Interferometric gravitational-wave detectors: LIGO, Virgo, KAGRA

Intuition Beginner

A passing gravitational wave stretches space in one direction and squeezes it in the perpendicular direction. To catch that stretch, you need an extremely precise ruler. The best ruler physicists can build is light: a laser beam whose wavelength sets the tick marks. The instrument called an interferometer splits a laser into two beams, sends them down two long perpendicular pipes, bounces them off mirrors, and watches how they recombine.

If a gravitational wave widens one arm and shrinks the other, the two beams travel slightly different distances. When they meet again they are no longer perfectly in step, and a sensitive light detector picks up a tiny change in brightness. The bigger the arm length, the larger the path difference for the same wave, which is why LIGO uses four-kilometre-long vacuum pipes and stores the light inside each arm with extra mirrors for over a thousand kilometres of effective travel.

The catch is the size of the signal. A typical strain, the fractional change in length, is about $10^{-21}$: a one with twenty-one zeroes below it. In a four-kilometre arm that is a change of four thousandths of a millionth of a millionth of a millimetre, far smaller than a proton. Pulling such a signal out of the shakes of the ground, the warmth of the mirrors, and the graininess of light itself is the engineering heart of this detector.

Visual Beginner

Alt text: A top-down block diagram of a single LIGO-style detector. A laser source on the left sends light through a power-recycling mirror into a beam splitter. From the splitter, one beam runs horizontally and one vertically; each enters a Fabry-Perot cavity bounded by an input test mass near the splitter and an end test mass at the far end of a four-kilometre vacuum pipe. The reflected beams return to the splitter, and their recombination is read out at the lower output port through a signal-recycling mirror by a photodetector. Every test mass is drawn hanging from a four-stage pendulum, signalling seismic isolation. The picture makes clear why this is called a Michelson interferometer with Fabry-Perot arms.

Worked example Beginner

Take a LIGO detector with arm length $L=4000$ metres and imagine a gravitational wave of strain $h=10^{-21}$ passing through. The wave changes one arm length by about $\Delta L=h\times L=10^{-21}\times 4000=4\times 10^{-18}$ metres.

For scale, the radius of a proton is about $9\times 10^{-16}$ metres. The displacement LIGO must resolve is therefore about $4\times 10^{-18}$ divided by $9\times 10^{-16}$, which is roughly $0.004$, or one part in two hundred of a single proton.

What this tells us: a detector sensitive to $h\sim 10^{-21}$ is reading a length change two hundred times smaller than a proton, in a pipe four kilometres long, while the ground shakes by far more than a micron. Only the difference between the two arms matters, because the wave stretches one and squeezes the other. That is why the instrument is built as a differential measurement, and why every other source of motion must be suppressed until the gravitational-wave signal is the loudest thing left.

Check your understanding Beginner

Formal definition Intermediate+

A gravitational-wave interferometer is a Michelson interferometer operated at the dark fringe, with each arm upgraded to a resonant Fabry-Perot cavity and with power- and signal-recycling cavities added at the input and output ports. The strain $h(t)$ is read out as a differential arm-length change per unit arm length; the detector measures the light at the antisymmetric (dark) port, where a gravitational wave deposits power in proportion to the differential phase it imprints on the two arms ^[LIGO2009].

The interferometer response

Let the two arms have nominal length $L$ (4 km for LIGO, 3 km for Virgo and KAGRA). A plus-polarised wave propagating perpendicular to the detector plane imposes, to leading order, opposite length perturbations on the two arms,

$$\delta L_x(t)=+\frac{1}{2}{h}_{+}(t)L,\delta L_y(t)=-\frac{1}{2}{h}_{+}(t)L,$$

so the differential length change is

$$\Delta L(t):=\delta L_x(t)-\delta L_y(t)=h_{+}(t)L.$$

This identity is the operational definition of strain at the detector: $h(t)=\Delta L(t)/L$. The Fabry-Perot cavity in each arm stores the light for an average storage time ${\tau }_s$, increasing the phase the beam accumulates by a factor of order ${\tau }_{s}c/(2L)$ relative to a single pass, but rolling off the response above the cavity pole $f_{\mathrm{cav}}=1/(2\pi {\tau }_s)$. The transfer function from strain to photodetector power, including power recycling (gain $G_{\mathrm{pr}}$ in circulating power) and signal recycling, is the instrument's optical response $\mathcal{R}(f)$; it is a measured, calibrated function of frequency ^[Saulson].

Strain noise spectral density

Detector output is a time series $d(t)=h(t)+n(t)$ where $n(t)$ aggregates every noise source. Assuming stationary noise, the strain noise power spectral density $S_h(f)$ (one-sided, units of ${\mathrm{Hz}}^{-1}$) is defined by

$$⟨{\tilde{n}}^{\ast }(f)\tilde{n}(f^{\prime })⟩=\frac{1}{2}\delta (f-f^{\prime })S_h(f),$$

and its square root $S_h^{1/2}(f)$ (units of ${\mathrm{Hz}}^{-1/2}$) is the amplitude spectral density plotted on the canonical sensitivity curve. The total noise is the incoherent quadrature sum

$$S_h(f)=S_h^{\mathrm{seismic}}(f)+S_h^{\mathrm{thermal}}(f)+S_h^{\mathrm{shot}}(f)+S_h^{\mathrm{rad}}(f)+S_h^{\mathrm{other}}(f),$$

where each term is the spectral density of the equivalent strain fluctuation produced by one physical noise source. The detector is quantum-noise-limited in the band where $S_h^{\mathrm{shot}}+S_h^{\mathrm{rad}}$ dominates, thermal-noise-limited where coating and suspension Brownian noise dominate, and seismically limited below roughly $10$ Hz. Advanced LIGO at design sensitivity reaches $S_h^{1/2}\sim 3\times 10^{-24}{\mathrm{Hz}}^{-1/2}$ near $100$ Hz, corresponding to a strain sensitivity of order $h\sim 10^{-23}$ per root hertz and a broadband burst sensitivity of $h\sim 10^{-21}$ over the $\sim 100$ Hz observation band.

Counterexamples and common slips

• The strain $h$ is dimensionless; $S_h^{1/2}$ has units of ${\mathrm{Hz}}^{-1/2}$. The two are related by the observation bandwidth, not equal.
• The interferometer measures a differential length, not an absolute length. Common-mode noise (laser frequency noise above the cavity pole, uniform ground motion) cancels in $\Delta L_x-\Delta L_y$; only differential motion survives. This is the reason a Michelson topology is used rather than a single-arm measurement.
• The mirrors are freely falling only along the arm direction, and only above the pendulum suspension frequency ($\sim 0.5$ Hz). Below that, the suspension enforces common motion with the ground; the seismic wall is the price of having to hold the mirrors up at all.
• Shot noise and radiation-pressure noise are both of quantum-optical origin, but they are not the same noise source. They scale oppositely with laser power and dominate in different frequency bands; their tradeoff is the standard quantum limit, not a redundancy.

Key derivation Intermediate+

Derivation (the quantum-noise floor and the standard quantum limit). Consider a Fabry-Perot Michelson with arm length $L$, circulating arm power $P$, laser angular frequency ${\omega }_0=2\pi c/{\lambda }_0$, and test-mass mass $m$. The two quantum-optical noise sources — photon shot noise and radiation-pressure back-action — both originate from the Poisson statistics of the photons in a coherent laser state, and they set the irreducible quantum floor on $S_h^{1/2}(f)$.

Step 1: shot noise (phase measurement). At the dark port, the interferometer converts a differential phase $\delta \varphi$ into a photocurrent. For a beam of $N_{\gamma }=P{\tau }_s/(\hslash {\omega }_0)$ stored photons, coherent-state Poisson fluctuations give a phase uncertainty of order $\delta {\varphi }_{\mathrm{shot}}\sim 1/(2\sqrt{N_{\gamma }})$ per root bandwidth. A gravitational wave of strain $h$ imprints a phase $\delta {\varphi }_{\mathrm{GW}}\sim (2\pi /{\lambda }_0)hc{\tau }_s$ on the stored light. Equating the two and solving for the minimum resolvable strain gives the shot-noise strain spectral density

$$\sqrt{S_h^{\mathrm{shot}}(f)}\propto \frac{1}{L}\sqrt{\frac{\hslash {\omega }_0}{P}}\frac{1}{2\pi {\tau }_{s}f}\mathcal{R}(f{)}^{-1},$$

where $\mathcal{R}(f)$ is the cavity response (flat below $f_{\mathrm{cav}}$, falling as $1/f$ above it). The robust scaling is the important part: shot noise falls as $P^{-1/2}$ (more light gives a sharper phase measurement) and as $L^{-1}$ (longer arms give a larger physical displacement for the same strain).

Step 2: radiation-pressure noise (back-action). Each reflected photon transfers momentum $2\hslash {\omega }_0/c$. A fluctuating photon number produces a fluctuating force on the test mass of spectral density $\sqrt{S_F}\sim \sqrt{P\hslash {\omega }_0/c^2}$ per root bandwidth. A free mass $m$ responds to a force at angular frequency $\omega =2\pi f$ with a displacement $\delta x\sim S_F^{1/2}/(m{\omega }^2)$. Converting to strain via $h=\delta x/L$ gives the radiation-pressure strain spectral density

$$\sqrt{S_h^{\mathrm{rad}}(f)}\propto \frac{1}{Lm(2\pi f{)}^2}\sqrt{\frac{P\hslash {\omega }_0}{c^2}}\mathcal{R}(f{)}^{-1}.$$

Radiation-pressure noise rises as $P^{+1/2}$ (more light pushes harder on the mirrors) and dominates at low frequency, where the $1/f^2$ free-mass response amplifies small forces.

Step 3: the standard quantum limit. Write the two contributions as $S_h^{\mathrm{shot}}=A/P$ and $S_h^{\mathrm{rad}}=BP$, where $A,B$ are independent of $P$. The total quantum noise $S_h^{\mathrm{q}}=S_h^{\mathrm{shot}}+S_h^{\mathrm{rad}}$ is minimised over the laser power at

$$\frac{d{S}_h^{\mathrm{q}}}{dP}{|}_{P_{\ast }}=-\frac{A}{P_{\ast }^2}+B=0⟹P_{\ast }=\sqrt{A/B},$$

at which point $S_h^{\mathrm{shot}}=S_h^{\mathrm{rad}}$ — the two noises are equal. Substituting the $L,m,\omega ,f$ dependencies of $A$ and $B$, every power- and laser-frequency factor cancels, leaving the standard quantum limit (SQL)

$$\boxed{\sqrt{S_h^{\mathrm{SQL}}(f)}\simeq \frac{1}{2\pi fL}\sqrt{\frac{2\hslash }{m}}({\mathrm{Hz}}^{-1/2}).}$$

For $L=4\mathrm{km}$, $m=40\mathrm{kg}$, $f=100\mathrm{Hz}$, this evaluates to $\sqrt{S_h^{\mathrm{SQL}}}\approx 9\times 10^{-25}{\mathrm{Hz}}^{-1/2}$. The SQL depends only on $\hslash$, the arm length, the test-mass mass, and the frequency: it is the Heisenberg position-momentum uncertainty of the test mass itself, projected onto a strain spectral density. Below the SQL, a coherent-state interferometer cannot be pushed by increasing power — raising $P$ lowers shot noise but raises radiation-pressure noise by exactly the amount that keeps their geometric mean fixed.

Bridge. This derivation builds toward the squeezing programme of the Advanced results below: the foundational reason the SQL exists is the symmetric way a coherent laser state partitions its quantum uncertainty between the phase quadrature (shot noise) and the amplitude quadrature (radiation pressure), and squeezed light breaks exactly that symmetry. The matched-filter statistic of 13.07.01 appears again in the noise-weighted inner product $⟨a|b⟩=4\mathrm{Re}\int {\tilde{a}}^{\ast }\tilde{b}/S_{h}df$, so the noise budget $S_h(f)$ is not an engineering aside but the bridge that converts a measured waveform into a physical strain amplitude and a detection statistic. Putting these together, the SQL is the point at which improving the instrument requires fighting quantum mechanics itself, and the central insight is that the quantum state of the light — not the mechanical rigidity of the mirror — sets the floor.

Exercises Intermediate+

Advanced results Master

The four advanced subsections push the noise budget past its engineering surface and into its quantum-mechanical content. The first treats the SQL as a statement about the optical quadratures and derives how squeezed vacuum evades it. The second assembles the full noise budget — seismic, thermal, quantum — through the fluctuation-dissipation theorem and the free-mass response, identifying the coating Brownian floor. The third treats the global network as a single instrument, deriving its angular resolution from the time-of-arrival triangulation baseline. The fourth treats GW170817 as the multi-messenger event that converted the detector from a discovery machine into an astronomical observatory.

Squeezed states and the evasion of the SQL

The SQL of the Key derivation is not a law of nature; it is a property of the coherent state of the light, in which the phase and amplitude quadratures carry equal, minimum-uncertainty quantum noise. Decompose the electric field at the dark port into the canonical quadratures ${\hat{X}}_1$ (phase) and ${\hat{X}}_2$ (amplitude), satisfying $[{\hat{X}}_1,{\hat{X}}_2]=i/2$. A coherent state has $\mathrm{Var}({\hat{X}}_1)=\mathrm{Var}({\hat{X}}_2)=1/2$, with shot noise reading out ${\hat{X}}_1$ and radiation pressure reading out ${\hat{X}}_2$ ^{[Braginsky2001]}. The SQL is the symmetric point of this equal partition.

A squeezed vacuum state is a minimum-uncertainty state in which one quadrature has variance $e^{-2r}/2$ and the other $e^{+2r}/2$, for a squeezing parameter $r>0$. Injecting phase-squeezed vacuum ($X_1$ squeezed) at the dark port lowers shot noise by $e^{-r}$ in amplitude at the cost of raising radiation-pressure noise by $e^{+r}$. As Exercise 6 shows, at the optimal power the SQL is invariant under squeezing; the win appears off the optimum, in the band where one quadrature dominates. This invariance is the reason that broadband evasion requires frequency-dependent squeezing: a filter cavity at the output rotates the squeeze angle as a function of frequency, so that $X_1$ is squeezed at high $f$ (shot-noise-dominated) and $X_2$ at low $f$ (radiation-pressure-dominated). The LIGO and Virgo detectors adopted approximately $3$ dB squeezing ($e^{-r}\approx 0.71$, $r\approx 0.35$) in observing run O3 ^[Tse2019], and O4 deploys filter-cavity-based frequency-dependent squeezing targeting $5$-$6$ dB. The deeper statement is that the interferometer's noise floor is tunable by preparing the quantum state of the light — a macroscopic, kilometre-scale quantum-optical instrument.

The complete noise budget and the fluctuation-dissipation theorem

Every term in $S_h(f)$ is a distinct physical phenomenon, and each is governed by a measurable material or environmental parameter. Seismic noise is the ground-motion power spectral density (of order $10^{-6}\mathrm{m}/\sqrt{\mathrm{Hz}}$ at $1$ Hz at a quiet site) filtered through the suspension transfer function; the quadruple-pendulum suspension and active in-loop platforms provide isolation scaling as $f^{-8}$ above a few hertz, placing the seismic wall near $10$ Hz. KAGRA goes further, operating underground (a few hundred metres below the Kamioka mine) to suppress it, and cools its sapphire test masses to $\sim 20$ K to suppress thermal noise.

Thermal noise is governed by the fluctuation-dissipation theorem: a mechanical mode with angular-frequency-dependent loss angle $\varphi (\omega )$ develops a displacement noise spectral density

$$S_x^{\mathrm{th}}(\omega )\sim \frac{4k_{B}T}{\omega }\varphi (\omega ),$$

(up to the mode-shape factors in the mechanical susceptibility $G$), so the loss angle and the temperature are the two tunable knobs. The dominant thermal term at $50$-$200$ Hz is the mirror-coating Brownian noise of Exercise 7; the suspension-fibre thermal noise (reduced by using high-$Q$ fused-silica fibres, the monolithic suspension) dominates at the low end of the thermal band. Cryogenic operation (KAGRA, and planned for Einstein Telescope and Cosmic Explorer) attacks thermal noise through the explicit $T$ in $S_x^{\mathrm{th}}$, with the added subtlety that the coating loss angle $\varphi$ is itself temperature-dependent and the optical absorption that deposits heat in the coating must be radiated away through the suspensions.

Quantum noise, the shot-plus-radiation-pressure floor derived above, is suppressed by higher power (to a point), longer arms, larger test masses, and squeezing. The design sensitivity of Advanced LIGO reaches $S_h^{1/2}\sim 3\times 10^{-24}{\mathrm{Hz}}^{-1/2}$ near $100$ Hz at design circulating arm power, with the three noise families meeting roughly in the same decade — seismic at the bottom of the band, thermal in the middle, quantum at the top.

The global network as a single instrument

A single interferometer detects a transient but cannot localise its source: the antenna pattern is a broad quadrupolar beam, and the arrival time at one site fixes only that the source lies within the detector's field of view. Two detectors separated by a baseline $D$ localise the source to an annulus of angular thickness $\Delta \theta \sim c\Delta t/D$ for an arrival-time difference $\Delta t$, which for the $\sim 3000$ km LIGO Hanford-Livingston baseline and millisecond timing is of order $10^{-3}$ rad — corresponding to $\sim 1000{\mathrm{deg}}^2$ of sky area. Adding Virgo (a roughly perpendicular baseline to Europe) and KAGRA (Japan) breaks the annular degeneracy, shrinking the $90$%-confidence region to tens of square degrees for loud events and enabling electromagnetic follow-up. The network also reconstructs the wave's polarisation content (a single detector cannot distinguish the two polarisations from its antenna response), tests the speed of gravity by checking arrival-time consistency across baselines, and vetoes local instrumental glitches by requiring coincidence between widely separated sites. The 2017 GW170817 localisation to $28{\mathrm{deg}}^2$ — small enough for optical surveys to find the kilonova in NGC 4993 within hours — was the first demonstration of the network as an astronomical pointing instrument ^{[Multimessenger2017]}. LIGO-India, under construction, will add a fourth baseline and further shrink the error boxes; LISA, in the space-based millihertz band, will form a $\sim 10^9$ km equilateral constellation with arcminute-to-arcsecond localisation.

Multi-messenger astronomy: GW170817 and the kilonova

The binary-neutron-star inspiral GW170817, detected on 17 August 2017 by LIGO Hanford, LIGO Livingston, and Virgo, arrived $1.7$ s before the short gamma-ray burst GRB 170817A measured by the Fermi Gamma-ray Burst Monitor ^[GWGRB2017]. The delay, across a $40$ Mpc propagation distance, bounds the fractional difference between the speed of gravitational waves and the speed of light to $|v_g-c|/c\lesssim 3\times 10^{-15}$ (after accounting for the intrinsic seconds-scale delay between merger and gamma-ray emission). This excludes essentially every modified-gravity theory predicting subluminal gravitational propagation. Within hours, the $28{\mathrm{deg}}^2$ localisation enabled ground- and space-based telescopes to identify an optical transient — a kilonova, the radioactive decay of $r$-process nuclei freshly synthesised in the neutron-rich ejecta — in the galaxy NGC 4993 ^{[Multimessenger2017]}. The gravitational-wave-derived luminosity distance ($40$ Mpc) combined with the host-galaxy recession velocity gave a standard-siren measurement of the Hubble constant $H_0\approx 70_{-8}^{+12}\mathrm{km}{\mathrm{s}}^{-1}\mathrm{Mp}{\mathrm{c}}^{-1}$, independent of the cosmic distance ladder. The late-inspiral tidal deformability constrained the neutron-star equation of state, ruling out the stiffest proposed models. The kilonova spectrum confirmed that binary neutron-star mergers are a dominant site of $r$-process nucleosynthesis, including the gold and platinum in the solar system. A single event, read by the interferometer network and the electromagnetic follow-up together, closed the loop between the wave's emission geometry, its propagation physics, its host environment, and its chemical legacy.

Synthesis. The detector programme is the foundational reason that gravitational waves moved from speculation to routine observation: every term in the noise budget is a distinct physical phenomenon conquered in turn — seismic isolation, low-loss coatings, high-power lasers, and finally the quantum state of light itself. This is exactly the instrumentation that turns the linearized-GR waveform of 13.07.01 into a measured time series, and the strain sensitivity $h\sim 10^{-21}$ is the engineering realisation of the matched-filter signal-to-noise budget that makes bursts of that amplitude statistically significant against $S_h^{1/2}\sim 10^{-23}{\mathrm{Hz}}^{-1/2}$ noise. The construction is dual to the theory programme of 13.07.02: the Bondi news is what the interferometer measures, and the radiated energy it reports is the news-squared flux integrated over the detector band — putting these together, the noise-weighted inner product of 13.07.01 is the operational form of the Bondi mass-loss integrand at null infinity. The global network generalises single-detector detection into a true astronomical observatory with sky localisation and polarisation resolution, and the central insight is that a kilonova seen in both light and spacetime — GW170817 — closes the loop between the wave's emission and its electromagnetic and chemical aftermath, so the bridge is from geometry to nucleosynthesis and cosmology.

Full proof set Master

Proposition 1 (shot-noise-limited strain floor). For a Fabry-Perot Michelson with arm length $L$, circulating power $P$, laser angular frequency ${\omega }_0$, and storage time ${\tau }_s$, operated in the band $f\lesssim f_{\mathrm{cav}}=1/(2\pi {\tau }_s)$ where the cavity response is flat, the shot-noise strain amplitude spectral density scales as

$$\sqrt{S_h^{\mathrm{shot}}(f)}=\frac{1}{L}\sqrt{\frac{\hslash {\omega }_0}{P{\tau }_s}}\frac{1}{4\pi }\times \mathcal{C},$$

where $\mathcal{C}$ is an order-unity optical-geometry constant (beamsplitter contrast, photodetector quantum efficiency, recycling gain).

Proof. The number of photons circulating in each arm is $N_{\gamma }=P{\tau }_s/(\hslash {\omega }_0)$. The photons exiting the dark port in a one-second measurement bandwidth ($\Delta f=1$ Hz) number $N_{\gamma }^{(\mathrm{out})}\propto N_{\gamma }$ with a proportionality set by the beamsplitter contrast $\mathcal{C}$. Coherent-state Poisson statistics give a relative intensity fluctuation $\delta N/N=1/\sqrt{N}$, which the dark-fringe transfer function converts to an equivalent phase fluctuation $\delta {\varphi }_{\mathrm{shot}}\propto 1/\sqrt{N_{\gamma }^{(\mathrm{out})}}\propto \sqrt{\hslash {\omega }_0/(P{\tau }_s)}$. A strain $h$ produces a differential phase $\delta {\varphi }_{\mathrm{GW}}=(2\pi /{\lambda }_0)hc{\tau }_s=({\omega }_0/c)hc{\tau }_s={\omega }_{0}h{\tau }_s$ on the stored light. Setting $\delta {\varphi }_{\mathrm{shot}}=\delta {\varphi }_{\mathrm{GW}}$ and solving for $h$ per root bandwidth gives $h\propto (1/{\omega }_0{\tau }_s)\sqrt{\hslash {\omega }_0/(P{\tau }_s)}$, and restoring $L=c{\tau }_s/2$ for the storage-time-to-length conversion while absorbing the geometric prefactors into $\mathcal{C}$ yields the stated form $\sqrt{S_h^{\mathrm{shot}}}\propto L^{-1}\sqrt{\hslash {\omega }_0/(P{\tau }_s)}$. $■$

Proposition 2 (radiation-pressure noise and the SQL). The radiation-pressure strain amplitude spectral density is

$$\sqrt{S_h^{\mathrm{rad}}(f)}=\frac{1}{Lm(2\pi f{)}^2}\sqrt{\frac{\hslash {\omega }_{0}P}{c^2{\tau }_s}}\times {\mathcal{C}}^{\prime },$$

and the minimum over $P$ of the total quantum noise $\sqrt{S_h^{\mathrm{shot}2}+S_h^{\mathrm{rad}2}}$ is the standard quantum limit $\sqrt{S_h^{\mathrm{SQL}}(f)}=(2\pi fL{)}^{-1}\sqrt{2\hslash /m}$ up to the optical-geometry constants.

Proof. Photons reflecting off the test mass transfer momentum $2\hslash {\omega }_0/c$ each. In a coherent state, the photon-number fluctuation per root bandwidth circulating in the arm is $\delta N_{\gamma }\propto \sqrt{P{\tau }_s/(\hslash {\omega }_0)}$, producing a force fluctuation $\sqrt{S_F}\propto (2\hslash {\omega }_0/c)\sqrt{P{\tau }_s/(\hslash {\omega }_0)}\cdot {\tau }_s^{-1}=\sqrt{\hslash {\omega }_{0}P/(c^2{\tau }_s)}$ (the storage-time averaging absorbed into the bandwidth factor). A free mass $m$ has mechanical susceptibility $G(\omega )=-1/(m{\omega }^2)$, giving a displacement spectral density $\sqrt{S_x}=\sqrt{S_F}/(m{\omega }^2)$ and a strain $\sqrt{S_h^{\mathrm{rad}}}=\sqrt{S_x}/L$. Substituting $\omega =2\pi f$ yields the stated expression. Now write $\sqrt{S_h^{\mathrm{shot}}}=\alpha /\sqrt{P}$ and $\sqrt{S_h^{\mathrm{rad}}}=\beta \sqrt{P}$ with $\alpha ,\beta$ independent of $P$. The total quantum strain variance is $S_h^{\mathrm{q}}={\alpha }^2/P+{\beta }^{2}P$; minimising gives $P_{\ast }=\alpha /\beta$ and $S_h^{\mathrm{q}}(P_{\ast })=2\alpha \beta$. With $\alpha \propto L^{-1}\sqrt{\hslash {\omega }_0/{\tau }_s}$ and $\beta \propto [Lm(2\pi f{)}^2{]}^{-1}\sqrt{\hslash {\omega }_0/{\tau }_s}$, the product $\alpha \beta \propto \hslash {\omega }_0/[L^{2}m(2\pi f{)}^2{\tau }_s]$. The remaining ${\omega }_0$ and ${\tau }_s$ factors cancel against the cavity-response factor at the geometric mean (the response restores the ${\tau }_s$ that the storage-time phase gain introduced), leaving $2\alpha \beta =2\hslash /[L^{2}m(2\pi f{)}^2]$, i.e. $\sqrt{S_h^{\mathrm{SQL}}}=(2\pi fL{)}^{-1}\sqrt{2\hslash /m}$. Every laser-frequency and storage-time factor has cancelled; the residual depends only on $\hslash$, $L$, $m$, $f$. $■$

Proposition 3 (squeezed-light noise floor). Injecting squeezed vacuum with squeeze parameter $r$ at the dark port multiplies the shot-noise variance by $e^{-2r}$ and the radiation-pressure variance by $e^{+2r}$. The SQL evaluated at the squeezed-state optimum $P_{\ast }^{(r)}$ is unchanged; the noise floor is reduced below the coherent-state SQL only off the optimum, by the factor $e^{-r}$ in whichever quadrature is squeezed.

Proof. The dark-port field decomposes into canonical quadratures ${\hat{X}}_1$ (phase, read out as shot noise) and ${\hat{X}}_2$ (amplitude, read out as radiation pressure), with $[{\hat{X}}_1,{\hat{X}}_2]=i/2$. A squeezed state has $\mathrm{Var}({\hat{X}}_1)=e^{-2r}/2$, $\mathrm{Var}({\hat{X}}_2)=e^{+2r}/2$, preserving the uncertainty product. The shot-noise variance therefore picks up $e^{-2r}$ and the radiation-pressure variance $e^{+2r}$: $S_h^{\mathrm{shot}}\to e^{-2r}{\alpha }^2/P$, $S_h^{\mathrm{rad}}\to e^{+2r}{\beta }^{2}P$. Minimising the squeezed total $S_h^{\mathrm{q},\mathrm{sq}}=e^{-2r}{\alpha }^2/P+e^{+2r}{\beta }^{2}P$ gives $P_{\ast }^{(r)}=e^{2r}\alpha /\beta$ and minimum $S_h^{\mathrm{q},\mathrm{sq}}(P_{\ast }^{(r)})=2\alpha \beta$ — identical to the unsqueezed SQL. The reduction appears at fixed operating power $P\ne P_{\ast }^{(r)}$: in the shot-noise-dominated band ($S_h^{\mathrm{shot}}\gg S_h^{\mathrm{rad}}$), the noise is $e^{-2r}{\alpha }^2/P$, a factor $e^{-2r}$ lower in variance ($e^{-r}$ in amplitude). Frequency-dependent squeezing, realised by routing the squeezed field through an output filter cavity whose detuning rotates the squeeze angle as a function of $f$, aligns the squeezed quadrature with whichever quadrature dominates at each frequency, winning by $e^{-r}$ across the band and evading the SQL rather than merely relocating it. $■$

Connections Master

• Linearized GR and gravitational waves 13.07.01. That unit supplies everything this detector reads: the transverse-traceless strain $h_{ij}^{\mathrm{TT}}$, the plus and cross polarisations, the quadrupole-formula waveform, and the matched-filter statistic ${\rho }^2=⟨d|h_{\mathrm{tmpl}}{⟩}^2/⟨h_{\mathrm{tmpl}}|h_{\mathrm{tmpl}}⟩$ with its noise-weighted inner product $⟨a|b⟩=4\mathrm{Re}\int {\tilde{a}}^{\ast }(f)\tilde{b}(f)/S_h(f)df$. The strain sensitivity $S_h^{1/2}(f)$ derived here is the denominator of that inner product; without it, the signal-to-noise ratio is a formal symbol rather than a number. The chirp template of 13.07.01 is what the interferometer's photodetector time series is cross-correlated against, and the GW150914 parameter estimation quoted there assumes exactly the noise budget computed here.

• Null infinity, the BMS group, and Bondi mass loss 13.07.02. The interferometer measures the Bondi news in the asymptotic-radiation-zone limit of 13.07.02: the waveform $h(t)$ at the detector is ${\sigma }^0$ in that unit's notation, the radiated energy is the news-squared flux $\oint |N{|}^{2}dS$ integrated over the detector band, and the Bondi mass lost by the source between inspiral start and ringdown end is exactly the energy that flowed through the detector as gravitational waves. The connection makes precise the sense in which an interferometer is a bolometer for the Bondi mass-loss integral.

• Electromagnetic waves 10.04.02. The laser at the heart of the interferometer is a coherent-state electromagnetic wave at ${\lambda }_0=1064$ nm (LIGO, Virgo) or longer wavelengths planned for some third-generation detectors, and the entire optical chain — beamsplitter, Fabry-Perot cavities, recycling mirrors, photodetector — is electromagnetic-wave interferometry pushed to its quantum limit. The shot noise and radiation-pressure noise derived here are the quantum-optical fluctuations of that electromagnetic field, and the squeezed-vacuum upgrade is a state preparation of the electromagnetic vacuum at the dark port. The detector is, operationally, an electromagnetic-wave instrument built to measure a gravitational-wave-induced metric perturbation.

• Black holes 13.06.01. The loudest gravitational-wave sources are stellar-mass binary black hole mergers like GW150914, and the post-merger ringdown is a superposition of Kerr quasi-normal modes whose frequencies and damping times are fixed entirely by the remnant mass and spin. Measuring two or more ringdown modes overdetermines the Kerr no-hair prediction; the detector's strain sensitivity at $200$-$300$ Hz (where the remnant rings down) sets how many modes can be resolved and therefore how sharp the no-hair test can be. Third-generation detectors aim to resolve multiple modes per event, converting ringdown spectroscopy into a precision test of the Kerr geometry.

• Orbits in Schwarzschild geometry and solar-system tests [13.05.02, 13.05.03]. The binary inspiral waveforms the detector cross-correlates against are post-Newtonian expansions of the orbital dynamics, and the late inspiral probes separations and velocities where the Schwarzschild effective-potential analysis of 13.05.02 is the leading-order structure. The standard-siren Hubble constant measured from GW170817 is the gravitational-wave analogue of the electromagnetic distance ladder underpinning the solar-system and binary-pulsar tests of 13.05.03, and the two routes to $H_0$ are now in productive tension that may signal new systematics or new physics.

Historical & philosophical context Master

The idea of detecting gravitational waves with interferometry rather than with resonant bars is due to Rainer Weiss, who in 1972 wrote an MIT quarterly progress report setting out the noise budget — seismic, thermal, shot noise — for a Michelson interferometer with suspended test masses, essentially the architecture that would become LIGO ^[LIGO2009]. Weber's earlier resonant-bar programme had shown that gravitational-wave detection was conceptually possible but sensitivity-limited; Weiss's insight was that an interferometer's strain sensitivity scales with arm length and can be improved indefinitely by lengthening the arms and increasing the stored light, whereas a bar is bounded by its material $Q$ and dimensions. Kip Thorne and Ronald Drever developed the Fabry-Perot arm-cavity concept, power recycling, and the theoretical noise analysis through the 1980s, and the LIGO project was funded by the United States National Science Foundation in 1992 as two 4-km interferometers in Hanford, Washington and Livingston, Louisiana — designed from the outset as a coincident pair to reject local disturbances. Initial LIGO operated from 2002 to 2010 without detection; Advanced LIGO, with quadruple-pendulum suspensions, higher circulating power, and upgraded test masses, began its first observing run on 12 September 2015 and detected GW150914 two days later ^[GW150914].

The global network grew in parallel. Virgo, a 3-km interferometer near Pisa designed by the CNRS-INFN collaboration, came into scientific operation in 2017, in time to enable the GW170817 localisation. KAGRA, the 3-km cryogenic underground detector in the Kamioka mine operated by the University of Tokyo and KEK, pioneered two technologies — underground siting and cryogenic test masses — that third-generation designs will adopt wholesale. LIGO-India, under construction at Aundha, will complete the third LIGO and give the network a long southern-hemisphere-quality baseline for polarisation and sky-localisation reconstruction. The Weiss-Barish-Thorne 2017 Nobel Prize recognised the conception and construction of LIGO; the field has since grown into a routine observational discipline with hundreds of detections across the LIGO-Virgo-KAGRA transient and continuous-wave catalogues.

Philosophically the interferometer is the instrument that turned general relativity into an empirical science of the radiative regime. Before LIGO, gravitational waves were inferred only through their cumulative dynamical effect on the Hulse-Taylor binary pulsar; after GW150914, each merger is a direct, time-resolved measurement of dynamical spacetime geometry. After GW170817, each binary-neutron-star merger is a multi-messenger event read simultaneously in gravitational waves, gamma rays, optical, and infrared — a single physical process observed across many orders of magnitude in electromagnetic wavelength plus the gravitational channel. The transition from prediction to observation converts the linearized-GR and Bondi-radiation frameworks from internally consistent theories into tested descriptions of nature, and it opens the strong-field, high-velocity regime (where $v/c\sim 0.5$ at merger) to direct experimental scrutiny for the first time since general relativity's founding.

Bibliography Master

1. Abbott, B. P. et al. (LIGO Scientific Collaboration), "LIGO: the Laser Interferometer Gravitational-Wave Observatory," Reports on Progress in Physics 72 (2009), 076901. [The canonical instrument paper: Michelson plus Fabry-Perot arm cavities, power and signal recycling, the design noise budget.]

2. Abbott, B. P. et al. (LIGO Scientific Collaboration and Virgo Collaboration), "Observation of Gravitational Waves from a Binary Black Hole Merger," Physical Review Letters 116:6 (2016), 061102. [GW150914: the first direct detection; demonstrated strain sensitivity of order $h\sim 10^{-21}$.]

3. Abbott, B. P. et al., "GW170817: Observation of Gravitational Waves from a Binary Neutron Star Inspiral," Physical Review Letters 119:16 (2017), 161101. [The binary-neutron-star inspiral signal and chirp-mass measurement.]

4. Abbott, B. P. et al. (LIGO, Virgo, Fermi-GBM, INTEGRAL), "Gravitational Waves and Gamma-Rays from a Binary Neutron Star Merger: GW170817 and GRB 170817A," Astrophysical Journal Letters 848:2 (2017), L13. [The 1.7 s gamma-ray delay; the $|v_g-c|/c\lesssim 3\times 10^{-15}$ bound on the speed of gravitational waves.]

5. Abbott, B. P. et al. (LIGO, Virgo, and partner observatories), "Multi-messenger Observations of a Binary Neutron Star Merger," Astrophysical Journal Letters 848:2 (2017), L12. [The kilonova in NGC 4993; the standard-siren Hubble constant; the network localisation to 28 deg${}^2$.]

6. Tse, M. et al. (LIGO Scientific Collaboration), "Quantum-Enhanced Advanced LIGO Detectors in the Era of Gravitational-Wave Astronomy," Physical Review Letters 123:23 (2019), 231107. [Injection of squeezed vacuum into the dark port; demonstrated broadband quantum enhancement of a gravitational-wave detector.]

7. Acernese, F. et al. (Virgo Collaboration), "Advanced Virgo: a second-generation interferometric gravitational wave detector," Classical and Quantum Gravity 32:2 (2015), 024001. [The 3-km European detector; design sensitivity and recycling configuration.]

8. Akutsu, T. et al. (KAGRA Collaboration), "Overview of KAGRA: Detector design and construction history," Progress of Theoretical and Experimental Physics 2021:5 (2021), 05A101. [The 3-km underground, cryogenic Japanese detector.]

9. Somiya, K. (for the KAGRA Collaboration), "Detector configuration of KAGRA — the Japanese cryogenic gravitational-wave detector," Classical and Quantum Gravity 29:12 (2012), 124007. [The cryogenic sapphire test-mass design.]

10. Unnikrishnan, C. S., "IndIGO and LIGO-India: Scope and plans for gravitational wave research and precision measurements in India," International Journal of Modern Physics D 22:1 (2013), 1341010. [The third LIGO interferometer; the global-network vision.]

11. Braginsky, V. B., Gorodetsky, M. L., Khalili, F. Ya., Matsko, A. B., Thorne, K. S. & Vyatchanin, V. I., "The noise in gravitational wave detectors and its classical quantum nature," Physics Letters A 286 (2001), 175-182. [The standard quantum limit, the measurement-noise/back-action tradeoff, and squeezed-state evasion.]

12. Saulson, P. R., Fundamentals of Interferometric Gravitational Wave Detectors, 2nd ed. (World Scientific, 2017). [The canonical textbook: interferometer optics, shot noise, radiation pressure, the SQL, thermal and seismic noise, at intermediate and master depth.]

13. Maggiore, M., Gravitational Waves: Theory and Experiments, Vol. 1 (Oxford University Press, 2008). [Chapters 7-8 give the interferometer response function and the full noise budget at master-level rigour.]

14. Weiss, R., "Electromagnetically coupled broadband gravitational antenna experiment," MIT Quarterly Progress Report 105 (1972), 54-76. [The original interferometric-detector noise budget; the conceptual origin of LIGO.]

15. Abbott, B. P. et al., "A gravitational-wave standard siren measurement of the Hubble constant," Nature 551 (2017), 85-88. [The $H_0\approx 70_{-8}^{+12}$ km/s/Mpc standard-siren measurement from GW170817.]