Derivation of the Spin Memory Effect

Celestial Sphere S2sourceCSv_A (velocity)DG after waveStokes Theorem:DG = |C DN_A dx^A= |S e^AB D_A DN_B d2SDN_A = angular momentum aspect change (spin memory)B-mode: curl component of the memory

1. The Observable: Velocity Circulation

Consider a ring of freely falling test particles on a closed curve $\mathcal{C}$ embedded in a 2-sphere at large radius $r \gg GM/c^2$ near future null infinity $\mathscr{I}^+$. Define the velocity circulation around this loop:

$$\Gamma(\mathcal{C}) = \oint_{\mathcal{C}} \dot{x}^A \, dl_A$$

where $\dot{x}^A$ is the transverse velocity of a particle (the angular components $A \in \{\theta, \phi\}$) and $dl_A$ is the line element along $\mathcal{C}$. By Stokes' theorem on $S^2$:

$$\Gamma(\mathcal{C}) = \oint_{\mathcal{C}} v_A \, dx^A = \int_{\Sigma} \epsilon^{AB} D_A v_B \, d^2\Sigma$$

The integrand $\epsilon^{AB} D_A v_B$ is the curl of the velocity field. This is precisely why the circulation detects the magnetic-parity (B-mode, odd-parity) component of the gravitational wave field. An E-mode (gradient) perturbation produces $v_A = D_A \Phi$ for some scalar $\Phi$, and $\epsilon^{AB} D_A D_B \Phi = 0$ identically on $S^2$, so the circulation vanishes for purely electric modes.

2. Geodesic Equation at O(1/r) in Bondi Coordinates

Work in Bondi coordinates $(u, r, x^A)$ where the metric takes the asymptotic form:

$$ds^2 = -du^2 - 2\,du\,dr + r^2 \gamma_{AB}\,dx^A dx^B + \frac{2m_B}{r}\,du^2 + r\,C_{AB}\,dx^A dx^B + D^B C_{AB}\,du\,dx^A + \ldots$$

The relevant Christoffel symbols at leading order in $1/r$ are:

$$\Gamma^A_{uu} = -\frac{1}{2r}\,\partial_u(D^B C_{AB}) + \mathcal{O}(r^{-2}), \qquad \Gamma^A_{uB} = \frac{1}{2}\,N^A{}_B + \mathcal{O}(r^{-1})$$

where $N_{AB} = \partial_u C_{AB}$ is the Bondi news tensor. For a particle initially at rest ($\dot{r} = 0$, $\dot{x}^A = 0$ at early times), the geodesic equation$\ddot{x}^\mu + \Gamma^\mu_{\alpha\beta}\dot{x}^\alpha\dot{x}^\beta = 0$ gives, at leading order with $\dot{u} \approx 1$:

$$\ddot{x}^A + \Gamma^A_{uu} + 2\,\Gamma^A_{uB}\,\dot{x}^B = 0$$

Writing $v_A = g_{AB}\dot{x}^B$ for the covariant transverse velocity and keeping only $\mathcal{O}(1/r)$ terms:

$$\partial_u v_A = \frac{1}{2}\,N_{AB}\,v^B + \frac{1}{2}\,\partial_u(D^B C_{BA}) + \mathcal{O}(r^{-2})$$

The first term is the news-velocity coupling (linear in $v^B$), and the second is the metric drift from the $g_{uA}$ component at $\mathcal{O}(1/r)$. Identifying$g_{uA}|_{1/r} = \frac{1}{2}D^B C_{BA}$, we can write:

$$\partial_u v_A = \frac{1}{2}\,N_{AB}\,v^B + \partial_u\!\left(g_{uA}\big|_{\mathcal{O}(r^{-1})}\right)$$

3. Integration from $u = -\infty$ to $u = +\infty$

Integrate both sides over the full burst duration. The boundary conditions are:$v_A(-\infty) = 0$ (particles at rest before the wave), and$C_{AB}(-\infty) = C_{AB}^{(-)}$, $C_{AB}(+\infty) = C_{AB}^{(+)}$. Thus:

$$v_A(+\infty) = \frac{1}{2}\int_{-\infty}^{+\infty} N_{AB}\,v^B\,du + \frac{1}{2}\Delta(D^B C_{BA})$$

where $\Delta(D^B C_{BA}) = D^B C_{BA}^{(+)} - D^B C_{BA}^{(-)}$. The first integral is nonlinear (it involves $v^B$, which itself evolves), so we iterate: at leading order,$v^B$ is $\mathcal{O}(1/r)$, and $N_{AB}$ is also $\mathcal{O}(1/r)$, so the integral is $\mathcal{O}(1/r^2)$. The dominant contribution to the transverse velocity is:

$$v_A^{(1)}(+\infty) = \frac{1}{2}\,\Delta(D^B C_{BA})$$

However, the angular momentum aspect $N_A$ also contributes at $\mathcal{O}(1/r)$ through the subleading metric. Carefully tracking all $\mathcal{O}(1/r)$ terms (Pasterski, Strominger, Zhiboedov 2016):

$$v_A(+\infty) = \frac{1}{2r}\,\Delta(D^B C_{BA}) + \frac{1}{3r}\,\Delta N_A + \mathcal{O}(r^{-2})$$

4. Applying the Curl Operator

Apply $\epsilon^{AB}D_A$ to extract the magnetic-parity part. Consider each term separately.

First term: $\epsilon^{AB}D_A \Delta(D^C C_{CB})$. Since $C_{AB}$ is symmetric trace-free (STF), decompose it into E and B modes:

$$C_{AB} = \left(D_A D_B - \frac{1}{2}\gamma_{AB} D^2\right)\Phi_E + \epsilon_{C(A}D_{B)}D^C \Phi_B$$

For the E-mode part, $D^C C_{CB}^{(E)}$ is a gradient on $S^2$, so:

$$\epsilon^{AB}D_A D^C C_{CB}^{(E)} = \epsilon^{AB}D_A \partial_B(\text{scalar}) = 0$$

This is the crucial identity: the curl of any gradient vanishes on $S^2$. More precisely, for any smooth scalar $f$ on $S^2$:

$$\epsilon^{AB}D_A D_B f = 0$$

because $D_A D_B f - D_B D_A f = 0$ for scalars on a 2D manifold (the Riemann tensor contribution$R^C{}_{DAB} f$ vanishes when contracted with $\epsilon^{AB}$ due to the symmetry$R_{CDAB} = K(\gamma_{CA}\gamma_{DB} - \gamma_{CB}\gamma_{DA})$ on $S^2$).

Therefore: the shear boundary term $\Delta(D^B C_{BA})$ contributes only through its B-mode part, which is subdominant. The dominant B-mode contribution comes from $\Delta N_A$:

$$\epsilon^{AB}D_A v_B\big|_{+\infty} = \frac{2}{3r}\,\epsilon^{AB}D_A \Delta N_B$$

The factor $2/3$ arises from the precise coefficient of $\Delta N_A$ in the subleading soft expansion (related to the soft graviton theorem at sub-subleading order).

5. The Spin Memory Formula (Stokes' Theorem)

Now apply Stokes' theorem on $S^2$ to convert the line integral to an area integral. For the circulation around $\mathcal{C} = \partial\Sigma$:

$$\Delta\Gamma(\mathcal{C}) = \oint_{\mathcal{C}} v_A(+\infty)\,dx^A = \frac{2}{3r}\oint_{\mathcal{C}} \Delta N_A\,dx^A$$

Using the generalized Stokes theorem $\oint_{\partial\Sigma} \omega = \int_\Sigma d\omega$, where $\omega = \Delta N_A\,dx^A$ is a 1-form on $S^2$:

$$(d\omega)_{AB} = D_A(\Delta N_B) - D_B(\Delta N_A) = \epsilon_{AB}\,\epsilon^{CD}D_C \Delta N_D$$

So the area integral becomes:

$$\boxed{\Delta\Gamma(\mathcal{C}) = \frac{2}{3r}\int_\Sigma \epsilon^{AB}D_A \Delta N_B \, d^2\Sigma}$$

This is the spin memory effect: the permanent change in the velocity circulation is determined by the curl of the change in the angular momentum aspect $\Delta N_A$.

6. Deriving $\Delta N_A$ from the News Tensor

The evolution equation for the angular momentum aspect follows from the Bianchi identity applied to the asymptotic Einstein equations. At $\mathscr{I}^+$:

$$\partial_u N_A = \frac{1}{4}\,D_A\!\left(N_{BC}N^{BC}\right) - D^B(N_{BA}m_B) - \frac{1}{2}\,D_A\!\left(C^{BC}N_{BC}\right) + \frac{1}{2}\,N_{BC}\,D_A C^{BC} - 4\pi G\,T_{uA}^{\rm matter}$$

Integrate from $u = -\infty$ to $u = +\infty$. The key nonlinear term is:

$$\Delta N_A = \frac{1}{4}\int_{-\infty}^{+\infty} D_A(N_{BC}N^{BC})\,du - \int_{-\infty}^{+\infty} D^B(N_{BA}m_B)\,du + \ldots$$

The term $N_{BC}N^{BC}$ is the GW energy flux density (the same quantity that drives Christodoulou displacement memory). When we take $\epsilon^{AB}D_A$ of $\Delta N_B$, the gradient terms $D_A(\text{scalar})$ vanish, and only the curl part survives. This picks out the angular momentum flux from the nonlinear terms $D^B(N_{BA}m_B)$ and the matter flux $T_{uA}^{\rm matter}$:

$$\epsilon^{AB}D_A \Delta N_B = -\epsilon^{AB}\int_{-\infty}^{+\infty} D^C(N_{CA})m_B\,du - 4\pi G\,\epsilon^{AB}\int_{-\infty}^{+\infty} T_{uA}^{\rm matter}\,du$$

7. E/B Mode Decomposition: Why Displacement = E, Spin = B

Any symmetric trace-free tensor $H_{AB}$ on $S^2$ admits a unique decomposition:

$$H_{AB} = \left(D_A D_B - \frac{1}{2}\gamma_{AB}D^2\right)\Phi_E + \frac{1}{2}\left(\epsilon_{AC}D^C D_B + \epsilon_{BC}D^C D_A\right)\Phi_B$$

Displacement memory measures $\Delta C_{AB}$, the permanent change in the shear. Christodoulou's result shows $\Delta C_{AB}$ is sourced by the energy flux$\int N^{CD}N_{CD}\,du$, which is a scalar on $S^2$. Its STF Hessian is pure E-mode:

$$\Delta C_{AB}^{\rm disp} = \left(D_A D_B - \frac{1}{2}\gamma_{AB}D^2\right)\Phi_E \quad \Longleftrightarrow \quad \text{E-mode (even parity)}$$

Spin memory measures $\epsilon^{AB}D_A \Delta N_B$, the curl of the angular momentum change. This is a pseudoscalar on $S^2$ (odd under parity), and therefore:

$$\underbrace{\Delta\xi^i \sim \Delta C_{AB}}_{\text{even parity, E-mode}} \quad \longleftrightarrow \quad \underbrace{\Delta\Gamma \sim \epsilon^{AB}D_A \Delta N_B}_{\text{odd parity, B-mode}}$$

In the spin-weighted spherical harmonic decomposition with $\,_{-2}Y_{lm}$, the E-mode gives$a_{lm}^E \neq 0$, $a_{lm}^B = 0$, while spin memory gives $a_{lm}^B \neq 0$. This is the gravitational analog of the CMB E/B decomposition.

Spin Memory from a Precessing Binary Merger: Skymap and E/B Comparison

Python
script.py169 lines

Click Run to execute the Python code

Code will be executed with Python 3 on the server

Rate this chapter: