If we look at the killing equation for a vector field $X$ in $\mathbb{R}^{(p,q)}$ (or on an open subset thereof) in coordinates with constant diagonal pseudo-metric we get:

$$X_{\mu,\nu}+X_{\nu,\mu}=0 \tag{1}$$

In the case $\mu=\nu$ it is clear that this implies $X^\mu$ does not depend on the coordinate $q^\mu$. In the book I am reading it is remarked that the equation also implies that $X^\mu$ in general can be put in the form:

$$X^\mu=c^\mu+\omega^\mu_\nu q^\nu$$

Where (1) implies for $\omega$:


where $g$ is the metric.

My question is, why does the Killing equation imply that $X$ is linear in the chosen coordinates?


This is essentially a consequence of the connection on $\mathbb{R}^n$ being flat.

One can give the $\omega$ in $X = c + \omega x$ explicitly as $\omega_{\mu\nu} = \partial_\mu X_\nu$, and the $c$ as $c = X - \omega x$.

For a Killing field $X$, one has that $\nabla_\mu\nabla_\nu X_\rho = {R^\sigma}_{\mu\nu\rho}X_\sigma$, where $R$ is the Riemann tensor, but $R=0$ for flat connections, so $\partial_\mu\omega_{\nu\sigma} = \partial_\mu\partial_\nu X_\sigma = 0$, i.e. $\omega$ is constant.

Evaluating $\partial_\mu c_\nu$, one also finds that $c$ is constant if $\omega$ is antisymmetric, and $\partial_\mu X_\nu$ is antisymmetric by virtue of the Killing equation.

Altogether, this gives $X= c+\omega x$ for $c,\omega$ constant and $\omega$ antisymmetric.

  • $\begingroup$ It is surprising to me that you have to use more sophisticated notions than just differentiation in $\mathbb{R}^n$ (curvature). If I look at the equation given it really looks like some elementary algebraic considerations should give me the result. Is there a more pedestrian way to arrive at $\partial \partial X = 0$? $\endgroup$ – s.harp Dec 31 '15 at 11:18

Here is one method:

  1. The Killing equation $$({\cal L}_X\eta)_{\mu\nu}~=~0\tag{1}$$ for a constant metric $$\eta_{\mu\nu}={\rm const}\tag{2}$$
    reads $$ X_{\mu,\nu}+X_{\nu,\mu}~=~0,\tag{3}$$ as OP correctly mentions.

  2. On the other hand consider a coordinate transformation $$ x^{\mu}~~\longrightarrow~~ x^{\prime \nu}~=~f^{\nu}(x), \tag{4}$$ that preserves the metric (2), i.e. $$ \eta_{\mu\nu}~=~\frac{\partial x^{\prime\kappa}}{\partial x^{\mu}}\eta_{\kappa\lambda}\frac{\partial x^{\prime\lambda}}{\partial x^{\nu}}. \tag{5}$$ Note that for an infinitesimal coordinate transformation of the form $$ \delta x^{\mu}~=~x^{\mu \prime}-x^{\mu}~=~ \varepsilon X^{\mu}, \tag{6}$$ eq. (5) becomes exactly the same eq. (3)!

  3. The solutions (4) to eq. (5) is proven to be affine transformations in e.g. this Phys.SE post using various methods.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.