1. Introduction
1.1. Development of linear theories for arbitrary Debye number
Electrophoresis is a fundamental phenomenon describing the motion of charged particles under the influence of an electric field. First observed experimentally by Reuss (Reference Reuss1809), it has become integral to numerous scientific and technological applications. Notably, electrophoresis plays a key role in particle and protein separation (Vanifatova Reference Vanifatova2003; Xu et al. Reference Xu, Caswell, Tucker, Kabisatpathy, Brodhacker and Scrivens2007; Liu & Mathies Reference Liu and Mathies2009; Zhu et al. Reference Zhu, Lu and Liu2012; Schneider & Tripathi Reference Schneider and Tripathi2022), drug delivery systems (Simó et al. Reference Simó, Gallardo, Román, Barbas and Cifuentes2002), DNA analysis (Huang et al. Reference Huang, Quesada and Mathies1992; Dolnikk Reference Dolnikk1999) and advancements in microfluidics (Erickson & Li Reference Erickson and Li2004).
The theoretical foundation for electrophoresis was established by Smoluchowski (Reference Smoluchowski1903) for a thin electric double layer (EDL), who extended earlier work (Helmholtz Reference Helmholtz1879) to derive a more accurate formula for the electrophoretic mobility
where
$\tilde {\varepsilon }_m$
is the dielectric permittivity of the medium,
$\tilde {\zeta }$
is the zeta potential and
$\tilde {\mu }$
is the dynamic viscosity of the fluid. Dimensional variables are marked with a tilde to distinguish them from dimensionless counterparts. This equation implies that the particle velocity
$\tilde {U}_{\infty }$
is directly proportional to the electric field strength
$\tilde {E}_{\infty }$
, while
$\tilde {M}$
remains independent of
$\tilde {E}_{\infty }$
. This direct proportionality underlies the concept of linear electrophoresis.
Equation (1.1) is classically introduced in the context of weak applied fields and a thin EDL. However, these are not the only conditions under which Smoluchowski behaviour can arise. More recent asymptotic analyses have shown that, in the thin EDL limit, the same leading-order mobility/slip relation may remain valid even for strong applied fields, including dimensionless field strengths of the order of the inverse Debye number, provided that the particle is only moderately charged and the relevant thin layer asymptotic assumptions are satisfied (Schnitzer & Yariv Reference Schnitzer and Yariv2012b ; Schnitzer et al. Reference Schnitzer, Zeyde, Yavneh and Yariv2013). Thus, the classical weak-field interpretation should not be viewed as its universally necessary condition of validity.
Smoluchowski demonstrated that the electrophoretic drift with constant velocity results from the balance between the electrostatic (Coulomb) force acting on the net charge
$\tilde {q}$
of the particle and the Stokes frictional force opposing its motion. Substituting expressions for these forces into the force balance leads to (1.1).
However, the thin EDL limit is not always a proper assumption. In aqueous electrolytes, the Debye length ranges from
$1 \rm\, nm$
in concentrated solutions up to
$100\rm \, nm$
in dilute solutions, while the particle radius can span from hundreds of microns (microparticles) down to a few hundred nanometres (nanoparticles). Hence, the Debye number
$\varepsilon = \tilde {\lambda }_D / \tilde {a}$
may vary from
$10^{-5}$
to
$10^{2}$
and, thus, can be large for nanoparticles. Even larger Debye lengths are encountered in non-polar fluids, where
$\tilde {\lambda }_D$
can reach micrometres or even tens of micrometres (Prieve et al. Reference Prieve, Hoggard, Fu, Sides and Bethea2008), implying that
$\varepsilon \sim O(1)$
or even
$\varepsilon \gg 1$
.
In 1924, Debye and Hückel examined the case of large Debye numbers. Under the same two assumptions as in Smoluchowski‘s derivation (small
$\zeta$
potential and weak electric field) they obtained an expression differing from Smoluchowski‘s equation (1.1) by a numerical factor of 2/3.
In 1931, Henry (Reference Henry1931) showed that the discrepancy arises from the extent to which the electric field lines are distorted by the particle. For large particles (
$\varepsilon \to 0$
) in aqueous solutions, the electric field lines are parallel to the surface and the Smoluchowski‘s equation applies. For very small particles (
$\varepsilon \to \infty$
), with a small distortion of the electric field, the Debye–Hückel equation is valid. Henry introduced a unified formula accounting for intermediate values of the Debye number
$\varepsilon$
, ensuring a smooth transition between these two limiting cases (see table 1).
Comparison of theoretical models for the electrophoretic velocity
$\tilde {U}_{\infty }$
in the weak electric field regime. Function
$f(\alpha )$
is defined by (1.2).

Table 1. Long description
The table compares three models for electrophoretic velocity in the weak electric field regime. The models listed are Smoluchowski, Henry, and Debye-Hückel. Each model is associated with a specific velocity expression. The Smoluchowski model is valid for large particles in aqueous solutions where the electric field lines are parallel to the surface. The Debye-Hückel model applies to very small particles with minimal distortion of the electric field. The Henry model provides a unified formula for intermediate values of the Debye number, ensuring a smooth transition between the Smoluchowski and Debye-Hückel cases.
The function
$f(\alpha )$
in the table 1 is the well-known Henry function:
Here
$\alpha = \tilde a/\tilde \lambda _D$
is the inverse Debye number.
When
$\alpha \to 0$
, Henry‘s function
$f(\alpha ) \to 1$
and this case corresponds to the Debye–Hückel limit. The Henry function increases monotonically with
$\alpha$
(see figure 1) and approaches
$3/2$
as
$\alpha \to \infty$
(Smoluchowski‘s limit).
The Henry function,
$f(\alpha )$
, illustrating the transition from the Debye–Hückel (D–H) limit (
$\alpha \to 0$
) to the Smoluchowski (SMO) limit (
$\alpha \to \infty$
).

Physically, the thin EDL limit (
$\varepsilon \to 0$
) often corresponds to electrophoresis in polar liquids such as water, where the Smoluchowski formula is valid. The thick EDL limit (
$\varepsilon \to \infty$
) is typical for non-polar liquids, such as decane, dodecane, benzene, toluene, where the EDL thickness can exceed the particle radius, and the Debye–Hückel formula applies. Henry‘s unifying approach bridges these two regimes via
$f(\alpha )$
. All these formula assume a weak electric field and a uniform (angularly symmetric) surface charge. Consequently, any nonlinear effects arising from stronger fields or non-uniform surface charge distributions remain beyond the scope of these classical theories.
1.2. Development of nonlinear theories for arbitrary Debye numbers
In real systems, a moderately strong electric field or relatively high
$\zeta$
potentials invalidate the assumptions behind (1.1), making it necessary to correct Smoluchowski‘s formula. The Nernst–Planck–Poisson–Stokes (NPPS) equations are fundamentally nonlinear; hence, even for small but finite
$\varepsilon$
, a sufficiently strong electric field can distort the spherical ion distribution near the particle surface. Three key effects are typically distinguished in this context – relaxation, retardation and surface conductance Shaw (Reference Shaw1969) – each introducing additional forces that alter the electrophoretic velocity relative to Smoluchowski‘s prediction. These effects, which are generally not mutually exclusive, highlight the limitations of the classical linear theories and emphasise the need for more comprehensive models. The specific roles of retardation, relaxation and surface conduction are examined in greater detail in the section addressing the forces acting on the particle.
The early developments to refine the Smoluchowski formula were focused on the weak electric field regime and on small but finite
$\varepsilon$
. The research was initiated independently and nearly simultaneously by Booth (Reference Booth1948) and Henry (Reference Henry1948). Later Overbeek (Reference Overbeek1943) and Booth (Reference Booth1950) confirmed and refined these results. All these studies assumed a weak electric field threshold and used linearised equations. Subsequent numerical works by Wiersema et al. (Reference Wiersema, Loeb and Overbeek1966) and O’Brien & White (Reference O’Brien and White1978) examined highly charged particles (
$\tilde \zeta /\tilde \varPhi _0 \ge O(1)$
) under weak electric fields, revealing that the electrophoretic velocity can deviate significantly from the simple linear dependence on
$\zeta$
potential. Although for small
$\varepsilon$
the mobility still followed (1.1), at a large
$\zeta$
potential the numerical solutions showed a marked deviation, even displaying a peak in velocity as a function of
$\varepsilon$
. This behaviour was attributed to surface conduction, a phenomenon defined and described by Dukhin (Reference Dukhin1965), Dukhin & Shilov (Reference Dukhin and Shilov1969), Dukhin & Semenikhin (Reference Dukhin and Semenikhin1970). Later, O’Brien & Hunter (Reference O’Brien and Hunter1981) simplified and revised Dukhin‘s derivations, proposing modifications to Smoluchowski‘s equation for a thin EDL, again linking the observed effects to surface conduction. O’Brien (Reference O’Brien1983) further extended it to accommodate multi-component electrolytes without imposing specific geometrical constraints. Importantly, all these studies remained within the weak electric field linearised framework, focusing on small deviations from equilibrium.
Works of Yariv (Reference Yariv2009), Yariv & Davis (Reference Yariv and Davis2010), Schnitzer & Yariv (Reference Schnitzer and Yariv2012a
,
Reference Schnitzer and Yarivb
,
Reference Schnitzer and Yarivc
), Schnitzer et al. (Reference Schnitzer, Zeyde, Yavneh and Yariv2013), Schnitzer & Yariv (Reference Schnitzer and Yariv2014) pioneered a systematic study of electrophoresis in a nonlinear formulation using asymptotic methods. Their results do not fall under the scope of either Smoluchowski‘s theory or the weak-field models of O’Brien & White (Reference O’Brien and White1978), O’Brien & Hunter (Reference O’Brien and Hunter1981). In Schnitzer & Yariv (Reference Schnitzer and Yariv2012b
) they developed an asymptotic solution for the limit
$\varepsilon \to 0$
(small Debye number) and a ‘logarithmically large’
$\zeta$
potential, namely
$\tilde \zeta / \tilde \varPhi _0 = O(\ln {1/\varepsilon }) \to \infty$
(highly charged surfaces). Importantly, they removed the earlier restriction
$E_{\infty } = \tilde E_{\infty } \tilde a/\tilde \varPhi _0 \to 0$
and focused on the moderate-field regime,
$E_{\infty } = O(1)$
. By solving the inner problem analytically, they derived effective boundary conditions for a macroscale model in the electroneutral bulk region. This model is described by the equations for the salt concentration, the electrostatic potential and the Stokes equations with a Coulomb body force (see Probstein Reference Probstein1994). Based on this model, Schnitzer et al. (Reference Schnitzer, Zeyde, Yavneh and Yariv2013) calculated the electrophoretic velocity of a microparticle in the ‘weakly nonlinear’ regime by performing a perturbation expansion in
$E_{\infty }$
for
$\sigma \ge O(1)$
and for an arbitrary Dukhin number,
${\textit{Du}} \sim \varepsilon \sigma$
, where
$\sigma = {\tilde {\sigma } \tilde {\lambda }_D}/{\tilde {\varepsilon }_m \tilde {\varPhi }_0}$
is the dimensionless surface charge density. Since
${\textit{Du}}$
characterises surface conduction, they found that, under these conditions,
$U_{\infty } \sim a_1 E_{\infty } + a_3 E_{\infty }^3$
, with
$a_2 = 0$
, at small Péclet number;
$a_1$
represents the linear electrophoretic mobility, corresponding to the first-order term in the expansion with respect to
$E_{\infty }$
. In the limit of weak electric fields, it coincides with the classical results of Debye–Hückel or Smolukhowski, depending on the value of the Debye number;
$a_3$
is the first nonlinear correction to the particle velocity, arising from effects such as relaxation, retardation and surface conductance.
When the Debye number is finite,
$\varepsilon = O(10^{-1}) - O(1)$
, the EDL can no longer be treated as asymptotically thin, even though its thickness may still be smaller than the particle radius. This regime is therefore intermediate between the classical thin EDL limit and the thick EDL limit. It is especially relevant for low conductivity or non-polar systems, where reduced ion concentrations and low solvent polarity may lead to significantly larger Debye lengths than in typical aqueous electrolytes. Under strong external electric fields, such conditions can produce marked deviations from Smoluchowski‘s formula, requiring a direct analysis of how relaxation, electrophoretic retardation, surface conduction and finite EDL thickness effects interact.
The seminal experimental study of Stotz (Reference Stotz1978) also addressed key theoretical aspects of high-field electrophoresis in non-polar liquids, where particles can carry large surface charges and maintain a moderately thick EDL under such conditions.
Asymptotic analyses provide valuable insight into electrokinetic motion, but they inherently rely on the existence of a small or large parameter that enables a systematic expansion. In a non-polar medium, the Debye number
$\varepsilon$
typically spans a broad range, from
$O(10^{-1})$
up to
$O(10^{2})$
. While the upper end of this range can be approached through a large-
$\varepsilon$
(formally
$\varepsilon \to \infty$
) limit, the lower end,
$\varepsilon = O(10^{-1})$
, does not correspond to any classical asymptotic regime. This intermediate-
$\varepsilon$
case is therefore particularly difficult to treat analytically: there is no natural small (or large) parameter, and neither the thin-double-layer (Smoluchowski-type) nor the thick-double-layer (Hückel-type) approximations apply. To the best of our knowledge, no analytical solutions (in the form of closed-form expressions or controlled asymptotics) are currently available for the nonlinear intermediate-
$\varepsilon$
regime, which motivates the primarily numerical approach adopted in this work.
Recent reviews have summarised the historical development and current state of nonlinear electrophoresis, emphasising that electrophoretic mobility should no longer be regarded as a field-independent material property once the applied field becomes sufficiently strong (Khair Reference Khair2022; Mendiola-Escobedo & Lapizco-Encinas Reference Mendiola-Escobedo and Lapizco-Encinas2025). In particular, Mendiola-Escobedo & Lapizco-Encinas (Reference Mendiola-Escobedo and Lapizco-Encinas2025) provide a broad overview of fifty years of nonlinear electrophoresis, from the early works of Dukhin and co-workers to modern experimental and numerical studies. They distinguish several nonlinear regimes, including the classical cubic regime for dielectric particles, second-kind or superfast electrophoresis for conducting and semiconducting particles, and the
$E^{3/2}$
regime associated with large Peclet numbers. This classification highlights that nonlinear electrophoresis is governed by the coupled action of electric-field-induced concentration polarisation, surface conduction, ionic-cloud deformation and hydrodynamic transport. The same review also identifies the motion of highly charged particles with thick or intermediate EDLs as an important open problem, which motivates direct numerical studies beyond the classical thin EDL and weak-field limits.
Khair (Reference Khair2022) identifies three main directions in the modern development of the field: experimental measurements of field-dependent mobility, asymptotic theories in the thin EDL limit and numerical solutions of the full electrokinetic equations. This classification is particularly relevant to the present work, which focuses on the latter direction and addresses a regime where direct numerical solution is required to resolve the coupled deformation of the ionic cloud, concentration polarisation and hydrodynamic response. Although thin EDL asymptotic models, especially the Schnitzer–Yariv macro-scale formulation, have provided substantial insight into nonlinear electrophoresis, such approaches are not sufficient to describe intermediate-Debye-number regimes in which the diffuse layer occupies a finite fraction of the particle scale. As noted in recent reviews, numerical computations are needed to bridge the gap between the thin and thick EDL limits Khair (Reference Khair2022). The present study therefore solves the full NPPS system in order to examine strong-field electrophoresis at an intermediate Debye number, where the ionic cloud remains attached to the particle but undergoes a qualitative restructuring.
The direct numerical simulation of the NPPS equations offers a reliable approach for studying every detail of electrophoresis under arbitrary parameters. A recent paper by Cobos & Khair (Reference Cobos and Khair2023) presented a numerical solution of the electrophoretic problem focusing on
$\varepsilon \sim O(1)$
and
$\varepsilon \gt \gt 1$
. In Cobos & Khair (Reference Cobos and Khair2023) a spectral element algorithm was employed to discretise the NPPS system under steady-state conditions and the resulting nonlinear algebraic system was solved using Newton‘s method. The previously obtained analytical solution (Khair Reference Khair2017) for
$\varepsilon \gt \gt 1$
was compared with the numerical solution: for a large coupling coefficient
$\varkappa$
between hydrodynamics and electrostatics, the quantitative agreement was good, while for a small
$\alpha$
, qualitative agreement was maintained. In this work, the authors also provided detailed charge density fields for various external electric fields, surface charges and Debye numbers. Additionally, the numerical solution showed good agreement with Stotz‘s experimental data.
Nevertheless, the work Cobos & Khair (Reference Cobos and Khair2023) faces serious limitations. The NPPS equations are very stiff, leading to numerical instability for both highly charged particles
$\sigma \gt \gt 1$
and a strong electric field
$E_{\infty } \gt \gt 1$
. In that study, numerical computations were limited to
$E_{\infty } = 10$
for
$\sigma \gt \gt 1$
due to numerical instabilities of the employed algorithm. The authors explicitly noted that extending the calculations to stronger fields would require substantially finer meshes and considerably longer computation times. As a result, the analysis did not cover the regime of very strong electric fields, higher zeta potentials or the detailed transient evolution of the ionic cloud surrounding the particle – effects that may play a decisive role in nonlinear electrophoresis.
Cobos & Khair (Reference Cobos and Khair2025) recently extended full electrokinetic simulations of nonlinear electrophoresis to unsteady electric fields, including suddenly applied, switched-off and oscillatory fields. They showed that, after a sudden field application, the initial particle response is governed by momentum diffusion, whereas the nonlinear approach to the steady mobility is controlled by the slower deformation of the Debye cloud. In oscillatory fields, the amplitude and phase of the mobility depend on whether the Debye cloud has sufficient time to deform during one field period. These results highlight the central role of ionic-cloud dynamics in nonlinear electrophoresis. The present work is complementary: we focus on the steady strong-field regime at an intermediate Debye number and analyse in detail the spatial restructuring of ion concentrations, charge density and ionic fluxes that determine the nonlinear mobility.
The present work can be regarded as a direct continuation and extension of the study by Cobos & Khair (Reference Cobos and Khair2023). We overcome limitations by employing a finite-difference numerical algorithm for solving the full, time-dependent nonlinear NPPS system, that allows stable and accurate solutions for highly nonlinear regimes. This was done based on a third-order accurate semi-implicit Runge–Kutta scheme developed by Nikitin (Reference Nikitin2006a , Reference Nikitinb ) for incompressible Navier–Stokes equations. This approach was applied to the Nernst–Planck equations. A distinctive feature of the method is that only part of the Nernst–Planck operator is solved using an implicit scheme – specifically, those terms of the operator that physically introduce instability into the numerical algorithm. While the remaining part of the operator is treated explicitly. This allows us to maintain a reasonable balance between computational cost, numerical stability and convergence of the algorithm.
As a result, this enables us to systematically explore the strongly nonlinear regimes (both
$E_{\infty } \gt \gt 1$
and
$\sigma \gt \gt 1$
), as well as to capture the full spatiotemporal evolution of the EDL. In particular, our computations resolve the dynamic formation and rearrangement of the ion cloud near the particle surface, providing insight into the interplay between electric field-induced distortion, ion migration and hydrodynamic flow. The calculated ion concentration is shown for the first time via ion-flux streamfunctions.
The present study is restricted to a non-polar medium, for which the electrohydrodynamic coupling is weak. We therefore fix it for all computations reported in the paper. Within this non-polar setting, we consider intermediate
$\varepsilon \sim O(10^{-1})$
. Our main interest is in nonlinear electrophoresis, which is promoted by sufficiently strong applied fields and surface charge; we thus focus on cases with
$E \sigma \gtrsim 1$
. Such conditions are also typically associated with
${\textit{Du}} = O(1)$
, indicating significant surface conduction and concentration polarisation.
2. Mathematical formulation
2.1. Nernst–Plank–Poisson–Stokes system of equations
A spherical dielectric microparticle of radius
$\tilde {a}$
is suspended in a binary univalent electrolyte subjected to an external electric field of intensity
$\tilde {E}_{\infty }$
. For a negatively charged particle, the electrophoretic motion occurs in the direction opposite to the electric field. But in a particle frame, the fluid is seen moving along the electric field direction (see figure 2).
Electrophoresis of a negatively charged particle moving with dimensionless velocity
$U_{\infty }$
opposite to the external electric field of strength
$E_{\infty }$
.

In the present study we restrict attention to a symmetric binary system of monovalent mobile charge carriers with equal diffusivities,
$D^+ = D^-$
. This assumption allows us to isolate the nonlinear effects associated with strong-field deformation of the ionic cloud, finite EDL thickness, concentration polarisation and electrohydrodynamic coupling. Allowing
$D^+ \neq D^-$
would introduce additional asymmetry between the transport of positive and negative charge carriers, including unequal electromigration–diffusion balances and the possible formation of diffusion potentials. These effects may alter both the spatial charge distribution and the nonlinear electrophoretic response. Their systematic investigation would require an additional parametric study of the diffusivity ratio and is therefore left for future work.
We adopt an axisymmetric formulation in spherical coordinates (
$\tilde r$
,
$\theta$
) as indicated in figure 2. The behaviour of ions in the medium is described by the system of Nernst–Planck equations, with respect to the ion molar concentrations
$\tilde {C}^{\pm }$
. The system is supplemented by the Poisson equation for the electric potential
$\tilde {\varPhi }$
and the Navier–Stokes equations for the velocity field
$\boldsymbol{\tilde {U}} = (\tilde {U}_r, \tilde {U}_{\theta })$
. Due to the smallness of the scale and, as a consequence, the smallness of the Reynolds number, the Stokes approximation for a creeping flow is used to describe the motion of a fluid. The electric field inside the microparticle is described by the Laplace equation for the potential
$\tilde {\varphi }$
. On the particle surface the existence of a surface charge
$\tilde {\sigma }$
is assumed, the density of which is uniform over the entire surface of the particle.
The statement of the problem is carried out in a coordinate system moving with particle velocity
$\tilde {U}_{\infty }$
.
To analyse the problem under general conditions, the system is non-dimensionalised using the characteristic length
$\tilde a$
, thermal voltage
$\tilde \varPhi _0 = \tilde R \tilde T / \tilde F$
, equilibrium electrolyte concentration
$\tilde {C}_{\infty }$
, dynamic fluid viscosity
$\tilde {\mu }$
and ion diffusion coefficient
$\tilde {D}$
. Here
$\tilde {R}$
is the universal gas constant,
$\tilde {T}$
is the absolute temperature and
$\tilde {F}$
is the Faraday constant. The characteristic velocity is given by
$\tilde {D} / \tilde {a}$
.
The system is described by the following dimensionless parameters: electric field strength
$E_{\infty } = \tilde {a} \tilde {E}_{\infty } / \tilde {\varPhi }_0$
; Debye number
$\varepsilon = \tilde {\lambda }_D / \tilde {a}$
, where
$\tilde {\lambda }^2_D = \tilde {\varepsilon }_m \tilde \varPhi _0/ 2 \tilde F \tilde {C}_{\infty }$
is the thickness squared of the EDL; the dimensionless surface charge
$\sigma =\tilde \sigma /\tilde \sigma _0$
, where
$\tilde \sigma _0 = \tilde \varepsilon _m \tilde \varPhi _0/\tilde {\lambda }_D$
; the coupling coefficient between the electrostatics and hydrodynamics
$\varkappa = \tilde {\varepsilon }_m \tilde {\varPhi }_0^2 / \tilde {\mu } \tilde {D}$
; the ratio of electric permittivities of particle and medium
$\delta = \tilde {\varepsilon }_p / \tilde {\varepsilon }_m$
, where
$\tilde {\varepsilon }_p$
and
$\tilde {\varepsilon }_m$
are, respectively, the electric permittivities of the particle and the electrolyte.
The dimensionless system of NPPS equations can be written as
The potential inside the particle
$\varphi$
satisfies the Laplace equation, assuming that the particle is electrically insulated and non-conductive,
When crossing the particle‘s surface, the electric potential is assumed to be continuous, while its derivative normal to the surface undergoes a jump due to the surface charge
$\sigma$
and different permittivities in the medium and inside the dielectric particle. On the particle surface, the flux of positive and negative ions is equal to zero (impermeability condition), and the velocity components satisfy the no-slip conditions:
Far from the particle‘s surface, at
$r \to \infty$
, the ion concentration tends to equilibrium, the electric field strength tends to the external field
$E_{\infty }$
and the fluid velocity tends to the particle velocity
$U_{\infty }$
:
For the non-stationary problem, the concentrations
$C^\pm$
are uniformly distributed over space:
The NPPS formulation employed here is not restricted to aqueous electrolytes. It provides a continuum description of a dilute concentration of mobile charged species in a Newtonian dielectric liquid. In non-polar media, such charge carriers may be ions formed by oil soluble ionic additives or charged reverse micelles. Provided that these carriers can be characterised by effective concentrations, diffusivities and electrical mobilities, their transport is governed by the same combination of diffusion, electromigration and advection as in a polar media. The electrostatic potential and fluid motion are then described by the Poisson and Stokes equations, respectively. Continuum electrodiffusion models of this type have previously been applied to electrophoresis and EDL deformation in low conductivity non-polar liquids (Strubbe et al. Reference Strubbe, Beunis, Marescaux, Verboven and Neyts2013; Khair Reference Khair2017). More generally, electrodiffusion theory provides a microscopic continuum basis for the electrohydrodynamic description of weakly conducting dielectric liquids (Mori & Young Reference Mori and Young2018).
In weakly dissociating electrolytes, neutral ion pairs and free ions may interconvert according to a reaction of the form
where
$S$
denotes a neutral undissociated species. Baygents & Saville (Reference Baygents and Saville1991) showed that such association/dissociation processes may modify electrophoretic transport by opposing perturbations of the diffuse ionic cloud. A more general electrodiffusion model would therefore contain source terms in the Nernst–Planck equations,
where
$R$
represents the local production or removal of free ions by dissociation and recombination.
The reaction term is neglected here under the assumption that association/dissociation kinetics are slow compared with the electrodiffusive redistribution of the mobile charge carriers. The model therefore treats the total population of free carriers as prescribed and describes only their spatial redistribution by diffusion, electromigration and advection. This approximation is appropriate when the applied field does not produce a substantial change in the bulk carrier concentration over the characteristic transport time. If the reaction and transport time scales become comparable, the source term
$R$
should be retained; the resulting relaxation toward local chemical equilibrium may weaken concentration polarisation and alter the quantitative nonlinear mobility response.
Accordingly, the present model assumes that the concentration of charge carriers in the solvent is low, which corresponds to a low conductivity liquid. Their fluxes are determined by diffusion, electromigration and hydrodynamic advection. The liquid permittivity and viscosity, as well as the ionic diffusivities and mobilities, are taken to be independent of the applied electric field. The total amount of mobile charge carriers is also assumed not to change through field-induced dissociation or recombination in the bulk liquid. Thus, the model does not include an additional field-dependent production of charge carriers.
It should be emphasised that these assumptions do not imply a spatially uniform conductivity. Within the NPPS formulation, the local ionic concentrations evolve in response to electromigration, diffusion and advection. Consequently, the local conductivity may vary substantially because of concentration polarisation and deformation of the charge cloud. What is neglected is an additional constitutive dependence of the carrier mobilities on the electric field magnitude.
2.2. Calculation of the electrophoretic velocity based on the force balance
In order to determine the electrophoretic velocity
$U_{\infty }$
, the system of equations (2.1)–(2.8) must be closed by imposing the force-free condition on the particle. This condition requires that the sum of hydrodynamic (viscous)
$F_V$
and electrostatic (Maxwell–Wagner)
$F_M$
stresses acting on the particle‘s surface
$\varSigma$
is zero. Mathematically, it is expressed as
\begin{equation} F_V + F_M = \int \limits _{\varSigma } \big(-\varPi \boldsymbol{n} + \tau _{nn}^V \boldsymbol{n} + \tau _{nl}^V \boldsymbol{l} + \tau _{nn}^M \boldsymbol{n} + \tau _{nl}^M \boldsymbol{l}\big) \, {\rm d} \varSigma = 0 .\end{equation}
Only projections to the
$x$
axis must be kept due to axial symmetry of the problem. Projection of the normal unit vector
$\boldsymbol{n}$
on the
$x$
axis is
$n_x = \cos \theta$
and that of the tangential unit vector
$l$
is
$l_x = -\sin \theta$
. Hence,
\begin{align} \tau _{rr}^M = \frac {\varkappa }{2}\left [\left (\frac {\partial \varPhi }{\partial r}\right )^2 -\frac {1}{r^2}\left (\frac {\partial \varPhi }{\partial \theta }\right )^2\right ]\!, \qquad \tau _{r \theta }^M = \varkappa \frac {1}{r} \frac {\partial \varPhi }{\partial r} \frac {\partial \varPhi }{\partial \theta }.\end{align}
To simplify the description of the axisymmetric flow around the particle, we introduce the streamfunction
$\varPsi (r,\theta )$
, which satisfies the incompressibility condition. The radial and angular components of the velocity field are expressed in terms of
$\varPsi$
as
The explicit forms of
$F_V$
and
$F_M$
, derived through the projection of the stress tensors onto the
$x$
axis, are given by
\begin{align} F_V &= \pi r^2\,\int _0^\pi \left ( -\sin {\theta }\,\frac {\partial ^3 \varPsi }{\partial r^3} - \frac {\sin {\theta }}{r^2} \,\,\frac {\partial ^3 \varPsi }{\partial r \partial \theta ^2} + \frac {2\sin {\theta }}{r} \,\,\frac {\partial ^2 \varPsi }{\partial r^2} +\right .\nonumber \\ & \left . + \frac {5\cos {\theta }}{r^2} \,\,\frac {\partial ^2 \varPsi }{ \partial r \partial \theta } - \frac {4\sin {\theta }}{r^2} \,\,\frac {\partial \varPsi }{ \partial r}-\frac {8\cos {\theta }}{r^3} \,\,\frac {\partial \varPsi }{ \partial \theta } \right ) \,{\rm d}\theta , \\[-12pt]\nonumber \end{align}
\begin{align} F_M &= -\pi \varkappa \,\int _0^{\pi } \left (r^2 \cos {\theta } \sin {\theta }\left (\frac {\partial \varPhi }{\partial r}\right )^2 + r^2 \sin ^2{\theta }\frac {\partial \varPhi }{\partial \theta }\frac {\partial ^2 \varPhi }{\partial r^2} + \sin ^2{\theta }\frac {\partial \varPhi }{\partial \theta }\frac {\partial ^2 \varPhi }{\partial \theta ^2}\right ) \,{\rm d} \theta . \end{align}
In our approach, both viscous and Maxwell stress tensors are computed directly from the numerical solution of the coupled NPPS system. Importantly, by calculating the full stress tensors, we inherently account for all four forces acting on the particle. So our formulation naturally captures the coupling between hydrodynamic and electrostatic phenomena, without invoking any simplifying assumptions such as potential flow or neglecting surface conduction effects. Thus, the electrophoretic velocity
$U_{\infty }$
obtained from the force-free condition incorporates the complete physics of the system, including nonlinear interactions between the fluid motion and the electric field distribution around the particle.
The system of equations (2.1)–(2.18) is solved numerically using a finite-difference method detailed in Frants et al. (Reference Frants, Amiroudine and Demekhin2024). The following section presents the numerical results and explores the physical implications under varying electric field strengths and Debye numbers.
3. Numerical results
To clarify the physical basis of the parameter selection, we discuss below how the dimensionless parameters used in the simulations are related to the real physical systems. The Debye number
$\varepsilon$
and the surface charge density
$\sigma$
was chosen to represent the values inferred from the measurements of Stotz (Reference Stotz1978). The applied field
$E_\infty$
was varied over a range covering both the experimental conditions and the stronger nonlinear regime investigated numerically. The parameters
$\kappa$
and
$\delta$
were kept fixed in the main parameter study so that the effects of
$E_\infty$
and
$\sigma$
could be examined separately. The parameter
$\kappa$
was also selected to be as close as possible to the experimental conditions reported by Stotz (Reference Stotz1978), within the framework of the present dimensionless model. Their physical interpretation is discussed below, and an additional sensitivity analysis is presented for the permittivity ratio
$\delta$
.
The Debye number
$\varepsilon = \lambda _D/a$
was chosen to reproduce the experimentally relevant ratio of the EDL thickness to the particle radius. In the experiments of Stotz (Reference Stotz1978), the characteristic values
$\lambda _D \sim 2 \,\unicode{x03BC} \mathrm{m}$
and
$a\sim 12\, \unicode{x03BC} \mathrm{m}$
yield
$\varepsilon = 0.166$
. Since
$\varepsilon$
directly controls the spatial extent of the ionic cloud, this value represents the intermediate EDL regime relevant to the experimental system.
3.1. Sensitivity to the particle to liquid permittivity ratio
$\delta$
The present model contains several independent dimensionless parameters. The main objective of the study is to determine the influence of the applied field
$E_\infty$
and the surface charge
$\sigma$
at fixed values of
$\varepsilon$
and
$\kappa$
. To avoid introducing an additional dimension into the parameter space, the permittivity ratio
$\delta$
was also kept fixed at
$\delta = 0.05$
used in our previous studies. This choice was based on the assumption that the electrophoretic response is only weakly sensitive to the dielectric contrast between the particle and the surrounding liquid.
Since this assumption is particularly important for comparison with non-polar liquids, an additional sensitivity analysis was performed by varying
$\delta$
over the interval
$0.05 \leqslant \delta \leqslant 1.5$
, while keeping all other parameters fixed. The selected interval spans the value used in the original calculations and the range relevant to a non-polar system (Stotz Reference Stotz1978). The results are shown in figure 3.
Electrophoretic velocity
$U_\infty$
as a function of the particle to liquid permittivity ratio
$\delta = \varepsilon _p / \varepsilon _m$
for
$\varepsilon = 0.166$
,
$\sigma = -20$
and
$E_\infty = 5$
,
$46.7$
and
$100$
. The right vertical axis shows the relative change
$\Delta U_\infty$
with respect to the value at
$\delta = 0.05$
.

The permittivity ratio
$\delta = \varepsilon _p/\varepsilon _m$
is a material parameter determined by the dielectric properties of the particle and the surrounding liquid. For dielectric particle and non-polar media, the permittivities of the two phases are typically of the same order, so that
$\delta$
is generally expected to be close to unity. In particular, for the system considered by Stotz, typical values
$\varepsilon _p \approx 2.4$
–
$2.6$
for polystyrene particle and
$\varepsilon _m \approx 1.9$
–
$2.2$
for the liquid give
$\delta \approx 1.1$
–
$1.3$
. Thus, the value
$\delta = 0.05$
used in our original calculations does not represent the dielectric permittivities ratio of the Stotz system and was retained only as the fixed value adopted in our previous studies.
Nevertheless, the results presented in figure 3 confirm that the electrophoretic response is only weakly sensitive to
$\delta$
over the considered range. At
$E_\infty = 5$
, the electrophoretic velocity decreases slightly with increasing
$\delta$
, and its total relative variation remains below approximately
$0.2\,\%$
. At
$E_\infty = 46.7$
, the dependence reverses and
$U_\infty$
increases by approximately
$1.4\,\%$
over the interval. At the stronger field
$E_\infty = 100$
, the influence of
$\delta$
becomes more pronounced, but the total increase remains limited to approximately
$6\,\%$
. This weak sensitivity is consistent with the structure of the model: the parameter
$\delta$
enters only through the electrostatic boundary condition at the particle surface and does not appear explicitly in the bulk NPPS equations. Therefore, the dominant effects discussed in the present work are controlled primarily by
$E_\infty$
and
$\sigma$
at fixed
$\varepsilon$
, while the main conclusions remain robust with respect to the precise value of
$\delta$
within the considered range.
3.2. Time evolution
The early stages of electrophoresis are inherently unsteady: immediately after the external field is switched on, the EDL deforms and relaxes towards its steady configuration. Accordingly, we study the transient response following the sudden application of the field
$E_{\infty }$
and illustrate it for an intermediate Debye number,
$\varepsilon = 0.166$
. Unlike most previous studies, which focus primarily on steady electrophoretic mobility, we resolve the full temporal evolution. It provides essential insight into the non-steady dynamics of electrophoresis. In particular, it reflects the relaxation behaviour of the EDL and the delayed hydrodynamic response of the surrounding medium. By analysing the time-dependent growth of
$U_{\infty }$
, one can identify the characteristic response time, which governs the transition to steady motion. This time scale is critical for understanding how quickly a system responds to an external stimuli and varies depending on key parameters such as the Debye length, particle size and ion diffusivity. Furthermore, the evolution of
$U_{\infty }(t)$
reveals the impact of relaxation forces and transient electrohydrodynamic phenomena that remain completely hidden in purely steady-state analyses.
Acceleration of the particle upon switching on the electric field at
$\tilde {t} \approx 10\rm \, s$
. Prior to this moment, the particle is kept in a field free liquid medium so that a symmetric equilibrium EDL can fully develop around it. The main dimensional parameters correspond to the particle radius
$12\rm\, \unicode{x03BC} m$
and a field strength
$\tilde {E}_\infty = 10^5\rm \, V\,m^{-1}$
(see figure 12 of Stotz Reference Stotz1978) with a characteristic response time
$\tilde {\tau } = 0.9172\rm \, s$
. The corresponding dimensionless parameters are
$E_{\infty } = 46.7$
,
$\varepsilon = 0.166$
,
$\varkappa = 0.05$
,
$\sigma = -20$
,
$\delta = 0.05$
.

To illustrate these general considerations about the temporal response, figure 4 shows the evolution of the dimensional electrophoretic velocity
$\tilde {U}_{\infty }(t)$
after the uniform external electric field is applied at
$\tilde {t} \approx 10\rm \, s$
. During the preceding time interval no electric field is applied. This allows the equilibrium EDL to fully establish around the particle. In this way, the subsequent evolution of
$\tilde {U}_{\infty }(t)$
and other functions capture the response of the equilibrium EDL to the electric field, rather than a simultaneous build-up of the EDL and the onset of particle motion. Such a protocol is consistent with experimental practice, where the suspension is typically allowed to reach equilibrium before the electric field is switched on. The simulation parameters correspond to the experimental conditions of Stotz (Reference Stotz1978), with particle radius
$\tilde {a} = 12\rm \, \unicode{x03BC} m$
and electric field strength
$\tilde {E}_{\infty } = 10^5 \rm \, V\,m^{-1}$
. The velocity
$\tilde {U}_{\infty }$
rises from zero and saturates to a steady value within
$\approx 0.12\rm \, s$
, while the characteristic system response time is
$\tilde {\tau } = 0.9172\rm \, s$
based on
$\tilde {D}/\tilde {a}^2$
. Notably, the particle reaches its terminal velocity faster than this characteristic time, indicating that the mechanical response of the particle to the applied electric field is faster than the complete restructuring of the ionic cloud. Although the EDL continues to evolve beyond this point (see supplementary material 1 is available at https://doi.org/10.1017/jfm.2026.12033), the net force acting on the particle stabilises earlier, allowing
$U_{\infty }$
to reach a plateau before full electrostatic equilibrium is re-established. This behaviour reflects the delayed EDL relaxation and the finite response time of the surrounding medium.
Charge density
$\rho = C^+ - C^-$
evolution at a moderate Debye number. The parameters used are
$\varepsilon = 0.166$
,
$E_{\infty } = 46.7$
,
$\sigma = -20$
,
$\varkappa = 0.05$
,
$\delta = 0.05$
.

Figure 5 demonstrates the EDL dynamics in response to the electric field. At the initial moment (
$t \sim 0$
), the system is in electrostatic equilibrium – the EDL is nearly symmetric. As early as
$t = 0.01$
, the EDL begins to expand and lose its symmetry, and by
$t = 0.02$
the slight accumulation of charge
$\rho$
on the right (downstream) side of the particle is seen. The ion cloud is displaced and stretched predominantly in the direction of electric field, indicating the onset of nonlinear effects. By
$t = 0.05$
, a distinct elongated structure of positive
$\rho$
develops in the wake of the particle. At
$t = 0.10$
, this structure becomes more pronounced and more strongly stretched behind the particle, reflecting both ion redistribution and the growing contribution of relaxation effects. At the final snapshot (
$t = 0.25$
), the EDL exhibits a complex, strongly asymmetric structure: a broadened front region with lower charge density and more narrow, high-density charge in the trailing region. Additionally, a pair of small negative charge regions are observed on the lateral sides of the particle, caused by convective ion redistribution and microvortical flow. These features are consistent with the formation of an extended space-charge (ESC) region, qualitatively similar to the space-charge region previously described near ion-selective interfaces (Zaltzman & Rubinstein Reference Zaltzman and Rubinstein2007; Mishchuk Reference Mishchuk2010; Rubinstein & Zaltzman Reference Rubinstein and Zaltzman2010).
3.3. Non-equilibrium EDL
An important manifestation of nonlinear dynamics in electrophoresis is the breakdown of the EDL spherical symmetry (see figure 6). As the particle begins to move under the influence of an electric field, the EDL cannot instantaneously adjust to the new flow and field conditions: it lags behind the moving particle, becomes asymmetric (elongated) and creates the relaxation force opposite to the direction of the particle motion. This effect becomes particularly pronounced at moderate to large values of the product
$\varepsilon E_{\infty }$
, where the EDL deformation is no longer a small perturbation of the equilibrium spherical state. Consequently, asymptotic approaches based on small deviations from the equilibrium EDL are insufficient for capturing the governing physics. This limitation is particularly important in non-polar systems, where the standard assumption of a thin EDL fails, so that a fully nonlinear numerical treatment is required.
Schematic illustration of the EDL structure around the negatively charged dielectric particle in an external electric field: (a) symmetric and nearly spherical EDL in a weak electric field; (b) deformed and asymmetric EDL in strong electric fields. The EDL is compressed in the direction of
$E_{\infty }$
and extended in the opposite direction. This distortion is caused by different electromigration and results in relaxation and retardation forces opposing the particle‘s motion.

Figure 6 schematically illustrates the deformation of the EDL during the electrophoresis of a positively charged dielectric particle of radius
$\tilde {a}$
. In figure 6(a), corresponding to a weak external electric field, the EDL remains nearly spherical and symmetric, with a uniform counterion distribution confined within the Debye layer of thickness
$\tilde {\lambda }_D$
. In figure 6(b), at higher electric field strength, the distribution of counterions becomes asymmetric due to the finite response time of the ionic cloud. The EDL is compressed on the upstream side of the particle and stretched in the opposite direction. This deformation results from the different electromigration of ions: cations are pushed forward along the field direction, while anions migrate in the opposite direction. As a result, the charge density
$\rho$
is redistributed around the particle, contributing to the relaxation force that opposes its motion. This schematic exemplifies the breakdown of spherical EDL symmetry in strong fields for a finite Debye length.
3.4. Steady-state structure of non-equilibrium ionic cloud
The schematic in figure 6 qualitatively illustrates how the EDL is deformed by the applied field and how a relaxation force arises. We now turn to the fully resolved numerical solution in order to quantify this non-equilibrium structure. After the transient has decayed and a steady asymmetric EDL has formed, we examine the corresponding distributions of ion concentrations and charge density around the particle, as well as the associated slip velocity along its surface.
When the EDL becomes asymmetric, with a predominance of counterions accumulating near the left (upstream) pole, the charge density
$\rho = C^+ - C^-$
near the particle surface
$r = 1$
reaches its maximum at the left pole and decreases gradually toward the right pole. This asymmetric ionic distribution is directly reflected in the slip velocity
$U_s$
(see figure 7). When variations along
$\theta$
are relatively slow compared with radial variations,
$U_s$
can be estimated using Smoluchowski formula, i.e.
which relates the tangential slip to the local zeta potential
$\zeta$
and the applied field
$E_{\infty }$
.
Tangential component of the velocity,
$U_{\theta }$
, as a function of angle
$\theta$
for
$\varepsilon = 0.166$
,
$E_{\infty } = 10$
,
$\sigma = -7$
,
$\varkappa = 0.05$
,
$\delta = 0.05$
, together with the Smoluchowski slip velocity given in (3.1). The numerical profile is given in the steady state at a distance
$7 \varepsilon$
from the particle surface, which is taken as an approximate location of the slipping plane in this case.

For constant
$\zeta$
, the Smoluchowski slip velocity
$U_s$
is symmetric with respect to the equator, since
$U_s(\pi /2 - \theta ) = U_s(\pi /2 + \theta )$
. In the nonlinear regime the numerical solution shows that both
$\rho$
and
$\zeta$
depend on angle
$\theta$
. In particular,
$\zeta (\theta )$
increases monotonically as
$\theta$
decreases from
$\pi$
to 0. The slip velocity deviates from the simple
$\sin \theta$
profile: the numerical tangential velocity
$U_{\theta }(\theta )$
in figure 7 has a higher magnitude than the Smoluchowski‘s prediction and its minimum is shifted at around
$\theta \approx 117^\circ$
compared with the symmetric reference curve. The numerical profile in figure 7 is shown at a distance
$7\varepsilon$
from the particle surface, which is taken as an approximate location of the slipping plane. This choice is based on the charge density distribution
$\rho$
along the line
$\theta = 180^\circ$
: at
$r-1\approx 7\varepsilon$
, the space charge has already decayed nearly to zero and the electrolyte becomes approximately quasineutral. This enhanced and non-uniform slip generates strong tangential shear along the surface and, in turn, drives a microvortex in the bulk at
$\theta \approx 45^{\circ }$
(see figure 5). This vortex redistributes the surrounding ionic charge and, in particular, generates a negatively charged region at this location at
$t = 0.25$
(see figure 5). A qualitatively similar vortex structure has been previously observed for an ion-selective granule and its formation mechanism was described in Frants et al. (Reference Frants, Artyukhov, Kireeva, Ganchenko and Demekhin2021).
The non-uniform slip distribution along the particle surface leads to a non-trivial pattern of ion and charge transport in the surrounding medium. Before discussing the resulting steady distributions of
$C^{\pm }$
and
$\rho$
, it is useful to examine the structure of the ion fluxes in the far field.
In the region far from the particle (
$r \gg 1$
), the influence of the surface charge
$\sigma$
becomes negligible, and the ion fluxes are entirely governed by advection with the electrophoretic velocity
$U_{\infty }$
and electromigration in the imposed electric field
$E_{\infty }$
. In this far-field region the liquid is electrically neutral,
$C^+ = C^- = 1$
. The normal components of ion fluxes across the spherical surface
$\varSigma _1$
enclosing the domain
$\varOmega$
(see figure 8) are given by
These expressions describe the normal to the surface fluxes of cations (
$J_R^+$
) and anions (
$J_R^-$
) through a spherical surface at a large distance (several particle radii) from the particle. They reveal that
$E_{\infty }$
enhances the cation flux and reduces the anion flux, resulting in an asymmetric ionic current in the far field, even though the liquid medium remains locally electroneutral there.
The volume region
$\varOmega$
is enclosed by two surfaces,
$\varSigma _1$
and
$\varSigma _2$
. The particle surface
$\varSigma _2$
is impenetrable for both ionic species, while the outer surface
$\varSigma _1$
is placed in the far field, where the electrolyte is electroneutral.

Both ion species enter the domain
$\varOmega$
(see figure 8) through the left hemisphere (
$90^\circ \lt \theta \lt 180^\circ$
) and leave it through the right hemisphere (
$0^\circ \lt \theta \lt 90^\circ$
), so that the radial fluxes are antisymmetric with respect to the equator. This net transport from left to right is accompanied by tangential fluxes, which are symmetric relative to the equator and given by
These expressions show that, in the far field, both cations and anions are advected by the electrophoretic flow while being simultaneously driven in opposite directions by the electric field, resulting in a charge-neutral but directionally biased ionic current.
As the ions approach the particle, they begin to experience the influence of the surface charge field
$\sigma$
. For the negatively charged surface, the behaviour of cations and anions diverges significantly. Cations are attracted to the negatively charged surface and can accumulate in its vicinity, while anions are repelled. Because the particle surface at
$r = 1$
is impermeable to both species, the ions must eventually reorient their motion from normal to tangential. As a result, two distinct charge-bearing layers form near the particle. The first is the counterion layer, the thickness of which is governed by the surface charge
$\sigma$
and that resembles a conventional EDL. The second is a broader charged layer with a weaker but non-zero net charge, whose origin and properties are discussed in detail below. The high cation concentration in the inner layer supports strong tangential cation flux. The non-uniformity of these fluxes induces significant normal ion transport, which, for high
$\sigma$
, can influence the electrokinetic behaviour even in the electroneutral bulk region.
The significant asymmetry between the cation and anion fluxes, evident from (3.2) and (3.3), highlights the fact that analysing only the charge density
$\rho = C^+ - C^-$
and the total salt concentration
$K = C^+ + C^-$
is not sufficient to fully capture the ion transport dynamics. While
$\rho$
characterises the net charge,
$K$
reflects the overall ionic strength (connected to electrical conductivity of electrolyte). Both are averaged over the contribution of individual species and, thus, hide the underlying directional imbalances of the cation and anion fluxes. In particular, the enhancement of the cation flux and simultaneous suppression of the anion flux by the
$E_{\infty }$
lead to a strongly anisotropic ionic transport, which can only be understood by examining the behaviour of
$C^+$
and
$C^-$
separately.
Figure 9 shows the distribution of
$C^+$
. The whole cation flow domain can be conveniently divided into three subregions, denoted as
$I_c$
,
$\textit{II}_c$
and
$\textit{III}_c$
in figure 9.
Space distribution of the cation concentration
$C^+$
for
$\varepsilon = 0.166$
,
$\sigma = -20$
,
$E_{\infty } = 46.7$
and
$\delta = 0.05$
along with corresponding cross-sections at
$\theta = 0^\circ$
,
$45^\circ$
,
$90^\circ$
,
$135^\circ$
and
$180^\circ$
, illustrating the variation in cation concentration near the particle surface.

In subregion
$I_c$
, which occupies the outer part of the domain away from the particle, the cation concentration remains equal to its bulk value,
$C^+ = 1$
. The influence of the surface charge field is negligible in this region, and the electrolyte is effectively electroneutral. Both the normal and tangential components of the cation flux reduce, to leading order, to the far-field expressions in relations (3.2) and (3.3) for the cations for the case of a negatively charged particle.
Inside subregion
$\textit{II}_c$
, the cation concentration increases sharply due to the influence of the surface charge. This region, of thickness
$O(\varepsilon )$
, resembles the classical EDL. The cation concentration reaches its maximum at the particle surface,
$C^+ = C^+_{\textit{max}}(\theta )$
, and then decays exponentially with the distance from the surface, approaching the bulk value
$C^+ = 1$
. The value
$C^{+}_{\textit{max}}(\theta )$
is not uniform along the surface and is largest near the downstream pole.
According to the equilibrium Gouy–Chapman theory, the maximum counterion concentration at the surface for these parameters is predicted to be
$C^+_{\textit{max}} = 411.3$
, independent of the polar angle
$\theta$
. However, our numerical results reveal a strongly non-uniform surface profile of
$C^+_{\textit{max}}(\theta )$
that deviates substantially from Gouy–Chapman theory. For the present parameters, we obtain
$C^+_{\textit{max}} = 87$
at
$\theta = 180^\circ$
,
$C^+_{\textit{max}} = 97.4$
at
$\theta = 135^\circ$
,
$C^+_{\textit{max}} = 184.2$
at
$\theta = 90^\circ$
,
$C^+_{\textit{max}} = 385.4$
at
$\theta = 45^\circ$
and
$C^+_{\textit{max}} = 456$
at
$\theta = 0^\circ$
(see cross-sections in figure 9). A notable exception is on the trailing edge of the particle, where the concentration is close to the Gouy–Chapman value and only slightly exceeds the equilibrium value. This deviation reflects the influence of electro-osmotic flow and non-equilibrium counterion accumulation behind the particle, effects that are absent in the Gouy–Chapman model.
The observed discrepancies are primarily due to the high Dukhin number in this system,
${\textit{Du}} = 3.69$
, indicating strong surface conduction and a non-equilibrium EDL. Unlike the assumptions of Gouy–Chapman theory, the counterions in the EDL here are not in equilibrium, but undergo significant tangential motion along the particle surface. This flux is directed counterclockwise (
$J_\theta ^+ \lt 0$
), as illustrated in figure 9 and imparts a pronounced non-equilibrium character to the EDL. As a result, the local ionic distributions deviate considerably from the predictions of the equilibrium theory. Importantly, all three transport mechanisms – diffusion, advection and electromigration – contribute comparably to the total ionic flux within the EDL. For comparison, the magnitude of the tangential cation flux
$J_\theta ^+$
inside the EDL reaches values of several thousand, while outside the EDL it drops to the order of tens. This sharp contrast underscores the dominance of surface transport phenomena and the inadequacy of equilibrium models in describing the ion dynamics under these conditions.
As cations in subregion
$\textit{II}_c$
are moved toward smaller angles
$\theta$
, they are continuously supplied from the outer region
$I_c$
, resulting in a net accumulation of cations within the EDL. Since the particle surface is impermeable, this accumulation must eventually be compensated by the release of excess cations into the surrounding bulk. This release occurs near the opposite pole of the particle (
$\theta = 0^\circ$
), where the concentrated outward flux of cations expells into the bulk, forming a local sink.
Indeed, at the angle
$\theta = \theta _1 \approx 45^\circ$
, the EDL detaches from the particle surface, and the accumulated cations are released into the bulk in the form of a narrow, focused jet, denoted as subregion
$\textit{III}_c$
in figure 9. Along the central axis of this jet (
$\theta = 0^\circ$
), the cation concentration decreases monotonically with distance from the particle: at
$r = 1$
one has
$C^+ \approx 456$
, at
$r = 1.2$
, it drops to
$C^+ = 46.3$
and by
$r = 1.5$
it is reduced to
$C^+ = 20.1$
.
The structure for anions is more intricate than for cations and can be divided into four distinct regions:
$I_a$
,
$\textit{II}_a$
,
$\textit{III}_a$
and
$\textit{IV}_a$
, as illustrated in figure 10. Importantly, these subregions do not necessarily coincide with the corresponding domains defined for cations due to the fundamentally different responses of positive and negative ions to the external electric field
$E_{\infty }$
and the surface charge
$\sigma$
.
(a) Space distribution of the anion concentration
$C^-$
for
$\varepsilon = 0.166$
,
$\sigma = -20$
,
$E_{\infty } = 46.7$
and
$\delta = 0.05$
along with cross-sections at
$\theta = 0^\circ$
,
$45^\circ$
,
$90^\circ$
,
$135^\circ$
and
$180^\circ$
, illustrating the variation in anion concentration near the particle surface.

Subregion
$I_a$
corresponds to the far-field region where the influence of the particle‘s surface charge
$\sigma$
is negligible. Here, the anion transport is fully governed by advection with the electrophoretic velocity
$U_{\infty }$
and electromigration induced by
$E_{\infty }$
. At sufficiently large radial distances (
$r \gg 1$
), the anion concentration approaches its bulk value,
$C^- = 1$
. The numerically computed normal component of the anion flux
$J_{r}^{-}$
across the outer spherical surface
$\varSigma _1$
(see figure 8) exhibits a sinusoidal angular dependence in agreement with the far-field approximation given by (3.2). Specifically, it is well described by the expression
$J_R^- = 11.6 \, \cos {\theta }$
. This flux profile is antisymmetric with respect to the equator (
$\theta = 90^\circ$
), reflecting the symmetric nature of the imposed field and flow (a direct consequence of the axial symmetry of the problem).
As the radius
$r$
of the domain
$\varOmega$
decreases and the region around the particle becomes smaller, anions begin to experience the repulsive influence of the particle‘s negative surface charge
$\sigma$
, which modifies both their concentration and trajectories. The corresponding transition zone is designated as subregion
$\textit{II}_a$
in figure 10. This zone is a layer of angularly dependent thickness
$\varDelta _1 (\theta )$
, within which the anion concentration decreases from its bulk value
$C^- = 1$
to
$C^- \approx 0$
. Anions entering region
$\textit{II}_a$
from the outer region
$I_a$
are deflected by the repulsive electrostatic field of the charged surface. As a result, their initially normal trajectories in the far field are diverted into tangential ones. The anion flux in this layer is directed counterclockwise along the surface of the particle, characterised by
$J^-_\theta \lt 0$
, and is oriented toward the right pole (see figure 10). For the parameters considered, this transition begins approximately at
$r \approx 3$
.
With a further decrease in the radius of the domain
$\varOmega$
to
$r = 1 + \varDelta _2$
(for the present parameters,
$\varDelta _2 \approx 1$
), we enter subregion
$\textit{III}_a$
, which is characterised by a strongly depleted anion concentration,
$C^- \approx 0$
, with thickness
$\varDelta _2(\theta )$
. This subregion occupies almost the entire spherical shell in the range
$1 \lt r \lt 1 + \varDelta _2$
within the angular sector
$\theta _1 \lt \theta \lt 180^\circ$
, where
$\theta _1 \approx \pm 45^\circ$
is a separation angle between regions
$\textit{III}_a$
and
$\textit{IV}_a$
. Due to the vanishing concentration, the anion flux within this region becomes negligible,
$\boldsymbol{J}^- \approx 0$
. Therefore, the entire tangential anion flux
$J^-_\theta \lt 0$
directed toward the right pole is completely formed by the movement of anions in subregion
$\textit{II}_a$
.
As anions move within the layer
$\varDelta _1(\theta )$
(subregion
$\textit{II}_a$
), their accumulation increases and the excess anions in the
$\theta = 0^\circ$
zone form an outward-directed jet, denoted as subregion
$\textit{IV}_a$
, analogous to the cation jet (subregion
$\textit{III}_c$
). This anion jet, with a cross-sectional width
$\varDelta _3(r)$
, emerges from the vicinity of the particle surface and generates a positive normal anion flux,
$J^-_R \gt 0$
.
The anion concentration along the jet axis exhibits a non-monotonic profile: starting from
$C^- = 5$
at the particle surface, it first increases to a local maximum
$C^- = 38$
at
$r = 1.1$
, then decreases to
$C^- = 19$
at
$r = 1.5$
and subsequently decays monotonically toward its bulk equilibrium value as
$r$
increases further.
Note that the charge density
$\rho$
magnitude in the upstream region (near
$\theta = 180^\circ$
) is significantly smaller than in the downstream region (near
$\theta = 0^\circ$
). The fact that the maximum of
$\rho$
is no longer tightly bound to the particle surface but is instead displaced and elongated into the bulk is a clear signature of partial EDL detachment.
To the best of our knowledge, there is no established terminology in the literature for situations where the
$\rho$
maximum is displaced away from the particle surface, while the near-surface Debye layer is still present. In the classical equilibrium layer picture (Poisson–Boltzmann/Gouy–Chapman models)
$\rho$
attains its maximum directly at the surface and then decays monotonically into the bulk. More developed non-equilibrium theories of a polarised EDL in a strong electric field likewise describe distortion, concentration polarisation and relaxation of the EDL, but do not, to our knowledge, introduce a distinct regime in which the principal
$\rho$
maximum is shifted several particle radii into the bulk while the immediate near-surface layer remains present. In the case of extended space charge near ion-selective interfaces, the region of enhanced charge also develops outside local equilibrium, yet it is still treated as an interfacial non-equilibrium layer adjacent to the ion-selective surface (Dukhin & Mishchuk Reference Dukhin and Mishchuk1993; Rubinstein & Zaltzman Reference Rubinstein and Zaltzman2010; Hardt & Baier Reference Hardt and Baier2011; Ganchenko et al. Reference Ganchenko, Frants, Shelistov, Nikitin, Amiroudine and Demekhin2019; Frants et al. Reference Frants, Amiroudine and Demekhin2024). In our case, by contrast, the maximum of
$\rho$
is shifted into the bulk by a distance of about two particle radii, i.e. it is located at
$r \approx 3$
, while a finite, though less pronounced, charged layer still persists in the immediate vicinity of the surface. This represents a qualitatively different configuration from the classical equilibrium picture and from the ion-selective surface.
We therefore use the term partial detachment of the EDL in a purely descriptive sense to denote this regime, in which the locus of maximum space charge is no longer bound to the particle surface with near-surface Debye layer. Full detachment, in the strict sense of a maximum space charge separated from the surface by an extended region of very low charge density, is not observed for the moderate Debye number considered here (
$\varepsilon = 0.166$
) as compared with the case of large Debye numbers. Our numerical study shows that the displacement of
$\rho$
increases systematically with
$\varepsilon$
: for smaller Debye numbers, the maximum remains close to the particle surface, while for larger Debye numbers, a fully detached space-charge region develops. Thus, the degree of (partial) detachment is controlled primarily by the Debye number, reflecting the competition between field-driven convection of ions and their diffusive relaxation back toward the surface. A rigorous theoretical criterion for the onset of such partial detachment has not yet been established and lies beyond the scope of the present work. Here we simply identify it empirically in terms of the location of the charge density maximum and its dependence on the Debye number.
(a) Space distribution of the charge density
$\rho$
for
$\varepsilon = 0.166$
,
$\sigma = -20$
,
$E_{\infty } = 46.7$
and
$\delta = 0.05$
together with cross-sections at
$\theta = 0^\circ$
,
$45^\circ$
,
$90^\circ$
,
$135^\circ$
and
$180^\circ$
, illustrating the variation of the charge density near the particle surface.

The characteristic thicknesses
$\varDelta _1$
,
$\varDelta _2$
and
$\varDelta _3$
for the present parameters are all of the order of unity, and the angular transition point is located at approximately
$\theta _1 \approx 45^\circ$
. However, both the thicknesses and the transition angle are sensitive to the system parameters and may vary substantially with changes in field strength, surface charge, Debye number or electrolyte properties.
Knowing the ion distributions
$C^-$
and
$C^+$
, one can readily obtain the charge density
$\rho = C^+ - C^-$
, as illustrated in figure 11. Two distinct charged regions with
$\rho \ne 0$
are clearly visible: the EDL of characteristic thickness
$O(\varepsilon )$
with high charge density
$\rho \gg 1$
and a broader outer charged region of thickness
$O(1)$
, where
$C^- \approx 0$
leading to
$\rho \approx C^+ \approx 1$
. This two-layer charge structure qualitatively resembles the configuration observed near ion-selective surfaces by Rubinstein & Shtilman (Reference Rubinstein and Shtilman1979), who demonstrated that in both limiting and overlimiting current regimes, a thin EDL (typically
$10 \rm\, nm$
thick) can coexist with a much thicker ESC layer reaching up to
$\sim 1 \rm \, \unicode{x03BC} m$
. The broader charged region in our simulations plays a role analogous to the ESC layer, although here it arises around a dielectric particle rather than a planar ion-selective surface.
The presence of both net charge and a tangential electric field within charged layers induces electro-osmotic flow in these regions. Since the net charge in the ESC region is greater than in the EDL, even though the local charge density is smaller, the electro-osmotic contribution associated with the ESC plays a dominant role in shaping the near-surface flow dynamics. The localised blue regions on the lateral sides of the particle (near
$\theta \approx \pm 45^\circ$
) correspond to negative values of
$\rho$
, where the anion concentration locally exceeds that of the cations. This local charge inversion arises from the field-induced distortion of the ionic atmosphere and highlights the strongly non-equilibrium character of ion transport in this regime, particularly the role of intense tangential ion fluxes in redistributing ions around the particle.
The ion motion within these charged layers gives rise to significant surface conductivity. This phenomenon was originally predicted by Dukhin and further developed theoretically in the works of Shilov et al. (Reference Shilov, Barany, Grosse and Shramko2003) and Schnitzer & Yariv (Reference Schnitzer and Yariv2014). In our case, the strong surface conduction is reflected in the large Dukhin number and the pronounced non-equilibrium deformation of the EDL and ESC layers described above. A qualitatively different picture of ion transport is observed near ion-selective particles, as discussed in earlier works (Ganchenko et al. Reference Ganchenko, Frants, Shelistov and Demekhin2018, Reference Ganchenko, Frants, Shelistov, Nikitin, Amiroudine and Demekhin2019).
The emergence of the cationic (
$\textit{III}_c$
) and anionic (
$\textit{IV}_a$
) jets in figures 9 and 10 leads to the formation of a net charged jet with
$\rho \ne 0$
in the bulk. At present, no analytical description is available for this jet corresponding to the strongly nonlinear regime.
3.5. Ionic-flux streamfunction
To understand how the EDL distortion and concentration polarisation are established, it is not sufficient to examine only the ion concentrations. The key quantity controlling the transient and steady redistribution of charge is the ionic flux: it reveals the transport pathways by which ions are delivered to (or removed from) different regions, and it allows us to identify near-surface transport routes, recirculation loops, stagnation zones and jet-like features. In what follows we therefore analyse the ionic-flux field, with particular emphasis on the anion flux
$\boldsymbol{J}^-$
, since for negatively charged particles it most clearly highlights the formation of depleted/enriched regions and the associated asymmetry responsible for the electrophoretic response.
The ionic fluxes are decomposed into radial and polar (tangential) components:
Each component includes contributions from convection, diffusion and electromigration. In non-dimensional variables,
In steady or quasisteady conditions, the ionic-flux field is divergence free in the electrolyte bulk (no volumetric sources/sinks):
This property motivates a
$J$
-flux streamfunction representation, which provides a compact visualisation of the full two-component flux field. In an axisymmetric geometry the relevant geometric factor is
$r^{2}\sin \theta$
, and we define the
$J$
-flux streamfunction
$\varPsi _{J}^{\pm }(r,\theta )$
such that the flux lines coincide with the level sets
$\varPsi _{J}^{\pm } = \mathrm{const}$
and
With this definition,
$\boldsymbol{\nabla }\boldsymbol{\cdot }\boldsymbol{J}^{\pm }=0$
is satisfied at steady state up to numerical errors, while
$\varPsi _{J}^{\pm }$
directly exposes the global topology of transport: closed contours correspond to recirculating transport loops and open contours indicate net transport between the particle and the far field.
Although computations are performed in
$(r,\theta )$
, all figures in this section are presented in the physical
$(x,y)$
plane using the standard mapping
$x = r\cos \theta , \, y = r\sin \theta$
applied to the
$J$
-flux streamfunction grid. We use two complementary visualisation modes. Contour lines of
$\varPsi _{J}^{\pm }(x,y)$
represent meridional streamlines of the ionic flux. They clearly show the topology and relative intensity of recirculation regions, jet-like transport and near-surface pathways. To visualise the local direction of transport, we transform spherical components into Cartesian components, i.e.
and plot
$(J^{\pm }_{x}, J^{\pm }_{y})$
as a vector field. Since
$\boldsymbol{J}^{\pm }$
may vary by orders of magnitude (especially near the particle), arrows are rescaled for readability; in some plots we additionally normalise vectors to emphasise direction.
Figure 12 visualises the steady-state anion flux
$\boldsymbol{J}^-$
as a vector field and streamfunction. The flux pattern is strongly inhomogeneous and can be understood by comparing it with the anion concentration map (see figure 10). In particular, the apparent ‘empty’ regions (almost no streamlines) and the closed streamline patterns directly reflect the formation of depleted and enriched zones of
$C^-$
.
Vector field and streamfunction of the steady-state solution for anion flux
$J^-$
around the particle for
$E_{\infty } = 46.7$
,
$\sigma = -20$
and
$\varepsilon = 0.166$
.

In region
$I_a$
(far-field transport), where
$C^- \approx 1$
, the flux field is dominated by a smooth, open transport around the particle: streamlines are open and connect the far field upstream to downstream. This region corresponds to nearly uniform concentration and, therefore, to a relatively regular balance between convection and electromigration.
Regions
$\textit{II}_a$
–
$\textit{III}_a$
are the depleted ‘dead’ zone. On the upstream side of the particle, a broad depleted region develops. In this zone the anion availability is low, so both the convective contribution (
$\sim U\,C^-$
) and the electromigration contribution (
$\sim C^- \boldsymbol{\nabla }\varPhi$
) are suppressed. Moreover, inside the depleted plateau the gradients of
$C^-$
are relatively small, so diffusion does not provide a compensating transport. As a result, the magnitude of
$\boldsymbol{J}^-$
becomes very small, the
$J^-$
-flux streamfunction varies weakly and the contour lines become sparse or visually disappear. This explains the ‘empty’ region in figure 12 corresponding to a depletion induced almost zero transport (quasi-stagnant) zone.
In region
$\textit{IV}_a$
downstream of the particle, the concentration map exhibits a pronounced enriched jet of anions. The high
$C^-$
concentration supports a large anion flux and the streamfunction contours become dense and aligned with the jet direction, indicating a focused transport channel that carries anions away from the particle. The jet boundaries coincide with sharp concentration gradients. Here the diffusion and electromigration cooperate to sustain the strong, directed flux.
A compact region of closed streamlines appears near the downstream side close to the particle surface, exactly where the concentration field transitions from the depleted domain (
$\textit{II}_a$
–
$\textit{III}_a$
) to the enriched jet (IV
$_a$
). The strong lateral concentration gradients at the jet root drive a local transverse flux, while the surrounding transport organises into closed loops. Physically, these loops indicate that part of the anion transport is recirculating locally (exchange between near-surface layers and the jet periphery) rather than contributing to net far-field transport. The separatrix between open streamlines (net transport) and closed contours (recirculation) is therefore controlled by the depletion/enrichment structure of
$C^-$
.
Vector field and streamfunction of the steady-state solution for cation flux
$J^+$
around the particle for
$E_{\infty } = 46.7$
,
$\sigma = -20$
and
$\varepsilon = 0.166$
.

Figure 13 shows the steady cation flux
$\boldsymbol{J}^+$
streamfunction contours and flux vectors. Its structure is qualitatively different from the anion because the particle surface is negatively charged. As a result, cations are attracted toward the surface and a strong near-surface transport layer forms, while the bulk concentration remains close to uniform (see figure 9 for
$C^+$
map).
In region
$I_c$
far from the particle, where
$C^+ \approx 1$
, the cation flux is a smooth, open field that bypasses the particle. Streamlines are open and deflected around the particle, indicating that the far-field transport is only weakly perturbed.
In contrast to anions, the
$C^+$
distribution does not develop a depleted region. Since cations are electrostatically attracted to the negatively charged surface,
$C^+$
near the particle is maintained at
$O(1)$
, so the convective and electromigration contributions to
$\boldsymbol{J}^+$
are not suppressed. The cation transport remains active everywhere in the vicinity of the particle.
In region
$\textit{II}_c$
on the upstream side, the concentration map indicates only a mild redistribution of
$C^+$
, yet the flux vectors close to the surface directed predominantly toward the particle. This reflects the dominant role of electromigration in drawing cations toward the negatively charged surface. The streamlines near the surface therefore represent the behaviour that supply cations into the near-surface layer, rather than a weak-flux depletion region.
In region
$\textit{III}_c$
downstream of the particle, the
$C^+$
field exhibits an enriched jet, which acts as a high transport zone: streamlines become more concentrated and align with the downstream direction, indicating an export of cations along the jet. The formation of this zone is controlled by the near-surface uptake of cations and their subsequent downstream advection and redistribution.
For
$\boldsymbol{J}^+$
, closed streamline loops are absent compared with the anion case. Physically, the strong attractive electromigration toward the negatively charged surface organises the transport into predominantly open pathways, rather than into pronounced recirculating flux loops.
In summary, the
$C^+$
map explains the topology of
$\boldsymbol{J}^+$
. Because the particle surface is negatively charged, cations do not form a broad depleted zone. Instead, cation transport is dominated by near-surface attraction (surface conductivity) together with a downstream enriched wake that acts as the main export pathway.
3.6. High surface charge at low-to-moderate fields
The analysis above demonstrates how at intermediate Debye numbers, the ionic cloud and corresponding fluxes around the particle show strong non-equilibrium behaviour. The next step is to identify the conditions at which these nonlinear features become most pronounced. In this section we therefore focus on the regime of high surface charge density
$\sigma$
(equivalently, high
$\zeta$
potential) at low-to-moderate electric fields, which is of particular interest because it maximises the role of surface conduction while remaining experimentally accessible.
Moreover, such conditions are experimentally relevant for colloidal systems in non-polar solvents and for particles with strongly charged or functionalised surfaces. This regime is typically associated with enhanced surface conductivity (
${\textit{Du}} \sim O(1)$
), leading to a non-trivial dependence of electrophoretic mobility on the applied field. As shown by our results, the ionic cloud behaviour becomes qualitatively different, and effects such as EDL deformation and partial detachment, relaxation-induced retardation and mobility saturation can be observed – none of which are captured by classical theories.
For high surface charge and, as a consequence, for significant Dukhin numbers, more accurate prediction is obtained using the analytical formula
originally proposed by O’Brien & Hunter (Reference O’Brien and Hunter1981) and later rigorously justified in the limit
$E_{\infty } \to 0$
by Schnitzer & Yariv (Reference Schnitzer and Yariv2012b
), Schnitzer et al. (Reference Schnitzer, Zeyde, Yavneh and Yariv2013). Note that the O’Brien–Hunter relation (3.10) is also a linear (weak-field) mobility formula. In this sense it is an analogue of the classical Smoluchowski expression, but extended to the regime of highly charged surfaces: it explicitly incorporates the Dukhin number, thereby accounting for significant surface conduction, which is neglected in the Smoluchowski formula.
In the high
$\sigma$
regime, strong surface conduction outside the thin EDL limit can substantially reshape the ionic cloud beyond the simple dipolar pattern predicted by linear asymptotics. We now turn to a moderate electric field, with nonlinear deformation of the EDL, and examine how the structure of the anion cloud changes as
$\sigma$
is increased.
Figure 14 shows the two-dimensional anion concentration field
$C^-$
around the particle for several values of
$\sigma$
at a fixed electric field
$E_{\infty } = 46.7$
. As
$\lvert \sigma \rvert$
increases, the anion cloud becomes more compressed, leading to a more pronounced depletion region at
$\theta = 180^\circ$
and enriched wake downstream at
$\theta = 0^\circ$
. At a lower surface charge (
$\sigma = -7$
), the distortion of
$C^-$
is significant. A slightly elongated depleted zone forms ahead of the particle, whereas an enriched anion wake develops behind it. This concentration pattern is typical for the case of mobility
$M = U_{\infty } / E_{\infty } \lt 1$
. However, due to the moderate Debye number
$\varepsilon$
, the spatial extent of both depleted and enriched regions is limited to approximately two particle radii.
Spatial distributions of the anion concentration
$C^-$
around a dielectric particle for three values of surface charge density: (a)
$\sigma = -7$
, (b)
$\sigma = -20$
, and (c)
$\sigma = -140$
, at a fixed Debye number
$\varepsilon = 0.166$
and electric field strength
$E_{\infty } = 46.7$
.

As the surface charge increases to
$\sigma = -20$
, the electrophoretic mobility exceeds unity, and the distribution of
$C^-$
becomes qualitatively similar to the regime analysed earlier for
$E_{\infty } = 46.7$
and
$\varepsilon = 0.166$
(see figure 10). A further increase in
$\lvert \sigma \rvert$
up to
$\sigma =-140$
leads to only minor qualitative changes in the spatial structure of
$C^-$
. This indicates the onset of a saturation regime, where the anion cloud structure is no longer strongly affected by further increases in
$\lvert \sigma \rvert$
.
At weak and moderate electric fields (
$E_{\infty } \ll 1$
and
$E_{\infty } = O(1)$
), and for a wide range of surface charge magnitudes
$\lvert \sigma \rvert$
, the numerical results can be meaningfully compared with available asymptotic theories. In this regime, the main reference frameworks are the classical analysis of O’Brien & Hunter (Reference O’Brien and Hunter1981), the theory developed by Shilov et al. (Reference Shilov, Barany, Grosse and Shramko2003) and the more recent asymptotic studies by Schnitzer & Yariv (Reference Schnitzer and Yariv2012b
), Schnitzer et al. (Reference Schnitzer, Zeyde, Yavneh and Yariv2013).
The solution derived by O’Brien & Hunter (Reference O’Brien and Hunter1981) corresponds to the leading-order term of the weak-field asymptotic expansion (
$E_{\infty } \ll 1$
), valid for high surface conductance. It provides a generalisation of Smoluchowski‘s formula to systems with large Dukhin numbers. According to O’Brien & Hunter (Reference O’Brien and Hunter1981), the electrophoretic mobility in the linear approximation with respect to
$E_{\infty }$
is given by (3.10).
The analytical solution by O’Brien & Hunter (Reference O’Brien and Hunter1981) agrees with our numerical results with graphical accuracy for
$E_{\infty } = 0.1$
. However, as the field strength increases (
$E_{\infty } \gt 1$
), the weak-field approximation of O’Brien and Hunter begins to underestimate the particle mobility. In particular, at
$E_{\infty } = 5$
, (3.10) predicts a particle velocity nearly 4.5 times lower than the numerical value.
Electrophoretic mobility versus surface charge
$\mid \sigma \mid$
at intermediate Debye number
$\varepsilon = 0.166$
. Results are shown for (a)
$E_{\infty } = 1$
and
$5$
, (b)
$E_{\infty } = 10$
.

Figure 15. Long description
Two line graphs depict the relationship between electrophoretic mobility and surface charge at intermediate Debye numbers. Panel A: The graph shows the ratio of U subscript infinity to E subscript infinity versus sigma. The x-axis represents sigma, ranging from 0 to 100. The y-axis represents the ratio U subscript infinity to E subscript infinity, ranging from 0 to 0.4. The graph includes data points for numerical results with E subscript infinity equal to 5 and 1, as well as theoretical models by Shilov and Yariv for the same E subscript infinity values. Panel B: The graph shows the ratio of U subscript infinity to E subscript infinity versus sigma. The x-axis represents sigma, ranging from 0 to 140. The y-axis represents the ratio U subscript infinity to E subscript infinity, ranging from 0 to 1.2. The graph includes data points for numerical results with E subscript infinity equal to 10, as well as theoretical models by Shilov, Schnitzer, and Pade for the same E subscript infinity value.
To extend the validity of the theory to the moderate-field regime, it is necessary to include higher-order terms in the electric field expansion. This refinement was undertaken independently by Shilov et al. (Reference Shilov, Barany, Grosse and Shramko2003) and Schnitzer et al. (Reference Schnitzer, Zeyde, Yavneh and Yariv2013), yielding the following weakly nonlinear expression for the electrophoretic mobility:
Here the leading-order coefficient
$a_0$
is the linear contribution originally obtained by O’Brien & Hunter (Reference O’Brien and Hunter1981) in the form
The first-order term vanishes due to symmetry, so the nonlinear correction to the mobility appears at
$O(E_{\infty }^2)$
and is represented in (3.11) through the coefficient
$a_2$
. The expressions for
$a_2$
obtained by Shilov et al. (Reference Shilov, Barany, Grosse and Shramko2003) and by Schnitzer et al. (Reference Schnitzer, Zeyde, Yavneh and Yariv2013) differ. In the formulation of Schnitzer et al. (Reference Schnitzer, Zeyde, Yavneh and Yariv2013),
$a_2$
is given by
where the coefficients
$k_i$
are polynomials that depend on
$\zeta _0$
and
$\varkappa$
(see Schnitzer et al. Reference Schnitzer, Zeyde, Yavneh and Yariv2013 for their explicit forms). In contrast, the approach by Shilov et al. (Reference Shilov, Barany, Grosse and Shramko2003) yields a more compact expression:
These corrected expressions given by Shilov et al. (Reference Shilov, Barany, Grosse and Shramko2003) and Schnitzer et al. (Reference Schnitzer, Zeyde, Yavneh and Yariv2013) are also shown in figure 15(a) for
$E_{\infty } = 1$
and
$E_{\infty } = 5$
, and both show excellent agreement with our numerical simulations. However, at higher fields (
$E_{\infty } = 10$
), these formulas begin to overestimate the mobility significantly. This is expected, since (3.11) is based on a truncated Taylor expansion in
$E_{\infty }$
: such an expansion is accurate only near
$E_{\infty }=0$
and its accuracy deteriorates as
$E_{\infty }$
increases. To improve the approximation beyond the weakly nonlinear regime, we replace the Taylor form in (3.11) by a Padé-type rational approximation, i.e.
where the coefficient
$a_0$
is taken in the form of (3.12) and coefficients
$a_2$
and
$b_2$
in the form of (3.14), but with unknown coefficients
$\alpha _1$
,
$\alpha _2$
,
$\alpha _3$
,
$\beta _1$
,
$\beta _2$
,
$\beta _3$
:
Here
${\textit{Du}} = -\varepsilon \sigma (1+2\varkappa ) \gt 0$
. The coefficient
$a_0$
in (3.15) was fixed by the weak-field analytical result in (3.12) from O’Brien & Hunter (Reference O’Brien and Hunter1981) and was not included among the fitting parameters. The coefficients
$a_2$
and
$b_2$
were represented by the quadratic functions of
${\textit{Du}}$
introduced in (3.16). The six constants
$\alpha _1$
,
$\alpha _2$
,
$\alpha _3$
,
$\beta _1$
,
$\beta _2$
and
$\beta _3$
were then determined simultaneously by a least squares fit of (3.15) to the numerical values of
$U_\infty /E_\infty$
obtained at
$E_\infty = 5$
and
$E_\infty = 10$
over the considered range of
$\sigma$
. For every numerical data point, the corresponding Dukhin number was evaluated as
${\textit{Du}} = -\varepsilon \sigma (1+2\kappa )$
, and a single common set of fitting coefficients was used for the entire data set. Substitution of the fitted constants into (3.16) yields (3.17) and (3.18).
For all field strengths considered, the mobility increases monotonically with
$\lvert \sigma \rvert$
and then approaches a saturation plateau at large
$\lvert \sigma \rvert$
. Notably, for
$E_{\infty } = 46.7$
, the mobility exceeds unity at
$\sigma \approx 14.85$
.
Based on the computed mobility curves, we therefore propose the following coefficients
$a_2$
and
$b_2$
for Padé-type approximation:
The Padé-type approximation in (3.15)–(3.18) is empirical, but not asymptotically exact at arbitrary field strength. The coefficients
$a_2$
and
$b_2$
were determined by fitting the numerical results at
$E_\infty = 5$
and
$E_\infty = 10$
over the considered range of
$\sigma$
. Comparison with additional numerical data indicates that the approximation remains quantitatively reliable up to approximately
$E_\infty = 15$
for the present values of
$\varepsilon$
and
$\kappa$
. For stronger fields, it should be interpreted only as a qualitative extrapolation, and its accuracy is not guaranteed.
3.6.1. Regime of high surface charge and high electric field
With increasing electric field strength
$E_{\infty }$
, the role of advection becomes increasingly important, leading to the breakdown of the symmetry between depleted and enriched regions. In the range
$E_{\infty } = 10{-}50$
the upstream depletion zone is compressed into a boundary-layer-like structure shown in figure 16(a) similar to subregion
$\textit{III}_a$
in figure 10, while the enriched region transforms into a narrow anionic jet analogous to subregion
$\textit{IV}_a$
.
Spatial distribution of the cation concentration
$C^+$
(a,b,c), anion concentration
$C^-$
(d,e,f) and charge density
$\rho$
(g,h,i) at intermediate Debye number
$\varepsilon = 0.166$
and high surface charge
$\sigma = - 20$
. Left column for
$E_{\infty } = 46.7$
, middle column for
$E_{\infty } = 93.4$
and right column for
$E_{\infty } = 140$
.

At higher field strengths (see figure 16
d–f), the anion concentration field
$C^-$
undergoes a pronounced qualitative restructuring. As
$E_{\infty }$
increases, the downstream enriched anion jet progressively weakens, while the depleted region on the upstream side expands and penetrates farther into the bulk. In the field range shown in figure 16(d–f), these two tendencies coexist: the enriched wake is still present, but it becomes increasingly compressed and less dominant, whereas the depleted zone broadens and begins to control the overall large-scale topology of the anion cloud. If the electric field strength is increased further, this tendency continues: at sufficiently large
$E_{\infty }$
(
$E_{\infty } \approx 250$
, not shown here), the enriched anion zone disappears altogether and is replaced by a fully depleted downstream region. This gradual reorganisation of the anion cloud suggests that the transition is dynamically important for the electrophoretic response. As will be shown below, the field interval in which the anion wake is being restructured corresponds to a noticeable change in the behaviour of the mobility curve.
Interestingly, this pronounced restructuring of the anion cloud is not accompanied by an equally strong change in the cation distribution. As seen in figure 16(a–c), the cation concentration field
$C^+$
preserves its overall topology as the electric field increases: the main effect is a gradual narrowing and sharpening of the downstream concentration jet, while the large-scale structure remains essentially unchanged.
As the electric field increases, the charge density field
$\rho$
in figure 16(g–i) also undergoes a noticeable restructuring. In particular, the localised regions of negative charge on the lateral sides of the particle progressively weaken and eventually disappear, while the positive charge region extends into the downstream wake. This evolution is consistent with the strong-field reorganisation of the ionic cloud discussed above and reflects the growing asymmetry of the charge distribution.
Vector field and
$J^-$
streamfunction around the particle for
$\sigma = -20$
and
$\varepsilon = 0.166$
at (a)
$E_{\infty } = 46.7$
, (b)
$E_{\infty } = 93.4$
, and (c)
$E_{\infty } = 140$
.

Since the most pronounced qualitative changes occur in the anion concentration field, it is natural to examine the anion flux
$\boldsymbol{J}^-$
in order to understand how this restructuring is formed dynamically. In particular, while the concentration plots reveal the evolving depleted and enriched regions, they do not by themselves show the transport pathways responsible for this reorganisation.
Figure 17 shows how the restructuring is reflected in the transport topology of the anion flux
$\boldsymbol{J}^-$
. As the electric field increases, the localised recirculating transport on the downstream side of the particle is progressively suppressed, and the flux field becomes increasingly dominated by open streamlines extending around and away from the particle.
At higher fields, however, the mobility does not approach a plateau, in contrast to the behaviour reported by Stotz (Reference Stotz1978). This difference is attributable to the intermediate (not large) Debye number considered here: the ionic cloud remains attached to the particle and does not detach into the bulk. Consequently, the assumptions underlying the Stotz scenario are not satisfied in the present case.
Electrophoretic mobility
$U_{\infty }/E_{\infty }$
as a function of electric field strength
$E_{\infty }$
for intermediate Debye number
$\varepsilon = 0.166$
and surface charge density
$\sigma = -20$
.

Figure 18 shows the electrophoretic mobility as a function of the applied field strength. The mobility increases monotonically and exceeds unity at approximately
$E_{\infty } = 11.5$
. In the weak-field regime, the dependence is nearly linear. At higher fields, however, the curve does not approach a true saturation plateau. Instead, it exhibits only a weak plateau in the approximate range
$E_{\infty }\approx 50$
–
$90$
, after which the mobility continues to increase. This short plateau occurs precisely in the field interval where the anion flux topology undergoes a qualitative restructuring. Physically, part of the field-induced increase is spent on reorganising the ion transport pattern itself, rather than directly amplifying the electrophoretic response, which leads to a temporary slowing down of the mobility growth.
3.7. Comparison with Stotz‘s experiments
In the experiments of Stotz (Reference Stotz1978), two classes of particles were examined, with relatively high and relatively low surface charge. For particles with a lower surface charge (corresponding to
$\sigma = -20$
in our dimensionless units), only a partial detachment of the EDL occurs, resulting in a field-dependent electrophoretic mobility, as illustrated in figure 19. In this figure, the experimental mobility curve (solid line) lies below the value predicted by Hückel‘s formula. This deviation can be attributed to the intermediate Debye numbers in the experiment, specifically
$\varepsilon = 0.166$
and
$\varepsilon = 0.062$
for 12 and 32
$\unicode{x03BC}$
m particles, respectively. Under such conditions, the retardation effect due to the EDL becomes significant and reduces the particle‘s mobility – an effect not accounted for in Hückel‘s theory.
Comparison of the dimensional electrophoretic mobility
$\tilde {M}$
with the experimental results of Stotz (Reference Stotz1978) for a particle of radius
$\tilde {a}=12\,\unicode{x03BC} \mathrm{m}$
and Debye length
$\tilde {\lambda }_D = 2\,\unicode{x03BC} \mathrm{m}$
, corresponding to
$\varepsilon = 0.166$
. The solid curve shows the experimental measurements of Stotz (Reference Stotz1978) and the symbols represent the numerical results. The simulation parameters are
$\varepsilon =0.166$
,
$\sigma =-20$
,
$\delta =0.05$
and
$\kappa =0.05$
.

For the particle sizes and Debye lengths used by Stotz, the system lies in the intermediate-Debye-number range, where the EDL induced retardation force and the relaxation force are both non-negligible. These two contributions reduce the net driving force and, therefore, shift the mobility below the Hückel prediction.
At the same time, the partial EDL detachment implies that advection significantly distorts the ionic cloud, weakening (but not eliminating) its hydrodynamic and electrostatic coupling to the particle. In terms of forces, this means that the balance between Coulomb force, electrophoretic retardation and relaxation force evolves with
$E_{\infty }$
, which naturally leads to a field-dependent mobility.
As the EDL becomes thicker, the system moves away from the Smoluchowski limit and the electric force acting on the diffuse charge is distributed over a larger volume farther from the particle surface. In this situation, a larger fraction of the electric input is transferred to the surrounding fluid through electrohydrodynamic coupling, which enhances the retardation mechanism and modifies the balance between Coulomb force, electrophoretic retardation and relaxation force. At sufficiently strong fields, the resulting advection and EDL polarisation can drive a qualitative reorganisation of the ionic cloud. In particular, the diffuse charged region may become strongly displaced and partially detached from the particle, leading to a new transport regime in which the hydrodynamic and electrostatic coupling to the particle is weakened but not fully lost.
An additional key factor is the smallness of the parameter
$\varkappa$
, which controls the strength of electrohydrodynamic coupling in the dimensionless momentum equation. For the non-polar systems considered here,
$\varkappa \ll 1$
so that the feedback of the electric body force on the flow is relatively weak. As a result, once the ionic cloud is strongly deformed by advection, its ability to generate strong electrohydrodynamic back coupling (and, hence, a strong retardation/relaxation response) is reduced. This weak-coupling property facilitates partial detachment and the associated field-dependent mobility regime observed in the Stotz experiments.
By contrast, in aqueous electrolytes the corresponding coupling is typically much stronger (effectively larger
$\varkappa$
in the present scaling), so the electrokinetic and hydrodynamic fields remain more tightly coupled. In this case, the ionic cloud transmits more efficiently the electric forcing to the surrounding liquid, and the partial detachment regime described here is not observed. Our numerical results reproduce the non-polar solvent behaviour, since the coupled NPPS model resolves both EDL deformation and the associated relaxation effects beyond the assumptions of classical linear theories.
Our numerical results demonstrate good agreement with the experimental data in the regime of stronger electric fields. The insets in figure 19 illustrate the accumulation of substantial charge near the particle surface across various field strengths and angular positions. Notably, a pronounced charge build-up is observed on the front side of the particle, extending up to approximately one particle radius into the surrounding medium.
Importantly, Stotz reported that the conductivity of the solution remained constant over the range of electric field strengths used in the experiments. On this basis, an increase in particle charge caused by additional field-induced dissociation was neglected (Stotz Reference Stotz1978). The experimental conditions considered in that work are therefore consistent, at least to leading order, with the present assumption of a fixed amount of mobile charge carriers and field-independent transport coefficients.
In summary, the Stotz experiments considered in this section are consistent with the electrophoretic regime studied in the present work, namely, the intermediate-Debye-number regime in which the diffuse ionic cloud remains only partially detached from the particle. Under these conditions, the EDL is strongly deformed by the applied field, but its hydrodynamic and electrostatic coupling to the particle is not lost completely. As a result, the electrophoretic mobility remains field dependent and stays below the Hückel prediction, in agreement with both the experimental results and our numerical results. This behaviour is naturally interpreted within the force-balance framework discussed above: at intermediate Debye numbers, the retardation and relaxation contributions remain significant and evolve with the field strength due to EDL polarisation and advection.
4. Conclusions
In this work we investigated nonlinear electrophoresis of a charged particle in a polar medium at an intermediate Debye number, with particular focus on regimes where classical linear theories become inaccurate. By solving the coupled NPPS system, we resolved the full ionic and hydrodynamic structure of the problem, including EDL deformation, concentration polarisation, ionic-flux topology and resulting changes in electrophoretic mobility.
The results show that the intermediate-Debye-number regime is governed by a strong interplay of advection, electromigration and diffusion, which produces pronounced asymmetry of the ionic cloud. In particular, the anion field develops depleted and enriched regions with distinct transport pathways, including high concentration jets. As the electric field increases, this structure undergoes a qualitative reorganisation: the enriched anion jet disappears and is replaced by an extended depleted region, while the EDL remains attached to the particle surface. This identifies a strongly nonlinear regime that is not described by existing weak-field asymptotic theories.
We also analysed the mobility dependence on surface charge and field strength. For a fixed intermediate Debye number, the mobility increases monotonically with
$\lvert \sigma \rvert$
and then saturates, while as a function of
$E_{\infty }$
it remains field dependent and exceeds unity in the strong-field regime. Classical and weakly nonlinear asymptotic formulas (O’Brien–Hunter, Shilov, Schnitzer) provide accurate predictions only at weak and moderately fields. To extend analytical predictability, we introduced a Padé-type approximation, which significantly improves the agreement with our numerical results and more than doubles the practical range of validity of the asymptotic mobility formula.
Finally, the comparison with the experiments of Stotz shows good qualitative and quantitative agreement in the regime of intermediate Debye numbers and strong fields. The observed field-dependent mobility is consistent with a partially detached ionic cloud. Within the force-balance interpretation, this corresponds to a regime where both electrophoretic retardation and relaxation remain significant and evolve with field strength. These findings clarify the physical mechanisms of nonlinear electrophoresis at intermediate Debye numbers and provide a framework for modelling electrophoretic transport in nonclassical regimes beyond the scope of standard linear theories.
The classical Poisson–Nernst–Planck (PNP) model is not expected to remain quantitatively valid at arbitrarily large electric fields. In particular, the point ion and ideal dilute solution assumptions may lead to unrealistically large counterion concentrations in regions of strong electrostatic accumulation. At sufficiently high local potentials, finite ion size effects, ion–ion correlations, dielectric saturation and concentration-dependent ionic transport may become important and limit the growth of the local charge density. Modified PNP models incorporating steric constraints generally predict saturation of the counterion concentration and a reduced nonlinear electrokinetic response at large voltages. In non-polar systems, field-dependent dissociation and recombination of charge carriers may introduce an additional departure from the present fully dissociated model. Consequently, the highest field results reported here should be viewed as predictions within the classical PNP framework; inclusion of these additional effects would most likely attenuate the calculated high field increase in electrophoretic mobility. On the other hand, the reasonably good agreement obtained with the experimental mobility measurements of Stotz suggests that, despite these limitations and the relatively high applied field strengths, the present model still captures the dominant mechanisms of the electrophoretic response and provides quantitatively reasonable predictions over the experimentally relevant parameter range considered here.
A natural continuation of the present study is to consider the same strong field and high surface charge regime in a polar medium, but at large Debye numbers rather than intermediate ones. In this case, the ionic cloud is expected to be much more weakly bounded to the particle. The electrohydrodynamic response may become qualitatively different, including stronger cloud displacement, possible large scale detachment and a different mobility scaling at high fields. Such a regime lies beyond the scope of the present paper and requires a separate analysis, since the mechanisms governing ion transport, force balance and a nonlinear electrophoretic response are no longer the same as those identified here for an attached or only partially detached ionic atmosphere.
As for the model used in the present study, several extensions would be valuable in future work. In particular, unequal diffusivities of the positive and negative charge carriers,
$D^+ \neq D^-$
, should be considered in order to quantify the additional transport asymmetry and diffusion effects that are excluded by the present symmetric approximation. It would also be important to incorporate incomplete dissociation by introducing a neutral ion-pair species together with association and dissociation source terms in the Nernst–Planck equations, especially when the population of free carriers becomes field dependent. These extensions would allow the model to address a broader class of weakly conducting systems and to assess how reaction kinetics modify concentration polarisation and nonlinear electrophoretic mobility. At the same time, the reasonably good agreement obtained here with the experimental measurements of Stotz indicates that, despite these simplifications, the present NPPS formulation already captures the dominant mechanisms of the electrophoretic response over the parameter range considered.
Future work may extend the present formulation to weakly deformed particles by combining a perturbative description of the particle shape with the fully nonlinear NPPS model (Ganguly & Gupta Reference Ganguly and Gupta2026). This would allow the influence of small shape deviations on nonlinear electrophoretic mobility to be quantified.
Supplementary material
Supplementary material is available at https://doi.org/10.1017/jfm.2026.12033.
Funding
The present work was supported by RSF (Russian Science Foundation) under Grant No. 24-79-00129.
Declaration of interests
The authors report no conflict of interest.
Appendix A. Force balance
For a particle subjected to a time-dependent electric field, the exact dimensional force balance must include the particle inertia,
rather than the quasisteady force-free condition (2.11)
Moreover, since the governing equations are formulated in a reference frame translating with the particle, this frame is non-inertial whenever
${\rm d}\tilde U_{\infty }/{\rm d} \tilde t \neq 0$
. The corresponding dimensional momentum equation therefore contains the frame acceleration contribution
where it is written as a fictitious body force on the right-hand side. Here
$\rho _m$
is the density of the surrounding liquid medium.
The present formulation employs a quasisteady momentum approximation. To assess its validity a posteriori, the complete local and convective acceleration terms were restored in the axisymmetric momentum equations and evaluated using the computed transient solution.
The radial momentum equation was written in the form
\begin{align} - \frac {\partial P}{\partial r} + \frac {\partial ^2 U_r}{\partial r^2} + \frac {2}{r} \frac {\partial U_r}{\partial r} + \frac {1}{r^2} \frac {\partial ^2 U_r}{\partial \theta ^2} + \frac {\cot \theta }{r^2}\frac {\partial U_r}{\partial \theta } - \frac {2U_r}{r^2} - \frac {2U_\theta \cot \theta }{r^2} - \frac {2}{r^2}\frac {\partial U_\theta }{\partial \theta } \nonumber \\[3pt] = \frac {\varkappa }{\varepsilon ^2}(C^+-C^-) \frac {\partial \varPhi }{\partial r} + \varepsilon _D \left ( \frac {\partial U_r}{\partial t} + U_r \frac {\partial U_r}{\partial r} + \frac {U_\theta }{r}\frac {\partial U_r}{\partial \theta } - \frac {U_\theta ^2}{r} \right )\!, \end{align}
whereas the polar component was written as
\begin{align} -\frac {1}{r}\frac {\partial P}{\partial \theta } + \frac {\partial ^2 U_\theta }{\partial r^2} + \frac {2}{r}\frac {\partial U_\theta }{\partial r} + \frac {1}{r^2}\frac {\partial ^2 U_\theta }{\partial \theta ^2} + \frac {\cot \theta }{r^2} \frac {\partial U_\theta }{\partial \theta } - \frac {U_\theta }{r^2\sin ^2 \theta } + \frac {2}{r^2}\frac {\partial U_r}{\partial \theta } \nonumber \\[3pt] \ = \frac {\varkappa }{\varepsilon ^2} (C^+-C^-) \frac {1}{r}\frac {\partial \varPhi }{\partial \theta } + \epsilon _D \left ( \frac {\partial U_\theta }{\partial t} + U_r\frac {\partial U_\theta }{\partial r} + \frac {U_\theta }{r}\frac {\partial U_\theta }{\partial \theta } + \frac {U_r U_\theta }{r} \right )\!. \end{align}
Here,
where
$\nu =\mu /\rho _m$
is the kinematic viscosity and
$Sc = \nu /D$
is the Schmidt number. Using the physical properties of the non-polar liquid corresponding to the experiments of Stotz,
we obtain
For comparison, the coefficient multiplying the Coulomb body force term is
and is therefore of order unity. Thus, even before the computed field magnitudes are taken into account, the acceleration contribution is strongly suppressed relative to the Coulomb body force.
A direct numerical comparison was additionally performed during an early transient stage, when the solution was still clearly time dependent and had not reached its steady state. The radial Coulomb force contribution,
was compared with the complete radial acceleration term,
\begin{equation} \varepsilon _D \frac {\textit{DU}_r}{\textit{Dt}} = \varepsilon _D \left (\frac {\partial U_r}{\partial t} + U_r \frac {\partial U_r}{\partial r} + \frac {U_\theta }{r} \frac {\partial U_r}{\partial \theta } - \frac {U_\theta ^2}{r} \right )\!. \end{equation}
The maximum magnitude of the radial Coulomb force term was approximately
$4.5\times 10^4$
, whereas the radial acceleration contribution remained of order unity. The latter was therefore approximately four to five orders of magnitude smaller.
The same comparison was carried out for the polar components. The maximum magnitude of the polar Coulomb force term,
was approximately
$1.4 \times 10^4$
, whereas the polar acceleration contribution,
remained of order
$10^{-1}$
. The polar acceleration term was therefore approximately five orders of magnitude smaller than the corresponding Coulomb force component.
These estimates provide consistent evidence that the acceleration contributions are dynamically insignificant under the conditions considered here. Consequently, the particle acceleration terms may be neglected within the quasisteady approximation used in the present calculations.
Appendix B. Numerical method
The coupled NPPS system was solved numerically in axisymmetric spherical coordinates. The complete electrokinetic formulation was retained, without invoking any simplifications. The numerical procedure combines finite-difference spatial discretisation with a semi-implicit time-integration scheme (for more details, see Frants et al. Reference Frants, Amiroudine and Demekhin2024).
A staggered grid was used in the electrolyte domain. The ionic concentrations
$C^\pm$
and the electric potential
$\varPhi$
were defined at the cell centres, whereas the radial and polar velocity components were evaluated at the corresponding cell faces. All spatial derivatives were approximated with second-order finite differences. Since the strongest gradients occur within the EDL, the radial grid was refined towards the particle surface. A uniform discretisation was used in the angular direction. The computations presented in this work were performed on a grid containing
$1024$
nodes in the radial direction and
$256$
nodes in the angular direction, with more than 100 radial nodes inside the EDL.
The electric potential was calculated separately in the particle and in the surrounding media. Inside the dielectric particle, it satisfies Laplace‘s equation and was represented by the regular Legendre expansion
\begin{equation} \varPhi _p(r,\theta ) = \sum _{k=0}^{K} a_k r^k P_k(\cos \theta ). \end{equation}
The coefficients
$a_k$
were determined by matching this representation to the numerical solution in the liquid through the electrostatic boundary conditions at the particle surface.
The axisymmetric Poisson equation was written as a sum of radial and angular operators:
The discrete angular operator was diagonalised once at the initialisation stage:
Transformation to the eigenvector basis reduces the two-dimensional Poisson problem to a set of independent one-dimensional radial problems. The resulting tridiagonal systems were solved using the Thomas algorithm.
The Stokes equations were formulated in terms of the axisymmetric streamfunction. Taking the curl of the momentum equation eliminates pressure and leads to a fourth-order equation for the streamfunction. The angular part of this equation was treated using the same eigendecomposition strategy as for the Poisson equation. For each angular mode, the resulting fourth-order radial discretisation produces a pentadiagonal linear system, which was solved by a direct band matrix algorithm. The velocity components were then reconstructed from the streamfunction.
The Nernst–Planck equations were advanced in time using a third-order semi-implicit Runge–Kutta method Nikitin (Reference Nikitin2006a ,Reference Nikitin b ). For each ionic species, the spatially discretised equation was written schematically as
where
$H^\pm$
contains the terms responsible for the strongest numerical stiffness and is treated implicitly, whereas the remaining terms
$G^\pm$
are evaluated explicitly. This splitting avoids the severe time-step restriction of a fully explicit method while retaining a substantially lower computational cost than a fully implicit solution of the coupled nonlinear system. The scheme also includes an estimate of the local temporal error and an adaptive time-step procedure.
At every Runge–Kutta stage, the ionic concentrations were updated first, after which the electric potential and the hydrodynamic fields were recalculated. The particle velocity was obtained from the force-balance condition using the updated electrostatic and hydrodynamic stresses. The calculation was continued until the variation of the solution between successive time steps became smaller than the specified convergence tolerance.



U~∞
f(α)
f(α)
α→0
α→∞
U∞
E∞
U∞
δ=εp/εm
ε=0.166
σ=−20
E∞=5
46.7
100
ΔU∞
δ=0.05
t~≈10s
12μm
E~∞=105Vm−1
τ~=0.9172s
E∞=46.7
ε=0.166
ϰ=0.05
σ=−20
δ=0.05
ρ=C+−C−
ε=0.166
E∞=46.7
σ=−20
ϰ=0.05
δ=0.05
E∞
Uθ
θ
ε=0.166
E∞=10
σ=−7
ϰ=0.05
δ=0.05
7ε
Ω
Σ1
Σ2
Σ2
Σ1
C+
ε=0.166
σ=−20
E∞=46.7
δ=0.05
θ=0∘
45∘
90∘
135∘
180∘
C−
ε=0.166
σ=−20
E∞=46.7
δ=0.05
θ=0∘
45∘
90∘
135∘
180∘
ρ
ε=0.166
σ=−20
E∞=46.7
δ=0.05
θ=0∘
45∘
90∘
135∘
180∘
J−
E∞=46.7
σ=−20
ε=0.166
J+
E∞=46.7
σ=−20
ε=0.166
C−
σ=−7
σ=−20
σ=−140
ε=0.166
E∞=46.7
∣σ∣
ε=0.166
E∞=1
5
E∞=10
C+
C−
ρ
ε=0.166
σ=−20
E∞=46.7
E∞=93.4
E∞=140
J−
σ=−20
ε=0.166
E∞=46.7
E∞=93.4
E∞=140
U∞/E∞
E∞
ε=0.166
σ=−20
M~
a~=12μm
λ~D=2μm
ε=0.166
ε=0.166
σ=−20
δ=0.05
κ=0.05