Doppler spectroscopy

• 1. Introduction

  ›

  Doppler spectroscopy is the technique that has revealed more than 1,100 worlds beyond our solar system, accounting for roughly 19 percent of all known extrasolar planets as of January 2026. It works by watching a star rather than a planet, listening for the faintest tremor in the light a star emits. That tremor tells astronomers something enormous is tugging at the star from the dark.

  The idea is disarmingly simple. When a planet orbits a star, the two objects do not hold still. They orbit their shared center of mass, and the star itself shifts slightly back and forth. As it moves toward Earth, its light compresses into slightly bluer wavelengths. As it moves away, the light stretches toward red. This cycle repeats with the same period as the planet's orbit, and a sensitive enough instrument can read it like a heartbeat.

  But the signals are extraordinarily small. Jupiter, a giant among planets, nudges our Sun by only about 12.4 meters per second over a twelve-year span. Earth's tug on the Sun amounts to just 0.1 meters per second across a year. For most of human history, no instrument existed that could hear those whispers. The questions at the heart of this story are about how we built ears sensitive enough to listen, what those ears found, and where their limits still lie.

• 2. A 1952 Prediction

  ›

  Otto Struve put the idea into print in 1952, decades before any instrument could act on it. He reasoned that a planet the size of Jupiter would pull its host star into a perceptible wobble, and that the resulting Doppler shifts would appear as tiny redshifts and blueshifts moving through the star's emission spectrum in a regular rhythm.

  His reasoning was sound, but the technology of his time returned radial-velocity errors of 1,000 meters per second or more. The signal Struve was hunting sat far below that noise floor. Jupiter's effect on the Sun is roughly 12.4 meters per second; Earth's effect is a hundred times smaller still. The gap between what spectrographs could measure and what planets actually produce was so vast that the idea lay mostly dormant for decades.

  Struve's paper nonetheless planted a precise and testable claim at the center of planet-hunting research. It specified what to look for, what instruments would need to achieve, and which class of planet would be easiest to detect first: large ones, orbiting close to their stars, producing the strongest gravitational tug. That framework would guide every spectrograph designer who followed him.

• 3. The ELODIE Discovery

  ›

  In 1993, a new spectrograph called ELODIE was installed at the Haute-Provence Observatory in Southern France. It could resolve radial-velocity shifts as small as 7 meters per second, a threshold that brought nearby Jupiter-scale planets into reach for the first time.

  Astronomers Michel Mayor and Didier Queloz trained ELODIE on a star in the constellation Pegasus. What they found was a planet orbiting astonishingly close to its host star, at a distance of just 0.05 astronomical units, completing one orbit in 4.23 days. The planet, designated 51 Pegasi b, produced a radial-velocity signal of 55.9 meters per second in the host star. Nothing like it had been anticipated in theories of how planetary systems form.

  In November 1995, Mayor and Queloz published their findings in the journal Nature. The paper has since been cited more than 1,000 times. Planets had been detected orbiting pulsars before this, but 51 Pegasi b was the first planet confirmed around a main-sequence star, the first detected through Doppler spectroscopy, and the discovery that forced astronomers to rethink what solar systems could look like. After that publication, search programs launched at the Keck, Lick, and Anglo-Australian Observatories, as well as the Geneva Extrasolar Planet Search, began adding candidates rapidly.

• 4. Generations of Instruments

  ›

  First-generation spectrographs like ELODIE were suited to finding hot Jupiters, massive planets orbiting close to their stars. Saturn, with a radial-velocity signature of 2.75 meters per second at 9.58 astronomical units, already sat near the edge of what those instruments could reliably detect.

  The second generation arrived in the early 2000s. HARPS, installed at the La Silla Observatory in Chile in 2003, pushed the threshold down to 0.3 meters per second. That precision brought super-Earths into view. A planet like Gliese 581c, a super-Earth orbiting just 0.07 astronomical units from its host star, produces a signal of 3.18 meters per second, well within HARPS's reach.

  A third generation was expected to come online around 2017, with measurement errors projected below 0.1 meters per second. At that precision, an extraterrestrial astronomer using one of these instruments could, in principle, detect Earth itself. Earth produces only 0.089 meters per second of wobble in the Sun over its 365.26-day orbit. By 2025, Doppler spectroscopy had successfully extended into the near-infrared, achieving sub-meter-per-second precision in that range as well.

• 5. Reading the Signal

  ›

  Each Doppler spectroscopy measurement starts with a series of observations of a star's light spectrum. Astronomers watch for periodic variations in the position of characteristic spectral lines, wavelengths shifting regularly back and forth over time.

  Raw data carries contamination from other sources, so statistical filters are applied before any planet interpretation is attempted. Then mathematical best-fit methods are used to isolate the periodic sine wave that a planet in orbit would produce. A circular orbit generates a clean sine curve; eccentricity distorts the curve and complicates the calculations.

  From the amplitude of that curve, astronomers can derive the minimum mass of any orbiting planet, using a calculation anchored in the binary mass function. To find the true mass, not just the minimum, requires knowing how the orbital plane is tilted relative to Earth's line of sight. One method for finding that inclination is to detect the planet's own spectral lines separately from the star's; in 2012, that approach was applied to Tau Boötis b, where carbon monoxide was identified in the infrared spectrum, making it the first non-transiting planet to have its true mass determined this way.

  For cases where a clear spectral separation is not possible, radial-velocity data can be combined with astrometric measurements that track the star's movement across the sky rather than toward and away from Earth. That pairing also helps distinguish genuinely massive planets from brown dwarfs.

• 6. Where the Method Struggles

  ›

  Doppler spectroscopy measures only the component of a star's motion along the line of sight. If a planet's orbital plane is tilted away from that line, the measured radial velocity is smaller than the true one, and the calculated planet mass is an underestimate.

  Stars themselves add noise. Some stars have gas envelopes that expand and contract; others vary in their emission on their own. Either effect can swamp the tiny planetary signal the method is trying to extract, making those stars poor candidates for this kind of search. Beyond variability, stellar activity from the host star, things like starspots and other surface phenomena, can drown out a planetary signature or produce patterns that mimic one. Gaussian Process modeling has emerged as a tool for addressing this; it models the radial-velocity time series alongside stellar activity indicators, allowing researchers to separate the stellar contribution from any underlying planetary signal.

  The Bayesian Kepler periodogram offers another layer of scrutiny. Applied to the HD 208487 system, the method appeared to detect a second planet with an orbital period of roughly 1,000 days; the candidate may instead reflect stellar activity. A similar analysis of HD 11964 produced an apparent planet with a period close to one year, but re-reduced data failed to confirm it, pointing to the Earth's own orbital motion as the likely artifact. The threshold for confirmation remains demanding, and the difference between a real planet and a convincing illusion can hinge on how data are reduced.