ 1 ∂ 2 i ~v + ∇h + ac − ∇ A~ = 0, (16) (uL)i ∝ ∂iφ, (uT)i ∝ A . (26) ∂t k c2 ∂t2 where It is noted that the constant with appropriate dimension in Eq.(26) can be recovered without difficulty. We finally ~ 0 h ≡ ac∇ · A, ~vk ≡ a∇A . (17) obtain the following equations: where a is the constant with appropriate dimension.  1 ∂2   1 ∂2  − ∇2 ~u = 0, − ∇2 ~u = 0, (27) Now, we introduce a following gauge-fixing condition: 2 2 L 2 2 T cs ∂t cs ∂t  2  1 ∂ 2 ∇ · − ∇ A~ = 0. (18) where ∇ · ~uT = 0 and ∇ × ~uL = 0 are satisfied. Con- c2 ∂t2 sequently, Eq. (23) becomes the wave equations for the 5 longitudinal mode ~uL and the transverse modes ~uT. On nates from the quasiparticle description of liquids that 2 the other hand, in the low energy regime ¯hω  mFcs , the specifies the number of modes at the melting tempera- scalar field φ becomes dynamically ineffective, because ture and at the critical temperature. A unified treatment 2 2 2 Eq. (25) can be approximated as (mFcs /¯h ) · φ ' 0, of the hydrodynamic and the elastic behaviors with a cor- which results in the recovery of hydrodynamics in this rect interpolation between them is realized through the energy regime by adopting the gauge-fixing condition, gauge symmetry which allows the natural transformation Eq. (18). It is noted that the Lagrangian, Eq. (22), of the dynamical equation into each regime. We con- can be transformed from the elastic regime to the hydro- structed a general, thermodynamically accessible model dynamic regime by choosing the gauge-fixing condition, for liquids, which incorporates the qualitative behaviors which is the consequence of a gauge symmetry in this the- at the melting point and at the critical point with natural ory. Therefore, this Lagrangian appropriately reproduces interpolation. This is as if the Debye model properly pre- the change in the number of modes depending on temper- dicts the T 3-law at the lower temperature limit, and the ature between the melting point and the critical point. Dulong-Petit law at the higher temperature limit with The relative weight of each mode in the Lagrangian can interpolation between two limiting cases. be now controlled by the mass mF, which decides whether the field φ is dynamically effective or not depending on the temperature. IV. CONCLUSIONS Now, we discuss heat capacity of liquids based on the Lagrangian, Eq. (22) . This Lagrangian allows us to eval- In the present paper, we propose a thermodynamical uate the partition function by means of a similar proce- model for evaluating the heat capacity of liquids. We sug- dure to the Debye model or finite-temperature field the- gest a microscopic interpretation of the hydrodynamic ory in statistical physics [23, 26]. As a result, we obtain equations and introduce a quasiparticle picture for the a unified partition function dynamics of particles in liquids. With the aid of a newly proposed equivalence between hydrodynamics and U(1)  2   gauge theory, this interpretation allows us to construct Y −βEA Y −βEφ Ztotal =  e ~k   e ~k  , (28) a hydrodynamic Lagrangian whose thermodynamic prop- ~k ~k erty is easily calculated by appropriate procedures in sta- tistical mechanics. By taking the elastic behavior in liq- A uids into account by means of a neutral scalar field, we where we define two energies of the quasiparticles, E~ = q k also propose a new Lagrangian for liquids as a possible ~ φ ~ 2 2 2 cs¯h|k| and E = (cs¯hk) + (mFc ) . According to this way to calculate the heat capacity of liquids. It should ~k s partition function, we obtain the internal energy of liq- be emphasized that the gauge symmetry of our theory uids in equilibrium, which approximately reproduces Eq. enables us to establish a unified treatment of the hydro- (10) up to the first order in β¯hωF. We give a detailed dynamic and the elastic equations, which appropriately derivation of the heat capacity from the partition func- reproduces the number of modes at the melting point tion, Eq. (28), in Supplemental Material. and at the critical point. Our theory can be regarded as As mentioned above, the expression obtained by our the counterpart in liquids to the Debye model in solids. Lagrangian qualitatively reproduces the continuous de- In the future study, we will apply our theory to other crease in the heat capacity from 3NkB to 2NkB, and thermodynamical properties of liquids. each limit corresponds to the number of quasiparticles at the melting point and at the critical point, respectively. Our theoretical standpoint is the hydrodynamic equation ACKNOWLEDGEMENT as opposed to the phonon theory of liquid [7–10], which approaches the liquid phase from the solid phase. Our We would like to thank Toshihiro Kawakatsu, Motoi theory, however, does not need any estimation for the Endo, Naoto Kan, Kiyoharu Kawana, Ryota Kojima, contributions to the total energy from each mode [7–10]. Katsumasa Nakayama, Hikaru Ohta, Sayuri Takatori, A clear derivation of the heat capacity crucially origi- and Sumito Yokoo for helpful discussions.

[1] P. A. Egelstaff, An introduction to the liquid state, 2nd [5] B. Sadigh and G. Grimvall, Phys. Rev. B 54, 15742 ed. (Clarendon Press, London, 1992). (1996). [2] J.-P. Hansen and I. R. McDonald, Theory of Simple Liq- [6] M. Forsblom and G. Grimvall, Phys. Rev. B 72, 132204 uids (Academic Press, London, 1986). (2005). [3] T. E. Faber, Introduction to the Theory of Liquid Metals [7] K. Trachenko, Phys. Rev. B 78, 104201 (2008). (1972). [8] D. Bolmatov and K. Trachenko, Phys. Rev. B 84, 054106 [4] G. Grimvall, Phys. Scr. 11, 381 (1975). (2011). 6

[9] D. Bolmatov, V. V. Brazhkin, and K. Trachenko, Sci. Rep. 2, 421 (2012). [10] K. Trachenko and V. V. Brazhkin, Rep. Progr. Phys. 79, 016502 (2016). [11] Y. Heo, M. Antoaneta Bratescu, D. Aburaya, and N. Saito, Appl. Phys. Lett. 104, 111902 (2014). [12] J. Frenkel, Kinetic Theory of Liquids (Clarendon Press, Oxford, 1946). [13] A. Altland and B. D. Simons, Condensed matter field theory (Cambridge University Press, 2010). [14] T. Egami, V. Levashov, R. S. Aga, and J. R. Morris, Mater. Trans. 48, 1729 (2007). [15] T. Iwashita, D. M. Nicholson, and T. Egami, Phys. Rev. Lett. 110, 205504 (2013). [16] J. V. Bellissard, arXiv:1708.06624v1 (2017). [17] J. Bellissard and T. Egami, Phys. Rev. E 98, 063005 (2018). [18] H. Marmanis, Phys. Fluids 10, 1428 (1998). [19] A. C. R. Mendes, C. Neves, W. Oliveira, and F. Takakura, Braz. J. Phys. 33, 346 (2003). [20] T. Kambe, Fluid Dyn. Res. 42, 055502 (2010). [21] E. M. C. Abreu, J. A. Neto, A. C. R. Mendes, and N. Sasaki, Phys. Rev. D 91, 125011 (2015). [22] K. Takagi and K. Negishi, J. Chem. Phys. 72, 1809 (1980). [23] L. D. Landau and E. M. Lifshitz, Statistical Physics (Pergamon Press, Oxford, 1969). [24] L. D. Landau and E. M. Lifshitz, Fluid Mechanics (Perg- amon Press, Oxford, 1959). [25] D. C. Wallace, Phys. Rev. E 57, 1717 (1998). [26] J. I. Kapusta and C. Gale, Finite-Temperature Field The- ory: Principles and Applications, 2nd ed., Cambridge Monographs on Mathematical Physics (Cambridge Uni- versity Press, 2006). 7

SUPPLEMENTAL MATERIAL FOR “HEAT CAPACITY OF LIQUIDS IN LIGHT OF HYDRODYNAMICS AS U(1) GAUGE THEORY”

This supplemental material provides (I) the equivalence between hydrodynamics and U(1) gauge theory and (II) the evaluation of the total energy of liquids.

(I)The equivalence between hydrodynamics and U(1) gauge theory.— In this section, we first discuss the possibility of our gauge-fixing condition. In general, gauge vectors can be expressed as

A~ = ∇χA + ∇ × X~ A. (29) In the equivalence between hydrodynamics and U(1) gauge theory, we adopt following gauge-fixing condition:

µ ∂µ∂ χA = 0. (30)

µ 0µ Even if ∂µ∂ χA 6= 0 is satisfied, we can obtain a suitable gauge vector A as

~0 0 ~ µ 0 A = ∇χA + ∇ × XA, ∂µ∂ χA = 0, (31) where we define as

0 χA = χA − α, (32) by the gauge transformation: Aµ → A0µ = Aµ + ∂µα. (33) Therefore, for a given gauge vector Aµ, we can always choose as Z 3 ~ ik·x α(x) = χA(x) + d kα˜(k)e , (34) whereα ˜ is an arbitrary function of ~k and we define kµ = (c|~k|,~k). This means that our gauge-fixing is possible at all times. Then, we obtain following relations:   µ 0 µ ∂µ∂ A~ = ∇ × ∂µ∂ X~ A , (35) Z ~0 0 3 ~ 2 ~ ik·x ∇ · A = ∇ · ∇χA = d k k α˜(k)e , (36) which is generally non-zero. Thus, we define following quantities:

h ≡ ac∇ · A~0, (37) 00 ~vk ≡ a∇A , (38) ∂ ~v ≡ ac∂ ∂µA~0, (39) ∂t ⊥ µ where ∇ × ~vk = 0 and ∇ · ~v⊥ = 0 are satisfied. As the result, we can obtain the equations of the motion of the hydrodynamics from the equation of motion of the U(1) gauge theory. (II)The evaluation of the total energy of liquids.— Let us evaluate the total energy of liquids. We consider the total energy of liquids for partition functions, Eq.(7) and Eq.(8). From these partition function, the energy of liquids in higher frequency regime is expressed as follows: We consider the total energy of liquids for partition functions, Eq.(7) and Eq.(8).

Z ωD 2 ˆ 3V ω ¯hω hHi (ωF < ω < ωD) = E0(ωF < ω < ωD) + 3 dω4π 3 βhω¯ , (40) (2π) ωF cs e − 1 where V is a volume of the system and E0(ωF < ω < ωD) is the zero point energy:

3V Z ωD ω2 ¯hω E0(ωF < ω < ωD) = 3 4π 3 dω . (41) (2π) ωF cs 2 8

By using the Debye’s function, we express Eq.(40) as following form:

ˆ V 3 3
hHi (ωF < ω < ωD) = E0(ωF < ω < ωD) + 2 3 kBT ωDD(xD) − ωFD(xF) , (42) 2π cs The Debye’s function D(x) is often expressed as

3 Z x y3dy D(x) ≡ 3 y , (43) x 0 e − 1 where we define as x ≡ β¯hω. Similar to the high frequency regime, we can evaluate the energy of liquid in the lower frequency regime as

Z ωF 2 ˆ 2V ω ¯hω hHi (0 < ω < ωF) = E0(0 < ω < ωF) + 3 dω4π 3 βhω¯ , (44) (2π) 0 cs e − 1 V = E0(0 < ω < ωF) + 2 3 kBTD(xF), (45) 3π cs where E0(0 < ω < ωF) is the zero point energy:

2V Z ωF ω2 ¯hω E0(0 < ω < ωF) = 3 4π 3 dω . (46) (2π) 0 cs 2 Combining Eq.(42) and Eq.(46), we obtain the total energy of liquids:

hHˆ i = hHˆ i (0 < ω < ωF) + hHˆ i (ωF < ω < ωD) , (47)  3 ! kBT 3 ωF = E0 + V 2 3 ωD 3D(xD) − D(xF) , (48) 6π cs ωD where E0 is the total zero-point energy, E0(0 < ω < ωF) + E0(ωF < ω < ωD). If we take the heat expansion effect V → V (1 + αT/2) into account, we finally obtain

   3 !! αT ωF hHˆ i = 1 + E0 + NkBT 3D(xD) − D(xD) , (49) 2 ωD

3 2 3 where we used a relation, N = V ωD/6π cs with the number of atoms N. This result is consistent with [10]. Next, we estimate the total energy of liquids from the partition function, Eq.(28). By definition of the total energy, we obtain as follows:

" φ φ ! A A !# X E~ E~ E~ E~ hHˆ i = k + k + 2 · k + k . (50) βEφ βEA 2 ~k 2 e ~k − 1 ~k e − 1

For convenience, we define two energies, hHˆ iφ and hHˆ iA as

" A A # X E~ E~ hHˆ iA = 2 k + k , (51) βEA 2 e ~k − 1 ~k " φ φ # X E~ E~ hHˆ iφ = k + k . (52) βEφ 2 ~k ~k e − 1 Substituting summations to integrations, we obtain following forms:

Z ωD 2 ˆ A A 2V ω ¯hω hHi = E0 + 3 dω 4π 3 βhω¯ , (53) (2π) 0 cs e − 1 Z ωD 2 p 2 2 ˆ φ φ V ω ¯h ω − ωF hHi = E0 + 3 dω 4π 3 βhω¯ , (54) (2π) ωF cs e − 1 9

A φ where E0 and E0 are the zero point energies: Similar to the Debye’s model in solids, in order to obtain a suitable interpolating function of the heat capacity of liquids which behaves as 3NkBT and 2NkBT in the low and high temperature regime, respectively, we adopt the cut-off scale of the energy, ωD in Eq.(53) and Eq.(54).

Z ωD 2 A 2V ω ¯hω E0 = 3 dω4π 3 , (55) (2π) 0 cs 2 Z ωD 2 p 2 2 φ V ω ¯h ω − ωF E0 = 3 dω4π 3 . (56) (2π) ωF cs 2 By using the Debye’s function, we express Eq.(53) as following form:

ˆ A A kBT 3 hHi = E0 + 2V 2 3 ωDD(xD). (57) 6π cs Besides, Eq.(54) is calculated as

Z ωD 2 p 2 ˆ φ φ V ω ¯hω 1 − (ωF/ω) hHi = E0 + 3 dω4π 3 βhω¯ , (58) (2π) ωF cs e − 1 Z ωD 2 φ V ω ¯hω  2  = E0 + 3 dω4π 3 βhω¯ 1 + O((¯hωFβ) ) , (59) (2π) ωF cs e − 1  3 ! φ kBT 3 ωF 2 = E0 + V 2 3 ωD D(xD) − D(xF) + O((¯hωFβ) ). (60) 6π cs ωD

Combining Eq.(57) and Eq.(60), we obtain

 3 ! ˆ kBT 3 ωF 2
hHi = E0 + V 2 3 ωD 3D(xD) − D(xF) + O (¯hωFβ) , (61) 6π cs ωD

By taking the heat expansion effect into account, we obtain the total energy:

   3 !! αT ωF 2
hHˆ i = 1 + E0 + NkBT 3D(xD) − D(xF) + O (¯hωFβ) , (62) 2 ωD

A φ where E0 is the zero-point energy, E0 + E0 .