Important Equations
The key equations of the masterclass, collected unit by unit. Each one is derived or motivated in its unit, with worked examples and the conditions under which it holds.
Relativity is a theory with few equations and many consequences. The Lorentz factor, the Lorentz transformation, the invariant interval, and the energy-momentum relation carry the whole of the special theory; the geodesic equation, the field equations, and the Schwarzschild metric carry the general theory; and the quadrupole formula and the chirp mass carry what the detectors hear. This page collects those equations in the order of the units, each with a note on what its symbols mean and under which conditions it holds, for example whether a formula is exact or the weak-field form.
The sheet is a reference for someone who has read the units, so it gives results and never the steps between them. The speed of light and the gravitational constant stay in every equation, and the sign convention follows the tier: Units 1 to 8 write the interval as the proper time, Units 9 to 16 use the line element with signature (−,+,+,+). The page grows as units are published, so an equation you cannot find yet belongs to a unit that is still to come.
Frames of Reference and the Physicist's Toolkit
Kinetic energy and the work-energy theorem
\[ K = \tfrac{1}{2}mv^2 , \qquad W_{\mathrm{net}} = K_2 - K_1 \] The kinetic energy \(K\) of a body of mass \(m\) moving at speed \(v\), and the work-energy theorem: the net work done on a body equals the change in its kinetic energy. The expression for \(K\) holds for speeds small compared with the speed of light; the theorem itself survives in relativity once \(K\) is replaced by the expression that Unit 7 derives.
Newtonian gravity
\[ F = \frac{GMm}{r^2} , \qquad U = -\frac{GMm}{r} \] The attraction \(F\) between two masses \(M\) and \(m\) whose centres are \(r\) apart, and the potential energy \(U\) of the pair, counted from zero at infinite separation. Both hold outside the bodies. Near the surface of a planet of radius \(R\), for heights \(h\) small compared with \(R\), the difference in \(U\) between two points \(h\) apart reduces to \(mgh\) with \(g = GM/R^2\).
The Galilean transformation
\[ x' = x - vt , \qquad y' = y , \qquad z' = z , \qquad t' = t \] The coordinates \((t', x', y', z')\) of an event in the frame \(S'\) in terms of its coordinates \((t, x, y, z)\) in the frame \(S\), when \(S'\) moves at the velocity \(v\) along the \(x\) axis and the origins coincide at \(t = 0\). The inverse is obtained by changing the sign of \(v\). The last equation, \(t' = t\), is the assumption of a universal time, and it is the one that Unit 4 will replace.
The Galilean addition of velocities
\[ u = u' + v \] A body moving at the velocity \(u'\) in the frame \(S'\) moves at the velocity \(u\) in the frame \(S\), where \(v\) is the velocity of \(S'\) relative to \(S\). Velocities along the \(x\) axis carry their signs; for velocities in any direction the vectors add, \(\vec{u} = \vec{u}' + \vec{v}\). The rule holds for speeds small compared with the speed of light.
The binomial approximation
\[ (1 + x)^n \approx 1 + nx \qquad \text{for } |x| \ll 1 \] Valid for any power \(n\), positive, negative, or fractional. The error is about \(\tfrac{1}{2}n(n - 1)x^2\). In relativity \(x\) is usually \(\pm v^2/c^2\) and \(n\) is \(\pm\tfrac{1}{2}\), so \(\sqrt{1 - v^2/c^2} \approx 1 - \tfrac{1}{2}v^2/c^2\) and \(1/\sqrt{1 - v^2/c^2} \approx 1 + \tfrac{1}{2}v^2/c^2\).
Light travel time
\[ t = \frac{d}{c} , \qquad d = c\,t \] The time \(t\) that light needs to cover the distance \(d\) in vacuum, and the distance light covers in the time \(t\), with \(c = 299{,}792{,}458\unit{m\,s^{-1}}\). A light-year is the distance \(d\) for \(t\) equal to one year.
Light, the Ether, and the Two Postulates
The speed of light in vacuum
\[ c = 299{,}792{,}458\unit{m\,s^{-1}} \qquad \text{(exact since 1983)} \] Light in vacuum covers 299,792 km in one second and about 30 centimetres in one nanosecond. In estimates this unit uses \(c \approx 3.00 \times 10^{8}\unit{m\,s^{-1}}\). In air light is slower by 0.03 per cent, and in water by a quarter.
Maxwell’s prediction
\[ c = \frac{1}{\sqrt{\varepsilon_0\,\mu_0}} \] The electric constant \(\varepsilon_0\) (the permittivity of free space) sets the force between charges and the magnetic constant \(\mu_0\) (the permeability of free space) sets the force between currents. Their product fixes the speed of an electromagnetic wave in vacuum. With the values measured by Wilhelm Weber (1804-1891) and Rudolf Kohlrausch (1809-1858) in 1856 the formula gives \(3.1 \times 10^{8}\unit{m\,s^{-1}}\), which Maxwell compared with Fizeau’s \(3.15 \times 10^{8}\); the two agreed to within two per cent, far too close to be a coincidence.
The expected fringe shift
\[ \Delta t \approx \frac{L v^2}{c^3} , \qquad N = \frac{2 L v^2}{\lambda\,c^2} \] \(\Delta t\) is the difference between the round-trip times along an arm of length \(L\) parallel and perpendicular to an ether wind of speed \(v\), and \(N\) is the shift of the fringes, in fringe spacings, when the instrument is rotated by 90 degrees; \(\lambda\) is the wavelength of the light. Both hold for \(v \ll c\).
Einstein synchronization
\[ t_B = t_A + \frac{L}{c} \] A light signal that leaves clock \(A\) when it reads \(t_A\) reaches clock \(B\), at rest relative to \(A\) at the distance \(L\), when \(B\) should read \(t_B\). Two clocks set this way are synchronized in the frame in which both are at rest. Equivalently, a signal sent from \(A\) to \(B\) and reflected back takes equal times out and back.
Simultaneity, Time Dilation, and Length Contraction
Relativity of simultaneity
Two events that are simultaneous in one inertial frame are not simultaneous in a second inertial frame that moves along the line joining them. Events that happen at the same place are simultaneous for everyone or for no one. The disagreement grows with the speed of the second frame and with the distance between the events, and it is far too small to notice at everyday speeds.
The Lorentz factor
\[ \gamma = \frac{1}{\sqrt{1 - v^2/c^2}} \] The Lorentz factor of a speed \(v\) is 1 at rest, stays within half a per cent of 1 up to a tenth of the speed of light, and grows without limit as \(v\) approaches \(c\). It is never smaller than 1.
Time dilation
\[ \Delta t = \gamma\,\Delta\tau \] A clock that moves at speed \(v\) through an inertial frame shows the time \(\Delta\tau\) between two of its ticks, while the synchronized clocks of that frame measure the longer time \(\Delta t\) between the same two ticks. Moving clocks run slow by the Lorentz factor.
Length contraction
\[ L = \frac{L_0}{\gamma} = L_0\sqrt{1 - \frac{v^2}{c^2}} \] A rod of proper length \(L_0\), its length measured in the frame in which it is at rest, has the shorter length \(L\) in a frame through which it moves at speed \(v\) along its own direction. Lengths at right angles to the motion do not change.
The Lorentz Transformation and the Addition of Velocities
The Lorentz transformation
\[ x' = \gamma\,(x - vt) , \qquad y' = y , \qquad z' = z , \qquad t' = \gamma\left(t - \frac{vx}{c^2}\right) , \qquad \gamma = \frac{1}{\sqrt{1 - v^2/c^2}} \] The coordinates \((t', x')\) of an event in the frame \(S'\), which moves at the speed \(v\) along the \(x\) axis of the frame \(S\) in the standard configuration, in terms of its coordinates \((t, x)\) in \(S\). The transformation holds for any pair of inertial frames and any speed below \(c\).
The inverse Lorentz transformation
\[ x = \gamma\,(x' + vt') , \qquad y = y' , \qquad z = z' , \qquad t = \gamma\left(t' + \frac{vx'}{c^2}\right) \] The coordinates in \(S\) in terms of those in \(S'\). The equations are the Lorentz transformation with the primed and unprimed coordinates exchanged and \(v\) replaced by \(-v\), because from \(S'\) the frame \(S\) moves backwards at the speed \(v\).
The velocity-addition formula
\[ u = \frac{u' + v}{1 + u'v/c^2} \] The velocity \(u\) in \(S\) of an object that moves at the velocity \(u'\) along the \(x'\) axis of \(S'\), where \(S'\) moves at \(v\) along the \(x\) axis of \(S\). Velocities against the direction of \(v\) are entered as negative numbers. For \(u'\) and \(v\) both much smaller than \(c\) the denominator is 1 and the Galilean sum \(u = u' + v\) returns.
Velocity across the motion
\[ u_y = \frac{u'_y}{\gamma\,(1 + u'_x v/c^2)} , \qquad u_z = \frac{u'_z}{\gamma\,(1 + u'_x v/c^2)} \] The components of the velocity at right angles to \(v\), where \(u'_x\), \(u'_y\), and \(u'_z\) are the components measured in \(S'\). An object that moves straight across \(S'\) at \(u'_y\) with \(u'_x = 0\) moves across \(S\) at the smaller \(u'_y/\gamma\), while it also drifts along \(x\) at \(v\).
The Fizeau drag
\[ u \approx \frac{c}{n} + v\left(1 - \frac{1}{n^2}\right) \] The speed of light in the laboratory when it travels through a medium of refractive index \(n\) that moves at the speed \(v\) along the light’s path, to first order in \(v/c\). The factor \(1 - 1/n^2\) is Fresnel’s drag coefficient, here a consequence of velocity addition. For light against the flow, \(v\) is negative.
The invariant combination
\[ c^2t'^2 - x'^2 = c^2t^2 - x^2 \] The difference between the square of the light travel distance \(ct\) and the square of the position \(x\) of an event has the same value in every inertial frame. For a light flash from the origin it is zero in every frame, which is the second postulate once more; for two events it applies to the differences \(c^2\Delta t^2 - \Delta x^2\).
Spacetime Diagrams and the Invariant Interval
The axes of a moving frame
\[ \tan\alpha = \frac{v}{c} \] In the spacetime diagram of \(S\), the \(ct'\) axis of a frame \(S'\) moving at \(v\) along \(x\) is tilted from the \(ct\) axis by the angle \(\alpha\) towards the light line, and the \(x'\) axis is tilted from the \(x\) axis by the same angle towards the light line. The lines of simultaneity of \(S'\) are parallel to its \(x'\) axis. For a frame moving in the negative \(x\) direction, \(v\) is negative and both axes tilt the other way, towards the light line \(x = -ct\).
The invariant interval
\[ c^2\Delta\tau^2 = c^2\Delta t^2 - \Delta x^2 - \Delta y^2 - \Delta z^2 \] The interval between two events separated by \(\Delta t\) in time and by \(\Delta x\), \(\Delta y\), \(\Delta z\) in space has the same value in every inertial frame. When it is positive, \(\Delta\tau\) is the proper time between the events, the time read by a clock that is present at both. When it is negative, \(\sqrt{-c^2\Delta\tau^2}\) is the proper length between them, the distance measured in the frame in which they are simultaneous. When it is zero, the events can be joined by a light flash.
Proper time along a world line
\[ \Delta\tau = \sum_i \sqrt{\Delta t_i^2 - \frac{\Delta x_i^2}{c^2}} = \sum_i \frac{\Delta t_i}{\gamma_i} \] The time read by a clock that follows a world line made of straight pieces, each covered at the speed \(v_i\) with the Lorentz factor \(\gamma_i\) during the coordinate time \(\Delta t_i\). Every piece contributes less than its coordinate time, so between two given events the straight world line, the one along which a clock is at rest in some inertial frame, has the longest proper time. Unit 6 writes the sum as an integral for a smoothly varying speed.
The relativistic Doppler formula
\[ f = f_0\sqrt{\frac{1 - v/c}{1 + v/c}} \quad\text{(receding)} , \qquad f = f_0\sqrt{\frac{1 + v/c}{1 - v/c}} \quad\text{(approaching)} , \qquad f = \frac{f_0}{\gamma} \quad\text{(transverse)} \] The frequency \(f\) received from a source that emits the frequency \(f_0\) in its own frame and moves at the speed \(v\) directly away from the observer, directly towards the observer, or across the line of sight. Wavelengths change by the inverse factors. The transverse formula holds for light that reaches the observer at right angles to the motion in the observer’s frame, and it is time dilation alone. For speeds much smaller than \(c\) the first two reduce to \(f \approx f_0\,(1 \mp v/c)\), with the minus sign for a receding and the plus sign for an approaching source, which is the formula astronomers use; the relativistic correction is proportional to \(v^2/c^2\).
The Twin Paradox and Accelerated Motion
Proper time along a world line
\[ \Delta\tau = \int \sqrt{1 - \frac{v^2}{c^2}}\;\dd t \] The time shown by a clock that moves with the speed \(v(t)\) in an inertial frame, between two events on its world line, with \(t\) the coordinate time of that frame. For a piece of the world line at constant speed the integral is \(\Delta t/\gamma\). Of all world lines between two events, the straight one, which is the world line of a clock moving at constant velocity, has the largest proper time.
The twin result
\[ \Delta\tau_{\mathrm{traveller}} = \frac{\Delta t_{\mathrm{home}}}{\gamma} = \Delta t_{\mathrm{home}} \sqrt{1 - \frac{v^2}{c^2}} \] For a trip at the constant speed \(v\) to a destination and back at the same speed, where \(\Delta t_{\mathrm{home}}\) is the time between the departure and the reunion on the home clock, which stays in one inertial frame, and \(\Delta\tau_{\mathrm{traveller}}\) is the time on the clock that makes the trip. The turnaround is taken to be short compared with the legs.
Hyperbolic motion
\[ v = c\tanh\frac{a\tau}{c} , \qquad x = \frac{c^2}{a}\left[\cosh\frac{a\tau}{c} - 1\right] , \qquad t = \frac{c}{a}\sinh\frac{a\tau}{c} \] The speed, the distance covered, and the elapsed time in the inertial frame in which a ship started from rest, after the proper time \(\tau\) on board, for a constant proper acceleration \(a\). The speed approaches \(c\) but never reaches it, and the world line is the hyperbola \((x + c^2/a)^2 - c^2 t^2 = c^4/a^2\), or \(x^2 - c^2t^2 = c^4/a^2\) with \(x\) measured from a point \(c^2/a\) behind the start. For \(a\tau \ll c\) the three formulas reduce to Newton’s \(v = a\tau\), \(x = \tfrac12 a\tau^2\), and \(t = \tau\).
The Rindler horizon
\[ d_{\mathrm{horizon}} = \frac{c^2}{a} \] The distance behind a ship of constant proper acceleration \(a\), measured in its momentarily comoving frame, at which the horizon lies; no signal from an event behind it ever reaches the ship. For \(a = g\) the horizon is \(0.968\) light-years behind the ship. A signal from the starting point sent later than \(c/a\) after the start, \(0.968\) years for \(a = g\), never arrives as long as the acceleration lasts.
Relativistic Energy, Momentum, and Mass-Energy Equivalence
Relativistic momentum
\[ \vec{p} = \gamma\,m\,\vec{v} = \frac{m\,\vec{v}}{\sqrt{1 - v^2/c^2}} \] The momentum of a body of mass \(m\) moving at the velocity \(\vec{v}\), with \(\gamma\) the Lorentz factor of its speed. At everyday speeds \(\gamma = 1\) and the formula turns back into Newton’s \(m\vec{v}\). It is this momentum, summed over all bodies, that is conserved in every inertial frame.
Kinetic energy and total energy
\[ K = (\gamma - 1)\,mc^2 , \qquad E = \gamma\,mc^2 = K + mc^2 \] The kinetic energy \(K\) of a body of mass \(m\) with the Lorentz factor \(\gamma\), and its total energy \(E\), which is the kinetic energy plus the term \(mc^2\) that remains when the body is at rest. Both formulas hold at every speed below \(c\).
Mass-energy equivalence
\[ E_0 = mc^2 \] The rest energy \(E_0\) of a body of mass \(m\). The conversion factor \(c^2 = 8.988 \times 10^{16}\unit{m^2\,s^{-2}}\) is so large that one kilogram of mass corresponds to \(8.99 \times 10^{16}\unit{J}\).
The energy-momentum relation
\[ E^2 = p^2c^2 + m^2c^4 , \qquad\text{and for a massless particle}\qquad E = pc \] The relation between the total energy \(E\), the magnitude \(p\) of the momentum, and the mass \(m\) of any particle, in any inertial frame. For a particle at rest it reduces to \(E = mc^2\); for a particle of zero mass the energy is \(c\) times the momentum. The speed follows from \(v/c = pc/E\).
The invariant mass of a system
\[ M^2c^4 = E_{\mathrm{tot}}^2 - p_{\mathrm{tot}}^2c^2 \] The invariant mass \(M\) of a collection of particles, computed from the sum \(E_{\mathrm{tot}}\) of their energies and the vector sum \(\vec{p}_{\mathrm{tot}}\) of their momenta, with \(p_{\mathrm{tot}}\) the magnitude of that vector. It is the same in every frame and equals the mass a single body would have if all the particles merged into it. It is conserved in every collision and decay, and it is not the sum of the masses of the particles.
The Equivalence Principle and Curved Spacetime
Gravitational frequency shift and clock rates
\[ \frac{\Delta f}{f} = \frac{gh}{c^2} = \frac{\Delta\Phi}{c^2} , \qquad \frac{\text{rate of the upper clock}}{\text{rate of the lower clock}} = 1 + \frac{\Delta\Phi}{c^2} \] Light that climbs through a difference \(\Delta\Phi = gh\) in gravitational potential arrives with its frequency lowered by the fraction \(\Delta f/f\), and a clock at the upper place runs faster than a clock at the lower place by the same fraction. On the Earth the fraction is \(1.09 \times 10^{-16}\) per metre of height. Both statements are the weak-field form, valid when \(|\Delta\Phi| \ll c^2\).
Rate of a satellite clock
\[ \frac{\Delta\tau}{\Delta t} \approx 1 - \frac{v^2}{2c^2} + \frac{\Phi_{\mathrm{sat}} - \Phi_{\mathrm{ground}}}{c^2} \] The proper time \(\Delta\tau\) that elapses on a satellite clock per time \(\Delta t\) on a clock at rest on the ground. The second term is the velocity time dilation of Unit 3 in the binomial approximation, the third is the gravitational shift of this unit with \(\Phi = -G\Mearth/r\); the ground clock’s own motion with the Earth’s rotation is neglected here. Valid for \(v \ll c\) and \(|\Phi| \ll c^2\).
Deflection of light at the Sun
\[ \alpha_{1911} = \frac{2G\Msun}{c^2 \Rsun} = 0.875\unit{arcsec} , \qquad \alpha_{1915} = \frac{4G\Msun}{c^2 \Rsun} = 1.75\unit{arcsec} \] The angle by which a ray of starlight that grazes the edge of the Sun is bent towards it: the value of the equivalence principle alone, from the Newtonian fall of a particle at speed \(c\), and the value of the full theory, which is twice as large and is the measured one. For a ray that passes the Sun at a larger distance \(b\), replace \(\Rsun\) by \(b\).
Tensors, Metrics, and Geodesics
The metric of a sphere
\[ \dd s^2 = R^2\,\dd\theta^2 + R^2\sin^2\theta\,\dd\phi^2 , \qquad L = \int \dd s \quad\text{along the path} \] The distance \(\dd s\) between two neighbouring points on a sphere of radius \(R\), in terms of their difference \(\dd\theta\) in colatitude and \(\dd\phi\) in longitude. The length \(L\) of any path on the sphere is the integral of \(\dd s\) along it.
The line element and the metric
\[ \dd s^2 = g_{\mu\nu}\,\dd x^{\mu}\dd x^{\nu} , \qquad v_{\mu} = g_{\mu\nu}\,v^{\nu} , \qquad v^{\mu} = g^{\mu\nu}\,v_{\nu} \] The metric \(g_{\mu\nu}\) turns coordinate steps \(\dd x^{\mu}\) into the distance or interval \(\dd s\); repeated indices are summed. The metric lowers an index and its inverse \(g^{\mu\nu}\) raises one. Under a change from the coordinates \(x^{\mu}\) to new coordinates \(x'^{\alpha}\) the metric components change as \(g'_{\alpha\beta} = \dfrac{\partial x^{\mu}}{\partial x'^{\alpha}}\dfrac{\partial x^{\nu}}{\partial x'^{\beta}}\,g_{\mu\nu}\), so that \(\dd s^2\) stays the same; this is the chain rule applied to both factors of \(\dd x^{\mu}\dd x^{\nu}\).
Flat spacetime and the four-velocity
\[ \dd s^2 = \eta_{\mu\nu}\,\dd x^{\mu}\dd x^{\nu} = -c^2\,\dd t^2 + \dd x^2 + \dd y^2 + \dd z^2 , \qquad u^{\mu}u_{\mu} = \eta_{\mu\nu}u^{\mu}u^{\nu} = -c^2 \] The line element of flat spacetime in an inertial frame, with the signature \((-,+,+,+)\), and the norm of the four-velocity \(u^{\mu} = \dd x^{\mu}/\dd\tau\) of any body with mass, the norm of a vector being its contraction \(u^{\mu}u_{\mu}\) with itself. The norm is the same number in every frame; it is the interval statement \(\dd s^2 = -c^2\dd\tau^2\) divided by \(\dd\tau^2\).
The geodesic equation and the Christoffel symbols
\[ \frac{\dd^2 x^{\lambda}}{\dd\tau^2} + \Gamma^{\lambda}_{\mu\nu}\,\frac{\dd x^{\mu}}{\dd\tau}\frac{\dd x^{\nu}}{\dd\tau} = 0 , \qquad \Gamma^{\lambda}_{\mu\nu} = \tfrac12\,g^{\lambda\sigma}\left(\partial_{\mu}g_{\sigma\nu} + \partial_{\nu}g_{\sigma\mu} - \partial_{\sigma}g_{\mu\nu}\right) \] A geodesic \(x^{\lambda}(\tau)\) satisfies one such equation for each coordinate, with \(\tau\) the proper time along a timelike world line, or the distance \(s\) along a path on a surface. Along a light ray no proper time passes, and \(\tau\) is then replaced by any parameter for which the equation keeps this form, called an affine parameter. The Christoffel symbols \(\Gamma^{\lambda}_{\mu\nu}\), symmetric in \(\mu\) and \(\nu\), are built from the first partial derivatives of the metric, \(\partial_{\mu}g_{\sigma\nu} = \partial g_{\sigma\nu}/\partial x^{\mu}\). In spacetime there are 40 of them, four choices of the upper index times the ten different pairs of lower indices.
The Riemann tensor, the Ricci tensor, and the Ricci scalar
\[ R^{\rho}{}_{\sigma\mu\nu} = \partial_{\mu}\Gamma^{\rho}_{\nu\sigma} - \partial_{\nu}\Gamma^{\rho}_{\mu\sigma} + \Gamma^{\rho}_{\mu\lambda}\Gamma^{\lambda}_{\nu\sigma} - \Gamma^{\rho}_{\nu\lambda}\Gamma^{\lambda}_{\mu\sigma} , \qquad R_{\sigma\nu} = R^{\rho}{}_{\sigma\rho\nu} , \qquad R = g^{\sigma\nu}R_{\sigma\nu} \] The Riemann tensor measures by how much a vector changes when it is parallel transported around a small loop; it vanishes everywhere if and only if the geometry is flat. The Ricci tensor is its contraction over the first and third indices, and the Ricci scalar is the contraction of the Ricci tensor with the inverse metric. On a two-dimensional surface the Ricci scalar is twice the Gaussian curvature, \(R = 2/R^2\) for a sphere of radius \(R\).
Geodesic deviation
\[ \frac{\dd^2\xi^{\mu}}{\dd\tau^2} = -R^{\mu}{}_{\nu\rho\sigma}\,u^{\nu}\xi^{\rho}u^{\sigma} , \qquad \text{in one dimension:}\quad \frac{\dd^2\xi}{\dd s^2} = -K\,\xi \] The separation vector \(\xi^{\mu}\) between two neighbouring geodesics with the common four-velocity \(u^{\mu}\) changes at a rate set by the Riemann tensor; the derivative is taken along the geodesic with the Christoffel corrections of the covariant derivative. On a surface of Gaussian curvature \(K\) the separation of two neighbouring geodesics obeys the one-dimensional form, where \(s\) is the distance along them.
Einstein's Field Equations
Poisson’s equation
\[ \nabla^2\Phi = \frac{\partial^2\Phi}{\partial x^2} + \frac{\partial^2\Phi}{\partial y^2} + \frac{\partial^2\Phi}{\partial z^2} = 4\pi G\rho \] The gravitational potential \(\Phi\) at a point has second derivatives whose sum, the Laplacian \(\nabla^2\Phi\), equals \(4\pi G\) times the mass density \(\rho\) at that point. Where there is no matter the right side is zero and the equation is called Laplace’s equation.
Stress-energy tensor of a perfect fluid
\[ T^{\mu\nu} = \left(\rho + \frac{p}{c^2}\right)u^{\mu}u^{\nu} + p\,g^{\mu\nu} \] \(\rho\) is the mass density measured in the fluid’s own rest frame (its energy density there divided by \(c^2\)), \(p\) is its pressure, \(u^{\mu}\) its four-velocity, and \(g^{\mu\nu}\) the inverse metric of Unit 9. In the rest frame of the fluid, where \(u^{\mu} = (c, 0, 0, 0)\) and \(g^{\mu\nu}\) is \(\eta^{\mu\nu} = \mathrm{diag}(-1, 1, 1, 1)\), the components are \(T^{00} = \rho c^2\), \(T^{11} = T^{22} = T^{33} = p\), and all others are zero. Dust is the case \(p = 0\). The pressure appears twice so that both rest-frame values come out right: in \(T^{00}\) the first term gives \((\rho + p/c^2)c^2\) and the second gives \(-p\), which cancel to \(\rho c^2\), while in \(T^{11}\) the first term is zero and the second gives \(p\).
The Einstein tensor
\[ G_{\mu\nu} = R_{\mu\nu} - \tfrac12\,R\,g_{\mu\nu} , \qquad \nabla^{\mu}G_{\mu\nu} = 0 \] The Ricci tensor \(R_{\mu\nu}\) and the Ricci scalar \(R\) are the contractions of the Riemann tensor of Unit 9. The Einstein tensor \(G_{\mu\nu}\) is symmetric, contains the metric and its first and second derivatives, is linear in the second derivatives, and has a vanishing covariant divergence for every metric, by the contracted Bianchi identity.
Einstein’s field equations
\[ G_{\mu\nu} + \Lambda\,g_{\mu\nu} = \frac{8\pi G}{c^4}\,T_{\mu\nu} \] The curvature of spacetime at a point, in the combination \(G_{\mu\nu} = R_{\mu\nu} - \tfrac12 R\,g_{\mu\nu}\), is proportional to the stress-energy tensor of the matter and energy at that point; \(G\) is Newton’s constant, \(c\) the speed of light, and \(\Lambda\) the cosmological constant, which is zero in Units 11 and 12 and in every later unit of this program. The constant \(8\pi G/c^4\) is fixed in Section 4. For \(\Lambda = 0\) the equations can be written with the trace \(T = g^{\mu\nu}T_{\mu\nu}\) in the equivalent trace-reversed form (contracting both sides with \(g^{\mu\nu}\) gives \(R = -(8\pi G/c^4)\,T\), and putting this back in place of \(R\) gives) \[ R_{\mu\nu} = \frac{8\pi G}{c^4}\left(T_{\mu\nu} - \tfrac12\,T\,g_{\mu\nu}\right) , \] which shows that in empty space, where \(T_{\mu\nu} = 0\), the Ricci tensor vanishes: \(R_{\mu\nu} = 0\).
The weak-field metric
\[ g_{00} = -\left(1 + \frac{2\Phi}{c^2}\right) , \qquad \frac{\dd\tau}{\dd t} = \sqrt{-g_{00}} \approx 1 + \frac{\Phi}{c^2} \] In a weak, static gravitational field with Newtonian potential \(\Phi\), the time-time component of the metric differs from its flat value \(-1\) by \(2\Phi/c^2\), and a clock at rest there runs slow against the coordinate time \(t\), which is the time of a clock far away, by the fraction \(|\Phi|/c^2\) (\(\Phi\) is negative, so \(\dd\tau/\dd t\) is less than one). The approximation needs \(|2\Phi/c^2| \ll 1\). The space components of the metric are also changed, by a comparable amount, but a slow body does not notice them; light does, which is the factor of two of Unit 8.
The Schwarzschild Solution and the Classical Tests
The Schwarzschild metric
\[ \dd s^2 = -\left(1 - \frac{r_{\mathrm{s}}}{r}\right)c^2\,\dd t^2 + \left(1 - \frac{r_{\mathrm{s}}}{r}\right)^{-1}\dd r^2 + r^2\,\dd\theta^2 + r^2\sin^2\theta\,\dd\phi^2 , \qquad r_{\mathrm{s}} = \frac{2GM}{c^2} \] The line element of spacetime outside a non-rotating spherical mass \(M\), in the signature \((-,+,+,+)\) of Unit 9. The length \(r_{\mathrm{s}}\) is the Schwarzschild radius of the mass; for the Sun it is \(2.95\) km and for the Earth \(8.87\) mm, the size of a marble. The metric holds for \(r\) larger than the radius of the body, and it is exact.
Gravitational time dilation, exact form
\[ \frac{\dd\tau}{\dd t} = \sqrt{1 - \frac{r_{\mathrm{s}}}{r}} = \sqrt{1 - \frac{2GM}{rc^2}} \] The proper time \(\dd\tau\) that passes on a clock at rest at the Schwarzschild coordinate \(r\), per coordinate time \(\dd t\), which is the time of a clock at rest far from the mass. Two clocks at rest at \(r_1\) and \(r_2\) run at the rate ratio \(\sqrt{(1 - r_{\mathrm{s}}/r_1)/(1 - r_{\mathrm{s}}/r_2)}\), and light sent from one to the other arrives with its frequency changed by the same ratio.
The effective potential and the innermost stable circular orbit
\[ \frac{1}{2}\left(\frac{\dd r}{\dd\tau}\right)^2 + V_{\mathrm{eff}}(r) = \mathcal{E} , \qquad V_{\mathrm{eff}}(r) = -\frac{GM}{r} + \frac{L^2}{2r^2} - \frac{GML^2}{c^2 r^3} , \qquad r_{\mathrm{ISCO}} = \frac{6GM}{c^2} = 3\,r_{\mathrm{s}} \] The radial motion of a body of angular momentum \(L\) per unit mass, in its own proper time. The first two terms are Newton’s gravitational and centrifugal potentials; the third is the relativistic correction, and it is attractive and dominates at small \(r\). Stable circular orbits exist only for \(L \geq \sqrt{12}\,GM/c\) and only outside \(r_{\mathrm{ISCO}}\), the innermost stable circular orbit.
The perihelion advance
\[ \Delta\phi = \frac{6\pi GM}{c^2\,a\,(1 - e^2)} \quad\text{per orbit} \] The angle by which the pericentre of a bound orbit of semi-major axis \(a\) and eccentricity \(e\) advances, in the direction of the motion, on every revolution around a mass \(M\). The formula is the first-order result and is exact enough for every orbit in this unit; the next correction is of order \((r_{\mathrm{s}}/a)^2\).
The deflection of light
\[ \theta = \frac{4GM}{c^2\,b} = 1.75\unit{arcsec}\times\frac{\Rsun}{b}\quad\text{for the Sun} \] The angle through which a light ray passing a mass \(M\) at the impact parameter \(b\) is bent towards it, to first order in \(GM/(c^2b)\). The star appears displaced away from the Sun by the same angle. For the Sun the angle at the solar edge is \(1.751\) arcseconds and falls as \(1/b\) outward.
The Shapiro delay of a radar echo
\[ \Delta t \approx \frac{4GM}{c^3}\,\ln\frac{4\,r_{\mathrm{E}}\,r_{\mathrm{P}}}{b^2} \] The extra round-trip time of a signal from the Earth, at the distance \(r_{\mathrm{E}}\) from the Sun, to a planet or spacecraft at the distance \(r_{\mathrm{P}}\) and back, when the signal passes the Sun at the distance \(b \ll r_{\mathrm{E}}, r_{\mathrm{P}}\). For the Sun \(4GM/c^3 = 19.7\) microseconds, and the logarithm is about 13 for a ray that grazes the solar edge.
Black Holes and Event Horizons
Gravitational redshift from a static source
\[ 1 + z = \frac{f_{\mathrm{emitted}}}{f_{\mathrm{received}}} = \left(1 - \frac{r_{\mathrm{s}}}{r}\right)^{-1/2} \] A source held at rest at the Schwarzschild coordinate \(r\) sends light of frequency \(f_{\mathrm{emitted}}\); a receiver at rest far away measures \(f_{\mathrm{received}}\). The redshift grows without limit as \(r\) approaches \(r_{\mathrm{s}}\), and the formula applies only to a source at rest; a moving source adds its Doppler shift.
Proper time of a radial fall
\[ \tau = \frac{\pi}{2}\,\frac{r_0}{c}\sqrt{\frac{r_0}{r_{\mathrm{s}}}} \] The time on the clock of a probe released from rest at the radius \(r_0\) until it reaches \(r = 0\), with \(r_{\mathrm{s}} = 2GM/c^2\). The formula is exact for any \(r_0 > r_{\mathrm{s}}\) and does not notice the horizon. The distant observer’s coordinate time \(t\) for the same fall is infinite.
Tidal acceleration across a falling body
\[ \Delta g \approx \frac{2GM\,\ell}{r^3} \] The difference between the gravitational accelerations at the two ends of a body of length \(\ell\) oriented along the radial direction at radius \(r\), which stretches the body. At the horizon, \(r = r_{\mathrm{s}} = 2GM/c^2\), it becomes \(\Delta g = c^6\ell/(4G^2M^2)\) and falls as \(1/M^2\).
Photon sphere and shadow
\[ r_{\mathrm{ph}} = \tfrac32\,r_{\mathrm{s}} , \qquad b_{\mathrm{c}} = \frac{\sqrt{27}}{2}\,r_{\mathrm{s}} , \qquad d_{\mathrm{shadow}} = 2b_{\mathrm{c}} = 2\sqrt{27}\,\frac{GM}{c^2} \approx 5.2\,r_{\mathrm{s}} \] Light can orbit a non-rotating black hole at the radius \(r_{\mathrm{ph}}\); any ray that approaches with an impact parameter below \(b_{\mathrm{c}}\) is captured. Seen from far away against a bright background, the hole therefore appears as a dark disc of diameter \(d_{\mathrm{shadow}}\), about 2.6 times the diameter of the horizon.
Horizon area and the area theorem
\[ A = 4\pi r_{\mathrm{s}}^2 = \frac{16\pi G^2 M^2}{c^4} , \qquad \dd A \ge 0 \] The area of the event horizon of a non-rotating black hole of mass \(M\) (a rotating one of the same mass has a smaller area), and the area theorem, proved by Hawking in 1971: in any process obeying the field equations with matter of positive energy, the total area of all event horizons never decreases. Because \(A\) grows as \(M^2\), two merging black holes always end with more area than they had, even though mass is lost to the waves, and a black hole can never split in two.
Linearized Gravity and Gravitational Waves
The linearized field equations
\[ g_{\mu\nu} = \eta_{\mu\nu} + h_{\mu\nu} , \qquad \Box\,\bar{h}_{\mu\nu} = -\frac{16\pi G}{c^4}\,T_{\mu\nu} , \qquad \Box = \eta^{\mu\nu}\partial_{\mu}\partial_{\nu} = -\frac{1}{c^2}\frac{\partial^2}{\partial t^2} + \nabla^2 \] In the Lorenz gauge, Equation , the ten components of the trace-reversed perturbation \(\bar{h}_{\mu\nu}\) each obey a wave equation whose source is the stress-energy tensor \(T_{\mu\nu}\) of Unit 10. The operator \(\Box\) is the d’Alembertian, the wave operator of flat spacetime, and \(\nabla^2\) is the sum of the second derivatives with respect to \(x\), \(y\), and \(z\). The equations hold wherever \(|h_{\mu\nu}| \ll 1\).
A plane gravitational wave
\[ \dd s^2 = -c^2\,\dd t^2 + \left(1 + h_+\right)\dd x^2 + \left(1 - h_+\right)\dd y^2 + 2h_\times\,\dd x\,\dd y + \dd z^2 , \qquad h_+, h_\times \propto \cos\!\left[\omega\!\left(t - \frac{z}{c}\right)\right] \] The metric of a gravitational wave travelling along \(z\) in the transverse-traceless gauge. The wave is transverse, since it changes only distances in the plane perpendicular to its direction, and it has two polarizations, plus and cross, which differ by a rotation of \(45^{\circ}\) about the direction of travel. The dimensionless amplitudes \(h_+\) and \(h_\times\) are called the strain.
The strain
\[ \frac{\Delta L}{L} = \frac{h}{2} \] A gravitational wave of amplitude \(h\) changes the proper distance \(L\) between two free masses placed across its direction of travel by the fraction \(h/2\), stretching along one transverse axis and squeezing along the other. The value \(h/2\) holds for a wave of one polarization with the masses placed along the axis of that polarization; for any other orientation the change is smaller by a geometrical factor between 0 and 1. The word strain is used both for the amplitude \(h\) of the wave and for the fractional change \(\Delta L/L\) that it produces, and the two differ by the factor one half.
The quadrupole formula
\[ \bar{h}_{ij}(t, r) = \frac{2G}{c^4\,r}\,\ddot{Q}_{ij}\!\left(t - \frac{r}{c}\right) , \qquad Q_{ij} = \int \rho\,x_i\,x_j\,\dd^3x \] The spatial components of the wave at a distance \(r\) that is large compared with the size of the source and with the wavelength, from a source that moves slowly compared with \(c\), are set by the second time derivative of the mass quadrupole moment \(Q_{ij}\) of the source, evaluated at the retarded time \(t - r/c\). The physical wave is the transverse-traceless part of this expression with respect to the direction from the source to the observer. The formula is called the quadrupole formula, and the derivation box sketches where it comes from.
The power radiated by a circular binary
\[ P = \frac{32}{5}\,\frac{G}{c^5}\,\mu^2 a^4\omega^6 = \frac{32}{5}\,\frac{G^4}{c^5}\,\frac{(m_1 m_2)^2\,(m_1 + m_2)}{a^5} \] The gravitational-wave power of two masses \(m_1\) and \(m_2\) on a circular orbit of separation \(a\) and angular frequency \(\omega\), with \(\mu = m_1 m_2/(m_1 + m_2)\). The second form follows from the first by Kepler’s third law. For an elliptical orbit of eccentricity \(e\) the power is larger by the factor \(f(e) = \left(1 + \tfrac{73}{24}e^2 + \tfrac{37}{96}e^4\right)/(1 - e^2)^{7/2}\), which Peters and Mathews derived in 1963 and which Section 5 uses.
The orbital decay of a binary
\[ \dot{P}_{\mathrm{b}} = -\frac{192\pi}{5}\,\frac{G^{5/3}}{c^5}\left(\frac{P_{\mathrm{b}}}{2\pi}\right)^{-5/3} \frac{m_1 m_2}{(m_1 + m_2)^{1/3}}\,f(e) \] The rate of change of the orbital period \(P_{\mathrm{b}}\) of two masses that lose energy to gravitational waves, a pure number (seconds per second) that is negative because the orbit shrinks and the period gets shorter. \(f(e)\) is the eccentricity factor of the key equation of Section 4, equal to 1 for a circular orbit. The accumulated shift of the time of periastron after a time \(t\) is \(\tfrac12\,(\dot{P}_{\mathrm{b}}/P_{\mathrm{b}})\,t^2\), which grows with the square of the time.
Merging Black Holes and Neutron Stars
The chirp mass and the chirp equation
\[ \mathcal{M} = \frac{(m_1 m_2)^{3/5}}{(m_1 + m_2)^{1/5}} , \qquad \dot{f} = \frac{96}{5}\,\pi^{8/3}\left(\frac{G\mathcal{M}}{c^3}\right)^{5/3} f^{11/3} \] The gravitational-wave frequency \(f\) of a circular binary rises at a rate \(\dot{f} = \dd f/\dd t\) that depends on the two masses only through the chirp mass \(\mathcal{M}\). The equation is the first term of a series in powers of \(v/c\), the orbital speed divided by the speed of light, and holds while the two bodies are well separated and slow; near the merger the correction terms of Section 3 matter.
The time to merger
\[ t_{\mathrm{merge}} = \frac{5}{256}\left(\frac{G\mathcal{M}}{c^3}\right)^{-5/3}\left(\pi f\right)^{-8/3} \] This is the time from the moment the gravitational-wave frequency is \(f\) until the two bodies merge, as the chirp equation predicts it. The chirp equation puts the merger at infinite frequency; the real merger comes a little earlier, at the frequencies of Section 2. The quantity \(G\mathcal{M}/c^3\) is a time, equal to \(4.925\unit{\mu s}\) per solar mass of chirp mass.
The strain amplitude of an inspiral
\[ h = \frac{4}{r}\left(\frac{G\mathcal{M}}{c^2}\right)^{5/3}\left(\frac{\pi f}{c}\right)^{2/3} \] The strain at a distance \(r\) from a circular binary of chirp mass \(\mathcal{M}\) whose gravitational-wave frequency is \(f\), for an observer on the axis of the orbit; for other directions and for a detector that is not optimally oriented the amplitude is smaller by a factor between 0 and 1. The quantity \(G\mathcal{M}/c^2\) is a length, \(1,477\unit{m}\) per solar mass of chirp mass.
The frequency of the innermost stable circular orbit
\[ f_{\mathrm{ISCO}} = \frac{c^3}{6^{3/2}\,\pi\,GM} \approx 4,400\unit{Hz} \times \frac{\Msun}{M} \] The gravitational-wave frequency of a test body on the innermost stable circular orbit around a non-rotating mass \(M\), at the separation \(6GM/c^2\). For two bodies of comparable mass there is no sharp last orbit; the computer simulations of Section 3 put the peak of the wave amplitude at two to three times this frequency, so the formula gives the scale at which the inspiral ends.
The ringdown of a black hole
\[ f_{\mathrm{ring}} \approx 1.2 \times 10^{4}\unit{Hz} \times \frac{\Msun}{M} , \qquad \tau \approx 0.055\unit{ms} \times \frac{M}{\Msun} \] The frequency and the damping time of the slowest-decaying quasi-normal mode of a non-rotating black hole of mass \(M\) (the mode with the angular pattern of the quadrupole, which a merger excites most strongly), from the perturbation theory of the Schwarzschild metric. A spinning black hole rings at a higher frequency and for longer: for the spin 0.67 that is typical of a merger remnant the frequency is 1.4 times higher, and for faster spins the factor grows beyond 2.
Laser Interferometer Observatories
Arm length change and phase shift
\[ \Delta L = \frac{h}{2}\,L , \qquad \Delta\phi = \frac{4\pi}{\lambda}\,\Delta L \] A gravitational wave of strain \(h\), with its polarization lined up with the arms and arriving along the normal to the detector plane, changes the length \(L\) of each arm by \(\Delta L\), with opposite signs in the two arms, and shifts the phase of light of wavelength \(\lambda\) by \(\Delta\phi\) per round trip. The phase difference between the arms, which the interference measures, is \(2\Delta\phi = 4\pi hL/\lambda\) per round trip. The formulas hold while the round-trip time \(2L/c\) is short compared with the period of the wave.
The Fabry-Perot arm cavity
\[ \mathcal{F} \approx \frac{2\pi}{T} , \qquad N_{\mathrm{eff}} = \frac{2\mathcal{F}}{\pi} , \qquad f_{\mathrm{p}} = \frac{c}{4L\mathcal{F}} \] A cavity of length \(L\) whose input mirror transmits the fraction \(T\) of the light, with an end mirror that transmits almost nothing, has the finesse \(\mathcal{F}\), multiplies the phase shift of a single round trip by \(N_{\mathrm{eff}}\), the number of effective round trips, and responds to signals below its pole frequency \(f_{\mathrm{p}}\); above \(f_{\mathrm{p}}\) the response falls as \(1/\sqrt{1 + (f/f_{\mathrm{p}})^2}\). These are the standard results of cavity optics, stated without derivation. For Advanced LIGO, \(T = 1.4\) per cent, \(\mathcal{F} \approx 450\), \(N_{\mathrm{eff}} \approx 286\), and \(f_{\mathrm{p}} \approx 42\) Hz.
The shot-noise limit
\[ \sqrt{S_h}\,(\text{shot}) = \frac{1}{L}\sqrt{\frac{\hbar c\lambda}{4\pi P}} \] The amplitude spectral density of the strain noise from the random arrival of photons, for a Michelson interferometer with arms of length \(L\), light of wavelength \(\lambda\) and power \(P\) at the beam splitter, read out at the dark fringe; \(\hbar\) is the reduced Planck constant. The noise is the same at every frequency and falls as \(1/\sqrt{P}\). With Fabry-Perot arms of \(N_{\mathrm{eff}}\) effective round trips it is smaller by \(N_{\mathrm{eff}}\) below the pole frequency.
The amplitude spectral density and the signal-to-noise ratio
\[ h_{\mathrm{rms}}(f, \Delta f) = \sqrt{S_h(f)\,\Delta f} , \qquad \rho^2 = 4\int_0^\infty \frac{|\tilde{h}(f)|^2}{S_h(f)}\,\dd f \] The noise power spectral density \(S_h(f)\), in \(\unit{Hz^{-1}}\), gives the root-mean-square strain noise \(h_{\mathrm{rms}}\) in a band \(\Delta f\) around \(f\); its square root is the amplitude spectral density of Figure 3. The signal-to-noise ratio \(\rho\) of a matched filter weights the Fourier transform \(\tilde{h}(f)\) of the signal by the inverse of the noise at each frequency, and is the height of the correlation peak in units of its scatter in noise alone. For noise that is flat across the band, \(\rho^2\) reduces to twice the integral of \(h(t)^2\) over the duration of the signal, divided by \(S_h\).
The time delay between two detectors
\[ \Delta t = \frac{d}{c}\cos\theta \le \frac{d}{c} \] A gravitational wave arrives at two detectors a straight-line distance \(d\) apart with a delay \(\Delta t\) that depends on the angle \(\theta\) between the direction to the source and the baseline; the delay can never exceed \(d/c\). For LIGO Hanford and Livingston, \(d = 3{,}002\) km and \(d/c = 10.0\) ms. A measured delay confines the source to a ring of the sky of angular radius \(\theta\) about the baseline.
Pulsar Timing Arrays and Space Detectors
The timing residual of a gravitational wave
\[ \frac{\delta\nu}{\nu} \sim h , \qquad \delta t \sim \frac{h}{2\pi f} \] A gravitational wave of strain \(h\) and frequency \(f\) changes the received frequency of a pulsar’s pulses by a fraction of order \(h\) and shifts their arrival times by a residual of order \(h/(2\pi f)\), with a geometric factor of magnitude between 0 and 1 that depends on the direction of the wave, and on whether the Earth term alone or both terms are counted. The residual grows in proportion to the period of the wave, so the slowest waves produce the largest residuals, as long as one period still fits into the length of the data set.
The Hellings-Downs curve
\[ \Gamma(\xi) = \frac12 + \frac32\,x\ln x - \frac{x}{4} , \qquad x = \frac{1 - \cos\xi}{2} \] The expected correlation between the gravitational-wave residuals of two pulsars whose lines of sight are the angle \(\xi\) apart, for an isotropic background of waves in general relativity, as a fraction of the variance, the mean squared size, of the residual of one pulsar. It is \(\tfrac12\) for two different pulsars in the same direction (a pulsar correlated with itself gives 1, because its pulsar term is then shared too), crosses zero at \(49.3^{\circ}\), reaches its minimum of \(-0.152\) at \(82.5^{\circ}\), crosses zero again at \(121.8^{\circ}\), and is \(\tfrac14\) for pulsars on opposite sides of the sky. The curve does not depend on the frequency or the strength of the background.
The characteristic strain of a background of binaries
\[ h_c(f) = A\left(\frac{f}{1\unit{yr^{-1}}}\right)^{-2/3} , \qquad A = 2.4^{+0.7}_{-0.6} \times 10^{-15} \] The characteristic strain of an isotropic background of binaries that evolve by gravitational radiation alone falls as the \(-2/3\) power of the frequency; the amplitude \(A\) at the reference frequency of one cycle per year is the value measured by NANOGrav in 2023 for that fixed slope (median and 90 per cent interval). The papers often quote the same spectrum in the timing residuals, where the corresponding power spectral density falls as \(f^{-13/3}\): the residual is the strain divided by \(2\pi f\), the power is its square, and a further factor \(1/f\) turns a characteristic strain into a spectral density.
The arm response of a space detector
\[ f_* = \frac{c}{2\pi L} , \qquad f_* = 19\unit{mHz} \ \text{for} \ L = 2.5 \times 10^{9}\unit{m} \] A detector with arms of length \(L\) responds to the full strain of a wave at frequencies well below the transfer frequency \(f_*\), the frequency at which the arm equals the wavelength divided by \(2\pi\) (the sixth of a wavelength of Section 1); above \(f_*\) the wave reverses sign while the light is still in the arm, and the response falls, roughly as \(1/f\). LISA’s band therefore runs from about 0.1 mHz, where the acceleration noise of the test masses takes over, to about 1 Hz, with its best sensitivity between a few millihertz and \(f_*\).
The gravitational-wave frequency of a known binary
\[ f_{\mathrm{gw}} = \frac{2}{P} \] A binary with the orbital period \(P\) radiates at twice its orbital frequency, as Unit 13 derived for the quadrupole of a dumbbell. For the binaries that optical and X-ray astronomy have found with periods of minutes, \(f_{\mathrm{gw}}\) lies in the millihertz band, and these binaries, called verification binaries because they will verify that the instrument works, are the signals LISA is guaranteed to find.