Geometry of the homogeneous and isotropic spaces

H. Sonoda September 2000; last revised October 2009

The purpose of this note is to summarize the geometry of maximally symmetric spaces, or homogeneous and isotropic spaces. We are interested in three dimensional spaces, but in the following we will study the spaces of an arbitrary dimension N = 2, 3, 4, ... . The three dimensional cases are obtained by taking N = 3.

Homogeneity and isotropy

We are interested in an N dimensional space for which the cosmological principle applies. The cosmological principle is the assumption that there is no special point or direction in our universe. We can formulate this more precisely as follows. If the distances between two arbitrary points are kept invariant under a map P f (P ), then the map is called an isometry. Given a particular point in space, isometries can be classied into two: those moving the point and those leaving the point xed. We call a space homogeneous if given arbitrary points P and Q, there exists an isometry mapping P to Q. We call a space isotropic 1) with respect to a point P if there exist N (N independent isometries xing 2 the point P . (Roughly speaking these are rotations around P .) We call a space maximally symmetric if it is homogeneous and isotropic. We note that due to the homogeneity a maximally symmetric space is isotropic with respect to any points. Actually a space which is isotropic with respect to any points is maximally symmetric. In the following we take it for granted that the homogeneous and isotropic spaces can be classied into three types: euclidean spaces, spheres, and hyperbolic spaces.1 Since we know everything about the geometry of euclidean spaces, we will only discuss spheres and hyperbolic spaces. In each case we rst consider the simplest two dimensions and then generalize the results to arbitrary N dimensions.
1 For

a proof, see Chapter 13 of S. Weinberg, Gravitation and Cosmology (Wiley).

Spheres

Let xi (i = 1, 2, 3) be three orthogonal coordinates of a three-dimensional euclidean space. A two-dimensional sphere of radius R is dened by the equation
2 2 2 x2 1 + x2 + x3 = R

(1)

In the euclidean space an SO(3) transformation is given by xi xi = Oij xj (2)

where the repeated indices j are summed over 1 to 3, and O is a three-by-three real orthogonal matrix of a unit determinant: O T O = 13 (3)

We note that the sphere is invariant under the SO(3) transformation group, i.e., under the transformations an arbitrary point on the sphere is mapped to a point on the same sphere. It is easy to see that the SO(3) transformations are 2 isometries of the sphere since dx2 i = dxi . Let us consider the innitesimal transformations explicitly. An arbitrary innitesimal transformation can be written as O = 13 + T where the three real 0 0 T1 0 0 0 1 antisymmetric matrices T are given by 0 1 0 0 0 1 0 1 , T2 0 0 0 , T3 1 0 0 0 0 0 1 0 0 0 (4)

(5)

Considering the actions of the isometries on the point P (0, 0, 1), we nd that T1,2 move P , and T3 leaves P invariant. Hence, T1,2 give homogeneity, and T3 gives isotropy to the three dimensional sphere.
x_3 T_3

r
T_2 x_2

x_1

T_1

Let us now consider the line element on the sphere. We rst consider the familiar spherical coordinates: x1 x2 x3 = R sin cos = R sin sin = R cos 2 (6)

where 0 < < and 0 < < 2 . The line element is given by (dl)2 =
3 i=1

( ) (dxi )2 = R2 (d)2 + sin2 (d)2

(7)

Around either the north pole N or the south pole S , we can alternatively use x1,2 as local coordinates. The third coordinate x3 is dependent on x1,2 as 2 x3 = R2 x2 (8) 1 x2 depending on whether the point is in the north or south hemisphere. The above implies x1 dx1 + x2 dx2 dx3 = (9) x3 Hence, the line element on the sphere is given by (dl)2 = (dx1 )2 + (dx2 )2 + (x1 dx1 + x2 dx2 ) 2 R2 (x2 1 + x2 )
2

(10) 2 x2 1 + x2 and the

Finally we express x1,2 in terms of the projected radius r angle as x1 x2 x3 = r cos = r sin = R2 r 2

(11)

Then, the line element is given by (dl)2 = (dr)2 + r2 (d)2 r2 1 R 2 (12)

This is the form most convenient to cosmology. (See the later section on the Robertson-Walker metric.) Let us make a nal note on the relation between the radial coordinate r and the proper radius l on the sphere. The proper radius l is the distance between the origin (north pole P in the gure) and a point at r. Since dl = we nd l=
0

dr 1
r2 R2

(13)

dr 1

= R arcsin
r2 R2

r R

(14)

Since the circumference at r is given by 2r, we nd

(15)

All of the above can be easily generalized to the N -dimensional sphere. It is dened in the N + 1 dimensional euclidean space as
2 2 x2 1 + ... + xN +1 = R

(16)

The isometries are given by the SO(N+1) transformations: xx =O x (17)

where O is an N -by-N real orthogonal matrix of a unit determinant. An innitesimal transformation can be written as ij O = 1N +1 + (18) ij T
i<j N +1

where T ij is a real antisymmetric matrix whose only non-vanishing elements are ( ij ) ( ) (19) T ij = T ij ji = 1 Let us consider a point P = (0, ..., 0, 1) on the sphere. T ij (i < j N ) give N (N 1) independent rotations around P , and T i,N +1 (i N ) move P in N 2 independent directions. In the spherical coordinates we obtain x1 x2 x3 xN xN +1 = R sin sin 1 sin 2 ... sin N 2 sin N 1 = R sin sin 1 sin 2 ... sin N 2 cos N 1 = R sin sin 1 sin 2 ... cos N 2 ..... = R sin cos 1 = R cos (20)

where 0 < < , 0 < 1,...,N 2 < , 0 < N 1 < 2 . The line element on the sphere is given recursively as [ ] (dlN )2 = R2 (d)2 + sin2 (dlN 1 )2 (21) where dlN 1 is the line element on the unit N 1 sphere. If we use r N 2 i=1 (xi ) , the distance from the N + 1-th axis, instead of the angle , we nd x1 x2 x3 xN xN +1 = r sin 1 sin 2 ... sin N 2 sin N 1 = r sin 1 sin 2 ... sin N 2 cos N 1 = = = r sin 1 sin 2 ... cos N 2 ......... r cos 1 R2 r 2 4 (22)

and the line element is given by (dlN )2 = (dr)2 2 + r2 dlN 1 r2 1 R 2 (23)

Hyperbolic spaces

To construct a hyperbolic surface, we start from a three-dimensional Minkowski space as opposed to a euclidean space. Let x1,2,3 be the coordinates of the Minkowski space with the metric (dl)2 = ij dxi dxj (dx1 )2 + (dx2 )2 (dx3 )2 (24)

Hence, x3 plays the role of time. The space-time distance is invariant under the Lorentz transformations SO(2, 1): xx =Lx where L is a three-by-three real matrix satisfying LT L = , det L = 1, L33 > 0 (26) (25)

We dene a hyperbolic surface of radius R by

(27)

This denes two separate space-like surfaces, and we take the one with point P = (0, 0, R) on.
x_3 r

P R
x_2 x_1

Since the surface is invariant under the SO(2, 1) transformations (i.e., Lorentz transformations in 2 + 1 dimensional space-time), the isometries are given by SO(2, 1). An innitesimal SO(2, 1) transformation is given in the form L = 13 + T where 0 1, 0 1 0, 0 0 1 0 T3 = 1 0 0 0 0 0 (28)

0 0 T1 = 0 0 0 1

0 0 T2 = 0 0 1 0 5

(29)

With respect to P , T1,2 correspond to the homogeneity, and T3 to the isotropy of the hyperbolic surface. Using x1,2 as local coordinates, the line element is given by (dl)2 = (dx1 )2 + (dx2 )2 (dx3 )2 = (dx1 )2 + (dx2 )2 With r (x1 dx1 + x2 dx2 ) 2 R2 + x2 1 + x2
2

(30)

2 x2 1 + x2 as the distance from the x3 axis, we nd x1 x2 x3 = r cos = r sin = R2 + r 2 (31)

The line element is given by (dl)2 = (dr)2 + r2 (d)2 r2 1+ R 2 (32)

A nal note on the on the relation between the radial coordinate r and the proper radius l on the hyperbolic surface. We obtain ( ) r r2 r dr 1 r = R sinh (33) l= = R ln + 1+ 2 2 R R R 0 1+ r
R2

Hence, we nd
r (circumference) r R = = 2 (proper radius) l sinh1 r R

>1

(34)

Now, let us summarize the properties of N dimensional hyperbolic spaces. We dene an N dimensional hyperbolic space as a space-like subspace in the N + 1 dimensional Minkowski space. Let the metric of the Minkowski space be (dl)2 = ij dxi dxj (dx1 )2 + ... + (dxN )2 (dxN +1 )2 We dene a hyperbolic space of radius R by
2 2 2 x2 1 + ... + xN xN +1 = R

(35)

(36)

The isometries of the hyperbolic space is given by the SO(N, 1) transformations xx =Lx where L is an N + 1-by-N + 1 real matrix satisfying LT L = , detL = 1, LN +1,N +1 > 0 (38) (37)

The innitesimal SO(N, 1) transformations are given in the form ij L = 1N +1 + ij T

(39)

where T ij (i < j N ) are the real antisymmetric matrices with the only nonvanishing elements (T ij )ij = (T ij )ji = 1, and T i,N +1 (i N ) are the real symmetric matrices with (T i,N +1 )i,N +1 = (T i,N +1 )N +1,1 = 1. With respect to the special point P = (0, ..., 0, 1) on the hyperbolic space, T ij correspond to the + ... + x2 isotropy, and T i,N +1 to the homogeneity of the space. With r x2 1 N as the distance from the xN +1 axis, we nd x1 x2 xN xN +1 The line element is given by (dl)2 = (dr)2 + r2 (dlN 1 )2 r2 1+ R 2 (41) = r sin 1 ... sin N 1 = r sin 1 ... cos N 1 ..... = r cos 1 = R2 + r 2

(40)

where dlN 1 is the line element on the unit N 1 sphere.

Robertson-Walker metric

If we adopt the cosmological principle that our universe is both isotropic and homogeneous, the space must be one of the three types discussed above: euclidean space, sphere, and hyperbolic space. Hence, the space-time metric of the universe must be given by (ds)2 = c2 (dt)2 (dl)2 (42)

where c is the light velocity, and dl is the line element in space. In the previous sections, we have found ( ) (dl)2 = (dr)2 + r2 (d)2 + sin2 (d)2 (43) if the space is at (or euclidean), (dl)2 = ( ) (dr)2 + r2 (d)2 + sin2 (d)2 r2 1 a(t)2 (44)

if the space is a sphere of radius a(t), and (dl)2 = ( ) (dr)2 + r2 (d)2 + sin2 (d)2 r2 1 + a(t)2 7 (45)

if the space is a hyperbolic space of radius a(t). By rescaling r by a(t), we can incorporate all the above as [ ] ( ) (dr)2 2 2 2 2 (46) (dl)2 = a(t)2 + r ( d ) + sin ( d ) 1 Kr2 where K = 0, 1, 1 for euclidean space, sphere, and hyperbolic space.

De Sitter space

In the above we have considered the maximally symmetric spaces. In this nal section we generalize the notion of maximal symmetry from spaces to spacetimes. We wish to construct 3 + 1-dimensional maximally symmetric spacetimes which have 4 translational isometries and 6 rotational isometries. We can construct such a space, known as the de Sitter space, as follows. We start from a ve dimensional Minkowski space with the metric (ds)2 = dx dx (dx1 )2 + (dx2 )2 + (dx3 )2 + (dx4 )2 (dx0 )2 We consider the four-dimensional subspace M dened by2
2 2 2 2 2 x2 1 + x2 + x3 + x4 x0 = R

(47)

(48)

The isometries of the subspace M is given by the SO(4, 1) transformations xx =Lx (49)

where L satises LT L = , det L = 1, and L00 > 0. Taking x = x1,2,3 and x0 as independent local coordinates, we nd (ds)2 = (dx)2 (dx0 )2 + (x dx x0 dx0 ) 2 R2 + x2 0x
2

(50)

The four coordinates x, x0 are not entirely free, but they are constrained by the condition 2 R 2 + x2 (51) 0x 0 We can rewrite the above line element using dierent choices of space-time coordinates. We give two examples here. The rst example is given by [ 2 ( ] 2 ) x0 x x0 x x0 = R cosh + 1+ sinh 2R2 R 2R2 R ( ) x0 x = x exp (52) R
2 Note the dierence in the sign of the right-hand side compared to the denition of the hyperbolic spaces.

The coordinates x and x0 are entirely free. The line element becomes ) ( 2x0 (dx )2 (ds)2 = (dx0 )2 exp R

(53)

Note that the space metric appears at so that this becomes a Robertson-Walker 2t metric with a particular scale parameter: a(t) = e R . The second example is given by [ ( )] 2 R x 2x0 x0 = x0 ln 1 2 exp 2 R R ( ) x0 x = x exp =x (54) R The coordinates x (and hence x) are constrained by x element becomes time independent as ( (ds)2 = 1 x R2
2) 2

< R2 . Then the line )2

( (dx0 )2 (dx )2

x dx

R2 x

(55)

The metric in this form was rst applied by de Sitter to a theory of the steady state universe.