Phospholipids form a bilayer in water because they are amphipathic: a bulky, water-loving head group attached to two hydrophobic tails. Water molecules hydrogen-bond to each other and exclude the tails, and the resulting hydrophobic effect, plus steric packing limits on the head groups, confines the molecules to two opposing sheets with tails buried and heads facing the water.

That answer has three parts, and most short explanations give only the first. The hydrophobic effect supplies the driving force, but it does not by itself dictate a bilayer. Head-group sterics and tail geometry decide whether the aggregate is a micelle, a bilayer, or something else. And the bilayer is not a static wall but a fluid two-dimensional solution whose stability depends on temperature, lipid composition, and the surrounding solution.

Here is the full mechanism, the numbers that matter, the cases where the rule breaks, and what the model does and does not predict.

The hydrophobic effect is an entropy story, not a repulsion story

Water is a poor solvent for hydrocarbons not because water pushes them away but because water pays a price to accommodate them. Around an exposed nonpolar surface, neighboring water molecules lose hydrogen-bonding partners and must adopt ordered, cage-like orientations. That ordering costs entropy. When tails cluster together, some of that structured water is released back into the bulk, and the gain in water entropy outweighs the loss of tail entropy.

The result is a free-energy minimum at aggregation, and the numbers are modest per lipid but large in sum. Each methylene group transferred from water into a hydrocarbon interior contributes roughly 1.5–3.5 kJ/mol to the transfer free energy at room temperature. A two-chain phospholipid with 28–36 tail carbons therefore buries a hydrophobic surface worth several tens of kJ/mol. The hydrophobic effect grows with temperature up to roughly 100–140 °C before weakening, which is why heating is not a reliable way to dissolve membranes.

Here is the part that complicates the usual sentence. The bilayer interior is not dry hydrocarbon. Water penetrates roughly the first 8–10 Å of each leaflet near the head-group region, and even the acyl-chain region holds some water at defect sites. The bilayer is better described as a hydrated, structured hydrophobic slab than as a sealed oil film.

One plain-language version: the tails are not actively repelled by water; water simply prefers its own company, and the lipid tails are excluded as a consequence.

Why two sheets and not a ball: head-group packing decides the outcome

If the hydrophobic effect were the only force, lipids would collapse into the smallest hydrophobic object possible, a micelle with all tails in one core. Phospholipids do not do that because their head groups are large relative to the tail volume. This is captured by a dimensionless packing parameter, the tail volume divided by the product of head-group area and tail length. Lipids with a large head-group area and modest tail volume cannot fit into a sphere without leaving an unfavorable space or over-straining the interface.

Empirically, the packing parameter predicts the aggregate. Values near or above 1 favor bilayers; the classic boundary for large bilayers is around one third to one half, with higher values corresponding to planar bilayers and eventually inverted phases. A common phospholipid such as dioleoylphosphatidylcholine has a packing parameter that sits comfortably in the bilayer regime at physiological temperature. Lysophospholipids, with only one tail, have a much smaller hydrophobic volume and readily form micelles; that difference is exactly the geometry argument.

The bilayer solves a boundary problem. An open edge of a bilayer would expose tails to water, so the sheet closes on itself into a vesicle, or the edges are stabilized by curvature and packing defects. In a cell membrane the problem is solved by having no edge at all: the bilayer is a closed surface, and the energy cost of the exposed margin is avoided.

Quantified context: what the membrane actually looks like

Bilayer dimensions are well constrained. A typical phospholipid bilayer is about 40–50 Å thick from head to head, with each leaflet's head-group region contributing roughly 8–10 Å and the acyl-chain region making up the core. The hydrophobic thickness, the region between the polar head-group shells, is commonly 25–35 Å, and that matters for how embedded proteins are matched to the membrane.

Lateral mobility is fast. Individual phospholipids move micrometers per second, with lateral diffusion coefficients around 0.1–10 µm²/s, so a lipid can cross an entire bacterium-sized cell in seconds to minutes. Flip-flop between leaflets is rare without help, with spontaneous half-times measured in hours to days, and that asymmetry is maintained by dedicated enzymes (flippases, floppases, scramblases) that consume ATP or run down gradients.

Aggregate Typical lipid Shape driver Size scale Stability in excess water
Micelle Lysophospholipid, bile salt Single tail, large head 3–8 nm diameter Stable, no bilayer interior
Flexible bilayer/vesicle Short-chain or double-chained lipid Packing parameter near 1 20 nm to many µm Stable, closes into vesicles
Planar supported bilayer Long-chain phospholipid Strong bilayer former 4–5 nm thick, stacked Stable on a substrate, edges need support
Inverted hexagonal phase Unsaturated lipid at low hydration Packing parameter above 1 Periodic tubes of head groups Forms at low water content or special composition
Helical ribbon Mixed lipid with high bending energy Frustration between leaflets Nanoscale width Kinetically trapped, not equilibrium

The table's point is that "bilayer" is one outcome among several. The same amphipathic molecules can assemble into a micelle not by changing the water but by changing the ratio of head-group area to tail volume. That ratio can be shifted by pH, salt, and tail length.

Grayscale rules, edge cases, and where the model breaks

The clean binary of "hydrophilic head, hydrophobic tail" softens at the boundary. At extremes of pH, the phosphate or amine groups of a phospholipid gain or lose charge, changing the effective head-group area and the electrostatic repulsion between neighboring lipids. That can shift the favored assembly, split the transition, or increase the bilayer's area per molecule.

Bilayers also undergo a thermal phase transition at a melting temperature specific to tail length and saturation. Below the melting temperature, saturated tails pack into a gel phase; above it, the membrane is liquid-crystalline. For dipalmitoylphosphatidylcholine, the transition is about 41 °C, while unsaturated lipids such as dioleoylphosphatidylcholine stay fluid well below room temperature. Cholesterol modulates this sharply, broadening and eventually abolishing the transition at high enough fractions.

Caveat worth stating clearly: the bilayer's stability depends on the lipid being above its critical micelle concentration, but not in the way a simple soap analogy suggests. Actual critical micelle concentrations for phospholipids are extraordinarily low, often in the nanomolar to micromolar range, so in any biological or laboratory system the lipid is far above the threshold. The relevant control is not concentration but geometry and the available water.

Common misconceptions worth correcting

The most persistent error is describing water as "repelling" the tails. The hydrophobic effect is entropically driven and arises from water–water interactions, not from a repulsive force between hydrocarbon and water. A related error is assuming the bilayer is rigid. Above the phase transition the membrane is a two-dimensional fluid with a lateral pressure profile that varies by depth and can reach hundreds of atmospheres locally, even though no single water molecule "pushes."

Another misconception treats the bilayer as a precise, uniform structure. Real membranes contain domains, curvature, and defects. Proteins insert with hydrophobic spans that must match the hydrophobic thickness, and mismatches cost energy.

Red flag: treating the bilayer as an impermeable barrier. Small uncharged molecules and water cross it readily, and ion permeability of a pure lipid bilayer is extremely low but not zero. Red flag: assuming every amphipathic molecule forms a bilayer. Detergents, bile salts, and single-chain lipids favor micelles. Red flag: expecting the bilayer to stay intact far below the lipid melting temperature; gel-phase membranes can crack and leak. Red flag: using the word "stable" as if assembly were irreversible. Bilayer formation is reversible and sensitive to temperature and composition.

Practical actions for researchers and students

Concrete steps exist for making the model useful in a lab or study plan.

Set the temperature relative to the lipid transition temperature. For a saturated 16-carbon phospholipid this is about 41 °C; running below it gives a gel-phase bilayer, which is stiffer and less permeable than the liquid-crystalline state. Match the hydrophobic thickness to your protein. Measure or look up the hydrophobic span of membrane proteins and compare it with the 25–35 Å core; a mismatch changes insertion efficiency and can promote aggregation. Choose the assembly by packing parameter, not by convenience. For micelles, use single-chain detergents with a known critical micelle concentration; for vesicles, use double-chain lipids above their transition temperature. Populate a supported bilayer above the phase transition, then cool. Formation is faster and defects anneal better when the lipid is fluid during deposition. For storage, keep phospholipid stocks below their transition temperature in a dry state, and avoid repeated freeze–thaw cycles, which can fuse vesicles and alter size distributions. Check the actual critical micelle concentration for your detergent before assuming micelles form at your working concentration.

Where the bilayer model ends

The model explains why pure phospholipids in excess water form sheets and closed vesicles. It predicts the roughly 40 Å thickness and the low flip-flop rate. It also predicts that changing the head-to-tail ratio changes the aggregate. What it does not predict is the behavior of real cell membranes, which are mixtures of hundreds of lipid species plus cholesterol and proteins. Those mixtures can produce coexisting liquid-ordered and liquid-disordered phases, domains on the 10–100 nm scale, and curvature that the simple bilayer picture ignores.

The most honest boundary is thermodynamic. Equilibrium assembly into a bilayer is well understood for simple systems. Nonequilibrium processes, such as vesicle budding, membrane fusion, and the maintenance of lipid asymmetry by enzymes, are not consequences of the hydrophobic effect alone. They depend on protein machinery and on the cell's energy budget. The bilayer is a necessary starting condition for those processes, not a sufficient explanation of them.

Read the rule this way: the hydrophobic effect gives phospholipids a strong reason to hide their tails from water, and packing constraints give them a strong reason not to hide them in a sphere. The bilayer is the solution that satisfies both. Every real membrane is that solution with additional complications layered on top.