a

r
X
i
v
:
h
e
p
-
t
h
/
0
0
0
1
0
7
8
v
1


1
2

J
a
n

2
0
0
0
A note on Riemann normal coordinates.
Agapitos Hatzinikitas

University of Athens,
Nuclear and Particle Physics Division,
Panepistimioupoli GR-15771 Athens, Greece.
Abstract
The goal of this note is to provide a recursive algorithm that allows one to calculate the
expansion of the metric tensor up to the desired order in Riemann normal coordinates. We
test our expressions up to fourth order and predict results up to sixth order. For an arbitrary
number of symmetric partial derivatives acting on the components of the metric tensor subtle
treatment is required since the degree of complication increases rapidly.
1 Introduction
Riemann normal coordinates (abbreviated as RNC from now on) have a vast eld of applications
to Physics. In particular they constitute the main ingredient of the background eld method [1]
widely used for nonlinear - model calculations in curved spacetime. They have the appealing
feature that the geodesics passing through the origin have the same form as the equations
of straight lines passing through the origin of a Cartesian system of coordinates in Euclidean
geometry.
The purpose of this letter is to bridge the gap one confronts when attempts to perform
computations that involve expansion of the metric tensor beyond the fourth order [2]. We
present the general method for breaking the barrier of the fourth order and at the same time
we supply the reader with formulae adequate to reach the order of its choice.
The starting point will be a geodesic, on a compact n-dimensional Riemannian manifold,
parametrized by x
l
(s) and satisfying the dierential equation:
d
2
x
l
ds
2
+
l
jk
(x)
dx
j
ds
dx
k
ds
= 0 (1)
where s is the arc length and
l
jk
denotes the Christoel symbol for the Levi-Civita connection.
Any integral curve of (1) is completely determined by a point P(x
1
0
, x
2
0
, , x
n
0
) and a direction
dened by the tangent vector
l
=
_
dx
l
ds
_
P
. The power series solution of (1) is:
x
l
(s) = x
l
0
+

k=1
1
k!
_
d
k
x
l
ds
k
_
P
s
k
(2)
and its coecients can be replaced using successive dierentiations by:
_
dx
l
ds
_
P
=
l

e-mail: ahatzini@cc.uoa.gr
1
_
d
2
x
l
ds
2
_
P
=
_

l
i
1
i
2
_
P

i
1

i
2
_
d
3
x
l
ds
3
_
P
=
_

(i
1

l
i
2
i
3
)
2
l
(i
1

i
2
i
3
)
_
P

i
1

i
2

i
3
_
d
4
x
l
ds
4
_
P
=
_

(i
1

i
2

l
i
3
i
4
)

l
(i
1
i
2

i
3
i
4
)
4
(i
1

l
i
2

i
3
i
4
)
2
(i
1

i
2
i
3

l
i
4
)
+ 4
l
k(i
1

k
i
2

i
3
i
4
)
+ 2
l

(i
1
i
2

i
3
i
4
)
_
P

i
1

i
2

i
3

i
4
.
(3)
where dots in (3) represent higher derivative terms at the point P and symmetrization exclu-
sively relates the lower i indices

. Apparently the domain of convergence of (2) depends on
the components g
ij
and the values of
i
. However, for suciently small values of s it denes an
integral curve of (1).
2 The computational procedure
In RNC the geodesics through P are straight lines dened by:
y
l
=
l
s. (4)
This set of coordinates cannot be used to cover the whole manifold and is valid only for a small
neighbourhood of the point P where the conditions of the existence and uniqueness theorem of
dierential equation of the geodesics are satised. In this region no two geodesics through P
intersect due to one-one correspondence of x
l
and y
l
. Incidently, this displays the local nature
of RNC.
Substituting (4) into eq. (2) one has:
x
l
= x
l
0
+

k=1
1
k!
_

l
i
1
i
2
i
k
_
P
y
i
1
y
i
2
y
i
k
(5)
where
_

l
i
1
i
2
i
k
_
P
are the generalized Christoel symbols at the point P used in eq. (3). The
Jacobian

x
l
y
m

P
= 0 and thus the series (5) can be inverted.
For this local system of coordinates the geodesic equation can be written as:

l
ij
(y)y
i
y
j
= 0 (6)
and the power series solution becomes:
y
l
=

k=1
1
k!
_

l
i
1
i
2
i
k
_
P

i
1

i
2

i
k
s
k
. (7)
Reduction of (7) to (4) for arbitrary values of
l
implies that at the origin P holds:

l
(i
1
i
2
i
k
)
= 0 (8)

Proof of the fourth expression in (3) would require the use of the identity
dx

ds
D

dx

ds
= 0
2
or, by induction, one can easily prove that eq. (8) is equivalent to:

(i
1

i
2

i
k2

l
i
k1
i
k
)
= 0 (9)
Paraphrasing eq. (9) means that all symmetric derivatives of the ane connection vanish at
the origin in RNC.
Generally speaking, a covariant second rank tensor eld on a manifold can be expanded
according to:
T
k
1
k
2
(

) =

n=0
1
n!
_

i
1

i
2

in
_
T
k
1
k
2
()
i
1

i
2

in
. (10)
The coecients of the Taylor expansion are tensors and can be expressed in terms of the
components R
l
mnp

of the Riemann curvature tensor and the covariant derivatives D

k
T
lm
and
D
k
R
l
mnp
. One without much eort can prove that:

(i
1

i
2

i
n1

l
in)k
=
_
n 1
n + 1
_
_
D
(i
1
D
i
2
D
i
n2

R
l
i
n1
kin)
+
(i
1

i
2

i
n2
_

i
n1
k

l
in)
_

(i
1

i
2

i
n3
_

l
i
n2

i
n1
kin)
l , k
_

(i
1

i
2

i
n4
_

l
i
n3

D
i
n2

i
n1
kin)
l , k
_
.
.
.

(i
1

l
i
2

D
i
3
D
i
n2

i
n1
kin)
l , k
_ _
(11)
where the interchange of covariant and contravariant indices act independently. Expression (11)
reproduces for various values of n

the following results:

(i
1

l
i
2
)k
=
1
3

R
l
(i
1
i
2
)k

(i
1

i
2

l
i
3
)k
=
1
2
D
(i
1

R
l
i
2
ki
3
)

(i
1

i
2

i
3

l
i
4
)k
=
3
5
_
D
(i
1
D
i
2

R
l
i
3
ki
4
)
+
2
9

R
l
(i
1
i
2

i
3
i
4
)k
_

(i
1

i
2

i
3

i
4

l
i
5
)k
=
2
3
_
D
(i
1
D
i
2
D
i
3

R
l
i
4
ki
5
)
D
(i
1

i
2
ki
3

R
l
i
4
i
5
)
_

(i
1

i
2

i
3

i
4

i
5

l
i
6
)k
=
5
7
_
D
(i
1
D
i
4

R
l
i
5
ki
6
)

1
5
_
7D
(i
1
D
i
2

R
l
i
3
i
4

i
5
i
6
)k
+ D
(i
1
D
i
2

i
3
ki
4

R
l
i
5
i
6
)
_
+
3
2
D
(i
1

i
2
ki
3
D
i
4

R
l
i
5
i
6
)

16
45

R
l
(i
1
i
2

i
3
i
4

i
5
i
6
)k
_
(12)
The coecients of (10) can be rewritten as:

(i
1

i
2

in)

T
k
1
k
2
= D
(i
1
D
i
2
D
in)

T
k
1
k
2

Our conventions are: R

l
mnp
=
n

l
mp
+
k
mp

l
kn
(n p), R

= R

and R = R

In [1] there is a misprint for the n = 4 case. A minus sign is needed in front of the
2
9
-term.
3
+
(i
1

i
2

i
n1
)
_

ink
1
)

T
k
2
+ k
1
k
2
_
+
(i
1

i
2

i
n2
)
_

i
n1
k
1
D
in)

T
k
2
+ k
1
k
2
_
.
.
.
+
(i
1

i
2
_

i
3
k
1
D
i
4
D
in)

T
k
2
+ k
1
k
2
_
+
(i
1
_

i
2
k
1
D
i
3
D
in)

T
k
2
+ k
1
k
2
_
. (13)
Expressions (11) and (13) compose the building blocks of the current recursive method which
produces the following results for dierent values of n:

(i
1

i
2
)

T
k
1
k
2
= D
(i
1
D
i
2
)

T
k
1
k
2

1
3
_

R

(i
1
k
1
i
2
)

T
k
2
+ k
1
k
2
_

(i
1

i
2

i
3
)

T
k
1
k
2
= D
(i
1
D
i
2
D
i
3
)

T
k
1
k
2
+
_

(i
1

i
2

i
3
)k
1

T
k
2
+ 2
(i
1

i
2
k
1
D
i
3
)

T
k
2
+ k
1
k
2
_

1
3
_

R

(i
1
k
1
i
2
D
i
3
)

T
k
2
+ k
1
k
2
_

(i
1

i
2

i
3
i
4
)

T
k
1
k
2
= D
(i
1
D
i
2
D
i
3
D
i
4
)

T
k
1
k
2
+
_

(i
1

i
2

i
3

i
4
)k
1

T
k
2
+ 3
(i
1

i
2

i
3
k
1
D
i
4
)

T
k
2
+ 3
(i
1

i
2
k
1
_
D
i
3
D
i
4
)

T
k
2

1
3
_

R

i
3
i
4
)

T
k
2
+ k
2
_
_
+ k
1
k
2
_

(i
1

i
2

i
3
i
4

i
5
)

T
k
1
k
2
= D
(i
1
D
i
2
D
i
3
D
i
4
D
i
5
)

T
k
1
k
2

_
10
3

(i
1
k
1
i
2
D
i
3
D
i
5
)

T
k
2
+ 5D
(i
1

i
2
k
1
i
3
D
i
4
D
i
5
)

T
k
2
+ 3D
(i
1
D
i
2

i
3
k
1
i
4
D
i
5
)

T
k
2
+
2
3
D
(i
1
D
i
2
D
i
3

i
4
k
1
i
5
)

T
k
2

2
3
D
(i
1

i
2
k
1
i
3

i
4
i
5
)

T
k
2
+ D
(i
1

i
2
k
1
i
3
_

R

i
4
i
5
)

T
k
2
+ k
2
_
+
2
3
_
D
(i
1

i
2
i
3

T
k
2
+ k
2
_

R

i
4
i
5
)k
1
k
1
k
2
_

(i
1

i
2

i
3
i
4

i
5

i
6
)

T
k
1
k
2
= D
(i
1
D
i
6
)

T
k
1
k
2
+
_

(i
1

i
5

i
6
)k
1

T
k
2
+ 6
(i
1

i
4

i
5
k
1
D
i
6
)

T
k
2
+ 15
(i
1

i
3

i
4
k
1
D
i
5
D
i
6
)

T
k
2
+ 20
(i
1

i
2

i
3
k
1
D
i
4
D
i
6
)

T
k
2
+ 14
(i
1

i
2
k
1
D
i
3
D
i
6
)

T
k
2
+ 10
(i
1

i
3

i
4
k
1
_

i
5

i
6
)

T
k
2
+ k
2
_
+ 10
(i
1

i
2

i
3
k
1
_

i
4

i
5

i
6
)

T
k
2
+ k
2
_
+ 36
(i
1

i
2

i
3
k
1
_

i
4

i
5

D
i
6
)

T
k
2
+ k
2
_
+ 5
(i
1

i
2
k
1
_

i
3

i
5

i
6
)

T
k
2
+ k
2
_
+ 24
(i
1

i
2
k
1
_

i
3

i
4

i
5

D
i
6
)

T
k
2
+ k
2
_
+ 45
(i
1

i
2
k
1
_

i
3

i
4

D
i
5
D
i
6
)

T
k
2
+ k
2
_
+ 15
_

(i
1

i
2
k
1

i
3

i
4

i
5

i
6
)

T
k
2
+ k
2
_
+ k
2
_
+
(i
1

i
2
k
1
D
i
3
D
i
6
)

T
k
2
+ k
1
k
2
_
. (14)
4
If the second rank tensor with components

T
k
1
k
2
is replaced by the metric components g
k
1
k
2
then the related covariant derivatives (provided we deal with a torsion free ane connection)
vanish and the above expressions are simplied. One could derive for n = 5 the result:

(i
1

i
5
)
g
k
1
k
2
=
4
3
_
D
i
1
D
i
3

R
k
1
i
4
i
5
k
2
+ 2
_
D
i
1

R
k
1
i
2
i
3

i
4
i
5
k
2
+ k
1
k
2
__
. (15)
On the other hand for n = 6 one gets:

(i
1

i
2

i
3
i
4

i
5

i
6
)
g
k
1
k
2
=
10
7
D
(i
1
D
i
4

R
k
1
i
5
i
6
)k
2
+
34
7
_
D
(i
1
D
i
2

R
k
1
i
3
i
4

i
5
i
6
)k
2
+ k
1
k
2
_
+
55
7
D
(i
1

R
k
1
i
2
i
3

D
i
4

i
5
i
6
)k
2
+
16
7

R
k
1
(i
1
i
2

i
3
i
4
l

R
l
i
5
i
6
)k
2
. (16)
Thus, plugging into (10) expressions (15) and (16) we end up with the following expansion of
the metric tensor in RNC:
g
k
1
k
2
= g
k
1
k
2
+
1
2!
2
3

R
k
1
i
1
i
2
k
2

i
1

i
2
+
1
3!
D
i
1

R
k
1
i
2
i
3
k
2

i
1

i
3
+
1
4!
6
5
_
D
i
1
D
i
2

R
k
1
i
3
i
4
k
2
+
8
9

R
k
1
i
1
i
2
m

R
m
i
3
i
4
k
2
_

i
1

i
4
+
1
5!
4
3
_
D
i
1
D
i
3

R
k
1
i
4
i
5
k
2
+ 2
_
D
i
1

R
k
1
i
2
i
3

i
4
i
5
k
2
+ k
1
k
2
__

i
1

i
5
+
1
6!
10
7
_
D
i
1
D
i
4

R
k
1
i
5
i
6
k
2
+
17
5
_
D
i
1
D
i
2

R
k
1
i
3
i
4

i
5
i
6
k
2
+ k
1
k
2
_
+
11
2
D
i
1

R
k
1
i
2
i
3

D
i
4

i
5
i
6
k
2
+
8
5

R
k
1
i
1
i
2

i
3
i
4
l

R
l
i
5
i
6
k
2
_

i
1

i
6
+ O(
i
1

i
7
). (17)
The expansion (17) is in perfect agreement with that quoted in [3]. The inverse of the metric
tensor (obeying g
k
1
k
2
g
k
2
k
3
=
k
1
k
3
) can be found to be:
g
k
1
k
2
= g
k
1
k
2

1
2!
2
3

R
k
1
k
2
i
1
i
2

i
1

i
2

1
3!
D
i
1

R
k
1
k
2
i
2
i
3

i
1

i
3

1
4!
6
5
_
D
i
1
D
i
2

R
k
1
k
2
i
3
i
4

4
3

R
k
1
i
1
i
2
m

R
m k
2
i
3
i
4
_

i
1

i
4

1
5!
4
3
_
D
i
1
D
i
3

R
k
1
k
2
i
4
i
5
3
_
D
i
1

R
k
1
i
2
i
3

R
k
2
i
4
i
5
+ k
1
k
2
__

i
1

i
5

1
6!
10
7
_
D
i
1
D
i
4

R
k
1
k
2
i
5
i
6
5
_
D
i
1
D
i
2

R
k
1
i
3
i
4

R
k
2
i
5
i
6
+ k
1
k
2
_

17
2
D
i
1

R
k
1
i
2
i
3

D
i
4

R
k
2
i
5
i
6
+
16
3

R
k
1
i
1
i
2

i
3
i
4
l

R
l k
2
i
5
i
6
_

i
1

i
6
+ O(
i
1

i
7
). (18)
As a simple check one could consider the symmetric space V
n
in RNC for which D
k

R
lmnp
= 0
and prove that indeed the R terms in (17) satisfy:
g

= g

+
1
2

k=1
(1)
h
2
2k+2
(2k + 2)!
f

2
f

k1

, h =
_
k + 1 if k is even
k if k is odd
(19)
where f

=

R

1
i
1
i
2

i
1

i
2
and f

k1

=

R
ip
k1
i
p+1

ip

i
p+1
.
5
3 Conclusions
We have shown the method one could rely on to evaluate the expansion of the components of
the metric tensor in RNC at a specic order. The recursive structure permits an answer the
derivation of which becomes cumbersome when one attempts to calculate higher order terms.
References
[1] L. Alvarez-Gaume, D. Z. Freedman and S. Mukhi, Ann. of Phys. 134 (1981) 85.
[2] L. Eisenhart, Riemannian Geometry, Princeton Univ. Press, Princeton, N. J., 1965.;
V. I. Petrov, Einstein Spaces, Pergamon, Oxford, 1969.
[3] U. M uller, C. Schubert and Anton M. E. van de Ven, J. Gen. Rel. Grav. 31 (1759) 1999.
6