Ask most people why soap bubbles are round and they'll say something about fragility — round shapes distribute stress evenly, so a bubble that is round doesn't break. This is true but backwards. Bubbles are not round because roundness prevents bursting. They are round because the physics that governs surface tension happens to produce the smallest possible surface, and the smallest surface enclosing a fixed volume is always a sphere.
Surface tension as minimization
A soap film is a thin layer of water molecules trapped between two layers of soap molecules. Water molecules attract each other. A molecule in the interior of the film is pulled equally in all directions. A molecule at the surface is pulled inward and sideways, but not outward — there are no water molecules on the outside to pull it that way. This imbalance creates a net inward force at the surface: surface tension.
Surface tension is not a static property. It is a dynamic tendency of the surface to minimize its area. Every square millimeter of soap film costs energy — the energy required to hold those surface molecules in a less favorable configuration than being interior molecules. A soap film will spontaneously contract toward the configuration that minimizes total surface area, because that configuration minimizes total surface energy. The film is a physical optimizer.
The sphere theorem
The problem of finding the shape that encloses a given volume with minimum surface area is one of the classical problems of the calculus of variations, called the isoperimetric problem (in three dimensions). The answer is the sphere.
This was known geometrically to the ancient Greeks — Zenodorus proved it in some form around 200 BCE, though a rigorous proof in the modern sense came much later. The full proof using calculus of variations was completed in the 19th century. The result: for any volume V, the surface area of the enclosing sphere is the smallest surface area achievable by any smooth surface. Any other shape — a cube, a cylinder, an irregular blob — has more surface area than the sphere with the same volume.
Soap film minimizes area. Minimum area for a closed surface enclosing a fixed volume is a sphere. Therefore, a soap bubble is a sphere. The geometry is not an approximation or an averaging — it is exact.
When bubbles are not spheres
Two bubbles joined together are not two spheres. The junction creates a flat or curved membrane between them, and each bubble deforms from a perfect sphere. The precise shape is still governed by the minimization principle — the total surface area of the entire configuration is minimized — but the constraint is now shared volume, not individual volume.
The mathematics here becomes the Plateau problem, named after Joseph Plateau, a 19th-century Belgian physicist who studied soap films experimentally after losing his eyesight (reputedly from staring at the sun). Plateau observed that soap films always meet at angles of 120 degrees when three films join at an edge, and at angles such that four edges meet at each vertex in foam. These angles are not aesthetic choices — they are the configurations that minimize total area at every junction simultaneously. The mathematical proof that solutions to the Plateau problem always exist came from Jesse Douglas in 1931, earning him one of the first Fields Medals.
The precision of physical optimization
What is striking about soap bubbles is that they solve a nontrivial optimization problem — the isoperimetric problem in three dimensions — in real time, without computation, using only physics. The film does not approximate the sphere. It is the sphere, to the precision of its molecular structure.
Physical systems that minimize energy do this routinely. Crystals minimize lattice energy. Light minimizes travel time (Fermat's principle). Soap films minimize surface area. The mathematics needed to prove these minimization results rigorously is often far harder than the physics needed to produce them. Soap films were solving the Plateau problem before Douglas proved it was solvable.
The bubble is a physical proof. It is also, if you are in the right frame of mind, a genuinely strange object: a surface of zero mean curvature variation, a solution to a free-boundary optimization problem, realized in a moment and gone in the next.