BMS Group and Asymptotic Symmetries at $\mathscr{I}^+$

1. Full Bondi–Sachs Metric with Sub-leading Terms

We adopt retarded Bondi coordinates $(u,r,x^A)$ with $x^A=(\theta,\phi)$ on the celestial sphere. The Bondi gauge conditions are:

$$g_{rr} = 0, \qquad g_{rA} = 0, \qquad \partial_r\!\left(\det\!\left(\frac{g_{AB}}{r^2}\right)\right) = 0$$

These fix $r$ to be the luminosity distance and $u$ labels null hypersurfaces. The most general metric consistent with these conditions, expanded through $O(r^{-2})$, reads:

$$ds^2 = -\left(1 - \frac{2m_B}{r} - \frac{1}{16r^2}C_{AB}C^{AB}\right)du^2 - 2\,du\,dr$$

$$\quad + r^2\gamma_{AB}\,dx^A dx^B + r\,C_{AB}\,dx^A dx^B$$

$$\quad + D^B C_{AB}\,du\,dx^A + \frac{1}{r}\left(-\frac{2}{3}N_A + \frac{1}{3}C_{AB}D_C C^{BC}\right)du\,dx^A + O(r^{-2})$$

Here $\gamma_{AB}$ is the round metric on $S^2$, $m_B(u,x^A)$ is the Bondi mass aspect, $C_{AB}(u,x^A)$ is the symmetric trace-free (STF) shear tensor, and $N_A(u,x^B)$ is the angular momentum aspect. The crucial $C_{AB}C^{AB}/(16r^2)$ correction to $g_{uu}$ arises from the determinant condition: expanding

$$\det\!\left(\frac{g_{AB}}{r^2}\right) = \det\!\left(\gamma_{AB} + \frac{1}{r}C_{AB} + \cdots\right) = \det(\gamma_{AB})\left(1 + \frac{1}{2r^2}C_{AB}C^{AB} + \cdots\right)$$

where we used $\gamma^{AB}C_{AB}=0$ (trace-free) so the linear term vanishes. Setting $\partial_r(\det(g_{AB}/r^2))=0$ at each order generates the sub-leading metric corrections.

2. News Tensor and Gauge Invariance

The Bondi news tensor is defined as:

$$N_{AB} := \partial_u C_{AB}$$

To prove gauge invariance, consider a residual diffeomorphism preserving Bondi gauge. Under such a transformation with supertranslation parameter $f(x^A)$ and Lorentz generator $Y^A$, the shear transforms as:

$$C_{AB} \;\to\; C_{AB} + 2D_{\langle A}D_{B\rangle}f + \mathcal{L}_Y C_{AB} + \frac{u}{2}(D_CY^C)\,N_{AB}$$

Taking $\partial_u$ of the supertranslation piece:

$$\partial_u\!\left(2D_{\langle A}D_{B\rangle}f\right) = 0$$

since $f$ is independent of $u$. Therefore $N_{AB}$ is invariant under supertranslations. Under Lorentz transformations, $N_{AB}$ transforms tensorially (as it should for a physical observable). Gravitational radiation is present if and only if $N_{AB} \neq 0$.

3. Bondi Mass Loss Formula from Einstein Equations

We derive the mass loss by imposing $G_{uu} = 8\pi T_{uu}$ (vacuum: $T_{uu}=0$) at order $r^{-1}$. Computing the Ricci tensor from the full Bondi-Sachs metric, the $uu$-component at leading non-trivial order gives:

$$R_{uu}\Big|_{r^{-1}} = -\frac{2}{r}\partial_u m_B - \frac{1}{4r}N_{AB}N^{AB} + \frac{1}{2r}D_AD_B N^{AB} + O(r^{-2})$$

Setting $R_{uu}|_{r^{-1}} = 0$ and solving for $\partial_u m_B$:

$$\boxed{\partial_u m_B = -\frac{1}{8}N_{AB}N^{AB} + \frac{1}{4}D_AD_B N^{AB}}$$

Integrating over $S^2$ and using $\oint D_AD_B N^{AB}\,d^2\Omega = 0$ (total divergence), the total Bondi mass $M(u) = \frac{1}{4\pi}\oint m_B\,d^2\Omega$ satisfies:

$$\frac{dM}{du} = -\frac{1}{32\pi}\oint N_{AB}N^{AB}\,d^2\Omega \;\leq\; 0$$

The inequality follows because $N_{AB}N^{AB} = 2|N_{AB}|^2 \geq 0$ (positive-definite norm of a real symmetric trace-free tensor). This is the Bondi mass loss theorem: gravitational waves always carry positive energy to $\mathscr{I}^+$.

4. Derivation of the BMS Vector Field

We seek the most general vector field $\xi = \xi^u\partial_u + \xi^r\partial_r + \xi^A\partial_A$ that preserves the Bondi gauge conditions. Start with the three requirements:

(i) $\mathcal{L}_\xi g_{rr} = 0$: Since $g_{rr}=0$ and $g_{ur}=-1$, the Lie derivative gives $-2\partial_r\xi^u = 0$, so $\xi^u$ is independent of $r$.

(ii) $\mathcal{L}_\xi g_{rA} = 0$: This yields $g_{AB}\partial_r\xi^B = -\partial_A\xi^u$. With $g_{AB} = r^2\gamma_{AB} + rC_{AB} + \cdots$, we solve order-by-order to get:

$$\xi^A = Y^A(x^B) + \frac{1}{r}D^A\xi^u + O(r^{-2})$$

(iii) $\partial_r\!\left(\det(g_{AB}/r^2)\right)=0$ preserved: This constrains $\xi^r$. Computing $\mathcal{L}_\xi(\det(g_{AB}/r^2))$ and using $\gamma^{AB}\mathcal{L}_\xi g_{AB}|_{\text{leading}} = 2(D_AY^A - 2\xi^r/r)$, we obtain:

$$\xi^r = -\frac{r}{2}D_AY^A + \frac{1}{2}D^2\xi^u + O(r^{-1})$$

Finally, writing $\xi^u = f(x^A) + \frac{u}{2}D_AY^A$ (the $u$-dependent piece is fixed by requiring the transformation to be well-defined for all $u$), the full BMS vector field is:

$$\boxed{\xi = \left(f + \frac{u}{2}D_AY^A\right)\partial_u - \frac{1}{2}\left(rD_AY^A - D^2 f\right)\partial_r + \left(Y^A - \frac{1}{r}D^A f\right)\partial_A}$$

Here $f(x^A)$ parametrises supertranslations and $Y^A(x^B)$ parametrises Lorentz transformations (conformal Killing vectors of $S^2$ for the standard BMS group, or arbitrary smooth vector fields for the extended BMS group).

5. Supertranslation Action on Shear via Lie Derivative

Under a pure supertranslation ($Y^A=0$), we compute $\delta C_{AB} = \mathcal{L}_\xi g_{AB}|_{O(r)}$. The relevant Lie derivative components are:

$$\mathcal{L}_\xi g_{AB} = \xi^\mu\partial_\mu g_{AB} + g_{\mu B}\partial_A\xi^\mu + g_{A\mu}\partial_B\xi^\mu$$

At order $r$, the $\xi^u\partial_u g_{AB}$ term gives $f \cdot r\,\partial_u C_{AB} = f\,r\,N_{AB}$. The $\xi^r\partial_r g_{AB}$ term with $\xi^r = \frac{1}{2}D^2 f + O(r^{-1})$ gives $\frac{1}{2}D^2 f \cdot 2r\gamma_{AB} = r\,D^2 f\,\gamma_{AB}$. The $g_{\mu B}\partial_A\xi^\mu$ terms contribute $-2r^2\gamma_{C(A}\partial_{B)}\left(\frac{-D^C f}{r}\right) = 2r\,D_{(A}D_{B)}f$. Collecting:

$$\delta C_{AB} = f\,N_{AB} + 2D_{(A}D_{B)}f + D^2 f\,\gamma_{AB}$$

For time-independent vacua ($N_{AB}=0$), and noting $D_{\langle A}D_{B\rangle} = D_{(A}D_{B)} - \frac{1}{2}\gamma_{AB}D^2$, we get $2D_{(A}D_{B)}f + D^2 f\,\gamma_{AB} = 2D_{\langle A}D_{B\rangle}f + 2\gamma_{AB}D^2 f$. But $C_{AB}$ is trace-free, so only the STF part contributes:

$$\boxed{C_{AB} \;\to\; C_{AB} + 2\,D_{\langle A}D_{B\rangle}f}$$

This infinite-dimensional family of vacua, labelled by the supertranslation-inequivalent shears $C_{AB}$, is the gravitational vacuum degeneracy underlying soft theorems and memory effects.

6. Superrotation Charge via Wald–Zoupas

The covariant phase space formalism begins with the pre-symplectic potential $\Theta$. For general relativity with Lagrangian $L = \frac{1}{16\pi G}\sqrt{-g}\,R$, a field variation $\delta g_{\mu\nu}$ yields:

$$\Theta^\mu[\delta g] = \frac{1}{16\pi G}\sqrt{-g}\left(g^{\alpha\beta}\delta\Gamma^\mu_{\alpha\beta} - g^{\mu\alpha}\delta\Gamma^\beta_{\alpha\beta}\right)$$

The symplectic current is $\omega = \delta_1\Theta[\delta_2 g] - \delta_2\Theta[\delta_1 g]$. Given a vector field $\xi$, the Noether current decomposes as:

$$J_\xi = \Theta[\mathcal{L}_\xi g] - i_\xi L = dQ_\xi + C_\xi$$

where $Q_\xi$ is the Noether charge 2-form (Wald charge) and $C_\xi$ vanishes on-shell. The Wald-Zoupas prescription corrects for the non-vanishing of $\Theta$ at $\mathscr{I}^+$ by subtracting a reference term:

$$\mathcal{H}_\xi = \oint_{S}\left(Q_\xi + i_\xi\Theta^{\rm ref}\right)$$

For a superrotation generated by $Y^A$, evaluating on a cross-section $S$ of $\mathscr{I}^+$ at retarded time $u$, the explicit result is:

$$\boxed{Q_Y = \frac{1}{8\pi G}\oint_{S^2}\!\left[Y^A N_A + \frac{u}{2}D_AY^B N^{AB} + \frac{1}{4}C_{AB}\!\left(2D^AY^B + D_CY^C\gamma^{AB}\right)\right]d^2\Omega}$$

For $Y^A$ restricted to $\ell=1$ conformal Killing vectors of $S^2$, $Q_Y$ reduces to the ADM angular momentum. The full Virasoro extension ($Y^A$ arbitrary) generates the superrotation Ward identity.

7. BMS Algebra with Explicit Commutators

Let $T_f$ denote the supertranslation generator for parameter $f$ and $R_Y$ the superrotation generator for $Y^A$. The BMS algebra is:

$$[T_{f_1}, T_{f_2}] = 0$$

Supertranslations commute (abelian ideal). For the mixed commutator, since $Y^A$ acts on functions via Lie derivative $\mathcal{L}_Y f = Y^A\partial_A f$:

$$[R_Y, T_f] = T_{\mathcal{L}_Y f - \frac{1}{2}(D_AY^A)f}$$

The extra $-\frac{1}{2}(D_AY^A)f$ arises from the conformal weight of $f$. For two superrotations:

$$[R_{Y_1}, R_{Y_2}] = R_{[Y_1,Y_2]_{\rm Lie}}$$

For the standard BMS group, $Y^A$ are the six conformal Killing vectors of $S^2$ (Lorentz algebra). In the extended BMS (Barnich-Troessaert), $Y^A$ are arbitrary, and using stereographic coordinates $z = e^{i\phi}\cot(\theta/2)$ the algebra becomes:

$$\mathfrak{bms}_4 = \mathfrak{supertrans} \;\rtimes\; \mathfrak{lorentz}$$

Extended BMS (Barnich–Troessaert):

$$\mathfrak{bms}_4^{\rm ext} = \mathfrak{supertrans} \;\rtimes\; (\mathfrak{Vir} \oplus \overline{\mathfrak{Vir}})$$

In mode notation with $L_n = R_{z^{n+1}\partial_z}$ and $T_{m,\bar{m}} = T_{z^m\bar{z}^{\bar{m}}}$, the Virasoro subalgebra reads $[L_m, L_n] = (m-n)L_{m+n}$ and $[L_n, T_{m,\bar{m}}] = \left(\frac{n+1}{2}-m\right)T_{m+n,\bar{m}}$.

Simulation: BMS Group Action on the Celestial Sphere

We visualise supertranslation modes $f = Y_{\ell m}$ as deformations of retarded time on $S^2$, and compute how the shear $C_{AB}$ transforms under the symmetric trace-free second derivative $2D_{\langle A}D_{B\rangle}f$.

BMS Supertranslation Modes and Shear Transformation

Python
script.py122 lines

Click Run to execute the Python code

Code will be executed with Python 3 on the server

Rate this chapter: