AI-generated illustration for Galileo Whispers. Conceptual test masses in an imagined chamber, not a measurement of gravitational-wave memory or a reconstruction of an instrument.
Gravitational-wave memory is a lasting change in the relative separation of ideal, freely falling test masses after a gravitational wave passes. The oscillating part of a signal can fade while a change between the early and late configurations remains. That residual is the memory; it is not light stored in space or a record kept by a computer.
The word ideal matters. A suspended mirror in a working observatory does not behave like an unconstrained marker forever. Understanding that difference separates a prediction about spacetime from a claim about what an instrument must visibly retain.
A ripple need not end where it began
A familiar sketch of a gravitational wave shows alternating stretching and squeezing. If that oscillation returns entirely to its starting level, freely falling markers can finish with their original relative separation. A signal with memory has an additional component whose early and late levels differ.
Marc Favata’s 2010 review distinguishes this nonoscillatory change from the faster waveform usually associated with an orbiting binary. A black-hole merger can produce both. Looking only for a succession of peaks can therefore miss part of the prediction.
The relevant comparison is between physical separations under specified conditions, not between arbitrary dots drawn on a changing coordinate grid. Coordinates are labels; they can be chosen in different ways. A detector calculation must connect the predicted signal to an actual relative-distance or timing observable.
Nor does a lasting relative displacement imply an unending local shove. The change can remain after the wave’s changing tidal influence has passed, much as a final position can differ without a force continuing to act. The analogy concerns the distinction between a change and continued acceleration, not a mechanical model of spacetime.
Radiation can contribute to gravity
There are different contributions to gravitational memory. The motion of matter can produce what is often called ordinary or linear memory. General relativity also predicts nonlinear memory associated with energy carried away by gravitational radiation itself.
That second contribution is sometimes called Christodoulou memory. It reflects the fact that gravitational waves carry energy and participate in the gravitational field, rather than being disturbances that travel on an entirely unaffected stage.
Favata’s work on post-Newtonian memory contributions develops the nonoscillatory part for inspiralling compact binaries. Its mathematical expansion is not needed to appreciate the conceptual distinction: the rapidly changing orbital motion and the accumulated radiation can leave different features in the received waveform.
Their sizes depend on the source, viewing direction and the component of motion being measured. The phrase “permanent change” does not assign one universal displacement to every detector or every merger.
A fractional change can be extraordinarily small
Consider a purely illustrative pair of freely falling markers initially separated by 1,000 meters. If their final separation differs by one part in a thousand billion billion, or 10^-21, the difference is 10^-18 meter.
That arithmetic is simply 1,000 meters multiplied by 10^-21. It is not a measurement, a prediction for a named merger, or a complete conversion from a particular gravitational-wave polarization to an instrument reading. The source geometry and detector response must supply that connection.
The example explains why lasting does not mean large. A residual can be physically meaningful yet remain extremely difficult to distinguish from drift, calibration errors or other low-frequency behavior.
Real detectors have restoring forces and noise
Ground-based interferometer mirrors are suspended, controlled and monitored. Their supports and feedback systems matter over long times. A permanent displacement of ideal free masses should not be pictured as an observatory’s mirrors staying indefinitely in a new, plainly visible position.
Memory also builds over a finite time. A detector’s response varies with frequency, so a slow change can be less accessible than the faster oscillations in the same event. Researchers need a model of the source and of the measuring system.
A 2016 study by Paul Lasky and colleagues examined combining information from multiple mergers. The predicted signals cannot simply be added without considering their orientation and sign: contributions can otherwise cancel rather than reinforce one another.
That study proposed a detection strategy under stated assumptions. Its forecasts are not a current event count or evidence that a particular detection has occurred. The durable idea is the measurement problem: a wave can have a lasting theoretical consequence without giving a real detector an easily readable permanent offset.
