Motivated by the goal of improving and augmenting stochastic Lagrangian models of particle dispersion in turbulent flows, techniques from the theory of stochastic processes are applied to a model transport problem. The aim is to find an efficient and accurate method to calculate the total tracer transport between a source and a receptor when the flow between the two locations is weak, rendering direct stochastic Lagrangian simulation prohibitively expensive. Importance sampling methods that combine information from stochastic forward and back trajectory calculations are proposed. The unifying feature of the new methods is that they are developed using the observation that a perfect strategy should distribute trajectories in proportion to the product of the forward and adjoint solutions of the transport problem, a quantity here termed the ‘density of trajectories’ $D(\boldsymbol{x},t)$. Two such methods are applied to a ‘hard’ model problem, in which the prescribed kinematic flow is in the large-Péclet-number chaotic advection regime, and the transport problem requires simulation of a complex distribution of well-separated trajectories. The first, Milstein’s measure transformation method, involves adding an artificial velocity to the trajectory equation and simultaneously correcting for the weighting given to each particle under the new flow. It is found that, although a ‘perfect’ artificial velocity $\boldsymbol{v}^{\ast }$ exists, which is shown to distribute the trajectories according to $D$, small errors in numerical estimates of $\boldsymbol{v}^{\ast }$ cumulatively lead to difficulties with the method. A second method is Grassberger’s ‘go-with-the-winners’ branching process, where trajectories found unlikely to contribute to the net transport (losers) are periodically removed, while those expected to contribute significantly (winners) are split. The main challenge of implementation, which is finding an algorithm to select the winners and losers, is solved by a choice that explicitly forces the distribution towards a numerical estimate of $D$ generated from a previous back trajectory calculation. The result is a robust and easily implemented algorithm with typical variance up to three orders of magnitude lower than the direct approach.

We study the centrifugal instability of non-axisymmetric vortices in the presence of an axial flow ($w$) and a background rotation (${\it\Omega}_{z}$) using the local stability approach. Analytically solving the local stability equations for an axisymmetric vortex with $w$ and ${\it\Omega}_{z}$, growth rates for wave vectors that are periodic upon evolution around a closed streamline are calculated. The resulting sufficient criterion for centrifugal instability in an axisymmetric vortex is then heuristically extended to non-axisymmetric vortices and written in terms of integral quantities and their derivatives with respect to the streamfunction on a streamline. The new criterion for non-axisymmetric vortices, which converges to the exact criterion of Bayly (Phys. Fluids, vol. 31, 1988, pp. 56–64) in the absence of background rotation and axial flow, is validated by comparisons with numerically calculated growth rates for two different anticyclonic vortices: the Stuart vortex (specified by the concentration parameter ${\it\rho},~0<{\it\rho}\leqslant 1$) and the Taylor–Green vortex (specified by the aspect ratio $E,~0<E\leqslant 1$). With no axial velocity and finite background rotation, the criterion predicts a lower and an upper threshold of $|{\it\Omega}_{z}|$ between which centrifugal instability is present. We further demonstrate that the criterion represents an improvement over the criterion of Sipp & Jacquin (Phys. Fluids, vol. 12, 2000, pp. 1740–1748). Finally, in the presence of both axial velocity and background rotation, the criterion is shown to be accurate for large enough ${\it\rho}$ and $E$.

The steady state of a forced Taylor–Green flow is investigated in a rotating frame of reference. The investigation involves the results of 184 numerical simulations for different Reynolds numbers $\mathit{Re}_{F}$ and Rossby numbers $\mathit{Ro}_{F}$. The large number of examined runs allows a systematic study that enables the mapping of the different behaviours observed to the parameter space ($\mathit{Re}_{F},\mathit{Ro}_{F}$), and the examination of different limiting procedures for approaching the large $\mathit{Re}_{F}$ small $\mathit{Ro}_{F}$ limit. Four distinctly different states were identified: laminar, intermittent bursts, quasi-two-dimensional condensates and weakly rotating turbulence. These four different states are separated by power-law boundaries $\mathit{Ro}_{F}\propto \mathit{Re}_{F}^{-{\it\gamma}}$ in the small $\mathit{Ro}_{F}$ limit. In this limit, the predictions of asymptotic expansions can be directly compared with the results of the direct numerical simulations. While the first-order expansion is in good agreement with the results of the linear stability theory, it fails to reproduce the dynamical behaviour of the quasi-two-dimensional part of the flow in the nonlinear regime, indicating that higher-order terms in the expansion need to be taken into account. The large number of simulations allows also to investigate the scaling that relates the amplitude of the fluctuations with the energy dissipation rate and the control parameters of the system for the different states of the flow. Different scaling was observed for different states of the flow, that are discussed in detail. The present results clearly demonstrate that the limits of small Rossby and large Reynolds numbers do not commute and it is important to specify the order in which they are taken.

The problem of a suspension droplet falling under gravity was examined for polydisperse droplets composed of a mixture of particles with different densities and sizes. The study was conducted using both simulations based on oseenlet particle interactions and laboratory experiments. The hydrodynamic interactions of the particles within the suspension droplet allow a polydisperse collection of particles to fall as a coherent droplet, even for cases where the difference in particle terminal velocity would cause them to separate quickly from each other in the absence of hydrodynamic interactions. However, a gradual segregation phenomenon is observed in which particles with lower terminal velocity preferentially leave the suspension droplet by entering into the droplet tail, whereas particles with higher terminal velocity remain for longer periods of time within the droplet. When computations and experiments are performed for bidisperse mixtures, a point is eventually reached where all of the lighter/smaller particles are ejected into the droplet tail and the droplet continues to fall with only the heavier/larger particles.

In highly stratified atmospheric and oceanic environments, a large fraction of energy input by various sources can be manifest as internal waves (IWs). The propagating nature of IWs results in the distribution of the energy over a large fraction of the air/water column. Wakes of translating bodies are one source of input energy that has been of continued interest. To further the understanding of wakes in strongly stratified environments, and particularly the near-field regime where strong coupling to the internal wavefield is evident, an extensive series of experiments on the internal wavefield generated by a towed sphere was performed, wherein the internal wavefield was measured over a Froude number range $0.1\leqslant \mathit{Fr}\leqslant 5$ (where $\mathit{Fr}=U/ND$, $U$ is the tow speed, $D$ the sphere diameter and $N$ the Brunt–Väisälä (BV) frequency). In a second series of experiments, the temporal wavefield evolution was studied over two BV periods. These measurements show that the body generation (lee wave) mechanism dominates at $\mathit{Fr}\lesssim 1$, while the random eddies in the turbulent wake become the dominant source at $\mathit{Fr}\gtrsim 1$. In the low-$\mathit{Fr}$ regime, $\mathit{Fr}\leqslant 1$, there is a resonant peak in the coupling of the input wake energy to the internal wavefield at a Froude number of ${\sim}0.5$, and at its maximum 70 % of the input energy is coupled into IW potential energy. In this regime it was also found that the spreading angle of the evolving wavefield was considerably broader than predicted by the classical point-source models for the wavefield further downstream, owing to the existence in the near field of a significant energy content in the higher-IW modes that deteriorate at later times. In the low-$\mathit{Fr}$ regime, it was found that, while the IW potential energy increases $\propto \mathit{Fr}^{2}$, the fraction of the total energy input is a weak function of $\mathit{Fr}$, varying as $\mathit{Fr}^{1/2}$.

Mixing and reaction between two scalars initially separated by scalar-free ambient fluid is important in problems ranging from ecology to engineering. Using a two-channel planar laser-induced fluorescence (PLIF) system the instantaneous spatial structure of two independent scalars emitted from parallel jets into a slow coflow is quantified. Of particular interest is the scalar covariance used to define the correlation coefficient. Joint probability distribution functions (JPDFs) and instantaneous images of the scalar fields demonstrate that initially the flow mainly consists of incursions of fluid from one jet into the other, and vice versa, before scalars have time to assemble in attracting regions of the flow and coalesce due to diffusive flux. Decomposing the joint probability distribution exhibits the effect these events have on scaler overlap and scalar covariance. Along the centreline near where the mean profiles of the jets meet, the scalar covariance is negative; however, the covariance becomes positive as the scalars converge in shared structure and diffusive flux bridges a reduced barrier of ambient fluid. The mixing path between scalar filaments can be probabilistically observed through the conditional diffusion of the two scalars at various points in the flow.

The dynamics of one-dimensional, piston-driven hydrogen–air detonations are predicted in the presence of physical mass, momentum and energy diffusion. The calculations are automatically verified by the use of an adaptive wavelet-based computational method which correlates a user-specified error tolerance to the error in the calculations. The predicted frequency of 0.97 MHz for an overdriven pulsating detonation agrees well with the 1.04 MHz frequency observed by Lehr in a shock-induced combustion experiment around a spherical projectile, thus giving a limited validation for the model. A study is performed in which the supporting piston velocity is varied, and the long time behaviour is examined for an initially stoichiometric mixture at 293.15 K and 1 atm. Several distinct propagation behaviours are predicted: a stable detonation, a high-frequency pulsating detonation, a pulsating detonation with two competing modes, a low-frequency pulsating detonation and a propagating detonation with many active frequencies. In the low-frequency pulsating mode, the long time behaviour undergoes a phenomenon similar to period-doubling. Harmonic analysis is used to examine how the frequency of the pulsations evolves as the supporting piston velocity is varied. It is found that the addition of viscosity shifts the neutral stability boundary by about 2 % with respect to the supporting piston velocity. As the supporting piston velocity is lowered, the intrinsic instability grows in strength, and the effect of viscosity is weakened such that the results are indistinguishable from the inviscid predictions.

This study amends the inviscid Wagner lift model for starting flow at relatively large angles of attack to account for the influence of additional leading edge and trailing edge vortices. Two methods are provided for starting flow of a flat plate. The first method is a modified Wagner function, which assumes a planar trajectory of the trailing edge vortex sheet accounting for a temporal offset from the original Wagner function given release of leading edge vortices and a concentrated starting point vortex at the initiation of motion. The second method idealizes the trailing edge sheet as a series of discrete vortices released sequentially. The models presented are shown to be in good agreement with high-fidelity simulations. Through the present theory, a vortex force line map is generated, which clearly indicates lift enhancing and reducing directions and, when coupled with streamlines, allows one to qualitatively interpret the effect of the sign and position of vortices on the lift and to identify the origins of lift oscillations and peaks. It is concluded that leading edge vortices close to the leading edge elevate the Wagner lift curve while a strong leading edge vortex convected to the trailing edge is detrimental to lift production by inducing a strong trailing edge vortex moving in the lift reducing direction. The vortex force line map can be employed to understand the effect of the different vortices in other situations and may be used to improve vortex control to enhance or reduce the lift.

We report on new patterns in high-speed flows of granular materials obtained by means of extensive numerical simulations. These patterns emerge from the destabilization of unidirectional flows upon increase of mass holdup and inclination angle, and are characterized by complex internal structures, including secondary flows, heterogeneous particle volume fraction, symmetry breaking and dynamically maintained order. In particular, we evidenced steady and fully developed ‘supported’ flows, which consist of a dense core surrounded by a highly energetic granular gas. Interestingly, despite their overall diversity, these regimes are shown to obey a scaling law for the mass flow rate as a function of the mass holdup. This unique set of three-dimensional flow regimes raises new challenges for extending the scope of current granular rheological models.

We develop a model of the steady exchange flows which may develop between two aquifers at different levels in the geological strata and across which there is an unstable density stratification, as a result of their connection through a series of fractures. We show that in general there are multiple steady exchange flows which can develop, depending on the initial conditions, and which may involve a net upwards or downwards volume flux. We also show that there is a family of equilibrium exchange flows with zero net volume flux, each characterised by a different interlayer flux of buoyancy. We present experiments which confirm our simplified model of the exchange flow. Such exchange flows may supply unsaturated water from a deep aquifer to drive dissolution of a structurally trapped pool of geologically stored $\text{CO}_{2}$, once the water in the aquifer containing the trapped pool of $\text{CO}_{2}$ has become saturated in $\text{CO}_{2}$, and hence relatively dense. Such exchange flows may also lead to cross-contamination of aquifer fluids, which may be of relevance in assessing risks of geological storage systems.

The non-Newtonian shear rheology of colloidal dispersions is the result of the competition and balance between hydrodynamic (dissipative) and thermodynamic (conservative) forces that lead to a non-equilibrium microstructure under flow. We present the first experimental measurements of the shear-induced microstructure of a concentrated near-hard-sphere colloidal dispersion through the shear thickening transition using small-angle neutron scattering (SANS) measurements made in three orthogonal planes during steady shear. New instrumentation coupled with theoretical derivations of the stress-SANS rule enable rigorous testing of the relationship between this non-equilibrium microstructure and the observed macroscopic shear rheology. The thermodynamic and hydrodynamic components of the stress that drive shear thinning, shear thickening and first normal stress differences are separately defined via stress-SANS rules and compared to the rheological behaviour of the dispersion during steady shear. Observations of shear-induced hydrocluster formation is in good agreement with Stokesian dynamics simulation results by Foss & Brady (J. Fluid Mech., vol. 407, 2000, pp. 167–200). This unique set of measurements of shear rheology and non-equilibrium microstructure of a model system provides new insights into suspension mechanics as well as a method to rigorously test constitutive equations for colloidal suspension rheology.

Interaction with an opposing current amplifies wave modulation and accelerates nonlinear wave focusing in regular wavepackets. This results in large-amplitude waves, usually known as rogue waves, even if the wave conditions are less prone to extremes. Laboratory experiments in three independent facilities are presented here to assess the role of opposing currents in changing the statistical properties of unidirectional and directional mechanically generated random wavefields. The results demonstrate in a consistent and robust manner that opposing currents induce a sharp and rapid transition from weakly to strongly non-Gaussian properties. This is associated with a substantial increase in the probability of occurrence of rogue waves for unidirectional and directional sea states, for which the occurrence of extreme and rogue waves is normally the least expected.

Sinusoidally oscillatory flow around four circular cylinders in an in-line square arrangement is numerically investigated at Keulegan–Carpenter numbers ($\mathit{KC}$) ranging from 1 to 12 and at Reynolds numbers ($\mathit{Re}$) from 20 to 200. A set of flow patterns is observed and classified based on known oscillatory flow regimes around a single cylinder. These include six types of reflection symmetry regimes to the axis of flow oscillation, two types of spatio-temporal symmetry regimes and a series of symmetry-breaking flow patterns. In general, at small gap distances, the four structures behave more like a single body, and the flow fields therefore resemble those around a single cylinder with a large effective cylinder diameter. With increasing gap distance, flow structures around each individual cylinder in the array start to influence the overall flow patterns, and the flow field shows a variety of symmetry and asymmetry patterns as a result of vortex and shear layer interactions. The characteristics of hydrodynamic forces on individual cylinders as well as on the cylinder group are also examined. It is found that the hydrodynamic forces respond in a similar manner to the flow field to the cylinder proximity and wake interference.

This work lays out a methodology for calculating the pressure distribution internal to a generic, deformable, axisymmetric body with an internal cavity region whose deformation expels/ingests finite jets of water. This work is partially motivated by a desire to model instantaneous jetting forces and total work required for jellyfish and cephalopod locomotion, both of which can be calculated from the internal pressure distribution. But the derivation is non-specific and can be applied to any axisymmetric, deformable body (organic or synthetic) driving fluid in or out of an internal cavity. The pressure distribution over the inner surface is derived by integrating the momentum equation along a strategic path, equating local surface pressure to known quantities such as stagnation pressure, and correlating unknown terms to the total circulation of characteristic regions. The integration path is laid out to take advantage of symmetry conditions, inherent irrotationality, and prescribed boundary conditions. The usefulness/novelty of this approach lies in the fact that circulation is an invariant of motion for inviscid flows, allowing it to be modelled by a series of vorticity flux and source terms. In this study we also categorize the various sources of circulation in the general cavity–jet system, providing modelling for each of these terms with respect to known cavity deformation parameters. Through this approach we are able to isolate the effect of different deformation behaviours on each of these circulation components, and hence on the internal pressure distribution. A highly adaptable, transparent, prototype jet actuator was designed and tested to measure the circulation in the cavity and the surrounding fluid as well as the dynamic forces acting on the device during operation. The circulation in both the jet and cavity regions shows good agreement with the inviscid modelling, except at the end of the refill phase where circulation is lost to viscous dissipation. The total instantaneous forces produced during actuation are accurately modelled by the pressure analysis during both expulsion and refilling phases of the jetting cycle for multiple deformation programs. Independent of the end goal, such as propulsion, mixing, feeding etc., the efficiency of the process will always be inversely proportional to the total energy required to drive the system. Therefore, given a consistent output, efficiency is maximized by the minimum required energy. Here it is observed (somewhat counter-intuitively) that, for both jetting and refilling, total work required to drive the fluid is lower for impulsive velocity programs with fast accelerations at the start and end of motion than sinusoidal velocity programs with smoother gradual accelerations. The underlying cause is that sinusoidal programs result in a peak in pressure (force) simultaneously with maximum deflection velocity of the deformable boundary driving fluid motion; for the impulsive programs these peaks are out of phase and overall energy consumption is reduced.

We address the problem of computing the hydrodynamic forces and torques among $N$ solid spherical particles moving with given rotational and translational velocities in Stokes flow. We consider the original fluid–particle model without introducing new hypotheses or models. Our method includes the singular lubrication interactions which may occur when some particles come close to one another. The main new feature is that short-range interactions are propagated to the whole flow, including accurately the many-body lubrication interactions. The method builds on a pre-existing fluid solver and is flexible with respect to the choice of this solver. The error is the error generated by the fluid solver when computing non-singular flows (i.e. with negligible short-range interactions). Therefore, only a small number of degrees of freedom are required and we obtain very accurate simulations within a reasonable computational cost. Our method is closely related to a method proposed by Sangani & Mo (Phys. Fluids, vol. 6, 1994, pp. 1653–1662) but, in contrast with the latter, it does not require parameter tuning. We compare our method with the Stokesian dynamics of Durlofsky et al. (J. Fluid Mech., vol. 180, 1987, pp. 21–49) and show the higher accuracy of the former (both by analysis and by numerical experiments).

We study, using both experiment and theory, the coupling of transport and shape dynamics for elastomeric fibres moving through an inhomogeneous flow. The cellular flow, created electromagnetically in our experiment, comprises many identical cells of counter-rotating vortices, with a global flow geometry characterized by a backbone of stable and unstable manifolds connecting hyperbolic stagnation points. Our mathematical model is based upon slender-body theory for the Stokes equations, with the fibres modelled as inextensible elastica. Above a certain threshold of the control parameter, the elasto-viscous number, transport of fibres is mediated by their episodic buckling by compressive stagnation point flows, lending an effectively chaotic component to their dynamics. We use simulations of the model to construct phase diagrams of the fibre state (buckled or not) near stagnation points in terms of two variables that arise in characterizing the transport dynamics. We show that this reduced statistical description quantitatively captures our experimental observations. By carefully reproducing the experimental protocols and time scales of observation within our numerical simulations, we also quantitatively explain features of the measured buckling probability curve as a function of the effective flow forcing. Finally, we show within both experiment and simulation the existence of short and long time scales in the evolution of fibre conformation.

This paper builds upon the investigation of Augier et al. (Phys. Fluids, vol. 26 (4), 2014) in which a strongly stratified turbulent-like flow was forced by 12 generators of vertical columnar dipoles. In experiments, measurements start to provide evidence of the existence of a strongly stratified inertial range that has been predicted for large turbulent buoyancy Reynolds numbers $\mathscr{R}_{t}={\it\varepsilon}_{\!K}/({\it\nu}N^{2})$, where ${\it\varepsilon}_{\!K}$ is the mean dissipation rate of kinetic energy, ${\it\nu}$ the viscosity and $N$ the Brunt–Väisälä frequency. However, because of experimental constraints, the buoyancy Reynolds number could not be increased to sufficiently large values so that the inertial strongly stratified turbulent range is only incipient. In order to extend the experimental results toward higher buoyancy Reynolds number, we have performed numerical simulations of forced stratified flows. To reproduce the experimental vortex generators, columnar dipoles are periodically produced in spatial space using impulsive horizontal body force at the peripheries of the computational domain. For moderate buoyancy Reynolds number, these numerical simulations are able to reproduce the results obtained in the experiments, validating this particular forcing. For higher buoyancy Reynolds number, the simulations show that the flow becomes turbulent as observed in Brethouwer et al. (J. Fluid Mech., vol. 585, 2007, pp. 343–368). However, the statistically stationary flow is horizontally inhomogeneous because the dipoles are destabilized quite rapidly after their generation. In order to produce horizontally homogeneous turbulence, high-resolution simulations at high buoyancy Reynolds number have been carried out with a slightly modified forcing in which dipoles are forced at random locations in the computational domain. The unidimensional horizontal spectra of kinetic and potential energies scale like $C_{1}{\it\varepsilon}_{\!K}^{2/3}k_{h}^{-5/3}$ and $C_{2}{\it\varepsilon}_{\!K}^{2/3}k_{h}^{-5/3}({\it\varepsilon}_{\!P}/{\it\varepsilon}_{\!K})$, respectively, with $C_{1}=C_{2}\simeq 0.5$ as obtained by Lindborg (J. Fluid Mech., vol. 550, 2006, pp. 207–242). However, there is a depletion in the horizontal kinetic energy spectrum for scales between the integral length scale and the buoyancy length scale and an anomalous energy excess around the buoyancy length scale probably due to direct transfers from large horizontal scale to small scales resulting from the shear and gravitational instabilities. The horizontal buoyancy flux co-spectrum increases abruptly at the buoyancy scale corroborating the presence of overturnings. Remarkably, the vertical kinetic energy spectrum exhibits a transition at the Ozmidov length scale from a steep spectrum scaling like $N^{2}k_{z}^{-3}$ at large scales to a spectrum scaling like $C_{K}{\it\varepsilon}_{\!K}^{2/3}k_{z}^{-5/3}$, with $C_{K}=1$, the classical Kolmogorov constant.

Recent experiments on coating flows and liquid drop impact both demonstrate that wetting failures caused by air entrainment can be suppressed by reducing the ambient gas pressure. Here, it is shown that non-equilibrium effects in the gas can account for this behaviour, with ambient pressure reductions increasing the mean free path of the gas and hence the Knudsen number $\mathit{Kn}$. These effects first manifest themselves through Maxwell slip at the boundaries of the gas, so that for sufficiently small $\mathit{Kn}$ they can be incorporated into a continuum model for dynamic wetting flows. The resulting mathematical model contains flow structures on the nano-, micro- and millimetre scales and is implemented into a computational platform developed specifically for such multiscale phenomena. The coating flow geometry is used to show that for a fixed gas–liquid–solid system (a) the increased Maxwell slip at reduced pressures can substantially delay air entrainment, i.e. increase the ‘maximum speed of wetting’, (b) unbounded maximum speeds are obtained, as the pressure is reduced only when slip at the gas–liquid interface is allowed for, and (c) the observed behaviour can be rationalised by studying the dynamics of the gas film in front of the moving contact line. A direct comparison with experimental results obtained from a dip-coating process shows that the model recovers most trends but does not accurately predict some of the high viscosity data at reduced pressures. This discrepancy occurs because the gas flow enters the ‘transition regime’, so that more complex descriptions of its non-equilibrium nature are required. Finally, by collapsing onto a master curve experimental data obtained for drop impact in a reduced pressure gas, it is shown that the same physical mechanisms are also likely to govern splash suppression phenomena.

Viscoplastic fluids do not flow unless they are sufficiently stressed. This property can be exploited in order to produce novel flow features. One example of such flows is viscoplastically lubricated (VPL) flow, in which a viscoplastic fluid is used to stabilize the interface in a multi-layer flow, far beyond what might be expected for a typical viscous–viscous interface. Here we extend this idea by considering the encapsulation of droplets within a viscoplastic fluid, for the purpose of transportation, e.g. in pipelines. The main advantage of this method, compared to others that involve capillary forces is that significantly larger droplets may be stably encapsulated, governed by the length scale of the flow and yield stress of the encapsulating fluid. We explore this set-up both analytically and computationally. We show that sufficiently small droplets are held in the unyielded plug of a Poiseuille flow (pipe or plane channel). As the length or radius of the droplets increases, the carrier fluid eventually yields, potentially breaking the encapsulation. We study this process of breaking and give estimates for the limiting size of droplets that can be encapsulated.

A buoyant vortex ring produced by an underwater bursting balloon was studied experimentally. The effect of dimensionless surface tension on characteristics including rise velocity, rate of expansion, circulation, trajectory, and lifetime of the vortex ring bubble was investigated. Results showed reasonable agreement with the literature on vortex rings produced by conventional approaches. It was observed that as the dimensionless surface tension increased, the rise velocity, the circulation and consequently the stability of the vortex ring bubble increased; however, the rate of expansion tends toward constant values. A semi-analytical model is proposed by modifying the drag-based model presented by Sullivan et al. (J. Fluid Mech., vol. 609, 2008, pp. 319–347) to make it applicable to buoyant vortex rings. The modified model suggests that the vortex ring expansion is essentially due to the buoyancy force. An expression is also derived for the circulation in terms of the initial volume of the balloon and the depth at which the balloon bursts.