nLab
Levi-Civita connection

Context

Riemannian geometry

Differential geometry

differential geometry

synthetic differential geometry

Axiomatics

Models

Concepts

Theorems

Applications

Contents

Idea

The Levi-Civita connection is the unique symmetric connection on the tangent bundle of a Riemannian manifold or pseudo-Riemannian manifold that is compatible with the metric or pseudo-metric. The curvature and geodesics on a pseudo-Riemannian manifold are taken with respect to this connection. The existence and uniqueness of the Levi-Civita connection is the so-called fundamental theorem of Riemannian geometry.

Definition

Definition

For (X,g)(X,g) a Riemannian manifold, the Levi-Civita connection g\nabla_g on XX is the unique connection on the tangent bundle TXT X that

  1. the covariant derivative of the metric vanishes, gg=0\nabla_{g} g = 0;

  2. g\nabla_g has vanishing torsion.

We say in detail what this means in “first order formalism”.

Definition

For (X,g)(X,g) a (dk,k)(d-k,k)-dimensional Riemannian manifold (for k=0k = 0) or pseudo-Riemannian manifold (for k=1k = 1), the Levi-Civita connection g\nabla_g on XX is the unique ISO(dk,k)ISO(d-k,k)-connection on a bundle \nabla (for ISO(dk)ISO(d-k) the Poincare group) such that

  1. metric compatibility: the vielbein ee gives the metric: g=eeg = e \cdot e;

  2. torsion freeness: the curvature of the vielbein (the torsion) vanishes: T=0T = 0.

More in detail, locally on a patch U iXU_i \subset X the ISO(dk,k)ISO(d-k,k)-connection \nabla is given by a Poincare Lie algebra-valued 1-form

(e,ω):TU i𝔦𝔰𝔬(dk,k) d𝔰𝔬(dk,k) (e, \omega) : T U_i \to \mathfrak{iso}(d-k,k) \simeq \mathbb{R}^d \ltimes \mathfrak{so}(d-k,k)

with

  1. e ie_i an d\mathbb{R}^d-valued form – the vielbein;

  2. ω i\omega_i a special orthogonal Lie algebra-valued form – the “spin connection”.

The curvature 2-form of this similarly decomposes into

  1. the torsion T a:=F e a=de a+ω a be bT^a := F_{e}^a = d e^a + \omega^a{}_b \wedge e^b,

    (this equation is also called the first Cartan structure equation)

  2. the Riemann curvature R g a b:=F ω a b=dω a b+ω a cω c dR_{g}{}^a{}_b := F_{\omega}^a{}_b = d \omega^a{}_b + \omega^a{}_c \wedge \omega^c{}_d.

    (this equation is also called the second Cartan structure equation)

The Bianchi identity satisfied by this curvature is

  1. dF e a+ω a bF e=F ω a be bd F_e^a + \omega^a{}_b F_e = F_\omega^a{}_b \wedge e^b;

  2. dF ω a b+ω a cF ω c bF ω a cω c b=0d F_\omega^a{}_b + \omega^a{}_c \wedge F_\omega^c{}_b - F_\omega^a{}_c \wedge \omega^c{}_b = 0.

The metric compatibility condition in the definition of Levi-Civita connection says that

g=e ae a. g = e^a \otimes e_a \,.

The torsion-freeness condition says that

F e=0. F_e = 0 \,.

In terms of Christoffel symbols

The Levi-Civita connection may be discussed in terms of its components – called Christoffel symbols – given by the canonical local trivialization of the tangent bundle over a coordinate patch. This has been the historical route and is still widely used in the literature.

Metric compatibility

Here a metric gg is compatible with the connection \nabla or preserved by it (here thought of in its incarnation as a covariant derivative) if and only if Xg=0\nabla_X g = 0 for all XX, which is equivalent to the preservation of the metric inner product of tangent vectors under parallel translation. Since

(1)X(g(X 1,X 2))=( Xg)(X 1,X 2)+g( XX 1,X 2)+g(X 1, XX 2), X(g(X_1,X_2)) = (\nabla_X g)(X_1,X_2) + g(\nabla_X X_1, X_2) + g(X_1, \nabla_X X_2) ,

by the fact that covariant differentiation commutes with contractions and satisfies the derviative identity, compatibility is equivalent to

(2)X(g(X 1,X 2))=g( XX 1,X 2)+g(X 1, XX 2), X (g (X_1,X_2)) = g(\nabla_X X_1, X_2) + g(X_1, \nabla_X X_2),

for all X,X 1,X 2X,X_1, X_2.

Uniqueness and existence on n\mathbb{R}^n

Now assume M nM \subset \mathbb{R}^n and we have such a connection associated to gg.
Then the connection is uniquely determined by its Christoffel symbols, which we can determine in terms of gg by a bit of elementary algebra. In other words, we just need to compute i j\nabla_{\partial_i} \partial_j. Now

(3) kg( i, j)=g( k i, j)+g( i, k j). \partial_k g( \partial_i, \partial_j) = g( \nabla_{\partial_k} \partial_i, \partial_j) + g( \partial_i, \nabla_{\partial_k}\partial_j).

We can get two other equations by cyclic permutation:

(4) ig( j, k)=g( i j, k)+g( j, i k) \partial_i g( \partial_j, \partial_k) = g( \nabla_{\partial_i} \partial_j, \partial_k) + g( \partial_j, \nabla_{\partial_i}\partial_k)
(5) jg( k, i)=g( j k, i)+g( k, j i) \partial_j g( \partial_k, \partial_i) = g( \nabla_{\partial_j} \partial_k, \partial_i) + g( \partial_k, \nabla_{\partial_j}\partial_i)

So let S ij:= i j= j iS_{i j} := \nabla_{\partial_i} \partial_j = \nabla_{\partial_j} \partial_i, by symmetry. Let T ijk:= ig( j, k)T_{i j k} := \partial_i g( \partial_j, \partial_k); these are smooth real functions. These equations can be written

(6)T kij=g(S ik, j)+g(S jk, i) T_{k i j} = g( S_{i k}, \partial_j) + g( S_{j k}, \partial_i)
(7)T ijk=g(S ij, k)+g(S ik, j) T_{i j k} = g( S_{i j}, \partial_k) + g( S_{i k}, \partial_j)
(8)T jki=g(S jk, i)+g(S ij, k) T_{j k i} = g( S_{j k}, \partial_i) + g( S_{i j}, \partial_k)

These are three linear equations in the unknowns g(S ik, j),g(S jk, i),g(S ij, k)g( S_{i k}, \partial_j), g( S_{j k}, \partial_i), g( S_{i j}, \partial_k). The system is nonsingular, so we get a unique solution, and consequently by nondegeneracy a unique possibility for the S ijS_{i j}.

Incidentally, we have in fact shown the uniqueness assertion of the general theorem, since that is local.

We shall now prove existence in this restricted case. Choose S ijS_{i j} to satisfy the system of three equations outlined above where i<j<ki \lt j \lt k. Then set S ji:=S ijS_{j i} := S_{i j}, and we have a connection \nabla with i j:=S ij\nabla_{\partial_i} \partial_j := S_{i j} since the vector fields i\partial_i are a frame (i.e. a basis at each tangent space on MM). It is symmetric, since the torsion TT vanishes (by S ij=S jiS_{i j}=S_{j i}) on pairs ( i, j)(\partial_i,\partial_j), and hence identically, since it is a tensor.

We must check for compatibility. The difference of the two terms in (1) vanishes when X,X 1,X 2X,X_1,X_2 are of the form i\partial_i. The vanishing holds generally because the difference of the two sides, which is ( Xg)(X 1,X 2)(\nabla_X g)(X_1,X_2), is a tensor. Hence compatibility follows.

Uniqueness and existence in the general case

We have already shown the uniqueness assertion, since that is local. Connections restrict to connections on open subsets.

We have proved the existence of \nabla when MM is an open submanifold of n\mathbb{R}^n (though not necessarily with the canonical metric i=1 ndx idx i\sum_{i=1}^n d x_i \otimes d x_i). In general, cover MM by open subsets U iU_i diffeomorphic to an open set in n\mathbb{R}^n. We get connections i\nabla_i on U iU_i compatible with g U ig|_{U_i}.

We claim that i U iU j= j U iU j\nabla_i|_{U_i \cap U_j} = \nabla_j|_{U_i \cap U_j}. This is an easy corollary of uniquness. So we can patch the connections together to get the one Levi-Civita connection on MM.

Applications

In physics

In the physics, the theory of general relativity models the field of gravity in terms of the Levi-Civita connection on a Lorentzian manifold. See there for more details.

References

A discussion in terms of synthetic differential geometry is in

  • Gonzalo Reyes, General Relativity:

    Metrics, connections and curvature (pdf)

    The Riemann-Christoffel tensor (pdf)

    Affine connections, parallel transport and sprays (pdf)

Revised on April 21, 2012 09:51:55 by Urs Schreiber (82.113.119.35)