Skip to content

Scrollable formula. Use Left and Right arrow keys to reveal hidden content.

Scrollable table. Use Left and Right arrow keys to reveal additional columns.

Scrollable figure. Use Left and Right arrow keys to inspect hidden content.

The ADM action is the Einstein–Hilbert action reorganized as a time integral over spatial metrics. Its bulk Lagrangian contains the intrinsic scalar curvature of a slice and a quadratic kinetic term in its extrinsic curvature. Lapse and shift enter without time derivatives, foreshadowing their role as multipliers of constraints rather than propagating canonical fields.

Required background. Presymplectic Potentials and Currents supplies the passage from a Lagrangian variation to canonical data, while The Gibbons-Hawking-York Boundary Term supplies the boundary term needed when induced metrics are fixed.

Helpful background. The Einstein Equation supplies the covariant field equation, Hypersurface Normals and Tangential Projectors supplies the slice geometry, and Formulations of General Relativity places the canonical description among equivalent continuum formulations.

Consider a region bounded initially and finally by spacelike slices and, when present, by a timelike side boundary. The gravitational action starts from

Sgrav=116πG∫Md4x−g((4)R−2Λ)+Sboundary.S_{\rm grav} =\frac1{16\pi G} \int_{\mathcal M}d^4x\sqrt{-g} \left({}^{(4)}R-2\Lambda\right) +S_{\rm boundary}.

Use the chapter conventions

(∂t)μ=αnμ+βμ,γμν=gμν+nμnν,Kij=−12Lnγij.(\partial_t)^\mu=\alpha n^\mu+\beta^\mu, \qquad \gamma_{\mu\nu}=g_{\mu\nu}+n_\mu n_\nu, \qquad K_{ij}=-\frac12\mathcal L_n\gamma_{ij}.

The metric determinant factors as

−g=αγ.\sqrt{-g}=\alpha\sqrt\gamma.

The remaining task is to rewrite the four-dimensional scalar curvature. The Gauss relation supplies its spatial part, while the two-normal Ricci relation supplies a normal derivative of KK and a spatial divergence involving the lapse.

For the selected sign of KijK_{ij}, the curvature identity can be arranged as

−g (4)R=αγ((3)R+KijKij−K2)−2∂t(γK)+2∂i[γ(Kβi−Diα)].\begin{aligned} \sqrt{-g}\,{}^{(4)}R ={}&\alpha\sqrt\gamma \left( {}^{(3)}R+K_{ij}K^{ij}-K^2 \right)\\ &-2\partial_t(\sqrt\gamma K) +2\partial_i \left[ \sqrt\gamma (K\beta^i-D^i\alpha) \right]. \end{aligned}

Here DiD_i is the spatial Levi–Civita derivative and γ=det⁡(γij)\gamma=\det(\gamma_{ij}). The volume element is a density, so its Lie derivative is Lβγ=∂i(βiγ)\mathcal L_\beta\sqrt\gamma=\partial_i(\beta^i\sqrt\gamma), not merely βi∂iγ\beta^i\partial_i\sqrt\gamma. The split can be checked from ∂⊥γ=−αKγ\partial_\perp\sqrt\gamma=-\alpha K\sqrt\gamma, where ∂⊥=∂t−Lβ\partial_\perp=\partial_t-\mathcal L_\beta. The normal derivative of KK first produces a +K2+K^2 term; integration by parts through the evolving volume element changes the surviving bulk combination to KijKij−K2K_{ij}K^{ij}-K^2.

The total time derivative is precisely the second-normal-derivative content that the Gibbons–Hawking–York terms on the initial and final caps cancel for a Dirichlet variational problem. The spatial divergence combines with the action on a timelike side boundary. It may not be discarded on a finite domain or at an asymptotic end: after the Legendre transform it becomes part of the surface Hamiltonian.

With the cap term treated and the still-relevant side-boundary contribution denoted by S∂ΣS_{\partial\Sigma}, the ADM form is

Sgrav=116πG∫dt∫Σtd3x αγ((3)R+KijKij−K2−2Λ)+S∂Σ.S_{\rm grav} =\frac1{16\pi G} \int dt\int_{\Sigma_t}d^3x\, \alpha\sqrt\gamma \left( {}^{(3)}R+K_{ij}K^{ij}-K^2-2\Lambda \right) +S_{\partial\Sigma}.

Gourgoulhon derives the scalar-curvature split and bulk action in Gourgoulhon 2007, §3.4.2, pp. 48–50 and §4.5.1, pp. 66–67, PDF. The original canonical construction, including the role of surface terms, is developed in Arnowitt, Deser, and Misner 1962/2004, §§3–5, PDF.

In adapted coordinates,

Kij=−12α(γ˙ij−Diβj−Djβi).K_{ij} =-\frac1{2\alpha} \left( \dot\gamma_{ij} -D_i\beta_j-D_j\beta_i \right).

Therefore the ADM bulk Lagrangian depends on γ˙ij\dot\gamma_{ij} through KijK_{ij} but contains neither α˙\dot\alpha nor β˙i\dot\beta^i. The Hessian with respect to all ten metric velocities is degenerate. In canonical language,

pα≡∂L∂α˙=0,pi≡∂L∂β˙i=0p_\alpha\equiv\frac{\partial\mathcal L}{\partial\dot\alpha}=0, \qquad p_i\equiv\frac{\partial\mathcal L}{\partial\dot\beta^i}=0

are primary constraints. Variation with respect to lapse and shift will yield the Hamiltonian and momentum constraints rather than evolution equations for α\alpha and βi\beta^i.

The kinetic combination is indefinite. Decompose

Kij=Aij+13Kγij,γijAij=0.K_{ij}=A_{ij}+\frac13K\gamma_{ij}, \qquad \gamma^{ij}A_{ij}=0.

Then

KijKij−K2=AijAij−23K2.K_{ij}K^{ij}-K^2 =A_{ij}A^{ij}-\frac23K^2.

The negative trace direction is the conformal mode of the DeWitt metric on spatial metrics. This indefiniteness is a structural property of the gravitational kinetic term; it is not evidence that coordinate time evolution is itself an instability.

For inertial Minkowski slices,

(3)R=0,Kij=0,Λ=0,{}^{(3)}R=0, \qquad K_{ij}=0, \qquad \Lambda=0,

so the ADM bulk density vanishes.

Milne slices of the same flat spacetime have

(3)R=−6τ2,Kij=−1τγij,K=−3τ.{}^{(3)}R=-\frac6{\tau^2}, \qquad K_{ij}=-\frac1\tau\gamma_{ij}, \qquad K=-\frac3\tau.

Hence

KijKij−K2=3τ2−9τ2=−6τ2,K_{ij}K^{ij}-K^2 =\frac3{\tau^2}-\frac9{\tau^2} =-\frac6{\tau^2},

and the bulk integrand is −12/τ2-12/\tau^2, not zero. This does not contradict (4)R=0{}^{(4)}R=0: the omitted total derivative contributes +12/τ2+12/\tau^2 locally. Milne slicing therefore gives a decisive warning that a boundary term may vanish after integration under specified conditions but is not pointwise negligible.

Discarding a divergence requires boundary conditions. A total derivative can change the variational principle and the value of the Hamiltonian even though it does not change bulk Euler–Lagrange equations.

The ADM Lagrangian does not make lapse and shift dynamical. Their absence from the velocity Hessian is the origin of primary constraints.

A nonzero bulk density need not imply spacetime curvature. On Milne slices, the divergence terms complete the identity (4)R=0{}^{(4)}R=0.

Exercise: verify the Milne curvature split

Section titled “Exercise: verify the Milne curvature split”

For the Milne data above with α=1\alpha=1 and βi=0\beta^i=0, evaluate the bulk term and the time derivative −2∂τ(γK)/γ-2\partial_\tau(\sqrt\gamma K)/\sqrt\gamma. Show that their sum vanishes.

Solution

Since γ=τ3q\sqrt\gamma=\tau^3\sqrt q and K=−3/τK=-3/\tau,

γK=−3τ2q,\sqrt\gamma K=-3\tau^2\sqrt q,

so

−2γ∂τ(γK)=−2τ3q(−6τq)=12τ2.-\frac2{\sqrt\gamma} \partial_\tau(\sqrt\gamma K) =-\frac2{\tau^3\sqrt q}(-6\tau\sqrt q) =\frac{12}{\tau^2}.

The bulk curvature combination is

(3)R+KijKij−K2=−6τ2+3τ2−9τ2=−12τ2.{}^{(3)}R+K_{ij}K^{ij}-K^2 =-\frac6{\tau^2}+\frac3{\tau^2}-\frac9{\tau^2} =-\frac{12}{\tau^2}.

The spatial divergence vanishes because the lapse is spatially constant and the shift is zero. Bulk plus time derivative is therefore zero, as required by flat four-dimensional spacetime.

  • Arnowitt, Richard, Stanley Deser, and Charles W. Misner. “The Dynamics of General Relativity.” In Gravitation: An Introduction to Current Research, edited by Louis Witten, 227–265. Wiley, 1962; reprinted as arXiv:gr-qc/0405109 (2004). arXiv.
  • Gourgoulhon, Éric. 3+1 Formalism and Bases of Numerical Relativity. Lecture notes, arXiv:gr-qc/0703035v1 (2007). arXiv.