Why the diamond latticeFive filters, finite enumeration, one survivor

Chapter 7 ended with a hint: on a 2D lattice, vortex pairs always eventually find each other and annihilate. The topological constraint that forces the BKT equilibrium dissolves with them. To derive physics that lives for longer than finite time, we need at least three dimensions. But which 3D lattice? There are infinitely many. The paper's answer is: diamond, and only diamond. It is not a choice; it is what you get when you ask five independent questions and insist on a yes to each.

Five filters

The paper distills the demand into five structural requirements. Each one is necessary for a piece of the α derivation to fire. All five together are satisfied by exactly one of the common crystal structures.

  1. Coordination z ≥ d+1 = 4. The Bravais-rank bound: in d = 3, the nearest-neighbour vectors must span the full 3D space. That requires at least four of them. Everything lower than four fails the span test.
  2. Bipartite. Every bond connects a site on sublattice A to a site on sublattice B, never A-A or B-B. This produces the chiral Bloch Hamiltonian with its two-band structure — the object whose nodal points yield the Dirac equation (Chapter 9).
  3. Oh symmetry. The point group must be the full octahedral symmetry. This is what pins the band crossings to symmetry-protected points in the Brillouin zone; without it, the nodes gap out and we lose the Dirac spectrum.
  4. Two sites per unit cell. One site is not enough for a chiral Bloch Hamiltonian; three or more over-constrain the structure factor. Exactly two. The structure factor f(k) becomes a scalar whose zeros protect Dirac crossings at codimension-2 loci.
  5. d ≥ 3. Vortex-line persistence. You just saw this fail in 2D. In d ≥ 3 the vortex is a line (or higher-dimensional defect); annihilation requires coherent collapse along the whole length. Topology survives.

Now run the filters against every 3D lattice you can think of. Exactly one survives.

Lattice z ≥ 4 bipartite Oh sym. 2 sites/cell d ≥ 3 Verdict
Click any column header above to see what that filter does and why it is necessary.
Pick a candidate from the dropdown and the 3D view above updates to show that lattice. Drag the canvas to rotate; shift+drag to pan; scroll to zoom. The filter table below shows which conditions each lattice satisfies. Click any column header for a description of what that filter enforces; click the dropdown (or a row's lattice name) for a per-lattice explanation. Burgundy/blue atoms mark A/B sublattices where a true bipartite structure exists; grey atoms indicate a non-bipartite lattice, since a two-colouring would be misleading. Seven of the most common lattices — and only the diamond row passes every column.

What diamond looks like

The paper is compact on the description; here is the picture. Diamond is two interpenetrating face-centred cubic (FCC) sublattices, one offset from the other by one-quarter of the body diagonal. Every atom has exactly four nearest neighbours, and those four neighbours sit at the vertices of a regular tetrahedron around it, all on the other sublattice. That is the bipartite structure in its most literal form.

The buttons below peel the structure apart one piece at a time.

Controls: drag to rotate, shift+drag to pan, scroll to zoom. The mode buttons below switch between views of the same underlying lattice.

Full lattice: the default oblique view. A atoms burgundy, B atoms blue; every bond connects a burgundy to a blue. A-sublattice only: hides all B atoms. What remains is a face-centred cubic (FCC) lattice — each A atom is at an FCC site, with 12 second-nearest A-neighbours at distance √2/2·a (drawn as faint grey lines). B-sublattice is the same picture, offset by (1/4, 1/4, 1/4) of the cube. One tetrahedron: isolates the central atom and its four bonded neighbours. The bond angle between any two of the four is exactly 109.47°, the tetrahedral angle. Nothing else in the lattice sits below that cage; it is the elementary building block. View along [111]: camera snaps to look down the body diagonal of the cube. The three-fold rotation symmetry becomes visible, and the honeycomb-like projection that diamond shares with graphite (but with tetrahedral sp3 bonds out of the page, not planar sp2) is clear. (111) honeycomb 1 / 2 / 3: diamond along the body-diagonal direction is a stack of buckled honeycomb sheets. Each button isolates one of the three consecutive honeycomb layers and snaps the camera to view it face-on — showing the A/B alternation that makes each sheet bipartite. Stack all three and you have the full diamond crystal.

A vortex line on the diamond lattice

With the lattice picked out, the last thing to see is the object that lives on it. In Chapter 7 we watched a vortex sit at a single point on a 2D grid. In three dimensions, a vortex is no longer a point: it is a line threading through the crystal, with phases winding 2π around any circle transverse to it. On diamond specifically, that line passes alternately through A-sublattice and B-sublattice atoms. As we'll see in Chapter 9, the bound state associated with the line localises on one of those sublattices — the chiral downfold — and that is what makes the vortex an electron rather than a generic topological defect.

Each atom's colour encodes its phase θ = atan2(y, x), which depends only on the two horizontal coordinates — not on z. That z-independence is what makes this a vortex line (along the z-axis through the origin) rather than a point. Burgundy atom outlines mark A-sublattice sites; blue outlines mark B-sublattice. Drag to rotate, shift+drag to pan, scroll to zoom.

View down [001]: camera snaps to look down the z-axis. You see the pinwheel face-on — a rainbow winding around the vortex core, exactly like Figure 2 of Chapter 7, only now painted on the diamond lattice rather than a 2D grid. Both sublattices carry the winding indifferently. View from side: camera snaps to look along the y-axis. The vortex line becomes vertical on screen, with the rainbow pinwheel repeating at every horizontal slice — a stack of rainbow disks threaded by the line. That repetition is the whole point: a vortex in 3D carries its topology along its entire length. Let them oscillate: advances every phase by a common ω·dt per frame. The colours cycle through the hue wheel — every atom's phase is evolving in time — but the pinwheel pattern stays put. The winding is a topological invariant: it doesn't care that the oscillators are oscillating, only about the spatial winding of phase around the core.

Rotate the lattice and look at the vortex from every angle. Down the z-axis: a rainbow pinwheel, just like §7 Figure 2. From the side: that pinwheel repeats in every horizontal slice of the crystal, stacked into a tube. The vortex is a line, and every plane perpendicular to that line carries the same 2π winding. Annihilation — which took out our 2D vortex pair at step 40k — would require this whole line to collapse into its antipartner along every slice simultaneously. The free energy cost is now proportional to the line's length; long lines are effectively immortal on CLR timescales.

Notice also how the rainbow is painted on top of the diamond's bipartite A/B structure. The vortex itself doesn't care about sublattice — phases wind equally around A and B atoms — but the bound state will. In Chapter 9, we'll see that the quantum state localised at this vortex line has its amplitude concentrated on just one of the two sublattices. That is the chiral downfold; it is what turns a U(1) phase defect on a bipartite lattice into a spin-½ fermion. In other words, it is what makes this line an electron.

Where this leaves us

Diamond has been selected, not chosen. Every filter is independent; every filter is necessary. A paper that tried to derive α on simple cubic, BCC, FCC, or any 2D lattice would fail at one filter or another. The Ad family (the simplex lattices) generalises the story to arbitrary dimension: d-diamond always has coordination z = d+1 and always satisfies the five filters in that dimension. We'll come back to this in Chapter 13 when we ask what changes if we turn a "d-dial."

For now, we have the lattice. In Chapter 9 we put the vortex on it and see the electron assemble: charge from winding, the Dirac equation from the bipartite Bloch structure, spin-½ set up by the frame sector, Clifford algebra as the geometry of non-commutativity. Then Chapter 10 finds where α lives, at the BKT wall.