# Lorentz group

In physics, the Lorentz group is the group of all Lorentz transformations of Minkowski spacetime, the classical setting for all (nongravitational) physical phenomena. The mathematical form of

are each invariant under Lorentz transformations. Therefore the Lorentz group can be said to express a fundamental symmetry of many of the known fundamental laws of nature.

## Basic properties

The Lorentz group is a subgroup of the Poincaré group, the group of all isometries of Minkowski spacetime. The Lorentz transformations are precisely the isometries which leave the origin fixed. Thus, the Lorentz group is an isotropy subgroup of the isometry group of Minkowski spacetime. For this reason, the Lorentz group is sometimes called the homogeneous Lorentz group while the Poincaré group is sometimes called the inhomogeneous Lorentz group. Lorentz transformations are examples of linear transformations; general isometries of Minkowski spacetime are affine transformations.

Mathematically, the Lorentz group may be described as the generalized orthogonal group O(1,3), the matrix Lie group which preserves a certain quadratic form defined on R4. This quadratic form is interpreted in physics as the metric tensor of Minkowski spacetime, so this definition is simply a restatement of the fact that Lorentz transformations are precisely the linear transformations which are also isometries of Minkowski spacetime.

The Lorentz group is a 6-dimensional noncompact Lie group which is not connected, and whose connected components are not simply connected. The identity component (i.e. the component containing the identity element) of the Lorentz group is often called the restricted Lorentz group and is denoted SO+(1,3).

In pure mathematics, the restricted Lorentz group arises in another guise as the Möbius group, which is the symmetry group of conformal geometry on the Riemann sphere. This observation was taken by Roger Penrose as the starting point of twistor theory. It has a fascinating physical consequence for the appearance of the night sky as seen by an observer who is maneuvering at relativistic velocities relative to the "fixed stars", which is discussed below.

The restricted Lorentz group arises in other ways in pure mathematics. For example, it arises as the point symmetry group of a certain ordinary differential equation. This fact also has physical significance.

## Connected components

Because it is a Lie group, the Lorentz group O(1,3) is both a group and a smooth manifold. As a manifold, it has four connected components. Intuitively, this means that it consists of four topologically separated pieces.

To see why, notice that a Lorentz transformation may or may not

Lorentz transformations which preserve the direction of time are called orthochronous. Those which preserve orientation are called proper, and as linear transformations they have determinant +1. (The improper Lorentz transformations have determinant −1.) The subgroup of proper Lorentz transformations is denoted SO(1,3). The subgroup of orthochronous transformations is often denoted O+(1,3).

The identity component of the Lorentz group is the set of all Lorentz transformations preserving both orientation and the direction of time. It is called the proper, orthochronous Lorentz group, or restricted Lorentz group, and it is denoted by SO+(1, 3). It is a normal subgroup of the Lorentz group which is also six dimensional.

Warning: some authors carelessly refer to SO(1,3) or even O(1,3) when they actually mean SO+(1, 3).

The quotient group O(1,3)/SO+(1,3) is isomorphic to the Klein four-group. Every element in O(1,3) can be written as the product of a proper, orthochronous transformation and an element of the discrete group

{1, P, T, PT}

where P and T are the space inversion and time reversal operators:

P = diag(1, −1, −1, −1)
T = diag(−1, 1, 1, 1)

The four elements of this isomorphic copy of the Klein four-group label the four connected components of the Lorentz group.

## The restricted Lorentz group

As stated above, the restricted Lorentz group is the identity component of the Lorentz group. This means that it consists of all Lorentz transformations which can be connected to the identity by a continuous curve lying in the group. The restricted Lorentz group is a connected normal subgroup of the full Lorentz group with the same dimension (in this case, 6 dimensions).

The restricted Lorentz group is generated by ordinary spatial rotations and Lorentz boosts (which can be thought of as hyperbolic rotations in a plane that includes a time-like direction). The set of all rotations forms a Lie subgroup isomorphic to the ordinary rotation group SO(3). The set of all boosts, however, does not form a subgroup, since composing two boosts does not, in general, result in another boost.

A boost in some direction, or a rotation about some axis, each generate a one-parameter subgroup. An arbitrary rotation is specified by 3 real parameters, as is an arbitrary boost. Since every proper, orthochronous Lorentz transformation can be written as a product of a rotation and a boost, it takes 6 real numbers (parameters) to specify an arbitrary proper orthochronous Lorentz transformation. This is one way to understand why the restricted Lorentz group is six dimensional. (We'll study this in more detail in a later section on the Lie algebra of the Lorentz group.) To specify an arbitrary Lorentz transformation requires a further two bits of information, which pick out one of the four connected components. This pattern is typical of finite dimensional Lie groups.

## Relation to the Möbius group

The restricted Lorentz group SO+(1, 3) is isomorphic to the Möbius group, which is, in turn, isomorphic to the projective special linear group PSL(2,C). It will be convenient to work at first with SL(2,C). This group consists of all two by two complex matrices with determinant one

$\displaystyle P = \left[ \begin{matrix} a & b \\ c & d \end{matrix} \right], \; ad - bc = 1$

We can write two by two Hermitian matrices in the form

$\displaystyle X = \left[ \begin{matrix} t+z & x-iy \\ x+iy & t-z \end{matrix} \right]$

This trick has the pleasant feature that

$\displaystyle \det \, X = t^2 - x^2 - y^2 - z^2$

Therefore, we have identified the space of Hermitian matrices (which is four dimensional, as real vector space) with Minkowski spacetime in such a way that the determinant of a Hermitian matrix is the squared length of the corresponding vector in Minkowski spacetime. But now SL(2,C) acts on the space of Hermitian matrices via

$\displaystyle X \rightarrow P X P^*$

where $\displaystyle P^*$ is the Hermitian transpose of $\displaystyle P$ , and this action preserves the determinant. Therefore, SL(2,C) acts on Minkowski spacetime by (linear) isometries. We have now constructed a homomorphism of Lie groups from SL(2,C) onto SO+(1,3), which we will call the spinor map. The kernel of the spinor map is the two element subgroup ±I. Therefore, the quotient group PSL(2,C) is isomorphic to SO+(1,3).

## Appearance of the night sky

This isomorphism has a very interesting physical interpretation. We can identify the complex number

$\displaystyle \xi = u+iv$

with a null vector in Minkowski space

$\displaystyle \left[ \begin{matrix} u^2+v^2+1 \\ 2u \\ -2v \\ u^2+v^2-1 \end{matrix} \right]$

or the Hermitian matrix

$\displaystyle N = 2\left[ \begin{matrix} u^2+v^2 & u+iv \\ u-iv & 1 \end{matrix} \right]$

The set of real scalar multiples of this null vector, which we can call a null line through the origin, represents "lines of sight" from an observer at a particular place and time (an arbitrary event which we can identify with the origin of Minkowski spacetime) to various distant objects, such as stars.

But by stereographic projection, we can identify $\displaystyle \xi$ with a point on the Riemann sphere. Putting it all together, we have identified the points of the celestial sphere with certain Hermitian matrices, and also with lines of sight. This implies that the Möbius transformations of the Riemann sphere precisely represent the way that Lorentz transformations change the appearance of the celestial sphere.

For our purposes here, we can pretend that the "fixed stars" live in Minkowski spacetime. Then, the Earth is moving at a nonrelativistic velocity with respect to a typical astronomical object which might be visible at night. But, an observer who is moving at relativistic velocity with respect to the Earth would see the appearance of the night sky (as modeled by points on the celestial sphere) transformed by a Möbius transformation.

## Conjugacy classes

Because the restricted Lorentz group SO+(1, 3) is isomorphic to the Möbius group PSL(2,C), its conjugacy classes also fall into four classes:

• elliptic transformations
• hyperbolic transformations
• loxodromic transformations
• parabolic transformations

(To be utterly pedantic, the identity element is in a fifth class, all by itself.)

In the article on Möbius transformations, it is explained how this classification arises by considering the fixed points of Möbius transformations in their action on the Riemann sphere, which corresponds here to null eigenspaces of restricted Lorentz transformations in their action on Minkowski spacetime.

We will discuss a particularly simple example of each type, and in particular, the effect on the appearance of the night sky of the one-parameter subgroup which it generates. At the end of the section we will briefly indicate how we can understand the effect of general Lorentz transformations on the appearance of the night sky in terms of these examples.

A typical elliptic element of SL(2,C) is

$\displaystyle P_1 = \left[ \begin{matrix} \exp(i \theta/2) & 0 \\ 0 & \exp(-i \theta/2) \end{matrix} \right]$

which has fixed points $\displaystyle \xi = 0, \infty$ . Writing out the action $\displaystyle X \rightarrow P_1 X {P_1}^*$ and collecting terms, we find that our spinor map takes this to the (restricted) Lorentz transformation

$\displaystyle Q_1 = \left[ \begin{matrix} 1 & 0 & 0 & 0 \\ 0 & \cos(\theta) & -\sin(\theta) & 0 \\ 0 & \sin(\theta) & \cos(\theta) & 0 \\ 0 & 0 & 0 & 1 \end{matrix} \right]$

This transformation represents a rotation about the z axis. The one-parameter subgroup it generates is obtained by simply taking $\displaystyle \theta$ to be a real variable instead of a constant. The corresponding continuous transformations of the celestial sphere (except for the identity) all share the same two fixed points, the North and South pole. They move all other points around latitude circles. In other words, this group yields a continuous counterclockwise rotation about the z axis as $\displaystyle \theta$ increases.

Notice the angle doubling; this phenomenon is a characteristic feature of spinorial double coverings.

A typical hyperbolic element of SL(2,C) is

$\displaystyle P_2 = \left[ \begin{matrix} \exp(\beta/2) & 0 \\ 0 & \exp(-\beta/2) \end{matrix} \right]$

which also has fixed points $\displaystyle \xi = 0, \infty$ . Under stereographic projection from the Riemann sphere to the Euclidean plane, the effect of this Möbius transformation is a dilation from the origin. Our homomorphism maps this to the Lorentz transformation

$\displaystyle Q_2 = \left[ \begin{matrix} \cosh(\beta) & 0 & 0 & \sinh(\beta) \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ \sinh(\beta) & 0 & 0 & \cosh(\beta) \end{matrix} \right]$

This transformation represents a boost along the z axis. The one-parameter subgroup it generates is obtained by simply taking $\displaystyle \beta$ to be a real variable instead of a constant. The corresponding continuous transformations of the celestial sphere (except for the identity) all share the same fixed points (the North and South poles), and they move all other points along longitudes away from the South pole and toward the North pole.

A typical loxodromic element of SL(2,C) is

$\displaystyle P_3 = P_2 P_1 = P_1 P_2 = \left[ \begin{matrix} \exp \left((\beta+i\theta)/2 \right) & 0 \\ 0 & \exp \left(-(\beta+i\theta)/2 \right) \end{matrix} \right]$

which also has fixed points $\displaystyle \xi = 0, \infty$ . Our homomorphism maps this to the Lorentz transformation

$\displaystyle Q_3 = Q_2 Q_1 = Q_1 Q_2$

The one-parameter subgroup this generates is obtained by replacing $\displaystyle \beta + i \theta$ with any real multiple of this complex constant. (If we let $\displaystyle \beta,\theta$ vary independently, we obtain a two-dimensional abelian subgroup, consisting of simultaneous rotations about the z axis and boosts along the z axis; in contrast, the one-dimensional subgroup we are discussing here consists of those elements of this two-dimensional subgroup such that the rapidity of the boost and angle of the rotation have a fixed ratio.) The corresponding continuous transformations of the celestial sphere (always excepting the identity) all share the same two fixed points (the North and South poles). They move all other points away from the South pole and toward the North pole (or vice versa), along a family of curves called loxodromes. Each loxodrome spirals infinitely often around each pole.

A typical parabolic element of SL(2,C) is

$\displaystyle P_4 = \left[ \begin{matrix} 1 & \alpha \\ 0 & 1 \end{matrix} \right]$

which has the single fixed point $\displaystyle \xi=\infty$ on the Riemann sphere. Under stereographic projection, it appears as ordinary translation along the real axis. Our homomorphism maps this to the Lorentz transformation

$\displaystyle Q_4 = \left[ \begin{matrix} 1+\alpha^2/2 & \alpha & 0 & -\alpha^2/2 \\ \alpha & 1 & 0 & -\alpha \\ 0 & 0 & 1 & 0 \\ -\alpha^2/2 & 0 & 0 & 1-\alpha^2/2 \end{matrix} \right]$

This generates a one-parameter subgroup which is obtained by considering $\displaystyle \alpha$ to be a real variable rather than a constant. The corresponding continuous transformations of the celestial sphere move points along a family of circles which are all tangent at the North pole to a certain great circle. All points other than the North pole itself move along these circles. (Except, of course, for the identity transformation.)

Note: parabolic Lorentz transformations are often called null rotations.

The Möbius transformations are precisely the conformal transformations of the Riemann sphere (or celestial sphere). It follows that by conjugating with an arbitrary element of SL(2,C), we can obtain from the above examples arbitrary elliptic, hyperbolic, loxodromic, and parabolic (restricted) Lorentz transformations, respectively. The effect on the flow lines of the corresponding one-parameter subgroups is to transform the pattern seen in our examples by some conformal transformation. Thus, an arbitrary elliptic Lorentz transformation can have any two distinct fixed points on the celestial sphere, but points will still flow along circular arcs from one fixed point toward the other. Similarly for the other cases.

Finally, arbitrary Lorentz transformations can be obtained from the restricted ones by multiplying by a matrix which reflects across $\displaystyle t=0$ , or by an appropriate orientation reversing diagonal matrix.

## The Lie algebra of the Lorentz group

As with any Lie group, the best way to study many aspects of the Lorentz group is via its Lie algebra. We can immediately write down an obvious set of six generators:

• vector fields on R4 generating three rotations
$\displaystyle -y \partial_x + x \partial_y, \; -z \partial_y + y \partial_z, \; -x \partial_z + z \partial_x$
• vector fields on R4 generating three boosts
$\displaystyle x \partial_t + t \partial_x, \; y \partial_t + t \partial_y, \; z \partial_t + t \partial_z$

It may be helpful to briefly recall here how to obtain a one-parameter group from a vector field, written in the form of a first order linear partial differential operator such as

$\displaystyle -y \partial_x + x \partial_y$

The corresponding initial value problem is

$\displaystyle \frac{\partial x}{\partial \lambda} = -y, \; \frac{\partial y}{\partial \lambda} = x, \; x(0) = x_0, \; y(0) = y_0$

The solution can be written

$\displaystyle x(\lambda) = x_0 \cos(\lambda) - y_0 \sin(\lambda), \; y(\lambda) = x_0 \sin(\lambda) + y_0 \cos(\lambda)$

or

$\displaystyle \left[ \begin{matrix} t \\ x \\ y \\ z \end{matrix} \right] = \left[ \begin{matrix} 1 & 0 & 0 & 0 \\ 0 & \cos(\lambda) & -\sin(\lambda) & 0 \\ 0 & \sin(\lambda) & \cos(\lambda) & 0 \\ 0 & 0 & 0 & 1 \end{matrix} \right] \left[ \begin{matrix} t_0 \\ x_0 \\ y_0 \\ z_0 \end{matrix} \right]$

where we easily recognize the one-parameter matrix group of rotations about the z axis. Differentiating with respect to the group parameter and setting $\displaystyle \lambda=0$ in the result, we recover the matrix

$\displaystyle \left[ \begin{matrix} 0 & 0 & 0 & 0 \\ 0 & 0 & -1 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 0 & 0 \end{matrix} \right]$

which corresponds to the vector field we started with. This shows how to pass between matrix and vector field representations of elements of the Lie algebra.

Reversing the procedure in the previous section, we see that the Möbius transformations which correspond to our six generators arise from exponentiating respectively $\displaystyle \frac{\theta}{2}$ (for the three boosts) or $\displaystyle \frac{i \theta}{2}$ (for the three rotations) times the three Pauli matrices

$\displaystyle \sigma_1 = \left[ \begin{matrix} 0 & 1 \\ 1 & 0 \end{matrix} \right], \; \; \sigma_2 = \left[ \begin{matrix} 0 & i \\ -i & 0 \end{matrix} \right], \; \; \sigma_3 = \left[ \begin{matrix} 1 & 0 \\ 0 & -1 \end{matrix} \right]$

For our purposes, another generating set is more convenient. We list the six generators in the following table, in which

• the first column gives a generator of the flow under the Möbius action (after stereographic projection from the Riemann sphere) as a real vector field on the Euclidean plane,
• the second column gives the corresponding one-parameter subgroup of Möbius transformations,
• the third column gives the corresponding one-parameter subgroup of Lorentz transformations (the image under our homomorphism of preceding one-parameter subgroup),
• the fourth column gives the corresponding generator of the flow under the Lorentz action as a real vector field on Minkowski spacetime.

Notice that the generators consist of

• two parabolics (null rotations),
• one hyperbolic (boost in $\displaystyle \partial_z$ direction),
• three elliptics (rotations about x,y,z axes respectively).

Vector field on R2 One-parameter subgroup of SL(2,C),
representing Möbius transformations
One-parameter subgroup of SO+(1,3),
representing Lorentz transformations
Vector field on R4
Parabolic
$\displaystyle \partial_u\,\!$ $\displaystyle \left[ \begin{matrix} 1 & \alpha \\ 0 & 1 \end{matrix} \right]$ $\displaystyle \left[ \begin{matrix} 1+\alpha^2 & \alpha & 0 & -\alpha^2/2 \\ \alpha & 1 & 0 & -\alpha \\ 0 & 0 & 1 & 0 \\ -\alpha^2/2 & \alpha & 0 & 1-\alpha^2/2 \end{matrix} \right]$ $\displaystyle X_1 = \,\!$
$\displaystyle x (\partial_t + \partial_z) + (t-z) \partial_x \,\!$
$\displaystyle \partial_v\,\!$ $\displaystyle \left[ \begin{matrix} 1 & i \alpha \\ 0 & 1 \end{matrix} \right]$ $\displaystyle \left[ \begin{matrix} 1+\alpha^2 & 0 & \alpha & -\alpha^2/2 \\ \alpha & 1 & 0 & -\alpha \\ 0 & 0 & 1 & 0 \\ -\alpha^2/2 & 0 & \alpha & 1-\alpha^2/2 \end{matrix} \right]$ $\displaystyle X_2 = \,\!$
$\displaystyle y (\partial_t + \partial_z) + (t-z) \partial_y \,\!$
Hyperbolic
$\displaystyle \frac{1}{2} \left( u \partial_u + v \partial_v \right)$ $\displaystyle \left[ \begin{matrix} \exp \left(\frac{\beta}{2}\right) & 0 \\ 0 & \exp \left(-\frac{\beta}{2}\right) \end{matrix} \right]$ $\displaystyle \left[ \begin{matrix} \cosh(\beta) & 0 & 0 & \sinh(\beta) \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ \sinh(\beta) & 0 & 0 & \cosh(\beta) \end{matrix} \right]$ $\displaystyle X_3 = \,\!$
$\displaystyle z \partial_t + t \partial_z \,\!$
Elliptic
$\displaystyle \frac{1}{2} \left( -v \partial_u + u \partial_v \right)$ $\displaystyle \left[ \begin{matrix} \exp \left( \frac{i \theta}{2} \right) & 0 \\ 0 & \exp \left( \frac{-i \theta}{2} \right) \end{matrix} \right]$ $\displaystyle \left[ \begin{matrix} 1 & 0 & 0 & 0 \\ 0 & \cos(\theta) & -\sin(\theta) & 0 \\ 0 & \sin(\theta) & \cos(\theta) & 0 \\ 0 & 0 & 0 & 1 \end{matrix} \right]$ $\displaystyle X_4 = \,\!$
$\displaystyle -y \partial_x + x \partial_y \,\!$
$\displaystyle \frac{v^2-u^2-1}{2} \partial_u - u v \, \partial_v$ $\displaystyle \left[ \begin{matrix} \cos \left( \frac{\theta}{2} \right) & -\sin \left( \frac{\theta}{2} \right) \\ \sin \left( \frac{\theta}{2} \right) & \cos \left( \frac{\theta}{2} \right) \end{matrix} \right]$ $\displaystyle \left[ \begin{matrix} 1 & 0 & 0 & 0 \\ 0 & \cos(\theta) & 0 & \sin(\theta) \\ 0 & 0 & 1 & 0 \\ 0 & -\sin(\theta) & 0 & \cos(\theta) \end{matrix} \right]$ $\displaystyle X_5 = \,\!$
$\displaystyle -z \partial_x + x \partial_z \,\!$
$\displaystyle u v \, \partial_u + \frac{1-u^2+v^2}{2} \partial_v$ $\displaystyle \left[ \begin{matrix} \cos \left( \frac{\theta}{2} \right) & i \sin \left( \frac{\theta}{2} \right) \\ i \sin \left( \frac{\theta}{2} \right) & \cos \left( \frac{\theta}{2} \right) \end{matrix} \right]$ $\displaystyle \left[ \begin{matrix} 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & \cos(\theta) & -\sin(\theta) \\ 0 & 0 & \sin(\theta) & \cos(\theta) \end{matrix} \right]$ $\displaystyle X_6 = \,\!$
$\displaystyle z \partial_y - y \partial_z \,\!$

$\displaystyle \sigma_2 = \left[ \begin{matrix} 0 & i \\ -i & 0 \end{matrix} \right]$

Exponentiate:

$\displaystyle \exp \left( \frac{ i \theta}{2} \, \sigma_2 \right) = \left[ \begin{matrix} \cos(\theta/2) & -\sin(\theta/2) \\ \sin(\theta/2) & \cos(\theta/2) \end{matrix} \right]$

This element of SL(2,C) represents the one-parameter subgroup of (elliptic) Möbius transformations:

$\displaystyle \xi \rightarrow \frac{ \cos(\theta/2) \, \xi - \sin(\theta/2) }{ \sin(\theta/2) \, \xi + \cos(\theta/2) }$

Next,

$\displaystyle \frac{d\xi}{d\theta} |_{\theta=0} = -\frac{1+\xi^2}{2}$

The corresponding vector field on C (thought of as the image of S2 under stereographic projection) is

$\displaystyle -\frac{1+\xi^2}{2} \, \partial_\xi$

Writing $\displaystyle \xi = u + i v$ , this becomes the vector field on R2

$\displaystyle -\frac{1+u^2-v^2}{2} \, \partial_u - u v \, \partial_v$

Returning to our element of SL(2,C), writing out the action $\displaystyle X \rightarrow P X P^*$ and collecting terms, we find that the image under the spinor map is the element of SO+(1,3)

$\displaystyle \left[ \begin{matrix} 1 & 0 & 0 & 0 \\ 0 & \cos(\theta) & 0 & \sin(\theta) \\ 0 & 0 & 1 & 0 \\ 0 & -\sin(\theta) & 0 & \cos(\theta) \end{matrix} \right]$

Differentiating with respect the $\displaystyle \theta$ at $\displaystyle \theta=0$ , we find that the corresponding vector field on R4 is

$\displaystyle -z \partial_x + x \partial_z \,\!$

This is evidently the generator of counterclockwise rotation about the $\displaystyle y$ axis.

## Subgroups of the Lorentz group

The subalgebras of the Lie algebra of the Lorentz group can be enumerated, up to conjugacy, from which we can list the closed subgroups of the restricted Lorentz group, up to conjugacy. (See the book by Hall cited below for the details.) We can readily express the result in terms of the generating set given in the table above.

The one-dimensional subalgebras of course correspond to the four conjugacy classes of elements of the Lorentz group:

• $\displaystyle X_1$ generates a one-parameter subalgebra of parabolics SO(0,1),
• $\displaystyle X_3$ generates a one-parameter subalgebra of boosts SO(1,1),
• $\displaystyle X_4$ generates a one-parameter of rotations SO(2),
• $\displaystyle X_3 + a X_4$ (for any $\displaystyle a \neq 0$ ) generates a one-parameter subalgebra of loxodromic transformations.

(Strictly speaking the last corresponds to infinitely many classes, since distinct $\displaystyle a$ give different classes.) The two-dimensional subalgebras are:

• $\displaystyle X_1, X_2$ generate an abelian subalgebra consisting entirely of parabolics,
• $\displaystyle X_1, X_3$ generate a nonabelian subalgebra isomorphic to the Lie algebra of the affine group A(1),
• $\displaystyle X_3, X_4$ generate an abelian subalgebra consisting of boosts, rotations, and loxodromics all sharing the same pair of fixed points.

The three dimensional subalgebras are:

• $\displaystyle X_1,X_2,X_3$ generate a Bianchi V subalgebra, isomorphic to the Lie algebra of Hom(2), the group of euclidean homotheties,
• $\displaystyle X_1,X_2,X_4$ generate a Bianchi VII_0 subalgebra, isomorphic to the Lie algebra of E(2), the euclidean group,
• $\displaystyle X_2,X_2,X_3 + a X_4$ , where $\displaystyle a \neq 0$ , generate a Bianchi VII_a subalgebra,
• $\displaystyle X_1,X_3,X_5$ generate a Bianchi VIII subalgebra, isomorphic to the Lie algebra of SL(2,R), the group of isometries of the hyperbolic plane,
• $\displaystyle X_4,X_5,X_6$ generate a Bianchi IX subalgebra, isomorphic to the Lie algebra of SO(3), the rotation group.

(Here, the Bianchi types refer to the classification of three dimensional Lie algebras by the Italian mathematician Luigi Bianchi.) The four dimensional subalgebras are all conjugate to

• $\displaystyle X_1,X_2,X_3,X_4$ generate a subalgebra isomorphic to the Lie algebra of Sim(2), the group of Euclidean similitudes.

These subalgebras each generate by exponentiation a closed subgroup of the restricted Lie group. From these, all subgroups of the Lie group can be constructed, up to conjugation, by multiplying by one of the elements of the Klein four-group.

As with any connected Lie group, the coset spaces of the closed subgroups of the restricted Lorentz group, or homogeneous spaces, have considerable mathematical interest. We briefly describe some of them here:

• the group Sim(2) is the stabilizer of a null line, i.e. of a point on the Riemann sphere, so SO+(1,3)/Sim(2) is the Kleinian geometry which represents conformal geometry on the sphere,
• the (identity component of the) Euclidean group SE(2) is the stabilizer of a null vector, so SO+(1,3)/SE(2) is the momentum space of a massless particle; geometrically, this Kleinian geometry represents the degenerate geometry of the light cone in Minkowski spacetime,
• the rotation group SO(3) is the stabilizer of a timelike vector, so SO+(1,3)/SO(3) is the momentum space of a massive particle; geometrically, this space is none other than three-dimensional hyperbolic space H3.

## Covering groups

In a previous section we constructed a homomorphism SL(2,C) $\displaystyle \rightarrow$ SO+(1,3), which we called the spinor map. Since SL(2,C) is simply connected, it is the covering group of the restricted Lorentz group SO+(1,3). By restriction we obtain a homomorphism SU(2) $\displaystyle \rightarrow$ SO(3). Here, the special unitary group SU(2), which is isomorphic to the group of unit norm quaternions, is also simply connected, so it is the covering group of the rotation group SO(3). Each of these covering maps are two-fold covers in the sense that precisely two elements of the covering group map to each element of the quotient. One often says that the restricted Lorentz group and the rotation group are doubly connected. This means that the fundamental group of the each group is isomorphic to the two element cyclic group Z2.

Warning: in applications to quantum mechanics the special linear group SL(2, C) is sometimes called the Lorentz group.

Two-fold coverings are characteristic of spin groups. Indeed, in addition to the double coverings

Spin+(1,3)=SL(2,C) $\displaystyle \rightarrow$ SO+(1,3)
Spin(3)=SU(2) $\displaystyle \rightarrow$ SO(3)

we have the double coverings

Pin(1,3) $\displaystyle \rightarrow$ O(1,3)
Spin(1,3) $\displaystyle \rightarrow$ SO(1,3)
Spin+(1,2) = SU(1,1) $\displaystyle \rightarrow$ SO(1,2)

These spinorial double coverings are all closely related to Clifford algebras.

## Topology

The left and right groups in the double covering

SU(2) $\displaystyle \rightarrow$ SO(3)

are deformation retracts of the left and right groups, respectively, in the double covering

SL(2,C) $\displaystyle \rightarrow$ SO+(1,3)

But the homogeneous space SO+(1,3)/SO(3) is homeomorphic to hyperbolic 3-space H3, so we have exhibited the restricted Lorentz group as a principal fiber bundle with fibers SO(3) and base H3. Since the latter is homeomorphic to R3, while SO(3) is homeomorphic to three-dimensional real projective space RP3, we see that the restricted Lorentz group is locally homeomorphic to the product of RP3 with R3. However, globally, our fiber bundle has a twist similar to the way in which the Möbius band may be regarded as a twisted line bundle over the circle S1.

To understand this, we need to return to the double covering of SO(3) by SU(2). The reason SO(3) is doubly connected is that a curve in SO(3) representing a full turn of some rigid body cannot be contracted to the identity element, while a curve representing two full turns can be. Hermann Weyl and Paul Dirac each offered vivid illustrations of why this is so. For Weyl's argument, see the book by Frankel cited below.