PHYSICAL REVIEW C VOLUME 37, NUMBER 1 JANUARY 1988

Fluctuations and the thermal mean-field approach

R. Rossignoli and A. Plastino

Physics Department, National University, La Plata 1900, Argentina

J. P. Vary*
W. K. Kellogg Radiation Laboratory, California Institute of Technology, Pasadena, California 91125
(Received 2 September 1987)
Fluctuations of selected observables and their dependence on the size of the system are studied
within the framework of an exactly solvable model that mimics the interplay between pairing and
long range forces. It is found that thermal mean-field results for fluctuations can yield an ade-
quate estimate of the true fluctuations provided one evaluates those fluctuations in a particular
manner.

I. INTRODUCTION creation (destruction) operators related to some suitable

set of "unperturbed" operators (c, , c;) by means of a
The various thermal mean field approximations' general Bogoliubov transformation
finite-temperature Hartree-Fock (FTHF), finite-
temperature Hartree-Fock-Bogoliubov (FTHFB), and (a]i tal ia]i ~ ~ ~
taL, )
finite-temperature Bardeen-Cooper-Schrieffer (FTBCS)
are appealing techniques for the evaluation of statistical
=(c], . . . , cL, c], . . . , cL )W . (2.2)
averages in a variety of different scenarios. They consti- Here L denotes the number of s.p. states in the model
tute optimal lowest order approaches that can be em- space and 8' is a unitary 2L )(2L transformation of the
ployed as first steps towards more elaborate treatments. form
The main idea is to replace the exponent of the pertinent
density (or statistical) operator with single-particle U V
operators and then to adjust parameters so as to mini- V' U' (2.3)
mize the corresponding grand potential. The grand
canonical ensemble provides a convenient framework to Both 8' and the e,' depend upon the temperature T
obtain the statistical properties of finite fermion systems. and are determined by extremalization of the appropri-
Accordingly, particle number fluctuations are an impor- ate grand canonical potential
tant feature of these mean field descriptions. At finite
temperature they are always present, even in the absence
5Q=0, (2.4)
of pairing correlations. with
Since nuclear experiments fix the number of nucleons
through the selection of projectile and target combina-
tions, it is important to assess the role of fluctuations in
(2.5)
theoretical results for various observables. It is the aim
of the present effort to undertake such a task in situa- in the conventional potation. The entropy depends sim-
tions for which exact results are also available. It is par- ply on the mean occupation numbers That is, f;.
ticularly important to have calibrations of the role of
fluctuations for observables which carry the signals for S = —g [f, ln(f, )+(1 f; )ln(1 f; )],— (—
2.6)
the onset of phase transitions.
with i running over the s.p. levels and

a;.
II. OF BASIC FEATURES
A REVIEW
OF THERMAL MEAN FIELD APPROACHES f, =(ata, &

The essential feature of thermal mean field approaches

= [1+exp( pe,' ) ] (2.7)
(TMFA's) is that of casting the statistical operator in The stationary condition (2.4) leads to B(Q)/Bf;=0,
terms of a single particle (s.p. ) operator implying (8 '=8 —pA')
p= exp A, ]] —p g e,'a, (2. 1) a&A &

a
where p is the inverse temperature (Boltzmann's con-
stant taken equal to unity), 1],o is a nor]nalization con- a&8& —p (2.8)
stant (Trp=1), and a,t (a,. ) are fermion quasiparticle 8&a ta, &

37 314 1988 The American Physical Society

37 FLUCTUATIONS AND THE THERMAL MEAN-FIELD APPROACH 315

which, together with the corresponding approximate density operator, but in

many situations So lies closer to the exact value. As for
a&8 l

the Hamiltonian, SH = — a(8)/a(p) is usually different

&

a&a,ta, &
from S& in the TMFA's, but on account of the station-
(2.9)
ary condition (2.4),
a&8 & a&8 &

a& a,ta,~) a(a, a, &'

a&8 &
= as ' (3.4)
ap ap
determine both W and the set of energies '(or
e; as in the exact treatment.
equivalently, the f s}.
III. FLUCTUATIONS —GENERAL IV. THE MODEL
In view of the nature of the underlying approxima-
tions TMFA's should provide fairly accurate expectation We will illustrate the preceding concepts with refer-
values of s.p. operators. The central issue here is what is ence to a well-known SU(2)XSU(2) model ' that deals
the quality of the results for the corresponding fluctua- with two 2Q-fold degenerate s.p. levels, whose s.p. states
tions. To be specific we investigate the particle number are labeled p, v), p =1, . . . , 2Q, v=+1. The following
~

fluctuations which have two different origins: those aris- quasi-spin operators play a dominant role:
ing from the Bogoliubov transformation (BCS-like) and
those of a pure thermal nature (traces taken within the
grand canonical framework).

ble, (8' ) —
modynamical
:
In an exact treatment of the grand canonical ensem-
(8)—
S„can be identified with the ther-
quantity
J~ —

Q+=X
I g vcp~cp„,
p~v

p+ p-=Q-
(4. 1)

a&A&
'ap Q, = —,
P

pc „c „—0= '8 —0,

'
—,
(4.2)

In the TMFA's S„and S„' do not necessarily coincide. p7v

Moroever, under certain circumstances, S„' may provide where each J operator commutes with all the Q opera-
one with a better estimate of the exact fluctuations than tors. The corresponding Hamiltonian reads
S„when both are evaluated in the TMFA. Quite gen-
erally, in an exact treatment, 8=~J, + (J'++J' ) ——
—Q+Q (4 3)
s, =&o,'& —&o, &' (3.2)
with the Q part acting like a pairing interaction between
is identical to states separated by an energy gap e. The part, in turn, J
simulates the effect of a long range interaction. This

So —T a&0, &
(3.3)
model exhibits both first-order and second-order local
P. and nonlocal phase transitions, both at finite and zero
temperature (the concept of phase transition is both val-
if p; is the chemical potential accompanying 0; in the id and useful even for a relatively small number of parti-
appropriate potential, and 0, is an operator which com- cles, as established by Gilmore, Feng, and others ).
mutes with p. This identity does not necessarily hold By recourse to the finite temperature Wick's theorem,
within the TMFA framework. So is the value given by one easily finds in TMFA's
l

(@&=2& ~&j )+ —
'(&j )'+&j )') —(g/2)[&y, )(q-)+((&)'/4 (j, )' —(j —))/2']
&&j

(4.4)
S = —20 g [f ln(f„}+(1 f„)ln(1 f„}],— — (4.5)

where U = V(20 —1), g =2QG &0, and where lower the unperturbed situation; (b) the deformed phase, with
case operators denote intensive quantities, e.g. , (J„)&0, (Q )=0, induced by the monopole term in
j, =J, /2Q. the Hamiltonian; and (c) the superconducting phase,
The present picture exhibits three well-defined, with (Q„)&0 and (J
) =0, arising from the effects of
different phases: (a) the normal phase, characterized by pairing terms in B.
(J ) =(Q„)=0 (v=+1), which can be considered as A general s.p. transformation of the type (2.3) can be
316 R. ROSSIGNOLI, A. PLASTINO, AND J. P. VARY 37

cast within the present context, in the form of two suc- In this last situation the critical temperature is to be
cessive, commuting unitary transformations determined from
atp+ cosy —e' siny
''a' '
8T,
(4.6a) cosh (E/4T, )=1 . (4. 13)
a e ' siny cosy g' g

—e '~sina
. t When the coupling constant equals zero the angle y 6
~p + cosa cp + vanishes and the deformed phase is to be assigned to the
(4.6b)
e '~sina cosa Cp ground state for U & e (at T =0). A local, second or-
I I

der phase transition ensues at

which lead to

( j, ) = 2a),
rcos(— Tc
2»[( I
U
I
+e)/( I
u I
—e)l (4. 14)

( j+ ) =r sin(2a)e'~, For T & T,' the system will be found in the normal
(g, ) = —q cos(2y),
(4.7) phase. As T +ec, b—oth f+ and —+ —,' and r, q ~0. f
When V and 6
are simultaneously finite, a competi-
&g~ ) =q sin(2y }e'e, tion arises between the long- and short-range interac-
tions. However, for a given T below the critical temper-
where ature, the system is either superconducting or deformed
(exclusive "or"), so that the behavior of the system can
r =-,'(f++f ), (4.8a)
be described in the terms presented above.
q = '(1 f+
—, f ) . —
— (4.8b) Exact averages can be computed by decomposing the
full set of accessible states into irreducible (J, Q) repre-
The occupation numbers f+ and f
can always be sentations, and then diagonalizing A', within each (J, Q)
chosen so as to comply with the constraint f+ & and f multiplet. The TMFA becomes exact in the thermo-
0& f++f & 1. The angle P is adjusted so as to yield dynamic limit (see Ref. 5) for intensive mean values (first
negative mean values of the monopole interaction in 8, order moments).
while the angle 8 does not play any relevant role in the
minimization (continuous degeneracy}. A minimization V. FLUCTUATIONS —RESULTS
of the potential 0
with respect to y yields two possible
solutions: (a) sin(2y)=0 (normal phase) or (b) cos(2y) A. Particle number fluctuations
= —2p/gq (superconducting phase), so that
Particle number fluctuations S„evaluated in TMFA's
(lY') =20(f++f ), sin(2y)=0; are depicted in Fig. 1 as a function of temperature and
(8') =2Q(4p/g+ I ), cos(2y) = —2p/(gq) . (4.9b) for different phases. We obtain

Within the framework we are discussing, the FTHFB

equations can be regarded as a set of just two coupled
equations, namely Eqs. (4.8} with (see Ref. 5)
1.0—
f, = [1+exp(e„/T)] (4. 10)
e„=ve„—p, sin(2y ) =0; (4. 11a)
e„=ve, +gq/2, cos(2y)= —2p/(gq); (4. 11b) (JZ2

e„=e/2, sin(2a) =0; (4. 12a)

0.5
v= 9~0
e„=vr; cos(2a) =e/(2vr) . (4. 12b}
Expression (4. 12b) corresponds to the deformed phase.
In Eqs. (4. 11), the (I/20) terms of Eq. (4.4) have been
neglected, as customary (they represent pairing contribu-
tions to the HF potential). Henceforth we shall restrict 0.0 I

ourselves to the N=(A') =20 situation, for which the Tc 1.0 Tc 2.0
chemical potential p vanishes [cf. Eq. (4.9)], independent
of the phase. TEMPERATURE
We sketch now the main features of our model. If the FIG. 1. Relative fluctuation of the number of particles
monopole coupling constant V equals zero, a will always o'z —((8 ') —(8 )')/N as a function of temperature according
vanish. At T =0 the superconducting phase is assigned to thermal mean field treatments, for different values of the
to the ground state for g &4e. The gap (proportional to coupling constants. The temperature and the coupling con-
q} decreases as T grows Aphase . transition takes place stants are in units of e. A curve with only one coupling con-
when the gap vanishes, which is nonlocal and first order stant indicated has the other set to zero. Critical temperatures
for g &g, =4. 47', but local and second order if g &g, . are also indicated.
37 FLUCTUATIONS AND THE THERMAL MEAN-FIELD APPROACH 317

S„=2Q[2(N/2Q) —(N/2Q)' —(f++f f+f

2— )1

(5 1) 0.20
an expression that reduces to
0.15
),
S„=2Q g f„(1 f„— (5.2)
0. 10
when f f
++ =N /2Q (normal phase). The results
show order of magnitude variations in S„depending on
the phase at low fixed temperature T. The results also 0.05
show order of magnitude changes in S„with increasing
T within fixed phases, particularly for T close to T, .
The smooth line with U =g =0 represents the unper- 0.0 0.5 1.0
turbed situation that becomes an "attractor" for the per- TEMPERATURE
turbed solutions for high enough temperatures. The
TMFA coincides in this case with an exact treatment in
FIG. 2. Mean value of j, = J, /N vs T for a pure pairing in-
the grand canonical ensemble, so that we can ascribe a
teraction with g/a=5, according to exact results for N =10 (a)
pure thermal origin to the particle number fluctuation. and N = 50 (b), and FTBCS (c).
In the presence of a monopole interaction (cf. the curve
labeled v =5 in Fig. 1), fluctuations diminish in the de-
formed phase (T & T,'), as the interaction favors a larger
degree of order at low temperatures. For the case g =5, i.e., proportionalto S„/N in the normal phase. FTBCS
U =0, FTHFB degenerates into FTBCS for the supercon- values are at variance with the exact ones, especially
ducting phase. Particle number fluctuations have a when N is large (say, N =50) in the critical region. This
quantum origin ["Cooper pairs, "
y&0 in (4.6a)] at suggests that when computing fluctuations, corrections
T =0, and, as temperature works against the interaction, to Wick's theorem cannot be neglected even as N in-
fluctuations decrease as T grows towards T, . creases. The quantity
S„coincides with S„' [cf. Eq. (3.1)] both for the normal
&&j,
and deformed phases (for N =2Q), but not in the super- S,'/2Q = —T E'
&

(5.5)
conducting one, for which [cf. Eqs. (4. 10) —(4. 12)]
8&8'& is identical to the fluctuation in the exact picture [cf. Eq.
(3.3)], and should coincide with (5.4) in the normal
p
=2Q(4T/g), T & T, . (5.3)

S„' possesses the opposite behavior of S„since it van-

ishes at T =0 and increases with T. For T =T„expres- h
sion (5.3) coincides with (5.2) [cf. Eq. (4. 13)], so that no io
discontinuity is found at the phase transition. In the I~

thermodynamic limit, (5.3) is the exact particle number 0.2— I~

I I

fluctuation in the grand canonical ensemble. It should

also yield a more reliable estimate than S„even for finite g2

0. 1—
B. Fluctuations of J, and 8
Figure 2 depicts the (intensive) mean value ( ) vs T j,
for a pure pairing (v =0) Hamiltonian, with g &g, . Ex-
act values for finite N computed in the full canonical en- 0.0 0.5 1.0 1.5
semble are shown together with FTBCS results (which TEMPERATURE
correspond to N~oo). The mean field values lie very
close to the exact ones for finite N except in the transi-
tion region. & ) reaches its maximum at T =T, and
j, FIG. 3. Intensive fluctuation o2=((J,')
—(J, ) )/N vs T
then decreases towards zero in the normal phase. with the same parameters quoted for Fig. 2. Solid lines corre-
Figure 3 depicts the corresponding fluctuations that, spond to exact results for N =10 (a) and N =50 (b), dashed
according to Wick's theorem, can be cast in FTBCS as lines to FTBCS. Curve c is a plot of the direct evaluation (5.4),
while d is the corrected expression (5.9a), which shows the
S„/2Q = g „(If f„)/4, — (5.4) right trend for high enough N. The dotted line indicates the
location of the discontinuity at T = T, .
318 R. ROSSIGNOLI, A. PLASTINO, AND J. P. VARY 37

phase. Using (4. 10)-(4.12) we obtain

R
B(e /T)
B
T =' g vR„(v+gBq/Be), (5.6a)
BE
—,
=f.(1 —f. ) . (5.7)
Hence,

T ='
—, g R„(v+gBq/Be), (5.6b) g vR„/T 1 (g/—4T) g R„ (5.8)
BE V V

where and

Sg/2Q=T

= ' gR, +(g/16T)

—, gvR„1
V
(g/4T—) gR„', T &T, ; (5.9a)

= 'gR„, T&T,
—, ; (5.9b)

so that the difference S, — S,' does not vanish for T & T, In the case of a pure monopole interaction (g =0), we
and equals the second term on the right-hand side (rhs) can evaluate the specific heat with the aid of Eqs.
of (5.9a). This represents the correction to Wick's (4. 10)-(4.12) (we assume in the following U & 0},
theorem (arising from the dependence of e, upon q)
which is explicitly g dependent and strongly peaked at = —20[2urB(r)/B(T)], '; T & T, (5. 12a)
T=T, . Expressions (5.9) yield the exact fluctuation in
the thermodynamic limit, and exhibit a discontinuity at
the critical point. Numerical computation shows that
=20 '
—, QR„e/T, T & T,'; (5. 12b)
(5.9a) resembles more closely the exact results than does
Eq. (5.4) for not too small values of N.
The specific heat B(P ) /B( T) also differs from SH /T,
within TMFA's, in the T &T, region. In the normal
phase, the equality holds after neglecting terms that van-
ish (in intensive quantities} for N ~
ao. Note that
SH/2fI can be particle number dependent in TMFA's if
8 contains two-body terms. Following the procedure
outlined above one finds
B(8)
BT 'BT-
T
BS

=2Q[ —eB(r)/B(T) —gqB(q)/B(T)], (5. 10)

with

= g( ,'R„e„/T )— —1 (g/4T) QR„—

V V

(5.11a)

= g ( ,' vR „e„/T ) —
(g —/4T) g vR—
„Bq/BT .
V V

(5. 11b) TEMPERATURE

FIG. 4. Intensive energy fluctuation aH =((8')

—(8)~)/
Both (5.8) and (5. 11a) diverge for T~T, . However, N for a pure monopole interaction with U/@=5. Exact results
the specific heat (5. 10) remains finite for g &g, . Diver- (solid lines) are shown for N =10 (a) and N=50 (b). The
gences in the specific heat (and in other types of fluctua- dashed line (c) depicts the direct FTHF evaluation while d cor-
tions) are encountered in the range 4e&g &g, within responds to the corrected expression (5.12a), both for N =50.
TMFA's, where the phase transition is hard (nonlocal) The dotted line indicates the location of the discontinuity at
and the derivatives (5. 11) and (5.6) diverge accordingly. T=T'
37 FLUCTUATIONS AND THE THERMAL MEAN-FIELD APPROACH 319

with cy of the TMFA estimate of first order moments is de-

creased.
gR urlT R For fluctuations, the TMFA can yield reasonably ac-
V V curate results, provided thermodynamic expressions
(derivatives with respect to a suitable chemical potential)
(5. 13) can be employed. This fact has been utilized in order to
Temperature dependent effects (of e ) account for the evaluate corrections to the s.p. picture. In this way, the
denominator of (5. 13). As in the previous situation, corrected expressions provide the exact fluctuation of
8
SH&SH —T di, ) IB(T) in the deformed phase. Figure
the corresponding observable in the thermodynamic lim-
it and we are able to predict the strong maxima some
4 depicts results for S& (computed in the thermodynam-
fluctuations possess around the critical temperature for
ic limit) and for SH. The exact energy 6uctuation (cal-
sufficiently high This has been shown in particular
culated in the full canonical ensemble) is also shown for
¹

for the fluctuation of a one-body observable and for the

different values of N.
specific heat.
VI. CONCLUSIONS
Fluctuations of selected observables have been corn- ACKNOWLEDGMENTS
puted in different phases (normal, superconducting, and
deformed), within an exactly solvable model in order to Two of us (R.R. and A. P.) are supported as members
illustrate particular aspects of the thermal mean field ap- of the National Research Council (CONICET) of Argen-
proach. tina. This work was supported in part by a joint U. S.-
The TMFA is exact within this context in the thermo- Argentina exchange program sponsored by the National
dynamic limit for first order moments. Accordingly, the Science Foundation under Grant No. INT84-13827 and
TMFA also provides accurate mean values of observ- by the Consejo Nacional de Investigaciones Cientificas y
ables for sufficiently large N, when compared with exact Tecnicas. Additional support for this work was provid-
canonical ensemble averages. This is especially true for ed by the National Science Foundation under Grant
s.p. observables. The intrinsic particle number fluctua- Nos. PHY85-05682 and PHY86-04197, and by the U. S.
tions of the TMFA does not significantly affect this gen- Department of Energy under Contract No. DE-AC02-
eral trend, even though it varies widely with temperature 82ER40068 and Grant No. DE-FG02-87ER40371 of the
and phase. In the critical regions, however, the accura- Division of High Energy and Nuclear Physics.

'Permanent address: Physics Department, Iowa State Univer- 4M. C. Cambiaggio and A. Plastino, Z. Phys. A 288, 153
sity, Ames, IA 50011. (1978).
'D. J. Thouless, The Quantum Mechanics of Many Body Sys 5R. Rossignoli and A. Plastino, Phys. Rev. C 32, 1040 (1985);
tems (Academic, New York, 1961). 30, 1360 (1984).
A. L. Goodman, Nucl. Phys. A352, 30 (1981); A352, 45 (1981). 6R. Gilmore and D. H. Feng, Nucl. Phys. A301, 189 (1978); D.
30. Civitarese, A. Plastino, and A. Faessler, Z. Phys. A 313, H. Feng, R. Gilmore, and L. M. Narducci, Phys. Rev. C 19,
197 (1983); G. Bozzolo and J. P. Vary, Phys. Rev. C 31, 1909 1119 (1979); R. Gilmore, Catastrophe Theory for Scientists
(1985); H. G. Miller, R. M. Quick, G. Bozzolo, and J. P. and Engineers (Wiley, New York, 1981).
Vary, Phys. Lett. 168B, 13 (1986).