Before an eruption, volcanoes do not accelerate smoothly. They stutter. Seismicity spikes, then drops to near-baseline. Gas emissions surge, then quiet. Ground deformation accelerates, pauses, resumes. This intermittency has been the central obstacle to eruption forecasting: the same patterns that precede catastrophic eruptions also precede nothing. Standard models fit power laws to the acceleration phase and extrapolate to a singular failure time, but the intermittent pauses and restarts violate the assumption of monotonic approach to criticality. False alarms accumulate. Confidence erodes.
Lei and Sornette analyzed 34 historical eruptions and found that the stutter is not noise. It is a log-periodic oscillation decorating the power law approach to failure — the signature of discrete scale invariance in the damage process. The key mathematical move: extend the power law exponent from real numbers to complex numbers. A real exponent produces smooth acceleration. A complex exponent produces acceleration modulated by oscillations that are periodic in the logarithm of time-to-failure — each cycle of activity and quiescence is compressed by a constant ratio as the eruption approaches. The pauses get shorter. The accelerations get sharper. The rhythm is precise.
The physical mechanisms generating this discrete scale invariance include magma-driven dyke propagation through heterogeneous rock, where stress concentrates at crack tips, triggers damage in a surrounding zone, and then pauses as the stress redistributes — each pause-restart cycle operating at a smaller spatial and temporal scale than the last. Stress corrosion, inertial effects, and the interplay of damage and partial healing all contribute to the same structure: a cascade that proceeds not continuously but in geometrically shrinking steps.
The practical implication is a reversal of what intermittency means. In the standard framework, a volcano that goes quiet after a seismic swarm is reassuring — the crisis may be passing. In the log-periodic framework, the quiet period is the next expected phase of the oscillation. Its duration, relative to the previous active phase, contains information about when the next acceleration will begin and how close the system is to failure. The stutter is the forecast.
The general principle: intermittency can encode more information than smooth progression. A system that accelerates monotonically tells you one thing — that it is accelerating. A system that accelerates in discrete, geometrically compressed bursts tells you the ratio between bursts, the number of remaining cycles, and the approximate time of failure. The irregularity is not the obstacle to prediction. It is the prediction.