PAIRING IN THE BOGOLIUBOV-de GENNES EQUATIONS

Yong-Jihn Kimy
Department of Physics, Purdue University, West Lafayette, Indiana 47907

Abstract
It is shown that the Bogoliubov-de Gennes equations pair the electrons in
states which are linear combinations of the normal states. Accordingly, the
BCS-like reduction procedure is required to choose a correct pairing. For a
homogeneous system, we point out that the kernel of the self-consistency equa-
tion derived from the Bogoliubov-de Gennes equations needs to be constrained
by the BCS pairing condition. In the presence of ordinary impurities, on the
other hand, the Bogoliubov-de Gennes equations should be supplemented by
Anderson's pairing condition to obtain the correct vacuum state by the cor-
responding unitary transformation. This results in localization correction to
the phonon-mediated interaction.

PACS numbers: 05.30.-d, 74.20.-z, 74.90.+n

y Present address: Department of Physics, Korea Advanced Institute of Science
and Technology, Taejon 305-701, Korea

1
1. Introduction
Recently it was shown1;2 that the Abrikosov and Gor'kov's3;4 Green's function theory
of impure superconductors is in serious con ict with Anderson's theorem.5 (Strong cou-
pling theory of impure superconductors was discussed elsewhere.6 ) For magnetic impurity
eects, Kim and Overhauser7 (KO) proposed a BCS type theory with dierent predictions
from the pair-breaking theory of Abrikosov and Gor'kov.3 In fact, there have been several
experiments8 12 which agree with KO's predictions. It was also pointed out13 that the failure
of Green's function theory comes from the intrinsic pairing problem in Gor'kov's formalism.14
The self-consistency equation needs to be supplemented by a pairing condition derived from
the physical constraint of the Anomalous Green's function. The resulting equation is nothing
but another form of the BCS gap equation. Then the discrepancy disappears.
In this letter we show that the same pairing constraint is required in the Bogoliubov-
de Gennes (BdG) equations.15 The equations are obtained by a real space version of the
Bogoliubov-Valatin (BV) transformation,16;17 which is a generalization18;19 of the BV trans-
formation. In this case, the pairing constraint is determined to nd a correct vacuum state
by the corresponding unitary transformation. The resulting pairing condition is the same
as that obtained from the physical constraint of the Anomalous Green's function.13
For a homogeneous system, choosing a constant pair potential gives rise to the BCS
pairing. However, the kernel of the self-consistency equation has not been xed by the BCS
pairing condition. In the presence of ordinary impurities, Anderson's pairing between time-
reversed states is not obtained from the BdG equations because the pair potential depends
on the position. Alternatively, pairing occurs between the states which are the linear combi-
nation of the scattered states. Consequently, the BdG equations predict incorrectly that Tc
doesn't change even if the scattered states are localized. Note that the linear combination
of localized states becomes extended one. Magnetic impurities will be considered elsewhere.

2. Homogeneous System

2
2.1 Bogoliubov-de Gennes Equations
Let's consider a homogeneous system. We follow rst de Gennes' derivation.15 The
Hamiltonian is given
Z X Z X
y(r)[ p~ 1
2
H = dr 2m
] (r) V dr y(r) y(r ) (r ) (r);
2
(1)
 

where the eld operator (r) is expanded by plane wave basis set ~k (r) = ei~k
r, that is,
X
(r) = ~k (r)a~k; (2)
~k
where a~k is a destruction operator for an electron with spin . Using the Gor'kov's
factorization14, we may get an eective Hamiltonian of the form
Z X
Heff = drf y(r)[ 2~pm ] (r) +
(r) y(r ") y(r #) +
(r) (r #) (r ")g;
2
(3)

where

(r) = V < (r #) (r ") > : (4)

(r) is called by the pair potential at the position r.

In order to nd the eigenstates and corresponding energies, we perform a unitary trans-
formation
X
(r ") = ( n"un(r) ny #vn (r));
n
X
(r #) = ( n#un(r) + ny "vn (r)); (5)
n
where the and y are quasiparticle operators satisfying the fermion commutation relations
y g=  ;
f n; m mn 

f n; m g = 0: (6)

By the transformation (5), the eective Hamiltonian may be diagonalized, that is,
X y ;
Heff = Eg + n n n (7)
n;

3
where Eg is the ground state energy of Heff and n is the energy of the excitation n. >From
the condition (7), we obtain the well-known Bogoliubov-de Gennes equations:

u(r) = Heu(r) +
(r)v(r);

v(r) = He v(r) +
 (r)u(r); (8)

where He = 2~pm2 . These equations can be written compactly in a matrix form

0 u 1 0 H
(r) 1 0 u 1
e
@ A = @  A@ A: (9)
v
(r) He  v
0u1 0 v 1
Notice that dierent eigenfunctions are orthogonal and @ A is orthogonal to @  A.
v u
Substituting Eq. (5) into Eq. (4) we nd
X

(r) = V vn (r)un(r)(1 2fn); (10)
n
where fn = exp(1 n )+1
: Eq. (10) is the self-consistency equation for the pair potential.

2.2 Vacuum State

Now we analyze the derivation. It is important to note that the unitary transformation
(5) is in fact a generalization18;19 of the BV transformation,

a~k" = uk ~k" + vk y ~k#;

a~k# = uk ~k# vk y ~k" ; (11)

where

u2k + vk2 = 1: (12)

Inserting Eq. (2) into Eq. (5) we obtain

X X
~k (r)a~k " =
0 0 ( n"un(r) ny #vn (r));
~k n
X 0
X
~k (r)a~k # =
0 0 ( n#un(r) + ny "vn (r)): (13)
~k 0 n

4
If we multiply the left side of Eq. (13) with ~k (r) and integrate over the position, we nd
X
a~k" = ( n"un;~k ny #vn;~k );
n
X
a~k# = ( n#un;~k + ny "vn;~k ); (14)
n

where
Z
un;~k = ~k (r)un(r)dr;
Z
vn;~k = ~k (r)vn (r)dr: (15)

Let's compare Eqs. (11) and (14). Only if both un(r) and vn(r) are proportional to the
normal state wavefunction ~k (r), then the transformation (14) becomes the same as the BV
transformation.
To understand the physical meaning of the unitary transformation (5), we express n"
by the creation and destruction operators for an electron. If we multiply un (r) to the rst 0

equation of Eq. (13) and integrate over the position, we nd

X X Z Z 
un ;~k a~k" =
0
n" un(r)un (r)dr ny # un (r)vn (r)dr :
0 0 (16)
~k n

By a similar process to the second equation, we also nd

X X Z Z 
vn ;~k a~yk# =
0 ny # un(r)vn (r)dr + n" vn(r)vn (r)dr :
0 0 (17)
~k n

Adding Eqs. (16) and (17) and by the orthogonality condition of eigenfunctions, one can
get
X
n " =
0 (un ;~k a~k" + vn ;~k a~yk# ):
0 0 (18)
~k
Similarly it is given
X
n # =
0 (un ;~k a~k# vn ;~k a~yk" ):
0 0 (19)
~k

Let's dene n (r) and n (r) by

5
 X
n (r) = U1 un(r) = U1 un;~k ~k (r); (20)
n n ~k

and
X
n (r) = 1 vn (r) = 1 vn;~k ~k (r); (21)
Vn Vn ~k
where
Z
jUnj2 = un(r)un(r)dr; (22)

and
Z
jVnj2 = vn (r)vn(r)dr: (23)

Notice that un(r) and vn(r) were expanded by the plane wave basis ~k (r) in Eqs. (20) and
(21). Then we nd
X
an = U1 un;~ka~k ;
n ~k
X
an = V 1 vn; ~k a~k: (24)
n ~k

Finally, by substituting Eq. (24) into Eqs. (18) and (19), we obtain

n" = Unan" Vnayn#;

n# = Unan# + Vnayn" : (25)

Note that we pair n(r)[= U1 un(r)] " and n (r)[= V1 vn (r)] # (instead of ~k (r) " and
n n

 ~k (r) #) by the unitary transformation (5). The generated vacuum state is

Y
~BdG = (Un + Vnayn"ayn# )j0 >; (26)
n

instead of the BCS ground state20

Y
~BCS = (u~k + v~k a~yk"ay ~k#)j0 > : (27)
~k

6
Therefore a pairing constraint is necessary for the unitary transformation (5) to generate
the BCS ground state; that is, both un(r) and vn(r) should be proportional to the normal
state wavefunction ~k (r) in order to pair ~k " and ~k #. For the current-carrying state, we
can pair ~k + ~q " and ~k + ~q #. Then, un(r) = U~k ei(~k+~q)
r and vn (r) = V~k ei( ~k+~q)
r.

2.3 Pairing Constraint on the Self-consistency Equation

The pairing constraint should be also supplemented in the self-consistency equation
X

(r) = V vn (r)un(r)(1 2fn): (10)
n
However, un and vn are usually expanded as a power series in
15

un = uon + u1n +


;
vn = vno + vn1 +


; (28)

where

uon = ~k ; vno = 0 (~k > 0);

uon = 0; vno = ~k (~k < 0); (29)

and
X
u1n = en~k ~k ;
0 0

~k
X 0

vn1 = dn~k ~k :

0 0
(30)
~k 0

en~k and dn~k are given

0 0

Z
(jnj ~k )en~k =
(r)~k (r)vno (r)dr
Z
0 0 0

(jnj + ~k )dn~k =
 (r)~k (r)uon(r)dr;

0 0 0 (31)

with ~k = h22m~k2 EF . In that case both un and vn are the linear combiantion of the normal
state wavefunction, which violates the pairing constraint. If we substitute Eq. (28) to the
self-consistency equation Eq. (10), we obtain
7
 Z

(r) = K (r; l)
(l)dl; (32)

where
X X ~k (l)~k (l)~k (r)~k (r)
K (r; l) = V T 0
: 0
(33)
! ~k~k (~k i! )(~k + i! )
0
0

The !'s are !n = (2n + 1)T for all integer n. Consequently, this equation should be
corrected by a pairing constraint. Then we nd
Z

(r) = K c(r; l)
(l)dl; (34)

where
X X ~k (l)~k (l)~k (r)~k (r)
 
K c(r; l) = V T 0
~k = ~k :0
(35)
! ~k~k (~k i! )(~k + i! )
0
0
0

Note that Eq. (32) is the Gor'kov's self-consistency equation

XZ

(r) = V T
(l)G! (r; l)G ! (r; l)dl: (36)
!
The revised self-consistency equation (34) can be written
XZ

(r) = V T
(l)fG! (r; l)G ! (r; l)gp:p:dl; (37)
!
where p.p. means proper pairing constraint which dictates pairing between ~k " and ~k #.
It was pointed out13 that Eq. (36) was obtained from the Anomalous Green's function
F (r; r0) which includes the terms violating the homogeneity constraint F (r; r0) = F (r r0).
Accordingly, the same pairing constraint may be obtained from the homogeneity constraint
of the Anomalous Green's function.13

3. Inhomogeneous System: Nonmagnetic impurity case

3.1 Correct Vacuum State
In the presence of ordinary impurities, Anderson5 proposed a BCS type theory which
employs time-reversed scattered state pairs. The exact scattered states n satisfy the equa-
tion
8
 [ p + U (r)] n (r) = n n(r);
2
(38)
2m
where
X
U (r) = V o  ( r Ri ) : (39)
i
fRig are the impurity sites. Accordingly, the ground state is15
Y
~Anderson = (un + vncyn"cny# )j0 >; (40)
n

where cyn# is the creation operator for an electron in the state n (r)j #>.
On the other hand, it has been claimed15 that the energy is lowered if we pair states ~ n
which are better choices than n by using the Bogoliubov-de Gennes equations. The state
~ n is basically a linear combination of the normal scattered states. The coupling comes
from the pair potential. However, we show that pairing ~ n " and ~ n # leads to the violation
of the physical constraint of the system.
The eective Hamiltonian is given
Z X
y(r)[ p~
2
H0eff = drf 2m + U ( r)] ( r ) +
( r) y(r ") y(r #) +
 (r) (r #) (r ")g; (41)


where

(r) = V < (r #) (r ") > : (42)

As in Sec. 2.1, the unitary transformation

X
(r ") = ( n"un(r) ny #vn (r));
n
X
(r #) = ( n#un(r) + ny "vn (r)); (43)
n

leads to the following Bogoliubov-de Gennes equations

u(r) = [He + U (r)]u(r) +
(r)v(r);

v(r) = [He + U (r)]v(r) +
 (r)u(r): (44)

9
 To nd the vacuum state for particles, we expand the eld operator by the scattered
states:
X
(r) = n (r)cn : (45)
n
Then it can be shown
X
n" = (un;n cn " + vn;n cyn # );
0 0 0 0

n
X 0

n# = (un;n cn # vn;n cyn ");

0 0 0 0 (46)
n 0

where
Z
n (r)un(r)dr;
un;n =0

Z
0

n (r)vn (r)dr:
vn;n =0
 0
 (47)

For the states ~ n and ~ n dened by

X
~ n (r) = U1 un(r) = U1 un;n 0
n (r);
0

n n n 0

X
~ n (r) = V 1 vn (r) = V 1 vn;n 0
n (r);
0 (48)
n n n 0

it is given
X
bn = U1 un;n cn  ; 0 0

n n 0

X 
bn = V 1 vn;n cn  : 0 0 (49)
n n 0

bn is the destruction operator for an electron in the state ~ n .

Finally, we obtain

n" = Unbn" Vnbyn# ;

n# = Unbn# + Vnbyn" ; (50)

and
Y
~BdG = (Un + Vnbyn" byn# )j0 > : (51)
n

10
Note that the Bogoliubov-de Gennes equations, Eq.(44) correspond to the vacuum state
where ~ n (r)[= U1 un(r)] " and ~ n (r)[= V1 vn (r)] # (instead of n (r) " and n (r) #) are
n n

paired.
Now we must decide which is the correct ground state in the presence of impurities.
Above all, the correct ground state should satisfy the physical constraint of the system. If
we average over the impurity positions, the system becomes homogeneous. Consequently,
the pair potential should be a constant after the impurity average, i.e.,

(r)imp = constant: (52)

imp means an average over impurity positions R~ i .

Let's rst check the constraint for the state ~BdG . Eq. (42) leads to the pair potential
at T = 0,
X

(r) = V vn (r)un(r)
nX
= V UnVn~ n(r)~ n (r): (53)
n
In terms of the scattered states, it is rewritten
XX

(r) = V un;m m (r)vn;m 0
m (r):
0 (54)
n m;m 0

Subsequently it is given

(r)imp  imp
m (r) m (r) :
0 (55)

Note that
imp X 1
m(~k)" (r) m (~k )# (r)
0 0 = ei(~k+~k )
r[1 + Vo2
0

(~k ~k+~q)(~k ~k ~q) +


]

~q;i 0 0

6= constant; (56)

and
imp X 1
m(~k)" (r) m ( ~k)# (r) = [1 + Vo2 (~k ~k+~q)2 +


]
~q;i
= constant: (57)
11
In deriving these relations, we used the scattered states m(~k) (r),
i~k
r X Vo [X e i~q
R~ i ]ei(~k+~q)
r +


:
m(~k) (r) = e + (58)
~q ~k ~k+~q i
Consequently the state ~ BdG gives the `averaged' pair potential

(r)imp 6= constant; (59)

which violates the physical constraint of the system.

>From Eq. (57), it is clear that the state ~ Anderson gives the correct `averaged' pair
potential which satisfy the physical constraint of the system. Therefore the correct ground
state is ~ Anderson. To obtain ~ Anderson from the Bogoliubov-de Gennes equations, we need
a pairing constraint:

~BdG = ~Anderson; (60)

which gives

un(r) / n(r);
vn (r) / n (r): (61)

3.2 Localization Correction

It has been a common practice to assume a constant pair potential to derive Anderson's
theorem.15;21;22 For the constant pair potential
o , we can also choose

u(r) = n(r)un;
v(r) = n(r)vn: (62)

Then Eq. (44), or Eq. (10) leads to

X

o = V unvn n(r) n (r)(1 2fn): (63)
n
To be consistent, the right hand side is substitituted by the impurity average of the square
of wavefunction, i.e.,
12
 Z

o = V No dnunvn (1 2fn); (64)

where
X
No = (n) n(r) n (r)imp: (65)
n

Eq. (64) is the same form as that of the homogeneous case. Consequently, the transition
temperature doesn't change in the presence of nonmagnetic impurities, which is called by
Anderson's theorem.
However, we should not assume a constant pair potential to derive Eq. (62), which is
just Anderson's pairing constraint. Then, we obtain the self-consistency equation
X

(r) = V unvn n (r) n (r)(1 2fn); (66)
n

instead of Eq. (63) which is inconsistent. Multiplying the both sides of Eq. (66) by
m (r) m (r), one nds that
 

X

m = Vmn unvn(1 2fn); (67)
n

where
Z
m (r) m (r) n (r) n (r)dr:
Vmn = V   (68)

And it is given

un = 12 (1 + q 2n 2 );
n +
n
vn = 21 (1 q 2n 2 ): (69)
n +
n
Comparing Eqs. (67) and (64), we nd that Anderson's theorem is valid only when Vmn
is not much dierent from V , which is the eective interaction without impurities. It was
pointed out that this quantity is almost the same as V up to the rst order of the impurity
concentration.1;2

13
 Now it is clear that localization correction is important in Eq. (68). For the strongly
localized states, the eective interaction is exponentially small,1;6 like the conductance.23;24
It is, then, expected that the same weak localization correction terms occur both in the
conductance and the eective interaction. Recently weak localization correction to the
phonon-mediated interaction was reported.13 The results are the following:

Vnn3d  1 `
= V [1 (k `)2 (1 L )];
0 (70)
F

Vnn2d  2
= V [1 k ` ln(L=`)];
0 (71)
F

Vnn1d  1
0 = V [1 (k a)2 (L=` 1)]; (72)
F
where ` and L are the elastic and inelastic mean free paths and a is the radius of the wire.
For thin lms, the empirical formula is given25
Tco Tc / R ; (73)
sq
Tco
where Tco is the unperturbed value of Tc and Rsq is the resistance of a square sample. Notice
that this formula is obtained if we substitute Eq. (71) into the BCS gap equation. More
details will be published elsewhere.

4. Conclusion
It is shown that the Bogoliubov-de Gennes equations need a pairing constraint to obtain
a correct vacuum state by the corresponding transformation. The constraint is the same as
that obtained from the physical constraint of the Anomalous Green's function.

Acknowledgments
I am grateful to Professor A. W. Overhauser for discussions. The early version of the
paper was circulated in the U.S.A., Japan, and Korea from early 1995 to mid 1995. This
work was supported by the National Science Foundation, Materials Theory Program.
14
 REFERENCES

1. Yong-Jihn Kim and A. W. Overhauser, Phys. Rev. B47, 8025 (1993).

2. Yong-Jihn Kim and A. W. Overhauser, Phys. Rev. B49, 12339 (1994).
3. A. A. Abrikosov and L. P. Gor'kov, Sov. Phys. JETP 12, 1243 (1961).
4. A. A. Abrikosov and L. P. Gor'kov, Phys. Rev. B49, 12337 (1994).
5. P. W. Anderson, J. Phys. Chem. Solids 11, 26 (1959).
6. Yong-Jihn Kim, Mod. Phys. Lett. B10, 353 (1996).
7. Yong-Jihn Kim and A. W. Overhauser, Phys. Rev. B49, 15779 (1994).
8. M. F. Merriam, S. H. Liu, and D. P. Seraphim, Phys. Rev. 136, A17 (1964).
9. G. Boato, M. Bugo, and C. Rizzuto, Phys. Rev. 148, 353 (1966).
10. G. Boato and C. Rizzuto, (to be published), (referenced in A. J. Heeger, (1969), in
Solid State Physics 23, eds. F. Seitz, D. Turnbull and H. Ehrenreich (Academic
Press, New York), p. 409)
11. W. Bauriedl and G. Heim, Z. Phys. B26, 29 (1977); M. Hitzfeld and G. Heim, Sol.
Sta. Com. 29, 93 (1979).
12. A. Hofmann, W. Bauriedl, and P. Ziemann, Z. Phys. B46, 117 (1982).
13. Yong-Jihn Kim, Mod. Phys. Lett. B10, 555 (1996).
14. L. P. Gor'kov, Sov. Phys. JETP 7, 505 (1958); ibid. 9, 1364(1959).
15. P. G. de Gennes, Superconductivity of Metals and Alloys (Benjamin, New York, 1966).
16. N. N. Bogoliubov, JETP USSR 34, 58 (1958).
17. J. G. Valatin, Nuovo Cim. 7, 843 (1958).
18. C. Bloch and A. Messiah, Nucl. Phys. 39, 95 (1962).
19. Fan Hong-Yi and J. VanderLinde, J. Phys. A: Math. Gen. 23, L1113 (1990).
20. K. Yosida, Phys. Rev. 111, 1255 (1958).
21. P. G. de Gennes and G. Sarma, J. Appl. Phys. 34, 1380 (1963).
22. M. Ma and P. A. Lee, Phys. Rev. B32, 5658 (1985).

15
23. N. F. Mott and M. Kaveh, Adv. Phys. 34, 329 (1985).
24. P. A. Lee and T. V. Ramakrishnan, Rev. Mod. Phys. 57, 287 (1985).
25. B. I. Belevtsev, Sov. Phys. Usp. 33, 36 (1990).

16