---
title: Asymmetric Kedem–Katchalsky Conditions
url: https://www.emergentmind.com/topics/asymmetric-kedem-katchalsky-boundary-conditions
type: topic
---

# Asymmetric Kedem–Katchalsky Conditions

Asymmetric Kedem–Katchalsky boundary conditions are interface laws for transport across a membrane or sharp internal boundary in which the two sides do not enter on an equivalent footing. In their linear diffusion form, they couple flux continuity to a membrane law such as $J=P(c_1-\sigma c_2)$, so that the driving jump is not the symmetric difference $c_1-c_2$ but a weighted difference determined by the partition coefficient $\sigma$; in other settings, asymmetry is expressed by distinct interface coefficients on the two sides, as in $\partial_{\nu_1}u_1=\gamma_1(u_2-u_1)$ and $\partial_{\nu_2}u_2=\gamma_2(u_1-u_2)$, or by nonlinear jumps such as $\llbracket \Pi(u)\rrbracket$ in thin-membrane limits. Recent work places these conditions at the intersection of membrane transport, stochastic simulation, elliptic and parabolic transmission problems, spectral theory, and singular perturbation analysis [2007.12939] [2507.18400] [2105.02063].

## 1. Canonical formulations

The most common linear setting is a multilayer diffusion problem in which each layer satisfies
$$
\frac{\partial c_i}{\partial t}=D_i\frac{\partial^2 c_i}{\partial x^2},
$$
while the interface enforces flux continuity together with a Kedem–Katchalsky law,
$$
-D_1\frac{\partial c_1}{\partial x}=-D_2\frac{\partial c_2}{\partial x}=P(c_1-\sigma c_2).
$$
Here $P$ is the permeability coefficient and $\sigma$ is the partition coefficient. This pair of relations is the complete interface law in the two-layer diffusion model developed for Langevin simulation, and it is asymmetric whenever $\sigma\neq 1$; in practice it is also asymmetric when $D_1\neq D_2$ [2007.12939].

A second, explicitly directional formulation appears in interface logistic systems. There the two populations live in adjacent subdomains and interact only through the common interface, where the membrane law is written
$$
\partial_{\nu_1}u_1=\gamma_1(u_2-u_1), \qquad \partial_{\nu_2}u_2=\gamma_2(u_1-u_2).
$$
The coefficients $\gamma_1$ and $\gamma_2$ are generally different, so the law is asymmetric in a stronger sense than the partition-weighted diffusion law: the two sides are assigned distinct transfer coefficients [2402.08984].

A third formulation arises in thin-membrane limits for porous-medium or Darcy systems. After collapse of a membrane of thickness $\varepsilon$ to a zero-thickness interface, the effective transmission condition becomes
$$
\tilde\mu_{1,3}\,\llbracket \Pi \rrbracket
=
\tilde\mu_1\,\tilde u_1\nabla \tilde p_1\cdot \tilde{\boldsymbol n}_{1,3}
=
\tilde\mu_3\,\tilde u_3\nabla \tilde p_3\cdot \tilde{\boldsymbol n}_{1,3},
$$
with $\Pi'(u)=u\,p'(u)$. For the porous-medium pressure law $p(u)=u^\gamma$,
$$
\Pi(u)=\frac{\gamma}{\gamma+1}u^{\gamma+1}.
$$
This law is compatible with nonlinear generalized Kedem–Katchalsky conditions, although it does not introduce separate one-sided membrane permeabilities [2105.02063].

| Setting | Interface law | Asymmetry mechanism |
|---|---|---|
| Two-layer diffusion | $-D_1 c_{1x}=-D_2 c_{2x}=P(c_1-\sigma c_2)$ | $\sigma\neq 1$, $D_1\neq D_2$ |
| Interface logistic system | $\partial_{\nu_1}u_1=\gamma_1(u_2-u_1)$, $\partial_{\nu_2}u_2=\gamma_2(u_1-u_2)$ | $\gamma_1\neq\gamma_2$ |
| Thin-membrane porous-medium limit | $\tilde\mu_{1,3}\llbracket \Pi \rrbracket =$ common normal flux | nonlinear jump, heterogeneous one-sided traces |

## 2. Mechanisms of asymmetry

The partition coefficient is the simplest asymmetry mechanism. In the diffusion law $J=P(c_1-\sigma c_2)$, the interface flux is driven by $c_1-\sigma c_2$ rather than by a plain concentration difference. If $\sigma\neq 1$, the two sides are weighted differently; physically this corresponds to a jump in chemical potential across the interface. The same framework makes clear that finite permeability and partitioning are distinct effects: when $P\to\infty$, the law reduces to the partition condition $c_1=\sigma c_2$, while for finite $P$ and $\sigma=1$ one obtains the pure membrane law $J=P(c_1-c_2)$, which still permits a concentration jump because the membrane has finite resistance [2007.12939].

In Brownian-diffusion formulations, the same asymmetry is expressed through the relation
$$
\sigma=\exp\!\left(-\frac{\Delta\mu}{k_B T}\right),
$$
where $\Delta\mu=\mu_2-\mu_1$ is the chemical-potential jump. The boundary condition
$$
J=P[p_1(0,t)-\sigma p_2(0,t)]
$$
is therefore direction-dependent in effect whenever $\sigma\neq 1$, even though the membrane permeability parameter $P$ itself is taken symmetric [2003.09685].

A stronger notion of asymmetry appears when the interface law itself uses different coefficients on the two sides. In the two-species stationary interface logistic system, adding the two interface relations yields
$$
\frac{\partial u_1}{\partial \nu_1}+\frac{\partial u_2}{\partial \nu_2}
=
(\gamma_1-\gamma_2)(u_2-u_1),
$$
so continuity of total normal flux fails unless $\gamma_1=\gamma_2$. In that setting the operator is non-self-adjoint, and the asymmetry is structural rather than merely a consequence of unequal bulk coefficients [2507.18400].

Not all generalized Kedem–Katchalsky laws are asymmetric in this explicit directional sense. In the porous-medium thin-membrane limit, the interface constitutive law uses a single effective permeability $\tilde\mu_{1,3}$, and the surviving asymmetry comes from unequal bulk mobilities $\tilde\mu_1\neq \tilde\mu_3$ and from the one-sided traces at the limit interface. The paper is explicit that this is not a law with distinct directional membrane coefficients [2105.02063].

A more mechanistic perspective comes from the kinetic-theory derivation of osmotic transport in porous media. There, $\sigma$ is interpreted as the fraction of solute molecules reflected at pore entrances and exits, and a further parameter $\beta$ measures momentum transfer from rebounding solute molecules to nearby solvent molecules. The paper does not derive asymmetric left/right boundary conditions directly, but this suggests that genuinely directional Kedem–Katchalsky laws could be built by allowing $\sigma$ or $\beta$ to depend on the side of approach [1403.8077].

## 3. Stochastic realization and computational implementation

A major development is the translation of asymmetric Kedem–Katchalsky interface laws into particle-based Langevin dynamics. In the two-layer diffusion model, particle motion is generated by the underdamped Langevin equation
$$
m\frac{dv}{dt}=-\alpha(x) v+\beta(t)+f(x),
$$
with Einstein relation $\alpha(x)D(x)=k_B T$, and numerically integrated by the GJF scheme. The two Kedem–Katchalsky parameters are represented separately. First, finite permeability is converted into a transmission probability
$$
p=\frac{2P}{2P+v_{\rm th}},
$$
where
$$
v_{\rm th}=\sqrt{\frac{2k_B T}{\pi m}}.
$$
Second, the partition coefficient is represented as a chemical-potential step
$$
\phi_{\rm step}=k_B T\ln(\sigma)H(x),
$$
approximated by a thin force layer together with the trajectory weight
$$
W(x)=\exp\!\left[\frac{\phi(x)-\phi_{\rm step}(x)}{k_B T}\right].
$$
When a trajectory attempts to cross the interface, a uniform random number determines reflection or transmission; after transmission, the step is recomputed with an effective friction
$$
\alpha_{\rm eff}=
\frac{\alpha(x^n)|x^n|+\alpha(x_b^{n+1})|x_b^{n+1}|}
{|x^n|+|x_b^{n+1}|}.
$$
The method reproduces the continuum law
$$
-D_1\frac{\partial c_1}{\partial x}
=
-D_2\frac{\partial c_2}{\partial x}
=
P(c_1-\sigma c_2),
$$
shows excellent agreement with the continuum diffusion problem, captures the concentration jump at the interface, and converges as the interface-layer thickness parameter $\gamma$ is reduced. In the drug-eluting-stent example, the error at $x=0$ on the right side at $t=10^5$ decreases from $0.0095$ for $\gamma=4$ to $0.0012$ for $\gamma=0.5$; the timestep must remain in the ballistic regime,
$$
dt\ll \tau_{\rm ballistic}^{\rm min}=\frac{m}{\max(\alpha_i)}=\min(D_i)\frac{m}{k_B T},
$$
so that particles resolve the interface layer [2007.12939].

A closely related Brownian-dynamics framework decomposes the general interface into three primitive discontinuities: a friction jump, a semi-permeable membrane, and a step chemical potential. In that formulation the membrane law is again
$$
J=P(p_1-\sigma p_2),
$$
with transmission probability
$$
\Pi=\frac{2P}{2P+v_{\rm th}},
$$
while the partition step is realized by a narrow interfacial potential layer and Boltzmann reweighting. The general Kedem–Katchalsky boundary is then implemented by combining ballistic averaging across the diffusivity jump, probabilistic transmission/reflection for permeability, and interface-layer forcing for the chemical-potential jump [2003.09685].

## 4. Variational structure, operator theory, and nonlinear steady states

A useful baseline is provided by symmetric membrane problems. In a reaction–diffusion system on two subdomains separated by a membrane, the condition
$$
\partial_{\boldsymbol n^1} u^1_i=\partial_{\boldsymbol n^1} u^2_i=k_i(u_i^2-u_i^1)
$$
is conservative and dissipative: flux continuity follows from $\boldsymbol n^2=-\boldsymbol n^1$, and the weak formulation contains the nonnegative jump penalty
$$
\int_\Gamma D_i k_i [u_i][\psi].
$$
The associated bilinear form has the structure
$$
B[w,z]
=
\int_\Omega \nabla w\nabla z
+
\int_\Gamma k_i (w^2-w^1)(z^2-z^1),
$$
which underlies the global weak-solution theory for $L^1$ data. This is Kedem–Katchalsky in a conservative symmetric form, not yet an asymmetric law [2007.11989].

Membrane spectral theory extends naturally to reaction–diffusion systems used for Turing analysis. For each species, the membrane conditions
$$
D_{\phi l}\nabla \phi_l\cdot n
=
D_{\phi r}\nabla \phi_r\cdot n
=
k_\phi(\phi_r-\phi_l),
\qquad \phi\in\{u,v\},
$$
lead to symmetric bilinear forms with positive jump terms on $\Gamma$, compact self-adjoint membrane operators, and permeability-dependent eigenvalue sequences. The resulting patterns may be discontinuous at the membrane because the law permits nonzero jumps in concentration for finite $k_\phi$ [2105.09805].

The fully asymmetric case changes the analytical picture. In the stationary two-species/interface logistic system with
$$
\frac{\partial u_1}{\partial \nu_1}=\gamma_1(u_2-u_1),
\qquad
\frac{\partial u_2}{\partial \nu_2}=\gamma_2(u_1-u_2),
$$
the interface term in the bilinear form becomes
$$
\int_\Gamma (u_2-u_1)(\gamma_2 v_2-\gamma_1 v_1),
$$
which is symmetric only when $\gamma_1=\gamma_2$. The paper establishes a unique positive principal eigenvalue
$$
\lambda_*:=\Lambda_m^\Omega[-\Delta;\mathcal I],
$$
an upper threshold
$$
\lambda_\infty:=\min\{\lambda_{m_1},\lambda_{m_2}\},
$$
and the sharp existence criterion
$$
0<\lambda_*<\lambda<\lambda_\infty.
$$
It also proves that $\lambda_*$ is the unique bifurcation point from extinction and that, when $\lambda_\infty=\lambda_{m_1}<\lambda_{m_2}$, the positive steady state exhibits non-simultaneous blow-up: $u_{1,\lambda}\to\infty$ uniformly on $\overline{\Omega}_{0,1}$ while $u_{2,\lambda}$ remains bounded in $\overline{\Omega}_2$ as $\lambda\uparrow\lambda_\infty$ [2507.18400].

## 5. Thin-membrane limits and singular-perturbation regimes

The most rigorous derivation of a generalized Kedem–Katchalsky interface law from a resolved membrane model is obtained for a porous-medium equation under Darcy’s law. The prelimit problem is posed on three subdomains, with a membrane of thickness $\varepsilon$ and mobility scaling
$$
\frac{\mu_{2,\varepsilon}}{\varepsilon}\to \tilde\mu_{1,3}>0.
$$
Before passing to the limit, the model imposes continuity of density and continuity of normal flux across both membrane faces. As $\varepsilon\to 0$, the membrane collapses to a zero-thickness interface and the effective law becomes
$$
\tilde\mu_{1,3}\,\llbracket \Pi \rrbracket
=
\tilde\mu_1\,\tilde u_1\nabla \tilde p_1\cdot \tilde{\boldsymbol n}_{1,3}
=
\tilde\mu_3\,\tilde u_3\nabla \tilde p_3\cdot \tilde{\boldsymbol n}_{1,3},
$$
where $\Pi'(u)=u\,p'(u)$. For $p(u)=u^\gamma$,
$$
\llbracket \Pi \rrbracket
=
\frac{\gamma}{\gamma+1}\bigl((u^+)^{\gamma+1}-(u^-)^{\gamma+1}\bigr).
$$
When $f(u)=\ln u$, one recovers the classical linear jump in $u$, so the result can be read as a nonlinear generalization of the classical Kedem–Katchalsky interface law [2105.02063].

Singular-perturbation analysis in elliptic interface logistic problems shows how membrane asymmetry persists in effective limits. For the asymmetric interface law
$$
\partial_{\bf n_1}u_1=\gamma_1(u_2-u_1),
\qquad
\partial_{\bf n_2}u_2=\gamma_2(u_1-u_2),
$$
the principal eigenvalue satisfies
$$
\lim_{d\to \infty}\Lambda_1(-d\Delta+c_1,-d\Delta+c_2)
=
\frac{\gamma_2\int_{\Omega_1}c_1+\gamma_1\int_{\Omega_2}c_2}
{\gamma_2|\Omega_1|+\gamma_1|\Omega_2|},
$$
while the nonlinear positive solution obeys
$$
\lim_{d\to \infty}u_{i,d}
=
\frac{\gamma_2 \int_{\Omega_1}\beta_1 +\gamma_1\int_{\Omega_2}\beta_2}
{\gamma_2 \int_{\Omega_1}\alpha_1+\gamma_1\int_{\Omega_2}\alpha_2}.
$$
By contrast, as $d\to 0$,
$$
u_{i,d}\to \frac{(\beta_i(x))_+}{\alpha_i(x)}
\quad\text{uniformly in }\overline\Omega_i.
$$
The small-diffusion regime is therefore dominated by local carrying capacities, whereas the large-diffusion regime produces a membrane-weighted homogenization law [2402.08984].

## 6. Applications, misconceptions, and scope

The interface law is not merely formal. In the drug-eluting-stent model, the coating occupies $[-L_1,0]$, the arterial wall $[0,L_2]$, and the membrane at $x=0$ is modeled by
$$
-D_1 \frac{\partial c_1}{\partial x}
=
-D_2 \frac{\partial c_2}{\partial x}
=
P(c_1-\sigma c_2),
$$
with
$$
L_1= 5 \cdot 10^{-4} \, cm,\quad
L_2= 10^{-2} \, cm,\quad
D_1= 10^{-13} \, cm^2/s,\quad
D_2=7 \cdot 10^{-11} \, cm^2/s,
$$
$$
P= 10^{-5} \, cm/s,\qquad \sigma=0.164.
$$
Because $\sigma=0.164<1$, the interface energetically disfavors the right side and produces a substantial concentration jump at the membrane [2007.12939].

In ecological interface problems, asymmetric Kedem–Katchalsky conditions determine existence thresholds, bifurcation from extinction, and singular behavior near refuge zones. In particular, the directional coefficients $\gamma_1\neq\gamma_2$ enter the principal eigenvalue, alter the effective weighted averages seen in large-diffusion limits, and permit non-self-adjoint coupling strong enough to support non-simultaneous blow-up [2507.18400].

In reaction–diffusion pattern formation, membrane conditions with species-dependent permeabilities $k_u$ and $k_v$ modify the membrane Laplacian spectrum and admit discontinuous stationary patterns. This shows that Kedem–Katchalsky conditions should not be conflated with continuity transmission: for finite permeability, the solution itself may jump across the membrane, while the interface law ties that jump to the normal flux [2105.09805].

A common misconception is that “asymmetric” necessarily means separate left-to-right and right-to-left permeability coefficients. The recent literature shows several inequivalent meanings. In the two-layer diffusion and Brownian formulations, a single permeability $P$ is retained and asymmetry is encoded by $\sigma\neq 1$ and by heterogeneous bulk transport coefficients. In the porous-medium thin-membrane limit, the generalized law is nonlinear and uses one effective membrane permeability, so asymmetry is implicit through traces and bulk mobilities. Only some interface logistic models are asymmetric in the strongest sense, with distinct side coefficients $\gamma_1$ and $\gamma_2$ [2003.09685] [2105.02063] [2402.08984].

Taken together, these formulations define a broad research program. Asymmetric Kedem–Katchalsky boundary conditions may describe weighted concentration jumps, side-dependent membrane exchange, or nonlinear effective transmission laws derived from thin layers. Their mathematical signatures include concentration discontinuities, non-self-adjoint interface operators, membrane-weighted homogenized limits, and interface spectra that differ fundamentally from those of standard continuity transmission problems.

Source: https://www.emergentmind.com/topics/asymmetric-kedem-katchalsky-boundary-conditions