In warm (ice-free) clouds, rain formation requires cloud droplets to grow by collision–coalescence across the 10–50 µm size gap within the ~20-minute lifetime of a convective cloud. Stochastic simulations with a turbulence-enhanced collection kernel show that a single runaway droplet (the 'gel' or raindrop embryo) emerges at a random gelation time, but the distribution of these times is known only empirically, for one turbulence intensity, a 1 cm³ volume, enhancement restricted to droplets of radius ≤ 21 µm, and without condensation growth. The open problem is to build a predictive theory of this sol–gel transition: an analytical or parameter-predictive model of the gelation-time distribution for realistic turbulent collection kernels, extended across turbulence intensities, droplet spectra, and volumes, and incorporating the combined effect of turbulent collection and condensation growth.

Background

Warm clouds produce rain without ice through collision and coalescence of small droplets, yet no complete explanation exists for how droplets cross the size gap between roughly 10 and 50 µm diameter fast enough to form raindrops within the ~20 min lifetime of a warm convective cloud. The classical description is the kinetic collection equation (KCE), a deterministic mean-field equation for the droplet mass distribution N(x,t)N(x,t) under a collection kernel K(x,y)K(x,y) giving the collision–coalescence rate of droplets of masses xx and yy; it is the cloud-physics form of the Smoluchowski coalescence equation. Because the KCE tracks only ensemble means, it misses stochastic fluctuations, which are one candidate explanation for rapid droplet growth (the 'lucky droplet' picture of Kostinski and Shaw, 2005: a few droplets experience unusually many collisions and grow much faster than the average). The fully stochastic alternative simulates individual collision events with the Gillespie stochastic simulation algorithm, so each realization resolves a finite droplet population.

In aggregation theory, a system with a multiplicative kernel K(x,y)=CxyK(x,y)=Cxy undergoes gelation (the sol–gel transition): the second mass moment diverges at a finite time Tgel=[CM2(0)]−1T_{\text{gel}}=[CM_2(0)]^{-1}, after which a single aggregate containing a finite fraction of the total mass separates from the continuous distribution. Alfonso, Raga and Baumgardner (2013) imported this concept into cloud physics: in stochastic simulations of a 1 cm³ cloud volume whose collection kernel is the gravitational hydrodynamic kernel (with tabulated collision efficiencies) multiplied by a turbulence-induced collision enhancement factor (computed by Pinsky, Khain and Krugliak, 2008 for a cumulonimbus with turbulent energy dissipation rate ε=0.1 m2 s−3\varepsilon=0.1\,\mathrm{m^2\,s^{-3}} and Taylor-scale Reynolds number — the Reynolds number based on the Taylor microscale, a standard measure of the intensity of turbulent fluctuations — of Reλ=2×104Re_\lambda=2\times10^4), a runaway droplet of mass comparable to the initial system mass appears, whereas the non-turbulent case shows no such transition. They estimated the gelation time per realization from the onset of growth of the mass ratio between the largest and second-largest droplet, and obtained the empirical distribution of gelation times from 1000 Monte Carlo realizations. Statistical-mechanical treatments of the sol–gel transition exist for special kernels (e.g., Matsoukas, 2015), but no analytical model was offered for this empirical distribution — the authors note that the fixed-collision-count convolution argument of Kostinski and Shaw (2005) does not transfer, because the number of collisions needed to build the runaway droplet differs from realization to realization.

Follow-up work clarified the picture but left the timing question open. Alfonso and Raga (2017) computed the post-gel droplet size distribution numerically and showed that KCE means deviate strongly from stochastic expectations after the transition, with correlations between size bins peaking near the critical time; they explicitly note that available analytical techniques apply only to special kernels. Alfonso, Raga and Baumgardner (2019) found observational evidence for gel formation: in early-stage orographic cloud data the mass of the largest droplet follows a mixture of Gaussian and Gumbel distributions, as expected in the pseudo-critical regime of a finite system — the regime near the gelation point where a finite population behaves analogously to a physical system near a critical point — an analytical model of the gel's mass distribution, not of its formation time. In parallel, modern observational and large-eddy-simulation studies (Chandrakar et al., 2024) confirm that turbulent coalescence must be included to reproduce observed rain formation, and Lagrangian-particle studies quantify thresholds for precipitation embryos to accelerate rain initiation (Lim et al., 2025) and simulate combined condensation–collection growth with and without turbulence (Iwashima and Onishi, 2026), but none of these derive or test gelation-time statistics.

Problem Statement

Establish a quantitative, predictive theory of the sol–gel transition in warm-cloud droplet growth by turbulent collision–coalescence, as introduced by Alfonso, Raga and Baumgardner (2013). The common objective is to predict when the runaway droplet (the gel or raindrop embryo) forms — its gelation time and the statistics thereof — as a function of the physically controlling parameters. Concretely, the requested results, which jointly serve this single objective, are: (1) an analytical, or otherwise parameter-predictive, model of the distribution of gelation times (equivalently its distribution function and key statistics such as mean and spread) for realistic turbulence-enhanced collection kernels, replacing the purely empirical Monte Carlo distribution of the 2013 study, for which the fixed-collision-count convolution rationale is known not to apply; (2) extension of the gelation-time statistics beyond the original setting — across turbulent energy dissipation rates and Reynolds numbers, cloud volumes, and initial droplet spectra, and with the turbulent collision enhancement applied across the full droplet-size range including radii above 21 µm, so that the role of turbulence in the later runaway-growth stage is not underestimated; and (3) incorporation of the combined effect of turbulent collection and condensation growth on the formation of the runaway droplet within the stochastic description.

References

  1. Alfonso, L., Raga, G. B., and Baumgardner, D.: The validity of the kinetic collection equation revisited – Part 3: Sol–gel transition under turbulent conditions, Atmos. Chem. Phys., 13, 521–529, 2013, DOI 10.5194/acp-13-521-2013, https://doi.org/10.5194/acp-13-521-2013
  2. Alfonso, L.: An algorithm for the numerical solution of the multivariate master equation for stochastic coalescence, Atmos. Chem. Phys., 15, 12315–12326, 2015, DOI 10.5194/acp-15-12315-2015, https://doi.org/10.5194/acp-15-12315-2015
  3. Alfonso, L., and Raga, G. B.: The impact of fluctuations and correlations in droplet growth by collision–coalescence revisited – Part 1: Numerical calculation of post-gel droplet size distribution, Atmos. Chem. Phys., 17, 6895–6905, 2017, DOI 10.5194/acp-17-6895-2017, https://doi.org/10.5194/acp-17-6895-2017
  4. Alfonso, L., Raga, G. B., and Baumgardner, D.: The impact of fluctuations and correlations in droplet growth by collision–coalescence revisited – Part 2: Observational evidence of gel formation in warm clouds, Atmos. Chem. Phys., 19, 14917–14932, 2019, DOI 10.5194/acp-19-14917-2019, https://doi.org/10.5194/acp-19-14917-2019
  5. Kostinski, A. B., and Shaw, R. A.: Fluctuations and luck in droplet growth by coalescence, Bull. Am. Meteorol. Soc., 86, 235–244, 2005, DOI 10.1175/bams-86-2-235, https://doi.org/10.1175/bams-86-2-235
  6. Matsoukas, E.: Statistical thermodynamics of irreversible aggregation: the sol–gel transition, Sci. Rep., 5, 8855, 2015, DOI 10.1038/srep08855, https://doi.org/10.1038/srep08855
  7. Pinsky, M., Khain, A., and Krugliak, H.: Collisions of cloud droplets in a turbulent flow. Part V: Application of detailed tables of turbulent collision rate enhancement to simulation of droplet spectra evolution, J. Atmos. Sci., 65, 357–374, 2008, DOI 10.1175/2007JAS2358.1, https://doi.org/10.1175/2007JAS2358.1
  8. Chandrakar, K. K., Morrison, H., Grabowski, W. W., and Lawson, R. P.: Are turbulence effects on droplet collision–coalescence a key to understanding observed rain formation in clouds?, Proc. Natl. Acad. Sci., 121, e2319664121, 2024, DOI 10.1073/pnas.2319664121, https://doi.org/10.1073/pnas.2319664121
  9. Lim, J.-S., Noh, Y., Lee, H., and Hoffmann, F.: The critical number and size of precipitation embryos to accelerate warm rain initiation, Atmos. Chem. Phys., 25, 5313–5329, 2025, DOI 10.5194/acp-25-5313-2025, https://doi.org/10.5194/acp-25-5313-2025
  10. Iwashima, M., and Onishi, R.: Direct Lagrangian tracking simulation of droplet growth in vertically-developing turbulent cloud, arXiv:2604.10564, 2026, https://arxiv.org/abs/2604.10564