The information geometric descriptions of

denormalized thermodynamic manifolds

Shicheng Zhang, Huafei Sun

Abstract. In view of information geometry, Pthe state space of thermody-

n
namic parameters S = {| = Z 1 exp{ i=1 i Fi , Tr = 1} has been
investigated. Here the -geometric structures of the denormalization of
S called Se = {e|e
= f ( ), f ( ) > 0, Tr = 1} is investigated. The co-
variant derivative and the -curvature tensor is studied. Therefore, the
relation of the -geometric structures between S and Se are obtained. The
results were obtained in [8] when = 0 and in [2] when f ( ) = 1.

M.S.C. 2000: 53C15, 62B10.

Key words: Thermodynamic parameters; information geometry; -curvature tensor.

1 Introduction
Information geometry ([1]) originated from investigating the geometric structures of
the manifold which consists of probability density functions and it has various appli-
cations such as in statistical inference and neural networks.
Recently, some authors ([7, 6, 5, 3, 4, 2]) considered the geometric structure of the
space of thermodynamic parameters, which forms a manifold called S. This manifold
characterizes a given physical system. One of the main results they obtained is to
give the Riemannian Gaussian curvature of the manifold S. In the present paper,
we firstly define the -connection and obtain the -Gaussian curvature which will
becomes the Riemannian Gaussian curvature when = 0. Ingarden ([3]), Janyzek
([4]), Zheng Z. [10] and other authors reaching a Riemannian metric by Pstatistical
n
method. For a given equilibrium density operator ([7]) = Z 1 exp{ Pn i=1 i Fi },
where F1 , F2 , , Fn are linear independent operators, Z = Tr exp{ i=1 i Fi } is
a partition function and = (1 , 2 , , n ) are classical real parameters (statistical
temperature, press, magnetic field), which describes the environment of a physical
system. The thermodynamic parameters set S = {| Tr = 1} can be regarded as a
differential manifold equipped with a Riemannian metric

gij = Tr[(i ln )(j ln )]

for the case of commuting operators Fi , where Tr denotes the trace of the matrix,
and i means the partial derivative with respect to the parameter i .
Applied Sciences, Vol.12, 2010, pp.
158-166.
c Balkan Society of Geometers, Geometry Balkan Press 2010.

The information geometric descriptions ... 159

In [1], Amari investigated the denormalization of statistical model. Here we con-

sider the denormalization of thermodynamic parameters model, a more general man-
ifold, namely
Se = {e
|e
= f ( ), f ( ) > 0, Tr = 1},
where f ( )(> 0) is an arbitrary-order differentiable function. Then Se is a manifold
which contains S as a submanifold with dim Se = dim S + 1. Se is called the denormal-
ization of S. By calculating the covariant derivative and the -curvature tensor, we
obtain the relation of the -geometric structures between S and S.e At last, we use a
example to illustrate our conclusions.

2 The information geometric structures of the man-

ifold Se
Parameterizing the elements of Se as e by a coordinate system [, ] = (1 , , n , )
and letting
2 e1 2 ( 6= 1)
e()
l := 1

ln e ( = 1),
e () are
the components of the -metric ge() and the coefficients of the -connection
represented respectively by

geij = Tr(ie
()
(2.1) l() j e
l() ),
(2.2) e () = Tr(i j e
l() k e
l() ).
ijk

We denote its natural basis by ei =

i
, e =

and use , , , to denote the
subscripts of the natural basis, that is, , , , {1, , n, n + 1}, while i, j, k, l
{1, , n}. For the case of commuting operators Fi , from (2.1), we see that

e ln ee ln e).
ge = Tr(e

By a calculation, we get

geij = f ( )gij , gei = 0, ge = f ( )1 ( f ( ))2 .

The Riemannian metric of Se and its inverse are given by

e f ( )G 0 e 1 f ( )1 G1 0
(2.3) G = , G = .
0 f ( )1 ( f ( ))2 0 f ( )( f ( ))2

The square of the arc length of S defined by ds2 = gij di dj . Thus, the square of
the arc length of Se is given by

s2 = f ( )ds2 + f ( )1 ( f ( ))2 d 2 .
de

Note that the volume element of S is given by

p
dV = det(G)d1 dn ,
160 Shicheng Zhang, Huafei Sun

we have the relation between the volume elements dV and dVe

q
dVe = det(G)de 1 dn d
n1 p
= (f ( )) 2 f ( ) det(G)d1 dn d
n1
= dV (f ( )) 2 df ( ).

e () for Se satisfy
Proposition 1. The -connection coefficients of

e () = f ( )() ,
e () = 1 + f ( )gij ,

ij,k ij,k ij,
2
(2.4) e () =
e () = 1 f ( )gik , e

() e
=
()
= 0, e () = 0,

i,k i,k i, i, ,k
2
e () 1+
, = f ( )2 ( f ( ))3 + f ( )1 f ( ) f ( ),
2
and

e ()k = f ( )()k ,
e () = 1 + f ( )( f ( ))1 gij ,

ij ij ij
2
(2.5) e ()k = 1 f ( )1 f ( )ik ,
e ()k = e

()
=e ()
= 0, e ()k
= 0,
i i i i
2
e () 1+
= f ( )1 f ( ) + ( f ( ))1 f ( ),
2
where the -connection coefficients of () for S satisfies

() 1
ijk () = i j k ln Z.
2

Proof. By (2.2) and notice that

1 1
ei ej e
l() = f ( ) 2 i j l() , f eie
l() = f ( ) f ( )i l() ,
2
f
f e 1+ 3+ 1 1+ 1
l() = f ( ) 2 ( f ( ))2 2 + f ( ) 2 f ( ) 2 ,
2
we can obtain proposition 1.
These equations enable us to verify that the following relations hold for the co-
e (on S),
variant derivatives (on S) and e respectively:

(2.6) e e Ye = (X Y ) 1 + ( f ( ))1 < X,

e Ye > e ,
X 2

(2.7) e e e =
ee X e = 1 f ( )1 f ( )X,
e
X 2

(2.8) e e e = 1 + f ( )1 f ( ) + ( f ( ))1 f ( ) e ,
2
The information geometric descriptions ... 161

]
where X and Y are arbitrary vector fields on S, and ( ) denotes ( ).
From proposition 1, we can obtain the following
e
Theorem 1. S is (-1)-autoparallel in S.

Theorem 2. Let M be a submanifold of S and M f be its denormalization. For

every R, the following conditions (i) and (ii) are equivalent.
(i) M is -autoparallel in S.
f is -autoparallel in S.
(ii) M e
Proof. Let and be the -connections on Se and S as above. Noting that
e () ()

every vector field on M f can be represented as hi X f | f by vector fields {Xi } on

fi + h0
M
i 0
M and functions {h } and h on M , we see from (2.7) and (2.8) that condition (ii) is
equivalent to stating that e () Ye T (M
f) for all X, Y T (M ). On the other hand,
Xe
we have for all X, Y T (M ),

e () Ye T (M
f) (() Y ) T (M
f) () Y T (M ),
Xe X X

where the first equivalence follows from (2.6) and the second one is obvious. Therefore,
(i) and (ii) are equivalent.
Remark 1. The result is Theorem 2.10 in [1] when f ( ) = .

Proposition 2. The components of the -curvature tensor of Se are given by

2
e() = f ( )R() 1 f ( )(gil gjk gjl gik ),
R ijkl ijkl
4
e () e ()
R ijk = R i k = 0,

where the components of -curvature tensor of S satisfy

() 1 2
Rijkl = (k m i ln Zj l n ln Z k m j ln Zi l n ln Z)g mn .
4

Proof. Since

R f
e() = ( e () f e () e () e () e () e ()
)g + ( ),

combining (2.3), (2.4) and (2.5), we obtain

2
e() = f ( )R() + 1 f ( )(gil gjk gjl gik ),
R e() = 0,
R
ijkl ijkl i k
4
e() = 1 + f ( )(i gjk j gik + () () ).
Rijk jkl ikj
2
From (2.2), we get

() 1 1
ijk () = Tr[(i j ln )(k ln )] + Tr[(i ln )(j ln )(k ln )] = i j k ln Z.
2 2
162 Shicheng Zhang, Huafei Sun

On the other hand, from (2.1), we get

k gij = k (Tr[i ln j ln ])
= Tr[k ln i ln j ln ] + Tr[k i ln j ln ] + Tr[i ln k j ln ]
= i j k ln Z.
From above we can get
e() = 0
Rijk

and
() 1 2
Rijkl = (k m i ln Zj l n ln Z k m j ln Zi l n ln Z)g mn .
4
This finishes the proof of proposition 2.
Remark 2. Clearly manifold S and Se are 1-flat manifolds.
By a direct calculation, we obtain the -sectional curvatures and the -scalar
curvature of Se
e()
R 1 2
e () = ijij ()
(2.9) K ijij = (f ( ))1 Kijij (f ( ))1 ,
geii gejj geij geji 4
and
2
e() ge ge = f ( )1 R() + 1 f ( )1 (n n2 ).
e() = R
R
4
When n = 2, from (2.9), we see that the -Gaussian curvature of Se satisfies

1 2
12 ln Z 13 ln Z 12 2 ln Z 1 2
e () = f ( ) (1 ) 1 2 ln Z 12 2 ln Z 1 22 ln Z f ( ) (1 ) .
K
4 det(G)2 22 ln Z 4
1 22 ln Z 23 ln Z
Remark 3. The result is the conclusion in [8] when = 0.
Here, we will use an example to verify our conclusion above.
P2
Example 1. Suppose that S = {| = Z 1 exp i=1 i Fi }, and the matrices
Fi are given by

0 0 0 1 1 0
F1 = 1 1 0 , F2 = 0 0 0 .
1 1 0 1 1 0
Taking f ( ) = > 0, we get
2
X
Se = {e = Z 1 exp
|e i Fi }.
i=1

An equilibrium density S can be represented by

2 2 0
1
= exp 1 1 0 ,
Z
1 + 2 1 2 0
The information geometric descriptions ... 163

and the partition functions of is Z = exp[1 ] + exp[2 ] + 1. Under the coordinate

system (1 , 2 ), the geometric metrics of S are given by

G = (gij ) = (i j ln Z),

exp 2(1 + 2 )
det(G) = 12 ln Z22 ln Z (1 2 ln Z)2 = ,
(exp(1 ) + exp(2 ) + exp(1 + 2 ))3
and
12 ln Z 13 ln Z 12 2 ln Z
() 1 2
R1212 = 1 2 ln Z 12 2 ln Z 1 22 ln Z =
4 det(G)
22 ln Z 1 22 ln Z 23 ln Z
(1 2 ) exp 2(1 + 2 )
= ,
4(exp(1 ) + exp(2 ) + exp(1 + 2 ))3
()
() R1212 1 2 2 () 1 2
K1212 = = , R() = R1212 = .
det(G) 4 det(G) 2
Under the coordinate system (1 , 2 , ), we obtain the geometric metrics of Se
corresponding to those of S:

12 ln Z 1 2 ln Z 0
e = 1 2 ln Z G 0
G 22 ln Z 0 = ,
0 1
0 0 1

and
2
e() = R() 1 det(G) = 0,
R 1212 1212
4
e () e () e () e() = R
e() = 0,
R1 1 = R121 = R212 = R1 2 2 2
2
e () = 1 K () 1 1 = 0.
K 1212 1212
4
The scalar curvature of Se satisfies

1 2
12 ln Z 13 ln Z 12 2 ln Z 1 2
e() = (1 ) 1 2 ln Z 12 2 ln Z 1 22 ln Z (1 ) = 0.
R
2 det(G)2 2
22 ln Z 1 22 ln Z 23 ln Z

3 The approximation of the equilibrium density

In this section, we firstly define a Kullback-Leibler divergence, which is different from
the general distance of two points in the manifold.
Definition 1. Let e and e the Kullback-Leibler
e be two points in manifold S,
divergence between two states is defined by

(3.1) D(e
|e
) = Tr[e
(ln e ln
e)].

So, we have the relation of the Kullback-Leibler divergences between S and Se

(3.2) D(e
|e
) = f ( )D(|).
164 Shicheng Zhang, Huafei Sun

Definition 2. Let
r
X
(3.3) f=
M
e|e
= f ( )Z1 exp{ j E j }
j=1

be an r-dimensional manifold, where r n and Ej is related to Fi . M f is a submanifold

e
of S.
f is a 1-flat manifold, where = (1 , , r ) is the eaffine
Clearly submanifold M
f
coordinate system of M .
e the projection onto M
Definition 3. For e S, f is defined by
Y
(3.4) f =
e = arg min D(e
|e
).
e
eS

So we have the following proposition.

Proposition 3. Suppose that the projection from a point e in Se onto Mf is .

Then can be considered as a optimal approximation of the equilibrium density e in
f, and is a solution of the following differential equation
M

(3.5) ln Z = Tr[Ei ].
i

Proof. From (3.1) and (3.2) we have

D(e
|e
) = Tr[e(ln e ln
e)] = f ( )D(|)
Xn Xr
= f ( ) Tr[( i Fi ln Z + j Ej + ln Z )]
i=1 j=1
Xn r
X
= f ( ) i Tr[Fi ] ln Z + j Tr[Ej ] + ln Z .
i=1 j=1

Since
D(e
|e
(, ))
= f ( ) Tr[ ln ] = f ( )(Tr[Ei ] + ln Z ) = 0,
i i i
we obtain (3.5). This finishes the proof of Proposition 3.
Example 2. Let S be a 3-dimensional manifold and take
3
X
Se = e = f ( )Z1 exp i Fi = e = f ( ) ,
i=1

where = (1 , 2 , 3 ) is the e-affine coordinate system of S, and the matrixes Fi are

given by

1 0 0 0 0 0 0 0 0
F1 = 0 0 0 , F2 = 0 1 0 , F 3 = 0 0 0 .
0 0 0 0 0 0 0 0 1
The information geometric descriptions ... 165

f of Se satisfies
Then submanifold M
2
X
f=
M e = f ( )Z1 exp
j Ej =
e = f ( ) ,
j=1

where = (1 , 2 ) is the e-affine coordinate system of M , and the matrixes Ej are

given by

1 0 0 0 0 0
E1 = 0 1 0 , E2 = 0 0 0 .
0 0 0 0 0 1

So the equilibrium density e and

e can be written as

exp{1 } 0 0
f ( ) ,
e = 0 exp{2 } 0
Z ()
0 0 exp{3 }

and

exp{1 } 0 0
f ( ) ,

e= 0 exp{1 } 0
Z ()
0 0 exp{2 }

respectively.
From (3.5), we have

2 exp{1 } exp{1 } + exp{2 }

= ,
2 exp{1 } + exp{2 } exp{1 } + exp{2 } + exp{3 }
exp{2 } exp{3 }
= .
2 exp{1 } + exp{2 } exp{1 } + exp{2 } + exp{3 }

So we obtain the coordinates i of the optimal approximation of e as

exp{1 } + exp{2 }
1 = ln , 2 = 3 .
2

4 Conclusions
In this paper, we investigate the structures of the state space of the thermodynamic
parameters based upon the information geometric approach to the denormalization
Se of S. By calculating the covariant derivative and the -curvature tensor, we ob-
tain the relation of the -geometric structures between S and S. e At last, we study
the approximation of the equilibrium density and use two examples to illustrate our
conclusions.

Acknowledgement. This subject is supported by the National Natural Science

Foundation of China (No. 10871218).
166 Shicheng Zhang, Huafei Sun

References
[1] S. Amari and H. Nagaoka, Methods of Information Geometry, AMS and Oxford
University Press, 2000.
[2] L. Cao, H. Sun and L. Wu, Information geometry of the thermodynamic manifold,
II Nuovo Cimento 123 (2008), 593-598.
[3] R.S. Ingarden, Differential geometry and physics, Tensor N.S. 30 (1976), 201-209.
[4] H. Janyszek and R. Mrugala, Riemannian geometry and the thermodynamics of
model magnetic systems, Phys. Rev. A 39 (1989), 6515-6523.
[5] D.A. Johnston, W. Janke and R. Kenna, Information geometry, one, two, three.,
Acta Phys. Pol. B 34 (2003), 4923-4937.
[6] M. Portesi, F. Pennini and A. Plastino, Geometrical aspects of a generalized
statistical mechanics, Phys. A 373 (2007), 273-279.
[7] M. Portesi, A. Plastino and F. Pennini, Information geometry and phase transi-
tions, Int. J. of Mod. Phys. B 20 (2006), 5250-5253.
[8] L. Wu, H. Sun and L. Cao, Geometrical description of denormalized thermody-
namic manifold, Chin. Phys. B 18 (2009), 3790-3794.
[9] Z. Zhang, H. Sun and F. Zhong, Geometrical structures of certain class of sta-
tistical manifolds, J. of BIT 17 (2008), 368-372.
[10] Z. Zhang, H. Sun and F. Zhong, Information geometry of the power inverse
Gaussian distribution, Appl. Sci. 9 (2007), 194-203.

Authors addresses:
Shicheng Zhang
Department of Mathematics, Beijing Institute of Technology,
Beijing, 100081 China.
School of Mathematics, Xuzhou Normal University,
Xuzhou, Jiangsu 221116, China.
E-mail: zhangshicheng@126.com

Huafei Sun
Department of Mathematics, Beijing Institute of Technology, Beijing, 100081 China
E-mail: huaf.sun@gmail.com