Cosmologists disagree about how fast the universe is expanding, and the disagreement has hardened past the point where better statistics alone will resolve it. Redshift drift offers a way out that does not require picking a cosmological model in advance. It measures expansion directly, and it is hard enough that no one has yet claimed a clean detection.

The Model-Dependent Expansion Picture

The Hubble tension is the persistent mismatch between expansion rates inferred from the early universe and those measured nearby. Values from the cosmic microwave background, when fitted through a standard cosmological model, land near 67 kilometres per second per megaparsec. Distance-ladder measurements using supernovae land nearer 73. The gap sits at roughly five sigma, which is the threshold particle physicists normally treat as a discovery.

Both numbers depend on assumptions. Baryon acoustic oscillations and supernova distances are interpreted through a model that specifies what dark energy is and how it behaves. If that model is wrong in some subtle way, the tension could be an artifact of the interpretation rather than a feature of the universe.

Allan Sandage proposed an escape route in 1962. A galaxy's redshift should change slowly as the universe expands, because the expansion alters the recession velocity between the light's emission and its arrival. Measuring that change over decades would give a direct handle on the expansion rate and its time derivative, with no dark energy model inserted by hand.

How Redshift Drift Works

The signal is a velocity change, not a velocity. A galaxy at cosmological distance recedes at some speed today, and a spectrograph measures that speed from the shift of known spectral lines. Twenty years later, the same galaxy recedes slightly faster or slower, and the shift has moved. The difference is the drift.

Its size is brutal. Typical estimates put the expected drift at centimetres per second per decade for distant quasars, and smaller still for lower-redshift targets. That is comparable to the reflex motion a spectrograph sees when an Earth-like planet tugs on a Sun-like star. The measurement demands stability over decades, not just precision on a single night.

Two observing routes dominate. The Lyman-alpha forest, the thicket of hydrogen absorption lines imprinted by intergalactic gas on quasar spectra, offers many lines at high redshift where the drift signal is largest. Twenty-one centimetre absorption against radio sources offers an independent path, with different systematics, but needs radio arrays with extraordinary spectral stability.

Current Instruments and Surveys

ESPRESSO, the high-resolution spectrograph on the Very Large Telescope, was designed partly for this problem. It reaches centimetres-per-second precision on bright stars and has been used for quasar monitoring, though the drift signal sits at the edge of what a single instrument can deliver in a realistic campaign.

The Square Kilometre Array and the Extremely Large Telescope are the next bets. SKA's low-frequency arrays could monitor the twenty-one centimetre line across large samples, while ELT instruments would extend high-resolution optical spectroscopy to fainter quasars. Designs such as CODEX and HiRES were proposed as dedicated drift machines, and their science cases shaped how the current generation was specified.

Several groups now run decade-scale monitoring programmes, reobserving the same quasars with the same instrument to suppress calibration differences. This is slow work. A programme started in 2015 is only now reaching the baseline where a drift signal could begin to separate from noise. A related piece on Rubin Observatory's repeated survey design makes a similar point about the value of cadence over one-off depth.

What the Test Can and Cannot Do

What redshift drift delivers is a measurement of the expansion history that does not presuppose dark energy. The drift rate depends on the cosmological parameters, but the observable itself is a kinematic fact about the light. That makes it a genuinely independent check on the model-dependent inferences.

What it cannot do, at present, is distinguish between competing models. A single drift measurement at one redshift constrains a combination of parameters, and several dark energy models predict drift values that differ by less than current error bars. Distinguishing them requires measurements across a range of redshifts, each with small errors.

Systematics are the real obstacle. Wavelength calibration drifts, instrument flexure, and changes in the Earth's atmosphere all mimic a slow velocity shift. The Lyman-alpha route has the advantage that intergalactic absorption lines are narrow and numerous, but the disadvantage that the lines themselves can shift if the gas clouds move. A piece on seafloor pressure sensors and their resolution trade-offs covers a similar tension between coverage and precision in another measurement context.

The Road to a Detection

Statistical error dominates today. Even with perfect instruments, the shot noise from finite numbers of photons and finite samples of quasars limits how well a drift can be pinned down. Expanding the sample is the most direct lever, which is why survey design matters as much as spectrograph precision.

Next-generation facilities should push the combined error budget down by an order of magnitude over the coming decades. That is the scale required for a first detection, and it explains why no group has announced one. The honest position is that the test is feasible, the instruments are close, and the baselines are not yet long enough.

A first detection, when it comes, may take twenty years of accumulated data. It would give a model-independent constraint on the expansion rate and its change with time. It would not, on its own, resolve the Hubble tension. It would tell cosmologists whether the expansion history looks the way the standard model says it should.

What to Watch in the Coming Decade

Follow ESPRESSO drift papers as their quasar baselines cross the ten-year mark. The first results will be error-dominated, but the trend in the error bars is the thing to read.

Watch SKA progress on twenty-one centimetre spectroscopy, particularly any published spectral stability figures. Those numbers determine whether the radio route can compete with optical campaigns.

Track ELT instrumentation milestones, especially the high-resolution spectrographs planned for first light. Their wavelength calibration architecture is a better predictor of drift performance than their aperture.

Support long-baseline spectroscopy in facility reviews. The scientific case for redshift drift rests on patience, and patience is the first thing budget cycles cut. The related question of how methods sections get weighted in review is covered in this site's piece on grant review panels and methods scoring.

Why the Signal Is So Faint

The drift is small because the universe expands slowly compared with the light travel time across a galaxy. For a source at redshift one, the expected velocity change is on the order of a few centimetres per second over a decade. That is roughly the speed of a tortoise, and it must be extracted from a spectral line that may be broadened by the galaxy's internal motions to hundreds of kilometres per second. The ratio of signal to line width is one part in ten million. No spectrograph can resolve that in a single exposure; the measurement relies on comparing two epochs separated by many years, and on the assumption that the instrument itself has not changed in any way that mimics a shift.

The Systematics Battle

Wavelength calibration is the dominant systematic for optical drift measurements. A spectrograph is a physical object that flexes, warms, and cools. A change in temperature of a fraction of a degree can shift the dispersion relation enough to imitate a drift signal. To beat this, instruments like ESPRESSO use a laser frequency comb, which provides a grid of equally spaced lines that can be tracked to high precision. Even with a comb, the calibration must be stable over the years between observations, and the comb itself must be monitored for drift. Radio measurements face a different set of problems: the ionosphere and the interstellar medium can both shift the apparent frequency of a spectral line, and the calibration of the array's bandpass must be stable to a part in a million or better. These are engineering challenges, not fundamental limits, but they explain why the first detection is taking so long.

What a Detection Would Mean

A clean detection of redshift drift would be a direct confirmation that the universe is expanding in the way general relativity predicts, without the need to assume a particular dark energy model. It would also provide a measurement of the deceleration parameter, which describes how the expansion rate changes with time. That quantity is currently inferred from supernova data and baryon acoustic oscillations, both of which depend on the cosmological model. A drift measurement would be independent, and if it disagreed with the model-dependent values, it would force a rethink of the standard model. If it agreed, it would strengthen the case that dark energy is a cosmological constant. Either way, the result would be significant.