Anda di halaman 1dari 8

IL NUOVO CIMENTO

VOL. 14 D, N. 5

Maggio 1992

Gauge Technique and BCS Theory of Superconductivity(*).


R. ACHARYA(1) and P. NARAYANA SWAMY(2)
(1) Physics Department, Arizona State University - Tempe, A Z 85287

(2) Physics Department, Southern Illinois University - EdwardsviUe, IL 62026 (ricevuto il 27 Settembre 1991; approvato il 16 Marzo 1992)

Summary. m The gauge technique in unbroken (exact) relativistic quantum electrodynamics is applied to the nonrelativistic BCS theory exhibiting spontaneously broken U(1) gauge invariance. In addition to the BCS-type solution, we find an interesting new solution for weak coupling exhibiting a high T c.
PACS 74.20.Fg- BCS theory; applications. PACS 11.15.Tk - Other nonperturbative techniques. PACS 12.20 - Quantum electrodynamics.

The pioneering classic theory of Bardeen, Cooper and Schrieffer [1] (BCS) was the first microscopic theory which provided a satisfactory description of superconductivity in metals, which also lead to several remarkable predictions. The two foremost results are: 1) the critical temperature Tr is of the order of 0D exp [--1/~e~], where 0D is the Debye temperature for the solid and ~eff ~--"N ( 0 ) V a dimensionless parameter characterizing the strength of attraction between the electrons in a superconductor near the Fermi surface. The value of Tc is thus determined by only two parameters. 2) The BCS formula relating Tr and the energy gap A(O)/kB Tc = =/],, where k s is the Boltzmann constant and ln~, is the Euler constant. An elegant reformulation of the BCS theory which parallels quantum electrodynamics (QED), incorporating the role of Ward identity, is due to Nambu[2]. In Nambu's formulation it becomes transparent that the BCS theory relies on the ladder approximation to the Dyson-Schwinger (DS) equation for the electron propagator. This aspect is clearly presented by Schrieffer[3]. The general features of superconductivity are indeed model-independent consequences of the spontaneous breakdown of electromagnetic gauge invariance, as has been demonstrated by Weinberg[4]. The BCS theory was further developed and refined by Anderson, Bogoliubov, Gorkov and Eliashberg [5]. The work of Eliashberg in particular deals with a more realistic interaction between two electrons which includes a fully retarded attraction (*) The authors of this paper have agreed to not receive the proofs for correction. 487

488

R. ACHARYA

and

P. NARAYANA SWAMY

due to phonons and the screened Coulomb repulsion, in contrast to the simple constant pairing interaction in the BCS theory [5, 6]. Nevertheless the strong-coupling theory [5] in its customary form ignores vertex corrections [7]. This neglect is usually justified by appealing to a theorem due to Migdal [8], according to which such corrections are of order ( m * / M ) 1/2, where m* is the effective electron mass and M is the mass of the ion. It has been emphasized by Mahan[7], however, that Migdal's theorem may be unreliable since superconductivity itself (i.e., the Cooper instability) is caused by a vertex correction. It is the purpose of this note to incorporate the vertex corrections due to phonon interactions by making use of the generalized Ward-Takahashi (WT) identity [3] for the superconductor, i.e. (1) q~F~(p + q,p) = ~sG-l (p) - G-l (p + q)~3 ,

where vi(i = 1,2,3) are the Pauli matrices and summation over repeated indices -- 0, 1, 2, 3 is implied. We use the notation convention in which the metric is defined by g,y = [1, -1, -1, -1] and h = c = 1. The inverse of the electron propagator is given by the form (2) G-1 (p) = Z(p) Po I - sp ~3 - Ap ~1

in the theory of superconductivity in Nambu's notation[9], where Ap is the gap function signifying broken symmetry and ~p is the nonrelativistic energy measured relative to the Fermi energy, ~p = p 2 / 2 m * - ~ . The renormalization factor Z(p) satisfies coupled integral equations [3] and has the general form (3) Z(p) = 1 + O(~e~).

In this work, we shall utilize the leading order approximation for the renormalization factory only, Z ~ 1 (pairing approximation [3]), which will enable us to decouple the equation satisfied by the gap function Ap. In the pairing approximation, G -I (p) is given by a sum of graphs represented by the equation (4) G-1 (p) = Po I - ~p~3 - i(27:) -4 f d4k
v3 G(p

+ k) ~8V(k).

This is the ladder approximation to the DS equation in field theory. We observe that the free vertex is given by [3]
(5) r~ ( p ' , p ) = ~3 (~ = 0).

We can write down the full DS equation analogous to QED, while still retaining the pairing form of the potential, by replacing one of the factors of v3 in eq. (4) with the dressed vertex function I'o(p, k). It is worth noting that the bare vertex ~3 is renormalized by the residual interaction[lO] and yields the dressed vertex /'0. Following Schrieffer's notation[3], the DS equation, which takes us beyond the ladder approximation, thus reads
(6)

G-1 (P) = P01 - ~pv3 - i(27:)-4 f dtk ~3G(p + k)Fo (p, k)V(k).

The salient features of the gauge method invented by Salam[ll, 12] consist of expressing the longitudinal part of the vertex function which is a solution of eq. (1) in

GAUGE TECHNIQUE AND BCS THEORY OF SUPERCONDUCTIVITY

489

terms of the electron propagator, thereby leading to a closed system of equations for the electron propagator. The gauge theory, based on this technique, is manifestly nonperturbative and this ensures its validity as a strong-coupling theory. The gauge approximation consists of discarding the unknown transverse piece, F w (not determined by the WT identity), which satisfies

(7)

q~'FT(p+q,p)=--O,

F =/'L+F T

and retaining the longitudinal part F L. We shall omit the superscript L in what follows for simplicity. We can write down the solution f o r / ' , in a straightforward manner, thus

(8)

(p' + p), F, (p', p) = [(~' - ~) 1 + (P0 - Pg ) r8 ] (p,2_p2)

(P' - P)~ ire (A + 4 ' ) (p' -- p)~

where ~'= ~p,. This invariant form, in addition to being unique, possesses all the desirable properties enjoyed by the general decomposition form in scalar QED [11]. The second term on the r.h.s, is singular at p ' - p = 0 and represents the Goldstone pole signifying spontaneously broken U(1) symmetry, whose presence is essential for the existence of a nonvanishing gap function A. We need only F0, the ~ = 0 component of the vertex function, which is given by (9) Fo (0, k) = ~-~ [(Sk + ~) 1 -- ko r8 + it2 (A + zl')].
ko

Inserting this in eq. (6) after setting p = 0 (with no loss of generality), we obtain (10) G-1 (0) = ~r8 - i(2=) -4 j dtk rs G(k)Fo (0, k) V(k).

Using eq. (2), with Z(p) = 1, and assuming a constant pairing potential V(k) = V and a constant gap function Ap = A, we thus derive (11)
AZl---- i(2r0-4 V

f dakr3G(k)~-22 {(s k + tz) l

- k o r 8 + 2i At2 } .

After some algebra, this leads to the superconductivity gap equation (12) f 1 = 3i(2=) -4 V ~ dak
J

ko 2 (ko - k 2 )(ko - E~ ) ' 2 2

where Ek = V ~ + A 2. This result represents the major consequence of the gauge theory at zero temperature and may be contrasted with the corresponding equation in the BCS case: (12a)
34 -

f 1 = i(270-4V | dak
J
(ko

1
'

II NuovoCimentoD

490

R. ACHARYA

and

P. NARAYANA SWAMY

which, after integration over ko leads to [10] (12b)

l =

V----~- d3k ~ f
2(2r:)3

To proceed with an extension to finite temperatures, we may employ the finite-temperature field theory developed by Dolan and Jackiw [13] in the real-time formalism wherein the boson and fermion propagators are respectively given by the extensions: (13) and (14) 1 ~ ko - E~ + i~ 2 1 + 2ir:~(k~ - E~) k~ - E~ + i~ exp [fllk01 + 1] "

1
k" k, + i~

1
k" k, + i~

2ir:~(k~'k~ )
exp [fllko I - 1]

We observe that the finite-temperature boson propagator, eq. (13), is consistent with eqs. (1) and (8). Inserting these in eq. (12), and after an integration over the variable k0, we obtain (15) 13___~V| d 3 k E~ - k 2 Ek tgh 2(2=)3 J

~Ek --

where ~ = l / k s T and k 2 = 2m* ~. Equation (15) contains our fundamental result of the gauge theory of superconductivity for any temperature. We shall henceforward omit the subscript k in E and ~ for convenience. We recover eq. (12) for T = 0 K. Equation (15) should be contrasted with the BCS theory result,

(16)

1= 2(2r:)a fd3kE tgh(lzE)

We shall now proceed to analyze the consequences of the gauge theory result, eq. (15), for superconductivity and describe the points of departure from the conclusions of the BCS theory. Introducing the density of states N(0) on the Fermi surface and a cut-off represented by the Debye frequency ~ [14], we can cast eq. (15) in the form

(17)

1=

VN(O)
--~o

d~ ~ 2 + A 2 _ 2 m .

We may now specialize to the critical temperature T = Tc, ~ = fie when the gap A =

GAUGE TECHNIQUE AND B C S THEORY OF SUPERCONDUCTIVITY

491

= zl(Tr vanishes. In this case we obtain


to

(18)

1 =1 f d~ [~tgh(1/~c~)_ 2 V ~ - ~ c t g h ( l ~ e 3VN(0) 2 r _ 2m* ~


--to

2V~-~) ]

We can now determine Tr from eq. (18) by resorting to the approximation [15] (19a) Ix, t g h x -- 1, x~l, x > 1,

(19b)

I1/x, ctgh x = [1,

x<~l, x > 1.

Straightforward integration gives, in this approximation,

(20)

1 3vN(0) - l + ~ m
-arctg

~cln 1;m* o + g In
~+arctg 1 m*flr lln 2 l+m*/~ 1 - m*~

1 In 1-m*2fl~l
2re*tic
1 7m'2~c

1
+ g In

~/-s ~
V ~ - ~

This complicated transcendental equation may be solved numerically in order to determine Tc but we shall instead use the following approach. It is evident that Tc is determined by ~, VN(O) and one additional parameter, namely the effective mass m*. We can thus obtain different solutions depending on the size of m*. It suffices to examine t h r e e distinct cases in terms of the dimensionless parameter m*flc: a) m*#e >> 1; b) m*flr ~ 1; c) re*fie << 1. The first case leads to

(21)

3VN(0)

~ _ O((m,flr

an evidently unphysical situation corresponding to the absence of a superconducting solution. In the case b), eq. (20) can be reduced by introducing the parameter 8 = I1 - m * ~ c I, where 8<< 1. Thus, for small 8, we obtain (22) - -1 ~ 1 3VN(0) = 4 5 In 2 + In o~flc+ 0(8). 2

For the case b), we thus determine the critical temperature to be (23) kB Tc ~ - - ~ 4 exp 1 4 1 3VN(0) "

If we assume[16] 0D~230K, VN(O)~ (1/3), then we obtain the value T r Thus we find that the case b) essentially reproduces the BCS-type result, within the order of magnitude. Now we shall turn to the case c) and show that this leads to a new solution. For

492

R. ACHARYA

and

P. NARAYANA SWAMY

m*flc << 1, eq. (20) reduces to (24) From eq. (24), we thus derive (25) kB Tr ~ ~ eo~exp 1 ~ 1 + In fl~(o 3VN(0) 2

3VN(0)

1]

'

where e is the transcendental number 2 + 1/2! + 1/3! + .... This expression, as in the BCS theory, is not analytical in the coupling constant, and has no power series expansion in ~eff, i.e. has an essential singularity at ~eff= 0. This means that the results are inaccessible to any method which depends on the perturbation theory. Equation (25) may be contrasted with the BCS result [1] (26) kB Tc ~ 1.14~oexp [

VN(O) 1 " 1

Hence, for VN(O)<~ 1/3, we obtain the gauge theory prediction (27)
-

(Tc)gauge > ~ - -1- - e a (Tc)BCS

2 1.14

~8.8.

Admittedly, this result has been made possible by the approximations employed and we should refrain from taking the numbers too seriously. The interesting conclusion, however, is that the gauge theory prediction for Tr is an order of magnitude higher than that of the BCS theory. For a Debye temperature 0D ~<230K[16], the BCS theory result is (28) (T~)BCS ~<13 K

which should be compared with the gauge theory prediction, for the same OD, (29) (Tc)gauge <~115 K.

This is well within the realm of high-Tc superconductivity [17, 18]. We may now return to the case of zero temperature by setting T - - 0 in eq. (17). The resulting equation involves integrals which can be evaluated exactly in closed forms. However, we shall be content to express the result in a series involving powers of Vm-7:
eo

(30)

1 3VN(O) =

1 fdr

j'

valid up to terms of order O(m *(s/z) ). This form is reminiscent of the framework of Migdal's theorem. Equation (30) can be calculated and we obtain the leading order result (31) 3VN(0)
1 ~ In 2 ~

A-

GAUGE TECHNIQUE AND BCS THEORY OF SUPERCONDUCTIVITY or

493

(32)

zl=2~exp[

3VN(0)I ]

which is the prediction for the gap function, or more precisely the zero temperature value, A(0). We can now combine this with eq. (25) and obtain (33)

A(O) kB Te

4 -- ~ 1.472, e

a result independent of the value of VN(O). This may be contrasted with the BCS theory result [ zl(0) ] ~ --~ ~ 1.764. k~ Jscs Y

(34)

It is amusing to note that eq. (33) is also in terms of a transcendental number. This work clearly demonstrates that the gauge technique, by going beyond the ladder approximation, can yield a high value of Tr even for weak coupling, Ze. ~< 1/3. In conclusion, we make the following interesting observation. The gauge theory is a nonperturbative method and is thus valid for all values of the effective coupling, ~e, = = VN(O). One is therefore free to consider the limit Ze,--~ 0% in contrast to the weak-coupling ladder approximation [19]. In this limit then, we obtain the saturated value of eq. (25), thus (35) Max (k B Tc) = ~ eo~,
1

which implies the limiting value (0D ~ 230 K): (36) Max (Te) = 312 K,

which allows the possibility of room temperature superconductivity. The extension of the present work to include a more realistic retarded pairing potential will lead to the generalized Eliashberg equations [20] which require extensive numerical computation and will be addressed elsewhere.

We thank Profs. A. Salam and J. Strathdee for their valuable comments.

REFERENCES [1] J. BARDEEN, L. COOPER and J. SCHRIEFFER:Phys. Rev., 108, 1175 (1957). [2] Y. NAlVIBU:Phys. Rev., 117, 648 (1960). [3] J. SCHRIEFFER:Theory of Superconductivity (W. Benjamin Inc., New York, N.Y., 1964), see especially p. 157, 232 and Chapt. 7; S. ENGELSBERGand J. SCHRIEFFER:Phys. Rev., 131, 993 (1963). 14] S. WEINBERG:Prog. Theor. Phys. Suppl., 86, 43 (1986), Festshcrift honoring Y. Nambu on his 65th birthday.

494

T. LETARDI, S. BOLLANTI,P. DI LAZZARO, F. FLORA, N. LISI and c. E. ZHENG

[5] P. W. ANDERSON: Phys. Rev., 112, 1900 (1958); N. BOGOLIUBOV,V. TOLMACHEVand D. SHIRKOV: New Method in the Theory of Superconductivity (Acad. Sci. USSR, Moscow, 1958); L. GORKOV:Soy. Phys. JETP, 7, 505 (1958); G. ELIASHBERG:Soy. Phys. JETP, 11, 696 (1960). [6] D. SCALAPINO,J. SCHRIEFFER and J. WILKINS: Phys. Rev., 148, 263 (1966). [7] G. MAHAN: Many Particle Physics (Plenum Press, New York, N.Y., 1990), 2nd edition. [8] A. MIGDAL: Soy. Phys. JETP, 7, 996 (1958). [9] P. ALLEN and B. MITROVIC: Solid State Phys., 37, 1 (1982). [10] P. LITTLEWOODand C. VARMA: Phys. Rev. B, 26, 4883 (1982). [11] A. SALAM: Phys. Rev., 130, 1287 (1963); R. DELBOURGO:Nuovo Cimento A, 49, 484 (1979). [12] See also R. ACHARYA and P. NARAYANA SWAMY: Phys. Rev. D, 26, 2797 (1982). [13] L. DOLAN and R. JACI~W: Phys. Rev. D, 9, 3320 (1974). [14] See, e.g., A. FETTER and J. WALECKA: Quantum Theory of Many Particle Systems (McGraw-Hill Book Company, New York, N.Y., 1971), p. 333. [15] C. TSUEI, D. NEWNS, C. CttI and P. PATTNAIK:Phys. Rev. Lett., 65, 2724 (1990). [16] M. SCADRON: Ann. Phys., 148, 257 (1983). See also, Y. NAMBU: From Symmetries to Strings, a Symposium to honor S. Okubo (World Scientific, Singapore, 1990), p. 1. [17] M. TINKHAlgand C. LOBB:Solid State Phys., 42, edited by H. EHRENREICHet al. (Academic Press, New York, N.Y., 1989), p. 91. [18] D. TILLEY and J. TILLEY:Superfluidity and Superconductivity, 3rd ed., Chapt. 11 (Adam Hilger Co., New York, N.Y., 1990). [19] We are of course aware of the fact that in this limit, one should really consider the coupled problem involving Z(p) and A(p). [20] M. CRISAN: Theory of Superconductivity (World Scientific, Singapore, 1989), p. 87.

Anda mungkin juga menyukai