Preprint
Article

This version is not peer-reviewed.

Uniform Source Stability for Wilson Fermions

Submitted:

14 November 2026

Posted:

16 November 2026

You are already at the latest version

Abstract

We prove source-stability bounds for Wilson lattice gauge theory with degenerate sea-quark pairs, at arbitrary gauge coupling and hopping parameters 0 < κf ≤ κ* < 1/6. For signed, gauge-dependent spatial sources of finite canonical cost, a determinant-minor expansion gives quadratic total-variation control of the full normalized gauge law without a uniform Dirac spectral gap. The same sea pair controls the corresponding fermionic source coefficients. Two identical original pairs give quartic forward relative entropy, and quartic bounds in both directions for nonnegative sources. Determinant models locate the limits of one-pair control, including a cubic entropy law at a regular zero. Spatial heat kernels give uniform costs at fixed positive smearing and an explicit joint source-removal schedule. For overlap fermions, separated-time source coefficients admit exact contact conversion; the full-spacetime Hilbert–Schmidt contact has a sharp smearing threshold. A sharp background-uniform Wilson propagation bound gives explicit energy floors for vacuum-generated flavor-charge sectors when κf < 1/8. A certificate using three positive-time correlations detects finite-energy spectral weight. These results extend source comparison beyond convergent hopping expansions and identify the physical information that a nontrivial continuum limit must retain.

Keywords: 
;  ;  ;  ;  

1. Introduction

Fermion sources perturb both the determinant weight and its normalization. A comparison useful for continuum limits must control the resulting gauge law and fermionic correlations under the same sea determinants. Configuration-uniform inverse bounds provide such control in a massive hopping regime, but lose their margin as the Dirac spectrum approaches zero. We obtain the comparison directly from the original flavor pairs and the positive transfer operator.
For the standard Wilson plaquette theory with degenerate sea-quark pairs, all with 0 < κ f ≤ κ * < 1 / 6 , we prove
TV ( ν z , ν 0 ) ≤ I 0 ( 2 | z | M ) − 1 = z 2 M 2 + O ( z 4 M 4 ) .
Here ν z is the normalized gauge law after sourcing one pair, ν 0 is the source-free law, z is real, and M is an explicit canonical source cost. The function I 0 ( 2 h ) = ∑ k ≥ 0 h 2 k / ( k ! ) 2 is the modified Bessel function. At fixed M, the bound is independent of volume and gauge coupling. It requires neither a uniform Dirac gap nor a Gibbs-contraction hypothesis. The admissible interval crosses the free value κ = 1 / 8 ; identifying an interacting critical trajectory remains a separate dynamical problem.
The physical Wilson transfer construction [1,2] controls products of squared inverse minors through the actual sea flavors. A signed Cauchy–Binet expansion then bounds the determinant increment before normalization. With p identical original pairs, this controls the first p absolute moments of the likelihood increment. Two pairs give quartic forward relative entropy, and both orientations for nonnegative sources. A regular determinant zero gives a sharp local moment threshold. Explicit models distinguish a cubic entropy law at a fixed zero from failure of uniform entropy convergence under a varying original law.
The sources are opposite imaginary Hermitian perturbations of Q = γ 5 D , of the familiar twisted-mass type [3]. Determinant fluctuations and normalized reweighting moments are established tools [4], and symmetric mass shifts already have quartic leading fluctuations [5]. The contribution here is a nonasymptotic bound for signed, gauge-dependent spatial sources, with the moment range supplied by flavors already present in the sea. The same argument controls actual Grassmann source derivatives. Spatial heat kernels give its cost in physical units.
We then determine which contact and spectral information these comparisons preserve. The Ginsparg–Wilson contact identity [6,7,8] gives exact conversion of separated-time overlap sources, while the full Hilbert–Schmidt contact has a sharp threshold as the smearing scale tends to zero. Each regulator is treated under its own sea law. The positive-measure strategy of Vafa and Witten [9] yields an explicit Wilson energy floor for vacuum-generated flavor-charge sectors. This estimate applies below κ = 1 / 8 [10], within the larger interval of the source-stability theorem. A final certificate, using three positive-time correlations, tests whether finite physical energy survives the comparison.

2. Wilson Measure and Determinant Minors

Work first on a finite four-dimensional torus, with gauge group SU ( 3 ) , spin dimension four, periodic gauge links, periodic spatial fermions and antiperiodic temporal fermions. The gauge action is the standard Wilson plaquette action at any β > 0 . Every fermion flavor belongs to an original degenerate pair with identical boundary conditions, and 0 < κ f ≤ κ * < 1 / 6 . For a selected pair write
D = I − κ ∑ μ = 0 3 ( I − γ μ ) T μ + ( I + γ μ ) T μ * , Q = γ 5 D = Q * , b = 1 − 6 κ .
The T μ are covariant unitary shifts. Include the gauge weight and all unchanged original pairs in d λ ≥ 0 . For one selected pair, d ν 0 = Z 0 − 1 ( det Q ) 2 d λ , with Z 0 > 0 .
The physical transfer construction gives the canonical slice map ψ = B − 1 / 2 χ , ψ ¯ = χ * B − 1 / 2 γ 0 , where
B = I − κ ∑ j = 1 3 ( T j + T j * ) ≥ b I .
Temporal gauge retains the final temporal holonomy; its integral is the Gauss projector. On the physical Hilbert space the transfer operator is positive. For gauge-invariant complete slice symbols, Schatten Hölder therefore gives
Tr ( T n 1 F ^ 1 … T n s F ^ s ) Tr T N t ≤ ∏ j ∥ F ^ j ∥ , n j ≥ 0 , ∑ j n j = N t .
All coincident-time factors are combined before canonical normal ordering. The estimate uses the positive physical transfer operator in addition to positivity of the sea measure.
Let L α be a spin-independent rectangular color row on one time slice. It depends continuously only on that slice’s spatial links, transforms at a specified root r α , and satisfies
L α ( U g ) = g ( r α ) L α ( U ) G g * , L α L α * ≤ s α I 3 .
The deterministic s α may depend on the cutoff. Stack the rows, with their spin and color indices, into L , and put G = L Q − 1 L * . Inverse-minor expressions below are understood after their determinant-weighted polynomial extension.
Lemma 1
(Actual flavors control products of squared minors). Suppose the original sea contains p identical pairs, hence 2 p identical flavors, with the same Q. For r ≤ p and ordered row and column lists R ℓ , C ℓ of lengths k ℓ , with spin labels fixed and colors summed,
E 0 ∏ ℓ = 1 r ∑ c , d | det G [ ( R ℓ , c ) , ( C ℓ , d ) ] | 2 ≤ ∏ ℓ = 1 r ( 3 / b ) 2 k ℓ ∏ i ∈ R ℓ s i ∏ j ∈ C ℓ s j .
Each factor uses a different original pair. The gauge expectation need not factorize.
Proof. 
Choose the spin basis with γ 0 diagonal and γ 5 a unitary signed permutation. Each spin component of a canonical field or its barred partner is then a pure creation or annihilation component. By the CAR and (3), a component of L α ψ or ψ ¯ γ 5 L α * has canonical norm at most ( s α / b ) 1 / 2 .
For the two flavors of pair ℓ, write ζ f = L ψ f and ζ ¯ f = ψ ¯ f γ 5 L * . Their conditional covariance is the Hermitian matrix G. At an output row form the gauge-invariant even symbols B i = ∑ c ζ 1 , i , c ζ ¯ 2 , i , c and C i = ∑ c ζ 2 , i , c ζ ¯ 1 , i , c . Wick contraction of ∏ B R i ∏ C C j is the color sum of squared minors, up to a fixed orientation sign. Use different original pairs for the r factors. On each slice, normal order the complete gauge-invariant symbol. Bound its operator norm on the full canonical space by the triangle inequality over the color expansion; every normally ordered monomial has norm at most the product of its component norms. The resulting complete slice operator preserves the physical Hilbert space. Apply (4) to these operators and sum the 3 2 ∑ k ℓ color terms to obtain (6).
Finally, det Q det G [ R , C ] is a polynomial in Q and the rows: up to sign it is the augmented determinant det Q L C * L R 0 . Multiplication by the original p pair determinants thus removes every inverse in the product. The Grassmann identity supplies the same polynomial at singular matrices. This proves the stated extension without adding a sea flavor. □
For a finite row family with deterministic real weights, define the source and its absolute-weight version by
J = ∑ α w α L α * L α , J + = ∑ α | w α | L α * L α , M = 12 b ∑ α | w α | s α .
The factor twelve is spin times color. If C k ( J + ) is the sum of weighted squared minors of order k, replacing subsets by ordered lists and dividing by ( k ! ) 2 in (6) gives
E 0 ∏ ℓ = 1 r C k ℓ ( J + ) ≤ ∏ ℓ = 1 r M 2 k ℓ ( k ℓ ! ) 2 , r ≤ p .
Summing the coefficientwise one-pair bound for a finite list of sources of costs M j gives the entire majorant I 0 ( 2 ∑ j | z j | M j ) . In particular their original insertion moments obey
| S 0 ( P 1 … P N ) | ≤ N ! ( ( N / 2 ) ! ) 2 ∏ j M j ≤ 2 N ∏ j M j ( N even ) , S 0 ( P 1 … P N ) = 0 ( N odd ) .
Here P j = − i ( ψ ¯ 1 γ 5 J j ψ 1 − ψ ¯ 2 γ 5 J j ψ 2 ) , and moments mean the corresponding Grassmann source derivatives. The odd identity follows by exchanging the two selected flavors. Coincident-time moments use the complete slice-symbol convention fixed above.

3. Uniform Stability and Flavor-Dependent Moments

Source one original pair by Q ± i z J , with real z, leaving all other pairs unchanged. Set
R z = | det ( Q + i z J ) | 2 ( det Q ) 2 , Z = E 0 R z , d ν z = Z − 1 R z d ν 0 ( Z > 0 ) , e ( h ) = I 0 ( 2 h ) − 1 , h = | z | M .
The normalization Z is relative to the original partition function. In the finite Wilson model the determinant zero set is Haar-null: the determinant is real analytic and is nonzero at the free-link configuration with the stated temporal boundary condition and κ < 1 / 6 . The ratio is consequently defined almost everywhere. Density comparisons can also be written directly in their polynomial form before division.
Theorem 1
(Uniform source stability). In the setting of Section 2, assume the original sea contains p identical pairs with matrix Q. Then, for every real signed row profile and every integer 1 ≤ r ≤ p ,
E 0 | R z − 1 | r ≤ e ( h ) r .
In particular | Z − 1 | ≤ e ( h ) and, whenever Z > 0 ,
TV ( ν z , ν 0 ) ≤ min { 1 , e ( h ) } .
Here total variation is one half of the L 1 distance. The condition e ( h ) < 1 guarantees Z > 0 . If p ≥ 2 , then
χ 2 ( ν z ∥ ν 0 ) ≤ e ( h ) 2 Z 2 , D ( ν z ∥ ν 0 ) ≤ log 1 + e ( h ) 2 Z 2 .
For signed profiles one may replace Z by 1 − e ( h ) when e ( h ) < 1 . For nonnegative row weights, R z ≥ 1 , Z ≥ 1 , and
TV ( ν z , ν 0 ) ≤ 1 − I 0 ( 2 h ) − 1 , D ( ν 0 ∥ ν z ) ≤ log I 0 ( 2 h ) ≤ h 2 ;
p ≥ 2 ⟹ D ( ν 0 ∥ ν z ) ≤ 1 2 e ( h ) 2 .
All constants are independent of gauge coupling and volume at fixed row cost. Only pairs already present in the original sea can be counted in p.
Proof. 
Let W repeat w α on its spin and color indices. Sylvester’s identity and Cauchy–Binet give
R z = det ( I + z 2 ( G W ) 2 ) = ∑ k ≥ 0 z 2 k ∑ | I | = | C | = k w I w C | det G [ I , C ] | 2 .
Indeed, a principal minor of ( G W ) 2 expands over an intermediate set C, and the two minors of G = G * are conjugate. Thus | R z − 1 | ≤ ∑ k ≥ 1 z 2 k C k ( J + ) pointwise. Raise this positive majorant to the rth power and apply (8). Summing proves (11). After multiplication by the original determinants, the expanded minor terms extend continuously as polynomials in Q and the row entries. Their positive majorant controls the integrated density difference across determinant zeros as well.
For normalization put u = E 0 ( R z − 1 ) + and v = E 0 ( 1 − R z ) + . Then Z = 1 + u − v . If Z ≤ 1 , TV = E 0 ( 1 − R z / Z ) + ≤ v ; if Z ≥ 1 , TV = E 0 ( R z / Z − 1 ) + ≤ u / Z . Both bounds are at most u + v ≤ e ( h ) , proving (12) without a denominator loss. When p ≥ 2 ,
χ 2 ( ν z ∥ ν 0 ) = Var 0 ( R z ) / Z 2 ≤ E 0 ( R z − 1 ) 2 / Z 2 .
Jensen’s inequality under ν z gives D ( ν z ∥ ν 0 ) ≤ log ( 1 + χ 2 ( ν z ∥ ν 0 ) ) .
If W ≥ 0 , Sylvester gives R z = det ( I + z 2 A 2 ) ≥ 1 , with A = W 1 / 2 G W 1 / 2 = A * . Hence v = 0 and TV ≤ ( Z − 1 ) / Z ≤ 1 − I 0 ( 2 h ) − 1 . Also
D ( ν 0 ∥ ν z ) = log Z − E 0 log R z ≤ log I 0 ( 2 h ) ≤ h 2 .
The last inequality follows from ( k ! ) 2 ≥ k ! in the Bessel series. Finally write U = R z − 1 ≥ 0 and use log ( 1 + U ) ≥ U − U 2 / 2 and log ( 1 + E 0 U ) ≤ E 0 U . For p ≥ 2 this gives (). □
Thus one pair gives a quadratic total-variation error. Two original identical pairs give quartic forward relative entropy, and quartic entropy in both orientations for a nonnegative source. Quartic small-shift ensemble fluctuations already occur in two-flavor mass reweighting [5]; the point of Theorem 1 is its nonasymptotic likelihood-moment bound in the stated transfer range, with the actual flavor count and without a uniform inverse-gap input.

3.1. Determinant Zeros and Entropy

Example 1
(One-pair bounds and uniform forward entropy). Take a one-dimensional Hermitian matrix Q ∈ { η , 1 } with equal reference weights and one determinant pair. With J = 1 the original and sourced probabilities of Q = η are
p η = η 2 1 + η 2 , q η , z = η 2 + z 2 1 + η 2 + 2 z 2 , R z = 1 + z 2 / Q 2 .
The only nonconstant averaged minor satisfies E 0 Q − 2 = 2 / ( 1 + η 2 ) ≤ 2 ; larger minors vanish. A uniform one-pair Bessel majorant therefore holds with M = 2 . Let the original law vary with z ↓ 0 by choosing η = exp ( − 1 / z 2 ) . Then
TV ( ν z , ν 0 ) ∼ z 2 , D ( ν 0 ∥ ν z ) ∼ z 2 , D ( ν z ∥ ν 0 ) ⟶ 2 .
These formulas follow directly from Bernoulli relative entropy: q η , z ∼ z 2 , while log ( q η , z / p η ) = 2 / z 2 + 2 log z + o ( 1 ) . The complementary term in the forward entropy tends to zero; in the reverse direction it is asymptotic to z 2 .
This abstract determinant model proves that one-pair minor control alone is insufficient for uniform forward-entropy convergence. Near-zero determinants with finite means and uncontrolled higher moments are familiar in fermion simulations [11]; the example isolates their consequence for normalized laws. A stronger Wilson-specific conclusion requires additional information about the original measure.
Proposition 1
(The moment threshold near a regular determinant zero). Suppose a smooth finite-dimensional Hermitian family has a simple eigenvalue t crossing zero transversely, all other eigenvalues stay nonzero, and the reference density is smooth and positive. With p identical original determinant pairs, source one pair by a fixed Hermitian perturbation that lifts this zero. For fixed sufficiently small z ≠ 0 , restrict to a neighborhood where the sourced determinant stays nonzero. For 0 < r < ∞ ,
R z ∈ L loc r ( ν 0 ) ⟺ r < p + 1 2 .
Proof. 
In transverse coordinates ( t , y ) , the original density is comparable to | t | 2 p d t d y . Lifting the zero makes the sourced determinant product nonzero there, so R z ≍ t − 2 . The local integral is therefore comparable to ∫ − δ δ | t | 2 p − 2 r d t . A rank-one positive source overlapping the null eigenvector lifts the zero by the determinant lemma. □
Thus the integer range r ≤ p is sharp under the regular-zero hypothesis. Applying this local criterion to a Wilson trajectory requires establishing the hypothesis on that trajectory.
Example 2
(Cubic entropy at a fixed determinant zero). Take Q = t ∈ [ − 1 , 1 ] , uniform reference measure, one original pair and J = 1 . Then d ν 0 = 3 2 t 2 d t , R z = 1 + z 2 / t 2 and Z = 1 + 3 z 2 . Although E 0 Q − 2 = 3 , the likelihood variance is infinite for every z ≠ 0 . Put s = | z | and
I ( s ) = ∫ 0 1 t 2 log ( 1 + s 2 / t 2 ) d t = 1 3 log ( 1 + s 2 ) + 2 3 s 2 − 2 3 s 3 arctan ( 1 / s ) , J ( s ) = ∫ 0 1 log ( 1 + s 2 / t 2 ) d t = log ( 1 + s 2 ) + 2 s arctan ( 1 / s ) .
Integration by parts gives these formulas and hence
D ( ν 0 ∥ ν z ) = log ( 1 + 3 s 2 ) − 3 I ( s ) = π s 3 + O ( s 4 ) , D ( ν z ∥ ν 0 ) = 3 [ I ( s ) + s 2 J ( s ) ] 1 + 3 s 2 − log ( 1 + 3 s 2 ) = 2 π s 3 + O ( s 4 ) .
This fixed reference model has finite but nonquartic entropy in both orientations. Together with Example 1, it separates ordinary convergence, a uniform rate and convergence under a varying original law.

3.2. Fermionic Source Coefficients

Proposition 2
(Stability of actual fermionic source coefficients). Let P 0 be a real background row insertion and P 1 , … , P N real probe insertions of the same original pair, with costs M 0 , … , M N . The normalized Grassmann source functional is
S z ( F ) = S 0 ( e z P 0 F ) S 0 ( e z P 0 ) .
Put h = | z | M 0 and assume e ( h ) < 1 . For even N ≥ 2 ,
| S z ( P 1 … P N ) − S 0 ( P 1 … P N ) | ≤ 2 N ∏ j = 1 N M j cosh ( 2 h ) − 1 + e ( h ) 1 − e ( h ) .
For odd N, the zero-source coefficient vanishes and
| S z ( P 1 … P N ) | ≤ 2 N ∏ j = 1 N M j sinh ( 2 h ) 1 − e ( h ) .
Proof. 
Expand the unnormalized insertion coefficient in z. Equation (9) bounds its kth derivative by 2 N + k M 0 k ∏ j M j and makes it vanish when N + k is odd. Summing the even or odd Taylor series gives respectively cosh ( 2 h ) − 1 or sinh ( 2 h ) . Divide by Z ≥ 1 − e ( h ) ; for even N, changing the normalization contributes at most 2 N ∏ j M j e ( h ) / ( 1 − e ( h ) ) . □
The direct source expansion controls the fermionic insertions, including their normalization. For uniformly bounded probe costs, background removal has quadratic error in even degree and linear error in odd degree.

4. Spatial Heat Sources in Physical Units

On a spatial slice let
Δ U = ∑ j = 1 3 ( 2 I − T j − T j * ) , H a , τ = e − τ Δ U / a 2 , L a , τ , r = a − 3 / 2 ( H a , τ ) r , . .
The heat operator acts trivially on spin and never mixes time slices. Each path in its expansion has a nonnegative scalar weight and unitary parallel transport. Its color-block norm is therefore dominated by the scalar lattice heat kernel, a discrete semigroup-domination principle [12]. On a torus with physical spatial periods ℓ j = a L j , the semigroup identity gives
L a , τ , r L a , τ , r * ≤ S τ , ℓ I 3 , S τ , ℓ = ∏ j = 1 3 ℓ j − 1 + π 4 2 τ .
Indeed, in one direction L − 1 ∑ q e − 4 u sin 2 ( π q / L ) ≤ L − 1 + π / ( 4 u ) , using sin ( π q / L ) ≥ 2 min ( q , L − q ) / L and a Gaussian integral. Apply this at u = 2 τ / a 2 in three directions. In infinite spatial volume, S τ , ∞ = C 3 τ − 3 / 2 with C 3 = π 3 / 2 / ( 128 2 ) .
Use the reflection-compatible time mesh t n = a ( n + 1 / 2 ) and set
q a , τ = 2 κ a a − 3 / 2 H a , τ ψ , P a , τ ( f ) = − i a 4 ∑ n , r f ( t n , a r ) ( q ¯ 1 γ 5 q 1 − q ¯ 2 γ 5 q 2 ) ( t n , r ) .
The barred field is q ¯ a , τ = 2 κ a a − 3 / 2 ψ ¯ H a , τ . For a real profile the exact source kernel on a slice is 2 κ a a H a , τ M f H a , τ . Its row weight is 2 κ a a 4 f ( t n , a r ) ; hence
M a , τ , ℓ ( f ) = 24 κ a 1 − 6 κ a S τ , ℓ ∥ f ∥ a , 1 , ∥ f ∥ a , 1 = a 4 ∑ n , r | f ( t n , a r ) | .
This fixes the physical normalization. Multiplying the bilinear by Z P , a multiplies its cost by | Z P , a | .
Corollary 1
(Source removal across the positive-transfer interval). Let all original hopping parameters remain below one common κ * < 1 / 6 , let the physical spatial periods be bounded below, and let ∥ f ∥ a , 1 be bounded. Gauge couplings are unrestricted within the stated Wilson theory. If the effective amplitude z a Z P , a is real, τ a ↓ 0 and
| z a Z P , a | τ a − 3 / 2 ⟶ 0 ,
then the full normalized gauge laws with and without the source converge in total variation. Every fixed-degree coefficient of the same-pair probe functional in Proposition 2, with uniformly bounded probe costs, is preserved. With two original identical pairs, forward relative entropy also tends to zero; both orientations do so for a nonnegative profile.
Proof. 
Equation (27) gives h a = | z a Z P , a | M a , τ a , ℓ ( f ) → 0 . Apply Theorem 1 and Proposition 2. □
This sufficient criterion gives a joint limit in which the source strength and smearing radius tend to zero. For κ a bounded away from zero, the displayed heat-cost majorant grows as τ a − 3 / 2 ; it alone therefore gives no fixed-amplitude conclusion. At fixed positive probe radii, bounded profile L 1 norms and bounded normalization factors give uniform costs throughout κ a ≤ κ * < 1 / 6 .

4.1. A Common Source Sector at Positive Heat Time

The estimates also give compactness of a common source sector. Keep the finite set of original flavor pairs fixed. Along any sequence a j ↓ 0 , first take a common thermodynamic subsequence at each a j , and then a subsequence in j. For any fixed positive heat times and smooth compactly supported spacetime profiles, the source functions and all their derivatives have a common subsequential limit, locally uniformly in the complex amplitudes of each finite list of fixed sources. The limiting coefficients are translation invariant and positive under physical-time reflection. Here is the passage, including the spatially nonlocal rows.
At fixed a, embed centered gauge tori periodically in the compact link space SU ( 3 ) E ( Z 4 ) . Choose even temporal extents and place the antiperiodic seam outside each fixed cylinder as the volume grows. Use point rows in Lemma 1, one actual pair for each squared minor. The same complete-slice proof applies to unequal pairs with their own Q f and b f . Group the two Wick minors of each original pair as A f , B f ; then E | ∏ f A f B f | ≤ ( E ∏ f | A f | 2 E ∏ f | B f | 2 ) 1 / 2 . Each fixed-color squared product is bounded by its nonnegative full color sum, which uses one copy of each original pair. A bounded gauge multiplier is removed first by its supremum norm in the absolute-value integral; the transfer bound is applied only to complete slice symbols. This bounds the total variation of every balanced Grassmann coefficient measure; unbalanced monomials vanish by flavor phases. Weak-* compactness and a diagonal argument for the countable component monomials give a common infinite-volume hierarchy. Fixed translations and reflections eventually avoid the periodic seams, so their identities and finite reflection Gram inequalities pass to the same limit.
To approximate a heat row by a cylinder row, kill paths leaving a cube of physical radius R, keeping the diagonal degree six. Writing L R for the killed row, scalar path domination gives
∥ L − L R ∥ 2 ≤ C τ − 3 / 2 p R , p R ≤ min { 1 , 6 e − min { R 2 / ( 8 τ ) , R / ( 2 a ) } } .
Indeed the full-minus-killed scalar row has mass p R and maximum at most the heat diagonal. Its squared row norm is at most their product. For one walk coordinate, E e θ X u = e 2 u ( cosh θ − 1 ) ; the exponential maximal inequality with θ = min { R a / ( 4 τ ) , 1 } gives the displayed exit bound. The source difference satisfies L * L − ( L R ) * L R = ( L − L R ) * L + ( L R ) * ( L − L R ) . The canonical norm estimate therefore bounds its cost by 12 b − 1 ∑ α | w α | ( ∥ L α ∥ + ∥ L α R ∥ ) ∥ L α − L α R ∥ . Telescoping moments and (9) make their error tend to zero as R → ∞ . Thus the same infinite-volume hierarchy contains the heat sources.
At fixed positive heat times, the costs in (27) are uniform. For a countable dense family of profiles and heat times, diagonal compactness of Taylor coefficients, together with the common Bessel majorant, gives a common entire limit, locally uniformly with all derivatives. The moment bound extends the limiting functional continuously in each L 1 profile. Heat-time continuity follows from
| e − τ λ − e − σ λ | 2 ≤ 4 e − 2 τ 0 2 | τ − σ | 2 e − τ 0 λ , λ ≥ 0 , τ , σ ≥ τ 0 > 0 .
Taking row diagonals retains a heat factor, so (25) controls the normalized row difference. For smooth compact profiles, Riemann sums and bounded first derivatives give
∥ f ∥ a , 1 → ∥ f ∥ 1 , ∥ f ( · − ξ a ) − f ( · − ξ ) ∥ a , 1 → 0 if ξ a → ξ .
Hence lattice translations approximating physical translations pass to the same limiting functional. The same approximation passes every finite positive- time reflection Gram matrix. In its GNS space, a positive-time polynomial F satisfies ∥ T t F ∥ 2 = 〈 F , T 2 t F 〉 ≤ ∥ F ∥ ∥ T 2 t F ∥ . The moment-cost bound controls all translated norms by one constant; iteration gives ∥ T t F ∥ ≤ ∥ F ∥ , including null vectors. Profile translation continuity gives strong continuity. Symmetry and T t = T t / 2 2 yield a positive contraction semigroup T t = e − t H , H ≥ 0 .
The construction gives a sector at fixed positive smearing. Its source function can equal one, so a nontrivial local limit requires additional control as the probe radius tends to zero. Corollary 1 supplies the source comparison; Section 6 states the finite-energy test.

5. Exact Contact Conversion and Contact Bounds

The preceding Wilson comparison is independent of any overlap construction. For a normal Ginsparg–Wilson operator on its full one-flavor space, let
D + D * = D * D = D D * , D m = ( 1 − m / 2 ) D + m I , m = a M ∈ ( 0 , 2 ) ,
and define R = a D m − 1 , G = a ( I − D / 2 ) D m − 1 and c = 1 − m / 2 . The classical contact identity [8] is
R = c G + ( a / 2 ) I .
It retains the exceptional D = 2 subspace: G vanishes there but R does not. We first use this identity coefficientwise, and then compare the stronger full-operator metric.
Theorem 2
(Exact separated-time source conversion). Repeat D on any fixed degenerate original flavor block. Let K 1 ( U ) , … , K N ( U ) be source matrices on the full space, uniformly bounded in U at each fixed cutoff, with bounds allowed to depend on a. Assume K i K j = 0 for i ≠ j and Tr K i = 0 . Under one fixed source-free sea law define
Z X ( z ) = E det ( I + ∑ i z i K i X ) , X = R , G .
For every nonempty subset I of source labels, differentiating each selected variable once gives
∂ I Z R ( 0 ) = c | I | ∂ I Z G ( 0 ) , ∂ I log Z R ( 0 ) = c | I | ∂ I log Z G ( 0 ) .
Both prescriptions use the same sea, without a commutation condition between the sources and D.
Proof. 
Work configurationwise in C [ z 1 , … , z N ] / ( z 1 2 , … , z N 2 ) and put K = ∑ i z i K i . The support condition gives K 2 = 0 . Equation (31) then gives
I + K R = ( I + ( a / 2 ) K ) ( I + c K G ) , det ( I + ( a / 2 ) K ) = exp ( ( a / 2 ) Tr K ) = 1 .
The square-free coefficients agree before averaging. Integrating them proves the first identity. The constant term is one, so the formal logarithm in this finite quotient ring is defined and gives the second. Its coefficients are the ordinary logarithmic derivatives near zero. □
For example take time-block spatial kernels K i = H a , τ i M h i H a , τ i ⊗ Γ i ⊗ T i , where Γ i is a fixed spin matrix, T i is a traceless flavor matrix, and the h i have disjoint time supports. Then the theorem applies to all τ i ≥ 0 , including zero. After the exact factor c − 1 per bilinear, separated moments and connected coefficients coincide at each cutoff. Their boundedness or nonvanishing can be checked in either prescription.
Tracelessness is essential for this averaged statement. In the abstract GW example D = 2 I , take K i = X ( U ) P i with disjoint rank-one projectors. The bare mixed connected coefficient is ( a 2 / 4 ) Var ( X ) ; the corrected one is zero. For general traces the theorem is restored by multiplying the bare integrand by exp [ − ( a / 2 ) ∑ i z i Tr K i ( U ) ] inside the expectation. This subtracts the identity-contact tadpole at its actual gauge configuration.
Whenever I + ( a / 2 ) J is invertible, in particular near zero source, the underlying finite-source coordinate map also remains exact:
det ( I + J R ) = det ( I + ( a / 2 ) J ) det I + c J ( I + ( a / 2 ) J ) − 1 G .
It follows by factoring I + ( a / 2 ) J . For gauge-independent scalar site profiles the first factor is a deterministic local contact determinant and can be taken outside the sea expectation. For gauge-dependent rows it generally cannot. This distinction explains why the coefficientwise argument above is needed.
Proposition 3
(Full-row contact control and its sharp heat window). For any rectangular source row matrix L, without a commutation assumption, set B = L R L * and S = L G L * . Then
B − S = a 2 L A L * , A = D D m − 1 , ∥ A ∥ ≤ 1 , ∥ B − S ∥ 2 2 ≤ a 2 4 Tr [ ( L L * ) 2 ] .
For L = χ 1 / 2 ⨁ n H a , τ , n with a nonnegative spacetime profile, spin four and color three,
∥ B − S ∥ 2 2 ≤ 3 a ∥ χ ∥ ∞ ∥ χ ∥ a , 1 ∏ j = 1 3 ℓ j − 1 + π 8 τ .
For bounded profile norms on a fixed physical time slab and spatial periods bounded away from zero, τ a ≫ a 2 / 3 removes this contact as τ a → 0 . The exponent is sharp for this full-spacetime Hilbert–Schmidt metric.
Proof. 
For an eigenvalue d of D, d + d ¯ = | d | 2 and | d | ≤ 2 . Thus | ( 1 − m / 2 ) d + m | 2 = m 2 + ( 1 − m 2 / 4 ) | d | 2 ≥ | d | 2 . The contact identity gives A = I − M G , and Schatten Hölder gives ∥ L A L * ∥ 2 ≤ ∥ L ∥ 4 2 . Equality in (34) holds for D = 2 I , for every L. With F = ⨁ n H a , τ , n ,
Tr [ ( L L * ) 2 ] = Tr ( χ F 2 χ F 2 ) ≤ ∥ χ ∥ ∞ Tr ( χ F 4 ) .
Apply the scalar heat bound at 4 τ / a 2 and sum spin and color. The factors are a 2 a − 4 a 3 = a , proving (35).
For sharpness fix M > 0 and use the free Wilson–overlap symbol with kernel mass one, fixed physical periods and antiperiodic time. Put w ( k ) = ∑ μ = 0 3 ( 1 − cos k μ ) , ρ ( k ) 2 = ( w ( k ) − 1 ) 2 + ∑ μ sin 2 k μ , and q ( k ) = 2 + 2 ( w ( k ) − 1 ) / ρ ( k ) . For χ = 1 , the exact squared contact is
3 a 2 ∑ k e − ( 16 τ / a 2 ) ∑ j = 1 3 sin 2 ( k j / 2 ) q ( k ) m 2 + ( 1 − m 2 / 4 ) q ( k ) .
If a 2 ≪ τ ≪ 1 , the spatial weights concentrate at k = 0 , where q ( k 0 , 0 ) = 4 sin 2 ( k 0 / 2 ) . Away from k 0 = 0 the final fraction tends uniformly to one; the excluded temporal fraction is arbitrarily small. Its weighted mean therefore tends to one. The spatial Gaussian sum is asymptotic to ℓ 1 ℓ 2 ℓ 3 / ( 16 π τ ) 3 / 2 , and there are ℓ 0 / a temporal modes. Hence
∥ B − S ∥ 2 2 ∼ 3 V 4 ( 16 π ) 3 / 2 a τ − 3 / 2 , V 4 = ∏ μ = 0 3 ℓ μ .
At τ = c τ a 2 / 3 , c τ > 0 , the limit is positive; below that scale, while a 2 ≪ τ , it diverges. This proves sharpness. □
All Schatten norms here are unnormalized. One time slice instead has cost O ( a 2 τ − 3 / 2 ) ; a fixed physical slab contains O ( a − 1 ) slices. A bilinear factor Z P multiplies the squared cost by | Z P | 2 . Theorem 2 shows why the separated source coefficients admit the exact conversion (32) without a lower bound on the heat time. Coincident-time contacts require the additional control developed below.

5.1. Comparison with Intrinsic Dirac Filtering

For an intrinsic Dirac filter, the same contact has a complementary rank estimate. Write D a = D , D a , M = D m and G a = G at spacing a.
Proposition 4
(A sharp rank threshold). Let D a also satisfy D a * = Γ D a Γ , where Γ is a self-adjoint involution. On P a H = ( ker ( D a − 2 I ) ) ⊥ , put
T a = − i a D a ( I − D a / 2 ) − 1 , F a = ( I + ε 2 T a 2 ) − σ / 2 , n a = rank P a .
Here ε , σ > 0 . Extend all functions by zero on the exceptional subspace. Set S a = F a G a F a , B ^ a = F a ( a D a , M − 1 ) F a , H a = ∥ S a ∥ 2 2 , and E a = ∥ B ^ a − S a ∥ 2 2 . For q = 2 σ + 1 and r = 2 σ / q ,
E a ≤ a 2 4 n a 1 / q ε − 4 σ / q H a r .
If C 0 , d > 0 , n a ≤ C 0 a − d , and H a is tight under specified positive laws, then E a → 0 in probability for fixed ε > 0 and σ > ( d − 2 ) / 4 . The rate and prefactor are optimal given these two inputs. For varying radii a sufficient condition is a 4 σ + 2 − d ε a − 4 σ → 0 with tight H a .
Proof. 
The normal GW calculus gives T a = T a * , D a = i a T a ( I + i a T a / 2 ) − 1 , and B ^ a = ( I + i a T a / 2 ) S a . On a frequency λ , write
x λ = ( 1 + ε 2 λ 2 ) − 2 σ M 2 + λ 2 .
Then H a = ∑ x λ , E a = ( a 2 / 4 ) ∑ λ 2 x λ , and λ 2 x λ ≤ ε − 4 σ / q x λ r . The finite-sum inequality ∑ x r ≤ n a 1 − r ( ∑ x ) r proves the bound.
For sharpness fix h > 0 , take an even n a ∼ C 0 a − d , λ a = ( n a ε − 4 σ / h ) 1 / ( 4 σ + 2 ) , and T a = diag ( λ a I , − λ a I ) , with Γ interchanging the two blocks. The Cayley transform gives an exact normal GW family. Direct substitution yields
H a → h , E a ∼ C 0 1 / q 4 ε − 4 σ / q h r a 2 − d / q .
For d > 2 , this gives a nonzero contact at the threshold and divergence below it within the stated range σ > 0 , with bounded H a . This is sharpness in the general GW class; the spatial-heat sharpness above is established in an actual free Wilson–overlap family. □

5.2. Coincident Pairs and Finite Contact Subtraction

At coincident arguments, separated-source identities leave a further question: when can a positive pair measure be made tight as a distribution by a fixed finite normal subtraction? The criterion below answers that question without assuming a physical counterterm classification.
Let X = T 4 , D = { ( x , x ) : x ∈ X } , and
W ( x , y ) = 4 ∑ j = 1 4 sin 2 π ( x j − y j ) , 16 d ( x , y ) 2 ≤ W ( x , y ) ≤ 4 π 2 d ( x , y ) 2 .
Here d is the flat distance on the unit torus. For random finite positive measures m a on X 2 , put K a , k = m a ( W 2 k ) . A contact distribution has normal order at most q if it has the form
C a ( ϕ ) = ∑ | α | ≤ q c a , α x ↦ ∂ 2 α ϕ ( x , x ) , c a , α ∈ D ′ ( X ) .
The coefficients may be random, cutoff dependent, and of unbounded tangential order; only the normal order is fixed.
Theorem 3 (Finite normal subtraction and a finite weighted cost).
The following existence statements are equivalent:
1.
For some integer k ≥ 1 , the nonnegative costs K a , k are tight.
2.
For some cutoff-independent integer q ≥ 0 , there are contact distributions C a of normal order at most q such that m a − C a has tight distribution laws.
More precisely, the second condition forces K a , k to be tight for every 4 k > q . The first permits q = 4 k − 1 and gives tight laws in H − t ( X 2 ) for every t > 4 k + 4 .
Proof. 
Every normal derivative of W 2 k through order 4 k − 1 vanishes on D . Therefore, if 4 k > q ,
K a , k = ( m a − C a ) ( W 2 k ) .
Evaluation at this fixed smooth test is continuous in the distribution topology, so tight distribution laws imply tight scalar evaluations. This argument does not require positivity of the subtracted distribution.
Conversely, set s = 4 k and choose a smooth cutoff χ ( h ) equal to one near zero and supported inside a normal coordinate chart, where h = y − x . Let J s − 1 ϕ be the normal Taylor polynomial through degree s − 1 , and define
C a ( ϕ ) = m a ( χ J s − 1 ϕ ) , T a ( ϕ ) = m a ( ϕ − χ J s − 1 ϕ ) .
At each finite cutoff C a is a contact distribution of normal order at most s − 1 . Taylor’s integral formula near the diagonal, and the positive lower bound for W away from that chart, give
| ϕ − χ J s − 1 ϕ | ≤ C W 2 k ∥ ϕ ∥ C s , | T a ( ϕ ) | ≤ C K a , k ∥ ϕ ∥ C s .
Choose s + 4 < t 0 < t . Sobolev embedding in dimension eight gives ∥ T a ∥ H − t 0 ≤ C K a , k . The embedding H − t 0 ↪ H − t is compact. A tight scalar bound therefore gives compact sets containing the distribution laws with arbitrarily high probability, proving the assertion. □
The strict inequality 4 k > q is necessary for this reverse implication. For example, take m ϵ = ϵ − ( 4 k + 1 / 2 ) δ ( x 0 , x 0 + ϵ e 1 ) . Subtracting its normal Taylor jet through degree q = 4 k makes the resulting distributions tend to zero, whereas m ϵ ( W 2 k ) ≍ ϵ − 1 / 2 diverges. This is a positive-measure example, without a Wilson-resolvent realization.
If one additionally wants the extension to be a continuous function of the weighted weak measure W 2 k m a , subtract the next Taylor coefficient, through degree 4 k . The quotient ( ϕ − χ J 4 k ϕ ) / W 2 k then extends continuously by zero on the diagonal. With only degree 4 k − 1 , its boundary value may depend on the approaching direction. This explains why tightness and continuity of an extension have different minimal Taylor degrees. The criterion characterizes tight distributional extensions of a positive pair measure. Gauge-local counterterms, Ward identities and compatibility at higher collision diagonals impose further conditions.

6. Charge Sectors and Finite Physical Energy

The source comparison holds throughout κ f < 1 / 6 . A separate estimate shows why a heavy hopping subregion cannot retain a finite-energy charged sector. We use the positive-sea mechanism of Vafa and Witten [9], with an explicit optimal Wilson propagation rate.
Lemma 2
(Uniform temporal propagation). For 0 < κ < 1 / 8 , put r κ = 2 κ / ( 1 − 6 κ ) and μ κ = − log r κ . If P n projects onto time slice n, then every gauge background satisfies
∥ P n D κ − 1 P m ∥ ≤ r κ d t ( n , m ) 1 − 8 κ .
Here d t = | n − m | on infinite time and is cyclic distance on a finite antiperiodic temporal circle.
Proof. 
Split the Wilson hopping matrix as M sp + M tm . Each directional term is 2 ( P μ − T μ + P μ + T μ * ) with orthogonal spin projections and has norm two. Thus ∥ M sp ∥ ≤ 6 , ∥ M tm ∥ ≤ 2 . The time-block diagonal matrix B 0 = I − κ M sp satisfies ∥ B 0 − 1 ∥ ≤ ( 1 − 6 κ ) − 1 . In
D κ − 1 = ∑ j ≥ 0 ( κ B 0 − 1 M tm ) j B 0 − 1 ,
every ( n , m ) block with j < d t ( n , m ) vanishes. Summing the remaining geometric series proves (38). Antiperiodic phases do not change these norms or graph distances. □
For the spectral interpretation, normalize the physical transfer operator by its largest eigenvalue: T ˜ a = T a / λ max ( T a ) and H a = − a − 1 log T ˜ a . The infinite-time limit can be a mixed ground state; its GNS representation supplies the cyclic vector Ω and the corresponding nonnegative generator used below.
Theorem 4
(Charge-resolved spectral exclusion). At fixed a and finite spatial volume, first take the infinite-time physical Wilson vacuum limit. Let O be a finite gauge-invariant slice polynomial with bounded coefficients depending only on that slice’s spatial links, homogeneous under the flavor phases, with charge q f = # ψ f − # ψ ¯ f . Every flavor appearing in the polynomial, including in neutral factors, has 0 < κ f < 1 / 8 . Other original pairs obey the positive-transfer assumptions. Then, for a finite cutoff-dependent A O , a ,
0 ≤ C O , a ( n ) = 〈 O n O 0 † 〉 ≤ A O , a e − n ∑ f | q f | μ κ f , n ≥ 1 .
The adjoint is the physical reflection adjoint. The positive spectral measure of this channel is supported in
[ g a ( q ) , ∞ ) , g a ( q ) = 1 a ∑ f | q f | log 1 − 6 κ f 2 κ f .
On the closed subspace generated by O † Ω for the stated charge-q polynomials and their time translates, H a ≥ g a ( q ) I .
Proof. 
Expand the two complete slice symbols into component monomials. Physical-adjoint normal-ordering terms preserve flavor charge and have bounded coefficients at fixed cutoff. In any nonzero Wick matching let x f count contractions from a field at time n to a barred field at zero, and y f the reverse crossings. Internal contractions remove one field and one barred field, so x f − y f = q f and x f + y f ≥ | q f | . Each crossing contributes at most ( 1 − 8 κ f ) − 1 r κ f d t ( n , 0 ) . Every other contraction is bounded by ( 1 − 8 κ f ) − 1 . The finite number of terms and bounded spatial coefficients go into A O , a .
This absolute-value bound is background uniform and survives integration under the full original positive sea. Taking temporal circumference to infinity first makes d t ( n , 0 ) = n for each fixed n, with the same prefactor. The physical transfer representation then supplies C O , a ( n ) = ∫ e − a n E d ρ O , a ( E ) with ρ O , a ≥ 0 , for either Grassmann parity. Positive mass below g a ( q ) − δ would contradict (39) as n → ∞ . Hence its spectral projection annihilates every such O † Ω and every time translate. The projection’s kernel is closed, proving the sector statement. The charge-sign convention is immaterial because the bound uses absolute values. □
The threshold passes to a subsequent spatial thermodynamic limit when the smeared correlations and their bounds are uniform there. Multiplicative field normalization changes spectral weight but cannot move its support. If g a ( q ) → ∞ and renormalized correlations remain bounded at earlier positive times, then C a ( t ) ≤ e − g a ( q ) ( t − s ) C a ( s ) for 0 < s < t proves that every positive-time limit in this channel is zero.
For an off-diagonal degenerate-pair meson, g a = 2 μ κ / a ; for a three-quark equal-hopping baryon, g a = 3 μ κ / a . The diagonal nonsinglet bilinear of Section 4 has the same meson threshold: its disconnected contractions cancel between the two equal flavors, leaving twice the off-diagonal contraction. Counting eight oriented hops with norm two gives the sufficient absolute path-expansion condition 16 κ < 1 . That entire region has meson threshold greater than 2 log 5 / a ; at κ = 1 / 64 it is 2 log 29 / a . Neither arbitrary spatial source design nor a finite, time-independent normalization creates finite-energy charged weight in that regime.
The common per-charge coefficient, and hence the meson exponent, is sharp among background-uniform estimates. In the free background at zero spatial momentum, M sp = 6 I , and the two temporal spin projections decouple: S ( n , 0 ) = r κ n P ± / ( 1 − 6 κ ) for n > 0 . The pseudoscalar meson has a positive contribution 2 N c r κ 2 n / ( 1 − 6 κ ) 2 . A strictly larger uniform exponent would fail at large n. This is the standard free Wilson pole [13]; it calibrates the sharp constant, not attainment by an interacting sea.
In particular, approaching from κ ( a ) < 1 / 8 , positive meson weight below a fixed E * > 0 requires
κ ( a ) ≥ 1 6 + 2 e a E * / 2 = 1 8 − a E * 64 + O ( a 2 ) .
The exact source-stability theorem remains available on both sides of this free threshold. It thus supplies a comparison tool in a parameter region not excluded by the heavy-sector argument.

6.1. A Finite-Energy Test Using Only Positive Times

Positive spectral functionals can be bounded from Euclidean data by duality methods [14]. The following closed-form certificate uses only two summaries of three positive-time values; its constant is optimal for those summaries.
Proposition 5
(Sharp positive-time certificate). Let t > 0 and C ( u ) = ∫ [ 0 , ∞ ) e − u E d μ ( E ) , with μ ≥ 0 , C ( t ) ≤ B < ∞ , and C ( 2 t ) − C ( 3 t ) ≥ d > 0 . Then d ≤ B / 4 . For 0 < η < d / B , put
x ± = 1 ± 1 − 4 η 2 , I t , η = ( − t − 1 log x + , − t − 1 log x − ) .
With d μ t = e − t E d μ , one has
μ ( I t , η ) ≥ μ t ( I t , η ) ≥ d − η B 1 / 4 − η .
The last bound is sharp given only B , d ; μ may have infinite total mass. If instead supp μ ⊆ [ M , ∞ ) with M ≥ 0 and C ( t ) > 0 , then
C ( 2 t ) − C ( 3 t ) C ( t ) ≤ Φ ( t M ) , Φ ( s ) = 1 / 4 , s ≤ log 2 , e − s ( 1 − e − s ) , s ≥ log 2 .
Proof. 
The finite measure μ t has mass at most B. For φ ( E ) = e − t E ( 1 − e − t E ) , 0 ≤ φ ≤ 1 / 4 and ∫ φ d μ t = C ( 2 t ) − C ( 3 t ) . Writing m = μ t ( { φ > η } ) gives d ≤ η ( B − m ) + m / 4 , which proves (42). Equality is attained by placing mass ( d − η B ) / ( 1 / 4 − η ) at E = t − 1 log 2 and the remaining mass B − m at either endpoint of I t , η . Multiplication by e t E converts these atoms to a positive μ . For (43), normalize μ t by C ( t ) and maximize φ on [ M , ∞ ) . An atom at max ( M , t − 1 log 2 ) attains the maximum. □
If A j approximate C ( j t ) with errors ϵ j , use B = A 1 + ϵ 1 and d = A 2 − A 3 − ϵ 2 − ϵ 3 . A positive uniform d together with a uniform finite B retains a finite physical band in a common positive limiting hierarchy. The probes must represent this same channel: for a positive-time polynomial F use C ( u ) = S 0 ( ( θ F ) τ u F ) = 〈 [ F ] , e − u H [ F ] 〉 . Proposition 2 controls its reflection-paired, translated source coefficients. At finite a, arbitrary real times refer to the canonical e − t H a interpolation; with lattice data use one t a = a n a → t > 0 and the common separations t a , 2 t a , 3 t a . By (43), a threshold tending to infinity is incompatible with bounded B, positive contrast and vanishing errors. This is the precise physical-survival test left after source stability has been proved.

7. Conclusions

The original Wilson determinants give quantitative source control beyond the region of convergent hopping expansions. Theorem 1 bounds the normalized gauge law, and Proposition 2 controls the corresponding fermionic coefficients. The number of identical sea pairs determines the available likelihood moments. The regular-zero criterion and the cubic entropy model explain why a first-moment estimate alone does not yield quartic entropy.
Spatial heat kernels make the source cost explicit in physical units and give the joint removal criterion (28). For overlap sources, separated-time contact conversion is exact; full-spacetime Hilbert–Schmidt convergence has its own sharp smearing threshold. The rank estimate and finite normal subtraction criterion cover complementary choices of filter and coincident source topology.
The Wilson energy floor excludes a finite-energy charged limit along the heavy hopping route. The positive-time certificate states what a viable limit must retain: bounded correlations and a positive spectral contrast. An interacting continuum QCD construction additionally requires critical tuning, local renormalized field bounds, nonfree interaction, and the required continuum symmetries and counterterms in one common hierarchy. Those dynamical estimates remain to be established. The source comparison proved here is available throughout the larger positive-transfer interval.
Declaration of Generative AI and AI-Assisted Technologies in the Manuscript Preparation Process: During the research and preparation of this manuscript, the author used OpenAI ChatGPT 5.6 for selected mathematical calculations and developments, writing and checking code, literature searches, translation, journal formatting, and spelling and language checks. The research idea, scientific direction, development of distinct lines of inquiry, and part of the coding were the author’s own work. The author also conducted a comprehensive independent review of the AI-assisted literature searches. Indeed, the author resisted ChatGPT 5.6’s attempts to steer the research toward familiar, established approaches in the existing literature, instead pursuing an independent line of reasoning and formulating the equations himself. The author reviewed and edited AI-assisted material as needed and bears sole responsibility for the manuscript, including its accuracy, originality, results, and conclusions.

Data Availability Statement

The results are analytical. No experimental data are used.

Conflicts of Interest

The author declares no conflicts of interest.

References

  1. Lüscher, M. Construction of a selfadjoint, strictly positive transfer matrix for Euclidean lattice gauge theories. Commun. Math. Phys. 1977, 54, 283–292. [Google Scholar] [CrossRef]
  2. Osterwalder, K.; Seiler, E. Gauge field theories on a lattice. Ann. Phys. 1978, 110, 440–471. [Google Scholar] [CrossRef]
  3. Frezzotti, R.; Grassi, P.A.; Sint, S.; Weisz, P. Lattice QCD with a chirally twisted mass term. J. High Energy Phys. 2001, 2001, 058. [Google Scholar] [CrossRef]
  4. Lüscher, M.; Palombi, F. Fluctuations and reweighting of the quark determinant on large lattices. PoS 2008, LATTICE2008(049), 0810.0946. [Google Scholar] [CrossRef]
  5. Finkenrath, J.; Knechtli, F.; Leder, B. One flavor mass reweighting in lattice QCD. Nucl. Phys. B;Erratum Nucl. Phys. B 2013, 877 880, 441–456, [1306.3962 574–575. [Google Scholar] [CrossRef]
  6. Ginsparg, P.H.; Wilson, K.G. A remnant of chiral symmetry on the lattice. Phys. Rev. D. 1982, 25, 2649–2657. [Google Scholar] [CrossRef]
  7. Neuberger, H. Exactly massless quarks on the lattice. Phys. Lett. B 1998, 417, 141–144. [Google Scholar] [CrossRef]
  8. Capitani, S.; Göckeler, M.; Horsley, R.; Rakow, P.E.L.; Schierholz, G. Operator improvement for Ginsparg–Wilson fermions. Phys. Lett. B 1999, 468, 150–160. [Google Scholar] [CrossRef]
  9. Vafa, C.; Witten, E. Restrictions on symmetry breaking in vector-like gauge theories. Nucl. Phys. B 1984, 234, 173–188. [Google Scholar] [CrossRef]
  10. Giordano, M. Impossibility of spontaneous vector flavor symmetry breaking on the lattice. Phys. Rev. D. 2023, 107, 114509, [2303.03109. [Google Scholar] [CrossRef]
  11. Yunus, C.; Detmold, W. Infinite variance in Monte Carlo sampling of lattice field theories. Phys. Rev. D. 2022, 106, 094506, [2205.01001. [Google Scholar] [CrossRef]
  12. Güneysu, B.; Milatovic, O.; Truc, F. Generalized Schrödinger semigroups on infinite graphs. Potential Anal. 2014, 41, 517–541, [1306.6062. [Google Scholar] [CrossRef]
  13. Dürr, S.; Weber, J.H. Dispersion relation and spectral range of Karsten–Wilczek and Borici–Creutz fermions. Phys. Rev. D. 2020 Erratum: Phys. Rev. D 108, 059903. 2003, 102, 014516. [Google Scholar] [CrossRef]
  14. Lawrence, S. Model-free spectral reconstruction via Lagrange duality. arXiv 2024, arXiv:2408.11766. [Google Scholar] [CrossRef]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.