Stationary and Axisymmetric Reduction

A stationary axisymmetric spacetime has two commuting Killing fields, but that fact alone does not make its metric block diagonal. The familiar separation into a two-dimensional Killing-orbit block and a two-dimensional meridional block requires orthogonal transitivity, equivalently Frobenius integrability of the two-planes orthogonal to the group orbits. On a connected region meeting a regular axis, vacuum and suitable nonconvective matter enforce this circularity; generic matter flow need not.

Required background. Use Killing fields and isometry groups for commuting isometries and the Einstein equation for the Ricci conditions below.

Helpful background. Hypersurface normals and tangential projectors supplies the integrable-distribution and projection language.

Two Killing directions and the orbit metric

Let $k^\mu$ generate stationarity and $m^\mu$ generate axial symmetry. Assume

$$\mathcal L_k g_{\mu\nu}=0, \qquad \mathcal L_m g_{\mu\nu}=0, \qquad [k,m]^\mu=0,$$

with $m$ having closed orbits and vanishing on a regular rotation axis. On a regular region where $k$ and $m$ are independent, define their Gram matrix

$$\lambda_{ab}=g(\xi_a,\xi_b), \qquad (\xi_0,\xi_1)=(k,m).$$

Its determinant measures the nondegeneracy of the group orbit. Away from axes and Killing horizons, the metric can always be written locally in Kaluza–Klein form,

$$ds^2=h_{AB}dx^A dx^B +\lambda_{ab}\bigl(dy^a+A^a{}_A dx^A\bigr) \bigl(dy^b+A^b{}_B dx^B\bigr),$$

with $y^a=(t,\phi)$. The connection one-forms $A^a{}_A$ encode the obstruction to choosing surfaces everywhere orthogonal to both Killing fields. Their removal is therefore a geometric claim, not an automatic coordinate simplification.

Circularity as a Frobenius condition

The orthogonal two-plane distribution consists of vectors annihilated by the one-forms $k_\mu dx^\mu$ and $m_\mu dx^\mu$. Frobenius’ theorem says that this distribution is integrable exactly when

$$k\wedge m\wedge dk=0, \qquad k\wedge m\wedge dm=0.$$

In index notation these are

$$k_{[\mu}m_\nu\nabla_\rho k_{\sigma]}=0, \qquad k_{[\mu}m_\nu\nabla_\rho m_{\sigma]}=0.$$

When they hold, coordinates $(x^1,x^2,t,\phi)$ can be chosen so that

$$g_{At}=g_{A\phi}=0,$$

and the metric has two $2\times2$ blocks. This is the geometric content of circularity. The remaining off-diagonal entry $g_{t\phi}$ is not forbidden; it records frame dragging within the Killing-orbit block.

The Killing identity converts derivatives of these Frobenius obstructions into Ricci contractions. A useful sufficient matter condition is

$$k_{[\mu}m_\nu R_{\rho]\sigma}k^\sigma=0, \qquad k_{[\mu}m_\nu R_{\rho]\sigma}m^\sigma=0.$$

Through Einstein’s equation, these become restrictions on stress-energy flux across the Killing orbits. Vacuum satisfies them. So does a source-free Maxwell field invariant under both symmetries, on a connected region meeting the regular axis. To see why the source-free assumption matters, use $dF=d{\star F}=0$ and the commuting symmetries: Cartan’s identity makes both $F(k,m)$ and $({\star F})(k,m)$ constant. They vanish on the axis, where $m=0$, and hence throughout the connected region. These two conditions give the required circular electromagnetic stress tensor. Arbitrary stationary electric currents do not supply the same argument.

A perfect fluid satisfies the Ricci conditions when its four-velocity lies in the span of $k$ and $m$—purely circular motion—but meridional convection can violate them. Carter states the orthogonal-transitivity criterion, including the role of an axis and convective flux, in Carter 1969, Theorem 2, pp. 70–81.

Boundary information is also required. Differential identities propagate a vanishing obstruction, but a constant need not vanish unless it is fixed on an axis, asymptotic region, or other appropriate set. A regular axis supplies this normalization in the standard black-hole setting.

What the reduced equations contain

Once circularity holds, the quotient is a two-dimensional meridional geometry $h_{AB}$ coupled to the symmetric matrix $\lambda_{ab}$. The determinant of $\lambda$ sets a natural area coordinate, while its off-diagonal component carries rotation. Vacuum Einstein equations become a nonlinear matrix-valued harmonic-map system on the quotient together with equations for its conformal factor. This reorganization is powerful because elliptic methods apply in stationary domains, but it is not a time-evolution formulation.

Several limits check the construction. In a static spacetime, $k$ is hypersurface orthogonal and one can set $g_{t\phi}=0$. In Minkowski cylindrical coordinates, $\lambda=\mathrm{diag}(-1,r^2)$ and the meridional block is $dr^2+dz^2$. In Kerr, $g_{t\phi}\ne0$ while $g_{rt}=g_{r\phi}=g_{\theta t}=g_{\theta\phi}=0$ in Boyer–Lindquist coordinates: rotation is compatible with circularity.

Circular coordinates can fail at an axis because $m=0$, at a horizon where the Gram determinant degenerates, or in matter configurations with meridional currents. Such degeneration need not signal curvature singularity. Conversely, writing an ansatz without cross terms assumes circularity and may discard valid stationary axisymmetric solutions.

Exercise

Consider Minkowski spacetime in a uniformly rotating chart, $\phi'=\phi-\Omega t$:

$$ds^2=-(1-\Omega^2r^2)dt^2+2\Omega r^2dt\,d\phi' +r^2d\phi'^2+dr^2+dz^2.$$

Show that the metric is stationary, axisymmetric, and circular even where $\partial_t$ becomes spacelike.

Solution

All components are independent of $t$ and $\phi'$, so $k=\partial_t$ and $m=\partial_{\phi'}$ are commuting Killing fields. There are no cross terms between $(t,\phi')$ and $(r,z)$, so surfaces of constant $t$ and $\phi'$ are orthogonal to the Killing orbits; equivalently both Frobenius four-forms vanish. The norm $g(k,k)=-(1-\Omega^2r^2)$ changes sign at $r=|\Omega|^{-1}$, but the spacetime curvature remains zero. That “light cylinder” is a property of the chosen stationary generator, not a failure of circularity or a spacetime singularity.

References

• Carter, Brandon. “Killing Horizons and Orthogonally Transitive Groups in Space-Time.” Journal of Mathematical Physics 10, no. 1 (1969): 70–81. DOI.