betaFeedback

WorksPhysics and Astronomy

The electrostatic-hydrodynamic interaction of two spheres in a linear compressional flow

Abstract

We investigate the collision dynamics of two like-charged dielectric spheres in a uniaxial compressional flow, accounting for non-continuum lubrication interactions, van der Waals attraction, and the complete finite-dielectric electrostatic interaction. Mutual polarization causes the electrostatic force to behave substantially differently from the point-charge approximation and can produce strong near-field attraction despite far-field Coulomb repulsion, depending on the particle size and charge ratios. We identify a critical electrostatic strength above which particles approaching from the far field cannot reach contact. This threshold corresponds to a saddle-node bifurcation of the relative-trajectory field on the compressional axis. Below this threshold, near-field polarization-induced attraction can increase the upstream interception area and substantially enhance the collision efficiency. We report that electrostatic interactions lead to non-monotonic variations of collision efficiency with electrostatic strength, particle size ratio, and charge ratio. When electrostatic interactions remain repulsive at all separations, increasing their strength suppresses collisions monotonically. Non-continuum lubrication and van der Waals interactions further modify the collision efficiency, although sufficiently strong polarization-induced attraction can dominate these near-field effects. For weak electrostatic interactions, a perturbation analysis expresses the leading correction to the collision efficiency as a weighted average of the electrostatic force along the limiting colliding trajectory, with the weight concentrated in the lubrication region; this prediction agrees closely with the numerical results. These results demonstrate that finite-size polarization and near-field hydrodynamic interactions jointly govern the collision behaviour of like-charged particles in compressional flows.

Banner appropriate to article type will appear here in typeset article

Introduction

The transport and collision of particles suspended in a gaseous medium are central to a wide range of atmospheric and geophysical processes, including droplet growth in clouds and particle aggregation in volcanic plumes. In such systems, the relative motion of particle pairs

Abstract must not spill onto p.2 is governed by background flow and gravity, while hydrodynamic and non-hydrodynamic interactions can substantially modify their motions and collision outcomes. Of particular interest are electrostatic interactions, since cloud droplets and volcanic ash particles can acquire charge through several mechanisms, including ion attachment, inductive charging, and triboelectrification [45, 22, 37]. The dominant charging mechanisms, however, differ between these environments. Droplets in warm clouds can acquire charge primarily through the attachment and field-dependent collection of atmospheric ions [21], whereas volcanic ash is commonly electrified through particle-particle contact (triboelectrification) and fragmentation-related processes. The latter can additionally depend on environmental conditions such as temperature and water content [36]. The electrostatic forces arising from surface charges can either promote or inhibit particle encounters, depending on particle size and charge, thereby influencing collision-coalescence of water droplets in clouds and the aggregation of ash particles in volcanic plumes. Electrification is particularly evident in convective clouds and ash-rich volcanic plumes, where intense electrical activity and lightning provide striking manifestations of particle charging [59, 5, 24, 47]. These observations motivate a fundamental fluid-mechanical question: how does the coupling between electrostatic forces and hydrodynamic interactions modify the relative motions and collision dynamics of finite-sized particles in a background flow?

The fluid-mechanical framework for understanding particle collisions in turbulent flows has its origins in the seminal work of Saffman & Turner [51], who considered particles much smaller than the Kolmogorov length scale and approximated the turbulent velocity field experienced by an approaching pair as a quasi-steady uniaxial compressional flow. For sub-Kolmogorov particles, the flow varies linearly over the interparticle separation, and the instantaneous relative motion induced by turbulence can therefore be described in terms of the local velocity-gradient tensor. The straining component of this tensor drives particles towards one another, thereby serving as a crucial mechanism for collisions. Using Gaussian statistics for turbulent strain rates, Saffman & Turner [51] obtained the collision rate to be (8π/15)1/2n1n2Γη(a1+a2)3(8\pi/15)^{1/2} n_1 n_2 \Gamma_\eta(a_1+a_2)^3 where a1a_1 and a2a_2 are the particle radii, n1n_1 and n2n_2 are their respective number densities, and Γη=(ϵ/νf)1/2\Gamma_\eta=(\epsilon/\nu_f)^{1/2} is the Kolmogorov shear rate, with ϵ\epsilon and νf\nu_f denoting the turbulent dissipation rate and the kinematic viscosity of the surrounding fluid, respectively. Zeichner & Schowalter [60] calculated the collision rate of two non-interacting spheres subjected to a steady uniaxial compressional flow to be (8π/33)n1n2γ˙(a1+a2)3(8\pi/3\sqrt{3})n_1n_2\dot{\gamma}(a_1+a_2)^3, where γ˙\dot{\gamma} is the compression rate. Equating this expression with the Saffman–Turner collision rate yields γ˙=(9/40π)1/2Γη\dot{\gamma}=(9/40\pi)^{1/2}\Gamma_\eta, thereby relating the prescribed compressional flow to the characteristic small-scale strain rate of turbulence. The consideration of uniaxial compressional flow is particularly salient, as the most probable local strain in turbulence has two extensional and one compressional principal direction (Ashurst et al. 1987), a state whose axisymmetric idealization is the uniaxial compressional flow considered here. This characteristic motivates us to perform a focused analysis of the mechanics underlying particle-pair interactions in a canonical flow configuration. In the present investigation, we adopt the assumption of a frozen uniaxial compressional flow as the background flow field.

A local linear flow approximation does not fully capture the complexities of turbulence dynamics. Notably, turbulence and particle inertia can influence the collision rate via (i) altering the radial relative velocity of particle pairs and (ii) preferential concentration (Sundaram & Collins 1997; Chun et al. 2005; Falkovich & Pumir 2007). Consequently, a collision kernel obtained by approximating the background flow as a steady compressional flow should not be interpreted as the turbulent collision kernel. Previous studies on charged particles in turbulence show that electrostatic effects remain weak when Coulomb interactions are small compared with turbulence-induced relative motion, but become important once the two are comparable [48] (Lu & Shaw 2015). Furthermore, since the Coulomb force increases rapidly as particle separation decreases, electrostatics can play an important role during the final approach of particles, even when turbulence controls the larger-scale encounter dynamics. These studies, however, considered point-charge Coulomb interactions and did not include the finite-size polarization effects examined here. The trajectory-based formulation employed here addresses a more specific fluid-mechanical problem: it quantifies how hydrodynamic interactions and non-hydrodynamic forces modify pair trajectories and the collision dynamics for a prescribed straining flow. This formulation allows us to isolate the coupling between electrostatic and hydrodynamic interactions without conflating it with turbulence-induced clustering and relative-velocity statistics. Nevertheless, researchers have extended trajectory-based approaches to turbulent flows by approximating the background turbulent flow in the vicinity of an inertialess particle pair as a stochastically varying linear flow [7, 8, 13].

When two particles approach one another, the disturbance flow generated by each particle modifies the flow field experienced by the other, giving rise to hydrodynamic interactions that become increasingly important at small interparticle separations. These interactions can substantially modify pair trajectories and, consequently, influence collision rates. The effect of these hydrodynamic interactions can be quantified through the collision efficiency, defined as the ratio of the collision rate in the presence of hydrodynamic interactions to that in their absence. Two-body hydrodynamic interactions under Stokes-flow conditions have been studied extensively [18, 29]. In the case of rigid spheres, continuum lubrication resistance to normal relative motion scales as O(1/ξ)O(1/\xi), thereby preventing contact in finite time. Conversely, for viscous drops, interfacial mobility weakens the lubrication resistance, leading to an O(1/ξ)O(1/\sqrt{\xi}) scaling [11]. Here, ξ\xi is the closest surface-to-surface separation distance normalized by the mean radius of the two particles. The role of continuum hydrodynamic interactions on the collisions of rigid particles or viscous drops with uniaxial compressional flow has been examined previously [60, 58].

For particles suspended in a gaseous medium, like air, the continuum approximation becomes invalid when the interparticle separation is less than the molecular mean free path. Under such situations, one needs to consider non-continuum lubrication resistance, which scales as O(ln⁡(ln⁡(Kn/ξ))/Kn)O(\ln(\ln(Kn/\xi))/Kn) [55]. This non-continuum lubrication resistance exhibits a weaker divergence than its continuum counterpart, thereby allowing collisions in finite time. Here, the Knudsen number KnKn, which measures the strength of non-continuum interaction, is defined as the ratio of the mean free path of the medium to the mean radius of the particles. Previous studies have quantified the influence of non-continuum hydrodynamic interactions on collision dynamics for particles undergoing thermal fluctuations [42], suspended in laminar flows [14, 40], and turbulent flow [13, 12].

To assess the influence of surface charge on collision dynamics, both the magnitude and the polarity of the particle charges must be specified. Field measurements have established empirical relations between droplet charge and size [see 45, 4, 62, 39], and observed cloud-droplet charges lie well below the Rayleigh limit, beyond which electrostatic stresses destabilize the interface [46, 45]. Even such subcritical charges can modify collision dynamics, because mutual polarization redistributes surface charge during close approach and the interaction then departs markedly from the point-charge Coulomb force.

The electrostatic interaction between two charged spheres has been studied extensively, beginning with the inverse-distance expansion of Maxwell (1873), followed by the capacitance formulation of Russell (1909) and the bispherical-coordinate solution for conducting spheres developed by Davis (1964). The far- and near-field asymptotic behaviour of the electrostatic force between perfectly conducting spheres is also well established [32, 33]. A notable consequence of charge redistribution is that two like-charged conducting spheres may experience an attractive interaction at sufficiently small separations, except for charge ratios corresponding to equal surface potentials at contact. In the near-contact limit, this attraction exhibits an O(ξ−1[ln⁡ξ]−2)O\left(\xi^{-1}[\ln\xi]^{-2}\right) singularity [32], which is sufficiently strong to overcome continuum lubrication resistance and allow finite-time contact.

The perfect-conductor approximation is sufficiently accurate for particles with very high dielectric constants, such as metallic particles. However, this approximation is not appropriate for many common dielectric materials, including water, which has a relative permittivity of approximately 80. Consequently, cloud droplets may exhibit an electrostatic response that differs appreciably from that of perfect conductors, particularly at small interdroplet separations. Previous studies have shown that, depending on the particle size and charge ratios, two like-charged dielectric spheres can experience an attractive electrostatic interaction at sufficiently small surface-to-surface separations [6, 38, 27]. The physical origin of this behaviour lies in mutual polarization. At large separations, the leading electrostatic interaction is the monopole–monopole Coulomb contribution, which is repulsive for like-charged particles. As the particles approach, however, each particle increasingly polarizes the other, generating a spatially non-uniform bound-charge distribution. The resulting higher-order polarization contributions become progressively stronger and can overcome the Coulomb repulsion, causing the net electrostatic force to change from repulsive at large separation to attractive at close separation.

We define the size ratio κ=a2/a1\kappa= a_2/a_1 and the charge ratio β=q2/q1\beta= q_2/q_1, where q1q_1 and q2q_2 are the charges carried by the larger and smaller spheres of radii a1a_1 and a2a_2, respectively. For finite dielectric permittivity, the near-contact interaction is repulsive only within a finite band of β\beta at each κ\kappa, outside which polarization-induced attraction prevails; in the perfect-conductor limit this band collapses to a single curve [41]. Although the interaction between like-charged dielectric spheres approaches that between conducting spheres at large separations, the two differ substantially in the lubrication regime, where polarization-induced attraction and hydrodynamic resistance both become strong and compete directly in determining whether an approaching pair reaches contact.

The pairwise collision dynamics of colloidal particles in linear flows, incorporating the combined effects of hydrodynamic and colloidal interactions, including screened double-layer repulsion, have been studied extensively in classical colloid science [60, 49]. In a gas, by contrast, electrostatic interactions are unscreened and long-ranged, and mutual polarization becomes important at close approach. Most previous studies of electrostatically modulated collisions in gases have considered conducting spheres settling under gravity in a quiescent fluid [52, 53, 31, 28, 19, 15, 61], and, with few exceptions [15], they have relied on simplified descriptions of the hydrodynamic or electrostatic interactions that do not resolve their coupling in the lubrication regime. Patra et al. (2023) resolved this coupling for like-charged dielectric spheres settling under gravity. In an imposed linear flow, electrostatic effects on collisions in a gas have, to our knowledge, been examined with full hydrodynamic interactions only for uncharged droplets polarized by an external electric field [44], the settling counterpart of which was studied by Patra & Roy [43]. The interplay between non-continuum lubrication and the complete electrostatic interaction between charged, finite-dielectric particles in an imposed linear flow has not been addressed. This study aims to address this gap by combining non-continuum hydrodynamic interactions with the complete finite-dielectric electrostatic interaction to quantify the effects of mutual polarization on pair trajectory and collision efficiency of finite-sized charged particles in a uniaxial compressional flow.

The remainder of the paper is organized as follows. In § 2, we present the problem formulation and the framework used to evaluate the collision rate and efficiency. We briefly describe the numerical method in § 3. Then, in § 4, we show pair trajectories and the collision efficiency arising from the interplay between hydrodynamic and electrostatic interactions. Finally, in § 5, we summarize the principal findings.

Problem formulation

Trajectory equations for a particle pair

We consider a dilute polydisperse suspension of charged spheres subjected to a uniaxial compressional flow described by the velocity field U∞(x)=(γ˙x1,γ˙x2,−2γ˙x3)\mathbf{U}^{\infty}(\mathbf{x})=(\dot{\gamma}x_1,\dot{\gamma}x_2,-2\dot{\gamma}x_3), where γ˙\dot{\gamma} is the compression rate, x1x_1 and x2x_2 correspond to the extensional axes, and x3x_3 corresponds to the compressional axis. In many natural and industrial systems, such as atmospheric clouds and aerosol reactors, particle volume fractions are typically very low, about O(10−6)O(10^{-6}) (see [17, 2]). This allows us to treat these systems as dilute, in which the likelihood that a third particle affects the relative motion between two interacting particles is negligible. Therefore, our focus will be on the binary interactions of two spherical particles with radii a1a_1 and a2a_2 (see Figure 1).

Schematic of two like-charged particles in a uniaxial compressional flow

Figure 1. Schematic of two like-charged particles of radii a1a_1 and a2a_2 in a uniaxial compressional flow. Here, r\boldsymbol{r} is the centre-to-centre separation vector, and θ\theta is the angle between the compressional axis (i.e., x3x_3-axis) and r\boldsymbol{r}. Sphere ‘1’ denotes the satellite sphere and sphere ‘2’ the test sphere, while sphere ‘3’ represents the collision sphere of radius a1+a2a_1+a_2, centred on the test sphere. Representative trajectories of the satellite sphere relative to the test sphere are shown in blue. We denote the unit vectors in the rr and θ\theta directions as e^r\hat{\boldsymbol{e}}_r and e^θ\hat{\boldsymbol{e}}_\theta, respectively.

Before proceeding with the detailed analysis, it is essential to justify the key assumptions made in this study. The present investigation does not consider the role of surrounding fluid inertia. While particle inertia can significantly influence collision dynamics for larger particles, this effect is negligible for relatively small particles. Additionally, we assume that the particles are sufficiently large, so Brownian diffusion is negligible. The particle Reynolds number, denoted as RepRe_p, is defined in terms of the compression rate γ˙\dot{\gamma} and the radius of the larger particle a1a_1 as follows: Rep=ρfγ˙a12/μfRe_p=\rho_f\dot{\gamma}a_1^2/\mu_f, where ρf\rho_f and μf\mu_f are the density and dynamic viscosity of the surrounding fluid, respectively. The Stokes number, denoted by StSt, captures the effects of particle inertia and is defined as: St=γ˙τpSt=\dot{\gamma}\tau_p, with τp=2a12ρp/(9μf)\tau_p=2a_1^2\rho_p/(9\mu_f) representing the viscous relaxation time of the larger particle, where ρp\rho_p is the particle density. The Péclet number, denoted by PePe, measures the relative strengths of flow advection and Brownian diffusion and is given by: Pe=3πμfγ˙a13/kBTPe=3\pi\mu_f\dot{\gamma}a_1^3/k_BT. Here, kB=1.381×10−23 J K−1k_B=1.381\times10^{-23}\,\mathrm{J\,K^{-1}} is the Boltzmann constant, and TT is the absolute temperature. Let us calculate these relevant dimensionless quantities for a water droplet in air with a1=15 μma_1=15\,\mu\mathrm{m}, ρp≈103 kg m−3\rho_p\approx10^3\,\mathrm{kg\,m^{-3}}, ρf≈1 kg m−3\rho_f\approx1\,\mathrm{kg\,m^{-3}}, μf≈1.8×10−5 Pa s\mu_f\approx1.8\times10^{-5}\,\mathrm{Pa\,s}, T=298 KT=298\,\mathrm{K}, and γ˙=25 s−1\dot{\gamma}=25\,\mathrm{s^{-1}} corresponding to ε≈0.16 m2 s−3\varepsilon\approx0.16\,\mathrm{m^2\,s^{-3}}, representative of vigorous convective clouds. Using this data, we obtain the following estimates: Rep≈3.125×10−4Re_p\approx3.125\times10^{-4}, St≈0.069St\approx0.069 and Pe≈3.478×103Pe\approx3.478\times10^3. These typical values of RepRe_p, StSt, and PePe substantiate the assumptions made in the present study.

The characteristic hydrodynamic and capillary stresses for a spherical water droplet of radius a1a_1 in a uniaxial compressional flow scale as μfγ˙\mu_f\dot{\gamma} and Υ/a1\Upsilon/a_1, respectively, where Υ\Upsilon is the surface tension of the droplet-air interface, giving the flow capillary number Ca=μfγ˙a1/ΥCa=\mu_f\dot{\gamma}a_1/\Upsilon. For a1=15 μma_1=15\,\mu\mathrm{m}, γ˙=25 s−1\dot{\gamma}=25\,\mathrm{s^{-1}} and Υ=0.073 N m−1\Upsilon=0.073\,\mathrm{N\,m^{-1}}, Ca≈9.2×10−8Ca\approx9.2\times10^{-8}, indicating negligible flow-induced deformation. Moreover, for water droplets in typical warm-cloud conditions, the high droplet-to-medium viscosity ratio (≈102\approx10^{2}) suppresses interfacial mobility and internal circulation, thereby justifying the approximation of small cloud droplets as rigid spheres. Consistent with this approximation, Dhanasekaran et al. [14] showed that non-continuum lubrication and van der Waals interactions dominate over interfacial mobility in determining droplet collisions (see their figure 2).

Gravitational settling is not included in the present formulation. Its importance relative to the imposed strain is measured by the settling parameter Sv=4ρpg(a1−a2)/(9μfγ˙)S_{\mathrm{v}} = 4\rho_{\mathrm{p}}g(a_{1}-a_{2})/(9\mu_{\mathrm{f}}\dot{\gamma}), where gg is the gravitational acceleration; SvS_{\mathrm{v}} is the ratio of the difference in Stokes settling velocities of the two particles to the strain-induced relative velocity γ˙(a1+a2)/2\dot{\gamma}(a_{1}+a_{2})/2. For water droplets with a1=10 μma_{1}=10\ \mu\mathrm{m} and γ˙=25 s−1\dot{\gamma}=25\ \mathrm{s}^{-1}, Sv≈10S_{\mathrm{v}}\approx10 at κ=0.9\kappa=0.9 and Sv≈48S_{\mathrm{v}}\approx48 at κ=0.5\kappa=0.5, whereas Sv=0S_{\mathrm{v}}=0 for equal-sized droplets. Differential settling therefore dominates the relative motion of strongly bidisperse cloud droplets at these strain rates. The present results describe the strain-dominated limit Sv≲O(1)S_{\mathrm{v}}\lesssim O(1), which applies to near-equal pairs, for which straining is the only mechanism of approach, and to strongly strained environments. The coupled problem of settling in a compressional flow has been solved for uncharged spheres by Dhanasekaran et al. [14].

When Rep≪1Re_{\mathrm{p}}\ll1, the fluid motion around the particle pair can be accurately described using the Stokes equations. In the limit where both RepRe_{\mathrm{p}} and StSt are zero, we can express the relative velocity between two particles as a linear superposition of their relative velocities induced by the background flow, electrostatic forces, and van der Waals forces. For the current problem, the relative velocity between the two spheres, denoted by V12\mathbf{V}_{12}, is given as:

V12=V1−V2=E∞⋅r−[Arrr2+B(I−rrr2)]⋅(E∞⋅r)+16πμf(1a1+1a2)[Grrr2+H(I−rrr2)]⋅(Fe+FvdW)(1)\mathbf{V}_{12}=\mathbf{V}_{1}-\mathbf{V}_{2}=\mathbf{E}^{\infty}\cdot\mathbf{r}-\left[A\frac{\mathbf{r}\mathbf{r}}{r^{2}}+B\left(\mathbf{I}-\frac{\mathbf{r}\mathbf{r}}{r^{2}}\right)\right]\cdot\left(\mathbf{E}^{\infty}\cdot\mathbf{r}\right)+\frac{1}{6\pi\mu_{\mathrm{f}}}\left(\frac{1}{a_{1}}+\frac{1}{a_{2}}\right)\left[G\frac{\mathbf{r}\mathbf{r}}{r^{2}}+H\left(\mathbf{I}-\frac{\mathbf{r}\mathbf{r}}{r^{2}}\right)\right]\cdot\left(\mathbf{F}_{e}+\mathbf{F}_{\mathrm{vdW}}\right) \tag*{(1)}

where V1\mathbf{V}_{1} and V2\mathbf{V}_{2} denote the velocities of the satellite sphere (with radius a1a_{1}) and the test sphere (with radius a2a_{2}), respectively, E∞=[(∇U∞)+(∇U∞)T]/2\mathbf{E}^{\infty}=[(\nabla\mathbf{U}^{\infty})+(\nabla\mathbf{U}^{\infty})^{T}]/2 represents the strain rate tensor, r\mathbf{r} is the separation vector that points from the center of the test sphere to the center of the satellite sphere, r=∣r∣r=|\mathbf{r}|, I\mathbf{I} is the unit second-order tensor, and Fe\mathbf{F}_{e} and FvdW\mathbf{F}_{\mathrm{vdW}} represent the electrostatic and van der Waals forces, respectively. The mobility functions AA, BB, GG, and HH capture the hydrodynamic interactions. Specifically, AA and GG are axisymmetric mobilities, while BB and HH are asymmetric mobilities corresponding to linear flow and non-hydrodynamic forces, respectively. For continuum hydrodynamic interactions, these mobilities depend on the size ratio κ=a2/a1\kappa=a_{2}/a_{1}, and the dimensionless center-to-center distance r/a∗r/a^{*} of the particle pair, where a∗=(a1+a2)/2a^{*}=(a_{1}+a_{2})/2 represents the average radius of the two spherical particles. The dimensionless surface-to-surface distance ξ\xi, which we have used in the previous section, can be expressed as ξ=(r−(a1+a2))/a∗=(r/a∗)−2\xi=(r-(a_{1}+a_{2}))/a^{*}=(r/a^{*})-2. Methods for computing the continuum hydrodynamic mobilities, and their asymptotic forms at small and large separations, are well documented in the literature ([3, 30, 26, 58]).

The continuum assumption for hydrodynamic interactions would not be valid for particles colliding in a gaseous medium, particularly when the separation distance between the surfaces of the two spheres is less than the mean free path of the medium, denoted as λ0\lambda_{0}. The mean free path of air increases with altitude and is approximately 0.1 μm0.1\ \mu\mathrm{m} under typical warm cumulus cloud conditions ([57]). Since we consider scenarios where suspending media are gases, we must account for the non-continuum lubrication interactions. In these cases, hydrodynamic mobilities additionally depend on the Knudsen number, defined as Kn=λ0/a∗Kn=\lambda_{0}/a^{*}. The expressions of non-continuum lubrication forces for asymmetric motion have been derived by Sundararajakumar & Koch [55]. Building upon this foundational work, Dhanasekaran et al. [14] have recently incorporated the non-continuum lubrication interactions for axisymmetric mobilities. They considered continuum lubrication interactions for ξ>O(Kn)\xi> O(Kn) and non-continuum lubrication interactions for ξ≤O(Kn)\xi\le O(Kn). In the current study, we will utilize the uniformly valid solutions they developed for AA and GG. Continuum asymmetric mobilities BB and HH approach a finite value as ξ→0\xi\to0. Consequently, we anticipate that the breakdown of the continuum assumption will exert a negligible influence on the asymmetric relative motions of an inertialess particle pair. Therefore, we will consider continuum hydrodynamics for asymmetric mobilities at all separations.

The van der Waals attraction force, FvdW\boldsymbol{F}_{\mathrm{vdW}}, always acts along the line connecting two spheres and can be expressed in terms of the gradient of the van der Waals potential. Hamaker [20] derived an expression for the van der Waals potential using the principle of pairwise additivity; however, his calculation neglects electromagnetic retardation, which becomes significant when the separation distance between particles is of the order of, or exceeds, the London wavelength λL\lambda_L (≈0.1 μm\approx0.1\ \mathrm{\mu m}). In this work, we determine the van der Waals force from the expression for the retarded van der Waals potential given in Zinchenko & Davis [63]. The retarded form of FvdW\boldsymbol{F}_{\mathrm{vdW}} depends on r/a∗r/a^*, κ\kappa, AHA_H (the Hamaker constant), and NL=2π(a1+a2)/λL=2πa1(1+κ)/λLN_L = 2\pi(a_1+a_2)/\lambda_L = 2\pi a_1(1+\kappa)/\lambda_L (quantifies the dimensionless particle separation distance beyond which retardation effects become important). The non-dimensional van der Waals force fv\boldsymbol{f}_v, which we will use in subsequent analysis, is defined as fv=FvdW/(AH/a∗)=−fve^r\boldsymbol{f}_v = \boldsymbol{F}_{\mathrm{vdW}}/(A_H/a^*) = -f_v\hat{\boldsymbol{e}}_r.

In line with the work of Patra et al. [41], we account for the finite dielectric constants of the particles in our collision calculations. We assume that the particles are perfect dielectrics carrying prescribed free charges uniformly distributed over their surfaces, with no free charge within their interiors. The prescribed free surface charge is assumed not to redistribute during a particle encounter, and the net charge carried by each particle therefore remains constant. This assumption does not, however, imply a spatially uniform electrostatic response. The electric field generated by one particle polarizes the other, giving rise to a spatially non-uniform bound surface charge that varies with the instantaneous particle configuration. Thus, while the prescribed free surface charge remains fixed and uniform, the total surface charge distribution is generally non-uniform and evolves as the particles approach one another.

For particles suspended in air, the assumption that the net particle charge remains effectively constant during a collision event can be justified by comparing the characteristic encounter time with the time scale for charge relaxation through the surrounding gas. In the present uniaxial compressional flow, the characteristic particle relative velocity is Vrel∼γ˙a∗V_{\mathrm{rel}} \sim\dot{\gamma}a^{*}. Consequently, the corresponding collision time scale is given by tcoll=a∗/Vrel=1/γ˙t_{\mathrm{coll}} = a^{*}/V_{\mathrm{rel}} = 1/\dot{\gamma}; for γ˙=25 s−1\dot{\gamma} = 25\ \mathrm{s}^{-1}, this gives tcoll≈0.04 st_{\mathrm{coll}} \approx0.04\ \mathrm{s}. In a weakly conducting gas, charge relaxation occurs on the Maxwell relaxation time τ=ϵ0/σ\tau= \epsilon_{0}/\sigma, where ϵ0=8.85×10−12 C2N−1m−2\epsilon_{0} = 8.85 \times10^{-12}\ \mathrm{C}^{2}\mathrm{N}^{-1}\mathrm{m}^{-2} is the permittivity of free space and σ\sigma is the electrical conductivity of the surrounding medium. For clear air at sea level, σ≈2.3×10−14 S m−1\sigma\approx2.3 \times10^{-14}\ \mathrm{S}\ \mathrm{m}^{-1}, corresponding to τ≈6.5 min\tau\approx6.5\ \mathrm{min}. The lower electrical conductivity within warm clouds increases this relaxation time to approximately 20 min to 2 h ([45]). Therefore tcoll≪τt_{\mathrm{coll}} \ll\tau, implying that charge leakage through the surrounding gas is negligible over the collision time scale. This separation of time scales supports the assumption that the prescribed net charges remain constant during the collision process. For an aqueous cloud droplet, finite internal conductivity may facilitate the redistribution of free charge along the interfacial region. However, this effect is not accounted for in the current perfect-dielectric formulation and would require the coupling of ionic charge transport to the evolving electrostatic field.

Within this framework, we employ the formulation of Khachatourian et al. [27], who calculated the electrostatic interaction force between two charged dielectric spheres by deriving the potential field in a bispherical coordinate system. The non-dimensional electrostatic force fef_{\mathrm{e}}, defined as fe=Fe/[q12/(4πϵ0a12)]=Fee^rf_{\mathrm{e}} = F_{\mathrm{e}}/[q_{1}^{2}/(4\pi\epsilon_{0}a_{1}^{2})] = F_{\mathrm{e}}\hat{e}_{r}, depends on the size ratio κ\kappa, charge ratio β\beta, dielectric constants of sphere 1, sphere 2, and the surrounding fluid medium, denoted as k1k_{1}, k2k_{2}, and kmk_{\mathrm{m}}, respectively. For completeness, we briefly outline below the procedure used to compute the electrostatic force; further details are given in Khachatourian et al. [27].

The expression for the electrostatic force is given by:

Fe=−e^4 4πϵ0∑n=0∞Fn[n2Φ1,n−1eη1−(n+12)Φ1,n+n+12Φ1,n+1e−η1]Φ2,n.(2)\mathbf{F}_{\mathrm{e}}=-\hat{\mathbf{e}}_{4}\,4\pi\epsilon_{0}\sum_{n=0}^{\infty}\mathcal{F}_{n}\left[\frac{n}{2}\Phi_{1,n-1}e^{\eta_{1}}-\left(n+\frac{1}{2}\right)\Phi_{1,n}+\frac{n+1}{2}\Phi_{1,n+1}e^{-\eta_{1}}\right]\Phi_{2,n}. \tag*{(2)}

The potential coefficients Φ1,n\Phi_{1,n} and Φ2,n\Phi_{2,n}, which account for the mutual dielectric polarization of the two spheres, are obtained from two coupled recursive relations. One of them is

[(n+12)Φ1,ncosh⁡η1−n2Φ1,n−1−n+12Φ1,n+1](km+k1)+sinh⁡η12(km−k1)Φ1,n\left[\left(n+\frac{1}{2}\right)\Phi_{1,n}\cosh\eta_{1}-\frac{n}{2}\Phi_{1,n-1}-\frac{n+1}{2}\Phi_{1,n+1}\right](k_{\mathrm{m}}+k_{1})+\frac{\sinh\eta_{1}}{2}(k_{\mathrm{m}}-k_{1})\Phi_{1,n}
−[(n+12)Φ2,nFncosh⁡η1−n2Φ2,n−1Fn−1−n+12Φ2,n+1Fn+1](km−k1)+Fnsinh⁡η12(km−k1)Φ2,n=2Aq1e−(n+12)η14πϵ0a12,(3)-\left[\left(n+\frac{1}{2}\right)\Phi_{2,n}\mathcal{F}_{n}\cosh\eta_{1}-\frac{n}{2}\Phi_{2,n-1}\mathcal{F}_{n-1}-\frac{n+1}{2}\Phi_{2,n+1}\mathcal{F}_{n+1}\right](k_{\mathrm{m}}-k_{1}) +\mathcal{F}_{n}\frac{\sinh\eta_{1}}{2}(k_{\mathrm{m}}-k_{1})\Phi_{2,n} =\frac{\sqrt{2}\mathcal{A}q_{1}e^{-\left(n+\frac{1}{2}\right)\eta_{1}}}{4\pi\epsilon_{0}a_{1}^{2}}, \tag*{(3)}

with the other one obtained by interchanging subscripts 1 and 2 in (3). The expressions for η1\eta_1, η2\eta_2, A\mathcal{A} and Fn\mathcal{F}_n involved in the above two recursive relations are given by

η1=cosh⁡−1[(1+κ)(ξ+2)4+(1−κ2)(1+κ)(ξ+2)],(4)\eta_1=\cosh^{-1}\left[\frac{(1+\kappa)(\xi+2)}{4}+\frac{(1-\kappa^2)}{(1+\kappa)(\xi+2)}\right], \tag*{(4)}
η2=cosh⁡−1[(1+κ)(ξ+2)4κ−(1−κ2)κ(1+κ)(ξ+2)],(5)\eta_2=\cosh^{-1}\left[\frac{(1+\kappa)(\xi+2)}{4\kappa}-\frac{(1-\kappa^2)}{\kappa(1+\kappa)(\xi+2)}\right], \tag*{(5)}
A=a1(1+κ)2(ξ+2)216+(1−κ2)2(1+κ)2(ξ+2)2−(1+κ2)2,(6)\mathcal{A}=a_1\sqrt{\frac{(1+\kappa)^2(\xi+2)^2}{16}+\frac{(1-\kappa^2)^2}{(1+\kappa)^2(\xi+2)^2}-\frac{(1+\kappa^2)}{2}}, \tag*{(6)}
Fn=e−(n+12)(η1+η2).(7)\mathcal{F}_n=e^{-(n+\frac{1}{2})(\eta_1+\eta_2)}. \tag*{(7)}

Solving the two coupled recursive relations for Φ1,n\Phi_{1,n} and Φ2,n\Phi_{2,n} and substituting the resulting coefficients into the force expression above yields the electrostatic interaction force at each particle separation.

To simplify the analysis, we express the particle relative velocity equation in a spherical coordinate system, with the origin located at the center of the test droplet. We nondimensionalize the relative velocity equation using characteristic scales for length, velocity, and time, which are defined as a∗a^*, γ˙a∗\dot{\gamma}a^*, and γ˙−1\dot{\gamma}^{-1}, respectively. Because of this scaling, the nondimensional radial separation between the centers of the two spheres can vary from 22 to ∞\infty, and henceforth we denote this nondimensional radial distance by rr. The sphere with a dimensionless radius of r=2r=2 is referred to as the collision sphere. In contrast, at r=∞r=\infty, the influence of one sphere on the other becomes negligible. We assume a1>a2a_1>a_2, allowing the size ratio to vary within the range (0,1](0,1]. Performing the vector and tensor operations in Eq. (1) yields the components of the dimensionless relative velocity defined by v=V12/γ˙a∗\mathbf{v}=\mathbf{V}_{12}/\dot{\gamma}a^*, expressed in the radial (rr), polar (θ\theta), and azimuthal (ϕ\phi) directions as follows:

vr=−r(1−A)(3cos⁡2θ−1)+NeGfe−NvGfv,(8)v_r=-r(1-\mathcal{A})\left(3\cos^2\theta-1\right)+N_eGf_e-N_vGf_v, \tag*{(8)}
vθ=3r(1−B)sin⁡θcos⁡θ,(9)v_\theta=3r(1-\mathcal{B})\sin\theta\cos\theta, \tag*{(9)}
vϕ=0.(10)v_\phi=0. \tag*{(10)}

Here, NeN_e and NvN_v are the dimensionless quantities that measure the strengths of the electrostatic and retarded van der Waals forces, respectively, relative to the background compressional flow:

Ne=q1212π2μfϵ0a14γ˙κ,(11)N_e=\frac{q_1^2}{12\pi^2\mu_f\epsilon_0a_1^4\dot{\gamma}\kappa}, \tag*{(11)}
Nv=2AH3πγ˙μfκ(1+κ)a13.(12)N_v=\frac{2A_H}{3\pi\dot{\gamma}\mu_f\kappa(1+\kappa)a_1^3}. \tag*{(12)}

To place these parameters in an atmospheric context, we estimate γ˙\dot{\gamma} from the turbulent dissipation rate ε\varepsilon using the relation introduced in § 1. As a result, the computed values of NeN_e and NvN_v correspond to strain rates representative of atmospheric turbulence, while the underlying calculations adhere to the assumptions of a steady, uniaxial compressional flow.

For warm clouds, reported dissipation rates are ε∼10−3\varepsilon\sim10^{-3}–10−1 m2 s−310^{-1}\ \mathrm{m^2\,s^{-3}} [17], corresponding to γ˙∼2\dot{\gamma}\sim2–20 s−120\ \mathrm{s^{-1}} for νf≈1.8×10−5 m2 s−1\nu_f\approx1.8\times10^{-5}\ \mathrm{m^2\,s^{-1}}. Cloud droplets with radii a1∼10a_1\sim10–20 μm20\ \mu\mathrm{m} may carry charges of order 1010–102e10^2 e [45]. Here, e=1.602×10−19 Ce=1.602\times10^{-19}\ \mathrm{C} is the elementary charge. For these ranges, Eq. (11) gives Ne=O(10−5)N_e = O(10^{-5})–O(100)O(10^{0}). Electrostatic interactions may therefore become appreciable for relatively highly charged small droplets subjected to weak local strain. Similarly, using AH∼4×10−20A_H \sim4 \times10^{-20} J for water-based particles [25], Eq. (12) gives Nv=O(10−3)N_v = O(10^{-3})–O(10−1)O(10^{-1}).

Volcanic plumes represent a markedly different energetic regime. Proximal volcanic plumes may exhibit dissipation rates of ε∼10−1\varepsilon\sim10^{-1}–102 m2 s−310^{2}\ \mathrm{m}^{2}\,\mathrm{s}^{-3} [56], corresponding to γ˙∼20\dot{\gamma} \sim20–700 s−1700\ \mathrm{s}^{-1}. Measurements of silicate ash particles with radii a1∼30a_1 \sim30–100 μm100\ \mu\mathrm{m} indicate charges of approximately q1∼10−15q_1 \sim10^{-15}–10−13 C10^{-13}\ \mathrm{C} [23, 37]. For these representative ranges, Eq. (11) gives Ne=O(10−3)N_e = O(10^{-3})–O(104)O(10^{4}). Similarly, using AH∼(3A_H \sim(3–10)×10−20 J10) \times10^{-20}\ \mathrm{J} for silicate particles, Eq. (12) gives Nv=O(10−7)N_v = O(10^{-7})–O(10−3)O(10^{-3}). Particle inertia becomes significant across much of the volcanic ash parameter range. For silicate ash with ρp≈2.5×103 kg m−3\rho_p \approx2.5 \times10^{3}\ \mathrm{kg}\,\mathrm{m}^{-3}, the Stokes number varies roughly between O(1)O(1) and O(102)O(10^{2}) for the above particle sizes and strain rates. Consequently, the current inertialess formulation applies solely to sufficiently fine ash particles. For example, St≲0.3St \lesssim0.3 requires a1≲20 μma_1 \lesssim20\ \mu\mathrm{m} at γ˙≈20 s−1\dot{\gamma} \approx20\ \mathrm{s}^{-1} and a1≲4 μma_1 \lesssim4\ \mu\mathrm{m} at γ˙≈700 s−1\dot{\gamma} \approx700\ \mathrm{s}^{-1}. Therefore, the estimates of NeN_e and NvN_v for coarser ash particles should be interpreted as indicative of the interaction strengths achievable in volcanic plumes, rather than as parameter ranges to which the current inertialess pair dynamics can be quantitatively applied.

These estimates demonstrate that electrostatic and van der Waals interactions depend strongly on particle size, charge, and local strain rate in both scenarios. They also highlight a fundamental contrast: while electrostatic interactions typically remain weak in warm clouds, they may become comparable to or exceed the local flow-induced contribution in electrically active volcanic environments.

It is important to note that temperature is not varied independently in the present formulation. While calculating the above non-dimensional parameters, the influence of temperature enters primarily through the temperature-dependent density, viscosity, and molecular mean free path of the suspending medium, as well as the particle/droplet density. Similarly, the charging process itself is not modeled dynamically; instead, q1q_1 and q2q_2 are prescribed parameters, with representative charge ranges chosen separately for cloud droplets and volcanic ash particles. Thus, the present calculations quantify the collision dynamics for specified particle charges rather than the temperature-dependent generation of those charges.

Expressions for the collision rate and efficiency

The collision rate, denoted by K12K_{12}, quantitatively measures the rate at which particles of radii a1a_1 and a2a_2 with their respective number densities n1n_1 and n2n_2 collide with each other per unit volume. Mathematically, it can be expressed as the flux of particles into the collision surface, as given by the following equation:

K12=−18n1n2γ˙(a1+a2)3∫(r=2) & (v⋅n<0)(v⋅n)P(r) dA(13)K_{12}=-\frac{1}{8}n_1n_2\dot{\gamma}(a_1+a_2)^3\int_{(r=2)\,\&\,(\mathbf{v}\cdot\mathbf{n}<0)}(\mathbf{v}\cdot\mathbf{n})P(r)\,\mathrm{d}A \tag*{(13)}

where P(r)P(r) denotes the pair distribution function and n\mathbf{n} is the outward unit normal vector at the collision surface. The condition v⋅n<0\mathbf{v}\cdot\mathbf{n}<0 ensures that only inward-directed radial relative velocities at the collision surface contribute to the collision rate. For a dilute dispersion, the pair-distribution function is governed by the quasi-steady Fokker–Planck equation applicable to the spatial region external to the contact surface:

∇⋅(Pv)=0.(14)\nabla\cdot(P\mathbf{v})=0. \tag*{(14)}

As the particle motions in the far field become uncorrelated, this leads to the boundary condition: P→1P \to1 as r→∞r \to\infty. For calculation purposes, we consider r=r∞r = r_{\infty}, representing a large, yet finite, separation distance.

The two equations above do not include any diffusive flux, permitting the evaluation of the collision rate through trajectory analysis. By applying Eq. (14) in conjunction with the divergence theorem, the integral in Eq. (13) can be reformulated over the closed surface that bounds all trajectories originating at r=r∞r = r_{\infty} and terminating at r=2r = 2. As a result, the flux through the cross-section of this volume at r=r∞r = r_{\infty} - referred to as the upstream interception area, AcA_c - defines the collision rate. Given that the pair-distribution function satisfies the condition P=1P = 1 at r=r∞r = r_{\infty}, Eq. (13) reduces to

K12=−18n1n2γ˙(a1+a2)3∫Ac(v⋅n′)∣r∞ dA,(15)K_{12} = -\frac{1}{8} n_1 n_2 \dot{\gamma} (a_1 + a_2)^3 \int_{A_c} \left(\mathbf{v}\cdot\mathbf{n}'\right) \big|_{r_{\infty}} \, \mathrm{d}A, \tag*{(15)}

where n′\mathbf{n}' denotes the outward unit normal vector on the area elements of AcA_c. It is important to note that the term (v⋅n′)∣r∞\left(\mathbf{v}\cdot\mathbf{n}'\right) \big|_{r_{\infty}} can be evaluated by considering the contribution from the background flow only, as electrostatic and van der Waals forces are negligible at r=r∞r = r_{\infty}.

The collision rate calculated without considering any interactions, whether hydrodynamic or non-hydrodynamic, is referred to as the ideal collision rate, denoted by K120K^0_{12}. For a uniaxial compressional flow, Zeichner & Schowalter [60] derived the following expression for the ideal collision rate:

K120=8π33n1n2γ˙(a1+a2)3.(16)K^0_{12} = \frac{8\pi}{3\sqrt{3}} n_1 n_2 \dot{\gamma} (a_1 + a_2)^3. \tag*{(16)}

Finally, the ratio of actual collision rate, K12K_{12}, to the ideal collision rate, K120K^0_{12}, defines the collision efficiency, denoted as E12E_{12}:

E12=K12K120.(17)E_{12} = \frac{K_{12}}{K^0_{12}}. \tag*{(17)}

Numerical method

To calculate the collision rate, it is essential to determine the upstream interception area, as evident from Eq. (15). We achieve this by employing a trajectory analysis methodology, which involves initializing a test sphere at the origin and evolving the satellite spheres to identify those that collide. Since the particle relative velocity does not depend on the azimuthal coordinate ϕ\phi, we can effectively describe the dynamics by analysing the relative trajectories in a generic rsin⁡θ−rcos⁡θr\sin\theta-r\cos\theta plane. We find the relative trajectories of a particle pair by integrating the following dimensionless trajectory equation:

dθdr=vθrvr=3(1−B)sin⁡θcos⁡θ−r(1−A)(3cos⁡2θ−1)+NeGfe−NvGfv.(18)\frac{\mathrm{d}\theta}{\mathrm{d}r} = \frac{v_{\theta}}{r v_r} = \frac{3(1-B)\sin\theta\cos\theta}{-r(1-A)(3\cos^2\theta-1)+N_eGf_e-N_vGf_v}. \tag*{(18)}

This equation captures the relative motion of two spherical particles, influenced by the combined effects of a uniaxial compressional flow, hydrodynamic interactions, electrostatic forces, and van der Waals forces.

Colliding trajectories are defined as the paths of satellite sphere centers originating far upstream and terminating at the collision surface. In the far-field, these colliding trajectories constitute the upstream interception area AcA_c. Direct computation of colliding trajectories by choosing initial conditions over a spherical shell of dimensionless radius r∞r_{\infty} is computationally expensive, as most trajectories originating from this spherical shell do not reach the collision sphere. To address this limitation, we exploit the quasi-steady nature of the relative trajectory equation and perform backward integration using a fourth-order Runge–Kutta method, with initial conditions specified on the collision surface. At exactly r=2r=2, however, the radial relative velocity vanishes because the relevant hydrodynamic mobilities A=1A=1 and G=0G=0 at the contact surface. To circumvent this singularity, we choose our initial conditions on a slightly expanded sphere of radius 2+δ2+\delta, where δ\delta is a slight offset from the collision surface. We obtain converged results with reasonable computational cost for δ=10−6\delta=10^{-6}. Furthermore, as evident from Eq. (13), only trajectories with inward radial velocity at the collision sphere contribute to collisions; therefore, we further reduce computational effort by choosing initial conditions on the collision sphere where vr<0v_r<0.

Among these computed colliding trajectories, the limiting colliding trajectories exhibit maximum horizontal shifts from the compressional axis in the far field. The upstream interception areas for this problem are two identical circles whose radius is the far-field horizontal shift of these limiting colliding trajectories. After identifying the upstream interception areas, we calculate the collision efficiency. We use this numerical methodology to compute the collision efficiency for a range of values for κ\kappa, KnKn, β\beta, NeN_e, and NvN_v.

Results and discussion

Electrostatic modification of charged-tracer collisions

Before turning to the fully coupled hydrodynamic-electrostatic problem, we first examine how electrostatic interactions alone modify the collision rate between two charged tracers in a uniaxial compressional flow. We consider two tracers of unequal size and charge and neglect hydrodynamic interactions by taking A,B→0A,B\to0 and G→1G\to1. Retaining the monopole-monopole interaction and the leading monopole-induced-dipole correction, the radial component of the relative velocity is given by

vr=−r(3cos⁡2θ−1)+4Ne(1+κ)2[βr2−16(β2+κ3)K(1+κ)3r5].(19)v_r=-r\left(3\cos^2\theta-1\right)+\frac{4N_e}{(1+\kappa)^2}\left[\frac{\beta}{r^2}-\frac{16\left(\beta^2+\kappa^3\right)\mathcal{K}}{(1+\kappa)^3r^5}\right]. \tag*{(19)}

In the above equation, the parameter K=(k−1)/(k+2)\mathcal{K}=(k-1)/(k+2), where k=k1=k2k=k_1=k_2, denotes the Clausius–Mossotti factor, which quantifies the relative polarizability of the particle in its surrounding medium; for water droplets in air, K≈0.96\mathcal{K}\approx0.96.

To compute the critical angle θc\theta_c associated with the limiting trajectory that reaches the collision surface, we set vr=0v_r=0 at r=2r=2 in Eq. (19). This gives

θc=cos⁡−1[13{1+N~e8(1−2(β2+κ3)Kβ(1+κ)3)}1/2],(20)\theta_c=\cos^{-1}\left[\frac{1}{\sqrt{3}}\left\{1+\frac{\widetilde{N}_e}{8}\left(1-\frac{2\left(\beta^2+\kappa^3\right)\mathcal{K}}{\beta(1+\kappa)^3}\right)\right\}^{1/2}\right], \tag*{(20)}

where we have defined the rescaled electrostatic parameter as N~e=4βNe/(1+κ)2\widetilde{N}_e=4\beta N_e/(1+\kappa)^2.

The purely Coulombic limit follows directly by setting K=0\mathcal{K}=0, thereby retaining only monopole-monopole repulsion. In this limit, the relative velocity field is solenoidal, so the pair-conservation equation ∇⋅(Pv)=0\nabla\cdot(P\mathbf{v})=0, together with P→1P\to1 in the far field, gives P=1P=1 for all values of rr. The collision rate can therefore be evaluated directly from the inward flux across the collision sphere, yielding

K12C=π2[1633(1+N~e8)3/2−N~e]n1n2γ˙(a1+a2)3.(21)K_{12}^{\mathrm{C}}=\frac{\pi}{2}\left[\frac{16}{3\sqrt{3}}\left(1+\frac{\widetilde{N}_e}{8}\right)^{3/2}-\widetilde{N}_e\right]n_1n_2\dot{\gamma}(a_1+a_2)^3. \tag*{(21)}

Here, K12CK_{12}^{\mathrm{C}} denotes the collision rate in the presence of Coulombic repulsion and is distinct from the ideal collision rate K120K_{12}^{0} defined in § 2.2, for which all kinds of interactions are absent.

Indeed, for N~e=0\widetilde{N}_e=0, K12C=K120K_{12}^{\mathrm{C}}=K_{12}^{0}, recovering the classical ideal collision rate for a uniaxial compressional flow. Coulombic repulsion therefore monotonically reduces the collision rate from its ideal value. At N~e=16\widetilde{N}_e=16, the region vr<0v_r<0 on the collision sphere collapses to the compressional axis, while for N~e>16\widetilde{N}_e>16 no trajectory originating in the far field reaches contact and hence K12C=0K_{12}^{\mathrm{C}}=0 (see Figure 2).

Figure 2. Variation of the scaled collision rate, K12C/[n1n2γ˙(a1+a2)3]K_{12}^{\mathrm{C}}/[n_1n_2\dot{\gamma}(a_1+a_2)^3], with the scaled electrostatic parameter N~e\widetilde{N}_e for a pair of like-charged tracers interacting through Coulombic repulsion in a uniaxial compressional flow. The collision rate decreases monotonically with increasing N~e\widetilde{N}_e and vanishes at the critical value N~e=16\widetilde{N}_e=16, beyond which Coulombic repulsion prevents particle contact. The horizontal dashed line represents the ideal collision rate for a uniaxial compressional flow alone, K120/[n1n2γ˙(a1+a2)3]=8π/(33)K_{12}^{0}/[n_1n_2\dot{\gamma}(a_1+a_2)^3]=8\pi/(3\sqrt{3}).

Polarization effects can significantly alter the collision rate due to Coulombic repulsion. At contact, the leading electrostatic contribution in (19) changes sign when

2(β2+κ3)K=β(1+κ)3,(22)2(\beta^2+\kappa^3)\mathcal{K}=\beta(1+\kappa)^3, \tag*{(22)}

which leads to the two distinct boundaries defined by

β±=(1+κ)3±(1+κ)6−16K2κ34K.(23)\beta_{\pm}=\frac{(1+\kappa)^3\pm\sqrt{(1+\kappa)^6-16\mathcal{K}^2\kappa^3}}{4\mathcal{K}}. \tag*{(23)}

The electrostatic force map shown in Figure 3 illustrates how the sign of the near-contact interaction depends on the particle size and charge ratios. Within the range where β−<β<β+\beta_-<\beta<\beta_+, the interaction is repulsive, whereas outside this interval, the polarization-induced attraction dominates. These two curves are superimposed on the full electrostatic force at contact, obtained from the finite-dielectric formulation of Khachatourian et al. [27]. The complete calculations reveal a substantially narrower repulsive region compared to the monopole-dipole approximation, demonstrating that higher-order polarization contributions strengthen the near-field attraction and significantly modify the boundary between attraction and repulsion. Additionally, in Figure 3, the yellow line represents the condition under which the interaction between two like-charged conducting spheres becomes repulsive, providing a point of reference for comparison.

Electrostatic interaction force near contact in the $(\xi = 10^{-6})$ parameter space for two like-charged water droplets in air

Figure 3. Electrostatic interaction force near contact (ξ=10−6)(\xi= 10^{-6}) in the (κ,β)(\kappa, \beta) parameter space for two like-charged water droplets in air. The two curves β−\beta_{-} and β+\beta_{+} represent the attraction-repulsion boundaries predicted by the monopole-dipole approximation, with repulsion occurring for β−<β<β+\beta_{-} < \beta< \beta_{+}. The background map shows the sign of the full electrostatic force calculated using the finite-dielectric formulation of Khachatourian et al. [27]; the dark-red region denotes repulsive interactions, whereas interactions are attractive outside this region. The yellow curve represents the condition for which the electrostatic interaction between two like-charged conducting spheres is repulsive.

In contrast to the purely Coulombic interaction, the polarization contribution renders the relative velocity field non-solenoidal, so that P≠1P\ne1 at finite separation. Consequently, the collision rate cannot be accurately estimated based solely on the electrostatic force at contact or the critical angle. Furthermore, the sign of the electrostatic force at contact does not exclusively dictate whether a pair of particles will collide; it merely indicates whether the electrostatic interactions locally facilitate or impede their radial approach at that point. In the complete problem, the electrostatic force further competes with the imposed compressional flow and lubrication resistance. We therefore employ the full finite-dielectric electrostatic interaction in the calculations that follow and determine the collision efficiency by analysing pair trajectories rather than from the truncated far-field interaction.

Critical electrostatic strength for collision suppression

The preceding analysis of charged tracers demonstrated that sufficiently strong Coulombic repulsion can entirely inhibit collisions in a uniaxial compressional flow. We now examine whether a similar threshold exists when considering hydrodynamic interactions and the complete electrostatic interaction between finite-size dielectric spheres. We denote this threshold by (Ne)c(N_e)_c, defined as the critical value of NeN_e above which no trajectory starting from the far field can reach the collision surface. Given that the compressional axis (i.e., θ=0\theta= 0) provides the maximum inward flow-induced radial velocity, the absence of inward motion along this axis determines the limiting condition for collision. For prescribed values of κ\kappa, β\beta, and KnK_n, we compute (Ne)c(N_e)_c by minimizing 2r(1−A)/(Gfe)2r(1-A)/(G f_e) over all separations for which fe>0f_e > 0. Mathematically, this is expressed as:

(Ne)c=min⁡r≥2fe(r)>0[2r(1−A)Gfe].(24)(N_e)_c = \min_{\substack{r \ge2 \\ f_e(r)>0}} \left[\frac{2r(1-A)}{G f_e}\right]. \tag*{(24)}

At the critical separation rcr_c, the radial relative velocity satisfies vr(rc,θ=0)=0v_r(r_c,\theta= 0) = 0 and ∂vr/∂r∣r=rc=0\left.\partial v_r/\partial r\right|_{r=r_c} = 0. It is important to note that the minimization is confined to those separations where the electrostatic force is repulsive, as an attractive electrostatic force reinforces rather than opposes the inward motion. Physically, (Ne)c(N_e)_c measures the strength at which the largest electrostatic barrier encountered during approach first balances the inward motion induced by the compressional flow. For Ne<(Ne)cN_e < (N_e)_c, the flow can overcome this barrier, thereby permitting the occurrence of collisions. Conversely, for Ne>(Ne)cN_e > (N_e)_c, a region with vr>0v_r > 0 develops during approach, effectively preventing particles arriving from the far field from reaching the collision sphere.

We first consider combinations of (κ,β)(\kappa,\beta) for which the electrostatic force is repulsive at large separations but becomes attractive in the near field. A representative example is κ=0.5\kappa= 0.5 and β=10\beta= 10, for which we obtain (Ne)c≈30.4(N_e)_c \approx30.4 at Kn=10−2Kn = 10^{-2} and Nv=0N_v = 0. Figure Figure 4 illustrates the corresponding radial relative velocity, vrv_r, as a function of the dimensionless surface-to-surface separation ξ\xi for values of NeN_e that are below, at, and above (Ne)c(N_e)_c. Although the near-field electrostatic attraction promotes the final approach to contact, a particle pair approaching one another from the far field must first navigate through an intermediate region where the electrostatic interaction is repulsive. For Ne<(Ne)cN_e < (N_e)_c, the compressional flow is strong enough to overcome this repulsive barrier, resulting in vrv_r remaining negative throughout the approach and allowing the particles to reach the near-field region where the electrostatic attraction further enhances the likelihood of collision. At Ne=(Ne)cN_e = (N_e)_c, the inward flow and electrostatic repulsion balance at the critical separation ξc=rc−2\xi_c = r_c - 2, such that vr(ξc)=0v_r(\xi_c) = 0, while vr≤0v_r \le0 elsewhere. As shown in the inset of Figure Figure 4, increasing NeN_e beyond (Ne)c(N_e)_c results in the emergence of a finite interval over which vr>0v_r > 0. This resulting outward relative motion prevents particles approaching from the far field from overcoming the repulsive electrostatic barrier, thus inhibiting their access to the attractive near-field region. Therefore, the presence of an attractive electrostatic force near contact does not automatically ensure a collision; rather, the likelihood of collision is determined by the interplay between the compressional flow and the separation-dependent electrostatic interaction throughout the particle approach.

Figure 4. The radial component of the relative velocity, vrv_r, along the compressional axis (θ=0\theta= 0) as a function of the dimensionless surface-to-surface separation ξ\xi for NeN_e less than, equal to and greater than (Ne)c(N_e)_c, with Kn=10−2Kn = 10^{-2}, κ=0.5\kappa= 0.5, β=10\beta= 10, k1=k2=80k_1 = k_2 = 80 and Nv=0N_v = 0. The inset magnifies the region near vr=0v_r = 0: at Ne=(Ne)cN_e = (N_e)_c, vrv_r reaches zero at a single separation, whereas for Ne>(Ne)cN_e > (N_e)_c it becomes positive over a finite interval of ξ\xi. This positive vrv_r region represents an electrostatic repulsive barrier.

A qualitatively distinct scenario emerges when the electrostatic interaction remains repulsive in the near field. For instance, at κ=0.6\kappa= 0.6, the electrostatic force becomes repulsive in the near field for approximately 0.3≲β≲0.50.3 \lesssim\beta\lesssim0.5. Taking β=0.4\beta= 0.4 gives (Ne)c≈70.1(N_e)_c \approx70.1 for Kn=10−2Kn = 10^{-2} and Nv=0N_v = 0. Under these conditions, electrostatic repulsion prevents particles from approaching each other at all separation distances. When Ne<(Ne)cN_e < (N_e)_c, the inward velocity caused by the compressional flow is sufficiently large that the total radial velocity remains negative, allowing collisions to occur. Conversely, when Ne>(Ne)cN_e > (N_e)_c, the electrostatic repulsion forces surpass the contribution from the inward flow across a finite range of separations, resulting in vr>0v_r > 0 and inhibiting collisions. Figure 5 illustrates these two regimes.

Figure 5. Radial component of the relative velocity, vrv_r, along the compressional axis (θ=0\theta= 0) as a function of the dimensionless centre-to-centre distance rr, for (a) Ne<(Ne)cN_e < (N_e)_c and (b) Ne>(Ne)cN_e > (N_e)_c. The flow-induced (black), electrostatic (blue), and total (red) relative velocities are shown separately. These results are based on the parameters Kn=10−2Kn = 10^{-2}, κ=0.6\kappa= 0.6, β=0.4\beta= 0.4, k1=k2=80k_1 = k_2 = 80 and Nv=0N_v = 0, for which (Ne)c≈70.1(N_e)_c \approx70.1.

The variation of (Ne)c(N_e)_c with charge ratio is shown in Figure 6 for four distinct values of κ\kappa. Across all examined cases, (Ne)c(N_e)_c exhibits a decreasing trend with increasing values of β\beta. Notably, the dependence of (Ne)c(N_e)_c on β\beta undergoes a significant alteration within the regime where near-field interactions are repulsive in the (κ,β)(\kappa,\beta) parameter space. In particular, the slope of the (Ne)c(N_e)_c curves reveals marked changes within this repulsive band, with its width increasing as κ\kappa increases. These results show that the threshold for collision suppression is not set by the charge magnitude alone but depends jointly on the particle size and charge ratios.

Figure 6. Critical electrostatic parameter (Ne)c(N_e)_c as a function of the charge ratio β\beta for different size ratios κ\kappa, with Kn=10−2Kn=10^{-2}, Nv=0N_v=0, and k1=k2=80k_1=k_2=80. For each κ\kappa, the gap between the two vertical dashed lines of the same colour indicates the near-field repulsive interval Δβ\Delta\beta, determined from the complete electrostatic interaction between finite-size dielectric spheres. The changes in slope of the (Ne)c(N_e)_c curves occur within these repulsive intervals. Although the repulsive intervals may appear narrower with increasing κ\kappa because β\beta is plotted on a logarithmic scale, their actual width Δβ\Delta\beta increases with κ\kappa.

These critical values of NeN_e provide a useful basis for assessing the importance of electrostatic collision suppression under atmospheric conditions. To discuss this in the context of water droplets in warm clouds, we consider a satellite droplet of radius a1=10a_1 = 10 μ\mum and γ˙=25\dot{\gamma} = 25 s−1^{-1}. We adopt two empirical relationships between droplet size and charge: qi∼ai2q_i \sim a_i^2 and qi∼ai1.3q_i \sim a_i^{1.3} (where i=1,2i = 1, 2), which lead to β=κ2\beta= \kappa^2 and β=κ1.3\beta= \kappa^{1.3}, respectively (see [45]). Utilizing these relationships, we determine (Ne)c(N_e)_c over a

Figure 7. (a) Critical electrostatic parameter, (Ne)c(N_e)_c, as a function of the size ratio κ\kappa for pairs of water droplets in air (λ0=0.1 μm\lambda_0=0.1\ \mu\mathrm{m}), with the charge ratio-size ratio relationships β=κ2\beta=\kappa^2 and β=κ1.3\beta=\kappa^{1.3} when a1=10 μma_1=10\ \mu\mathrm{m}, γ˙=25 s−1\dot{\gamma}=25\ \mathrm{s}^{-1} and Nv=0N_v=0. (b) Corresponding critical charge on the test droplet as a function of its radius, above which electrostatic repulsion suppresses collision.

range of size ratios κ\kappa. Figure 7a shows the resulting variation of (Ne)c(N_e)_c with κ\kappa for both β−κ\beta-\kappa relationships. Figure 7b recasts these critical values of NeN_e in terms of the charge on the test droplet, corresponding to the threshold for collision suppression. Based on our estimates, two equal-sized droplets, each with a radius of 10 μm10\ \mu\mathrm{m}, would require a charge of approximately 5×10−16 C5 \times10^{-16}\ \mathrm{C}, or 3.1×103e3.1 \times10^3 e, on each droplet for electrostatic repulsion to suppress their collision. This charge threshold is substantially larger than the representative average charge of a 10 μm10\ \mu\mathrm{m} cloud droplet (see [45]), suggesting that electrostatic repulsion is unlikely to suppress collisions between such weakly charged droplets under the conditions considered here. Volcanic ash particles exhibit a significantly broader range of electrostatic behaviors. For the representative particle sizes, charges and strain rates in volcanic plumes discussed in § 2.1, the resulting NeN_e can approach or exceed (Ne)c(N_e)_c for some like-charged particle pairs, although the present inertialess calculation applies quantitatively only to fine ash. Under sufficiently strong charging conditions, repulsive forces may inhibit particle collisions, while polarization-induced attractive forces can facilitate collisions in various regions of the size- and charge-ratio parameter space. These results highlight that (Ne)c(N_e)_c should not be interpreted as a universal charge threshold; rather, it is contingent upon the complete electrostatic force profile and its competition with the imposed flow and hydrodynamic interactions.

Trajectory topology and the critical threshold

Having identified the critical electrostatic strength required to suppress collisions, we now proceed to investigate how increasing NeN_e modifies the relative trajectories leading to this transition. Given that the relative velocity between the droplet pair remains unaffected by the azimuthal coordinate ϕ\phi, we show relative trajectories within a representative rsin⁡θ−rcos⁡θr\sin\theta-r\cos\theta plane. To illustrate typical pair trajectories, we will use the parameter values κ=0.5\kappa=0.5, β=10\beta=10, Kn=10−2Kn=10^{-2}, k1=k2=80k_1=k_2=80 and Nv=0N_v=0. For these parameters, the analysis from the previous subsection gives (Ne)c≈30.4(N_e)_c\approx30.4.

To establish a clear context, we first briefly review the pair trajectories when a uniaxial compressional flow alone (Ne=0N_e=0) drives the relative motion between two hydrodynamically interacting spherical particles. Figure 8(a) shows the relative trajectories for the Ne=0N_e=0 case. With non-continuum lubrication interactions, which permit particle contact in finite time, two distinct types of relative trajectories exist: open trajectories that approach from and return to the far field without reaching the collision sphere (blue lines), and colliding trajectories that approach from the far field and reach the collision sphere (red lines). These trajectories are fore-aft symmetric. Integrating Eq. (18) with Ne=Nv=0N_e=N_v=0, gives

Relative trajectories of two like-charged spheres in a uniaxial compressional flow

Figure 8. Relative trajectories of two like-charged spheres in a uniaxial compressional flow for Kn=10−2Kn = 10^{-2}, κ=0.5\kappa= 0.5, β=10\beta= 10, k1=k2=80k_1 = k_2 = 80 and Nv=0N_v = 0 when (a) Ne=0N_e = 0, (b) Ne=1N_e = 1, (c) Ne=25N_e = 25, (d) Ne=29N_e = 29, (e) Ne=40N_e = 40 and (f) Ne=∞N_e = \infty. For these parameters, (Ne)c≈30.4(N_e)_c \approx30.4. The sphere at the centre is the test sphere, and the thin black circle denotes the projection of the collision sphere. Blue, red and golden lines denote open, colliding and limiting colliding trajectories, respectively. For Ne>(Ne)cN_e > (N_e)_c, the black lines delineate the isolated shell-like region created by the electrostatic barrier, the pink lines represent near-field trajectories driven inward by polarization-induced attraction, and the green lines denote trajectories outside this shell that diverge towards the far field. Colliding trajectories exist for Ne<(Ne)cN_e < (N_e)_c, as shown in (a-d), whereas no trajectory originating from the far field reaches the collision sphere for Ne>(Ne)cN_e > (N_e)_c, as shown in (e-f). In (f), the background flow is absent, and the relative motion is driven solely by electrostatic interactions.

sin⁡2θ(r)cos⁡θ(r)=Cφ(r),(25)\sin^2\theta(r)\cos\theta(r)=C\varphi(r), \tag*{(25)}

where CC specifies an individual trajectory and

φ(r)=exp⁡[∫2r3(B(r′)−1)r′(1−A(r′)) dr′].(26)\varphi(r)=\exp\left[\int_{2}^{r}\frac{3(B(r')-1)}{r'(1-A(r'))}\,\mathrm{d}r'\right]. \tag*{(26)}

Open and colliding trajectories correspond to ∣C∣>2/(33)|C|>2/(3\sqrt{3}) and ∣C∣<2/(33)|C|<2/(3\sqrt{3}), respectively, while the limiting colliding trajectories (golden lines) satisfy ∣C∣=2/(33)|C|=2/(3\sqrt{3}). These limiting trajectories act as separatrices between the open and colliding trajectories and intersect the collision sphere at θ=θc\theta=\theta_c, π−θc\pi-\theta_c, π+θc\pi+\theta_c, and 2π−θc2\pi-\theta_c, where θc=cos⁡−1(1/3)\theta_c=\cos^{-1}(1/\sqrt{3}). Accordingly, the limiting trajectory in the first quadrant can be obtained by initiating the backward integration from the collision sphere at θ=θc\theta=\theta_c. These limiting colliding trajectories form the boundaries of the two upstream interception areas, which are circles whose centres lie on the compressional axis. The corresponding analytical expression for the collision efficiency is given by Wang et al. [58] as

E12=exp⁡[∫2∞3(B(r)−A(r))r(1−A(r)) dr].(27)E_{12} = \exp\left[\int_{2}^{\infty} \frac{3(B(r)-A(r))}{r(1-A(r))}\,\mathrm{d}r\right]. \tag*{(27)}

At finite NeN_e, electrostatic interactions modify the relative particle motion and, consequently, the topology of the pair trajectories and the locations at which the limiting colliding trajectories hit the collision sphere. Figure 8(b) shows the relative trajectories for Ne=1N_e = 1, which is well below (Ne)c(N_e)_c. Since the trajectories retain fore-aft symmetry at finite NeN_e, it is sufficient to analyse the relative trajectory equation numerically only over the interval 0≤θ≤π/20 \le\theta\le\pi/2. Unlike the Ne=0N_e = 0 case, the critical angle at which the limiting colliding trajectory hits the collision sphere is no longer prescribed by the purely hydrodynamic solution. We therefore determine θc\theta_c numerically by scanning the interval 0≤θ≤π/20 \le\theta\le\pi/2 on the collision sphere and integrating (18) backward in rr. For the parameters considered here, the limiting colliding trajectory originates at θ=π/2\theta= \pi/2. Consequently, colliding trajectories starting from the far field can hit the collision sphere at any point in the interval 0≤θ≤π/20 \le\theta\le\pi/2.

Furthermore, the upstream interception areas for Ne=1N_e = 1 are larger than those for Ne=0N_e = 0, indicating an enhancement of the collision rate by electrostatic interactions for this case.

The influence of electrostatic repulsion becomes increasingly pronounced as NeN_e approaches (Ne)c(N_e)_c. At Ne=25N_e = 25, the family of colliding trajectories forms a narrow neck region in the far field (see Figure 8(c)). Trajectories originating from this narrow neck region in the far field can reach the collision sphere upon forward integration by overcoming electrostatic repulsion at intermediate separations. An increase in NeN_e to 29 results in a further constriction of this neck region (see Figure 8(d)). This progressive narrowing of the neck region elucidates the mechanism for collision suppression within this trajectory-phase representation.

For Ne>(Ne)cN_e > (N_e)_c, the trajectory topology changes qualitatively. Figure 8(e) illustrates this regime for Ne=40N_e = 40. In this case, a pair approaching from the far field initially has vr<0v_r < 0 because the inward radial motion induced by the compressional flow exceeds the opposing contribution from electrostatic repulsion. As the separation decreases, the repulsive electrostatic contribution becomes increasingly important until the two contributions balance at a finite separation, where vr=0v_r = 0. A further decrease in separation results in vr>0v_r > 0, which prevents the approaching pair from continuing up to the collision sphere. Correspondingly, backward integration from the collision sphere fails to yield trajectories that can reach the far field. Instead, the neck formation observed for Ne<(Ne)cN_e < (N_e)_c closes, resulting in the emergence of an isolated shell-like region. The black lines in Figure 8(e) delineate this region, while the pink lines represent the dynamics associated with the near-field attractive electrostatic interaction. Trajectories shown by the green lines that originate outside this shell will diverge toward the far field upon forward integration. In conclusion, this case demonstrates that no trajectory originating from the far field can cross the repulsive barrier and reach the collision sphere, consistent with the collision suppression criterion Ne>(Ne)cN_e > (N_e)_c.

A dynamical system interpretation of Figure 8 can be derived by analyzing the motion of particle pairs as an autonomous two-dimensional system described by the following equations:

r˙=vr=−r(1−A)(3cos⁡2θ−1)+NeGfe,(28)\dot{r} = v_r = -r(1-A)(3\cos^2\theta- 1) + N_e G f_e, \tag*{(28)}
θ˙=vθr=3(1−B)sin⁡θcos⁡θ,(29)\dot{\theta} = \frac{v_\theta}{r} = 3(1-B)\sin\theta\cos\theta, \tag*{(29)}

for Nv=0N_v = 0. To find the equilibrium points, we require that vr=0v_r = 0 and vθ=0v_\theta= 0. The latter condition restricts the possible equilibrium points to the compressional axis, θ=0,π\theta= 0,\pi, and the equatorial plane, θ=π/2\theta= \pi/2.

On the equatorial plane, the equilibrium condition becomes

r(1−A)+NeGfe=0.(30)r(1-A) + N_e G f_e = 0. \tag*{(30)}

For the parameter combination considered in Figure 8, the electrostatic interaction is attractive in the near field and repulsive in the far field. This near-field attraction can balance the outward flow-induced radial motion in the equatorial plane, resulting in an equatorial equilibrium at r=rsr = r_s. Linearization about this equilibrium yields two eigenvalues:

λr=∂∂r[r(1−A)+NeGfe]∣r=rs,λθ=∂θ˙∂θ∣θ=π/2=−3(1−B(rs)).(31)\lambda_r = \left.\frac{\partial}{\partial r}\left[r(1-A) + N_e G f_e\right]\right|_{r=r_s}, \qquad\lambda_\theta= \left.\frac{\partial\dot{\theta}}{\partial\theta}\right|_{\theta=\pi/2} = -3(1-B(r_s)). \tag*{(31)}

For the equilibrium shown in the present phase portraits, λr>0\lambda_r > 0 while λθ<0\lambda_\theta< 0, identifying it as a saddle point: radial perturbations are unstable because trajectories just inside the equilibrium are driven towards contact, whereas those just outside it move outward; conversely, angular perturbations are restoring toward θ=π/2\theta= \pi/2. The portion of the limiting colliding trajectory can be interpreted as a branch of the stable manifold of this saddle point, which separates colliding from non-colliding initial conditions. The inward part of this separatrix forms part of the unstable manifold leading towards the collision sphere.

At Ne=1N_e = 1, this equatorial saddle point lies relatively close to the collision sphere. As NeN_e increases to 25 and 29, this equatorial saddle point moves away from the collision sphere, and its stable manifold becomes increasingly deformed by the intermediate-range electrostatic repulsion. This results in a progressively narrower neck observable along the compressional direction. For Ne=25N_e = 25 and 29, both of which satisfy Ne<(Ne)c≈30.4N_e < (N_e)_c \approx30.4, no equilibrium exists on the compressional axis. Nonetheless, the radial velocity along this axis diminishes considerably over an intermediate range of separations as NeN_e approaches (Ne)c(N_e)_c, giving rise to a dynamical bottleneck through which colliding trajectories can still pass.

At Ne=(Ne)cN_e = (N_e)_c, the equilibrium condition along the compressional axis is governed by the equation

−2r(1−A)+NeGfe=0.(32)-2r(1-A) + N_e G f_e = 0. \tag*{(32)}

This equation first yields a double root at r=rcr = r_c (see the inset of Figure 4), satisfying simultaneously

vr(rc,0)=0and∂vr∂r∣r=rc=0,(33)v_r(r_c,0) = 0 \quad\text{and} \quad\left.\frac{\partial v_r}{\partial r}\right|_{r=r_c} = 0, \tag*{(33)}

which is precisely the criticality condition used to define (Ne)c(N_e)_c.

For Ne=40>(Ne)cN_e = 40 > (N_e)_c, the degenerate equilibrium splits into two axial fixed points on each compressional half-axis. Linearizing the system about an axial fixed point r=rar = r_a, θ=0\theta= 0, gives

λr=∂∂r[−2r(1−A)+NeGfe]∣r=ra,λθ=∂θ˙∂θ∣θ=0=3(1−B(ra)),(34)\lambda_r = \left.\frac{\partial}{\partial r}\left[-2r(1-A) + N_e G f_e\right]\right|_{r=r_a}, \quad\lambda_\theta= \left.\frac{\partial\dot{\theta}}{\partial\theta}\right|_{\theta=0} = 3\left(1-B(r_a)\right), \tag*{(34)}

where λθ\lambda_\theta is positive because the background flow drives particle pairs away from the compressional axis. At the outer fixed point (ra≈6r_a \approx6), λr<0\lambda_r < 0, so the fixed point is a saddle, wherein radial perturbations are restorative while transverse perturbations are unstable. Conversely, at the inner fixed point (ra≈4.2r_a \approx4.2), λr>0\lambda_r > 0, so it is an unstable node.

The axial fixed points occurring for Ne>(Ne)cN_e > (N_e)_c notably do not eliminate the pre-existing equatorial saddle points. For the parameters considered here, the equatorial saddle point persists at Ne=40N_e = 40 and therefore coexists with the newly created axial fixed points. Consequently, the transition across (Ne)c(N_e)_c modifies the structure of equilibria specifically along the compressional axis, while the equatorial saddle points continue to emerge from the interplay between the outward flow-induced motion and the attraction induced by near-field polarization. In the full axisymmetric configuration, these equatorial saddle points correspond to a ring of equivalent equilibria, whose intersections with the meridional plane appear as symmetric saddle points. It is important to note, however, that the existence of these points is not universal; it requires sufficiently strong near-field electrostatic attraction to balance the outward motion in the equatorial plane.

The invariant manifolds of the axial fixed points reorganize the global trajectory topology and prevent trajectories arriving from the far field from penetrating into the near-field attractive region. In particular, the change in global topology is governed by the stable manifold of the equatorial saddle, which constitutes the limiting colliding trajectory and therefore separates colliding from non-colliding initial conditions. For Ne<(Ne)cN_e < (N_e)_c, this manifold extends to the far field, thereby allowing a finite set of far-field trajectories to reach the collision sphere. In contrast, for Ne>(Ne)cN_e > (N_e)_c, the manifold originates at the inner axial node rather than in the far field. Consequently, the near-contact attractive region becomes dynamically separated from particles approaching from infinity, and trajectories leading to contact are confined to the bounded shell-like region, as shown in Figure 8(e). Thus, the transition from Ne=29N_e = 29 to Ne=40N_e = 40 can be interpreted as a saddle-node bifurcation on the compressional axis. Just below criticality, there is a narrow dynamical bottleneck, whereas just above criticality, newly created saddle points and their separatrices disconnect the near-contact collision region from the far field, thereby causing the collision efficiency to vanish.

The trajectories shown in Figure 8(b-Figure 8e) correspond to size and charge ratios for which the electrostatic interaction becomes attractive in the near field. When the electrostatic interaction remains repulsive at all separations, the near-field attractive region and the associated equatorial saddle points are absent. Nevertheless, the axial mechanism of collision suppression remains qualitatively similar: increasing NeN_e progressively reduces the upstream interception area, and, for Ne>(Ne)cN_e > (N_e)_c, the electrostatic barrier prevents trajectories originating in the far field from reaching the collision sphere. Therefore, we expect the collision efficiency to decrease monotonically with increasing NeN_e and vanish once the imposed compressional flow can no longer overcome the repulsive barrier.

The same dynamical system framework remains applicable for finite NvN_v, because the van der Waals interactions modify only the radial relative velocity equation. Consequently, the equilibrium locations remain confined to the compressional axis and the equatorial plane, although the additional attractive interaction modifies their radial positions. The van der Waals attraction leads to an increase in (Ne)c(N_e)_c, necessitating a stronger electrostatic repulsion to counteract both the imposed compressional motion and the attractive van der Waals contribution. Similarly, the equatorial fixed points shift as NvN_v varies. Thus, while the same dynamical system analysis is applicable for finite NvN_v, the locations and stability of the fixed points, the bifurcation threshold, and, in cases of sufficiently strong van der Waals attraction, the detailed phase-space topology may differ from the case where Nv=0N_v = 0.

The limiting case Ne=∞N_e = \infty, shown in Figure 8(f), represents a situation wherein particle motion is governed solely by electrostatic interactions, with no influence from the background flow. In this scenario, the dimensionless radial relative velocity can be expressed as:

vr(e)=(1+κκ)Gfe.(35)v_r^{(e)} = \left(\frac{1+\kappa}{\kappa}\right)Gf_e. \tag*{(35)}

In the above equation, the relative velocity is scaled by q12/(4πϵ0a12)/[6πμfa1]q_1^2/(4\pi\epsilon_0a_1^2)/[6\pi\mu f a_1], which is the characteristic relative velocity associated with purely electrostatically driven relative motion.

Since the electrostatic force acts purely in the radial direction, the relative trajectories do not depend on the angular coordinates θ\theta and ϕ\phi. In Figure 8(f), the thick black circle delineates the boundary at which fe=0f_e = 0. Inside this circle, polarization-induced attraction leads to radially inward trajectories, whereas outside it, the electrostatic interactions are repulsive, directing trajectories radially outward. Consequently, particles initially located outside this critical separation cannot reach the collision sphere in the absence of an imposed flow. This limiting case highlights the role of the compressional flow at finite NeN_e, especially for Ne<(Ne)cN_e < (N_e)_c, where the flow facilitates particle pairs to traverse the intermediate repulsive region and enter the near-field attractive regime, where collision can occur.

Non-monotonic variation of the collision efficiency

The preceding analysis demonstrates that changes in the topology of the relative trajectories directly modify the upstream interception area and, consequently, the collision rate. We therefore examine the dependence of E12E_{12} on the key parameters κ\kappa, β\beta, NeN_e, KnKn, and NvN_v, which collectively govern the collision dynamics. In the limit Ne→0N_e \to0, electrostatic interactions become negligible, and E12E_{12} approaches a value determined by the combined effects of the imposed compressional flow, non-continuum lubrication interactions, and van der Waals forces. In this flow-dominated regime, the collision efficiency increases with increasing κ\kappa. For small κ\kappa, the large disparity in particle sizes causes the smaller particle to follow the imposed flow streamlines more closely and to move around the larger particle rather than collide with it. Consequently, collisions at small κ\kappa are restricted to particles approaching along streamlines that pass sufficiently close to the larger particle. As κ\kappa increases, this restriction becomes less pronounced, resulting in a larger upstream interception area and, hence, a higher collision efficiency. As NeN_e increases, electrostatic interactions play an increasingly prominent role, thereby modifying the collision dynamics in ways that can either enhance or attenuate E12E_{12}, contingent upon the values of κ\kappa and β\beta. Nevertheless, for sufficiently strong electrostatic interactions, repulsion ultimately prevents particles from reaching the collision sphere, and E12E_{12} decreases to zero as NeN_e approaches (Ne)c(N_e)_c.

Figure 9(a-Figure 9c) shows the variation of E12E_{12} with NeN_e for κ=0.3\kappa= 0.3, 0.50.5, 0.70.7, and 1.01.0, with β=10−1\beta= 10^{-1}, 11, and 1010 in panels (a-c), respectively, for Kn=10−2Kn = 10^{-2} and Nv=10−3N_v = 10^{-3}. For the small charge ratio β=10−1\beta= 10^{-1} [Figure 9(a)], the near-field electrostatic attraction becomes stronger with increasing κ\kappa for κ≳0.32\kappa\gtrsim0.32. Consequently, the impact of near-field electrostatic attraction on E12E_{12} becomes increasingly significant as the size ratio extends beyond this critical value. For κ=0.3\kappa= 0.3, electrostatic interactions remain repulsive at all separations, causing E12E_{12} to decrease monotonically from its Ne=0N_e = 0 value as NeN_e increases. However, for larger size ratios κ=0.5\kappa= 0.5, 0.70.7, and 1.01.0, the presence of strong near-field attraction results in a non-monotonic variation of E12E_{12}. In these cases, the collision efficiency exhibits an initial increase above its value at Ne=0N_e = 0, followed by a subsequent decline as the electrostatic repulsive barrier becomes sufficiently strong. Ultimately, the collision efficiency vanishes as NeN_e approaches the corresponding critical value (Ne)c(N_e)_c. It is noteworthy that for β=10−1\beta= 10^{-1} and κ\kappa values ranging from 0.30.3 to 11, the electrostatic effects favour the variation of collision efficiency with κ\kappa in the Ne→0N_e \to0 limit. As a consequence, the curves representing different κ\kappa values in Figure 9(a) do not intersect with one another.

Figure 9. (a-c) Collision efficiencies as a function of NeN_e for Kn=10−2Kn = 10^{-2}, κ=0.3,0.5,0.7,1.0\kappa= 0.3, 0.5, 0.7, 1.0, NL=500N_L = 500, Nv=10−3N_v = 10^{-3} and k1=k2=80k_1 = k_2 = 80 when (a) β=10−1\beta= 10^{-1}, (b) β=1\beta= 1, and (c) β=10\beta= 10. (d) Collision efficiency as a function of NeN_e for Kn=10−2Kn = 10^{-2}, k1=k2=80k_1 = k_2 = 80, β=1\beta= 1, and κ=0.4\kappa= 0.4 and 1.01.0 in the absence of van der Waals interactions (Nv=0N_v = 0). The dashed lines in (d) denote the corresponding asymptotic predictions from Eq. (42).

A qualitatively different dependence on κ\kappa emerges for β=1\beta= 1 [Figure 9(b)]. In the flow-dominated regime, the collision efficiency increases with increasing κ\kappa. As NeN_e increases, however, this trend reverses because the near-field electrostatic attraction becomes progressively weaker with increasing size ratio when β=1\beta= 1. Notably, for κ=1\kappa= 1, the near-field electrostatic interaction remains repulsive. Consequently, the curves corresponding to different size ratios intersect at intermediate values of NeN_e, reflecting the contrasting dependence of the flow-driven and electrostatically modified collision dynamics on κ\kappa. Consistent with the presence of near-field electrostatic attraction, the collision efficiencies for κ=0.3\kappa= 0.3, 0.50.5, and 0.70.7 exceed their respective Ne=0N_e = 0 values over an intermediate range of NeN_e before decreasing sharply as NeN_e approaches the corresponding (Ne)c(N_e)_c. In contrast, no such enhancement occurs for κ=1\kappa= 1, owing to the repulsive nature of the electrostatic interaction.

For β=10\beta= 10 [Figure 9(c)], the variation of E12E_{12} with NeN_e for different values of κ\kappa is qualitatively similar to that observed for β=1\beta= 1. However, the underlying electrostatic interactions that drive this behaviour differ in two important respects. First, the near-field electrostatic interaction is attractive for all values of κ\kappa considered. Second, the magnitude of this attraction is substantially greater than that for β=1\beta= 1. These characteristics result in a more pronounced electrostatic enhancement of particle collisions, resulting in E12E_{12} exceeding its corresponding value at Ne=0N_e = 0 over an intermediate range of NeN_e for all values of κ\kappa. Notably, for smaller values of κ\kappa, this enhancement can be as much as an order of magnitude relative to the corresponding collision efficiency at Ne=0N_e = 0.

Weak electrostatic interactions

Figure 9(a-c) shows that weak charges can either raise or lower the collision efficiency, depending on κ\kappa and β\beta. To determine which separations control the sign and magnitude of this initial change, we treat the electrostatic contribution as a regular perturbation of the purely hydrodynamic relative trajectories for Ne≪1N_e \ll1. We set Nv=0N_v = 0, so that contact is enabled by non-continuum lubrication alone, and write

θ(r)=θ0(r)+Neθ1(r)+O(Ne2),(36)\theta(r) = \theta_0(r) + N_e\theta_1(r) + O(N_e^2), \tag*{(36)}

where θ0\theta_0 denotes the leading-order trajectory in the absence of electrostatic interactions and θ1\theta_1 represents the first-order electrostatic correction.

The limiting colliding trajectory satisfies the condition vr=0v_r = 0 at the collision surface. At leading order, this gives

θ0(2)=cos⁡−1(13),(37)\theta_0(2) = \cos^{-1}\left(\frac{1}{\sqrt{3}}\right), \tag*{(37)}

whereas the first-order condition yields

θ1(2)=−Gfe6r(1−A)sin⁡θ0cos⁡θ0∣r→2+≡C1.(38)\theta_1(2) = -\left.\frac{G f_e}{6r(1-A)\sin\theta_0\cos\theta_0}\right|_{r\to2^+} \equiv C_1. \tag*{(38)}

The limit r→2+r \to2^+ is understood because both 1−A1-A and GG vanish at exact contact, while their ratio remains finite. At leading order, the limiting colliding trajectory is the purely hydrodynamic trajectory (25) with C=2/(33)C = 2/(3\sqrt{3}). Defining I(r)=2φ(r)/(33)I(r) = 2\varphi(r)/(3\sqrt{3}), it satisfies

sin⁡2θ0(r)cos⁡θ0(r)=I(r),(39)\sin^2\theta_0(r)\cos\theta_0(r) = I(r), \tag*{(39)}

and selecting the root corresponding to the limiting colliding trajectory yields

θ0(r)=cos⁡−1{23cos⁡[13cos⁡−1(−332I(r))]}.(40)\theta_0(r)=\cos^{-1}\left\{\frac{2}{\sqrt{3}}\cos\left[\frac{1}{3}\cos^{-1}\left(-\frac{3\sqrt{3}}{2}I(r)\right)\right]\right\}. \tag*{(40)}

This solution satisfies θ0(2)=cos⁡−1(1/3)\theta_0(2)=\cos^{-1}(1/\sqrt{3}) and approaches θ0→0\theta_0 \to0 as r→∞r \to\infty, as required for the upstream limiting trajectory.

The first-order correction to the collision efficiency can be written as a single quadrature along this trajectory. Let s(θ)=sin⁡2θcos⁡θs(\theta)=\sin^2\theta\cos\theta, so that C=s(θ)/φ(r)C=s(\theta)/\varphi(r) is the constant of (25) when Ne=0N_e=0. For Nv=0N_v=0, the trajectory equation (18) gives, exactly,

ddrln⁡[s(θ)φ(r)]=3(1−B)r(1−A)NeGfevr,(41)\frac{\mathrm{d}}{\mathrm{d}r}\ln\left[\frac{s(\theta)}{\varphi(r)}\right]=\frac{3(1-B)}{r(1-A)}\frac{N_eGf_e}{v_r}, \tag*{(41)}

so that CC changes along a trajectory only through the electrostatic contribution to vrv_r. The collision efficiency is proportional to the far-field value of CC on the limiting colliding trajectory. The shift (38) of its contact point changes the value of CC at contact only at O(Ne2)O(N_e^2), because s(θ)s(\theta) is stationary at θ0(2)=cos⁡−1(1/3)\theta_0(2)=\cos^{-1}(1/\sqrt{3}). Integrating (41) along the leading-order trajectory (40), on which vr=−r(1−A)(3cos⁡2θ0−1)v_r=-r(1-A)(3\cos^2\theta_0-1), gives

E12=E12(0)[1−Ne∫2∞W(r)G(r)fe(r) dr]+O(Ne2),(42)E_{12}=E_{12}^{(0)}\left[1-N_e\int_{2}^{\infty}W(r)G(r)f_e(r)\,\mathrm{d}r\right]+O(N_e^2), \tag*{(42)}

where E12(0)E_{12}^{(0)} is the collision efficiency of uncharged, hydrodynamically interacting spheres given by (27) and

W(r)=3(1−B)r2(1−A)2(3cos⁡2θ0−1).(43)W(r)=\frac{3(1-B)}{r^2(1-A)^2(3\cos^2\theta_0-1)}. \tag*{(43)}

The weight WW is positive, so that electrostatic repulsion lowers the collision efficiency at first order and attraction raises it. Although W(r)W(r) becomes singular as r→2+r\to2^+, this singularity is integrable. In this limit, G/(1−A)G/(1-A) remains finite and 3cos⁡2θ0−1=O(J1/2)3\cos^2\theta_0-1=O(J^{1/2}), where J(r)=∫2rdr′/(1−A(r′))J(r)=\int_{2}^{r}\mathrm{d}r'/(1-A(r')). Owing to the weaker divergence of the non-continuum lubrication resistance, J(r)J(r) converges, consistent with finite-time contact. Consequently, the integrand in (42) behaves as O(J−1/2 dJ)O(J^{-1/2}\,\mathrm{d}J), whose integral is finite. The leading correction is thus a weighted average of the electrostatic force along the limiting trajectory, with the weight concentrated in the lubrication region; this is why near-contact attraction can raise E12E_{12} even when the interaction is repulsive at larger separations. The expansion requires the electrostatic contribution to vrv_r at contact to remain small compared with the straining contribution, Ne∣Gfe/[r(1−A)]∣r→2+≪1N_e\left|Gf_e/[r(1-A)]\right|_{r\to2^+}\ll1, rather than merely Ne≪1N_e\ll1. When the near-contact force is attractive and this ratio reaches unity, the inflow region on the collision sphere extends to the equator and the equatorial saddle point discussed in § 4.3 emerges from the collision sphere; beyond this point, the limiting colliding trajectory is the stable manifold of that saddle point, which a regular expansion about the purely hydrodynamic trajectory cannot describe.

Figure 9(d) compares the numerically calculated collision efficiencies with the asymptotic prediction (42) for Kn=10−2Kn=10^{-2}, β=1\beta=1, Nv=0N_v=0, and κ=0.4\kappa=0.4 and 1.01.0. In the limit Ne≪1N_e\ll1, the asymptotic results closely match the numerical solutions for both size ratios, including the increase of E12E_{12} with NeN_e for κ=0.4\kappa=0.4, for which the near-contact interaction is attractive, and its decrease for κ=1.0\kappa=1.0, for which it is repulsive. The asymptotic and numerical results separate at Ne≈10−1N_e\approx10^{-1} for κ=0.4\kappa=0.4 but only at Ne≈1N_e\approx1 for κ=1.0\kappa=1.0, consistent with the condition stated above; a full trajectory calculation is therefore required for moderate and large values of NeN_e.

Effects of non-continuum lubrication, van der Waals attraction and charge ratio

We now consider how non-continuum lubrication effects, measured by the Knudsen number KnKn, modify the collision efficiency in the presence of van der Waals and electrostatic interactions. Figure 10(a) shows the variation of E12E_{12} with KnKn for κ=0.5\kappa= 0.5, Ne=1N_e = 1, NL=250N_L = 250, and Nv=10−3N_v = 10^{-3}. For comparison, we also show results without electrostatic interactions. With non-continuum (NC) hydrodynamic interactions alone, E12E_{12}, obtained by evaluating the integral in (27), decreases monotonically with decreasing KnKn and approaches zero as Kn→0Kn \to0, since the relative thickness of the non-continuum lubrication region progressively diminishes. The inclusion of van der Waals (vdW) attraction changes this asymptotic behaviour: for the NC+vdW case, E12E_{12} approaches a finite value as Kn→0Kn \to0, corresponding to collisions driven by the imposed compressional flow and vdW attraction in the presence of continuum hydrodynamic interactions. Accounting for electrostatic interactions arising from particle surface charges further enhances E12E_{12} when polarization gives rise to an attractive near-field interaction. For β=1\beta= 1, the dependence of E12E_{12} on KnKn remains qualitatively similar to that for NC+vdW. However, for β=10\beta= 10, the substantially stronger near-field electrostatic attraction dominates the collision dynamics, rendering E12E_{12} nearly independent of KnKn over the range considered.

Figure 10. (a) Collision efficiency, E12E_{12}, as a function of KnKn for non-continuum hydrodynamic interactions alone (NC only; solid black line), non-continuum hydrodynamic and van der Waals interactions (NC+vdW; solid red line), and non-continuum hydrodynamic, van der Waals, and electrostatic interactions (NC+vdW+EF) for β=1\beta= 1 (dotted blue line) and β=10\beta= 10 (dash-dotted blue line), with κ=0.5\kappa= 0.5, Ne=1N_e = 1, NL=250N_L = 250, and Nv=10−3N_v = 10^{-3}. (b) E12E_{12} as a function of NvN_v for continuum hydrodynamic and van der Waals interactions (Continuum+vdW; solid black line), NC+vdW (solid red line), and NC+vdW+EF for β=1\beta= 1 (dotted blue line) and β=10\beta= 10 (dash-dotted blue line), with κ=0.5\kappa= 0.5, Kn=10−2Kn = 10^{-2}, Ne=1N_e = 1, and NL=250N_L = 250. (c) E12E_{12} as a function of the charge ratio β\beta for Ne=10−2N_e = 10^{-2}, 10−110^{-1}, 1, and 10, with κ=0.5\kappa= 0.5, Kn=10−2Kn = 10^{-2}, NL=250N_L = 250, and Nv=10−3N_v = 10^{-3}. (d) E12E_{12} as a function of the size ratio κ\kappa for water droplets in air with a1=10a_1 = 10 µm and q1=200eq_1 = 200e, for β=5\beta= 5 and 10, with NC+vdW+EF. For comparison, the collision efficiencies of uncharged droplets in a uniaxial compressional flow reported by Dhanasekaran et al. [14] are also shown.

The sensitivity of the collision efficiency to van der Waals interactions is quantified by varying NvN_{\mathrm{v}} from 10−710^{-7} to 10−210^{-2}. Figure 10(b) shows the variation of E12E_{12} with NvN_{\mathrm{v}} for κ=0.5\kappa= 0.5, Ne=1N_{\mathrm{e}} = 1, and NL=250N_{\mathrm{L}} = 250. In the absence of electrostatic interactions, decreasing NvN_{\mathrm{v}} reduces E12E_{12}, with the reduction being particularly pronounced for continuum hydrodynamics. For non-continuum lubrication interactions (Kn=10−2Kn = 10^{-2}), the decrease is more gradual, and E12E_{12} approaches the value obtained in the absence of van der Waals interactions. Accounting for electrostatic interactions enhances E12E_{12} when the near-field interaction is attractive, as for β=1\beta= 1 and 10. For β=10\beta= 10, the near-field electrostatic attraction is sufficiently strong that E12E_{12} becomes nearly insensitive to variations in the strength of the van der Waals interaction, implying that electrostatic forces dominate over other collision-inducing mechanisms.

To examine the influence of the relative particle charges on the collision dynamics, in Figure 10(c), we present E12E_{12} as a function of the charge ratio β\beta for four different values of NeN_{\mathrm{e}}. For weak electrostatic interactions (Ne=10−2N_{\mathrm{e}} = 10^{-2}), E12E_{12} depends only weakly on β\beta, indicating predominantly flow-driven collision dynamics. As NeN_{\mathrm{e}} increases, the dependence on β\beta becomes progressively stronger. For Ne=1N_{\mathrm{e}} = 1 and 10, E12E_{12} initially decreases with increasing β\beta, reflecting the suppression of collisions by electrostatic repulsion, before increasing as polarization-induced attraction becomes increasingly important. This competition is particularly pronounced for Ne=10N_{\mathrm{e}} = 10, for which E12E_{12} exhibits a distinct minimum at small β\beta. At larger β\beta, the rapid increase in E12E_{12} demonstrates the increasing influence of near-field electrostatic attraction on the collision dynamics. For any prescribed β\beta, the dependence of E12E_{12} on NeN_{\mathrm{e}} is consistent with the trends discussed earlier.

Finally, we estimate the collision efficiency of like-charged droplet pairs with unequal sizes suspended in air. Figure 10(d) shows E12E_{12} as a function of κ\kappa for water droplets in air with a1=10 μma_1 = 10\ \mu\mathrm{m}, q1=200eq_1 = 200e, γ˙=25 s−1\dot{\gamma} = 25\ \mathrm{s}^{-1} and λ0=0.1 μm\lambda_0 = 0.1\ \mu\mathrm{m}. For these conditions, the relevant dimensionless parameters vary with size ratio as Kn=0.02/(1+κ)Kn = 0.02/(1+\kappa), Ne=0.22/κN_{\mathrm{e}} = 0.22/\kappa, NL=200π(1+κ)N_{\mathrm{L}} = 200\pi(1+\kappa), and Nv=1.9×10−2/[κ(1+κ)]N_{\mathrm{v}} = 1.9 \times10^{-2}/[\kappa(1+\kappa)]. The charge ratios β=5\beta= 5 and 10, for which the smaller droplet carries the larger charge, depart from the empirical relations β=κ1.3\beta= \kappa^{1.3} and β=κ2\beta= \kappa^2 for natural cloud droplets used in § 4.2, and are more representative of artificially charged droplets ([28]). In the absence of electrostatic interactions, E12E_{12} increases with κ\kappa, consistent with the findings of Dhanasekaran et al. [14]. Electrostatic interactions qualitatively reverse the dependence of E12E_{12} on κ\kappa observed for uncharged droplets: for both β=5\beta= 5 and 10, the collision efficiency decreases monotonically with increasing κ\kappa. This behaviour arises from the stronger polarization-induced attraction experienced by droplet pairs with greater size disparity at these charge ratios, which facilitates particle approach and thereby counteracts the comparatively low collision efficiency associated with hydrodynamic interactions alone at small κ\kappa. Furthermore, E12E_{12} is systematically larger for β=10\beta= 10 than for β=5\beta= 5 over the range of κ\kappa considered, reflecting that the near-field electrostatic attraction force becomes increasingly pronounced with increasing charge ratio. These results demonstrate that, under the conditions considered, electrostatic interactions can substantially increase the collision efficiency of like-charged droplet pairs with large disparities in both size and charge compared with their uncharged counterparts.

Conclusions

We have investigated the collision dynamics of like-charged dielectric particles subjected to a uniaxial compressional flow, accounting for the coupled effects of hydrodynamic and electrostatic interactions. The formulation incorporates non-continuum lubrication effects, retarded van der Waals attraction, and the complete electrostatic interaction between finite-sized dielectric spheres. This approach effectively addresses the mutual polarization effects, which become increasingly important as the particles approach one another. Our analysis of charged tracers elucidated the competing influences of Coulomb repulsion and polarization-induced attraction on collision dynamics. For finite-sized, hydrodynamically interacting particles, we identified a critical value of the dimensionless electrostatic parameter, (Ne)c(N_e)_c, which signifies the transition from collision to collision suppression. When the parameter exceeds this threshold, the repulsive electrostatic barrier cannot be overcome by the flow-induced inward relative motion, thereby precluding collisions between particles approaching from the far field. The value of (Ne)c(N_e)_c depends strongly on the particle size ratio κ\kappa and charge ratio β\beta, reflecting the sensitivity of the finite-dielectric electrostatic interaction to particle size and charge asymmetry.

The representative pair trajectories shown in Figure 8 reveal how changes in trajectory topology govern the transition from collision to collision suppression. For Ne<(Ne)cN_e < (N_e)_c, the inward relative motion induced by the compressional flow is sufficient to overcome the intermediate electrostatic repulsive barrier, allowing the particles to reach the near-field region where, when present, polarization-induced attraction further promotes their approach to contact. As NeN_e approaches (Ne)c(N_e)_c, the upstream interception area associated with colliding trajectories progressively contracts and the collision efficiency decreases. For Ne>(Ne)cN_e > (N_e)_c, trajectories originating in the far field can no longer reach the collision surface, and the collision efficiency vanishes. We have shown that electrostatic interactions can either enhance or inhibit collisions depending on κ\kappa, β\beta, and NeN_e. When polarization-induced attraction becomes sufficiently strong, it can substantially enhance the collision efficiency and diminish the influence of non-continuum lubrication and van der Waals interactions on the collision dynamics. In contrast, sufficiently strong electrostatic repulsion prevents particles from coming into contact, causing collision efficiency to vanish once NeN_e exceeds (Ne)c(N_e)_c.

Two analytical results support this picture. The threshold (Ne)c(N_e)_c corresponds to a saddle-node bifurcation on the compressional axis, along which the inward straining is strongest. When the near-contact interaction is attractive, the limiting colliding trajectory is the stable manifold of an equatorial saddle point; above (Ne)c(N_e)_c this manifold originates at the newly created axial node rather than in the far field, so that the trajectories reaching contact occupy a bounded region disconnected from infinity and the collision efficiency vanishes. For weak electrostatic interactions, a regular perturbation of the limiting colliding trajectory expresses the O(Ne)O(N_e) change in the collision efficiency as a weighted average of the electrostatic force along that trajectory, with the weight concentrated in the lubrication region, so that the near-contact interaction weighs most heavily in deciding whether weak charges enhance or suppress collisions. This prediction agrees closely with the full trajectory calculations.

As discussed in § 1, the present study is relevant to electrically active particulate flows encountered in both atmospheric and geophysical settings. In particular, electrostatic interactions may influence droplet collisions in strongly electrified clouds, and particle aggregation in the ‘dirty thunderstorms’ associated with explosive volcanic eruptions, where ash particles can acquire substantial electrical charge. The present results establish the underlying particle scale physics by showing how the competition between hydrodynamic and electrostatic forces determines whether charged particles collide. Although these processes occur within complex turbulent environments, resolving their interactions at the particle level is a necessary step towards constructing physically based collision kernels for charged particles in turbulence.

The present trajectory-based framework can be extended to turbulent flows by approximating the velocity gradient experienced by a sub-Kolmogorov particle pair as a linearly varying stochastic field. This approximation will allow us to integrate the equations governing the relative motion of finite-size charged particles over many statistically independent realizations of the velocity-gradient tensor. Following the Monte Carlo integration approach developed by Dhanasekaran et al. [13], the turbulent collision rate can be obtained by ensemble averaging the collision flux over these realizations. This extension would therefore quantify how temporal fluctuations in the local strain and rotation rate tensors modify the collision efficiency calculated for the steady compressional flow.

Future work should additionally incorporate particle inertia and gravitational settling, both of which can alter relative velocities and, in turbulent flows, the spatial distribution of particle pairs. In particular, particle inertia can lead to preferential concentration and caustics, which cannot be captured by the deterministic linear-flow formulation considered here. An approach combining these effects with finite-size hydrodynamic and electrostatic interactions would enable the relative importance of turbulence, gravity, and electrostatic force to be quantified systematically across relevant particle sizes and charges. The resulting collision kernels could subsequently be used in population-balance or particle-based models to examine the consequences of electrostatic forces on droplet growth and precipitation initiation in thunderstorms, and on ash aggregation and fallout in volcanic ‘dirty thunderstorms’.

Funding. P.P. would like to acknowledge financial support from the Prime Minister’s Research Fellows (PMRF) scheme, Ministry of Education, Government of India (Project Number: SB22230184AMPMRF008746). A.R. acknowledges the financial support received from the Anusandhan National Research Foundation (ANRF), Government of India, through the Advanced Research Grant (ANRF/ARG/2025/010666/ENS).

Declaration of interests. The authors report no conflict of interest.

Data availability statement. The data and code supporting the findings of this study are available from the corresponding author upon reasonable request.

References

  1. [2]Balthasar, M., Mauss, F., Knobel, A. & Kröner, D. 2002 Detailed modeling of soot formation in a partially stirred plug flow reactor. Combustion and Flame 128 (4), 395–409.
  2. [3]Batchelor, G. K. & Green, J. T. 1972 The hydrodynamic interaction of two small freely-moving spheres in a linear flow field. Journal of Fluid Mechanics 56 (2), 375–400.DOI
  3. [4]Beard, K. V., Ochs III, H. T. & Twomey, C. H. 2004 Aircraft measurements of high average charges on cloud drops in layer clouds. Geophysical Research Letters 31 (14).DOI
  4. [5]Behnke, S. A., Edens, H. E., Thomas, R. J. & McNutt, S. R. 2018 Impulsive volcanic plumes generate volcanic lightning and vent discharges: A statistical analysis of Sakurajima volcano. Journal of Geophysical Research: Atmospheres 123 (8), 4131–4156.
  5. [6]Bichoutskaia, E., Boatwright, A. L., Khaçhatourian, A. V. & Stace, A. J. 2010 Electrostatic analysis of the interactions between charged particles of dielectric materials. The Journal of Chemical Physics 133 (2), 024105.DOI
  6. [7]Brunk, B. K., Koch, D. L. & Lion, L. W. 1998 Turbulent coagulation of colloidal particles. Journal of Fluid Mechanics 364, 81–113.DOI
  7. [8]Chun, J. & Koch, D. L. 2005 Coagulation of monodisperse aerosol particles by isotropic turbulence. Physics of Fluids 17 (2), 027102.DOI
  8. [11]Davis, R. H., Schonberg, J. A. & Rallison, J. M. 1989 The lubrication force between two viscous drops. Physics of Fluids 1 (1), 77–81.DOI
  9. [12]Dhanasekaran, J. & Koch, D. L. 2025 The effect of turbulence, gravity, and noncontinuum hydrodynamic interactions on the drop size distribution in clouds. Journal of the Atmospheric Sciences 82 (8), 1647–1659.
  10. [13]Dhanasekaran, J., Roy, A. & Koch, D. L. 2021a Collision rate of bidisperse, hydrodynamically interacting spheres settling in a turbulent flow. Journal of Fluid Mechanics 912, A5.DOI
  11. [14]Dhanasekaran, J., Roy, A. & Koch, D. L. 2021b Collision rate of bidisperse spheres settling in a compressional non-continuum gas flow. Journal of Fluid Mechanics 910, A10.DOI
  12. [15]Dubey, A., Bewley, G. P., Gustavsson, K. & Mehlig, B. 2024 Critical charges for droplet collisions. Physical Review Fluids 9 (7), 074302.DOI
  13. [17]Grabowski, W. W. & Wang, L.-P. 2013 Growth of cloud droplets in a turbulent environment. Annual Review of Fluid Mechanics 45, 293–324.DOI
  14. [18]Guazzelli, E. & Morris, J. F. 2011 A Physical Introduction to Suspension Dynamics. Cambridge University Press.DOI
  15. [19]Guo, S. & Xue, H. 2021 The enhancement of droplet collision by electric charges and atmospheric electric fields. Atmospheric Chemistry and Physics 21 (1), 69–85.DOI
  16. [20]Hamaker, H. C. 1937 The London-van der Waals attraction between spherical particles. Physica 4 (10), 1058–1072.DOI
  17. [21]Harrison, RG & Carslaw, KS 2003 Ion-aerosol-cloud processes in the lower atmosphere. Reviews of Geophysics 41 (3).
  18. [22]Harrison, R. G. 2004 The global atmospheric electrical circuit and climate. Surveys in Geophysics 25 (5), 441–484.
  19. [23]Harrison, R. G., Nicoll, K. A., Ulanowski, Z. & Mather, T. A. 2010 Self-charging of the Eyjafjallajökull volcanic ash plume. Environmental Research Letters 5 (2), 024004.DOI
  20. [24]Ichihara, M., Mininni, P. D., Ravichandran, S., Cimarelli, C. & Vagasky, C. 2023 Multiphase turbulent flow explains lightning rings in volcanic plumes. Communications Earth & Environment 4 (1), 417.
  21. [25]Israelachvili, J. N. 2011 Intermolecular and Surface Forces, 3rd edn. Academic Press.
  22. [26]Jeffrey, D. J. 1992 The calculation of the low Reynolds number resistance functions for two unequal spheres. Physics of Fluids A: Fluid Dynamics 4 (1), 16–29.
  23. [27]Khachatourian, A., Chan, H.-K., Stace, A. J. & Bichoutskaia, E. 2014 Electrostatic force between a charged sphere and a planar surface: A general solution for dielectric materials. The Journal of Chemical Physics 140 (7), 074107.
  24. [28]Khain, A., Arkhipov, V., Pinsky, M., Feldman, Y. & Ryabov, Ya. 2004 Rain enhancement and fog elimination by seeding with charged droplets. Part I: Theory and numerical simulations. Journal of Applied Meteorology 43 (10), 1513–1529.
  25. [29]Kim, S. & Karrila, S. J. 2013 Microhydrodynamics: Principles and Selected Applications. Courier Corporation.
  26. [30]Kim, S. & Mifflin, R. T. 1985 The resistance and mobility functions of two equal spheres in low-Reynolds-number flow. The Physics of Fluids 28 (7), 2033–2045.
  27. [31]Klimin, N. N., Rivkind, V. Ya. & Pachin, A. A. 1994 Collision efficiency calculation model as a software tool for microphysics of electrified clouds. Meteorology and Atmospheric Physics 53 (1), 111–120.
  28. [32]Lekner, J. 2012 Electrostatics of two charged conducting spheres. Proceedings of the Royal Society A 468 (2145), 2829–2848.
  29. [33]Lekner, J. 2016 Regions of attraction between like-charged conducting spheres. American Journal of Physics 84 (6), 474–477.
  30. [36]Méndez Harper, J., Courtland, L., Dufek, J. & McAdams, J. 2020 Microphysical effects of water content and temperature on the triboelectrification of volcanic ash on long time scales. Journal of Geophysical Research: Atmospheres 125, e2019JD031498.
  31. [37]Méndez Harper, J. & Dufek, J. 2016 The effects of dynamics on the triboelectrification of volcanic ash. Journal of Geophysical Research: Atmospheres 121 (14), 8209–8228.
  32. [38]Munirov, V. R. & Filippov, A. V. 2013 Interaction of two dielectric macroparticles. Journal of Experimental and Theoretical Physics 117 (5), 809–819.DOI
  33. [39]Nicoll, K. A. & Harrison, R. G. 2016 Stratiform cloud electrification: comparison of theory with multiple in-cloud measurements. Quarterly Journal of the Royal Meteorological Society 142 (700), 2679–2691.DOI
  34. [40]Patra, P., Koch, D. L. & Roy, A. 2022 Collision efficiency of non-Brownian spheres in a simple shear flow—the role of non-continuum hydrodynamic interactions. Journal of Fluid Mechanics 950, A18.DOI
  35. [41]Patra, P., Koch, D. L. & Roy, A. 2023 Collision efficiency of like-charged spheres settling in a quiescent environment. Journal of Fluid Mechanics 968, A22.DOI
  36. [42]Patra, P. & Roy, A. 2022 Brownian coagulation of like-charged aerosol particles. Physical Review Fluids 7 (6), 064308.DOI
  37. [43]Patra, P. & Roy, A. 2025 Gravity-induced collisions of uncharged cloud droplets in an electric field. Journal of the Atmospheric Sciences 82 (11), 2357–2373.arxiv.org/abs/2503.06688
  38. [44]Patra, P., Roy, A. & Wettlaufer, J. S. 2025 Electric field effects on the collision efficiency of uncharged water droplets in a linear flow. Physical Review Fluids 10 (10), 103601.arxiv.org/abs/2505.16010
  39. [45]Pruppacher, H. R. & Klett, J. D. 1997 Microphysics of Clouds and Precipitation. Kluwer Academic Publishers.
  40. [46]Rayleigh, L. 1882 XX. On the equilibrium of liquid conducting masses charged with electricity. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science 14 (87), 184–186.
  41. [47]Reeve, C. P., Williamson, C., Shelton, E., Stace, A. J. & Besley, E. 2025 Electrostatic aggregation of charged, polarizable particles in extreme atmospheric environments. The Journal of Physical chemistry A 129 (32), 7461.DOI
  42. [48]Ruan, X., Gorman, M. T. & Ni, R. 2024 Effects of electrostatic interaction on clustering and collision of bidispersed inertial particles in homogeneous and isotropic turbulence. Journal of Fluid Mechanics 980, A29.
  43. [49]Russel, W. B., Saville, D.A. & Showalter, W. R. 1991 Colloidal Dispersions. Cambridge University Press.
  44. [51]Saffman, P. G. & Turner, J. S. 1956 On the collision of drops in turbulent clouds. Journal of Fluid Mechanics 1 (1), 16–30.DOI
  45. [52]Schlamp, R. J., Grover, S. N., Pruppacher, H. R. & H amielec, A. E. 1976 A numerical investigation of the effect of electric charges and vertical external electric fields on the collision efficiency of cloud drops. Journal of the Atmospheric Sciences 33 (9), 1747–1755.
  46. [53]Schlamp, R. J., Grover, S. N., Pruppacher, H. R. & Hamielec, A. E. 1979 A numerical investigation of the effect of electric charges and vertical external electric fields on the collision efficiency of cloud drops: Part II. Journal of the Atmospheric Sciences 36 (2), 339–349.
  47. [55]Sundararajakumar, R. R. & Koch, D. L. 1996 Non-continuum lubrication forces between particles colliding in a gas. Journal of Fluid Mechanics 313, 283–308.
  48. [56]Textor, C & Ernst, G. G. J. 2004 Comment on “Particle aggregation in volcanic eruption columns” by Graham Veitch and Andrew W. Woods. Journal of Geophysical Research: Solid Earth 109 (B5).
  49. [57]Wallace, J. M. & Hobbs, P. V. 2006 Atmospheric Science: An Introductory Survey, , vol. 92. Elsevier.
  50. [58]Wang, H., Zinchenko, A. Z. & Davis, R. H. 1994 The collision rate of small drops in linear flow fields. Journal of Fluid Mechanics 265, 161–188.DOI
  51. [59]Wang, P. K. 2013 Physics and Dynamics of Clouds and Precipitation. Cambridge University Press.DOI
  52. [60]Zeichner, G. R. & Schowalter, W. R. 1977 Use of trajectory analysis to study stability of colloidal dispersions in flow fields. AIChE Journal 23 (3), 243–254.
  53. [61]Zhang, M., Yang, Z., Li, C., Li, J., Li, D., Fu, Q. & Yu, K. 2025 Collision characteristics of neutral and highly charged droplets in uniform electric fields. Physics of Fluids 37 (1), 012005.
  54. [62]Zhou, L. & Tinsley, B. A. 2007 Production of space charge at the boundaries of layer clouds. Journal of Geophysical Research: Atmospheres 112 (D11).
  55. [63]Zinchenko, A. Z. & Davis, R. H. 1994 Gravity-induced coalescence of drops at arbitrary Péclet numbers. Journal of Fluid Mechanics 280, 119–148.

Paper details

vlnc:W_wbxgzduy
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents