Data-driven scale identification in oscillatory dynamos

Abstract

ABSTRACT Parker’s mean-field model includes two processes generating large-scale oscillatory dynamo waves: stretching of magnetic field lines by small-scale helical flows and by differential rotation. In this work, we investigate the capacity of data-driven modal analysis, dynamic mode decomposition (DMD), to identify coherent magnetic field structures of this model. In its canonical form, the only existing field scale corresponds to the dynamo instability. To take into account multiscale nature of the dynamo, the model was augmented with coherent in time flow field, forcing small-scale magnetic field with a faster temporal evolution. Two clusters of DMD modes were obtained: the ‘slow’ cluster, located near the dynamo wave frequency and associated with its non-linear self-interaction, and the ‘fast’ cluster, centred around the forcing frequency and resulting from the interaction between the wave and the flow. Compared to other widely used methods of data analysis, such as Fourier transform, DMD provides a natural spatiotemporal basis for the dynamo, related to its non-linear dynamics. We assess how the parameters of the DMD model, rank, and delay, influence its accuracy, and finally discuss the limitations of this approach when applied to randomly forced, more complex dynamo flows.