Spectral cutoffs, finite rank and the order of limits

Name the operation before using its spectrum

For a self-adjoint D, the spectral projection P_I=1_I(D) selects a subspace. A smooth filter χ(D) instead weights spectral components. A windowed trace Tr[χ(D)²e^−tD²] measures a weighted part of the original spectrum. Finally D_χ=χ(D)Dχ(D) is a new operator whose eigenvalues are changed.

These expressions can all be useful, but they are not interchangeable. Their domains, kernels and limiting traces can differ dramatically. A smooth function’s regularity does not make its values equal to zero or one, and the notation “spectral window” does not itself specify a retained Hilbert space.

The temporal NCG source uses smooth scale windows to formulate possible dimensional flow and later places them around protected projections. The calculations here make the additional distinctions explicit. They do not discard the window proposal; they identify the exact mathematical object that a subsequent geometric or physical claim must use.

Source: Noncommutative Geometry, PDF pp. 25, 30.

Check idempotence in three dimensions

Take D=diag(0,1,2), P=diag(1,1,0) and χ(D)=diag(1,1/2,0). These matrices commute. Nevertheless the weighted band W=χ(D)Pχ(D) is diag(1,1/4,0), so W²=diag(1,1/16,0).

The filter has preserved the support of the two selected levels while attenuating one of them. It has not produced the orthogonal projector onto that support. Repeatedly applying it changes relative amplitudes, whereas repeated application of P leaves every retained vector unchanged.

If 0≤χ≤1, it can be a measurement Kraus operator with success probability ∥χ(D)ψ∥². A complementary operator √(I−χ(D)²) completes a trace-preserving two-outcome instrument. Conditioning on success includes state-dependent normalization. That is a declared measurement protocol, with a different meaning from choosing an energy subspace once and evolving inside it.

Source: Noncommutative Geometry, PDF p. 30. The matrix and measurement calculations are developed here.

A contour gives a genuine protected projection

Suppose a contour Γ encloses one isolated cluster of H(q) and stays in the resolvent set for every q in a parameter neighborhood. The Riesz formula gives

Functional calculus makes P²=P and P†=P. Its rank stays constant while the contour continues to separate the same cluster. A threshold crossing breaks that assumption. For example H(q)=diag(q,2), selected by negative energy, has rank one for q<0 and rank zero for q>0. No constant-rank bundle extends through that selection change.

In finite matrices, differentiating the resolvent is straightforward. For unbounded families it requires a compatible domain or resolvent regularity hypothesis. The resulting protected bundle is the correct input to Kato transport and projector curvature. Replacing it by a smooth filter can be useful for estimates, but it does not retain the same exact bundle structure without another projection step.

Source: Vacuum Obstructions, PDF pp. 22–23.

A gap alone does not control geometric derivatives

Let R(z)=(z−H)⁻¹. Differentiation gives ∂R=R(∂H)R. If the contour length is L and its distance from the spectrum is at least d, then the finite-matrix estimate is

The numerator matters. In the rotating model P(q)=u(q)u(q)† with u(q)=(cos kq,sin kq), the Hamiltonian H=Δ(I−P) has gap Δ for every k. Yet ∥∂P∥=|k| and ∥∂H∥=Δ|k|. Keeping the gap fixed does not prevent arbitrarily rapid movement of the protected space.

Curvature, projection loss and geometric kinetic terms depend on those derivatives. A useful uniform band theorem must therefore bound the variation as well as the separation, and must cover the intended parameter and gauge charts. The graph model supplies another exact witness: changing its positive gap leaves the projector and curvature unchanged.

Source: Sevenfold Gauge, PDF pp. 24–27.

Source: Vacuum Obstructions, PDF pp. 22–23.

Windowing an operator can create infinitely many zero modes

On ℓ²(Z), let D|n⟩=n|n⟩ and let P_N select |n|≤N. The operator D_N=P_NDP_N, viewed on the full infinite Hilbert space, sends every |n|>N mode to zero. Its kernel is infinite dimensional.

Consequently Tr e^−tD_N² is infinite for every positive t: each excluded mode contributes one. Restricting D_N to P_NH gives a different operator on a finite Hilbert space, with a finite heat trace. The weighted trace Tr[P_Ne^−tD²] is finite as well, because it discards the excluded modes in the trace rather than converting them to zero-energy states.

The same issue occurs for a smooth compactly supported filter. This is an explicit caution when reading χ(D)Dχ(D) as a new full-space Dirac operator. Its relation to a windowed spectral density, a compressed domain or a physical low-energy theory must be supplied. A formal operator expression cannot carry all three meanings simultaneously.

Source: Noncommutative Geometry, PDF p. 25. This infinite-kernel consequence is calculated here.

Calculate the finite heat trace and its dimension

On the retained circle modes, the exact heat trace is Z_N(t)=Σ_n=−N^Ne^−tn². At fixed N it tends to 2N+1 as t→0. Differentiating its logarithm gives the scale-dependent spectral dimension

It tends to zero in the extreme ultraviolet at fixed cutoff. It also tends to zero at very large t because only the single zero mode remains. A plateau near one can occur between those regimes, when N⁻²≪t≪1.

For the untruncated circle, the Gaussian Fourier identity gives Z_∞(t)=√(π/t)Σ_k∈Ze^−π²k²/t, and hence its small-t dimension is one. The infinite spectrum and the finite spectrum have different ultraviolet limits. In a product of three identical mode directions the heat traces multiply and the intermediate dimensions add, conditional on that supplied product structure.

Source: Noncommutative Geometry, PDF pp. 24–25. The circle calculation exposes the regulator crossover behind a spectral-dimension plateau.

Specify how the cutoff moves with the probing scale

Let N increase while t decreases. Writing u=n√t turns √t Z_N(t) into a Riemann sum. If N√t→a with finite nonnegative a, its limit is ∫_−a^ae^−u²du. Recovering the full coefficient √π requires N√t→∞.

A quantitative tail estimate is available for N>0:

Thus a plotted plateau can be accompanied by a chosen truncation tolerance. A visual slope alone does not establish a continuum law. One must state whether derivatives defining an effective dimension hold the regulator fixed or vary it with t; these are different procedures at cutoff transitions.

Fractal or anomalous spectral exponents require their own infinite-sector counting or trace asymptotics. They cannot follow solely from a protected band’s finite rank, nor can a finite number of retained modes determine a physical spacetime dimension without a separate reconstruction map.

Source: Noncommutative Geometry, PDF pp. 24–25. The joint limit and tail bound are worked calculations.

Keep the cutoff profile separate from physical matching

Even after a valid heat expansion is available, an action profile supplies further parameters. For f(u)=a e^−bu, with a,b>0, the spectral moments are F₀=a, F₂=a/b and F₄=a/b². The same operator and retained group can therefore receive different gauge, gravitational and volume coefficients.

In the source’s four-dimensional spectral benchmark, gauge inverse squares scale with F₀ times representation indices. The gravitational term also depends on F₂ and the cutoff scale. A choice of filtering or spectral weighting does not determine observed coupling strengths unless the physical rule selecting that profile is established.

The working sequence is therefore explicit: select the operator and trace, define the cutoff operation, prove the appropriate asymptotics, and only then match a normalized physical action. Smooth geometry, heat-kernel hypotheses and physical continuation remain inputs. The six inner generative levels and seventh completion role do not replace any of these operator or limiting conditions.

Source: Interactions Unification, PDF pp. 15–17, 21–24.

Continue reading

• Check the retained observable products
• Control transport of a selected band
• Match heat coefficients to an action
• State the continuum comparison requirements