Two approaches for describing the Casimir interaction with graphene: density-density correlation function versus polarization tensor G. L. Klimchitskaya,1,2 V. M. Mostepanenko,1,2 and Bo E. Sernelius3 4 1 0 1Central Astronomical Observatory at Pulkovo of the 2 Russian Academy of Sciences, St.Petersburg, 196140, Russia b e F 2Institute of Physics, Nanotechnology and Telecommunications, 2 St.Petersburg State Polytechnical University, St.Petersburg, 195251, Russia 2 3Division of Theory and Modelling, ] r e Department of Physics, Chemistry and Biology, h t Link¨oping University, Link¨oping, SE-58183, Sweden o . t a Abstract m - Thecomparison studies of theoretical approaches to the description of the Casimir interaction in d n layeredsystemsincludinggrapheneisperformed. Itisshownthatatzerotemperaturetheapproach o c [ using the polarization tensor leads to the same results as the approach using the longitudinal 2 density-density correlation function of graphene. An explicit expression for the zero-temperature v 6 transverse density-density correlation function of graphene is provided. We further show that the 8 8 6 computational results for the Casimir free energy of graphene-graphene and graphene-Au plate . 1 interactions at room temperature, obtained using the temperature-dependent polarization tensor, 0 4 deviate significantly from those using the longitudinal density-density correlation function defined 1 : v at zero temperature. We derive both the longitudinal and transverse density-density correlation i X functions of graphene at nonzero temperature. The Casimir free energy in layered structures r a including graphene, computed using the temperature-dependent correlation functions, is exactly equal to that found using the polarization tensor. PACS numbers: 78.67.Wj, 65.80.Ck, 12.20.-m, 42.50.Ct 1 I. INTRODUCTION During the last few years graphene and other carbon-based nanostructures have at- tracted the particular attention of many experimentallists and theorists due to their remark- able properties [1, 2]. These investigations have provided further impetus to technological progress. One of the topical subjects, which came to the experimental attention very re- cently [3], is the van der Waals and Casimir interaction of graphene deposited on a substrate with the test body made of an ordinary material. Theorists have already undertook a number of studies of graphene-graphene and graphene-material plate interactions using the Dirac model of graphene [1], which assumes the linear dispersion relation for the graphene bands at low energies. Specifically, in Ref. [4] the van der Waals coefficient for two graphene sheets at zero temperature was calculated using the correlation energy from the random phase approximation (in Ref. [5] the obtained value was improved using the nonlocal dielectric function of graphene). In Ref. [6], the van der Waals and Casimir forces between graphene and ideal metal plane were calculated at zero temperature using the Lifshitz theory, where the reflection coefficients of the electro- magnetic oscillations were expressed via the polarization tensor in (2+1)-dimensions. The important progress was achieved in Ref. [7], where the force at nonzero temperature between two graphene sheets and between a graphene and a material plate was expressed via the Coulomb coupling between density fluctuations. The density-density correlation function in the random-phase approximation has been used. It was shown [7] that for graphene the relativistic effects are not essential, and that the thermal effects become crucial at much shorter separations thaninthecase ofordinarymaterials. InRef. [8], thegraphene-graphene interaction was computed under an assumption that the conductivity of graphene can be described by the in-plane optical properties of graphite. It was shown [9] that for sufficiently large band gap parameter of graphene the thermal Casimir force can vary several-fold with temperature. In Ref. [10], the reflection coefficients in the Lifshitz theory were expressed via the polarization tensor at nonzero temperature whose components were explicitly cal- culated. The detailed computations of graphene-graphene and graphene-real metal Casimir interactions using this method were performed [10–12]. Finally, in Ref. [13] the reflection coefficients of the Lifshitz theory were generalized for the case of planar structures including two-dimensional sheets. The graphene-graphene and graphene-real metal interactions at 2 both zero and nonzero temperature were computed by using the electric susceptibility (po- larizability)ofgrapheneexpressed viathedensity-density correlationfunction. Itwasargued [13] that the zero-temperature form of polarizability can be used also at room temperature. We underline that there is no complete agreement between the results of different papers devoted to the van der Waals and Casimir interactions with graphene (see Ref. [12] where some of the results obtained are compared). In fact, all the approaches go back to the Lifshitz theory [14–16], but with different approximations made and with various forms of the reflection coefficients used. By and large the approaches based on the density-density correlation function used its longitudinal version, i.e., neglected by the role of (small [7]) relativistic effects. Furthermore, dependence of the correlation function on temperature which was unknown until the present time, was obtained by means of scaling [7] or even neglected [13]. Bycontrast, calculations basedonthepolarizationtensor arefullyrelativistic and include an explicit dependence of its components on the temperature [10–12]. This is the reason why it would be useful to establish a link between the two approaches and to test the validity of the approximations used. In this paper, we find a correspondence between the reflection coefficients of the elec- tromagnetic fluctuations on graphene expressed in terms of electric susceptibility (polariz- ability) of graphene and components of the polarization tensor. On this basis, we derive explicit expressions for both longitudinal and transverse electric susceptibilities of graphene, density-density correlation function and conductivities at arbitrary temperature. Then we consider the limiting cases of the obtained expressions at zero temperature and find that the longitudinal version coincides with that derived within the random phase approximation. Furthermore, we compare the computational results for graphene-graphene and graphene- real metal interactions at room temperature obtained using the polarization tensor [11, 12] with those obtained using the density-density correlation function in Ref. [13]. In doing so we pay special attention to contributions of the transverse electric susceptibility of graphene and explicit temperature dependence of the longitudinal density-density correlation function to the Casimir free energy. The paper is organized as follows. In Sec. II we establish a link between the two ap- proaches and derive the density-density correlation functions at nonzero temperature. Sec- tion III is devoted to the case of zero temperature. In Sec. IV the computational results for graphene-graphene and graphene-real metal thermal Casimir interactions using the zero- 3 temperature correlation function and the polarization tensor at room temparature are com- pared. In Sec. V the reader will find our conclusions and discussion. II. COMPARISON BETWEEN THE REFLECTION COEFFICIENTS IN TWO THEORETICAL APPROACHES As discussed in Sec. I, all theoretical approaches to the van der Waals and Casimir interaction between two graphene sheets or between graphene and material plate go back to the Lifshitz theory representing the free energy per unit area at temperature T in thermal equilibrium in the form [14–16] k T ∞ ′ ∞ F(a,T) = B k dk ln 1−r(1) (iξ ,k )r(2) (iξ ,k )e−2aql 2π ⊥ ⊥ TM l ⊥ TM l ⊥ Xl=0 Z0 n h i +ln 1−r(1)(iξ ,k )r(2)(iξ ,k )e−2aql . (1) TE l ⊥ TE l ⊥ h io Here, k is the Boltzmann constant, k is the projection of the wave vector on the plane B ⊥ of graphene, ξ = 2πk Tl/~ with l = 0, 1, 2, ... are the Matsubara frequencies, q = (k2 + l B l ⊥ ξ2/c2)1/2, and the prime on the summation sign indicates that the term with l = 0 is divided l by two. The reflection coefficients on the two boundary planes separated by the vacuum gap of width a for the two independent polarizations of the electromagnetic field, transverse (1) (2) magnetic (TM) and transverse electric (TE), are notated as r and r . TM,TE TM,TE Let the first boundary plane be the freestanding graphene. There are two main repre- (1) (g) sentations for the reflection coefficents r ≡ r on graphene. We begin with the TM,TE TM,TE TM coefficient. Within the first theoretical approach, the longitudinal electric susceptibility (polarizability) of graphene at the imaginary Matsubara frequencies is expressed as 2πe2 α||(iξ ,k ) ≡ ε||(iξ ,k )−1 = − χ||(iξ ,k ), (2) l ⊥ l ⊥ l ⊥ k ⊥ where χ||(iξ ,k ) is the longitudinal density-density correlation function. The latter is con- l ⊥ nected with the dynamical conductivity of graphene by [13] e2ξ σ||(iξ ,k ) = − lχ||(iξ ,k ), (3) l ⊥ k2 l ⊥ ⊥ where e is the electron charge. Then the TM reflection coefficient of the electromagnetic oscillations on graphene can be expressed as [13, 17, 18] q α||(iξ ,k ) (g) l l ⊥ r (iξ ,k ) = . (4) TM l ⊥ k +q α||(iξ ,k ) ⊥ l l ⊥ 4 The explicit form for α|| is discussed below. Within the second theoretical approach, the TM reflection coefficient is expressed via the 00-component Π of the polarization tensor in (2+1)-dimensional space-time [10–12] 00 q Π (iξ ,k ) (g) l 00 l ⊥ r (iξ ,k ) = . (5) TM l ⊥ 2~k2 +q Π (iξ ,k ) ⊥ l 00 l ⊥ The analytic expression for Π is known [10–12]. It depends on the temperature both 00 implicitly (through the Matsubara frequencies) and explicitly, as on a parameter. For the pristine (undoped) gapless graphene one has [10–12] π~αk2 8~αc2 1 k T Π (iξ ,k ) = ⊥ + dx B (6) 00 l ⊥ f(ξ ,k ) v2 ~c l ⊥ F Z0 ( ×ln 1+2cos(2πlx)e−θT(ξl,k⊥,x) +e−2θT(ξl,k⊥,x) ξ sin(2πlx) − (cid:2)l (1−2x) (cid:3) 2c coshθ (ξ ,k ,x)+cos(2πlx) T l ⊥ ξ2 x(1−x) cos(2πlx)+e−θT(ξl,k⊥,x) + l , c2f(ξ ,k ) coshθ (ξ ,k ,x)+cos(2πlx) p l ⊥ T l ⊥ ) where α = e2/(~c) is the fine structure constant, v is the Fermi velocity, and the following F notations are introduced v2 ξ2 1/2 f(ξ ,k ) ≡ Fk2 + l , (7) l ⊥ c2 ⊥ c2 (cid:18) (cid:19) ~c θ (ξ ,k ,x) ≡ f(ξ ,k ) x(1−x). T l ⊥ l ⊥ k T B p Now we equate the right-hand sides of Eqs. (4) and (5) and obtain the expression for the longitudinal polarizability of graphene at nonzero temperature via the 00-component of the polarization tensor 1 α||(iξ ,k ) = Π (iξ ,k ). (8) l ⊥ 2~k 00 l ⊥ ⊥ Using Eq. (2), for the longitudinal density-density correlation function one obtains 1 χ||(iξ ,k ) = − Π (iξ ,k ), (9) l ⊥ 4πe2~ 00 l ⊥ where Π is given by Eq. (6). Similar to the polarization tensor, the density-density corre- 00 lation function depends on T both implicitly and explicitly. The longitudinal conductivity of graphene at any T is given by Eq. (3). 5 Wecontinue withtheTEreflection coefficient. NotethatEqs. (2)and(3)remainvalidfor the transverse quantities: the polarizability of graphene α⊥(iξ ,k ), the transverse permit- l ⊥ tivity ε⊥(iξ ,k ), the density-density correlation function χ⊥(iξ ,k ), and the conductivity l ⊥ l ⊥ σ⊥(iξ ,k ). The TE reflection coefficient on graphene in terms of the transverse polarizabil- l ⊥ ity was found in Ref. [13] ξ2α⊥(iξ ,k ) r(g)(iξ ,k ) = − l l ⊥ . (10) TE l ⊥ c2k q +ξ2α⊥(iξ ,k ) ⊥ l l l ⊥ Note that according to our knowledge no explicit expression of α⊥ for graphene is available in the published literature. In terms of the polarization tensor, the TE reflection coefficient takes the form [10–12] k2Π (iξ ,k )−q2Π (iξ ,k ) r(g)(iξ ,k ) = − ⊥ tr l ⊥ l 00 l ⊥ , (11) TE l ⊥ 2~k2q +k2Π (iξ ,k )−q2Π (iξ ,k ) ⊥ l ⊥ tr l ⊥ l 00 l ⊥ where the index tr denotes the sum of spatial component Π1 and Π2. For the undoped 1 2 gapless graphene the analytic expression for Π is the following [10–12]: tr π~α ξ2 Π (iξ ,k ) = Π (iξ ,k )+ f2(ξ ,k )+ l (12) tr l ⊥ 00 l ⊥ f(ξ ,k ) l ⊥ c2 l ⊥ (cid:20) (cid:21) 1 ξ sin(2πlx) +8~α dx l(1−2x) c coshθ (ξ ,k ,x)+cos(2πlx) Z0 ( T l ⊥ x(1−x) ξ2 cos(2πlx)+e−θT(ξl,k⊥,x) − f2(ξ ,k )+ l . f(ξ ,k ) l ⊥ c2 coshθ (ξ ,k ,x)+cos(2πlx) p l ⊥ (cid:20) (cid:21) T l ⊥ ) By equating the right-hand sides of Eqs. (10) and (11), one obtains the expression for the transverse polarizability of graphene at any nonzero temperature c2 α⊥(iξ ,k ) = k2Π (iξ ,k )−q2Π (iξ ,k ) . (13) l ⊥ 2~k ξ2 ⊥ tr l ⊥ l 00 l ⊥ ⊥ l (cid:2) (cid:3) The respective result for the transverse density-density correlation function is found from equation similar to Eq. (2) c2 χ⊥(iξ ,k ) = − k2Π (iξ ,k )−q2Π (iξ ,k ) . (14) l ⊥ 4π~e2ξ2 ⊥ tr l ⊥ l 00 l ⊥ l (cid:2) (cid:3) Then the transverse conductivity of graphene is given by Eq. (3) where the index || is replaced with ⊥. Weemphasize that Eqs. (4), (5) and (10), (11) arethe exact consequencies of theMaxwell equations and electrodynamic boundary conditions imposed on the 2D graphene sheet. For 6 this reason, the obtained connections (8), (9) and (13), (14) between the polarizabilities and density-density correlation functions for graphene, on the one hand, and the components of the polarization tensor, on the other hand, are the exact ones. Keeping in mind that Eqs. (6) and (12) for the polarization tensor are calculated in the one-loop approximation [10], the specific expressions for the polarizabilities and density-density correlation functions obtained after the substitution of Eqs. (6) and (12) in Eqs. (8), (9) and (13), (14) should be also considered as found in the same approximation. In the next section we compare them with those contained in the literature. To conclude this section, we present an explicit expression for the quantity k2Π − ⊥ tr q2Π entering the transverse polarizability, the density-density correlation function and the l 00 conductivity of graphene. Substituting Π from Eq. (6) and Π from Eq. (12), one obtains 00 tr after identical transformations k2Π (iξ ,k )−q2Π (iξ ,k ) = π~αk2f(ξ ,k ) (15) ⊥ tr l ⊥ l 00 l ⊥ ⊥ l ⊥ 8~αc2 1 k Tξ2 − dx B l ln 1+2cos(2πlx)e−θT(ξl,k⊥,x) +e−2θT(ξl,k⊥,x) v2 ~c3 F Z0 ( (cid:2) (cid:3) ξ2 ξ sin(2πlx) − 2f2(ξ ,k )− l l (1−2x) l ⊥ c2 2c coshθ (ξ ,k ,x)+cos(2πlx) (cid:20) (cid:21) T l ⊥ cos(2πlx)+e−θT(ξl,k⊥,x) + x(1−x)f3(ξ ,k ) . l ⊥ coshθ (ξ ,k ,x)+cos(2πlx) T l ⊥ (cid:27) p This expression is used in below calculations. III. ENERGY OF THE CASIMIR INTERACTION BETWEEN TWO GRAPHENE SHEETS AT ZERO TEMPERATURE In the limiting case T → 0 the summation over the discrete Matsubara frequencies in Eq. (1) is replaced with integration over the imaginary frequency axis, and for two graphene sheets one arrives to the Casimir energy per unit area ~ ∞ ∞ E(a,T) = k dk dξ ln 1−r(g) 2(iξ,k )e−2aq 4π2 ⊥ ⊥ TM ⊥ Z0 Z0 n h i +ln 1−r(g)2(iξ,k )e−2aq . (16) TE ⊥ h io Here, the reflection coefficients are given by either Eqs. (4) and (10) or (5) and (11), where the discrete frequencies ξ are replaced with the continuous ξ. l 7 We begin from the contribution of the TM mode, E , to the total energy (16). In terms TM (g) of the polarization tensor, the reflection coefficient r is given by Eqs. (5) and (6) with TM the notation (7). As can be seen in Eq. (7), the quantity θ (ξ ,k ,x) → ∞ when T → 0. T l ⊥ Because of this, from Eq. (6) at T = 0K one obtains [6] π~αk2 Π (iξ,k ) = ⊥ . (17) 00 ⊥ f(ξ,k ) ⊥ Using the notation (7), one obtains from Eq. (8) the longitudinal polarizability of graphene at zero temperature πe2 k α||(iξ,k ) = ⊥ , (18) ⊥ 2~ v2k2 +ξ2 F ⊥ and from Eq. (9) the respective density-density correlation function p 1 k2 χ||(iξ,k ) = − ⊥ . (19) ⊥ 4~ v2k2 +ξ2 F ⊥ The longitudinal conductivity of graphene at T p= 0K is obtained from Eqs. (3) and (19). The density-density correlation function (19) at T = 0K, derived from the polarization tensor, coincides withtheclassical result[19,20]whichwasusedincomputationsofRef.[13]. Then, for the TM reflection coefficient on graphene at T = 0K we obtain one and the same result either from Eqs. (4) and (18) or from Eqs. (5) and (17) πe2 c2k2 +ξ2 r(g) (iξ,k ) = ⊥ . (20) TM ⊥ 2~c v2k2 +ξ2 +πe2 c2k2 +ξ2 F ⊥ p ⊥ This reflection coefficient coincides withpthat used in Ref.p[13]. We continue by considering the contribution of the TE mode, E , to the Casimir energy TE (g) (16). In terms of the polarization tensor, the reflection coefficient r is given by Eq. (11). TE The combination of the components of the polarization tensor, k2Π − q2Π , entering ⊥ tr l 00 Eq. (11), is given by Eq. (15). In the limiting case T → 0, one obtains from Eq. (15) k2Π (iξ,k )−q2Π (iξ,k ) = π~αk2f(ξ,k ). (21) ⊥ tr ⊥ 00 ⊥ ⊥ ⊥ Substituting Eq. (21) in Eq. (13), we find the transverse polarizability of graphene at zero temperature πe2k α⊥(iξ,k ) = ⊥ v2k2 +ξ2. (22) ⊥ 2~ξ2 F ⊥ q In a similar way, substituting Eq. (21) in Eq. (14), we find the transverse density-density correlation function at T = 0K k2 χ⊥(iξ,k ) = − ⊥ v2k2 +ξ2. (23) ⊥ 4~ξ2 F ⊥ q 8 The TE reflection coefficient at T = 0K is obtained either substituting Eq. (21) in Eq. (11) or Eq. (22) in Eq. (10). The result is πe2 v2k2 +ξ2 r(g)(iξ,k ) = − F ⊥ . (24) TE ⊥ 2~c c2k2 +ξ2 +πe2 v2k2 +ξ2 ⊥ p F ⊥ p p (g) As is seen from the comparison of Eqs. (20) and (24), the reflection coefficient r has the TE (g) opposite sign, as compared with r , and its magnitude is obtained from the latter by the TM interchanging of c and v . F Now we compare the computational results for the Casimir energy per unit area of two parallel graphene sheets at zero temperature obtained in Ref. [13] by means of the density- density correlation function and here using the polarization tensor. In both cases the Fermi velocity v = 8.73723 × 105m/s is employed [13, 21, 22]. In Fig. 1 the computational F results of Ref. [13] for E(a) normalized for the Casimir energy per unit area of two parallel ideal-metal planes π2 ~c E (a) = − (25) im 720 a3 are shown as black dots over the separation region from 10nm to 5µm. In making compu- tations it was assumed [13] that χ⊥(iξ,k ) = χ||(iξ,k ). The gray line shows our computa- ⊥ ⊥ tional results for E(a)/E (a) using the polarization tensor at T = 0K given by Eqs. (17) im and (21). In this case the contribution of the TE mode was calculated precisely. As can be seen in Fig. 1, both sets of computational results are in a very good agreement. ThisisexplainedbythefactthatE (a)contributes99.6%ofE(a)andE (a) = 0.004E(a) TM TE at all separation distances. Furthermore, the relative differences between the computational results of Ref. [13] for E (a) (obtained under the assumption that χ⊥ = χ||) and our results TE (g) here computed with the exact reflection coefficient r are of about 0.1%. Thus, the role of TE the TE contribution to the Casimir energy of two graphene sheets is really negligibly small [7], and it is not critical what form of the transverse density-density correlation function is used in computations. Physically this is connected with the fact that the TE contribution is missing in the nonrelativistic limit, whereas the relativistic effects contain additional small factors of the order of v /c. F 9 IV. CASIMIR INTERACTION WITH GRAPHENE AT NONZERO TEMPERA- TURE In this section we compare the computational results for the Casimir free energy of two graphene sheets and a freestanding graphene sheet interacting with an Au plate obtained using the approach of Ref. [13] and using the polarization tensor. All computations here are done at room temperature T = 300K. In this way we find the role of explicit dependence of the density-density correlation function and polarization tensor on the temperature. A. Two graphene sheets The free energy of the Casimir interaction between two sheets of undoped graphene was (1) (2) (g) computed at T = 300K using Eq. (1) with r = r = r . All computations TM,TE TM,TE TM,TE were performed using the following two approaches: the approach of Ref. [13] using the reflection coefficients (4) and (10), expressed via the zero-temperature longitunidal density- density correlationfunction(19)andtheapproachofRef.[12]using thereflectioncoefficients (5) and (11) expressed via the components of the polarization tensor (6) and (12). Within the approach Ref. [13], the dependence of the free energy on T is determined by the T- dependent Matsubara frequencies, whereas in the approach of Ref. [12] there is also explicit dependence of the polarization tensor on T as a parameter. InFig. 2 we present the computational results for the Casimir free energy oftwo graphene sheets at T = 300K as functions of separation over the interval from 10nm to 1µm. The results obtained using the polarization tensor at T = 300K are shown as the upper solid line, andthe results obtained using the longitudinal density-density correlation function (19) definedatT = 0Kareshownasdots. FromFig.2itisseenthattheuppersolidlinedeviates from dots significantly even at short separations. This is explained by the dependence of the polarization tensor on T as a parameter in addition to the implicit T-dependence through the Matsubara frequencies. The lower (gray) solid line in Fig. 2 shows the computational results obtained by means of the polarization tensor (17) and (21) at T = 0K. This solid line is in a very good agreement with dots computed using the formalism of Ref. [13], as it should be according to the results of Secs. II and III. Note that the dominant contribution to the free energy of graphene-graphene interaction 10