Controlling State Selection in Multistable Turbulent Convection

Ziel

Turbulent convection in engineered, geophysical, and astrophysical systems can self-organise into multiple large-scale stable states that coexist under identical control parameters, yet transport heat and momentum very differently. The range of admissible states, the geometry of their basins of attraction, and the mechanisms of transition between them remain poorly understood, and no general framework exists to predict or control which state is realised. This unpredictability limits performance and reliability, from industrial heat exchangers and reactors to weather and climate models. The CoSMiC project will turn multistability from a liability into a capability. Focusing on canonical buoyancy-driven configurations, including magnetoconvection, Rayleigh-Benard, oscillatory-rotating, centrifugal, double-diffusive, and internally heated convection, CoSMiC will establish a framework to identify, predict, and actively select among coexisting turbulent states. CoSMiC will integrate theory with high-fidelity simulations to (i) map attractor landscapes and quantify transport across wide parameter ranges; (ii) uncover transition threshold ranges; and (iii) demonstrate minimal, physics-anchored feedback control that steers the flow into desired states. The CoSMiC methodology will combine targeted step-response tests and sensitivity mapping with a proportional feedback loop that utilises small, localised actuations (boundary patches, gentle geometry tweaks, or body-force inputs). When needed, a two-loop scheme will enhance robustness in strongly nonlinear regimes. The CoSMiC outcome will be a set of transferable design rules for locking turbulent convection into chosen states with vanishingly small ongoing input. The CoSMiC results will advance fundamental turbulence theory and provide practical levers for more energy-efficient heat-transfer devices and safer reactors, and in the longer term, improved representations of regime variability in weather and climate models.