Parallel Transport and Covariant Derivatives

Hello! Let's continue our journey into the mathematical heart of General Relativity.

In our last lesson, we accomplished a crucial task: we derived the Levi-Civita connection, which gives us a unique, unambiguous way to calculate the covariant derivative $\nabla$ on a manifold equipped with a metric $g_{μν}$ . We now have the machinery to discuss how tensor fields change from point to point in a coordinate-independent manner.

Today, we will use this machinery to answer a fundamental question: How can we compare a vector at one point in spacetime with a vector at another? In flat Euclidean space, we can simply slide one vector to the other's location without changing its components. On a curved manifold, this is no longer straightforward. The very act of "moving" a vector introduces changes that depend on the path taken. This lesson will introduce the rigorous procedure for this process, known as parallel transport.

Our learning outcome is to define parallel transport along a curve and explain its relation to the covariant derivative. This concept is the geometric foundation for understanding motion in curved spacetime and will lead us directly to the geodesic equation, which describes the paths of free-falling particles.

1. The Problem of Comparing Vectors

In flat space, the idea of a "constant" vector field is trivial—all the vector arrows have the same length and point in the same direction. We can check if two vectors at different points are the same by simply sliding one over to the other.

On a curved manifold, this simple procedure fails. The tangent spaces at different points are distinct, and there's no natural way to identify a vector in one tangent space with a vector in another.

To build your intuition for this, let's start with a video that vividly illustrates the problem using the surface of a sphere.

Tensor Calculus 18: Covariant Derivative (extrinsic) and Parallel Transport

The YouTuber eigenchris provides an excellent visual explanation of why comparing vectors on a curved surface is problematic and introduces the intuitive idea of parallel transport as a solution.

Please watch from the beginning to 06:50. 00:00 - 02:11: This part sets up the contrast between comparing vectors in flat space and on a curved surface. 02:11 - 04:04: Focus on the sphere example. It clearly shows why just 'sliding' a vector is not a well-defined operation from the perspective of an observer on the surface. 04:04 - 06:50: This section introduces the concept of parallel transport as keeping a vector 'as straight as possible' and demonstrates the crucial property of path dependence. Notice how transporting a vector around a closed loop can change its orientation.

As the video shows, the result of transporting a vector from point P to point Q depends on the path taken between them. This path dependence is a deep and fundamental manifestation of curvature. Our goal now is to turn the intuitive idea of "keeping a vector as straight as possible" into a precise mathematical statement.

2. The Equation of Parallel Transport

The intuitive notion of "keeping a vector constant" as we move it along a curve means that its rate of change along that curve should be zero. On a manifold, the correct, coordinate-independent notion of a derivative is the covariant derivative.

Let's consider a curve $γ$ parameterized by $λ$ , so its coordinates are given by $x^{μ} (λ)$ . The tangent vector to this curve is $T^{μ} = \frac{d x ^{μ}}{d λ}$ . Now, consider a vector field $V^{ν}$ defined along this curve. The covariant derivative of $V^{ν}$ along the curve $γ$ is the contraction of the full covariant derivative $\nabla_{μ} V^{ν}$ with the curve's tangent vector:

\frac{D V ^{ν}}{d λ} \equiv T^{μ} \nabla_{μ} V^{ν} = \frac{d x ^{μ}}{d λ} \nabla_{μ} V^{ν}

This object measures how $V^{ν}$ changes as we move infinitesimally along the curve $γ$ .

With this, we can now give the formal definition of parallel transport.

Definition: A vector field $V$ is said to be parallel transported along a curve $γ$ if its covariant derivative along the curve is zero.

\frac{D V ^{ν}}{d λ} = \frac{d x ^{μ}}{d λ} \nabla_{μ} V^{ν} = 0

This is the equation of parallel transport.

Let's expand this equation using the definition of the covariant derivative and the chain rule:

\frac{d x ^{μ}}{d λ} (\frac{\partial V ^{ν}}{\partial x ^{μ}} + Γ_{μ σ}^{ν} V^{σ}) = 0

Recognizing that $\frac{d x ^{μ}}{d λ} \frac{\partial V ^{ν}}{\partial x ^{μ}} = \frac{d V ^{ν}}{d λ}$ (the ordinary derivative of the vector's components with respect to the curve parameter), we get the standard form of the parallel transport equation:

\frac{d V ^{ν}}{d λ} + Γ_{μ σ}^{ν} \frac{d x ^{μ}}{d λ} V^{σ} = 0

This is a system of $n$ coupled first-order linear ordinary differential equations for the $n$ components of the vector $V^{ν} (λ)$ . A fundamental theorem of ODEs guarantees that if we specify an initial vector $V^{ν} (λ_{0})$ at a starting point $x^{μ} (λ_{0})$ , there exists a unique solution for $V^{ν} (λ)$ all along the curve.

This equation is the mathematical embodiment of "keeping the vector as constant as possible". Any change in the vector's components ( $\frac{d V ^{ν}}{d λ}$ ) is precisely cancelled by the term involving the Christoffel symbols, which accounts for the "twisting" of the coordinate system.

The following resources formalize this definition.

General Relativity

This section from the UCDavis lecture notes provides a concise and direct definition of parallel transport, showing clearly how it is formulated as a first-order differential equation.

Please read the first part of Lecture 7, titled 'Today we’ll discuss parallel transport.' Focus on the definition given by Eq. (7.1) and its expansion into the ODE in Eq. (7.3). Notice how the text points out that this is a first-order differential equation that can be solved by iteration.

Lecture Notes on General Relativity - S. Carroll

Sean Carroll's notes offer a slightly more abstract but powerful perspective, defining the covariant derivative along a path as an operator D/d\lambda. This viewpoint is very common in more advanced literature.

Please read the section '3.4 Parallel Transport, Geodesics, and Covariant Derivatives.' Focus on the definitions in equations (3.30) and (3.31), which introduce the operator D/d\lambda and the general equation of parallel transport. This will solidify the abstract concept behind the coordinate-based equation we derived.

3. Properties of Parallel Transport

The properties of the Levi-Civita connection directly translate into important physical properties of parallel transport.

Preservation of Inner Products

In the last lesson, we established that the Levi-Civita connection is metric-compatible, meaning $\nabla_{σ} g_{μν} = 0$ . This has a profound consequence for parallel transport. Consider two vectors, $V^{μ}$ and $W^{ν}$ , that are both parallel transported along the same curve $γ$ . Let's see how their inner product $g_{μν} V^{μ} W^{ν}$ changes along the curve:

\frac{D}{d λ} (g_{μν} V^{μ} W^{ν}) = \frac{d x ^{σ}}{d λ} \nabla_{σ} (g_{μν} V^{μ} W^{ν})

Using the Leibniz rule for the covariant derivative:

= \frac{d x ^{σ}}{d λ} [(\nabla_{σ} g_{μν}) V^{μ} W^{ν} + g_{μν} (\nabla_{σ} V^{μ}) W^{ν} + g_{μν} V^{μ} (\nabla_{σ} W^{ν})]

= (\frac{d x ^{σ}}{d λ} \nabla_{σ} g_{μν}) V^{μ} W^{ν} + g_{μν} (\frac{D V ^{μ}}{d λ}) W^{ν} + g_{μν} V^{μ} (\frac{D W ^{ν}}{d λ})

Now we examine each term:

$\frac{d x ^{σ}}{d λ} \nabla_{σ} g_{μν} = 0$ because the connection is metric-compatible.
$\frac{D V ^{μ}}{d λ} = 0$ because $V$ is parallel transported.
$\frac{D W ^{ν}}{d λ} = 0$ because $W$ is parallel transported.

Therefore, the entire expression is zero:

\frac{d}{d λ} (g_{μν} V^{μ} W^{ν}) = 0

This means that parallel transport preserves inner products. Consequently, the lengths of vectors and the angles between them remain constant when they are parallel transported. This perfectly matches our physical intuition of moving an object without stretching or rotating it.

Test your understanding!

A vector $V^{μ}$ is parallel transported along a timelike curve $γ$ parameterized by its proper time $τ$ . At $τ = 0$ , the vector is spacelike, meaning $g_{μν} V^{μ} V^{ν} > 0$ . What can you say about the vector $V^{μ}$ at any other point on the curve, $τ > 0$ ? Is it possible for it to become timelike or null?

Show answer

It is not possible. Since parallel transport preserves inner products, the norm of the vector, $g_{μν} V^{μ} V^{ν}$ , must be constant along the curve. If it starts with a positive value (spacelike), it must maintain that same positive value for all $τ$ . It can never become zero (null) or negative (timelike).

The Formal Solution: Path-Ordered Exponential

For those who appreciate the complete mathematical structure, it's worth knowing that the ODE for parallel transport has a formal, explicit solution. By rewriting the equation as $\frac{d}{d λ} V = A (λ) V$ (in matrix notation) and solving it iteratively, one arrives at an infinite series solution. This series can be written compactly as a path-ordered exponential.

This is analogous to the time-evolution operator in quantum mechanics, which can be expressed as a time-ordered exponential of the Hamiltonian. Sean Carroll's notes (in the same section you just read, from eq. 3.35 onwards) provide a nice derivation of this, arriving at the formal solution:

V^{μ} (λ) = [P exp (- \int_{λ_{0}}^{λ} Γ_{σ ρ}^{μ} \frac{d x ^{σ}}{d η} d η)]_{ν}^{ρ} V^{ν} (λ_{0})

where $P$ is the path-ordering operator. You are not expected to use this formula for calculations at this stage, but it is important to know that a closed-form solution exists and that it depends intrinsically on the entire path of integration.

4. The Bridge to Geodesics

We have defined a procedure for transporting a vector along a pre-defined curve. Now, let's ask a deeper question: what if a curve is defined by the property that it parallel transports its own tangent vector?

Let the curve be $x^{μ} (λ)$ with tangent vector $T^{μ} = \frac{d x ^{μ}}{d λ}$ . The condition that the curve parallel transports its own tangent vector is:

\frac{D T ^{μ}}{d λ} = T^{ν} \nabla_{ν} T^{μ} = 0

This equation defines a very special class of curves. A curve that satisfies this condition is moving "as straight as possible" through spacetime—it experiences no covariant acceleration. Such a curve is called a geodesic.

Expanding this equation gives:

\frac{d x ^{ν}}{d λ} (\frac{\partial T ^{μ}}{\partial x ^{ν}} + Γ_{ν σ}^{μ} T^{σ}) = 0

\frac{d T ^{μ}}{d λ} + Γ_{ν σ}^{μ} \frac{d x ^{ν}}{d λ} T^{σ} = 0

Substituting $T^{μ} = \frac{d x ^{μ}}{d λ}$ , we get the geodesic equation:

\frac{d ^{2} x ^{μ}}{d λ ^{2}} + Γ_{ν σ}^{μ} \frac{d x ^{ν}}{d λ} \frac{d x ^{σ}}{d λ} = 0

This second-order differential equation describes the paths of freely falling particles in a gravitational field described by the metric (via the Christoffel symbols). It is one of the most important equations in General Relativity.

The UCDavis notes you read earlier make this connection explicit at the end of the section on parallel transport. It's a beautiful and direct link: a geodesic is simply a curve whose tangent vector is constant with respect to its own flow.

Conclusion

In this lesson, we have made a significant conceptual leap, moving from the static definition of a derivative to the dynamic process of transport along a path.

Key Takeaways:

Comparing vectors at different points on a curved manifold is non-trivial and requires a rule for transport.
Parallel transport is the process of moving a vector along a curve while keeping it "covariantly constant".
The mathematical definition of parallel transport for a vector $V$ along a curve with tangent $T$ is the equation $\frac{D V}{d λ} = T^{μ} \nabla_{μ} V = 0$ .
This is a system of first-order ODEs, which guarantees a unique solution for a given initial vector.
Because the Levi-Civita connection is metric-compatible, parallel transport preserves the lengths of vectors and the angles between them.
A curve that parallel transports its own tangent vector is called a geodesic. This provides a fundamental, geometric definition for the paths of freely falling particles.

We have now laid all the necessary groundwork. In our next lesson, we will dive deep into the geodesic equation, deriving it from a different perspective—the principle of extremal proper time—and analyzing its properties. This will be our first major step in calculating the motion of particles in curved spacetime.

Recommended resources