Fig. 2

Freezing bubbles under chilled and isothermal conditions. a Freezing of a 10-mL bubble deposited on an ice disk (Supplementary Fig. 3) in a walk-in freezer chilled to T_∞ = −18.5 ± 0.5 °C with RH = 60 ± 5%. b The freeze front induced a Marangoni flow, which detached and entrained some of the growing ice particles. c Time-lapse thermographic images, where dotted arcs clarify the bubble–air interface. The liquid portions of the bubble assumed the freezer’s temperature shortly after deposition, while the freeze fronts were warmer (i.e., near the melting temperature) due to the release of latent heat. The emissivity coefficient of ice was calibrated at ε = 0.98. Time zero corresponds to the bubble’s first contact with the icy substrate, where the top of the bubble is still adhered to the pipette (first frames of a–c). d Displacement (δ) of four thermal plumes (different colors) was measured over time when T_w ≈ T_∞ = −19.6 °C. The inset shows that the radius of plumes was of order R_p ~1 mm. The scale bar represents 2 mm. e Growth rate of freeze fronts coming from the substrate (red data points, v_i ≈ 300 μm s⁻¹) or from ice crystals suspended in the liquid film (green, v_i ≈ 250 μm s⁻¹) when T_∞ = −18.4 ± 1.7 °C. Error bars represent a standard deviation from an average of three trials. f The ice radius, R_i, was estimated from the tip of ice crystals growing from the substrate