Frosted Pattern: When Growth Leaps Forward ✛ ✎
The frost pattern on a windowpane doesn't spread uniformly: it sprouts branches, captures new areas, sometimes sealing off voids. This growth is not a smooth expansion but a series of leaps. This image resonates surprisingly with a mathematical model devised by Charles Loewner a century ago. It describes how domains in the complex plane expand under the influence of a one-dimensional signal — the driving function W(t). This signal acts as an invisible 'finger' guiding the boundary: Loewner's equation \(\partial_t^+ g_t(z) = \frac{2}{g_t(z)-W(t)}\) dictates how the conformal map $g_t$ evolves, carving the growing hull out of the upper half-plane.
But the classical approach required continuity of W(t). What if the signal jumps? Mathematicians Eveliina Peltola and Anne Schreuder provided the answer. It turns out that discontinuities don't break the construction: they give birth to new geometry. The main result: every locally growing hull corresponds to exactly one càdlàg function (right-continuous with left limits), and vice versa. At jump times, not just individual points but entire chunks are added. Sometimes these are bubbles — open regions suddenly swallowed by the hull, like air pockets in a frost pattern. Other times, they are compact clusters of boundary points $P_t$, rigidly attached to the real line (Theorem 1.4). The topology of these additions is governed by subtle properties of W(t).
A particularly interesting question is: can a hull be generated by a single curve? Yes, if the union of the hull with the real line is path-connected (Theorem 1.6). Then there exists a function η(t)—the 'trace' of growth — which turns out to be càglàd (left-continuous with right limits). And if η itself is left-continuous, the hulls become locally connected. These connections reveal the structure of fractal boundaries.
The practical significance of this work goes far beyond pure analysis. Loewner theory has experienced a second birth thanks to SLE (Schramm–Loewner evolution)—a universal stochastic process that describes the boundaries of random clusters in two-dimensional lattice models. For κ > 8, the SLE trace ceases to be a simple curve, and proving its existence remains an open problem. This new classification of hulls provides a tool: perhaps discontinuous driving functions will help construct approximations. Moreover, the work directly connects jumps in W(t) to phase transitions in probabilistic models, where the driving function can be a Lévy process and Markov chains describe its statistics. Understanding how bubbles and compact sets are born is critical for numerical simulations of rare events in random media, where nonlinear effects amplify jumps.
The frost pattern on glass is never perfectly smooth: random fluctuations in temperature and humidity are frozen into it. Similarly, mathematical hulls born from discontinuous driving possess a natural, rather than idealized, beauty. The results of Peltola and Schreuder turn an annoying feature into a working tool. Now Loewner theory is ready to describe jerky processes — from cracks in materials to the growth of biofilms. And the icy filigree on your window gains a rigorous algebraic subtext.
Although Loewner's equation looks like an innocuous ODE, it can generate mind-boggling geometric shapes. For instance, the spiral from Example 1.10, winding around a circle, at the moment of closure abruptly adds an entire disk, and this with an absolutely continuous driving function!