How We Know
Physics as Experiment
Cleverness Is the Tool
A popular image of physics is equations on a chalkboard. The reality is engineering. To measure a gravitational wave, you build a four-kilometer interferometer, suspend forty-kilogram mirrors from pendulums filtering out seven stages of seismic noise, reduce the vacuum inside to one-trillionth of atmospheric pressure, stabilize a laser to a wavelength that shifts less than the size of a proton over the course of a run, inject squeezed quantum light to dip below the shot-noise limit, and then wait for years for the right astrophysical event to produce a signal above your systematic uncertainties. Each of those engineering problems could have killed the project. All of them had to be solved simultaneously for any of them to matter.
Physics is an experimental science. Every number on every page of this site – the mass of a proton, the age of universe, the distance to a galaxy, the polarization of the cosmic microwave background – exists because somebody figured out how to measure it. Sometimes the experiment is simple and elegant. Often it requires decades of effort and inventing new physics to make the instruments work. This page is a tour of how some of the most famous modern measurements actually happen. The underlying principle is always the same: find a subtle, reliable way for the thing you want to measure to leave an imprint on something you can detect with enough precision.
Laser Interferometry
LIGO is the best illustration of modern experimental physics. Two perpendicular laser arms, each four kilometers long, have mirrors at each end. A laser beam is split and sent down each arm. The beams bounce back, recombine, and interfere. Normally the path lengths are tuned so the recombined beam is destructive – dark. When a gravitational wave passes through, it stretches one arm and shrinks the other by a tiny amount, the beams recombine out of phase, and a faint flash of light leaks out of the dark port. That flash is the signal.
The numbers are astonishing. LIGO detects strains of about one part in 1021. Over 4 kilometers, that is a length change of about 10-18 meters – less than one-thousandth the diameter of a proton. To measure this, every mirror hangs on a chain of pendulums that filter seismic noise. The laser power is amplified in resonant optical cavities to push past photon shot noise. The beam path is enclosed in one of the largest ultra-high vacuums ever built, 10,000 cubic meters pumped to 10-9 atmospheres so air molecules cannot scatter the beam. Each of these solves a different noise source that would otherwise drown the signal. The trick of the interferometer is to turn a measurement of absolute length into a measurement of interference fringes, which are much more sensitive than direct length measurements. Every experimental physicist owes something to Michelson and Morley, who built the first one in 1887.
Neutrino Detectors
Neutrinos almost never interact with matter. Roughly 1014 solar neutrinos pass through every square meter of your body every second, and essentially none of them are stopped. To detect them, you have to build a target so large that even one-in-a-quintillion interactions produce detectable events. Super-Kamiokande in Japan is 50,000 tons of ultra-pure water a kilometer underground, lined with 11,000 photomultiplier tubes that can detect single photons. When a neutrino interacts in the water, it knocks out a charged particle moving faster than light can move through water. That particle emits a cone of Cherenkov radiation, visible to the PMTs as a ring. The shape and timing of the ring tell you the neutrino's direction and flavor.
IceCube goes even bigger. A cubic kilometer of Antarctic ice, seeded with 5,160 photomultiplier modules on strings running 2.5 km deep. Any high-energy neutrino that interacts anywhere in that volume produces a Cherenkov trail, and the timing of photon arrivals at different modules reconstructs the trajectory. IceCube was the first detector to reliably identify astrophysical neutrinos arriving from sources outside our solar system and even outside our galaxy. Neutrinos cannot be focused or imaged in the normal sense; the telescope is the ice itself, and the detector infrastructure reads its flickers. This is neutrino astronomy, and it has become the first routine application of particles other than photons to observe the cosmos.
Particle Colliders
The LHC is a 27-kilometer ring of superconducting magnets cooled to 1.9 kelvin, colder than outer space, steering protons at 99.999999% of the speed of light. Beams crossing at designated interaction points produce collisions at a combined energy of 13.6 TeV. Each collision generates a spray of particles, and detectors like ATLAS and CMS – cathedrals of silicon, scintillator, and calorimeter material – record every charged particle's trajectory, every neutral particle's energy deposition, and every muon that punches through to the outer layers. Roughly one billion collisions happen per second at each interaction point. Trigger systems filter this down to about a thousand per second that might contain something interesting, which are then saved for analysis.
Finding the Higgs boson in 2012 meant identifying a few dozen events per year matching a specific signature among those trillions of routine collisions. It was done by carefully calibrating the detector, understanding the Standard Model backgrounds, and looking for a small statistical excess above them at the energies where the Higgs should decay. The entire apparatus – the machine, the detectors, the computing grid, the analysis pipeline – exists to turn "we want to know if this specific particle is real" into "here is a bump in a histogram." Modern particle physics looks like a vast analysis of digital data; the engineering is mostly about producing clean enough data for that analysis to reveal the signal.
Atomic Clocks
The most precise measurements in physics are of frequencies. A cesium-133 atom transitioning between two hyperfine states emits microwave radiation at an extremely well-defined frequency: 9,192,631,770 hertz. This is so stable that the second is defined by it. Modern optical atomic clocks have done even better, using transitions in trapped single ions of aluminum, ytterbium, or strontium. State-of-the-art clocks reach fractional frequency stability better than one part in 1018. They would drift by less than one second over the current age of universe.
This precision enables measurements that would be hopeless otherwise. Two atomic clocks placed one meter apart in the same building tick at measurably different rates, because Earth's gravity makes time flow slightly slower closer to the ground. The difference is a part in 1016, which is detectable in about a day of integration. Geodesy – mapping Earth's gravity field – can now be done by comparing clocks at different locations. GPS depends on atomic clocks on satellites applying general relativity corrections to nanoseconds of timing; without those corrections, your phone's position would drift by kilometers per day. The precision of the clocks is what turns GPS from a mathematical idea into engineering you carry in your pocket.
The Noise Floor
Every measurement is a contest between signal and noise, and noise has a fundamental floor that no engineering can eliminate. Thermal noise: any object above absolute zero has atoms jittering, producing random electrical and mechanical fluctuations. Photon shot noise: light arrives as discrete photons, so even a perfectly clean beam has statistical fluctuations proportional to one over the square root of the photon count. Quantum vacuum noise: empty space itself fluctuates, setting a hard limit on how precisely you can measure conjugate observables simultaneously, exactly as the uncertainty principle demands.
These limits are not problems to engineer past. They are properties of nature. What experimenters do instead is structure the measurement so the noise becomes irrelevant. Cool detectors to nanokelvin temperatures to suppress thermal noise. Use squeezed light to reduce shot noise in one observable at the cost of more in another – LIGO already does this routinely. Average over millions of trials so random fluctuations shrink with the square root of the count. Modulate the signal at a frequency where local noise is quiet and use lock-in detection to pull it out.
Knowing the noise floor of your apparatus is more useful than knowing its peak performance. It tells you what cannot be detected with what you have. It tells you when more averaging is wasted effort because you have already reached the fundamental limit. It tells you exactly how much improvement is possible before quantum mechanics itself becomes the bottleneck. Modern precision physics has been pushed to that bottleneck on multiple fronts – gravitational waves, atomic clocks, gravimeters – and the techniques developed to push past classical limits have become a sub-discipline of their own.
Spectroscopy From Distant Stars
A star is a ball of glowing plasma hundreds of light-years away. We have never and will never bring a sample of it back to a lab. Yet astronomers know, often to high precision, what stars are made of. The method is spectroscopy. Pass starlight through a prism or diffraction grating and you get a rainbow, but with thin dark lines at specific wavelengths where atoms in the star's atmosphere absorbed the light as it escaped. Each element leaves a characteristic pattern of lines. Match the pattern to lab measurements and you know what atoms are in that star. Hydrogen, helium, iron, magnesium, barium – they all have unmistakable fingerprints.
Spectroscopy works across every branch of astronomy. The CMB's temperature is extracted from the shape of its blackbody spectrum. Exoplanet atmospheres are characterized by measuring how starlight is filtered when the planet transits in front of its star. The expansion of universe is calibrated by the redshift of absorption lines in distant quasars. Helium was discovered in the Sun 27 years before it was found on Earth, purely by spotting spectral lines nobody recognized. The physics is simple: atoms absorb and emit specific wavelengths. The engineering is building telescopes, gratings, and detectors sensitive enough to recover the full spectrum even from the faintest sources.
Dating the Earth and the Universe
Geologists date rocks by measuring the ratio of radioactive isotopes to their decay products. Uranium-238 decays into lead-206 with a half-life of 4.5 billion years. A rock containing zircon crystals – which incorporate uranium but exclude lead at formation – can be precisely dated by measuring the U/Pb ratio. The oldest known terrestrial minerals are 4.4 billion years old. The oldest meteorites, presumed to date from solar system formation, come in at 4.567 billion years, and that number is the age of our solar system to better than 0.1%.
Carbon-14 dating works for shorter timescales. Atmospheric carbon includes a tiny fraction of carbon-14, produced by cosmic rays and decaying with a half-life of 5,730 years. Living things continuously exchange carbon with the atmosphere, maintaining the same ratio. When they die, the exchange stops, and the carbon-14 slowly decays. Measure the ratio in an organic sample and you get the time of death. This method has dated the Shroud of Turin, the Dead Sea Scrolls, and countless archaeological samples. The age of universe itself comes from fitting cosmological models to the cosmic microwave background, Type Ia supernovae, baryon acoustic oscillations, and large-scale structure, with multiple independent methods converging on 13.8 ± 0.02 billion years. We know when the big bang happened more precisely than we know many details of ancient history.
Cross-Checks and Triangulation
No single measurement is ever trusted in isolation. Every important number in physics has been measured by multiple independent methods that in principle could have disagreed. The speed of light has been measured by interferometry, by resonating cavities, by stellar aberration, by Roemer's observations of Jupiter's moons, and by the ratio of electromagnetic constants. All agree. The fine structure constant has been measured via the electron's magnetic moment, via atomic spectra, via recoil from photon absorption, via the quantum Hall effect. All agree. The Hubble constant has been measured by Type Ia supernovae, by Cepheid variables, by the CMB, by gravitational wave "standard sirens," by the tip of the red giant branch. These mostly agree, and where they do not agree – the Hubble tension – it is treated as a live research question rather than dismissed.
This redundancy is how science catches systematic errors. If two methods disagree by more than their stated uncertainties, at least one of them is wrong. The discrepancy triggers work to find the error. Sometimes the original method turns out to have a subtle bias nobody spotted; sometimes the newer method has a calibration issue; sometimes both were right and the disagreement reveals genuinely new physics. The Hubble tension might resolve in any of these ways; we do not know yet. The important point is that modern physics does not trust any single measurement, no matter how famous the experiment or how careful the analysis. Confirmation requires triangulation.
What We Cannot Measure
Some things look measurable but turn out not to be. The graviton, the hypothetical quantum of gravity, is consistent with every measurement made but will likely never be detected individually. Freeman Dyson pointed out that a detector sensitive enough to register a single graviton interaction would need to be so massive it would collapse into a black hole before detection. This is a hard physical obstacle, not an engineering one. The interior of a black hole past the event horizon is causally disconnected from us; no signal can carry information out. The state of universe beyond the cosmological horizon – the region from which light has not had time to reach us – is similarly inaccessible, and may remain so forever because the accelerating expansion keeps more of it receding out of reach.
Knowing what cannot be measured is itself useful physics. It tells you where to stop looking for direct evidence and start relying on theoretical consistency. A quantum theory of gravity is unlikely to be confirmed by detecting a graviton; it will be confirmed, if at all, by its correct predictions for things we can measure, like modifications to black hole evaporation or structures in the CMB. The map of the measurable and the unmeasurable is as important as the map of the known and the unknown.
The Bigger Picture
Modern physics is the product of a million small experimental victories over noise. Every detector is a device for separating a tiny signal from enormous backgrounds. Every measurement is a triumph over some source of error that could have ruined it. The equations on the chalkboard are the finished product; the labs, the telescopes, the accelerators, the atomic clocks, the detectors buried under mountains of rock and ice, are what let those equations be compared with universe. Without that engineering, physics would be speculation. With it, physics is a rigorously tested account of reality at every scale we have been able to reach.
Every page of this site presents physics as a set of ideas that can be understood. What this page adds is that those ideas have been relentlessly, exhaustively checked, and every one of them has passed. The fine structure constant has been measured to twelve decimal places. The cosmic microwave background's temperature fluctuations agree with inflation's predictions to sub-percent precision. The orbital decay of the Hulse-Taylor binary pulsar matches general relativity to better than 0.2% over thirty years of observation. This is not accidental. This is what it looks like when a community of experimenters, spanning continents and generations, does its work. Physics is not primarily a theoretical field. It is the experimental discipline that figured out how to discover the theoretical field.
