SOVIET PHYSICS JETP VOLUME 34 (7), NUMBER 1 JULY, 1958

A NEW METHOD IN THE THEORY OF SUPERCONDUCTIVITY. III

N. N. BOGOLIUBOV
Mathematics Institute, Academy of Sciences, U.S.S.R.
Submitted to JETP editor October 17, 1957
J. Exptl. Theoret. Phys. (U.S.S.R.) 34, 73-79 (January, 1958)

The Hamiltonian of Bardeen is analyzed in this paper and the method of summation of the prin-
cipal diagrams is applied. It is shown that the same results are obtain in this way as in Refs.
1 and 2.

RECENTLY great successes have been attained In the Frohlich model, we must set2
in the solution of problems of statistical physics
by means of summation of the principal diagrams.
In the present paper we shall show that in the We shall take into account the value N of the
theory of superconductivity, we can also obtain (by total number of electrons by means of the chemi-
this method) those results which were found in pre- cal potential ;>.., for which purpose we add the term
vious researches 1 •2 with the aid of a canonical - ;>..N to HB. Then we obtain the Hamiltonian
transformation and the principle of compensation
of diagrams with "dangerous" energy denominators.
H =r= H 0 +Hint.
As has been shown by Tolmachev and Tiablikov ,2
we can consider the Hamiltonian of Bardeen in H o =] {£ (k)- ),} a:Saks.
k, s
place of that of Frohlich, since they both, up to a
known degree, give equivalent results for the effect
of electron-phonon interaction on the dynamics of
electrons close to the Fermi surface. In our case
the Bardeen Hamiltonian is considerably simpler.
Therefore, for a more graphic description and
to establish the connection with the ideas of the
work of Bardeen, Cooper, and Schrieffer, 3 we shall for which we shall also consider the question of
start out from the Hamiltonian of Bardeen: the_ summation of the principal diagrams.
Since the interaction is effective only in a small
Ha = ~ E (k) aitaks region of t.he Fermi sphere and only between par-
k,s ticles (electrons or holes) with oppositely directed
spins, we see that a very important role will be
played by diagrams of the type shown in Fig. 1.
These diagrams were constructed from an "irre-
ducible complex" (see Fig. 2), consisting of a pair
of particles with momenta ± k and spiRs ± ~.
where

I' E (kp) -w<E(k) <E (kp) +w

6 (k) = { 0, \E (k) -E (kp) \>w
FIG. 1

I, k=O,
{
/::,. (k) = 0, k =I= 0

and where E (k) is the radially symmetric func- FIG. 2

tion representing the energy of an electron of mo- To sum the diagrams, we make use of the
mentum k; I and w are the Bar de en parameters. method of approximate second quantization, i.e.,

51
52 N. N. BOGOLIUBOV

we construct a simplified Hamiltonian, for which We proceed to the construction of an asymp-

the diagrams will only be of that class which we totically exact solution of this latter problem, neg-
desire to sum, and furthermore, with the same lecting only the quantities which vanish in the limit:
contribution which is made in the present Hamil- v- 00,
tonian.* We shall distinguish between pairs of electrons
Since the complexes of pairs of particles ( ± k, ±!) and pairs of holes, for which purpose we intro-
are not broken up .in the diagrams considered by us, duce the new Pauli operators, setting
it is natural to compare their quantum amplitudes
~~<= bt, E(k)<A,
~. ~ with the commutation relations
~k = bk, E (k) >A.
[bk, bk'] = 0, [bt, bt] = 0, [b;i' bk,] = 0; k =I= k'. ( 2)
We get
Furthermore, since there do not exist several pairs
with the same value of k, we must have
H = U +.2 ~IE (k) -AI ~t~k
( 3) k

We note further that the eigenenergy of the com-

- ~ ~ fJ (k) 9 (k') {fJo (k) ~t
plex will be k+k'

+ 9p (k) ~k} {6 0 (k') ~k' + fJp (k') ~;i;}, ( 5)

and that the matrix element of the Hamiltonian ( 1)
for the transition k - k' will be proportional to where
-I/V.
From these considerations, we obtain a simpli-
I, E(k)<A
fied Hamiltonian of the form {
fJp (k) = 0, E (k) >A
H=Ho"+Hint. Ho = ~2(E(k)-A)btbk, U = 2] {E (k)- A} 9p (k), 60 (k) + fJp (k) = I.
k
k

Hint=--~ ] btbk,fJ (k) fJ (k'),

(4)
Let us consider the wave function C for which
(k+k') all the filling factors

which contains the operators hk· b~ [with the

nk.= ~t~k
commutation relations ( 2), ( 3)] which we shall are equal to zero. Then
call the Pauli operators.
Taking expressions of arbitrary order
~kc = o.
We shall show that this wave function is an as-
Hint (Ho- Efl Hint ••• (Ho- E f1Hint.
ymptotically exact eigenfunction of the Hamilton-
it is now easy to verify directly that the sum of ian H, giving it the value U.
contributions from the diagrams of the type con- In fact, we have
sidered for the Hamiltonian ( 1) will be equal to
the sum of contributions of all diagrams for the '•
simplified Hamiltonian ( 4). H = H' + H" + U,
Thus, the problem of the summation of a spe-
cial class of diagrams for the Hamiltonian ( 1) is H' =2 ~IE (k)- AI ~t~k- ~ ] fJ (k) fJ (k') {fJo (k) ~t
k k+k'
shown to be equivalent to the problem of the model
of the dynamical system that is characterized by
the Hamilto:qian {4) . + fJp (k) ~k} 60 (k') ~k'- -~ ~ fJ (k) fJ (k') fJp (k) fJp (k') ~t~k·
h+h'

*We emphasize that the new meaning of "summation" in- H" = - ~ ~ fJ (k) 9 (k') 90 (k) fJp (k') ~t~;t.
troduced here cannot be taken in the universally accepted h+k'

sense. Strictly speaking, we do not sum here a series of terms

of a given class, bu! we compare the Hamiltonian for which
the expansion of perturbation theory exactly coincides with
But, obviously,
the given series. From. the mathematical viewpoint, we are H'C = 0.
dealing here with concepts close to those of the theory of
quasi-analytic functions. On the other hand,
 NEW METHOD IN THE THEORY OF SUPERCONDUCTIVITY 53

+ or
<C* I H" /2C) = <C* H" H"C> =
I2
= V2 ~
< const when V->- :x>. "' k2
LJ 6 (k) 0 (k') 00 (k) Op (k') 1= _£_ \ dk 2z dz
k+k' 2 j dEz 2 -E 2 j4'
0

But, in the limit V - 00 , H must be propor-

tional to V, while Ill" I~ as we have noted, re- where
mains finite. Therefore, in fact, C is an asymp-
totically exact eigenfunction of H, giving it the
_
p - 2n:2
I (k (ifF
dk)
2
k=kF • ( 8)
value U.
We have also As is seen, this equation, in the case ( 7) under
consideration, always has a negative root for E 2 •
N = (C*NC) = ~ 2.
E (k) <I. Consequently we obtain a purely imaginary value
for the energy E:
Equating this expression to the total number of
electrons in the Fermi sphere E ~ + i2we-IIP. ( 9)

~ 2,
E (k) < EF Thus the state C is found to be unstable.
In order to find the stable ground state with
we see that minimum energy, we introduce the new Pauli am-
/.. = EF = E (kp).
plitudes f3k, {3fc in non-trivial fashion, as earlier,
with the aid of the relations
We now analyze the problem of the stability of the
state C. We consider first the case in which
I<O. ( 6) ( 10)

We supplement the double sum in ( 5) with where Uk and Vk are real numbers satisfying
terms for which k = k', terms that make no con- the relation
tribution in the transition to the limit V- oo, We u~ + v~ = 1. ( 11)
then note that H - U is essentially a positive
form. The value U will consequently be a min- It is not difficult to note that the amplitudes ( 10)
imum and the state C will be stable from the actually satisfy all the commutation relations of
same considerations. the Pauli operators.
The situation will be different in the case Turning to the transformation ( 10), we find:

I>O. ( 7) bk = Ut..Vk ( 1 - 2~;~k) + u~~ - v~~;,

We note that since all the filling facts nk =

b[ = ukvk ( 1 - 2~;~~<)- v~~k + u~~t, ( 12)

(3~(3k in the state C will be equal to zero, we b'tbk = v~ + (u~- v~) ~ii~k + uhvh (~k + ~ii).
can, upon computation of the energy of the elemen-
tary excitations, consider the Pauli operators (3, Substituting these expression in the Hamiltonian
13+ to be Bose in character. ( 4 ) , we find:
There then remains only the diagonalization to
quadratic form of the operators (3, 13+ which rep- H = U + ~{2(E(k)-J..)ukvk
resent H - U of ( 5). This diagonalization can be
achieved, for example, with the aid of a method
set forth in our monograph.4 - ~ 6 (k) (u~- v:) ~ uk'v",6 (k')} (~k + ~ii)
k'
For the determination of the energy E of the
elementary excitation, we get the following secu- + ~ 2Ee (k) ~~~k- ~ ~ (ut~:,- v~;~k,- 2uk,vk,~i,~k,}
lar equation: k k,+k•

I ~ { eF (k) 60 (k) }
1= V f 6 (k) ek-E + ek + E '

ek = 2J E (k) - E (kF) /, where

whence, upon simplifying, we get
1=__£_ ~ ~- U = ~ 2 (£ (k) - /..) V~- ~ ~ 6 (k1) 6-(k2) Uk,Vk,Uk,Vk,;
v lEp< E (k) <Ep+O>) e~- £2 ' k (k,, k,)
54 N. N. BOGOLIUBOV

E. (k) = (E (k)- A) (u~- vV + 6 (k) ukvk 2~ L 6 (k') uwvh'· (H 0 -f-H")C=0

k' ( 15)
and
We let the coefficients for Wk + ,B~ ) in Eq.
( 13) vanish, and get

2 (E (k)- A) ukv"- ~ ll (k) (u~,- vV ~ 0 (k') uk'vh' = 0,

k' ( 16)
which was found in Ref. 2 with the help of the prin-
< const, V _,. oo.
ciple of compensation of dangerous diagrams.
Noting that A= E (kF) (with the accuracy re- We now proceed to the consideration of elementary
quired here), we have, just as in Ref. 2: excitations. Since all the filling factors vk in the
state C are equal to zero, we can consider the
1 { E (k)- E (kp) } Pauli operators to be Bose operators in the calcu-
uz (k) = 2 1+ V<E (k)- E (kp)) + 6 (k) c• ' 2 lation of the energy of the elementary excitations.
1 { E (kp)- E (k) }
Therefore, in the equation for the Hamiltonian ( 18)
vz.(k) = 2 1 + y (E (k)----;- E (kp)) 2 6 (k) + c• ' we can limit ourselves to the quadratic form H0
+ H'.
C = 2we-l/P,
Carrying out the diagonalization by the method
E. (k) = V<E (k)- E (kp)) 2 + (j (k) C2 • described earlier, 4 we obtain a set of linear equa-
( 17) tions:
The Hamiltonian ( 13) can now be written in the
form
k'
H = U + H + H' + H" 0

where
- (E + 2Ee (k)) xh = U! ~ 6 (k) ~ {u~,xk'- v~,<pk,} 6 (k')
k'

H 0 = ~ 2Ee (k) ~;i~k• - ~ a(k) v~ ~ {u%,'f'k'- v~,xk,} ll (k')

k ( 19)
k'

H' = - ~ ~ ll (k 1 ) ll (k 2) {uf,,~t,
(k,+k,)
with the normalization condition

~{I cp" 12 -I xk 12 } = I. (20)

( 18) k

Then we obtain the secular equation:

H" = ~ ~ a(kl) a(kz) {u~.~k,- vt~ii;
(kdh,)
J 1.( u~ v~ . l
l 1 + V ~ \2Ee+
I
E + 2Ee-E )o (k)J
u! ) }
X {1 + VI ~ (' 2Eevf,+ E + 2Ee- E O(k)
+~ ~ a(k1) 6 (k2) {ut~t,- v7,,~
(k 1 +kz) -{f~G(k)u!v~(2E ~E +2£ ~£)}2=0.
e e (21)

It is easy to see that for

IE I< 2 min Ee (k) = 2Ee (kF)
We select the wave function C for which all the
filling factors this equation has no solution, since the subtrahend
in Eq. (21) is then l_ess than the minuend.
For
IE I> 2 min Ee (k)
are zero. We shall show, as before, that, with ac-
curacy to quantities which vanish in the limit v we have a continuous spectrum
- oo, C is ~ eigenfunction of the Hamiltonian H,
giving it a value U. We actually have E= -:J~ 2Ee (!<) + 0 (! jV), 0 (I jV) _,.Q when V _,. oo.
 NEW METHOD IN THE THEORY OF SUPERCONDUCTIVITY 55

As is seen from ( 19), the minus sign does not over, it achieves various generalizations, for ex-
agree with the normalization condition (20·). ample, in the calculation of thermodynamic quan-
Thus all the E are positive (this can be seen tities.
directly from the fact that the quadratic form In conclusion, I should thank D. N. Zubarev, V.
under consideration is positive definite) and are V. Tolmachev, S. V. Tiablikov, and Iu. A. Tserkov-
separated from zero by the gap nikov for their valued discussion.
E = 2Ee (k) :;?-:- 2£. (kF) = 2C = 4we·-l/P. (22) 1 N. N. Bogoliubov, J. Exptl. Theoret. Phys.
Here again we obtain the results of Bardeen as (U.S.S.R.) 34, 58 (1958), Soviet Phys. JETP 7, 41
in the previous papers . 1 •2 ( 1958) (this issue).
Since we have confined ourselves only to dia- 2 V. V. Tolmachev and S. V. Tiablikov, J. Exptl.
grams consisting of pairs, we cannot decide di- Theoret. Phys. (U.S.S.R .. ) 34, 66 ( 1958 ), Soviet
rectly from (22) that the excitation with energy Phys. JETP 7, 46 ( 1958) (this issue).
2Ee (k) (Eq. ( 22)] consists indeed of two excita- 3 Bardeen, Cooper, and Schrieffer, Phys. Rev.
tions of the Fermi type, which was shown in Ref. 1. 106, 162 (1957).
As we see, the method of summation of dia- 4 N. N. Bogoliubov, JleK.Qii 3 KBaHTOBoi cTaTHCTHKH
grams is shown to be quite lucid and permits us to (Lectures on Quantum Statistics), (Kiev, 1949).
establish the connection with the ideas of the work In Ukrain ian.
of Bar de en, Cooper, and Schrieffer.
However, in our opinion, the method of canoni-
cal transformation is more flexible, allowing us Translated by R. T. Beyer
easily to obtain the higher approximations. More- 12

SOVIET PHYSICS JETP VOLUME 34 (7), NUMBER 1 JULY, 1958

BREMSSTRAHLUNG OF PI MESONS AND PRODUCTION OF PI-MESON PAIRS BY

GAMMA QUANTA IN COLLISION WITH NONSPHERICAL NUCLEI

I. T. DIA TLOV
Leningrad Physico-Technical Institute, Academy of Sciences, U.S.S.R.
Submitted to JETP editor October 22, 1957
J. Exptl. Theoret. Phys. (U.S.S.R.) 34, 80-86 (January, 1958)

The effective cross-sections are calculated for a number of radiative processes occurring
in the interaction of high-energy 1r mesons with nonspherical nuclei. The nonspherical
shape of the nuclei leads to a change of the angular distributions and the appearance in the
cross-sections of factors that depend only on the geometrical shape of the nuclei.

IN papers by Landau and Pomeranchuk, 1 Pomer- that very large distances from the nucleus ( reff
anchuk,2 and Vdovin3 treatments have been given ....., E/1-' 2 » R) contribute to the matrix elements
of the processes of bremsstrahlung in the inter- that give the probabilities of the processes, and
action of 1r mesons with nuclei, production of one can use in the calculation the asymptotic form
1r-meson pairs from nucfei by 'Y quanta, and pro- of the 1r-meson wave functions outside the region
duction of nuclear stars by 'Y quanta, for very of their interaction with the nucleus. At large en-
large 1r-meson energies E and y-quantum ener- ergies one can take these functions to be diffrac-
gies w ( E » ,.,.; w » ,.,., where ,.,. is the mass tion functions. The functions used in the papers
of the 1r meson; we set n = c = 1 throughout). A mentioned are those of the diffraction by a black
peculiarity of these processes at such energies is or a gray sphere.