G4 · Architecture · Dimensional Theory · Book 4 · Ch 20 · no operator assignment — see §20.0
← Ch 19 · The Acoustic Lattice Companion · Book 6 · Quantum Weave →
Principia Orthogona · Volume IV · Architecture Arc
Chapter 20 · Codimension · What a Lattice Cannot Carry

The Defect Lattice

Torsion as the density of a broken lattice — and a codimension count that decides whether two mechanisms can be the same object

$T^a \sim \sum_i b_i^a\,\delta^{(2)}(\Sigma_i)$  ·  $A(N) \propto N^{-1}$
op2.py · op2b.py 2+1 dimensions Katanaev–Volovich · Teo–Kane One result, one refutation
§ 20.0 · In Plain Terms

Before the Mathematics

The Architecture arc built a lattice, gave it spin, shook it, and listened to it. This chapter breaks it. A crystal with a fault line in it is not a failed crystal; it is a different geometry, and the mathematics that describes the fault is the same mathematics that describes a curved and twisted spacetime. That correspondence is old and well established. What is new here is a number and a refutation.

This chapter deliberately makes no operator assignment. The Chapter 7 post-mortem in the repository ledger records what happens when a chapter re-reads G = UFKC in a new domain while leaving the Volume I §3 signatures behind: the operators get re-glossed as a map, an event, a property and a theorem, and composability quietly dies. Defect geometry is a domain where that temptation is strong — a dislocation looks like a threshold crossing. The chapter declines the reading. Nothing below depends on the operator chain, and nothing below should be cited as extending it.

What follows: how a broken lattice becomes a curved geometry, the one thing that survives coarse-graining exactly, the one thing that only survives statistically and at what rate, and a codimension count that kills a conjecture.

The claim of this chapter
Coarse-graining a lattice with defects yields a torsion 2-form that is exactly independent of how you partition space, because the Burgers vector is a winding number. Its rotational covariance is not exact — it emerges statistically, with residual point-group anisotropy falling as $1/N$ in the number of defects per coarse-graining cell. Continuum torsion is therefore a high-density statistical object, not a property any single defect possesses. Separately: a codimension count shows that two mechanisms often identified in the literature cannot be the same object.

§ 20.1

A Broken Lattice Is a Curved Geometry

Take a perfect crystal and remove a half-plane of atoms, then close the gap. Walk a circuit around the edge of that missing half-plane, counting lattice steps: north four, east four, south four, west four. In a perfect crystal you return to where you started. Here you do not. The circuit fails to close, and the failure is a vector — the Burgers vector $b^a$. This is a dislocation, and the failure of a loop to close under parallel transport is precisely the definition of torsion.

Kleinert, and Katanaev and Volovich in covariant form, made this exact. A medium containing a continuous distribution of line defects is described, in the continuum limit, by a Riemann–Cartan geometry: an affine connection carrying both curvature and torsion. The dictionary is not loose. Dislocations, characterised by a Burgers vector — a failure of closure under transport — correspond to torsion:

$$T^a \;\sim\; \sum_i b_i^a\,\delta^{(2)}(\Sigma_i)$$

a surface density concentrated on the defect lines $\Sigma_i$. Disclinations, characterised by a Frank vector — a failure of frames to close rather than points — correspond to curvature in the same way. A defect-free crystal is, in this dictionary, exactly the torsion-free flat limit. Ruggiero and Tartaglia made the corresponding statement in a general-relativistic rather than condensed-matter setting: Einstein–Cartan theory can be read as a theory of continuously distributed spacetime defects.

[VERIFIED] — this correspondence is established literature, not a contribution of this chapter. What follows is.


§ 20.2

What Survives Coarse-Graining Exactly

The dictionary above is written in the continuum. The physical substrate is discrete. Between them sits a coarse-graining step that is almost always assumed rather than performed: partition space into cells, average, and read off a density. The obvious worry is that the answer depends on the partition — that a different cell size or a shifted origin gives a different torsion. If so, the object is not a geometry at all; it is an artefact of bookkeeping.

It does not. Integrating the torsion 2-form over a fixed region, across cell counts from 4 to 64 and three independent origin offsets, the deviation from the exact enclosed value is:

LatticeMaximum deviationStatus
Square (4-fold)0.000 × 100exact
Triangular (6-fold)1.155 × 10−14machine precision

This is not numerical luck, and it should not be reported as an empirical finding. The Burgers vector is a winding number of the discrete frame around a circuit — a topological invariant of the lattice connectivity. A topological invariant cannot depend on how the region containing it is subdivided. The computation confirms a theorem rather than discovering a fact, which is the correct relationship between the two.

Result 20.1 · scheme independence  [VERIFIED]
The integrated torsion 2-form over a fixed region is independent of the coarse-graining cell size and origin, exactly, because the enclosed Burgers vector is a winding number of the discrete frame.

§ 20.3

What Survives Only Statistically — and How Fast

Scheme independence is necessary but nowhere near sufficient. General covariance requires the frame index $a$ on $T^a$ to be a genuine frame index. In a lattice it is not. The Burgers vector is quantised in the lattice translation group: for a square lattice it takes one of four values, for a triangular lattice one of six. A discrete set of vectors is not closed under continuous rotation. A single defect's torsion is therefore not a Lorentz vector, and no amount of care in the coarse-graining will make it one.

What can rescue it is density. If a cell contains many defects with independently drawn Burgers vectors, the coarse-grained sum may forget the point group even though no individual term does. Whether it does, and how fast, is a computation.

Measuring the residual $m$-fold anisotropy $A$ of the coarse-grained torsion against $N$, the number of defects per cell — three million samples per point, debiased in quadrature against a matched isotropic control so that the sampling floor is not mistaken for signal:

LatticePoint groupTail exponent ($N \geq 16$)$A$ at $N=1$
Square4-fold−0.9901.000
Triangular6-fold−0.9861.000

The exponent is $-1$, and it is not fitted from the numerics alone. For the square set the characteristic function is $\varphi(k) = (\cos k_x + \cos k_y)/2$, whose expansion carries an anisotropic quartic term $k_x^4 + k_y^4 = \tfrac{3}{4}k^4 + \tfrac{1}{4}k^4\cos 4\theta$. Summing $N$ independent defects raises $\varphi$ to the $N$th power; evaluating at the typical scale $k \sim N^{-1/2}$ leaves a relative anisotropy of order $1/N$. Analytics and numerics agree to better than two percent.

Result 20.2 · emergent rotational covariance  [VERIFIED] analytics · [SIMULATION] numerics
The rotational covariance of coarse-grained torsion is not exact but emergent, with residual point-group anisotropy $A \propto N^{-1}$ in the defect count per coarse-graining cell. At $N = 1$, $A = 1$: maximally anisotropic.
What this buys
Continuum torsion is a high-defect-density statistical object. It is not a property of a defect; it is a property of a population of them. Any substrate proposing to generate spacetime geometry from discrete defects therefore acquires a quantitative condition — $N \gg 1$ defects per coarse-graining cell — and a prediction: the residual anisotropy of the resulting geometry is of order $1/N$. That is a falsifiable number where previously there was an assumption.
What this does not establish  [OPEN]

The test is rotational, in Euclidean 2+1. It says nothing about boosts. The averaging argument works because the rotation group is compact and the point group's orbit admits a uniform average; the Lorentz group is non-compact and admits none. The argument does not extend, and should not be reported as though it does.

The consequence is structural. A programme listing "no discrete-to-continuum map" and "emergence of continuous symmetry from discrete data" as independent next steps, to be taken in order, has mis-stated its own dependency graph. The compact half of the first is now closed; the non-compact half is the second. The first cannot be finished before the second, and the second is a known-hard problem in loop quantum gravity generally.


§ 20.4

A Codimension Count That Kills a Conjecture

The defect dictionary invites a further step, and it has been taken in the literature more than once: if torsion is carried by defect lines, and if matter can also be modelled as defects in a discrete pre-geometric substrate — braided ribbon excitations, in the preon programme — then perhaps the defects sourcing torsion and the defects carrying matter quantum numbers are the same objects, described in two languages.

It is an attractive conjecture. It fails on a count that requires no dynamics, no action principle, and no commitment to any framework: codimension.

IngredientDefectCodim in 3 space dimsBound modeStatistics
Torsion (Katanaev–Volovich, Kleinert)dislocation line2
Ran–Zhang–Vishwanath (2009)screw dislocation line in a 3D TI21D helicalAbelian
Teo–Kane (2010)point defect in 3D3Majorana zero modenon-Abelian

Torsion requires codimension-2 line defects. The non-Abelian statistics that make the identification interesting come from Teo and Kane, whose construction is codimension 3 — point defects in three spatial dimensions, where the non-Abelian character is generated by a framing degree of freedom that can rotate nontrivially. The only codimension-2 result available is Ran–Zhang–Vishwanath, and its bound mode is helical: Abelian.

A conjecture identifying the two is identifying objects of different codimension. It is worth being precise about the arithmetic, because the error is easy to make and appears in print: a point in three spatial dimensions has codimension 3, not 2. A line has codimension 2.

Refutation 20.3  [VERIFIED — negative]
In three spatial dimensions, the defect class carrying torsion (codim 2) and the defect class carrying non-Abelian zero modes in the cited construction (codim 3) are distinct. Any conjecture identifying them must either restrict to the Abelian codim-2 case and abandon the non-Abelian claim, or supply a codimension-2 framing mechanism, which does not exist in the cited literature. This is a refutation of an identification, not of either ingredient.

This is the kind of result the arc is for. It costs one afternoon of counting and it closes a direction that would otherwise absorb years. A negative result with a clean argument is worth more than a positive one with a hedge — and it leaves both underlying mechanisms entirely intact, which is the honest thing to say about it.


§ 20.5

Honest Inventory

ClaimStatusBasis
Defect ↔ Riemann–Cartan dictionary[VERIFIED]Kleinert; Katanaev–Volovich; Ruggiero–Tartaglia
Scheme independence of integrated $T^a$[VERIFIED]topological; confirmed to $10^{-14}$
$A \propto N^{-1}$ anisotropy decay[VERIFIED] analytic + [SIMULATION]characteristic function; 3×106 samples/point
Codimension mismatch[VERIFIED]Teo–Kane; Ran–Zhang–Vishwanath
Boost covariance of coarse-grained $T^a$[OPEN]non-compact; argument does not extend
Extension of 2+1 result to 3+1[OPEN]not attempted here
Any action principle for a defect substrate[OPEN]none written down

Reproducibility: the producing scripts are op2.py (scheme independence; falsification criterion recorded in the file before the run) and op2b.py (anisotropy scaling, chunked and noise-debiased). The criterion was fixed in advance precisely so that a scheme-dependent result would have been recorded as a refutation rather than reinterpreted.


§ 20.6

Where This Sits

Chapter 16 built the lattice and tuned it onto a resonance. Chapter 17 gave it spin, Chapter 18 shook it, Chapter 19 listened to it. This chapter asks what the lattice does where it fails, and finds that the failure is not an absence of structure but a different structure — with a geometry of its own, and a density threshold below which that geometry does not exist.

The companion reading is Book 6 · Quantum Weave Topology, whose §9 treats the weave as a substrate and whose §7 treats holonomy of holonomies; a weave is the object this chapter's coarse-graining acts on. Readers coming from Chapter 7 · Topological Orthogenesis should note the correction recorded there: the braid-group identification in that chapter's Theorem 7.1 was withdrawn, and the anyon material was re-scoped. This chapter touches the same literature — Teo–Kane, non-Abelian statistics, framing — and reaches a negative result about it. The two are consistent, and both are more careful than the version that preceded them.

Exercises

20.1  Show that the set of finite paths on a square lattice is countable, and that the Burgers vector of a closed circuit depends only on the homotopy class of the circuit in the punctured plane.

20.2  Compute the anisotropy exponent for a hexagonal lattice with a 3-fold Burgers set. Predict it before running the code; the characteristic-function argument gives it in two lines.

20.3  A weave contains one defect per coarse-graining cell. What is the residual anisotropy, and what does that imply about the smallest scale at which a continuum geometry can be said to exist on that substrate?

G6 LLC  ·  g6llc@proton.me  ·  +1 (646) 342-3751
Proved · kernel-checked
discriminant book21/Spiral.lean:75 Each name above is declared in this repository at the line shown and appears in an axiom report with no sorryAx. A clean axiom report is not a reading of the statement: per R20, a theorem can assume its conclusion and still report clean. Follow the link before citing one as evidence.