3D Mean-Field Bubble Growth Law

• Three-dimensional mean-field bubble growth law is a framework that replaces detailed bubble dynamics with averaged pressure, concentration, or hydrodynamic fields to predict coarsening behavior.
• Distinct transport regimes, such as diffusion-limited (t^(1/3)) and film-mediated (t^(1/2)), lead to different growth exponents and scaling laws in bubble coarsening.
• Mean-field models provide closure principles by linking individual bubble growth to global constraints, reconciling experimental, simulation, and theoretical findings across foams, gases, and dense fluids.

Three-dimensional mean-field bubble growth law denotes a reduced description for the evolution of a bubble radius $R(t)$ or a characteristic coarsening length $\ell(t)$ in a three-dimensional medium, obtained by replacing bubble-resolved transport and geometry with averaged pressure, concentration, or hydrodynamic fields. The literature does not support a single universal 3D law. Instead, the exponent and even the governing variable depend on the dominant transport mechanism and morphology: diffusion-limited Ostwald ripening gives $R\sim t^{1/3}$, film-mediated dry-foam coarsening gives $R\sim t^{1/2}$, and viscous hydrodynamic bubble coarsening in dense fluids gives $\ell\sim t$ (Chieco et al., 2023, Morgan et al., 1 Aug 2025, A et al., 25 Sep 2025).

1. Mean-field formulation in three dimensions

In coarsening theory, “mean field” usually means that the growth rate of an individual bubble depends only on its own size and a small set of global quantities, rather than on the full local neighborhood. A standard formulation is a continuity equation for the bubble-size distribution $f(R,t)$,

$$\frac{\partial f}{\partial t}=-\frac{\partial}{\partial R}\bigl(\dot R(R,t)\,f(R,t)\bigr),$$

together with conservation of total gas volume. In three dimensions, if the characteristic radius scales as $R_c(t)\sim t^\alpha$, then the total bubble number scales as $n(t)\sim t^{-3\alpha}$ (Watanabe et al., 2016).

A central object is the critical radius $R_c(t)$, defined so that bubbles with $\ell(t)$0 grow and bubbles with $\ell(t)$1 shrink. In solvable mean-field models, $\ell(t)$2 is treated as a global variable linked to the average radius or to a low-order moment of the size distribution. One compact closure tracks only the average radius $\ell(t)$3 and the average critical radius $\ell(t)$4, with $\ell(t)$5 used as a proxy for the distribution width (Chieco et al., 2023).

The same mean-field logic also appears outside ensemble coarsening. For a single bubble in a supersaturated liquid, in electrolysis, or in Rayleigh–Plesset dynamics, the reduced law is an ordinary differential equation for $\ell(t)$6 obtained after replacing the surrounding medium by an averaged far-field concentration, pressure, or effective interaction term. The common structure is therefore not a single exponent, but a closure principle.

2. Classical three-dimensional coarsening laws

The classical 3D mean-field laws are distinguished by the rate-limiting transport process.

Regime	Representative law	Typical setting
Diffusion-limited	$\ell(t)$7	dilute droplets or bubbles, bulk diffusion
Interface- or reaction-limited	$\ell(t)$8	low-temperature LJ bubble ripening
Dry-foam film diffusion	$\ell(t)$9	space-filling bubbles, transport through films

For dilute wet systems, the Lifshitz–Slyozov–Wagner limit gives the individual-bubble law

$R\sim t^{1/3}$0

which leads in the scaling state to $R\sim t^{1/3}$1. For dry foams, where the diffusion length is set by the film thickness rather than by the bubble radius, the corresponding law is

$R\sim t^{1/3}$2

and the scaling state gives $R\sim t^{1/3}$3 (Chieco et al., 2023).

Large-scale molecular dynamics of 3D Lennard–Jones bubble ripening confirms both classical exponents in different parameter regimes. After depressurization into the coexistence region, the total number of bubbles decays as $R\sim t^{1/3}$4. At higher initial temperature, $R\sim t^{1/3}$5, implying $R\sim t^{1/3}$6; at lower temperature, $R\sim t^{1/3}$7, implying $R\sim t^{1/3}$8. The same work directly confirmed self-similarity of the bubble-size distribution in the scaling regime (Watanabe et al., 2014).

A later molecular-dynamics study measured individual bubble growth rates and showed that the low-temperature case is genuinely reaction-limited and well described by classical LSW theory for that limit. The high-temperature case exhibits the $R\sim t^{1/3}$9 law expected for diffusion-limited growth, but its detailed agreement with diffusion-limited LSW becomes less reliable as gas volume fraction increases, indicating that the diffusion-limited mean-field approximation is much more sensitive to finite-volume-fraction effects (Watanabe et al., 2016).

3. Dry foams, wet foams, and film-mediated closures

For 3D dry foams, a simple solvable mean-field model closes the average dynamics with

$R\sim t^{1/2}$0

so that

$R\sim t^{1/2}$1

In the scaling state, $R\sim t^{1/2}$2, and the average radius obeys

$R\sim t^{1/2}$3

with asymptotic law $R\sim t^{1/2}$4. Away from the scaling state, the same model yields an explicit transient correction through a shift parameter $R\sim t^{1/2}$5, which encodes whether the initial size distribution is more monodisperse or more polydisperse than the eventual scaling distribution (Chieco et al., 2023).

A more refined 3D wet-foam model introduces a radius-dependent film-coverage factor. For non-adhesive foams with liquid fraction $R\sim t^{1/2}$6, the proposed mean-field law is

$R\sim t^{1/2}$7

where $R\sim t^{1/2}$8 is a scaled osmotic pressure linked to $R\sim t^{1/2}$9. This law reduces to the Lemlich dry-foam law in the dry limit and predicts a scaling state with the same exponent $\ell\sim t$0 for all $\ell\sim t$1, but with a $\ell\sim t$2-dependent scaling distribution $\ell\sim t$3. A notable consequence is that for any $\ell\sim t$4 the predicted scaling distribution has $\ell\sim t$5, implying a finite population of very small bubbles in the scaling state (Morgan et al., 1 Aug 2025).

The same wet-foam model is explicitly a border-blocking theory: gas is transported only through films, not through Plateau borders or bulk liquid. It is therefore a film-dominated 3D mean-field law, not an interpolation to bulk-diffusive LSW ripening.

4. Hydrodynamic mean-field laws in dense fluids

In dense single-component fluids, the classical $\ell\sim t$6 law can fail completely. Large-scale 3D molecular dynamics of a single-component Lennard–Jones fluid quenched to $\ell\sim t$7 at overall density $\ell\sim t$8 shows nucleation and growth of vapor bubbles inside a dense liquid matrix. The morphology exhibits dynamic scaling in both the equal-time correlation function and the structure factor, with Porod behavior $\ell\sim t$9, indicating sharp interfaces and a one-length description $f(R,t)$0 (A et al., 25 Sep 2025).

The raw growth data suggest an intermediate effective law $f(R,t)$1, but finite-size scaling with

$f(R,t)$2

gives optimal parameters $f(R,t)$3, $f(R,t)$4, and $f(R,t)$5. The asymptotic 3D coarsening law is therefore

$f(R,t)$6

which the authors identify as a viscous hydrodynamic regime.

The mean-field scaling argument differs from LSW. Diffusion-limited Ostwald ripening gives

$f(R,t)$7

whereas capillary–viscous balance gives

$f(R,t)$8

A further inertial hydrodynamic balance gives

$f(R,t)$9

but that regime was not resolved numerically in the available system sizes and time windows. In this dense-fluid setting, the relevant 3D mean field is therefore hydrodynamic rather than diffusive.

5. Single-bubble 3D laws and interacting-population extensions

Not all 3D mean-field bubble growth laws describe coarsening ensembles. For a single bubble in a supersaturated solution, the steady-diffusion approximation gives

$$\frac{\partial f}{\partial t}=-\frac{\partial}{\partial R}\bigl(\dot R(R,t)\,f(R,t)\bigr),$$0

with distinct driving terms for Henry and Sievert solubility laws: $$\frac{\partial f}{\partial t}=-\frac{\partial}{\partial R}\bigl(\dot R(R,t)\,f(R,t)\bigr),$$1 Both reduce at large radius to the same asymptotic 3D law,

$$\frac{\partial f}{\partial t}=-\frac{\partial}{\partial R}\bigl(\dot R(R,t)\,f(R,t)\bigr),$$2

after an initial and an intermediate stage in which Laplace pressure matters (Gor et al., 2012).

For hydrogen bubbles at an electrode, 3D direct numerical simulation of electrolysis recovers diffusion-controlled growth

$$\frac{\partial f}{\partial t}=-\frac{\partial}{\partial R}\bigl(\dot R(R,t)\,f(R,t)\bigr),$$3

at late times, while the initial exponent is overpredicted by the simulations. The same simulations show that nucleation-site multiplicity alters the relevant Sherwood number, reduces the effective growth rate through competition for dissolved gas, and changes detachment through coalescence; single-bubble detachment radii agree with Fritz’s formula, indicating buoyancy-dominated departure (Qin et al., 21 Jul 2025).

For dense bubble clouds in supersaturated liquid under microgravity, the short-time law is the Epstein–Plesset form

$$\frac{\partial f}{\partial t}=-\frac{\partial}{\partial R}\bigl(\dot R(R,t)\,f(R,t)\bigr),$$4

but later growth is screened by neighboring bubbles. A point-sink interaction model predicts an intermediate competition regime and a long-time husk regime in which only bubbles near the cloud surface continue to grow significantly (Vega-Martínez et al., 2020).

For flash-boiling microdroplets with many interacting nuclei, a modified Rayleigh–Plesset equation introduces interaction terms proportional to $$\frac{\partial f}{\partial t}=-\frac{\partial}{\partial R}\bigl(\dot R(R,t)\,f(R,t)\bigr),$$5, where $$\frac{\partial f}{\partial t}=-\frac{\partial}{\partial R}\bigl(\dot R(R,t)\,f(R,t)\bigr),$$6 is droplet radius, $$\frac{\partial f}{\partial t}=-\frac{\partial}{\partial R}\bigl(\dot R(R,t)\,f(R,t)\bigr),$$7 bubble number density, and $$\frac{\partial f}{\partial t}=-\frac{\partial}{\partial R}\bigl(\dot R(R,t)\,f(R,t)\bigr),$$8 bubble radius. Neglecting the acceleration term yields a semi-analytical algebraic growth law for $$\frac{\partial f}{\partial t}=-\frac{\partial}{\partial R}\bigl(\dot R(R,t)\,f(R,t)\bigr),$$9, intended as a 3D mean-field closure for monodisperse interacting bubbles in a finite droplet (Saha et al., 2023).

Outside diffusion-controlled contexts, classical 3D single-bubble dynamics is governed by Rayleigh–Plesset-type laws. In the $R_c(t)\sim t^\alpha$0-dimensional generalization, the 3D case gives

$R_c(t)\sim t^\alpha$1

with Minnaert frequency $R_c(t)\sim t^\alpha$2, Rayleigh collapse time $R_c(t)\sim t^\alpha$3, and pressureless inertial scaling $R_c(t)\sim t^\alpha$4. In porous-media bubble growth governed by Darcy flow, the reduced field is the generalized Newtonian potential of the bubble indicator function, and global filling of space occurs if and only if the initial internal potential is quadratic (Klotz, 2012, Karp, 2010).

6. Scope, non-universality, and open issues

A recurring misconception is that 3D bubble growth has a single mean-field exponent. The literature summarized here does not support that view. The same spatial dimension supports at least four distinct asymptotic coarsening laws: $R_c(t)\sim t^\alpha$5 for diffusion-limited Ostwald ripening, $R_c(t)\sim t^\alpha$6 for reaction-limited ripening and dry-foam border blocking, $R_c(t)\sim t^\alpha$7 for viscous hydrodynamic dense-fluid coarsening, and $R_c(t)\sim t^\alpha$8 as a hydrodynamic inertial crossover law (A et al., 25 Sep 2025).

A second limitation is that mean-field validity depends strongly on mechanism. Molecular dynamics shows that reaction-limited 3D bubble ripening remains mean-field-like up to gas volume fractions of order $R_c(t)\sim t^\alpha$9, whereas diffusion-limited behavior becomes increasingly inconsistent with classical LSW as the gas volume fraction rises. Likewise, bubble clouds and multi-site electrolysis show that competition for diffusing gas can change prefactors, early-time exponents, and detachment statistics even when the late-time exponent remains diffusive (Watanabe et al., 2016, Vega-Martínez et al., 2020, Qin et al., 21 Jul 2025).

A third issue is morphological closure. The wet-foam mean-field model predicts a large population of small bubbles for any $n(t)\sim t^{-3\alpha}$0, but it excludes rattlers and adhesion by construction. The resulting scaling distribution therefore differs qualitatively from some simulations and experiments, which show stronger accumulation of very small bubbles and interpret that feature in terms of contact-free bubbles that drop out of the film-mediated dynamics (Morgan et al., 1 Aug 2025).

The open problems are correspondingly mechanism-specific. Dense-fluid bubble coarsening still lacks a fully analytic Navier–Stokes mean-field theory that predicts not only exponents but also prefactors and distribution shapes. The viscous-to-inertial crossover in three-dimensional hydrodynamic coarsening has not yet been resolved numerically in that setting. For foams, the effect of adhesion, rattlers, and bulk-diffusive corrections remains unresolved within a comparable analytic framework. For electrolysis and flash boiling, the natural next step is electrode- or droplet-scale closure laws in which nucleation-site density, coalescence, and interaction-modified Sherwood numbers enter explicitly as mean-field parameters. The term “three-dimensional mean-field bubble growth law” therefore denotes a family of 3D closures, unified by coarse-graining but differentiated by transport physics, geometry, and the mechanism that limits gas transfer.