Membrane: Structure, Mechanics & Applications

• Membrane is a thin, flexible interface that couples mechanics, transport, and chemistry across different systems, exemplified by lipid bilayers and graphene.
• Biological membranes are nanometer-scale lipid bilayers that separate compartments, maintain ion gradients, and support electrical signaling through channels and pumps.
• Physical membranes are modeled as low-dimensional elastic surfaces with bending and in-plane strain, providing insights into phenomena like vesicle wrapping and graphene behavior.

A membrane is a thin, flexible interface that separates regions while coupling mechanics, transport, and chemistry. In cell biology it is typically a lipid bilayer that defines cells and organelles, isolates ion-rich aqueous compartments, and supports signaling through channels, pumps, and transmembrane potentials (Zhu et al., 2021, Phillips, 2017). In soft-matter physics and statistical mechanics, a membrane is more broadly a low-dimensional flexible system embedded in a higher-dimensional space; in this sense, fluid lipid bilayers and crystalline sheets such as graphene are both membranes, but with different in-plane constitutive behavior (Arroyo et al., 2010, Katsnelson et al., 2013). Taken together, these works suggest that “membrane” names a class of interfaces whose defining features are compartmentalization, curvature elasticity, lateral organization, and selective coupling between adjacent media.

1. Biological and physical meanings

In biological usage, membranes are lipid bilayers present in all cells and tissues. They define intracellular and extracellular compartments, maintain distinct ionic concentrations, and provide the excitable boundary on which transmembrane potentials and action potentials are formed (Zhu et al., 2021). Quantitatively, biological membranes are nanometer-thin, flexible, electrically active barriers that are simultaneously a 2D fluid of lipids and proteins, a deformable elastic shell, a selective permeation barrier, and a capacitor with ion-selective resistors embedded in it (Phillips, 2017).

In statistical physics, a membrane is a low-dimensional flexible object fluctuating in a higher-dimensional space. The paradigmatic case is a two-dimensional surface embedded in three dimensions. A central distinction is between fluid membranes, which have no shear modulus and are described primarily by bending and tension, and crystalline membranes, which retain fixed in-plane connectivity and therefore resist shear and stretch (Katsnelson et al., 2013). Graphene is the prototype crystalline membrane: it realizes a truly two-dimensional system with atomic resolution, while biological lipid bilayers are the canonical fluid membranes (Katsnelson et al., 2013, Arroyo et al., 2010).

This dual usage is not contradictory. It reflects two limiting constitutive classes of membranes: laterally fluid bilayers, whose dominant mesoscale variable is curvature, and laterally elastic sheets, whose dominant variables include both curvature and in-plane strain.

2. Geometry, elasticity, and mechanical response

For fluid lipid membranes, the standard continuum description is the Helfrich free energy,

$$\Pi = \int_{\Gamma} \frac{\kappa}{2}(H - C_0)^2 \, dS + \int_{\Gamma} \kappa_G K\, dS,$$

where $H$ is the mean curvature, $K$ is the Gaussian curvature, $\kappa$ is the bending rigidity, $\kappa_G$ is the Gaussian rigidity, and $C_0$ is the spontaneous curvature (Arroyo et al., 2010). Closely related formulations appear in coarse-grained membrane simulations and in models of vesicles, tubes, and spherical membranes (Bonazzi et al., 2024, Sadeghi et al., 2017). Typical bending rigidities for phospholipid bilayers are reported as $\kappa_B \sim 10$–25 $k_B T$, with values up to $\sim 100\,k_B T$ with sterols; the area stretch modulus is $K_A \approx 200$–250 mN/m (Phillips, 2017).

Membranes are not characterized by bending alone. Their in-plane dynamics are governed by surface viscosity $H$0, and their relaxation can be limited either by membrane dissipation or by the viscosity of the surrounding bulk fluid. A central result for vesicles is a crossover length

$H$1

with relaxation time scaling as $H$2 for small vesicles and $H$3 for large vesicles (Arroyo et al., 2010). This establishes that membrane viscosity is not a minor correction: for sufficiently small objects, it is the dominant dissipative mechanism.

For crystalline membranes, the continuum elastic Hamiltonian includes both bending and in-plane strain,

$H$4

with the geometric nonlinearity in the strain tensor coupling out-of-plane and in-plane modes (Katsnelson et al., 2013). In graphene, this coupling produces scale-dependent elastic constants and validates the broader membrane-physics view that long-range fluctuations renormalize mechanical response.

3. Dielectric barriers, capacitance, and selective transport

A defining biological role of membranes is ionic isolation. In the optic nerve formulation, ions cannot readily enter the hydrophobic lipid core because they must lose both their ionic atmosphere and much of their dielectric stabilization; this creates a large dielectric energy barrier (Zhu et al., 2021). The lipid bilayer is therefore treated as essentially impermeable to ions, while channels and pumps provide selective conductive pathways. Charges can nonetheless accumulate on the two sides of the membrane, so the bilayer behaves as a capacitor (Zhu et al., 2021).

At the level of basic electrical parameters, membrane capacitance per unit area is of order $H$5, bare membrane conductance is about $H$6–$H$7 nS/cm$H$8, typical membrane potentials are of order $H$9 mV, and the resulting electric field across a $K$0 nm membrane is $K$1 V/m (Phillips, 2017). Water permeability for pure phospholipid bilayers is $K$2–$K$3 $K$4m/s, whereas ion permeability is many orders of magnitude smaller unless channels are present (Phillips, 2017).

The optic-nerve multidomain model replaces a purely conductive membrane condition by a dielectric membrane condition in which capacitive charging is explicitly included. For ion species $K$5 on membrane $K$6,

$K$7

with $K$8 and $K$9 in the simulations (Zhu et al., 2021). This term is not diffusion through lipid; it is a bookkeeping device that associates capacitive charge storage with species conservation.

Synthetic membranes can also implement regulated transport. In graphene-oxide/polyamine laminates, the membrane is a 250–500 nm stack of GO sheets and polyamine monolayers, with interlayer spacings of about 1.9–2.5 nm, and water and monovalent-ion permeability can be tuned by pH and by specific ions (Andreeva et al., 2020). In the GO–PA system, K$\kappa$0 permeability is reported to be about $\kappa$1 larger than for Na$\kappa$2, Li$\kappa$3, or Cs$\kappa$4, and the regulation mechanism relies on specific interactions between membrane internal components and ions rather than on fixed pore size alone (Andreeva et al., 2020).

4. Membrane-mediated interactions, curvature coupling, and domains

Membranes do not merely host inclusions; they mediate interactions between them. In the elastic description, inclusions constrain membrane shape, membrane fluctuations, or membrane thickness, and these perturbations generate effective forces that are generic and non-specific (Bitbol et al., 2019). For isotropic conical inclusions on a tensionless membrane, the mean-field curvature-mediated interaction scales as $\kappa$5, while the fluctuation-induced term is

$\kappa$6

and anisotropic inclusions can interact through a longer-ranged $\kappa$7 term with strong angular dependence (Bitbol et al., 2019). Hydrophobic mismatch adds shorter-ranged thickness-mediated interactions on the nanometer scale (Bitbol et al., 2019).

Coarse-grained simulations of arc-shaped particles on vesicles and tubes show that membrane-mediated interactions depend strongly on global curvature. For 60° and 90° particles, the strongest side-to-side attractions occur on weakly curved spherical membranes, whereas interactions become minimal when the membrane curvature approaches the particle curvature, $\kappa$8 (Bonazzi et al., 2024). For 90° particles, free-energy minima on spheres reach $\kappa$9 to $\kappa_G$0, while at curvature matching they are only about $\kappa_G$1 to $\kappa_G$2 (Bonazzi et al., 2024). This establishes a curvature-matching principle: cooperative aggregation is weakest when the ambient membrane already realizes the curvature favored by the inclusion.

Anisotropic curvature-inducing proteins can also reorganize membranes into higher-order structures. In a model of “membrane nematogens,” elongated inclusions with directional spontaneous curvatures oligomerize through membrane-mediated interactions and commonly produce tubular and sometimes sheet-like equilibrium shapes; the resulting tubes possess calculable compressional stiffness and persistence length (Ramakrishnan et al., 2013). When curvature-inducing molecules bind to both leaflets of a bilayer, the coupled membrane can support checkerboard, stripe, kagome-lattice, and thread-like domains, with unbound membrane stabilizing the vertices of checkerboard patterns (Noguchi, 2022).

A distinct interfacial regime appears near membrane criticality. Minimal simulations of coacervating bulk polymers coupled to an Ising membrane and membrane-bound tethers show “surface densities,” a phase reminiscent of pre-wetting in which a molecularly thin three-dimensional liquid forms on a membrane rather than on a rigid solid substrate (Rouches et al., 2021). Near-critical membrane behavior dramatically broadens the parameter regime in which this prewetting-like transition occurs, and the model exhibits three surface-phase coexistence even though the membrane and the polymer bulk each support only two-phase coexistence on their own (Rouches et al., 2021).

5. Active, viscous, and chemically driven membranes

Many biologically relevant membranes are nonequilibrium systems. Active protein forces can enhance or suppress membrane fluctuations, protein binding and unbinding can be activated or inhibited by other proteins and by chemical reactions such as ATP hydrolysis, and reaction–diffusion coupling can generate traveling waves, Turing patterns, and membrane shape changes (Noguchi, 2024). In review form, these dynamics are represented either by continuum reaction–diffusion equations on Helfrich surfaces or by discrete lattice and particle models with state flips (Noguchi, 2024).

A representative continuum form is

$\kappa_G$3

with a companion equation for $\kappa_G$4, where $\kappa_G$5 and $\kappa_G$6 are membrane-bound species and the reaction terms are coupled to curvature-dependent free-energy densities (Noguchi, 2024). In such models, Turing regions can be enlarged by curvature–composition coupling, and traveling concentration waves can drive large-amplitude shape oscillations of vesicles (Noguchi, 2024).

Membranes are also viscous–elastic surfaces in the plane. For protein adsorption that imposes local spontaneous curvature, a purely elastic Helfrich model is insufficient; a viscous–elastic formulation is required to describe lipid flow, dynamic tension redistribution, and the steady residual tension that remains after adsorption (Rangamani et al., 2014). The local incompressibility condition,

$\kappa_G$7

links tangential lipid flow $\kappa_G$8 and normal velocity $\kappa_G$9, while the tangential force balance shows that tension is generally a field $C_0$0, not a single global scalar (Rangamani et al., 2014).

Membrane mechanics can also be controlled by attached viscoelastic networks. In a supported Lo/Ld bilayer coupled to an actin network, increasing actin density transforms nearly circular domains into elongated and then triangular or polygonal shapes, because local anisotropic stresses overcome isotropic line tension (Arnold et al., 2024). Differential dynamic microscopy shows that domain coarsening accelerates actin-network relaxation: $C_0$1 s on a single-phase membrane, versus $C_0$2 s on a phase-separated membrane, a factor of $C_0$3 change in the characteristic relaxation time (Arnold et al., 2024).

6. Remodeling, rupture, and multiscale modeling

Membranes wrap, bud, rupture, and reorganize under external loads. For floppy giant unilamellar vesicles interacting with colloids, the wrapping state is controlled by the particle radius $C_0$4, vesicle radius $C_0$5, adhesion length $C_0$6, and the finite curvature of the vesicle (Spanke et al., 2020). The curved-vesicle theory predicts three regimes—no wrapping, activated wrapping, and spontaneous wrapping—with phase boundaries

$C_0$7

and these boundaries match experiments on microparticles and floppy vesicles more accurately than flat-membrane theory (Spanke et al., 2020).

At the opposite extreme, very sharp local curvature lowers the traction force needed for rupture. Coarse-grained simulations of a lipid bilayer bent over nanopillars show that high curvature dramatically lowers the traction force necessary to achieve membrane rupture, and experiments show that sharp silicon pillars with $C_0$8 nm produce about $C_0$9 permeabilization, while smooth pillars with $\kappa_B \sim 10$0 nm and comparable micron-scale diameter produce $\kappa_B \sim 10$1 permeabilization (Capozza et al., 2018). The controlling variable is therefore the local edge curvature rather than pillar diameter or aspect ratio (Capozza et al., 2018).

These phenomena motivate coarse-grained membrane models spanning nanometer to micrometer scales. A particle-based solvent-free model representing each leaflet by interacting particles reproduces the $\kappa_B \sim 10$2 thermal-undulation spectrum, recovers the target bending rigidity, allows independent tuning of area compressibility and in-plane viscosity, and reproduces continuum nanoparticle-wrapping behavior (Sadeghi et al., 2017). A later one-particle-thick additive model, MesoMem, implemented in LAMMPS, self-assembles into lamellae and stable vesicles, maps a gel/fluid/gas phase diagram, reproduces the $\kappa_B \sim 10$3 fluctuation spectrum for tensionless membranes, and permits osmotic pressure, spontaneous curvature, and adhesive colloid wrapping to be incorporated in a single mesoscale framework (Sillano et al., 27 Feb 2026).

7. Crystalline membranes and the graphene limit

The membrane concept extends beyond lipid bilayers. In statistical mechanics, graphene is a prototype crystalline membrane: a one-atom-thick two-dimensional crystal embedded in three-dimensional space (Katsnelson et al., 2013). Its continuum Hamiltonian contains both bending and in-plane elastic terms, and the nonlinear coupling between the height field $\kappa_B \sim 10$4 and in-plane strain renormalizes the bending rigidity according to

$\kappa_B \sim 10$5

with corresponding scale dependence of the in-plane elastic constants (Katsnelson et al., 2013).

Graphene thereby validates a central membrane-physics result: long-wavelength fluctuations are governed by interacting long-range modes rather than by a simple harmonic approximation. At the same time, it exposes where phenomenological membrane theories are incomplete. The various scaling properties of correlation functions agree well between microscopic and phenomenological descriptions, but the temperature dependence of the bending rigidity cannot be understood on the basis of phenomenological approaches alone (Katsnelson et al., 2013). At very high temperature, graphene does not melt by a simple two-dimensional fluid-membrane scenario; instead, its melting is described as decomposition into entangled carbon chains (Katsnelson et al., 2013).

This crystalline limit clarifies that “membrane” is not restricted to amphiphilic bilayers. It denotes a broader class of thin, flexible manifolds whose physics is controlled by the coupling of geometry, fluctuations, and material response. In biological membranes that coupling appears together with dielectric exclusion, selective transport, and active chemistry; in crystalline membranes it appears together with lattice elasticity and atomistic bonding. Across both classes, membrane behavior is fundamentally governed by the interplay of curvature, lateral organization, and multiscale dynamics.