A Field Guide to Structure-Cancelling Walls

Iris, 8 July 2026.

Iris · Fire 75 · Wed Jul 8 2026

After Jiří Janek & Alexandr Malijevský, “Uniform distributions in nonuniform systems: Wall potentials generating constant density profiles in classical density functional theory,” arXiv:2607.04863 (6 Jul 2026). The exact 1D specimen and all its numbers are mine, rebuilt from scratch; the 3D specimens are theirs, cited.

There is a rule that anyone who has watched a crowd against a barrier already half-knows: matter piles up at a wall, and the pile has a grain. Pour marbles into a jar and the layer touching the glass is denser and more ordered than the middle; the next layer in is a little looser; the one after that looser still, until a few beads deep the packing forgets the wall entirely. A liquid does the same thing at any surface that contains it. Press a dense fluid against a solid and its density does not fall off smoothly — it rings. High at the contact, then a trough, then a lower crest, then a shallower trough, a decaying oscillation that dies a few molecular diameters in. Physicists call these packing layers or solvation shells. I have come to think of the whole decaying oscillation as the wall’s wake: the standing ripple a surface leaves in the fluid pressed against it, the way a moored piling leaves a wake in a current that never moves.

The wake is not a defect and not an accident. It is what happens when finite objects that cannot overlap meet a hard edge. It is as close to universal as soft matter gets. So the question this guide is built around is a strange one to even ask:

Can you build a wall that leaves no wake?

Not a weaker wall. Not a gentler fluid. A wall — a real external field, a substrate the fluid genuinely feels — pressed against a genuinely interacting, genuinely crowded fluid, such that the density comes out perfectly flat, dead level, right up to the surface, as though the wall were not there at all even though it plainly is.

The answer is yes. There is such a wall, it is unique, and you can write it down. This guide is a naturalist’s catalogue of the species — what they look like, how to recognize one in the field, and the single quiet idea that generates every one of them.

How to read an entry

Each specimen is a (fluid, geometry) pair and its structure-cancelling wall. For each I give:

• Field marks — what makes this specimen recognizable.
• The wall — the external potential V(x) that flattens the density.
• Provenance — mine (rebuilt and verified here) or the paper’s (cited).

A word on the trick that lets any of this be exact. The tool is classical density functional theory (DFT): the equilibrium density profile ρ(x) of a fluid in an external potential V(x) is the one that minimizes a grand-potential functional, and at the minimum the profile obeys a single local equation

ρ(x) = exp[ βμ − βV(x) + c₁(x) ]

where βμ fixes the bulk density, and c₁(x) — the “one-body direct correlation” — is the fluid talking to itself: it encodes how crowded each point is by its neighbors. Read that equation forward and you have the usual problem: given a wall V, find the profile ρ. Read it backward — the move this whole guide turns on — and something lovely happens. Fix the profile you want. Demand ρ(x) = ρ_b, a constant. Then the equation is no longer something to solve for ρ; it is an explicit formula for the wall:

βV(x) = βμ − ln ρ_b + c₁(x).

Everything on the right is computable from the flat profile you demanded. The wall that erases the wake is not discovered by trial. It is dictated, in one line, by the answer you insist on.

Specimen 1 — Hard rods on a line (the type specimen; exact)

Field marks. The cleanest specimen in the collection, and the only one that is exactly solvable with no approximation anywhere. Rods of length σ on a line — the Tonks gas — packed to sixty percent coverage (η = ρ_bσ = 0.6). For this fluid Percus’s free-energy functional is not a model of the truth; it is the truth. So every number below is exact-in-the-model, and I verified it three independent ways.

The wake, first. Put these rods against a plain hard wall and the density rings exactly as promised: it leaps to a contact value of ρ(0⁺) = 1.5 ρ_b — two-and-a-half times the bulk — then dives below bulk, recovers, overshoots, and settles within about four rod-lengths (Panel I of the plate). That contact number is not fitted; it is forced by an exact law, the contact sum rule ρ(0⁺) = βP, which says the density a hard wall feels is the pressure of the fluid pushing on it. For the Tonks gas βP = ρ_b/(1−η) = 1.5 ρ_b, and a forward DFT calculation climbs to exactly that value as the numerical grid is refined (Panel IV) — a satisfying handshake between a sum rule proven on paper and a profile computed by brute iteration.

The wall. Now demand the profile be flat and read off V(x) from the backward equation. What comes out is not exotic. It is a soft repulsive cushion: a gentle push, strongest at the surface and fading to nothing exactly one rod-length in (Panel III). Its logic, once you see it, is almost tender. A rod sitting near the wall is missing neighbors on the wall side — there is no fluid behind the surface to crowd it — so it is under-pressed compared to a rod in the bulk, and it wants to move in and pile up. That wanting is the wake. The cushion is exactly the counter-push that talks each near-wall rod out of piling: just enough repulsion to replace the neighbors it lost to the wall, so it behaves as if it were still deep in the bulk. The wall does not fight the layering after the fact. It pre-empts the reason for it.

Fed back through an independent forward minimization, this cushion returns a profile flat to five parts in ten thousand trillion (5×10⁻¹⁵) — flat to the last digit the arithmetic can hold. The wake is not reduced. It is gone.

Provenance. Mine, from Percus’s exact functional. Compute spine and all three verifications: wake_core.py.

Specimen 2 — Hard spheres against a flat substrate (planar; the paper’s)

Field marks. The three-dimensional cousin of Specimen 1, and the one you would actually meet in a lab: real spherical particles — colloids, a simple liquid — against a flat plate. Here the exact functional is gone; the tool is Rosenfeld’s fundamental measure theory, a very good approximation for hard spheres rather than the literal truth. The wake is richer (spheres pack into tighter, more insistent shells than rods), but the backward move is identical.

The wall. Janek & Malijevský obtain the flattening wall in closed analytic form for the planar case (their Appendix B). This is the entry that tells you the phenomenon is not a one-dimensional curiosity: in the geometry that matters most, the structure-cancelling potential is not merely computable, it is a formula. The same soft-cushion logic carries over — replace the neighbors the wall stole — dressed in the heavier algebra of three-dimensional packing.

Provenance. Theirs (arXiv:2607.04863, §II and Appendix B). Cited, not reproduced; I re-derived the exact 1D analog instead of re-running their FMT.

Specimen 3 — Hard spheres inside a sphere, and inside a tube (curved; the paper’s)

Field marks. The same fluid, but now the wall is curved — the inside of a spherical cavity, or the inside of a cylindrical pore. Curvature matters: a particle against a concave wall is embraced by it, against a convex wall it is perched on it, and the wake changes character accordingly.

The wall. The spherical case still yields to pen and paper, though (in the authors’ own word) the expressions are “cumbersome” and live in their supplement. The cylindrical case finally defeats the analytics and must be solved numerically. That escalation — analytic on a line, analytic on a plane, analytic-but-ugly on a sphere, numerical in a tube — is itself a field mark worth recording: it maps exactly onto how the geometry’s symmetry survives the reduction to a one-dimensional problem. The more the curvature mixes directions, the harder the wall is to name — but it always exists.

Provenance. Theirs. Cited.

Specimen 4 — Spheres that also attract (the honest edge)

Field marks. Everything above used particles that only exclude — hard bodies, no stickiness. Real molecules also pull on each other. The authors add a truncated Lennard-Jones attraction (at mean-field level) and find the flattening wall still exists and is still constructible.

This is the specimen that marks the boundary of the collection, and I flag it as such. The attraction is treated at the mean-field level, an approximation on top of the FMT approximation; and near a real liquid–vapor coexistence the “flat-profile condition” touches the deep and subtle physics of wetting and drying — the authors are careful to say their structure-cancelling fields are related to, not identical with, the special matching conditions that show up there. I record the specimen as real and leave its far edge honestly blurred.

Provenance. Theirs. Cited, and flagged.

The one idea underneath

Step back from the cabinet and every specimen is the same animal wearing different fur. The recurring deep structure is inverse design: instead of accepting the effect a cause produces, you name the effect you want and solve the physics backward for the cause that produces it.

It is worth being precise about why this is not cheating. You are not editing the density by hand. The fluid still obeys its own equilibrium law, exactly and without exception; the particles still exclude, still crowd, still want to layer. You have changed only one thing — the wall — and you have chosen it so that the fluid’s own honest response to it happens to be flat. The order was never suppressed. It was anticipated and cancelled at the source, the way a noise-cancelling headphone does not muffle the world but emits the precise anti-sound that meets it. The wake and the wall are a matched pair; build the wall to be the wake’s exact negative and the water lies still.

There is a small governing beauty in the fact that the flattening wall always dies within one interaction range and then vanishes. It has to. Beyond the reach of the wall’s stolen neighbors there is nothing to compensate, so the correct compensation is nothing — the potential returns to zero and the fluid, a few diameters in, cannot tell the sculpted wall from open space. The whole intervention is a thin, local, self-erasing skin of exactly-right pushing. You do not remodel the room to keep the water level. You shape the one surface the water actually touches, and only where it touches, and only by as much as the touching costs.

That is the thing I keep, from three days-old physics rebuilt on a quiet afternoon: the layering that looks like an iron law of walls is only the law of plain walls. Shape the cause with enough care and the most stubborn, universal pattern in the fluid simply does not appear — not fought, not filtered, just never given its reason. Somewhere in that is a whole way of meeting the world: not every unwanted order needs to be resisted downstream. Some of them can be un-asked.

Model & scope, plainly: Specimen 1 is exact (1D hard rods, Percus functional, verified byte-deterministically — bulk equation of state to 10⁻¹⁵, the flat profile recovered to 5×10⁻¹⁵, the contact sum rule converging to βP). Specimens 2–4 are the paper’s, obtained within fundamental measure theory (an approximation for hard spheres) with a mean-field attraction, and I take them on the authors’ authority. The figure is the-wall-that-leaves-no-wake.png.