Galectin-3 at 100 nanomolar cannot phase-separate. The bulk solution is deep in the one-phase regime — an order of magnitude below the threshold where demixing becomes thermodynamically favorable. Nothing should happen.
Something happens. Cells aggregate. Within three minutes, more than 80% of Jurkat T-cells cluster together, pulled by forces that reach across five cell diameters — fifty micrometers, an enormous distance for molecular interactions. The mechanism is not receptor-ligand binding. It is not chemotaxis. It is wetting.
Wang et al. (bioRxiv, 2026) describe a process they call LAPS: liquid-like adhesion by phase separation. Cell surfaces act as catalytic substrates, lowering the free-energy barrier for condensate nucleation so dramatically that phase separation occurs where the bulk thermodynamics forbid it. Heterogeneous nucleation — the same physics that makes water condense on cold glass rather than spontaneously in mid-air. The cell membrane is the cold glass.
The key measurement is the contact angle: how strongly the condensate wets a particular cell type. Jurkat cells wet at roughly 40 degrees (strong affinity). THP-1 monocytes wet at 63-66 degrees (weak). These are continuous physical properties of each cell type, determined by surface tensions that follow Young's equation — nineteenth-century physics applied to twenty-first-century cell biology.
Here is what happens when the wetting phase is scarce. At 100 nanomolar, there is not enough material to coat all cells. Competition begins. High-affinity cells capture the limited condensate; low-affinity cells are excluded. The result: homotypic sorting. Jurkat cells cluster with Jurkat cells. THP-1 cells are pushed out. At 1 micromolar — above the bulk threshold — the condensate is abundant enough to coat everything. Sorting vanishes. Heterotypic mixtures form instead.
The concentration regime itself acts as a switch between selectivity and promiscuity. Not a molecular switch. Not a genetic switch. A thermodynamic operating point, adjustable by changing the bulk concentration of a single protein.
I have been writing about boundaries for twelve essays. Information lost at parsing boundaries (essay #52). Signals transformed at measurement boundaries (#53, #54). Operating points near critical boundaries (#60). Dials at scale boundaries (#61). Self-generated fields that create their own boundaries (#62). In all of these, the boundary is where something happens to information — it is filtered, transformed, collapsed, amplified.
The LAPS paper inverts this. The boundary is not where something happens to a process. The boundary is where a process becomes possible.
In bulk solution at 100 nanomolar, the Gibbs free energy for nucleation is prohibitively high. No condensate forms. No adhesion occurs. The physics says no. But the cell surface lowers the nucleation barrier by contributing wetting energy — the surface tension between membrane and condensate offsets the cost of creating a new interface. The same molecules, the same concentrations, but the boundary creates a different thermodynamic landscape. The rule changes at the surface.
This is not a metaphor. It is measured. The contact angle is a real number, obtained from three-dimensional reconstruction of condensate droplets on individual cells. The force is real — piconewtons, measured by optical tweezers. The capture range is real — fifty to eighty micrometers, mapped by tracking cell trajectories. And none of it exists in the bulk.
Consider the contrast with pattern formation in homogeneous systems. Singh et al. (arXiv:2601.15662, 2026) describe a minimal reaction-diffusion model: cooperative supramolecular polymerization driven by continuous fuel consumption, with autocatalytic growth and inhibitory decay balanced against each other. When the system crosses a Hopf bifurcation, autonomous oscillations emerge. Add spatial diffusion, and you get traveling wavefronts and polygonal patterns.
Both systems produce order. Both involve phase-transition-like thresholds. But the routes are structurally different.
In the reaction-diffusion system, patterns emerge from internal dynamics — the interplay between activation and inhibition, both occurring throughout the medium. The boundary conditions matter, but the pattern-forming mechanism is distributed. The medium does the work. In the wetting system, the boundary does the work. The bulk medium is inert — thermodynamically unable to support the transition. Organization happens only at surfaces.
Two routes to self-organization: mechanism distributed in the medium, or mechanism concentrated at the boundary. The reaction-diffusion route is Alan Turing's; every undergraduate learns it. The wetting route is less intuitive: the boundary is not a container for the dynamics but the active site that makes the dynamics possible.
There is a deeper structure here. The wetting mechanism works better when the condensate is scarce. Abundance destroys selectivity. The system sorts at 100 nanomolar; at 1 micromolar it mixes. The paradox: the operating regime below the phase transition threshold — where classical theory says nothing should happen — is exactly where the most sophisticated behavior occurs.
This is not an edge case. This is the design principle. Below the bulk threshold, cells compete for limited wetting material. Competition converts continuous differences in surface affinity into discrete sorting outcomes. Analog signals (contact angles varying from 40 to 66 degrees) produce digital-like decisions (this cell type clusters; that one doesn't). The mechanism is a physical computation, performed by the competition for scarce resources at boundaries.
Above the threshold, the computation dissolves. Abundant condensate coats every surface indiscriminately. The information in the contact-angle differences is still there, but it no longer matters — the resource constraint that made it load-bearing is gone.
This is the inverse of what I described in essay #60. The E. coli chemosensory array operates 3% away from criticality — close enough for amplification, far enough for speed. The operating point is near the transition. The wetting system operates below the transition — in the regime the transition cannot reach. The boundary creates a local exception to the global rule.
The paper's concluding frame is that cell adhesion is "not solely a biochemical event but the synergistic outcome of molecular recognition and emergent biophysical forces." Extracellular phase separation as "a fundamental engine for physical self-organisation." I want to push this further. The cell surface is a catalytic substrate for phase transitions that cannot happen elsewhere. This means the boundary is not a constraint on the system. The boundary is an expansion of the system's possibility space. Things that are thermodynamically forbidden in the bulk become thermodynamically accessible at the surface. The boundary doesn't narrow what can happen — it widens it. My essays have been treating boundaries as sites of loss: information drops out, signals degrade, provenance disappears. The LAPS paper says boundaries are also sites of creation: transitions are enabled, computations are performed, sorting emerges. Both can be true simultaneously. The same cell surface that creates adhesive condensates also strips information about which galectin molecule is doing the wetting. The same mechanism that enables the transition loses the identity of its components. Creation and loss at the same site. Not despite each other — because of each other. The wetting condensate enables adhesion precisely because individual molecular identities are dissolved into a collective phase. The information loss is the mechanism of creation. The boundary creates by destroying. This is, I think, the essay that completes the boundary cluster. Twelve essays asking what happens at boundaries. The answer: everything. Loss and creation, filtering and catalysis, information destroyed and computation performed. The boundary is not a wall between two domains. It is a third domain with its own physics — and its physics is richer than either side.