arXiv:cond-mat/9711203v1 [cond-mat.soft] 20 Nov 1997 Mean Field Theory for Lossy Nonlinear Composites K. W. Yu Department of Physics, The Chinese University of Hong Kong, Shatin, New Territories, Hong Kong Abstract The mean-field theory for lossy nonlinear composites, described by complex and field-dependent dielectric functions, is presented. By using the spectral representation of linear composites with identical microstructure, we develop self-consistent equations for the effective response. We examine two types of microstructure, namely, the Maxwell-Garnett approximation and the effective medium approximation to illustrate the theory. PACS Numbers: 72.20.Ht, 64.60.Ak, 72.60.+g Typeset using REVTEX 1
I. INTRODUCTION Recently, we proposed a mean-field theory (MFT) for nonlinear composites [1]. The the- ory was applied to nonlinear transport properties of conducting composites at zero frequency, described by real conductivities and provided good agreement with numerical simulation [1]. However, for dielectric materials near relaxation, the loss may be important. It is tempting to extend the MFT to complex and nonlinear permitivities. The local constitutive relation of the composite system is given by D=χ(E∗·E)E,(1) whereχ=χ(x) is the (position dependent) complex third-order nonlinear coefficient.In what follows, we denote|E|2=E∗·EandE2=E·Efor convenience. The effective response is defined as the volume average ofD χeE2 0E0= 1 V ∫ V dVD(x),(2) whereVis the volume of composites andE0is the applied (average) field and is taken to be real: E0= 1 V ∫ V dVE(x).(3) The corresponding boundary-value problem of nonlinear composite media consists of the field equations: ∇ ·D= 0,(4) ∇ ×E= 0.(5) There exists a potentialϕso thatE=∇ϕ. The boundary condition forϕis ϕ=ϕ0=E0·x(6) on surface (S) of the composite andϕ0is real. Consider the integral (1/V)∫ VdVD·E. Through the field equations, we can convert it into an integral over the surface (S) of the 2
composite: (1/V)∫ SdS·(Dϕ), which is equal to (1/V)∫ SdS·(Dϕ0) becauseϕ=ϕ0on S. Converting the surface integral back to the volume integral, we obtain: 1 V ∫ V dVD·E= 1 V ∫ V dVD·E0.(7) Hence we find an equivalent expression for the effective response: χeE4 0= 1 V ∫ V dVD·E.(8) Let us consider a lossy nonlinear composite in which inclusions of complex nonlinear coef- ficientχ1, present at volume fractionp1, are randomly embedded in a host medium ofχ2, present at volume fractionp2, with the application of an external uniform fieldE0.Note thatp1+p2= 1.From Eqs.(1) and (8), we have therefore an expression for the effective response: χeE4 0=p1χ1〈|E1|2E2 1〉+p2χ2〈|E2|2E2 2〉,(9) where 〈|Ei|2E2 i〉= 1 Vi ∫ Vi dV|E|2E2,i= 1,2.(10) denotes the local field average within theith component. Similar to the derivation of Eq.(7), asϕ∗=ϕ0onS, we arrive at 1 V ∫ V dVD·E∗= 1 V ∫ V dVD·E0,(11) which implies an alternative expression for the effective response: χeE4 0=p1χ1〈(|E1|2)2〉+p2χ2〈(|E2|2)2〉,(12) where 〈(|Ei|2)2〉= 1 Vi ∫ Vi dV(|E|2)2,i= 1,2.(13) We should remark that the different types of local field averages defined in Eqs.(10) and (13) are generally not equal to each other. It is interesting to note that they both give the same 3
effective response [Eqs.(9) and (12)]. However, the nonlinear partial differential equations pertaining to the boundary-value problem cannot be solved analytically.We shall invoke the mean-field theory [1] to obtain an approximate expression for the effective response. According to previous work [1], we consider the linear constitutive relation: D=ǫE,(14) whereǫ=ǫ(x) is the (position dependent) complex dielectric constant.By definition, we find the effective linear response: ǫeE0= 1 V ∫ V dV ǫE.(15) From Eq.(7), we have therefore ǫeE2 0=p1ǫ1〈E2 1〉+p2ǫ2〈E2 2〉.(16) From Eq.(11), we have alternatively ǫeE2 0=p1ǫ1〈|E1|2〉+p2ǫ2〈|E2|2〉.(17) This expression will be useful for calculating〈|Ei|2〉. In order to estimateχe, we invoke the decoupling approximation [2] 〈|Ei|2E2 i〉=〈|Ei|2〉〈E2 i〉,i= 1,2.(18) within theith component.This assumption is good in microstructure for which the local electric field is nearly uniform within theith component, but less accurate when these fluctuations are large, as in a random composite near the percolation threshold [2]. With this assumption, χeE4 0=p1χ1〈|E1|2〉〈E2 1〉+p2χ2〈|E2|2〉〈E2 2〉.(19) Comparing Eqs.(16) and (19), it is tempting to writeǫ1=χ1〈|E1|2〉,ǫ2=χ2〈|E2|2〉and ǫe=χeE2 0.Hence, within the decoupling approximation, we may interpret the nonlinear 4
component as a linear dielectric material with a field-dependent dielectric constant. Suppose ǫeis known as a function of its constituent dielectric constants, ǫe=F(ǫ1, ǫ2, p1),(20) then we find the local field averages [1] p1〈E2 1〉=∂ǫe ∂ǫ1 E2 0,(21) p2〈E2 2〉=∂ǫe ∂ǫ2 E2 0.(22) However, we have to determine〈|Ei|2〉in theith component, which is generally not equal to〈E2 i〉. Thus a straightforward application of the previous formalism [1] is impossible. We resort to an alternative approach based on the spectral representation [3]. In what follows, we shall consider two important types of microstructures: (1) Dispersion microstructures as in the Maxwell-Garnett approximation [4] and (2) symmetric microstruc- tures as in the Bruggeman effective medium approximation [5]. II. MAXWELL-GARNETT APPROXIMATION The Maxwell-Garnett approximation (MGA) is good for dispersion microstructures and the theory is inherently non-symmetrical [4]. For convenience, we consider the case in which component 1 is embedded in component 2. For a linear composite ofǫ1andǫ2at volume fractionsp1andp2respectively, the MGA reads [4]: ǫe−ǫ2 ǫe+ (d−1)ǫ2 =p1 ǫ1−ǫ2 ǫ1+ (d−1)ǫ2 ,(23) wheredis the dimensionality of composites. Solving Eq.(23), we obtain ǫe=ǫ2[ǫ1(dp1+p2) +ǫ2(d−1)p2] p2ǫ1+ (d−p2)ǫ2 .(24) From Eqs.(21) and (22), we can calculate the local field averages: p1〈E2 1〉=p1d2ǫ2 2 [p2ǫ1+ (d−p2)ǫ2]2E2 0.(25) 5
p2〈E2 2〉= ( 1−p1d[dǫ2 2−p2(ǫ1−ǫ2)2] [p2ǫ1+ (d−p2)ǫ2]2 ) E2 0.(26) In MGA, the local field in the inclusion is uniform, we can determineE1and henceE∗ 1 explicitly, we find p1〈|E1|2〉=p1d2|ǫ2|2 |p2ǫ1+ (d−p2)ǫ2|2E2 0.(27) From Eq.(17) and using Eqs.(24) and (27), we find p2〈|E2|2〉= ( 1−p1d[d|ǫ2|2−p2|ǫ1−ǫ2|2] |p2ǫ1+ (d−p2)ǫ2|2 ) E2 0.(28) We should remark that although|〈E2 1〉|=〈|E1|2〉in MGA,|〈E2 2〉|is generally not equal to〈|E2|2〉.In fact,|〈E2 2〉| ≤ 〈|E2|2〉.A common error in the literature is to treat these averages equal even whenχis complex. If we writeǫ1=χ1〈|E1|2〉andǫ2=χ2〈|E2|2〉, then Eqs.(27)–(28) can be solved self-consistently for〈|E1|2〉and〈|E2|2〉and hence the effective nonlinear response can be calculated. In Fig.1, we present the MGA results in three dimensions (3D). We letχ1= 1 + 3i andχ2= 3 +iand computeχeas a function ofp1. Asp1increases, Re(χe) decreases from Re(χ2) = 3 towards Re(χ1) = 1 while Im(χe) increases from Im(χ2) = 1 towards Im(χ1) = 3. Since the contrast between the two components is relatively small, the local field averages 〈|E1|2〉and〈|E2|2〉remain close to unity. However, unlike|〈E2 1〉|,|〈E2 2〉|shows a significant deviation from〈|E2|2〉, indicating the nonuniformity of local field in the host. For two-component composites, it has proved convenient to adopt the spectral represen- tation of the effective linear response [3]: Letv= 1−ǫ1/ǫ2,w= 1−ǫe/ǫ2, ands= 1/v, we find w(s) = ∫1 0 m(s′)ds′ s−s′,(29) wherem(s) is the spectral density which is obtained through a limiting process: m(s) = lim η→0+−1 πImw(s+iη).(30) Furthermore,m(s) obeys the sum rule 6
∫1 0 m(s′)ds′=p1.(31) Eqs.(24)–(28) can readily be converted into the spectral representation. We find w=p1 s−s1 ,(32) wheres1=p2/d. From Eq.(30), we obtain the spectral density in MGA: m(s) =p1δ(s−s1).(33) Similarly, the local field averages [Eqs.(25)–(28)] are given by 〈E2 1〉=E2 0 (1−vs1)2,(34) 〈|E1|2〉=E2 0 |1−vs1|2,(35) p2〈E2 2〉= ( 1−p1(1−v2s1) (1−vs1)2 ) E2 0,(36) p2〈|E2|2〉= ( 1−p1(1− |v|2s1) |1−vs1|2 ) E2 0.(37) If we letv= 1−χ1〈|E1|2〉/χ2〈|E2|2〉and solve Eqs.(35) and (37) self-consistently for〈|E1|2〉 and〈|E2|2〉, the effective nonlinear response can be determined.Although the solution appears somewhat simpler, the same results are obtained. III. EFFECTIVE MEDIUM APPROXIMATION The Bruggeman effective-medium approximation (EMA) is known to be symmetrical with respect to interchanging components 1 and 2 [5].Note that there is a percolation threshold in the theory. For a random linear composite ofǫ1andǫ2at volume fractionsp1 andp2respectively, the self-consistency equations reads [5]: p1 ǫ1−ǫe ǫ1+ (d−1)ǫe +p2 ǫ2−ǫe ǫ2+ (d−1)ǫe = 0,(38) wheredis the dimensionality of composites. Solving Eq.(38), we obtainǫeand hencewand the spectral density 7
m(s) =dp1−1 d−1δ(s)θ(dp1−1) +m1(s),(39) whereδ(s) andθ(dp1−1) denote the Dirac delta function and the Heavyside step function respectively, and m1(s) =d 2π(d−1)s √ (s−s1)(s2−s),s1< s < s2,(40) = 0,otherwise, wheres1ands2are given by s1= 1 d ( 1 + (d−2)p1−2 √ (d−1)p1(1−p1) ) , s2= 1 d ( 1 + (d−2)p1+ 2 √ (d−1)p1(1−p1) ) .(41) Whenp1>1/d, which is the percolation threshold in EMA, component 1 percolates the composite and there is aδfunction contribution tom(s) ats= 0. Eqs.(34)–(37) can readily be generalized to [6]: p1〈E2 1〉= ∫1 0 m(s′)ds′ (1−vs′)2E2 0,(42) p1〈|E1|2〉= ∫1 0 m(s′)ds′ |1−vs′|2E2 0,(43) p2〈E2 2〉= ( 1− ∫1 0 (1−v2s′)m(s′)ds′ (1−vs′)2 ) E2 0,(44) p2〈|E2|2〉= ( 1− ∫1 0 (1− |v|2s′)m(s′)ds′ |1−vs′|2 ) E2 0.(45) One can check that when the MGA spectral density [Eq.(33)] is used, Eqs.(34)–(37) are recovered. Again we letv= 1−χ1〈|E1|2〉/χ2〈|E2|2〉. Eqs.(43) and (45) are coupled integral equations.They can be solved numerically for〈|E1|2〉and〈|E2|2〉and hence the effective nonlinear response can be determined. In Fig.2, we present the EMA results in 3D. We letχ1= 1 + 3iandχ2= 3 +i. Fig.2 exhibits similar behavior as Fig.1, however, with some differences. Here Re(χe) and Im(χe) cross atp1= 0.5 because EMA describes the symmetric microstructure and we have the 8
100%
Scan to connect with one of our mobile apps
Coinbase Wallet app
Connect with your self-custody wallet
Coinbase app
Connect with your Coinbase account
Open Coinbase Wallet app
Tap Scan
Or try the Coinbase Wallet browser extension
Connect with dapps with just one click on your desktop browser
Add an additional layer of security by using a supported Ledger hardware wallet