Nuclear Physics North-Hollandpublishing Co., Amsterdam: Abstract

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

1.D.2 [ Nuclear Physics All0 (1968) 241--256; (~ North-HollandPublishing Co.

, Amsterdam
N o t to be reproduced by photoprint or microfilm without written permission from the publisher

COMPARISON BETWEEN A BOSON-EXPANSION M E T H O D


AND THE EXACT SOLUTION IN A TWO-LEVEL M O D E L

R. A. B R O G L I A and B. S O R E N S E N
The Niels Bohr Institute, University of Copenhagen, Denmark

Received 24 November 1967


Abstract: A detailed comparisonis made, betweenthe boson method developedby Sorensenfollowing
the ideas of Beliaevand Zelevinskyand an exact solution of a two-levelsystemof like particles
interacting via a pairing force.

1. Introduction

The increasing interest in the study of complex nuclear excitations of essentially


non-harmonic character has stimulated the development of non-linear boson methods
(refs. 1, z)). Such methods will, at least in their applications, involve approximations, the
nature of which can usually be interpreted as an emphasis on certain correlation effects
accompanied by the negligence of others. It may be useful to support the underlying
physical intuition with the confrontation of the predictions of a given method with an
exact result, and thus some attention should be paid to the search for solvable models
which contain enough structure to yield non-trivial answers, especially concerning
the questions of the relative importance of various anJaarmonic effects. We feel that
one model, which may be useful in this sense, is the two-level model with state-
independent pairing force 3). The model contains a region with anharmonic features,
namely when the force strength has almost the critical value causing the transition
between superconductive and normal phase. If one further considers the spectra
corresponding to different numbers of particle pairs present, it is possible to calculate
the two-particle transfer amplitude, a quantity which is a natural measure of the
correlations produced by the pair field and which thus can provide a test of the
wave functions involved. In this way, the regions of validity of the RPA based on
normal and superconductive representations has been studied 4). Here we shall be
concerned with some basic features of a particular non-linear boson method which
has been investigated by one of the authors 2,s). This method is based on the particular
type of boson expansion suggested by Beliaev and Zelevinsky 1) (referred to as the
B-type expansion), which is constructed in such a way that it satisfies those impli-
cations of the Pauli principle which can be expressed through commutation relations
among pair operators, but at the same time causes fluctuations in the matrix elements
of other constraints imposed by the Pauli principle 5). The usefulness for realistic
241
March 1968
242 R. A. BROGLIA AND B. SORENSEN

calculations is achieved through two further features, one being a collectivity ap-
proximation (which decreases the number of degrees of freedom) and the other being
a prescription for a self-consistent determination of the "best" boson basis. The
simplicity of the two-level model, when only seniority-zero states are considered,
allows us to avoid the use of additional approximations so that we can provide a
test of the B-type expansion separately t.
In sect. 2, the model is defined and, in sect. 3, the boson expansion built on the par-
ticle-hole fermion basis is presented. In sect. 4, we discuss the numerical results and,
in sect. 5, boson expansions built on other fermion representations are discussed.
Sect. 6 contains concluding remarks.

2. The two-level model

We consider a system of fermion pairs, each consisting of particles occupying


time-reversed orbits. Two j-shells are available, and we introduce the operators

A~= (ai+ a,+ )o, N + = (a+~)0, i = 1, 2, (2.1)


where ai = z a ( c - 1 differs from as by a phase factor (_)~,+m, and by a change in
sign of the projection of the angular momentum. No further operators are necessary
for the description of seniority-zero states in the model. Assuming a pairing inter-
action with constant matrix elements, the Hamiltonian for the system becomes

= Az)(,]1Al-l-)2A2), (2.2)

where the symbol j stands for (2j+ 1)~. Following H6gaasen-Feldman a), we in-
troduce an orthonormal basis for the states with n pairs present

Ira, n - m ) = ~fm
- t (At)
+ m(A2)
+ ,,--,,, 10>. (2.3)
The two-particle transfer operator is some linear combination of A + and A~, but
since we do not want to introduce the kinematics of the process, we shall use a
transfer operator which weights equally the two levels (which is the obvious choice
when the quantum numbers of the two levels are equal and the assumption of a
similar radial shape of the two shells is made)

T = 2--~(A + +A+), (2.4)


Exact diagonalization and evaluation of transition matrix elements can now be
performed a,4). The lowest excited state is the pairing vibration 7). The BCS model
predicts the transition to the superconductive phase when G exceeds the critical value
4z
Gc - (Jr +)2) 2 . (2.5)
The relation of the present work to the general development of the boson-expansion method has
been presented recently 6).
BOSON-EXPANSION METHOD 243

3. The B-type boson expansion from particle-hole basis


The operators (2.1) characterizing the system are now expanded t in terms of
pure, monopole boson operators b + = b&=o(j~ji
) . +" "
The expansions obtained from the requirement that all commutation relations valid
among the operators (2.1) are reproduced by the boson power expansion series are

A + = - ( 2 - 4y°)~bl + 2 {1 1 }
)2(2_4yO)~ 431(2_4yi)^2
o +''" b+bibl

+ ^4 2 o - ~ { i + . .}b+b+blblbl+
. . . ., (3.1a)
Jl(2-4yi)

A~ = (2--4y °) b + ^2"
2 {
0 ~- 1
1 +
} b+2 +b 2 b2
dE(2_4y2) ^2 - 4 y 2o)
4j2(2 """
2
.~24(2_a,,%}
=-~*, {1 + . . .}b~ b 2+ b 2+ b 2 b 2 -b .... (3.1b)

Ni -- 3 i ( 1 - y ° ) - 2 b+bl ' (3.1c)


Ji

N2 -- ) 2 y ° + 2 b~b2. (3.1d)
J2

For convenience, we have given the boson operator bl hole character and b2 particle
character. The expansions (3.1.) contain two parameters yO0 1 = 31yO and yOO = 3ayO
which give the occupation probability of the single-particle levels in the vacuum of
the boson operators. Particle number conservation for the boson ground state requires
yOO = yOO. We shall tt use yOO = yOO = 0. The particular simplicity of the monopole
model has two effects on the form of the boson expansions. One is that the expan-
sions o f A + and N1 only depend on bi and those of A + and N 2 only on b 2. The
other is that the expansions of N 1 and N 2 are finite, a fact which is connected with the
neglect of higher-order constraints imposed by the exclusion principle (to be discussed
below). The induced expansions of H and T are cut at a finite order. We here consider
a truncation of H at fourth order (terms with up to four boson operators retained)
and of T (consistently) at third order. The corresponding expansions are given
explicitly in ref. a).
The basis in which the boson Hamiltonian is diagonalized is
In, k ) = {(k + p)!(k +q)!}-~(b+)k+'(b+)k+qtO), (3.2)
for a system with n = (Jl + ½+ P - q) pairs.
* This is only possible for the B-type b o s o n expansions. T h e exact b o s o n transcription f o r m u l a t e d
by M a r m u m o r i et al. 1) requires b o s o n operators o f all t h e possible total a n g u l a r m o m e n t a J.
tt This choice, w h i c h implies identity between t h e fern-don v a c u u m a n d the m o n o p o l e b o s o n
v a c u u m , is t h e natural one w h e n only seniority zero states are considered.
244 R.A. B R O G L I A A N D B. SORENSEN

The incomplete antisymmetrization of the boson images of many-fermion states,


which is implied by the use of the B-type expansions 5), can also be visualized in
terms of the allowance for fluctuations in the values of the boson images of some
many-fermion operators, such fluctuations being in fact essential for obtaining rapidly
converging expansions. The inaccuracy inherent in this method can thus be described
as the introduction of spurious states generated by the operator differences of these
constraints from their average values. Via the basic B-type plus the truncation errors
involved in the expansion of H, there may in this way occur an admixture of spurious
components in the physical states. Within the monopole model with complete particle
number conservation considered here, a number of the constraints imposed on the
B-type expansions are automatically fulflUed, and only constraints of high order remain,
the most important being the non-vanishing of the boson images of the operators
(A+) ja+~, (A~-)j2+~. (3.3)
A direct implication of these constraints is that the physical eigenstates should have
only components of those basis states (3.2) which satisfy

(k+p) < Jl+~, (k+q) < j2+~. (3.4)


In order to check whether the physical eigenstates are orthogonal to the spurions
generated by the operators (3.3), we performed the diagonalizations in a space which
was larger than the one given by eq. (3.4), namely allowing k to exceed these limits
by a few units. The amplitudes of such non-physical basis states gives a measure of
the admixture of spuriosity.
In order to decide whether such spuriosity is implied by the B-type expansion
directly or rather by the finite truncation of the H-expansion, one should in principle
vary the truncation point. However, the particular symmetry of the representation of
H for the symmetric model (Jl = J2) implies an exact orthogonality between physical
and spurious states for the closed-shell system (p = q = 0). In this case, we can
check the truncation error independently, and for p or q ~ 0 estimate the influence of
the spurion admixtures which are directly due to the use of B-type expansions.

4. Numerical comparison
In figs. 1-4, the energies of the exact solution are compared with those obtained
with the fourth-order boson expanded Hamiltonian discussed in the previous section.
The basis (3.2) was here cut at k = 10. In the two symmetric cases, Jl = J2 =
(fig. 1) and Jl =J2 = 39_ (fig. 2), the dependence on the number of pairs n is correctly
reproduced, and the energies of the low-lying states are in reasonable agreement for
j = { and in good agreement for j = 2¢9_.This is surprising for large values of the pairing
strength G. The reason is that Hcannot change k [in eq. (3.2)] by more than one unit t,
so that the influence of the neglected off-diagonal terms is directly determined by the
t See eq. (7.4) of ref. 3).
I i i i i l l l l

/ "
,I ,',"~
0
~ o

II
II
< <E II II

r ~'~ ~

b~ O~

I I ,..:,
i r~
<I b

i I t I I i I I I I
o
o'1 b>
0.02 0.04 0.06 0.08 0.10 0.12

1.5 , ,,, 1.5

248 R. A. BROGLIA AND B. SORENSEN

j = { case with the n = 20 results in the j = - ~ case, where, as mentioned, only


truncation errors are present. In the asymmetrical cases Jl = ~ , J2 = 39 (fig. 3)
and vice versa (fig. 4), the n-dependence of the excitation energies is wrong in the
superconductive region, although the general fit is quite good. In order to understand
this effect we display, in figs. 5 and 6, the calculated and the exact wave functions.
The spuriosity phenomenon can be studied because of the increase in size of the
boson space to k = 13. For n = 8 (fig. 5), the physical basis is nine-dimensional and
for n = 12 (fig. 6) it is 11-dimensional. The calculated wave functions contain com-
ponents outside the physical basis, but it is seen that these spurious admixtures are
small when G is small, and that for n = 12 they are even small in the superconductive
region. The incomplete overlap with the exact wave functions is in this case mainly

0.2
Jl : Jz = 5/z

0.1 n =2or~.

o.o ~ ~ ~ ~ g r st
0.0 0.5 1 1.5 2 13~z
6
Fig. 7. D e p e n d e n c e o f s p u r i o u s amplitudes for exciting three pairs in t h e case o f two o r f o u r pairs
in two ] = t levels o n t h e interaction strength.

due to the truncation error, which also appears in the energy spectrum. In the super-
conductive situation for n = 8, the picture is completely changed. The spurious
components are large and cause a general shift of the maxima and minima in the wave
function. The reason that this does not lead to major discrepancies in the spectrum
and transition cross sections is that the shape of the wave functions is correct and that
the shift is the same for all the wave functions. The energies relative to, for instance,
the ground state of the n = 12 system will be wrong. The asymmetry found between
these two cases elucidates the onset of inaccuracies in the B-type boson expansions.
BOSON-EXPAN$ION
METHOD 247

coefficients o f the expansions (3.1), i.e. by the magnitudes o f j.21, and ]A2
2 - Such a de-
pendence on the degeneracies can he seen by comparing the result for n = 3 in the
• ~6-2p
<J~
"2p
; Jz I>
<j'°-%j~)>
1.0~ 1.0
(a) (b)

gr.st.
/~ 2.exc.
°~I-L h \ / ~ e x c 0.5
1.ext. o,750.73
~i ~/~,'Wv "g~~t [ ..... gr.
' .....
st.
~ 8 \~' \\ ~ k38- gr,
1.ext.
st.

0.0x ~ 6
S
P- 0.0 _~:r-.._ 1,2P._

\\I/'/ /.<I
-o.5_ ~ _ p _ _ _ _ ~ S ,,- -0.5
Fig. 5. Direct comparison of exact (full lines) and boson (dashed lines) wave functions in the
] ~ = ~ , J 2 = -~g case with eight pairs present (a) G/e = 0.02 (non-superconductive region) and
(b) G/e = 0.10 (superconductive region). The energy spectra are inserted. The components of the
wave functions are given as function of the number p of pairs excited relative to a completely un-
correlated ground state. Only p ~ 8 corresponds to physical components (P region), the higher ones
are spurious (S region).

( "20-2p;J2"4+2pI>
<J~
1.0
(a) (b)

L /,gF. S~.
0.5
~ t 2'1.e;~,
exc" o5-A4 (,
~1.56= - -t 52 2.exc. ~' ,~, A %e-d,/-, ~~- ~ ' ~'-~- ~ o ~ .
I 0.750.73 1.exc.
..... gr. st.

0.0 "1 , ,P
2 4 6 8 10

I
I
-0.5
P ---S . -0.5 - P ~ S )

Fig. 6. Same as fig. 5 excep~ that now 12 pairs are present and hence the components up to p ~ 10
belong to the physical region.
248 R. A. BROGLIA AND B. SORENSEN

j = { case with the n = 20 results in the j = ~ case, where, as mentioned, only


truncation errors are present. In the asymmetrical eases Jt = ~ , Jz = -~- (fig. 3)
and vice versa (fig. 4), the n-dependence of the excitation energies is wrong in the
superconductive region, although the general fit is quite good. In order to understand
this effect we display, in figs. 5 and 6, the calculated and the exact wave functions.
The spuriosity phenomenon can be studied because of the increase in size of the
boson space to k = 13. For n = 8 (fig. 5), the physical basis is nine-dimensional and
for n = 12 (fig. 6) it is 11-dimensionai. The calculated wave functions contain com-
ponents outside the physical basis, but it is seen that these spurious admixtures are
small when G is small, and that for n = 12 they are even small in the superconductive
region. The incomplete overlap with the exact wave functions is in this case mainly
<b~ b~zl> <b~b~l>
~(3!
- - ~4:)
'~ or (3! 4!)"~

0.2
J~ = Jz : s/2
n =2or4

0.1

o.C
0.0 0.5 1 1,5 2 G~'_J
Fig. 7. Dependence of spurious amplitudes for exciting three pairs in the case of two or four pairs
in two j = ~ levels on the interaction strength.

due to the truncation error, which also appears in the energy spectrum. In the super-
conductive situation for n = 8, the picture is completely changed. The spurious
components are large and cause a general shift of the maxima and minima in the wave
function. The reason that this does not lead to major discrepancies in the spectrum
and transition cross sections is that the shape of the wave functions is correct and that
the shift is the same for all the wave functions. The energies relative to, for instance,
the ground state of the n = 12 system will be wrong. The asymmetry found between
these two cases elucidates the onset of inaccuracies in the B-type bos0n expansions.
At a pairing strength G = 0.1 5, the correlations are strong enough to make the am-
plitude for exciting eight pairs considerable, while the amplitude for exciting ten pairs
is still negligible. Thus, in the n = 8 case, the interaction can fairly easily admix
spurious excitations of nine pairs, while it is not strong enough to admix excitations
of 11 pairs, which would be the first spurion in the n = 12 case. In order that the
wave functions of the lower states do not have appreciable amplitudes up in the region
where spurious components begin to be present, it is necessary to include as large a
1.5
~ E (a)
0.00
--o.oo~
(b)
0.00 ~
0.00 ~0.00
(c)
0.00
o.oo
0.00--

0 0 0 0

1.0

1.19 1.19 0 0 "~1.13 116 0-000"01 ~1.1., 1.150.00


~ 0.01

0.5

0.0 1.69 1.19 1.19 1.69 ~ . ~ . . -~.1 " 6 31.19 1.18 1.63 1.62 1.18 1.18 1.62
18 19 20 21 22 18 19 20 21 22 18 19 20 21 22
n
Fig. 8. Excitation spectrum and two-particle transfer amplitudes for n pairs in two j = - ~ levels and
G]z~/8= 0.5 (non-superconductive region). The number attached to a state [n~ is I ( n l T I n - 1 ; gr.st.~ I .
(a) and (b) correspond to boson expansions based on particle-hole operators and H(T) truncated at
2.(1.) and 4.(3.) order, respectively, (c) is the exact solution.
E
1.5

(a) (b) (c) (d)

0.02 0.01

1.o -0 0.10 0.04


0.09 0.03
0

0 0 0 0 0.16 0.14

0 0

0.25 0.22
0.5 0.25 0.22 - - - - 1.11 0.33
- - 1.07 0.38
1.11 0.42 1.10 0.36
0.83 0.83 1.11 1.20

2.37 2.15 2,15 2.37 2.37 2.15 2.15 2.37 2.83 2.632.63 2.83 2.73 2./,7 2./I,7" 2,73
O.O
18 19 20 21 22 18 19 20 21 22 18 19 20 21 22 18 19 20 21 22
rl
Fig. 9. S a m e quantities are plotted as in fig. 8 b u t for G.~z2]e= 1.2 (transition region). (a) a n d (b) are
results obtained with b o s o n e x p a n s i o n s b a s e d o n t h e appropriate quasi-particle representation a n d
t r u n c a t e d at 2.(H) a n d 1. order (T). I n (a) all y 0 = 0, whereas in (b) t h e self-consistent values discussed
in sect. 5 have been used, w h i c h are y 0 = 0.03 for n 5& 20 a n d y 0 = 0 for n = 20. (c) c o r r e s p o n d s to
the b o s o n expansions with particle-hole operators cut at 4.(H) a n d 3.(T) order, while (d) gives t h e
exact solutions.
250 R. A. B R O G L I A A N D B. S O R E N S E N

part of the correlations as possible already in the starting representation, i.e. in the
present case to start from a quasi-particle representation. This is even more important
in general applications of the boson method, where the reduction of the dimension of
the boson basis is essential for numerical evaluations. For this reason, great care
should be taken in the determination of a boson representation, in which the vacuum
already includes as large a part of the necessary correlations as can be put there. We
show, in fig. 7, the grow of spuriosity with increasing correlation or when going up
in the excitation spectrum. It is seen that spurious components are not only present

(a) (b) (c)



0
- - 0 0 0
0.02
~0.0._...~.~ 0.03 0.03_~

1.5

1.0
0.&3 0.39 0.35 0.31 O,&60 43 0360'32- O 45 0 -'-1 0 37 0.32

0.5

3.49 3.&5 3A5 3.&9 &.O7 &.O7 4.07 &.07 3,96 3.96 3.96 3.96
0.0
18 19 20 21 22 18 19 20 21 22 18 19 20 21 22
13
Fig. 10. Same quantities are plotted as in fig. 8 b u t for GJlzfe = 2 (superconductive region). (a) uses
quasi-particle basis a n d 2.(1.) order expansions, (b) uses particle-hole basis a n d 4.(3.) order ex-
p a n s i o n s , (c) is exact.

for values of the coupling constant larger than the critical one G~, but that they grow
quite smoothly already from the onset of the interaction, and that the phase transition
is not felt at all.
The direct comparison of calculated with exact wave functions is not likely to show
whether small systematic deviations are present. In this sense, it is much more fruitful
to look at the coherent sums of the amplitudes or of their squares, i.e. to look at the
two-particle transfer amplitudes. This is done in figs. 8-10 in three typical cases for the
Jl = Jz = 39_ system, corresponding to the non-superconductive (fig. 8), transition
BOSON-EXPANSION METHOD 251

(fig. 9) and Superconductive region (fig. 10). For the non-superconductive case, the
second-order boson Hamiltonian can be diagonalized (this is the RPA, corresponding-
ly T is linear in the boson operators). The results are given for comparison in fig. 8.
In all regions, the transfer amplitudes predicted by the fourth-order boson calculation
are in as good agreement with the exact ones, as are the energies. In the non-super-
conductive case, the comparison with the second-order results reveals that the an-
harmonic corrections are extremely small. In the transition region, the situation is
opposite, the contributions to the excitation energy from the second,order terms more
or less cancel out so that the main contributions in this case are those coming from the
fourth-order terms. The superconductive region again shows almost harmonicity
features, a fact which is not taken advantage of with the boson representation em-
ployed here.

5. Other boson representations


We shall here discuss three ways of modifying the convergence properties of the
boson expansions of physical quantities. Two of them are concerned with canonical
transformations, either in the fermion space enabling us to include correlation effects
in the one-fermion representation, or in the boson space corresponding to the in-
clusion of certain two-body correlations in the one-boson (two-fermion) represen-
tation. If the canonical transformations are linear, only harmonic correlations will
in this way become present in the basic boson vacuum.
The third possibility consists in distorting the basic boson vacuum by means of the
parameters yOO of the B-type expansion. In principle, any kind of correlation can be
introduced in this way. The condition for optimizing the choice of boson represen-
tation can be stated in a phenomenological way: assuming each boson operator to be
associated with a definite mode of excitation, the splittings between groups of states
belonging to the same mode must be of the same order of magnitude as the corre-
sponding boson frequency. The optimum boson vacuum is thus considered as that
which minimizes the dimension of the boson representation necessary for describing
the low excitation spectrum.
In the case of pairing force, we thus expect that the change of representation in the
fermion space will be useful in the superconductive region and that the change of
representation in the boson space will be useful whenever the excitation energy
of the elementary mode (pairing vibration) is lowered considerably compared with
the corresponding independent (quasi)-particle energy, i.e. in the transition region and
especially when the levels have large degeneracies. The linear, canonical transfor-
mation of RPA-type can be used when the elementary excitation energy is not near to
zero, in which case the introduction of suitable yOO(j) values is the only possible way
of improvement.
We first consider the one-fermion transformation

a-j"m : "l,l jO~fn'-Ir- Vj~jm , (5.1 t


252 R . A . BROGLIA AND B. SORENSEN

which diagonalizes the one-fermion spectrum when (us, v~) fulfill the BCS equations.
The corresponding pair operators for the two-level model are
d ~ + = (~+~+)o, dV', = (~:~,)o, i = 1, 2, (5.2)
for which the B-type boson expansions read *
d + (2-4y°)~b +-2(2-4y°)-~);'2b{-b+b,+ ..., (5.3a)
~A/.' = ), yO + 23 [- Xb + b, . (5.3b)

Since this case was not considered in ref. 2), we shall give the corresponding ex-
pansions of Hamilton,an, number and transition operators

H = E [26iv j ; 1hi(i) + ( 2 - 4y°)~(2- 4y~,)~h5(ii') + 4 E yO, )v'


ii" i"

-1 ..tt -2 ..t +
X3, h6(tt )6.,+4), h6(tt)6w]b, by
+ E [ ( 2 - 4y°)~(2- 4y°) ~ - 2)? 26//,]ha(ii')b + b + + E ( 2 - 4y°)~y ° h4(ii')) r b +
ii" ii"

+ ~ [2(2-4y°)&Jvlh4(ii')- ~. h 4 ( i i " ) 3 , . y ° , 2 ( E - g y ° ) -½
ii" i"

x )7 26,,' - 4(2 - 4y ° ) - ~)~-36u, h4(ii')] b + b,+, by


+ 4 E [ -)7, 2hs(ii')(2 - 4Y°)~( 2 - 4Y°)-~3vv ' + (2 - 4y °)-1
ii'i"

x J,^ - - 4 6u,,fu, h s ( t•,°!) + ) , --


ij,TXh6(ii')f,r.]b+b,+bv.bv
"F 2
[ - 2 j v, - 2 h 3 ( ".,
t t ) ( 2 - 4 y , 0) ½( 2 - 4 y v0) --½"
6,.v.-2)i~ - 2 h 3 ( t-.tt )
ii'i"

x (2 - 4y°) - ~(2 - 4y~)'~fu,, + 8(2 " 4 y ° ) - ~J~-*hafi i') x 6 w, 3w] b+ b~ b~ by, + . . .


+h.c., (5.4)

where
hl(i ) ~ 1 ~ji(Ei+Gu~v2),
^ (5.5a)

h3(ii') = ~(,,
1 2v,,2 s,~ s,,), (5.5b)
h4(ii') = - ~ G1 J , j,,
*- ,,, v,,(~, 2 - ~?), (5.5c)
hs(ii') = - ~Gj,3,,(~,
~ , 2~,,2 + ~,2~,),
2 (5.50)
h6( ii') = - ~G}, ),, u i v, uv vv , (5.5e)

N = n+ E3~(uZ-v~)y°+2 E (uZi-v~)b+b, - E u,v,

× 3,((2 - 4y°)~(b + + b~) - ( 2 - 4y °)-.3;- 2(b+ b + b, + b + b, b,)) + . . . . (5.6)

where n is the number of particles.


* In accordance with the notations of refs. 3,0, we have put y OO= ¢~yO
BOSON-EXPAN~ONMETHOD 253

The two-particle transfer operator (2.4) is

ri = 2-~A + = (1 --2y~)
o , (uibi
2 + - v i b2 i )
-- (1 - 2y °)- ~2~-2(u~ b + b + b i - v 2 b + b, bi) + u, v,(2),y ° + 4)[ "1b + b,) + . . . . (5.7)

The nearly harmonic spectra found for the two-level model in the superconductive
region suggest that the expansions (5.4)-(5.7) will be rapidly converging. In this case,
the convergence parameters differ formally from those characterizing the particle-
hole expansions of sect. 3 by powers of the factors u~ and v~ which are smaller than
one in the superconductive cases. Therefore we try to cut the expansion of H already
at second order and consistently T at first order. We thus have a simple RPA matrix
diagonalization t, the result of which, assumingy ° = 0 in (5.3), is shown in figs. 9 and
10 in column (a). It is seen that the lower-order truncation gives good results even
when G is rather close to Gc (at Go the RPA energies are zero). The change of rep-
resentation has enabled us to describe the excited states in terms of one single boson
component in contrast to the situation with the particle-hole representation, where
the wave functions had components of many boson numbers.
Retaining a fixed, superconductive representation, one can also go to the region
G < Go. The higher-order terms in the expansions then become important, but even
the second-order method can give qualitatively reasonable results, as is indicated in
fig. 11 where the G)~/e = 1.2 representation is used for all G smaller than this one.
This is in contrast to the second-order particle-hole expansion of H corresponding
to the fixed representation of G = 0, which leads to imaginary energies when applied
outside its natural area, i.e. for G > Go. The main error in the description of the non-
superconductive region by the second-order expansion based on ui and v~ of G)~/e =
1.2 comes from the neglect of those parts of H which would project out spurious com-
ponents. For G = 0, the spurious amplitude in the wave functions is around 50 ~/o.
In addition, fig. 11 collects the various approximations under consideration.
The RPA diagonalization of (5.4) corresponds to a canonical transformation

b+ = Z (r.(i) c+ + s.(i)c.) (5.8)


n

in the boson space. As we have seen, the anharmonic corrections will be small in this
(c +, c) representation, which thus provides a better starting point for adding higher-
order terms than the (b +, b) representation. A similar canonical transformation
could be applied to the particle-hole boson representation considered in sects. 3 and 4
before diagonalizing the fourth-order Hamiltonlan, but since the anharmonic ad-
mixtures are of essential importance when G is appreciable, this is only an improve-
ment for G << G¢. For larger G, the frequency of the c-bosons will be lower than the
excitation energy and, thus, almost the entire excitation energy must come from the
higher-order terms, which clearly is an undesirable situation.
* The f o r m (5.4) o f H does n o t allow a reduction to a single dispersion relation.
254 R . A. B R O G L I A A N D B. S O R E N S E N

We finally tried to use yO = 0 in the expansions (5.3). The self-consistency condi-


tion differed f r o m that of ref. 5) in the following way. The constraints, which fluctuate
in the B-type expansion, produce spurious states which m a y be recognized by their
entering in the particle n u m b e r counting. We thus considered the correctness of the
particle n u m b e r as a criterion for having a representation with minimized spurion
admixture. In the second-order ground state (defined by c,10) = 0), the particle

E
E~

2.0

1.5

. ~ . 1,1

1.0
\ ~ I//
... ',"~'~ / /
0.5
"-~.._ " ~ / .I
-"'-L,,.\ i ,//

0.0 I I I I I I
o o.~ o.a 1.2 1.s 2.0 G~'.__z~
~,
E

Fig. 11. Excitation energies for 20 pairs in two j = ~ levels. Heavy, full line: exact solution; thin, full
line: fourth-order expansion from particle-hole basis; dashed line: second-order (RPA) expansion
from particle-hole basis; dot-and-dash line: second-order (RPA) expansion from BCS quasi-
particle representation and, double-dot-and-dash line: same, but based on a representation fixed at
G = 1.2 e/J1~.

n u m b e r (5.6) is given by
<0lN[0> = n + ~ (u~- v2){)~y ° + 2 ~" s,(i)2}. (5.9)
i n

In order to study the j~ = Jz = 39 case, we t o o k yO = yO and iterated until corre-


sponding sets (y o, r,(i), s,(i)) satisfied the requirement that the quantity in eq. (5.9)
BOSON-EX]PANSION METHOD 255

be equal to n. The results for a case in the transition region are shown in column (b)
of fig. 9. One notices a slight improvement compared with the results in column (a).

6. Conclusions
The results obtained with the two-level model indicate that the B-type boson ex-
pansion is a rapidly converging method for obtaining a simple description of non-
harmonic excitations. As expected, the lack of fulfilment of the Pauli principle ap-
pears most clearly in connection with levels having a small degeneracy, where the
degree of filling may become near to zero or one. Although the expansion starting
directly from the particle-hole representation was found to produce good agreement
for energies as well as two-particle transfer amplitudes for all interaction strengths,
it turned out by a more detailed analysis of the wave functions that the use of an
initial representation which included already correlations was a preferable improve-
ment. In this way, the ground state will not be very different from the boson vacuum,
and the low excited states will be largely few-boson states in consistency with the
neglect of higher-order terms in H and T. The rich variety of correlations which be-
comes possible in systems with more than two levels certainly makes the determina-
tion of a suitable starting representation an essential problem, especially in connection
with the collectivity approximation which neglects a large part of the Hamiltonian.
A general criterion which such a correlated basis should fulfill is that the frequency
of each boson mode should be larger than the lowest excitation energy of that mode *.
In most of the cases considered here this is fulfilled for the expansions with yO = 0.
This choice ceases to be the natural one in the transition region, where low-energy ex-
citations occur for large degeneracies of the two levels, thus leading to a large devi-
ation of the uncorrelated particle-hole energy from the correlated energy.
The success of the low-order truncated boson expansions for the two-level model
cannot directly be taken as a basis for choosing the truncation orders for more com-
plicated systems. One can easily imagine level structures and interactions which imply
correlations even in the lowest states, which cannot be reproduced by the limited de-
grees of freedom available within, e.g., the fourth-order expansion of the Hamiltonian.
It is yet believed that the results of the present investigation can be useful in the fur-
ther application of the B-type boson expansion not only by supporting the validity
of approximations made but also by providing insight into the way in which the physi-
cal correlations enter into the boson wave functions, information which can usually
not directly be extracted for the more complicated models.

The authors wish to thank Professor A. Bohr for reading the manuscript; the hos-
pitality of N O R D I T A is highly appreciated by one of the authors (B.S.), who stayed
there during a large fraction of this work.
* It is a fact that the third-order anharmonic terms decrease the excitation energy of the one-
phonon state so that a choice of uncorrelated boson frequency too low will shift all the anharmonic
correlation effects to the fourth or higher order terms.
256 R.A. B R O G L I A A N D B. SI~IRENSEN

References
1) S. T. Beliaev and V. G. Zelevinsky, Nuclear Physics 39 (1962) 582;
T. Tamura and T. Udagawa, Nuclear Physics 53 (1964) 33;
T. Marumori et al., Progr. Theor. Phys. 31 (1964) 1009;
K. Hara, Progr. Theor. Phys. 32 (1964) 88;
M. Yamamura et aL, Prog. Theor. Phys. 37 (1967) 336;
J. da Providencia, Nuclear Physics A108 (1968) 589
2) B. S¢rensen, Nuclear Physics A97 (1967) 1
3) J. H6gaasen-Feldman, Nuclear Physics 28 (1961) 258
4) R. Broglia et aL, Nuclear Physics A107 (1968) 1
5) B. S~rensen, Progr. Theor. Phys., to be published
6) B. Sarensen, in Proc. Int. Conf. on nuclear structure 1967, Tokyo
7) D. R. B~s and R. A. Broglia, Nuclear Physics 80 (1966) 289

You might also like