The proof without technical machinery

This chapter gives the whole argument in the language of patches, bounded orbit transport, and holonomy. It is intended to be read before the formal proof. Nothing proved later is being replaced: the two technically difficult bridges are stated here as named engines and proved in Part III.

Reader card. This chapter assumes familiarity with tiling hulls or Cantor dynamical systems, clopen patch cylinders, and invariant measures. It does not assume \(KK\)-theory. Its output is a map of the proof and a precise explanation of where the technical machinery enters.

1. The question from gap labelling

Let \(\Sigma\) be the Cantor transversal of an aperiodic pattern and let \(\mathbb Z^p\) act by translations. A finite-range covariant Hamiltonian belongs to, or is represented in, the crossed product

\[ A_{\Sigma,\Theta} =C(\Sigma)\rtimes_{\sigma_\Theta}\mathbb Z^p. \]

If a Fermi energy lies in a spectral gap, its Fermi projection determines a class in \(K_0(A_{\Sigma,\Theta})\). Pairing this class with an invariant measure gives the integrated density of states. The constructive problem is therefore not merely to predict a number, but to exhibit a finite pattern-equivariant projective representative having that trace.

The labels targeted in this book have the form

\[ \text{magnetic Pfaffian}\times\text{pattern frequency}. \]

The two factors come from different pieces of geometry. The pattern frequency is absorbed into a coefficient module. A Rieffel--Heisenberg module supplies the magnetic factor. Its abbreviated phase-space and lattice construction is given in A short introduction to Rieffel--Heisenberg modules.

2. Choose one height direction

Split the translation lattice as

\[ \mathbb Z^p=G\oplus\mathbb Zh, \qquad G\cong\mathbb Z^{p-1}, \]

and assume that the homeomorphism \(T_h\) is minimal. The height algebra is

\[ B_h=C(\Sigma)\rtimes_{T_h}\mathbb Z. \]

At the level of \(K\)-theory, integer-valued locally constant functions are identified when their difference is a height coboundary:

\[ K_0(B_h) \cong C(\Sigma,\mathbb Z)_{T_h} ={} \frac{C(\Sigma,\mathbb Z)}{(T_h-1)C(\Sigma,\mathbb Z)}. \]

Geometrically, a term \((T_h-1)a\) moves a finite, locally determined amount of integer mass along height orbits. Coinvariants therefore remember the amount of pattern mass but forget bounded redistributions in the height direction.

The input to the theorem is a class

\[ [f]\in \left(C(\Sigma,\mathbb Z)_{T_h}\right)^G. \]

The superscript does not say that a particular function \(f\) is fixed by every transverse translation. It says that translating \(f\) changes it by a height coboundary.

3. Turn a signed pattern weight into a projection

The function \(f\) may take negative values. Choose \(N\) so that

\[ g=f+N\geq1. \]

From a nonnegative integer-valued function \(g\), form a diagonal clopen projection \(P(g)\) whose pointwise rank is \(g(x)\). Then

\[ [P(g)]-[1_N]=[f]. \]

This is why the natural output is sometimes a graded module: an arbitrary integer-valued pattern class is a difference of two positive finite projective modules. For positive inputs, the construction can produce an ordinary ungraded module.

4. What is new: the transverse structure, not the projection

The projection in the preceding section is not by itself the new object. A clopen projection already represents a positive pattern-frequency class in the untwisted height algebra, and a graded difference represents an arbitrary integer-valued class. If we write this coefficient module as \(F_f\), then

\[ [F_f]=x_f\in K_0(B_h), \qquad \tau_\mu([F_f])=\mu(f). \]

The real construction is to enhance \(F_f\) by strict transverse transport. For every \(s\in G\), we seek a finite-propagation PE operator \(V_s\) such that

\[ V_s(\xi b)=V_s(\xi)\alpha_s(b), \qquad V_sV_t=V_{s+t}. \tag{O.A} \]

The first identity says that \(V_s\) covers the transverse action on the height algebra. The second says that the transports form an honest group action, rather than a collection of maps agreeing only in \(K\)-theory or up to a phase. Chapters 1--5 are devoted to constructing these operators and removing their holonomy. Thus the first substantive output is the pair

\[ (F_f,V), \]

not merely the underlying projection defining \(F_f\).

This enhancement is designed to accept a transverse magnetic module. Put

\[ A=B_h\rtimes_{\alpha,\sigma}G, \qquad D=C^*(G,\sigma), \qquad X_f=F_f\widehat\otimes_{B_h}A. \]

The coupled operators

\[ L_s(\xi\otimes a)=V_s\xi\otimes U_sa \]

give \(X_f\) a left \(D\)-action. Hence a transverse Rieffel--Heisenberg module \(E_T\) can be balanced against it:

\[ [f] \longmapsto F_f \longmapsto (F_f,V) \longmapsto \mathcal E_{T,f}=E_T\widehat\otimes_DX_f \longmapsto \mu(f)\tau_D([E_T]). \tag{O.B} \]

Keeping the two actions independent would give only an external product in which the pattern factor is passive. Balancing over \(D\) identifies each time-frequency shift on \(E_T\) with the same transverse translation that applies the pattern-dependent transport \(V_s\). This is what makes \(\mathcal E_{T,f}\) a genuinely mixed quasicrystalline module.

This perspective also clarifies the comparison with Benameur--Mathai. Their index theorem detects the relevant trace labels abstractly, while Hypothesis (H) supplies a rigid finite-sheet model for the strict transports needed in their explicit matrix homomorphism. The construction here replaces those sheets by class-dependent PE transports. When Hypothesis (H) holds, its sheet permutations should be regarded as a special source of the \(V_s\). After forgetting transverse equivariance, both constructions represent the same coefficient class in \(K_0(B_h)\). The book does not, however, infer equality of the descended modules merely from equality of their traces. A complete comparison would have to prove, with compatible choices and conventions,

\[ [\Phi_\Lambda(e)]=[\lambda_{F_\Lambda}(e)] \quad\text{in }K_0(A) \]

for the relevant transverse projection \(e\).

Finally, allowing every invariant height-coinvariant \(f\) and every transverse torus class realizes the full subgroup

\[ D_h(\mu),\tau_D(K_0(D)) \]

associated with the chosen height-one splitting, where

\[ D_h(\mu) =\mu!\left( \bigl(C(\Sigma,\mathbb Z)_{T_h}\bigr)^G \right). \]

The product denotes the additive group generated by products of elements from the two displayed groups. In arbitrary dimension this is a constructively realized height-one layer of the Benameur--Mathai frequency group, not automatically the whole predicted group. The remaining coefficient layers require higher-rank absorption.

5. Transverse invariance becomes bounded flow

For every generator \(T_i\) of \(G\), invariance of \([g]\) means that a locally constant integer function \(h_i\) satisfies

\[ (T_i-1)g=(T_h-1)h_i. \tag{O.1} \]

After adding a constant, take \(h_i\geq0\). Equation (O.1) is a conservation law. The discrepancy between \(T_i(g)\) and \(g\) is exactly the boundary of a bounded flow in the height direction.

Because the functions are locally constant and bounded, the matching can be performed on a finite clopen partition. It produces a partial isometry

\[ W_i:\alpha_i(P)B_h^n\longrightarrow PB_h^n \]

whose entries are finite sums of clopen-weighted height shifts \(1_CU_h^m\). In tiling language, \(W_i\) is a finite-range rule: a bounded patch around a point determines how far that component is moved along its height orbit.

The construction proves more than finite Fourier support. Each \(W_i\) is supported on a compact-open bisection, so it has a well-defined geometric symbol and normalizes the diagonal algebra of clopen sets.

6. The transports may have holonomy

Individual transports do not yet define a \(G\)-action. For two transverse generators, compare the two routes around a square. Their ratio is

\[ \Omega_{ij} =W_i\alpha_i(W_j) \bigl(W_j\alpha_j(W_i)\bigr)^*. \tag{O.2} \]

If \(\Omega_{ij}=1\), the two transports commute. In general, \(\Omega_{ij}\) is the holonomy around one transverse square.

The transfer functions have a parallel, purely integer-valued curvature:

\[ \kappa_{ij} =(T_j-1)h_i-(T_i-1)h_j. \tag{O.3} \]

Applying \(T_h-1\) and using (O.1) shows that

\[ (T_h-1)\kappa_{ij}=0. \]

Thus \(\kappa_{ij}\) is constant on each height orbit. Minimality of \(T_h\) makes it constant on all of \(\Sigma\). An invariant measure gives

\[ \int_\Sigma\kappa_{ij}\,d\mu=0, \]

so the constant is zero:

\[ \kappa_{ij}=0. \tag{O.4} \]

This short calculation is the dynamical reason the theorem works.

7. Technical Engine A: certify that the operator obstruction vanishes

Equations (O.2) and (O.3) live in different worlds. The first is a unitary in a crossed-product corner; the second is an integer-valued function. The load-bearing comparison theorem proves

\[ \partial_h[\Omega_{ij}] =\pm[\kappa_{ij}], \tag{O.5} \]

with the target and quotient interpreted precisely. It follows from (O.4) that

\[ [\Omega_{ij}]=0 \quad\text{in the actual corner }K_1. \tag{O.6} \]

This is where the PV sequence, Barlak's \(d_2\), the equivariant Toeplitz triangle, and enhanced \(KK\)-theory enter. They do not create the transports. They certify that the concrete holonomy built above has zero index. The full proof of this bridge is isolated in Technical Engine A.

8. Replace zero-index holonomy by exact commutation

Vanishing in \(K_1\) is still weaker than the equality \(\Omega_{ij}=1\). The next step uses the geometry of a single minimal Cantor homeomorphism.

The full diagonal corner can be written as the transformation groupoid of a coloured minimal successor map. The bisection symbols of the \(W_i\) normalize its topological full group. The Giordano--Putnam--Skau retraction separates such a normalizer into an index-zero orbit rearrangement and a part commuting with the successor.

Because (O.6) says that the orbit rearrangement has zero index, finite-PE gauges can absorb it. One obtains corrected transports whose underlying bisection symbols commute simultaneously. This is the strictification step, developed in the GPS chapter.

9. Separate permutation curvature from magnetic phase

For a block-supported magnetic form,

\[ \Theta_{ih}=0 \qquad(i\in G), \]

transverse automorphisms do not rotate the height unitary. Equality of the corrected bisections then gives equality of the corresponding operators. The coefficient module carries a strict \(G\)-equivariant structure.

For a general magnetic matrix, the corrected operators may differ by a locally constant diagonal phase. The remaining datum is a cocycle

\[ \nu\in Z^2\bigl(G,U(C_{\mathrm{lc}}(Y))\bigr). \]

In a fixed corner, diagonal gauges remove it exactly when \([\nu]=0\). The book does not claim that this phase always vanishes, nor that the phases from all choices form a choice-independent stable invariant.

10. Technical Engine B: attach the magnetic module

Let \(F_f\) be the strict coefficient module just constructed, and let \(E_T\) be a Rieffel--Heisenberg module over the transverse twisted group algebra. Equivariant descent forms a correspondence \(X_f\), and then

\[ \mathcal E_{T,f} =E_T\widehat\otimes X_f. \]

Conceptually, every transverse magnetic time-frequency shift is accompanied by the finite-PE height transport determined by the local pattern. The analytic details prove that the balanced action is well defined, adjointable, compact, and finite projective. They also prove

\[ \tau_\mu([\mathcal E_{T,f}]) =\mu(f)\tau_{\Theta_G}([E_T]). \tag{O.7} \]

The conceptual construction is explained in Attaching the magnetic Heisenberg module; the complete verification is in Technical Engine B.

11. What the theorem contributes

If the torus class isolates a Pfaffian term, (O.7) becomes

\[ \operatorname{Pf}(\Theta_I)\mu(f). \]

For nonconstant clopen data \(f\), this is not merely a noncommutative-torus class pulled back to the quasicrystal algebra. The projective module itself contains local pattern-dependent transport.

The result is a constructive lower bound for magnetic gap labelling. It does not by itself compute all of \(K_0\), prove the upper containment, or show that every allowed label is an open gap of one particular Hamiltonian.

The proof in one diagram

flowchart TB
    patch["locally constant pattern weight f"]
    coinv["G-invariant height coinvariant [f]"]
    flow["bounded height flows h_i"]
    transport["finite PE bisection transports W_i"]
    holonomy["square holonomy Omega_ij"]
    curl["integer curl kappa_ij"]
    kk["PV / Barlak comparison"]
    gps["GPS finite-PE gauge correction"]
    phase["residual diagonal magnetic phase"]
    strict["strict equivariant coefficient module F_f"]
    rieffel["Rieffel--Heisenberg module E_T"]
    mixed["mixed PE projective module"]
    trace["Pfaffian times pattern frequency"]

    patch --> coinv --> flow --> transport --> holonomy
    flow --> curl
    curl -->|"minimality: zero"| kk
    holonomy --> kk -->|"zero K_1 index"| gps
    gps --> phase
    phase -->|"block support, or phase coboundary"| strict
    strict --> mixed
    rieffel --> mixed --> trace

The running Sturmian example makes the flow and curvature steps explicit before the general proof is read.