SERIES  ARTICLE

Snippets of Physics
10. Thermodynamics of Self-Gravitating Particles

T Padmanabhan

T h e sta tistic a l m e c h a n ic s o f a s y ste m o f p a rti-

c le s in te r a c tin g th r o u g h g ra v ity le a d s to se v e ra l
c o u n te r -in tu itiv e fe a tu r e s. W e e x p lo re o n e o f
th e m , c a lle d A n to n o v in sta b ility , in th is in sta ll-
m e n t.
S u p p o se w e p u t a la rg e n u m b er (N ) o f p a rticles, ea ch T Padmanabhan works at
o f m a ss m a n d in tera ctin g th ro u g h a tw o -b o d y p o ten - IUCAA, Pune and is
interested in all areas
tia l U (x ¡ y ) in to a sp h erica l b ox o f ra d iu s R . W e w ill
of theoretical physics,
a rra n g e m a tters su ch th a t th e p a rticles m ov e ra n d o m ly especially those which
to sta rt w ith a n d b o u n ce o ® th e su rfa ce o f th e sp h ere have something to do with
ela stica lly. L et th e to ta l en erg y o f th e sy stem b e E gravity.
w h ich , o f co u rse, w ill rem a in a co n sta n t. W e a re in ter-
ested in th e b eh av io u r o f th e sy stem a t la te tim es, w h en
th e p a rticles w ill h av e h a d su ± cien t tim e to in tera ct
w ith ea ch o th er a n d ex ch a n g e en erg y.
T h e resu lt w ill clea rly d ep en d o n th e n a tu re o f th e in -
tera ctio n , sp eci¯ ed b y U (x ¡ y ) a s w ell a s th e o th er
p a ra m eters. If U (x ¡ y ) is a sh o rt ra n g e p o ten tia l rep -
resen tin g in term o lecu la r fo rces a n d if E is su ± cien tly
h ig h , th en th e sy stem w ill rela x tow a rd s a M a x w ellia n
d istrib u tio n o f v elo cities a n d n ea rly u n ifo rm d en sity in
sp a ce. (T h e v elo city d istrib u tio n w ill h av e a ch a ra cter-
istic tem p era tu re T ' 2 E = 3 N a n d w e a re a ssu m in g th a t
th is is h ig h er th a n th e `b o ilin g p o in t' o f th e `liq u id ' m a d e
o f th ese p a rticles. If n o t, th e ev en tu a l eq u ilib riu m sta te
w ill b e a m ix tu re o f m a tter in liq u id a n d va p o u r sta te.
A lso n o te th a t w e u se u n its w ith k B = 1 th ro u g h o u t.)
A ll th is is p a rt o f sta n d a rd lo re in sta tistica l m ech a n ics.
Keywords
W h a t h a p p en s if th e U (x ¡ y ) is d u e to g rav ita tio n a l in - Thermodynamics, gravitation.
tera ctio n o f th e p a rticles? W h a t a re th e d i® eren t p h a ses .

RESONANCE  October 2008 941

 SERIES  ARTICLE

in w h ich m a tter ca n ex ist in su ch a ca se? I w ill d iscu ss

so m e o f th e p ecu lia r e® ects th a t a rise in th is co n tex t.
T o d o th is, let u s b eg in b y q u ick ly rev iew in g th e w ay
o n e in tro d u ces th e eq u ilib riu m co n ¯ g u ra tio n in sta tisti-
ca l m ech a n ics. C o n sid er a sy stem d escrib ed b y a d istrib -
u tio n fu n ctio n f (x ;p ;t) su ch th a t f d 3 x d 3 p d en o tes th e
to ta l m a ss in a sm a ll p h a se sp a ce v o lu m e. W e a ssu m e
th a t th e ev o lu tio n o f th e d istrib u tio n fu n ctio n is g iv en
b y so m e eq u a tio n (u su a lly ca lled th e B o ltzm a n n eq u a -
tio n ) o f th e fo rm d f = d t = C (f ). T h e term C (f ) o n th e
rig h t h a n d sid e d escrib es th e e® ect o f co llisio n s. W h ile
th e p recise fo rm o f C (f ) ca n b e q u ite co m p lica ted , w e
ca n u su a lly a ssu m e th a t th e co llisio n a l ev o lu tio n o f f ,
d riv en b y C (f ), sa tis¯ es tw o rea so n a b le co n d itio n s:
(a ) T h e to ta l m a ss a n d en erg y o f th e sy stem a re co n -
serv ed a n d (b ) th e m ea n ¯ eld en tro p y, d e¯ n ed b y
Z
S = ¡ f ln f d 3 x d 3 p (1 )

1
For those who are unfamiliar d o es n o t d ecrea se (a n d in g en era l in crea ses).1 F o r a n y
with this expression, here is a su ch sy stem , w e ca n o b ta in th e eq u ilib riu m fo rm o f f b y
recap: In the standard deriva-
ex trem isin g th e en tro p y w h ile k eep in g th e to ta l en erg y
tion of Boltzmann distribution,
one extremises the function
a n d m a ss co n sta n t u sin g tw o L a g ra n g e m u ltip liers. T h is
S = – ni ln ni of the occupation is a sta n d a rd ex ercise in sta tistica l m ech a n ics a n d th e
numbers ni subject to the con- resu ltin g d istrib u tio n fu n ctio n is th e u su a l B o ltzm a n n
straint on total energy and num- d istrib u tio n g ov ern ed b y :
ber. In the continuuum limit one
· µ ¶¸ Z
works with f rather than ni and 1 2
the summation over i becomes
f (x ;v ) / ex p ¡ ¯ v + Á ; Á (x ) = d 3 y U (x ;y )½ (y ):
2
an integral over the phase space (2 )
leading to (1).
In teg ra tin g ov er v elo cities, w e g et th e clo sed sy stem o f
eq u a tio n s fo r th e d en sity d istrib u tio n :
Z
½ (x ) = d 3 v f = A ex p (¡ ¯ Á (x ));
Z
Á (x ) = d 3 y U (x ;y )½ (y ): (3 )

T h e ¯ n a l resu lt is q u ite u n d ersta n d a b le: It is ju st th e B o l-

tzm a n n fa cto r fo r th e d en sity d istrib u tio n : ½ / ex p (¡ ¯ V ),

942 RESONANCE  October 2008

 SERIES  ARTICLE

w h ere V is th e p o ten tia l en erg y a t a g iv en lo ca tio n d u e One could have

to th e d istrib u tio n o f p a rticles. O n e co u ld h av e a lm o st almost written
w ritten th is d ow n `b y in sp ectio n '! (S ee A p p en d ix fo r down equation (3)
m o re d eta ils) 'by inspection'.
E v ery th in g th a t w e h av e sa id so fa r is in d ep en d en t o f
th e n a tu re o f th e p o ten tia l U (ex cep t fo r o n e im p o rta n t
cav ea t w h ich w e w ill d iscu ss rig h t a t th e en d ). In th e
ca se o f g rav ita tio n a l in tera ctio n , (3 ) b eco m es:
Z
½ (y )d 3 y
½ (x ) = A ex p (¡ ¯ Á (x )); Á (x ) = ¡ G : (4 )
jx ¡ y j

T h e in teg ra l eq u a tio n (4 ) fo r ½ (x ) ca n b e ea sily co n -

v erted to a d i® eren tia l eq u a tio n fo r Á (x ) b y ta k in g th e
L a p la cia n o f th e seco n d eq u atio n { lea d in g to r 2 Á =
4 ¼ G ½ { a n d u sin g th e ¯ rst eq u a tio n . W e th en g et, fo r
th e sp h erica lly sy m m etric ca se, th e iso th erm a l sp h ere
eq u a tio n :
à !
1 d dÁ
2
r Á = 2 2
r = 4 ¼ G ½ c e ¡¯ [Á (r )¡Á (0 )]: (5 )
r dr dr

T h e co n sta n ts ¯ a n d ½ c (th e cen tra l d en sity ) h av e to

b e ¯ x ed in term s o f th e to ta l n u m b er (o r m a ss) o f th e
p a rticles a n d th e to ta l en erg y. G iv en th e so lu tio n to th is
eq u a tio n , w h ich rep resen ts a n ex trem u m o f th e en tro p y,
a ll o th er q u a n tities ca n b e d eterm in ed . A s w e sh a ll see,
th is sy stem sh ow s sev era l p ecu lia rities.
T o a n a ly se (5 ), it is co n v en ien t to in tro d u ce len g th , m a ss
a n d en erg y sca les b y th e d e¯ n itio n s
G M 0
L 0 ´ (4 ¼ G ½ c ¯ )1 = 2 ; M 0 = 4 ¼ ½ c L 30 ; Á 0 ´ ¯ ¡1 =
L0
(6 )
A ll o th er p h y sica l va ria b les ca n b e ex p ressed in term s o f
th e d im en sio n less q u a n tities x ´ (r = L 0 ); n ´ (½ = ½ c );
m = (M (r) =M 0 ); y ´ ¯ [Á ¡ Á (0 ))]; w h ere M (r ) is th e
m a ss in sid e a sp h ere o f ra d iu s r. T h ese va ria b les sa tisfy

RESONANCE  October 2008 943

 SERIES  ARTICLE

th e eq u a tio n s:

y 0= m = x 2 ; m 0
= n x2; n 0= ¡ m n = x 2 : (7 )

In term s o f y (x ) th e iso th erm a l eq u a tio n , (5 ), b eco m es

2
We have assumed that the
system is spherically symmet-
ric; it turns out that this is indeed
the extremal solution.
à !
1 d dy
x 2
= e¡ y (8 )
x2 dx dx
w ith th e b o u n d a ry co n d itio n y (0 ) = y 0(0 ) = 0 .2 L et u s
co n sid er th e n a tu re o f th e so lu tio n s to th is eq u a tio n .
B y d irect su b stitu tio n , w e see th a t n = (2 = x 2 ) ;m =
2 x ;y = 2 ln x sa tisfy (7 ) a n d (8 ). T h is so lu tio n , h ow ev er,
is sin g u la r a t th e o rig in a n d h en ce is n o t p h y sica lly a d -
m issib le. T h e im p o rta n ce o f th is so lu tio n lies in th e fa ct
th a t { a s w e w ill see { a ll o th er (p h y sica lly a d m issib le)
so lu tio n s ten d to th is so lu tio n [1 , 2 ] fo r la rg e va lu es o f x .
T h is a sy m p to tic b eh av io r o f a ll so lu tio n s sh ow s th a t th e
d en sity d ecrea ses a s (1 = r 2 ) fo r la rg e r im p ly in g th a t th e
m a ss co n ta in ed in sid e a sp h ere o f ra d iu s r in crea ses a s
M (r ) / r a t la rg e r . O f co u rse, in o u r ca se, th e sy stem
is en clo sed in a sp h erica l b ox o f ra d iu s R w ith a g iv en
m a ss M .
E q u a tio n (8 ) is in va ria n t u n d er th e tra n sfo rm a tio n y !
y + a ; x ! k x w ith k 2 = e a . T h is in va ria n ce im p lies
th a t, g iv en a so lu tio n w ith so m e va lu e o f y (0 ), w e ca n
o b ta in th e so lu tio n w ith a n y o th er va lu e o f y (0 ) b y sim -
p le resca lin g . T h erefo re, o n ly o n e o f th e tw o in teg ra tio n
co n sta n ts n eed ed in th e so lu tio n to (8 ) is rea lly n o n -
triv ia l. H en ce it m u st b e p o ssib le to red u ce th e d eg ree
o f th e eq u a tio n fro m tw o to o n e b y a ju d icio u s ch o ice o f
va ria b les. O n e su ch set o f va ria b les is:
m nx3 nx2
v ´ ; u ´ = : (9 )
x m v
In term s o f v a n d u , (5 ) b eco m es
u dv (u ¡ 1 )
= ¡ : (1 0 )
v du (u + v ¡ 3 )

944 RESONANCE  October 2008

 SERIES  ARTICLE

T h e b o u n d a ry co n d itio n s y (0 ) = y 0(0 ) = 0 tra n sla te in to

th e fo llow in g : v is zero a t u = 3 , a n d (d v = d u ) = ¡ 5 = 3 a t
(3 ,0 ). (Y o u ca n p rov e th is b y ex a m in in g th e b eh av io u r
o f (7 ) n ea r x = 0 reta in in g u p to n ecessa ry o rd er in
x ; try it o u t!) T h e so lu tio n v (u ) to eq u a tio n (1 0 ) ca n
b e ea sily o b ta in ed n u m erica lly : it is p lo tted in F igu re
1 a s th e sp ira llin g cu rv e. T h e sin g u la r p o in ts o f th is
d i® eren tia l eq u a tio n a re g iv en b y th e lo ca tio n in th e u v
p la n e a t w h ich b o th th e n u m era to r a n d d en o m in a to r o f
th e rig h t h a n d sid e o f (1 0 ) v an ish . S o lv in g u = 1 a n d
u + v = 3 sim u lta n eo u sly, w e g et th e sin g u la r p o in t to
b e u s = 1 , v s = 2 . U sin g (9 ), w e ¯ n d th a t th is p o in t
co rresp o n d s to th e a sy m p to tic so lu tio n n = (2 = x 2 );m =
2 x . It is o b v io u s fro m th e n a tu re o f th e eq u a tio n th a t
th e so lu tio n cu rv e w ill sp ira l a ro u n d th e sin g u la r p o in t
a sy m p to tica lly a p p ro a ch in g th e n = 2 = x 2 so lu tio n a t
la rg e x .
T h e n a tu re o f th e so lu tio n sh ow n in F igu re 1 a llow s u s Figure 1. Bound on RE/GM2
to p u t in terestin g b o u n d s o n va rio u s p h y sica l q u a n tities for the isothermal sphere.

RESONANCE  October 2008 945

 SERIES  ARTICLE

in clu d in g en erg y. T o see th is, w e sh a ll co m p u te th e to ta l

en erg y E o f th e iso th erm a l sp h ere. T h e p o ten tia l a n d
k in etic en erg ies a re
ZR
G M (r) d M G M 02 Z x 0
U = ¡ dr = ¡ m n xdx;
0 r dr L0 0
3M 3 G M 02 G M 02 3 Z x 0 2
K = = m (x 0 ) = n x dx;
2 ¯ 2 L0 L0 2 0
(1 1 )

w h ere x 0 = R = L 0 is th e b o u n d a ry a n d th e ex p ressio n
fo r K fo llow s fro m th e v elo city d ep en d en ce o f f in (2 ).
T h e to ta l en erg y is, th erefo re,

G M 02 Z x 0
E = K + U = d x (3 n x 2 ¡ 2 m n x )
2L 0 0
G M 02 Z x 0 d
= d x f2n x 3 ¡ 3m g
2L 0 0 dx
2 ½ ¾
G M 0 3 3
= n 0x 0 ¡ m 0 ;
L0 2
(1 2 )

w h ere n 0 = n (x 0 ) a n d m 0 = m (x 0 ). T h e d im en sio n less

q u a n tity (R E = G M 2 ) is g iv en b y
½ ¾
RE 1 3
¸ ´ 2
= u0 ¡ : (1 3 )
GM v0 2
N ote that the com bin ation (R E = G M 2 ) is a fu n ction on ly
of (u ;v ) at the bou n dary. L et u s n ow co n sid er th e co n -
stra in ts o n ¸ . S u p p o se w e sp ecify so m e va lu e fo r ¸ b y
sp ecify in g R ;E a n d M . T h en su ch a n iso th erm a l sp h ere
m u st lie o n th e cu rv e
µ ¶
1 3 R E
v = u ¡ ; ¸ ´ 2
(1 4 )
¸ 2 G M
w h ich is a stra ig h t lin e th ro u g h th e p o in t (1 :5 ;0 ) w ith
a slo p e ¸ ¡1 . O n th e o th er h a n d , sin ce all iso th erm a l
sp h eres m u st lie o n th e u ¡ v cu rv e, an isotherm al sphere

946 RESONANCE  October 2008

 SERIES  ARTICLE

can exist on ly if the lin e in equ ation (14) in tersects the

u ¡ v cu rve.
F o r la rg e p o sitiv e ¸ (p o sitiv e E ) th ere is ju st o n e in -
tersectio n . W h en ¸ = 0 , (zero en erg y ) w e still h av e a
u n iq u e iso th erm a l sp h ere. (F or ¸ = 0 , (1 4 ) rep resen ts
a v ertica l lin e th ro u g h u = 3= 2 .) W h en ¸ is n eg a tiv e
(n eg a tiv e E ), th e lin e ca n cu t th e u ¡ v cu rv e a t m o re
th a n o n e p o in t; th u s m o re th a n o n e iso th erm a l sp h ere
ca n ex ist w ith a g iv en va lu e o f ¸ . (O f co u rse, sp ecify in g
M ;R ;E in d iv id u a lly w ill rem ov e th is n o n -u n iq u en ess).
B u t a s w e d ecrea se ¸ (m o re an d m o re n eg a tiv e E ) th e
lin e in (1 4 ) w ill slo p e m o re a n d m o re to th e left; a n d
w h en ¸ is sm a ller th a n a critica l va lu e ¸ c , th e in tersec-
tio n w ill cea se to ex ist. S o w e rea ch th e k ey co n clu sio n
th a t n o isotherm al sphere can exist if (R E = G M 2 ) is be-
low a critical valu e ¸ c . T h is fa ct3 fo llow s im m ed ia tely 3
This derivation is due to the
fro m th e n a tu re o f th e u ¡ v cu rv e a n d (1 4 ). T h e v a lu e o f author [3,1]. It is surprising that
¸ c ca n b e fo u n d fro m th e n u m erica l so lu tio n a n d tu rn s Chandrasekhar, who has worked
out the isothermal sphere in u–v
o u t to b e a b o u t ¡ 0 :3 3 5 . coordinates as early as 1939,
missed discovering the energy
W h a t d o es th is resu lt m ea n ? T o u n d ersta n d its im p lica -
bound shown in Figure 1.
tio n s, co n sid er co n stru ctin g su ch a sy stem w ith a g iv en Chandrasekhar [2] has the u–v
m a ss M , ra d iu s R a n d a n en erg y E = ¡ jE j w h ich is curve but does not over-plot lines
n eg a tiv e. (T h e la st co n d itio n m ea n s th a t th e sy stem is of constant  . If he had done
g rav ita tio n a lly b o u n d .) In th is ca se, ¸ = R E = G M 2 = that, he would have discovered
Antonov instability decades be-
¡ R jE j= G M 2 is a n eg a tiv e n u m b er b u t let u s a ssu m e
fore Antonov did [4].
th a t it is a b ov e th e critica l va lu e; th a t is, ¸ > ¸ c . T h en
w e k n ow th a t a n iso th erm a l sp h ere so lu tio n ex ists fo r th e
g iv en p a ra m eter va lu es. B y co n stru ctio n , th is so lu tio n
is th e lo ca l ex trem u m o f th e en tro p y a n d co u ld rep re-
sen t a n eq u ilib riu m co n ¯ g u ra tio n if it is a lso a g lo b a l
m a x im u m o f en tro p y.
But for the system we
B u t fo r th e sy stem w e a re co n sid erin g , it is a ctu a lly are considering, it is
q u ite ea sy to see th a t th ere is n o g lo b a l m a x im u m fo r actually quite easy to
en tro p y. T h is is b eca u se, fo r a sy stem o f p o in t p a rticles see that there is no
in tera ctin g v ia N ew to n ia n p oten tia l, th ere is n o low er global maximum for
b o u n d to th e g rav ita tio n a l p oten tia l en erg y. If w e ta k e entropy.

RESONANCE  October 2008 947

 SERIES  ARTICLE

Suggested Reading a n a m o u n t o f m a ss m < M a n d fo rm a co m p a ct co re o f

ra d iu s r in sid e th e sp h erica l cav ity, th en b y d ecrea sin g
[1] T Padmanabhan, Physics
r o n e ca n su p p ly a rb itra rily la rg e a m o u n t o f en erg y to
Reports, Vol.188, p.285,
1990.
th e rest o f th e p a rticles. V ery so o n , th e rem a in in g p a r-
[2] S Chandrasekhar, An In- ticles w ill h av e v ery la rg e k in etic en erg y co m p a red to
troduction to the Study of th eir g rav ita tio n a l p o ten tia l en erg y a n d w ill essen tia lly
Stellar Structure, Dover b o u n ce a ro u n d in sid e th e sp h erica l cav ity lik e a n o n -
,1939.
in tera ctin g g a s o f p a rticles. T h e co m p a ct co re in th e
[3] T Padmanabhan, Astro-
phys. Jour. Supp., Vol.71,
cen ter w ill co n tin u e to sh rin k th ereb y su p p ly in g en erg y
p.651, 1989. to th e rest o f th e p a rticles. It is ea sy to see th a t su ch a
[4] V A Antonov, Vest. Lenin- co re{ h a lo co n ¯ g u ra tio n ca n h av e a rb itra rily h ig h va lu es
grad Univ., Vol.7, p.135, fo r th e en tro p y. A ll th is g o es to sh ow th a t th e iso th er-
1962. Translation: IAU
m a l sp h ere ca n n o t b e a global m a x im u m fo r th e en tro p y.
Symposium, Vol.113,
p.525, 1985.
(T h is w a s th e cav ea t in th e ca lcu la tio n w e p erfo rm ed
to d eriv e th e iso th erm a l sp h ere eq u a tio n ; w e ta citly a s-
su m ed th a t th e ex trem u m co n d itio n ca n b e sa tis¯ ed fo r
a ¯ n ite va lu e o f en tro p y.)
If w e in crea se th e ra d iu s o f th e sp h erica l b ox (w ith so m e
¯ x ed va lu e fo r E = ¡ jE j), th e p a ra m eter ¸ w ill b eco m e
m o re a n d m o re n eg a tiv e a n d fo r su ± cien tly la rg e R , w e
w ill h av e a situ a tio n w ith ¸ < ¸ c . N ow th e situ a tio n
g ets w o rse. T h e sy stem d o es n o t ev en h av e a lo ca l ex -
trem u m fo r th e en tro p y a n d w ill ev o lv e d irectly tow a rd s
a co re{ h a lo co n ¯ g u ra tio n . T h is is clo sely rela ted to a
p h en o m en o n ca lled A n to n ov in sta b ility [4 , 3 ].
In rea l life, o f co u rse, th ere is a lw ay s so m e sh o rt d is-
ta n ce cu t-o ® b eca u se o f w h ich th e co re ca n n o t sh rin k
to a n a rb itra rily sm a ll ra d iu s. In su ch a ca se, th ere is
a g lo b a l m a x im u m fo r en tro p y a ch iev ed b y th e (¯ n ite)
Address for Correspondence
co re{ h a lo co n ¯ g u ra tio n w h ich co u ld b e th o u g h t o f a s th e
T Padmanabhan
IUCAA, Post Bag 4 ¯ n a l sta te in th e ev o lu tio n o f su ch a sy stem . It w ill b e
Pune University Campus h ig h ly in h o m o g en eo u s a n d , in fa ct, is v ery sim ila r to a
Ganeshkhind sy stem w h ich ex ists a s a m ix tu re o f tw o p h a ses. T h is is
Pune 411 007, India. o n e k ey p ecu lia rity in tro d u ced b y lo n g ra n g e a ttra ctiv e
Email:
in tera ctio n s in sta tistica l m ech a n ics.
paddy@iucaa.ernet.in
nabhan@iucaa.ernet.in

948 RESONANCE  October 2008

 SERIES  ARTICLE

A p p e n d ix

In th e tex t, I d eriv ed eq u a tio n (3 ) fro m th e ex p ressio n fo r en tro p y in eq u a tio n (1 ).

G iv en th e p ecu lia rities o f g rav ita tin g sy stem s o n e m ay w o n d er h ow tru stw o rth y
th is a p p ro a ch is. H ere I d escrib e b rie° y a m o re b a sic d eriva tio n o f th e ex p ressio n
in (3 ).
T h e eq u ilib riu m sta te is th e o n e th a t m a x im izes th e en tro p y o f th e sy stem . W h en
w e stu d y th e sy stem in th e m icro ca n o n ica l en sem b le, th is en tro p y S is th e lo g a -
rith m o f th e v o lu m e g (E ) o f th e p h a se sp a ce ava ila b le to th e sy stem if th e to ta l
en erg y is E . T h a t is:

Z
S 1
e = g (E ) = d 3 N x d 3 N p ± (E ¡ H ); (1 5 )
N !
w h ere H is th e H a m ilto n ia n fo r th e sy stem o f N p a rticles g iv en b y th e su m o f
th e k in etic en erg ies (p 2i = 2 m ) (i = 1 ;2 ;:::;N ) a n d th e p o ten tia l en erg y o f p a irw ise
in tera ctio n . T h e D ira c d elta fu n ctio n tells u s th a t th e sy stem is co n ¯ n ed to th e
b o u n d a ry o f a 3 N d im en sio n a l sp h ere in m o m en tu m sp a ce g iv en b y th e eq u a tio n
2 3
XN 1X
p 2i = 2 m 4E ¡ U (x i;x j )5 ´ l2 : (1 6 )
i= 1 2 i6= j

O b v io u sly, th e m o m en tu m in teg ra tio n in (15 ) w ill lea d to a term p ro p o rtio n a l to

l3 N ¡1 ; so , p erfo rm in g th e m o m en tu m in teg ra tio n s a n d u sin g N À 1 , w e g et
2 33 N
Z 2
1 1X
e S = g (E ) / d 3N x 4E ¡ U (x i ;x j )5 : (1 7 )
N ! 2 i6= j

T h e in teg ra l in (1 7 ) is im p o ssib le to eva lu a te fo r a n y rea listic p o ten tia l b u t th ere

is a sta n d a rd a p p rox im a tio n u sin g w h ich w e ca n m a p th is p ro b lem to a m o re
tra cta b le o n e.
L et u s d iv id e th e th e sp a tia l v o lu m e V in to J (w ith J ¿ N ) cells o f eq u a l size,
la rg e en o u g h to co n ta in m a n y p a rticles b u t sm a ll en o u g h fo r th e p o ten tia l to b e
trea ted a s a co n sta n t w ith in ea ch cell. (W e w ill a ssu m e th a t su ch a n in term ed ia te
sca le ex ists, w h ich u su a lly d o es.) In stea d o f in tegratin g ov er a ll th e p a rticle
co o rd in a tes (x 1 ;x 2 ;:::;x N ) w e sh a ll su m over th e n u m b er o f p a rticles n a in th e
cell cen tered a t x a (w h ere a = 1 ;2 ;:::;J ) th ereb y a p p rox im a tin g th e in teg ra l b y

RESONANCE  October 2008 949

 SERIES  ARTICLE

a d iscrete su m . U sin g th e o b v io u s fa ct th a t th e in teg ra tio n ov er (N !)¡ 1 d 3 N x ca n

b e rep la ced b y
1
X µ ¶ X1 µ ¶ X1 µ ¶ µ ¶N
1 1 1 V
::: (1 8 )
n 1= 1 n 1! n 2= 1 n 2! nJ= 1 nJ! J
P
(su b ject to th e co n stra in t th a t n a = N ) a n d a p p rox im a tin g th e n ! s b y S tirlin g 's
a p p rox im a tio n : ln n ! = n ln (n = e ) w e g et, a fter so m e stra ig h tfo rw a rd a lg eb ra th e
resu lt: 1 1 1
X X X
eS ¼ ::: ex p S [f n a g]; (1 9 )
n 1= 1 n 2= 1 nJ= 1

w h ere
2 3
µ ¶
3N 1 XJ XJ n aJ
S [f n a g ] = ln 4E ¡ n a U (x a ;x b )n b 5 ¡ n a ln : (2 0 )
2 2 a6
= b a= 1 eV
If th e n u m b er o f p a rticles in ea ch cell is su ± cien tly la rg e { a s u su a lly is th e
ca se { th is is n o t a d ra stic a p p rox im a tio n . W e a re, o f co u rse, in terested in th e
co n ¯ g u ra tio n o f f n a g fo r w h ich th e su m m a n d in (1 9 ) rea ch es th e m a x im u m va lu e,
su b ject to th e co n stra in t o n th e to ta l n u m b er. T h is w o rk s in sta n d a rd sta tistica l
m ech a n ics b eca u se, in m o st ca ses o f in terest, th e la rg est term a ctu a lly d o m in a tes
th e su m a n d th e erro r in v o lv ed in ig n o rin g th e rest is sm a ll. T h a t is, to a h ig h
o rd er o f a ccu ra cy S = S [n a ;m a x ], w h ere n a ;m a x is th e so lu tio n to th e va ria tio n a l
p ro b lem (± S = ± n a ) = 0 w ith th e su m o f p a rticle n u m b ers n a b ein g k ep t eq u a l to
N . Im p o sin g th is co n stra in t b y a L a g ra n g e m u ltip lier a n d u sin g th e ex p ressio n
(2 0 ) fo r S , w e o b ta in th e eq u a tio n sa tis¯ ed b y n a ;m a x :
µ ¶
1 XJ n a ;m a x J
U (x a ;x b )n b;m ax + ln = co n sta n t; (2 1 )
T b= 1 V
w h ere w e h av e de¯ n ed th e tem p era tu re T th ro u g h th e rela tio n :
0 1 ¡1
1 3N @ 1 XJ
= E ¡ n a U (x a ;x b )n b A = ¯ (2 2 )
T 2 2 a 6= b
w ith n a = n a ;m a x . T o see th a t th is is n o t a s stra n g e a s it lo o k s, y o u o n ly n eed to
n o te, fro m (2 0 ) th a t th is ¯ is a lso eq u a l to (@ S = @ E ); th erefo re o n e ca n th in k o f
T a s th e co rrect th erm o d y n a m ic tem p era tu re. W e ca n n o w retu rn b a ck to th e
co n tin u u m lim it o f (2 1 ) b y w ritin g n a ;m a x (J = V ) = ½ (x a ) a n d rep la cin g th e su m
ov er p a rticles b y in teg ra tio n w ith th e m ea su re J = V . In th is co n tin u u m lim it, th e
ex trem u m so lu tio n in (2 1 ) is g iv en p recisely b y (3 ).

950 RESONANCE  October 2008