Bifurcations and Convection Onset in a Horizontally Rotating Cylinder Partially Filled with Liquid

The motion of a liquid in a horizontally rotating cylinder seems straightforward, almost mundane. One might expect the liquid simply to spin with the container in a quiet, uniform way. But once air and liquid share the same space, that simplicity disappears. The rotation, gravity, and viscosity begin to compete, and the flow takes on a life of its own. Even when the cylinder turns slowly enough that the surface looks nearly flat, faint circulations start to form beneath it. Over time, these circulations organize into convection cells that roll steadily along the cylinder’s length, their axes cutting across the direction of rotation. What makes this so striking is how such order can arise from what appears, at first, to be a perfectly stable pool. This phenomenon has important applications and similar configurations are used routinely in coating, mixing, and other industrial systems, where understanding internal flow is essential. Convection cells can either help by improving mixing or, in some cases, hinder by creating unwanted gradients. Predicting when they appear has proven difficult, mostly because several factors act together. The Reynolds number governs the balance between inertia and viscosity; the filling ratio controls how much of the cylinder is actually occupied by liquid; and the Froude number describes how gravity interacts with centrifugal force. Changing one of these slightly can shift the flow from stable rotation to gentle oscillation, and finally to a state of fully developed convection. Capturing that transition, with all its subtle dependencies, remains one of the most intriguing challenges in rotating multiphase flow. To this account, new research paper published in Journal of Fluid Mechanics and conducted by Dr. Daiki Watanabe, Dr. Kento Eguchi and Professor Susumu Goto from the Graduate School of Engineering Science at The University of Osaka in Japan, the researchers developed two complementary models: an experimental visualization model capturing real-time convection onset in partially filled rotating cylinders, and a numerical bifurcation model using S-CLSVOF multiphase DNS to replicate and generalize the observed transitions. Together, these frameworks quantify the relationship between filling ratio, Reynolds number, and convection wavelength with unprecedented precision. The numerical model, validated against experiment, newly distinguishes supercritical and subcritical bifurcation regimes and accurately predicts the wavelength of the most unstable mode in infinite cylinders. This dual-model approach sets a new standard for analyzing coupled gas–liquid dynamics in rotating geometries.

The authors used in their experiments a transparent acrylic cylinder of radius 50 mm and length 800 mm, partially filled with silicone oil of kinematic viscosity 50 mm² s⁻¹. The vessel was rotated about its horizontal axis at angular velocities between 0.5π and 1.8π rad s⁻¹, corresponding to Reynolds numbers of 90–280 and Froude numbers below 0.1, ensuring minimal interface deformation. Tracer particles illuminated by a vertical laser sheet enabled visualization of cross-sectional motion and quantitative velocity mapping by PIV. Under low rotation, the flow behaved as a simple rotating pool, but as angular velocity increased beyond a critical value, pairs of counter-rotating convection cells spontaneously appeared. For a filling ratio Ψ = 0.2, these structures emerged near Re ≈ 200, whereas at Ψ = 0.4, the transition occurred near Re ≈ 110, implying that increasing liquid volume reduces the threshold. When the researchers performed velocity field analysis, they found that cell intensity, measured by the vertical velocity amplitude , rose sharply once Re exceeded a critical value. Although the transition resembled a pitchfork bifurcation, finite cylinder length caused asymmetries, producing a gradual onset rather than a sharp critical point. Flow visualizations also revealed that the axial wavelength of the convection pattern increased with filling ratio, a feature later quantified in the simulations. The authors simulated the two-phase Navier–Stokes equations with either no-slip or slip end-wall conditions using the validated S-CLSVOF method. The DNS demonstrated that in an infinitely long cylinder, convection cells arise via a supercritical pitchfork bifurcation for low to moderate filling ratios (Ψ < 0.8), whereas at larger fillings (Ψ ≈ 0.9), the transition becomes subcritical. The computed critical Reynolds number showed a non-monotonic dependence on Ψ, reaching a minimum near Ψ = 0.6, where Rec ≈ 100. The axial wavelength λ* of a convection pair increased monotonically with Ψ, from 0.86 R at Ψ = 0.05 to about 3.6 R at Ψ = 0.9. Moreover, the team performed numerical runs with variable Froude numbers which showed that the critical condition depends only weakly on Fr within the small-Fr regime, validating the experimental assumption of an undeformed surface. Additionally, comparisons between DNS and experiments yielded nearly identical onset thresholds and cell counts which confirms the reliability of both methods.

In conclusion, Professor Susumu Goto and colleagues’ study provided a novel coherent framework for understanding the emergence of convection in partially filled rotating cylinders and successful bridged conceptual gap between classical rimming flow and solid-body rotation. The new discovery that convection cells can persist even when the free surface remains essentially flat challenges previous intuition that strong surface deformation is necessary for axial motion. The authors have defined the physical limits of stability in an accessible laboratory system by linking the onset to specific ranges of Reynolds number and filling ratio. Furthermore, we believe the bifurcation analysis has deeper implications for pattern formation and symmetry breaking in rotating multiphase flows and the transition from solid-body rotation to ordered convection through a supercritical pitchfork bifurcation illustrates how subtle imbalances between inertial and viscous forces can self-organize into coherent structures. The identification of an imperfect bifurcation in finite-length cylinders clarifies why real systems display smooth rather than abrupt transitions. Also, the observed crossover to subcritical behavior at high filling ratios suggests a qualitative change in the dominant instability mechanism, potentially linked to shear-layer coupling and end-wall effects.

From an engineering standpoint, the authors’ findings delineate design principles for efficient bladeless mixing devices. Because the convection cells enhance mixing without moving components, systems tuned near the critical Reynolds number can achieve rapid homogenization with minimal mechanical complexity. The results also inform industrial processes—such as coating, polymer curing, or crystal growth—where rotation and partial filling naturally coexist and where flow uniformity governs product quality. The scaling laws for cell wavelength versus filling ratio provide direct guidance for selecting vessel length and operating speed to sustain optimal mixing patterns. Beyond applications, the work extends the theoretical landscape of rotating flows. It introduces a well-characterized benchmark system for testing computational models of multiphase convection and highlights how dimensionless parameters (Re, Fr, Ψ) jointly control the onset of motion in geometrically simple yet dynamically rich configurations. The reported comparison between experiment and DNS—achieved with sub-percent agreement—demonstrates the maturity of modern interface-tracking algorithms for free-surface hydrodynamics. Overall, Watanabe et al have transformed simple configuration into a canonical platform for studying hydrodynamic bifurcations under rotation. Their synthesis of careful experimentation and numerically resolved physics not only elucidates a long-standing problem but also opens a route toward predictive control of convection-driven transport in rotating systems.