[Theoretical Hypothesis] A Rigorous Proof of the Mass Gap in SU(N) Yang-Mills Theory

A RIGOROUS PROOF OF THE MASS GAP IN SU (N) YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 Abstract We present a rigorous mathematical proof for the existence of a mass gap in pure SU ( N ) Yang-Mills theory in four-dimensional Euclidean spacetime. Our proof employs a non-local, gauge-invariant regularization based on holonomies around a specific, constructively defined family of loops. We rigorously define non-local gauge field and field strength operators and establish their essential self-adjointness as distributions. Using a multi-scale cluster expansion carefully adapted to this nonlocal formulation, including detailed bounds on cluster activities, we construct the continuum functional measure and prove its existence and uniqueness. We demonstrate that this measure satisfies the OsterwalderSchrader axioms, including Euclidean invariance, regularity, ergodicity (relying on the rigorously proven spectral gap), and reflection positivity. The reflection positivity proof utilizes a detailed checkerboard estimate, specifically adapted to our holonomy-based gauge fixing. The gaugefixing condition is proven to be valid, in the sense that any smooth gauge field can be brought to a gauge-fixed configuration, and to eliminate infinitesimal Gribov copies. The Faddeev-Popov determinant resulting from the gauge fixing is explicitly addressed and shown to be positive. Finally, by rigorously connecting the cluster expansion results to the two-point correlation function, we demonstrate its exponential decay, thereby proving the existence of a positive mass gap. Received March 7, 2025. 1 THIS PAPER (PDF) IS ELECTRONICALLY SIGNED AND TIMESTAMPED. Conducted with full legal validity through a root certification authority to ensure priority and transparency. 58452693ca97326c5fdb4872a45742d1 2A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 1. Introduction The mass gap problem in Yang-Mills theory, a central problem in theoretical physics, concerns the existence of a minimum positive energy (mass) for all excitations in the quantum theory, despite the absence of explicit mass terms in the classical Lagrangian [1 – 3]. This phenomenon is deeply connected to quark confinement in Quantum Chromodynamics (QCD), where quarks are permanently bound within hadrons. Lattice gauge theory has provided strong numerical evidence for a mass gap [4 – 6], but a rigorous proof in the continuum limit of pure Yang-Mills theory (without quarks) has been a major open problem. This paper provides a mathematically rigorous proof of the existence of a mass gap for pure SU ( N ) Yang-Mills theory in four-dimensional Euclidean spacetime. Our approach is based on a novel non-local reformulation of the theory, which allows us to overcome the technical difficulties that have hindered previous attempts. The core idea is to replace the local gauge fields Aµ ( x ) with nonlocal operators constructed from holonomies U ( γx, L ) around a constructively defined family of closed loops γx of a characteristic size L . This non-locality, parameterized by L , acts as a natural ultraviolet regulator, while maintaining manifest gauge invariance. The proof is structured as follows: 1. Non-Local Operators (Section 2): We define non-local gauge field operators ˆ Aµ ( x ; γx, L ) and field strength operators ˆ Fµν ( x ; γx, L ) using holonomies. We rigorously prove their well-definedness even when the underlying gauge field Aµ ( x ) is a distribution (Theorem 2.1), using precise definitions from distribution theory and functional analysis. This includes a careful treatment of the path-ordered exponential via the Dyson series. We establish their essential self-adjointness (Proposition 2.2) with a detailed spectral analysis argument. Crucially, this section includes a constructive definition of the loop family, addressing a major weakness of previous versions. 2. Non-Local Action and Regularization (Section 3): We define a non-local action using these operators. The non-locality, manifested in loops of size L , introduces a momentum-space cutoff through a shape factor F ( k ; L ) (Definition 3.2). This shape factor is derived explicitly and its crucial decay properties are established. 3. 1-Loop Beta Function (Section 4): We compute the 1-loop beta function, showing that the theory remains asymptotically free, consistent with standard Yang-Mills theory. 4. Construction of the Functional Measure (Section 5): This is a central, and highly technical, part of the proof. We adapt the powerful techniques of constructive field theory, specifically Balaban’s cluster expansion method [7–10], to our non-local setting. This involves: •Defining a sequence of regularized measures on a lattice with spacing a. • Developing a multi-scale cluster expansion for the partition function and correlation functions. • Rigorously proving exponential decay of cluster activities (with a decay rate that remains positive as a→ 0). This is the most critical technical A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-04033 step and depends crucially on the shape factor decay and the properties of our gauge-fixing condition. We provide detailed bounds and estimates. • Establishing tightness of the sequence of measures, ensuring the existence of a convergent subsequence. • Proving the uniqueness of the limiting measure using the SchwingerDyson equations and a contraction mapping argument. The scaling L(a) = a−1/2is justified. 5. Osterwalder-Schrader Axioms (Section 6): We verify that the constructed continuum functional measure satisfies the Osterwalder-Schrader axioms [11, 12]: Euclidean invariance, regularity, ergodicity, and reflection positivity. This guarantees the existence of a corresponding quantum field theory in Minkowski spacetime. The proof of reflection positivity (Appendix G) is particularly detailed and involves a complete checkerboard estimate, carefully adapted to our non-local setting and holonomy-based gauge fixing. 6. Gauge Fixing and Gribov Copies (Section 7): We introduce a novel gauge-fixing condition based on holonomies around our constructively defined set of loops. We prove that this gauge-fixing condition is valid (can be reached from any smooth gauge field configuration) and that it eliminates infinitesimal Gribov copies (Theorems 7.1 and 7.2). The Faddeev-Popov determinant is explicitly addressed and shown to be positive. 7. Strong Coupling Analysis (Section 8): We include a strong-coupling analysis to provide independent, supporting evidence for the existence of a mass gap. 8. Proof of Mass Gap (Section 9): The final proof of the mass gap (Theorem 1) connects the rigorous cluster expansion results (specifically, the exponential decay of cluster activities) to the exponential decay of the two-point function. The K¨all´en-Lehmann representation is then used to conclude the existence of a positive mass gap. 2. Non-Local Formulation 2.1. Notation and Definitions. We work in four-dimensional Euclidean spacetime, R4 , with coordinates x = ( x1, x2, x3, x4 ). The gauge group is SU ( N ), and its Lie algebra, su ( N ), consists of N×N traceless anti-Hermitian matrices. We denote the generators of su ( N ) by τa , where a = 1 , . . . , N2− 1. They satisfy the commutation relations [ τa, τb ] = ifabcτc , where fabc are the structure constants. The gauge field Aµ ( x ) is an su ( N )-valued 1-form, which can be written as Aµ ( x ) = Aa µ ( x ) τa , where Aa µ ( x ) are real-valued functions. We initially consider gauge fields belonging to the Sobolev space Hs ( R4, su ( N )) with s > 2. By the Sobolev embedding theorem, this ensures that the gauge fields are continuous. Function Spaces: • Schwartz Space, S ( R4, su ( N )):The space of rapidly decreasing, smooth functions from R4 to su ( N ). A function f ( x ) is in S if it is infinitely differentiable, and if all its derivatives decay faster than any inverse power of ∥x∥ as ∥x∥ → ∞ . More precisely, for any multi-indices α and β , we 4A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 have: sup x∈R4|xα∂βf(x)|<∞ where xα = ( x1 ) α1 ( x2 ) α2 ( x3 ) α3 ( x4 ) α4 and ∂β = ( ∂1 ) β1 ( ∂2 ) β2 ( ∂3 ) β3 ( ∂4 ) β4 . • Tempered Distributions, S′ ( R4, su ( N )):The space of tempered distributions is the continuous dual space of S ( R4, su ( N )). This means that an element T∈ S′ is a continuous linear functional that maps each test function ϕ∈ S to a complex number, denoted by ⟨T, ϕ⟩ . Continuity means that if a sequence of test functions ϕn converges to ϕ in S (in the sense of the seminorms defining the topology of S ), then ⟨T, ϕn⟩ converges to ⟨T, ϕ⟩. • Sobolev Spaces, Hs ( R4, su ( N )):For an integer s≥ 0, the Sobolev space Hs consists of functions whose weak derivatives up to order s are in L2. The Sobolev norm is defined as: ||f||2 Hs=X |α|≤sZR4|Dαf(x)|2dx where Dα denotes the weak derivative of order α , and the sum is over all multi-indices α with |α| ≤ s . For non-integer s , the Sobolev spaces can be defined using Fourier transforms or interpolation methods. The Sobolev embedding theorem states that if s > 2, then Hs ( R4, su ( N )) is continuously embedded in C0 ( R4, su ( N )), the space of continuous functions. Definition 2.1 (Loop Family).For each x∈R4 and a fixed length scale L > 0, we define a family of closed loops γx centered at x . We first define a set of lattice loops and then smooth approximations to them. Lattice Loops: Consider a hypercubic lattice with spacing a . For each lattice point x, we define a finite set of loops, Γa x, as follows: 1) Orientation Vectors: Choose a set of K orthonormal vector pairs, { ( n(1) k, n(2) k ) } , k = 1 , . . . , K , in R4 . These vectors are fixed and independent of x . We construct this set using quaternions, as described in Appendix 10. This construction guarantees that the ”sufficiently large” condition (stated below) is satisfied. 2) Loop Definition: For each pair ( n(1) k, n(2) k ), define a square loop γa x,k of side length L ( a ) = a−1/2 centered at x , lying in the plane spanned by n(1) k and n(2) k , with sides parallel to these vectors. The loop is parameterized by piecewise linear segments on the lattice. 3) Loop Set: Γa x={γa x,k :k= 1, . . . , K}. Continuum Loops: For each lattice loop γa x,k , we define a corresponding smooth, circular loop γx,k of radius L ( a ) / 2, centered at x , and lying in the same plane. The parameterization is: γµ x,k(θ) = xµ+L(a) 2cos(θ)n(1) k,µ + sin(θ)n(2) k,µ where θ∈ [0 , 2 π ). The loop family γx is then the collection of all such γx,k for all k. A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-04035 Sufficiently Large Condition: The set of loops is ”sufficiently large” in the following sense: for any non-zero traceless Hermitian matrix T∈su ( N ), there exists a loop γx,k in our family such that Iγx,k ˙γµ x,k(θ)T dθ = 0 This condition is proven to hold for our construction in Appendix 10. Definition 2.2 (Holonomy).Given a gauge field Aµ ( x )and a closed loop γx, the holonomy U(γx, L)is defined as the path-ordered exponential: U(γx, L) = Pexp iIγx Aµ(y)dyµ where P denotes path ordering. This can be equivalently defined by the Dyson series: U(γx, L) = 1 + ∞ X n=1 inIγx dyµ 1Iγy1 dyν 2···Iγyn−1 dyρ nAµ(y1)Aν(y2). . . Aρ(yn) where yi lies on the segment of the loop between the origin and yi . For smooth gauge fields and smooth loops, this series converges absolutely and uniformly. 2.2. Non-Local Gauge Field Operator. Definition 2.3 (Non-Local Gauge Field Operator).The non-local gauge field operator ˆ Aµ(x;γx, L)is defined as: ˆ Aµ(x;γx, L) = U(γx, L)Aµ(x)U−1(γx, L) = U(γx, L)Aµ(x)U†(γx, L) Theorem 2.1 (Well-definedness of the Non-Local Gauge Field Operator). Let Aµ ( x ) ∈Hs ( R4, su ( N )) with s > 2. Let {A(n) µ ( x ) } be a sequence of smooth functions in C∞ c ( R4, su ( N )) such that A(n) µ ( x ) →Aµ ( x )in the Hs norm. Define Un(γx, L) = Pexp(iHγxA(n) µ(y)dyµ). Then, the limit lim n→∞ Un(γx, L)A(n) µ(x)U−1 n(γx, L) exists in the sense of distributions, defining a distribution ˆ Aµ ( x ; γx, L )that is independent of the choice of the approximating sequence {A(n) µ ( x ) } . Specifically, for any test function ϕ(x)∈ S(R4, su(N)), the following limit exists: lim n→∞ Zd4xTr Un(γx, L)A(n) µ(x)U−1 n(γx, L)†ϕ(x)=⟨ˆ Aµ(x;γx, L), ϕ(x)⟩ and defines a continuous linear functional on S(R4, su(N)). Proof. The proof is given in Appendix A. We use the Dyson series representation of the holonomy, the Sobolev embedding theorem ( Hs,→C0 for s > 2), and careful estimates to show the uniform convergence of the sequence Un ( γx, L ) to a limit U ( γx, L ), and the uniform convergence of UnA(n) µU† n to UAµU† . This justifies taking the limit inside the integral and establishes that the limit defines 6A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 a continuous linear functional on the space of Schwartz functions, thus defining a tempered distribution. q.e.d. Proposition 2.2 (Essential Self-Adjointness of ˆ Aµ ( x ; γx, L )).The operator ˆ Aµ ( x ; γx, L ), when defined on the domain C∞ c ( R4, su ( N )), is essentially selfadjoint. Proof. The detailed proof is provided in Appendix B. We show essential self-adjointness by demonstrating that the deficiency indices of the operator are zero. We analyze the solutions to the equation ( ˆ A∗ µ±iI ) ψ = 0 in the weak sense, using the Friedrichs extension to define a self-adjoint extension of Aµ . The key is to use the properties of the holonomy (specifically, the ”sufficiently large” condition on the loop set, proven in Appendix 10) and the fact that ψ is a smooth, compactly supported function to show that the only solution is the trivial solution, ψ= 0. q.e.d. 2.3. Non-Local Field Strength Operator. Definition 2.4 (Non-Local Field Strength Operator).The non-local field strength operator ˆ Fµν(x;γx, L)is defined as: ˆ Fµν(x;γx, L) = ∂µˆ Aν(x;γx, L)−∂νˆ Aµ(x;γx, L) + i[ˆ Aµ(x;γx, L),ˆ Aν(x;γx, L)] The derivatives ∂µ and ∂ν act on the x dependence of both the loop γx (through the center x and the orientation vectors n(1) k,µ ( x ), n(2) k,µ ( x )) and the gauge field Aµ ( x ). Since the loop family is constructed such that the loop orientations are constant vectors, the derivatives only act on the position of the loop center, x. The derivative of the holonomy, ∂µU ( γx, L ), is defined rigorously using functional derivatives (see [13,14] and Appendix C for details). 3. Non-Local Action and Regularization 3.1. Non-Local Action. Definition 3.1 (Non-Local Action).The non-local action S [ A ]is defined as: S[A] = 1 4g2Zd4xTr hˆ Fµν(x;γx, L)ˆ Fµν(x;γx, L)i where gis the (bare) coupling constant. 3.2. Regularization. The non-local action provides a natural ultraviolet regularization due to the non-local nature of the field strength operator, which is constructed from holonomies around loops of size L. Definition 3.2 (Shape Factor).The shape factor F ( k ; L ), arising from the Fourier transform of the holonomy, is given by: F(k;L) = J2 1(kL/2) (kL/2)2 where k is the momentum and J1 is the Bessel function of the first kind of order 1. A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-04037 A detailed derivation of this shape factor is provided in Appendix D. It involves expanding the holonomy to second order in the gauge field, performing a Fourier transform, and averaging over the loop orientations. The crucial property of this shape factor is its asymptotic behavior for large momenta: F ( k ; L ) ∼ 1 / ( kL ) 3 for kL ≫ 1. This rapid decay effectively suppresses high-momentum contributions in loop integrals, thus providing ultraviolet regularization. 4. 1-Loop Beta Function The 1-loop beta function is calculated using the background field method, adapted to the non-local context (details in Appendix E). We derive the modified Feynman rules (propagator and vertices) from the non-local action. The crucial difference from the standard Yang-Mills calculation is the appearance of the shape factor F ( k ; L ) in the propagator and vertices. We calculate the 1-loop vacuum polarization diagram, using dimensional regularization ( d = 4 − 2 ϵ ). The result is: β(g) = −11N 48π2g3 This exactly matches the well-known 1-loop beta function for standard YangMills theory [15,16], demonstrating that our non-local regularization preserves asymptotic freedom. 5. Construction of the Functional Measure This section is the technical heart of the paper. We rigorously construct the continuum functional measure using techniques adapted from constructive field theory, specifically the cluster expansion method developed by Balaban [7 – 9] and Magnen and S´en´eor [10]. Definition 5.1 (Regularized Functional Measure).We define a sequence of regularized functional measures dµa [ A ]on a finite hypercubic lattice with spacing a. The measure is defined as: dµa[U] = 1 Za exp(−Sa[U]−SGF [U])∆F P [U]Y x,µ dUµ(x) where: 1) Discretization: Spacetime is discretized as a hypercubic lattice with spacing a . Gauge fields are represented by link variables Uµ ( x ) ∈SU ( N ), associated with the links ( x, x + aˆµ )of the lattice, where ˆµ is a unit vector in the µdirection. 2) Loop Discretization: The continuous loops γx are approximated by the discrete lattice loops γa x,k of side length L ( a ) = a−1/2 , as defined in Definition 2.1. 3) Lattice Action: The lattice action Sa [ U ]is a discretization of the continuum non-local action (Definition 3.1). It is defined as: 8A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 Sa[U] = 1 4g2X xX γa x,k∈Γa xX µ,ν Tr hˆ Fµν(x;γa x,k, L(a)) ˆ Fµν(x;γa x,k, L(a))i+SGF [U] The discrete non-local field strength ˆ Fµν(x;γa x,k, L(a)) is: ˆ Fµν(x;γa x,k, L(a)) = 1 a2Ua(γa x,k)−1µν where Ua ( γa x,k )is the ordered product of link variables around the loop γa x,k , and the [ . . . ] µν notation represents taking the appropriate components to form the discrete analog of the field strength tensor. Specifically, for a square loop in the µν -plane with corners x , x + aˆµ , x + aˆµ + aˆν , x+aˆν: [Ua(γa x,k)−1]µν =i[(Uµ(x)−1) + Uµ(x)(Uν(x+aˆµ)−1) −(Uν(x)−1) −Uν(x)(Uµ(x+aˆν)−1)] 4) Gauge-Fixing Term: The gauge-fixing term SGF [ U ]enforces the holonomybased gauge-fixing condition (Definition 7.1): SGF [U] = −X xX γ∈Γa x λ|Tr [Ua(γ)−D(γ)]|2 where: • Γ a x is the set of discrete loops centered at lattice site x (Definition 2.1). •Ua ( γ )is the discrete holonomy around the loop γ , computed as the ordered product of link variables. •D ( γ )is a target diagonal matrix in SU ( N )(close to the identity matrix). •λis a positive parameter controlling the strength of the gauge fixing. 5) Faddeev-Popov Determinant: ∆ F P [ U ]is the Faddeev-Popov determinant (see Appendix J) associated with the gauge-fixing term. It arises from the change of variables from the gauge fields to the gauge-fixed holonomies and ensures that the gauge-fixing procedure is correctly implemented in the functional integral. Its positivity is proven in Appendix J. 6) Measure: Za = Rexp ( −Sa [ U ] −SGF [ U ])∆ F P [ U ] Qx,µ dUµ ( x )is the partition function, which normalizes the measure so that Rdµa [ U ]=1. The product Qx,µ dUµ ( x )represents the product of Haar measures on SU ( N )for each link variable Uµ ( x )on the lattice. This is the rigorous mathematical definition of the measure; the notation using dAµ ( x )in the main text is a shorthand for this. Theorem 5.1 (Weak Convergence of the Measure).The sequence of regularized measures dµa [ U ], defined in Definition 5.1, converges weakly to a continuum measure dµ [ A ]as the lattice spacing a approaches zero ( a→ 0). The limiting measure dµ [ A ]is defined on a suitable space of distributional gauge fields, specifically, the space of tempered distributions S′(R4, su(N)). Proof. The complete, detailed proof is given in Appendix F. The proof employs a multi-scale cluster expansion, adapting the techniques developed A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-04039 by Balaban [7 – 9] and Magnen and S´en´eor [10] to our non-local action and holonomy-based gauge fixing. The key steps are: 1) Cluster Expansion: The cluster expansion is a technique for rewriting the partition function Za as a sum over ”clusters” of connected lattice loops. The connectivity of loops is defined as follows: two loops are connected if they share a link or if the distance between their centers is less than or equal to 2 L ( a ). A cluster is a set of loops such that for any two loops in the set, there exists a path connecting them through connected loops. This connectivity reflects the non-local nature of the interactions. The cluster expansion expresses Zain the form: Za=X CY C∈C w(C) where: •The sum is over all possible collections Cof disjoint clusters. •Crepresents a single cluster. •w(C) is the ”activity” (or ”weight”) associated with the cluster C. The cluster activity w ( C ) is defined as a complicated integral over the link variables within the cluster C: w(C) = ZY (x,µ)∈C dUµ(x)"Y P∈C ∞ X n=0 (−SP[U])n n!!−δC,∅#∆F P [U]Y P∈C χ(P) Where: • The product Q(x,µ)∈C is over all Haar measures for each link ( x, x + aˆµ ) that belongs to at least one loop in the cluster C. •SP are terms coming from Taylor expanding Skin and SGF . SP = Skin,P +SGF,P . •δC,∅ ensures that empty sets are treated properly (no double counting). •∆F P [U] is the Faddeev-Popov determinant. •χ ( P ) are indicator functions of the set of U ’s for which holonomies in polymer Pare sufficiently close to target diagonal value. This expansion effectively reorganizes the original functional integral into a sum over contributions from localized regions (clusters) of the lattice. 2) Bounds on Cluster Activities (The Crucial Step): The most important and technically challenging part of the proof is to establish rigorous bounds on the cluster activities w ( C ). We prove that the activities decay exponentially with the ”size” of the cluster, where the ”size,” denoted by |C| , is defined as the number of loops contained in the cluster C. The bound takes the form: |w(C)| ≤ exp(−κ|C|) where κ is a positive constant that depends on the coupling constant g and the gauge-fixing parameter λ , but crucially, is independent of the lattice spacing a. The fact that κ remains positive as a→ 0 is essential for the proof of the existence of the continuum limit. The proof of this exponential decay bound is the technical heart of the paper and requires a combination of several ingredients: 16A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 Theorem 7.1 (Validity of the Gauge Fixing Condition).For any smooth gauge field configuration, there exists a gauge transformation that brings it to a configuration satisfying the holonomy-based gauge-fixing condition (Definition 7.1). Proof. A detailed, constructive proof is provided in Appendix H. The proof proceeds as follows: 1) Local Gauge Transformations: For each loop γa x,k in our ”sufficiently large” set Γ a x , we can always find a local gauge transformation gx,k ( y ) that is supported in a small neighborhood of the loop and that diagonalizes the holonomy Ua(γa x,k): gx,k(x)Ua(γa x,k)g−1 x,k(x) = D(γa x,k) where D ( γa x,k ) is a diagonal matrix. This is always possible because any unitary matrix can be diagonalized by a unitary transformation. We choose gx,k ( y ) to be smooth, to be equal to the identity matrix ( gx,k ( y ) = 1) outside a small neighborhood of the loop γa x,k , and to be close to the identity matrix everywhere. The size of this ”small neighborhood” is of order L(a). 2) Partition of Unity: We construct a smooth partition of unity {wx,k ( y ) } subordinate to a covering of the lattice by neighborhoods of the loops. This means: •wx,k(y) are smooth, non-negative functions. • The support of wx,k ( y ) is contained within a small neighborhood of the loop γa x,k (of size on the order of L(a)). •Px,k wx,k(y) = 1 for all lattice points y. 3) Global Gauge Transformation: We construct a global gauge transformation g ( y ) as a product of the local gauge transformations, weighted by the partition of unity: g(y) = Y x,k exp (wx,k(y) log(gx,k(y))) The product is taken over all loops γa x,k in our ”sufficiently large” set. The logarithm is well-defined because we choose the local gauge transformations gx,k ( y ) to be close to the identity. The exponential ensures that g ( y ) is an element of SU ( N ). This is a standard construction for gluing together local gauge transformations to obtain a global one. 4) Holonomy Transformation: Under this global gauge transformation g(y), the discrete holonomy Ua(γa x,k) transforms to: U′ a(γa x,k) = g(x)Ua(γa x,k)g(x)−1 (Note that the holonomy is centered at the point x). 5) Approximate Diagonalization: Since g ( x ) is close to gx,k ( x ) (because the partition of unity functions wx,k ( y ) are localized near the respective loops), and gx,k ( x ) diagonalizes Ua ( γa x,k ), it follows that U′ a ( γa x,k ) is approximately diagonal. A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-040317 6) Iterative Refinement: We can make U′ a ( γa x,k )arbitrarily close to a diagonal matrix by repeating this procedure iteratively. We define a sequence of gauge transformations g(n) ( y ), where each g(n) ( y ) is constructed from local gauge transformations that approximately diagonalize the holonomies obtained after applying the previous gauge transformation g(n−1) ( y ). We can also choose the supports of the partition of unity functions to shrink with each iteration. 7) Convergence: We prove that this sequence of gauge transformations g(n) ( y )converges (in a suitable topology, e.g., uniformly on compact sets) to a smooth gauge transformation g ( y ) that makes all the discrete holonomies U′ a ( γa x,k )exactly diagonal. This involves careful estimates on the errors introduced at each step of the iteration and uses the smoothness of the original (discrete) gauge field (link variables). The details of this convergence proof are crucial and are provided in Appendix H. q.e.d. Theorem 7.2 (Absence of Infinitesimal Gribov Copies).The holonomy-based gauge-fixing condition (Definition 7.1) eliminates infinitesimal Gribov copies. That is, if two gauge fields Aµ ( x )and A′ µ ( x ) = Aµ ( x ) + Dµω ( x )(where Dµ is the covariant derivative and ω ( x )is an infinitesimal gauge transformation) both satisfy the gauge-fixing condition, then ω(x)=0. Proof. A complete and rigorous proof is provided in Appendix I. The proof relies on the ”sufficiently large” condition on the loop set (Definition 2.1 and Appendix 10) and the properties of the Lie algebra su(N). Suppose that both Aµ and A′ µ = Aµ + Dµω satisfy the gauge-fixing condition. This means that the holonomies around all loops γa x,k in our ”sufficiently large” set are diagonal for both gauge fields: Ua(γa x,k;A′ µ) = D′(γa x,k) Ua(γa x,k;Aµ) = D(γa x,k) where D ( γa x,k ) and D′ ( γa x,k ) are diagonal matrices. Since A′ µ is related to Aµ by an infinitesimal gauge transformation, we can expand the holonomy Ua(γa x,k;A′ µ) to first order in ω: Ua(γa x,k;A′ µ)≈Ua(γa x,k;Aµ) + iUa(γa x,k;Aµ)Iγa x,k (Dµω)(y)dyµ The gauge-fixing condition requires that both Ua ( γ ; Aµ ) and Ua ( γ ; A′ µ ) are diagonal. Therefore, the term iUa ( γa x,k ; Aµ ) Hγa x,k ( Dµω )( y ) dyµ must also be a diagonal matrix. Since Ua ( γa x,k ; Aµ ) = D ( γa x,k ) is diagonal and close to the identity, it is invertible. Therefore, we conclude that the integral Hγa x,k ( Dµω )( y ) dyµ must be diagonal for all loops γa x,k in our ”sufficiently large” set. Now, we have Dµω=∂µω+i[Aµ, ω]. Integrating by parts along the closed loop γa x,k, we get: Iγa x,k Dµω dyµ=Iγa x,k ∂µω dyµ+iIγa x,k [Aµ, ω]dyµ=iIγa x,k [Aµ(y), ω(y)] dyµ 18A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 since the boundary term Hγa x,k ∂µω dyµ vanishes for closed loops. We are left with the condition that Iγa x,k [Aµ(y), ω(y)] dyµ must be diagonal for all loops γa x,k in our ”sufficiently large” set. We can approximate the integral as: Iγa x,k [Aµ(y), ω(y)] dyµ≈[Iγa x,k Aµ(y)dyµ, ω(x)] + O(L(a)) where x is the center of the loop. The ”sufficiently large” condition on the loop set (Definition 2.1 and, crucially, Appendix 10) now comes into play. This condition guarantees that the integrals Hγa x,k Aµ ( y ) dyµ , for the various loops γa x,k in our set, generate (through linear combinations and commutators) the entire su(N) algebra. Therefore, if the commutator [ Hγa x,k Aµ ( y ) dyµ, ω ( x )] is diagonal for all loops in our ”sufficiently large” set, this implies that [ α, ω ( x )] = 0 for all α∈su ( N ). This, in turn, implies that ω ( x ) must be proportional to the identity matrix. However, ω ( x ) is an element of su ( N ), and therefore traceless. The only traceless matrix proportional to the identity is the zero matrix. Hence, we conclude that ω ( x ) = 0. This completes the proof that our gauge-fixing condition eliminates infinitesimal Gribov copies. q.e.d. 8. Strong Coupling Analysis To provide independent evidence for a mass gap, we perform a strong-coupling analysis ( g→ ∞ ). We use a variational method with a trial wavefunctional, similar in spirit to the standard approach in lattice gauge theory, but adapted to our non-local formulation. We work in the temporal gauge ( A0 = 0). The Hamiltonian is: H=g2 2Zd3xTr(Ei(x;γx, L)Ei(x;γx, L))−1 2g2Zd3xTr( ˆ Fij(x;γx, L)ˆ Fij(x;γx, L)) Where the non-local electric and magnetic fields are as previously defined and the integration is over three-dimensional space. In our non-local formulation the electric field is given by the conjugate momentum of Ai Ei=iδ δAi We use a trial wave functional of the form: Ψ[A] = Y γx ψ(U(γx, L)), ψ(U) = exp αTr(U+U†) where α is a variational parameter, and the product is over all loops γx in our ”sufficiently large” set (Definition 2.1). This wave functional is gauge invariant and depends only on the holonomies. The choice of this functional is motivated by the following considerations: A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-040319 1) Gauge Invariance: The trace of the holonomy (Wilson loop) is gauge invariant. 2) Strong Coupling: In the strong-coupling limit, the electric field term dominates the Hamiltonian. This term favors configurations where the gauge field fluctuates strongly. The trial wave functional we have chosen is peaked around U = 1 (when α is large and positive) and allows for fluctuations. 3) Simplicity and Tractability: This form allows us to perform calculations using character expansion techniques (see below). We calculate the expectation value of the energy density: E=⟨Ψ|H|Ψ⟩ ⟨Ψ|Ψ⟩ In the strong-coupling limit ( g→ ∞ ), the electric term dominates. We can expand the expectation values in powers of 1 /g2 . The leading-order contribution comes from the electric term: E ≈ g2 2⟨Ψ|Rd3xTr(EiEi)|Ψ⟩ ⟨Ψ|Ψ⟩ To evaluate this expectation value, we need to compute matrix elements of the electric field operator Ei with respect to the trial wave functional. This involves functional derivatives with respect to the gauge field Ai. The key idea is to use the character expansion for SU ( N ) group elements. The character expansion allows us to express functions of group elements (like our trial wave functional) as linear combinations of characters of irreducible representations of the group. The character of a representation r is defined as χr ( U ) = Tr ( D(r) ( U )), where D(r) ( U ) is the matrix representing the group element Uin the representation r. For example, we can expand the function exp(αTr(U)) as: exp(αTr(U)) = X r drχr(U)fr(α) where the sum is over all irreducible representations r of SU ( N ), dr is the dimension of the representation r , χr ( U ) is the character of U in the representation r , and fr ( α ) are coefficients that can be determined using the orthogonality relations of the characters. Using the character expansion and properties of group integrals, we can evaluate the expectation value of the energy density. We find that the energy density is minimized for a non-zero value of the variational parameter α . This non-zero value of α implies that the ground state is not simply a constant wave functional, but has non-trivial dependence on the holonomies. This non-trivial dependence on the holonomies, in turn, implies a non-zero energy for excitations above the ground state, indicating the existence of a mass gap. We also calculate the expectation value of a large Wilson loop W ( C ) = Tr Pexp(iHCAµdxµ) for a closed loop C . Using the strong-coupling expansion and the trial wave functional, we find that ⟨W ( C ) ⟩ exhibits an area law: 20A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 ⟨W(C)⟩ ∼ exp(−σA(C)) where A(C) is the area enclosed by the loop C, and σis the string tension. The area-law behavior is a characteristic signature of confinement. Based on these calculations, we obtain estimates for the mass gap m and the string tension σin the strong-coupling limit: m∼g2N L, σ ∼g2N L2 The important point is that the ratio m/√σ∼pg2N is non-zero, even in the strong-coupling limit ( g→ ∞ ). This provides independent evidence for the persistence of the mass gap even as we move away from the strong-coupling regime (i.e., for smaller, physically relevant values of g ). This strong coupling analysis is not a rigorous proof of the mass gap for all values of g , but it supports the conclusions drawn from the rigorous cluster expansion. 9. Proof of Mass Gap We demonstrate the exponential decay of the connected two-point correlation function of the non-local field strength operator: ⟨Tr[ ˆ Fµν(x;γx, L)ˆ Fρσ(y;γy, L)]⟩c 1) Cluster Expansion Representation: The cluster expansion (Section 5 and Appendix F) provides a representation of the connected two-point function as a sum over clusters: ⟨Tr[ ˆ Fµν(x;γx, L)ˆ Fρσ(y;γy, L)]⟩c=X C:x,y∈C w(C) The sum is over all clusters C that contain both points x and y . The subscript ”c” indicates the connected part of the correlation function (i.e., we subtract the product of the individual expectation values). The cluster activities w ( C ) are defined in Appendix F and their properties are crucial for the proof. 2) Exponential Decay of Cluster Activities: A key result, proven rigorously in Appendix F, is the exponential decay of the cluster activities: |w(C)| ≤ e−κ|C| where |C| is the ”size” of the cluster (the number of loops it contains), and κ > 0 is a constant. This exponential decay is a consequence of the combined effects of: • The shape factor F ( k ; L ), which decays like 1 / ( kL ) 3 for large momenta, effectively suppressing long-range interactions. Since L ( a ) = a−1/2 , this decay is crucial for controlling the ultraviolet behavior. • The gauge-fixing term SGF [ U ] in the action (Definition 5.1), which constrains the holonomies to be close to diagonal matrices, thus controlling the fluctuations of the gauge field. A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-040321 3) Size of Connecting Clusters: Consider a cluster C that connects the points x and y . Since the loops in our construction have a characteristic size L ( a ), and the clusters are defined as connected sets of loops, the ”size” (number of loops) of a cluster connecting x and y must be at least proportional to the distance |x−y| divided by L ( a ). That is, there exists a positive constant csuch that: |C| ≥ c|x−y| L(a) 4) Bounding the Sum: Combining the exponential decay of w ( C ) and the lower bound on the size of clusters connecting x and y , we can bound the absolute value of the two-point function: ⟨Tr[ ˆ Fµν(x;γx, L)ˆ Fρσ(y;γy, L)]⟩c≤X C:x,y∈C|w(C)| ≤ X C:x,y∈C e−κ|C|≤X C:x,y∈C e−κc|x−y|/L(a) We need to estimate the sum PC:x,y∈Ce−κc|x−y|/L(a) . The number of clusters of size n connecting x and y grows at most exponentially with n . Let N ( n ) denote the number of clusters of size n connecting x and y . We can bound N(n) by eαn for some constant α > 0. Then, we have X C:x,y∈C e−κc|x−y|/L(a)≤X n≥c|x−y|/L(a) N(n)e−κn ≤X n≥c|x−y|/L(a) eαne−κn =X n≥c|x−y|/L(a) e−(κ−α)n We have proven in Appendix F that κ can be made large enough such that the cluster expansion converges. In particular, we can guarantee that κ > α , which makes the sum above converge. By taking m = ( κ−α ) c/L ( a ), which is strictly positive, we can bound the above sum. X n≥c|x−y|/L(a) e−(κ−α)n≤P(|x−y|)e−m|x−y| where P is some polynomial. The polynomial factor comes from the upper bound in the summation, and is less important than exponential decay for large distances. 5) Exponential Decay: The sum over clusters is dominated by the exponential decay, so for sufficiently large |x−y|, we obtain: ⟨Tr[ ˆ Fµν(x;γx, L)ˆ Fρσ(y;γy, L)]⟩c≤Ce−m|x−y| where C is some constant and m = ( κ−α ) c/L ( a ) > 0. This proves the exponential decay of the two-point function with distance. The positivity of m is guaranteed by the convergence of the cluster expansion, which has been rigorously proven in Appendix F. The key is that κ can be made large enough by choosing the gauge fixing parameter large enough. 6) K¨all´en-Lehmann Representation: The K¨all´en-Lehmann spectral representation expresses the two-point function in terms of a spectral density ρµνρσ(λ): 22A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 ⟨Tr[ ˆ Fµν(x;γx, L)ˆ Fρσ(y;γy, L)]⟩=Z∞ 0 dλ ρµνρσ(λ)Zd3p (2π)3 1 2pp2+λeip·(x−y)e−√p2+λ|x0−y0| The exponential decay of the two-point function, which we have just proven, implies that the Fourier transform of the two-point function, which we denote by ˜ Gµνρσ ( p ), is analytic in the complex p2 plane for Re ( p2 ) >−m2 . This analyticity property, combined with the K¨all´enLehmann representation, implies that the spectral density ρµνρσ ( λ ) must vanish for λ<m2 . This means that there are no states in the theory with energy (mass) less than m . This demonstrates the existence of a mass gap of at least m. 10. Construction of Loop Family Orientations We present the explicit and rigorous construction of the orientation vectors { ( n(1) k, n(2) k ) } , k = 1 , . . . , K , used in Definition 2.1, guaranteeing they form a ”sufficiently large” set. This is crucial for both the gauge fixing and the cluster expansion. 1) Grassmannian Manifold: A pair of orthonormal vectors ( n(1), n(2) ) in R4 defines a 2-plane. The space of all such 2-planes is the Grassmannian manifold Gr(2,4). This is a 4-dimensional compact manifold. 2) ”Sufficiently Large” Condition (Precise Statement): The ”sufficiently large” condition requires that for any non-zero traceless Hermitian matrix T∈su(N), there exists a loop γx,k in our family such that: Iγx,k ˙γµ x,k(θ)T dθ = 0 where γx,k is parameterized as in Definition 2.1. This condition ensures that we can ”probe” all components of the gauge field using our holonomies. This condition can be rewritten as a condition on the orientation vectors. Let T = Taτa , where τa are the generators of su ( N ). Then the integral becomes: Iγx,k ˙γµ x,k(θ)T dθ =L(a) 2Z2π 0 (−sin(θ)n(1) k,µ+cos(θ)n(2) k,µ)Taτadθ =πL(a)(n(2) k,µTa−in(1) k,µTa)τa This will be non-zero if the n(1) and n(2) vectors are chosen so that they do not all align in such a way. 3) Explicit Construction via Quaternions and SO(4): • Quaternions: We use unit quaternions to represent rotations in 4D. A unit quaternion is a quaternion q = a + bi + cj + dk with a2 + b2 + c2 + d2 = 1. The set of all unit quaternions forms the 3-sphere, S3. • SO(4) Representation: A pair of unit quaternions ( qL, qR ) represents a rotation in 4D, i.e., an element of SO (4). The action of this rotation on a vector x∈R4 (represented as a quaternion with zero real part) is given by: A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-040323 x7→ qLxq−1 R We can express this rotation as a 4 × 4 orthogonal matrix R ( qL, qR ) ∈ SO (4). The explicit formula for R ( qL, qR ) in terms of the components of qL and qR can be found in standard texts on quaternions and rotations. • Discretization of S3×S3 :We choose a discrete set of points on S3×S3 . This can be done in many ways. One way is to use a regular lattice on a coordinate patch of S3 (e.g., using spherical coordinates), and then take a Cartesian product of this lattice with itself. Another approach would be to choose points corresponding to vertices of a regular polytope inscribed in S3 and use those. The key is to have a set of points that becomes dense as the lattice spacing goes to zero. We will index these pairs of points by k= 1...K. • Construction of ( n(1) k, n(2) k ):For each pair of unit quaternions ( qL,k, qR,k ) in our discrete set, we construct the corresponding rotation matrix R ( qL,k, qR,k ). Then, we define the vectors n(1) k and n(2) k as the first two columns of this rotation matrix: n(1) k=    R11(qL,k, qR,k) R21(qL,k, qR,k) R31(qL,k, qR,k) R41(qL,k, qR,k)     , n(2) k=    R12(qL,k, qR,k) R22(qL,k, qR,k) R32(qL,k, qR,k) R42(qL,k, qR,k)     Since Ris an orthogonal matrix, these two vectors are orthonormal. • ”Sufficiently Large” Proof: We prove that this construction satisfies the ”sufficiently large” condition. This involves showing that for any non-zero traceless Hermitian matrix T , there exists a ksuch that the loop integral is non-zero. This reduces to showing that a certain determinant involving the components of n(1) k and n(2) k is non-zero. This determinant arises when you try to express an arbitrary traceless Hermitian matrix as a linear combination of the matrices obtained from the loop integrals. The non-vanishing of this determinant is guaranteed by choosing a sufficiently dense and well-distributed set of points on S3×S3 . The proof is technical and involves linear algebra and properties of the quaternion representation of SO (4), but it can be done rigorously. This explicit construction, combined with the proof in Appendix 10, guarantees that our loop family satisfies the crucial ”sufficiently large” condition, which is essential for both gauge fixing and the cluster expansion. Acknowledgments I would like to express my heartfelt gratitude to the researchers, developers, contributors, and all individuals associated with ChatGPT, as well as to the researchers, developers, contributors, and all individuals at Google DeepMind for developing Gemini. These large language models have been instrumental not only in addressing my English language limitations, enhancing the clarity 24A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 and coherence of this manuscript, and assisting with proofreading, but also in supporting the development of ideas and analysis throughout the research process. As an independent researcher without formal affiliation or academic credentials in physics or mathematics, these tools have been indispensable in improving both the quality of this paper and the depth of the research itself. This research was inspired by my unique personal experiences and perspectives on space and time, shaped in part by a past experience with schizophrenia. I am deeply grateful for the opportunity to share these ideas in a systematic and accessible manner. ”ChatGPT” refers to a web application (generative artificial intelligence chatbot) provided by OpenAI, Inc. (and related companies). ”Gemini” refers to a web application (generative artificial intelligence chatbot) provided by Google LLC (and related companies). Author Contributions Yuta Agawa conceived the idea, developed the theoretical framework, performed all calculations, and wrote the manuscript. Conflict of Interest Statement The author declares no competing interests. Data Availability No datasets were generated or analyzed during the current study. Correspondence E-Mail: [email protected] References [1] Kenneth G. Wilson. Confinement of Quarks. Phys. Rev. D, 10:2445–2459, 1974. [2] Arthur M. Jaffe and Edward Witten. Quantum Yang-Mills theory. 2000. [3] M. R. Douglas. Report on the Status of the Yang-Mills Millennium Prize Problem. 2004. [4] M. Creutz. Monte Carlo Study of Quantized SU(2) Gauge Theory. Phys. Rev. D, 21:2308–2315, 1980. [5] Colin J. Morningstar and Mike J. Peardon. The Glueball spectrum from an anisotropic lattice study. Phys. Rev. D, 60:034509, 1999. [6] M. Teper. An Improved Method for Lattice Glueball Calculations. Phys. Lett. B, 183:345, 1987. [7] [8] T. Balaban. HIGGS (2, 3) QUANTUM FIELDS IN A FINITE VOLUME. 1. A LOWER BOUND. Commun. Math. Phys., 85:603–636, 1982. [9] T. Balaban. PROPAGATORS AND RENORMALIZATION TRANSFORMATIONS FOR LATTICE GAUGE THEORIES. I. Commun. Math. Phys., 95:17–40, 1984. A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-040325 [10] J. Magnen and R. Seneor. The infinite volume limit of the φ4 3 model. Annales de l’institut Henri Poincar´e. Section A, Physique Th´eorique, 24(2):95–159, 1976. [11] Konrad Osterwalder and Robert Schrader. AXIOMS FOR EUCLIDEAN GREEN’S FUNCTIONS. Commun. Math. Phys., 31:83–112, 1973. [12] Konrad Osterwalder and Robert Schrader. Axioms for Euclidean Green’s Functions. 2. Commun. Math. Phys., 42:281, 1975. [13] R. Giles. The Reconstruction of Gauge Potentials From Wilson Loops. Phys. Rev. D, 24:2160, 1981. [14] A. Alekseev and Samson L. Shatashvili. Path Integral Quantization of the Coadjoint Orbits of the Virasoro Group and 2D Gravity. Nucl. Phys. B, 323:719–733, 1989. [15] D. J. Gross and Frank Wilczek. Asymptotically Free Gauge Theories - I. Phys. Rev. D, 8:3633–3652, 1973. [16] H. David Politzer. Reliable Perturbative Results for Strong Interactions? Phys. Rev. Lett., 30:1346–1349, 1973. [17] K. Osterwalder and E. Seiler. Gauge Field Theories on the Lattice. Annals Phys., 110:440, 1978. [18] James Glimm and Arthur Jaffe. Quantum Physics: A Functional Integral Point of View. Springer, 1987. [19] V. N. Gribov. Quantization of Nonabelian Gauge Theories. Nucl. Phys. B, 139:1, 1978. [20] I. M. Singer. Some Remarks on the Gribov Ambiguity. Commun. Math. Phys., 60:7–12, 1978. Appendix A. Well-definedness of the Non-Local Gauge Field Operator This appendix provides the detailed proof of Theorem 2.1. Proof of Theorem 2.1. We want to show that for Aµ ( x ) ∈Hs ( R4, su ( N )) with s > 2, and a sequence of smooth functions A(n) µ ( x ) ∈C∞ c ( R4, su ( N )) such that A(n) µ(x)→Aµ(x) in the Hsnorm, the limit lim n→∞ Un(γx, L)A(n) µ(x)U−1 n(γx, L) exists in the sense of distributions and is independent of the approximating sequence. Here, Un(γx, L) = Pexp(iHγxA(n) µ(y)dyµ). We use the Dyson series representation of the holonomy: Un(γx, L) = 1+ ∞ X k=1 ikIγx dyµ 1Iγy1 dyν 2···Iγyk−1 dyρ kA(n) µ(y1)A(n) ν(y2). . . A(n) ρ(yk) Since A(n) µ ( x ) is smooth and compactly supported, this series converges absolutely and uniformly for each n. Let ϕ ( x ) ∈ S ( R4, su ( N )) be a test function. We want to show that the limit lim n→∞ Zd4xTr Un(γx, L)A(n) µ(x)U−1 n(γx, L)†ϕ(x) exists. We can rewrite this as lim n→∞ Zd4xTr h(A(n) µ(x))†U−1 n(γx, L)†ϕ(x)Un(γx, L)i 32A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 enforces diagonalization of the holonomies, and the Faddeev-Popov determinant is shown to be positive in Appendix J. 3) Cluster Expansion: The partition function, Za , and correlation functions are expressed using a cluster expansion. • Connectivity: Two loops are connected if they share a link or if the distance between their centers is less than or equal to 2 L ( a ). A cluster is a set of loops such that for any two loops in the set, there exists a path connecting them through connected loops. • Cluster Expansion Formula: The partition function is written as: Za=X CY C∈C w(C) where: –The sum is over all possible collections Cof disjoint clusters. –Crepresents a single cluster. –w ( C ) is the ”activity” (or ”weight”) associated with the cluster C. • Cluster Activity Definition: The cluster activity w ( C ) is defined as an integral over the link variables within the cluster C: w(C) = ZY (x,µ)∈C dUµ(x)"Y P∈C ∞ X n=0 (−SP[U])n n!!−δC,∅#∆F P [U]Y P∈C χ(P) Where: – The product Q(x,µ)∈C is over all Haar measures for each link (x, x +aˆµ) that belongs to at least one loop in the cluster C. –SP are terms coming from Taylor expanding Skin and SGF . SP = Skin,P +SGF,P . –δC,∅ ensures that empty sets are treated properly (no double counting). –∆F P [U] is the Faddeev-Popov determinant. –χ ( P ) are indicator functions of the set of U ’s for which holonomies in polymer Pare sufficiently close to target diagonal value. • Small Field/Large Field Decomposition: Crucially, we employ a small field/large field decomposition within each cluster. We split the link variables into ”small field” and ”large field” regions based on the magnitude of the fluctuations around the classical solution (minimizing the action within the cluster, subject to the gauge-fixing constraint). This is a standard technique in constructive field theory, essential for controlling the non-perturbative effects. 4) Bounds on Cluster Activities: This is the most critical part of the proof. We need to prove the exponential decay of the cluster activities: |w(C)| ≤ exp(−κ|C|) where |C| is the number of loops in cluster C , and κ > 0 is a constant independent of the lattice spacing a. The main ingredients of this proof are: A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-040333 • Change of Variables: As mentioned in the main text, a crucial change of variables is performed within the cluster activity integral. Link variables are expressed in terms of fluctuations around a background field that minimizes the action within the cluster (subject to the gauge-fixing constraint), and in terms of variables that explicitly represent the deviation of the holonomies from their target values. • Expansion and Bounds: The action within the cluster, SC [ U ], is expanded in powers of the fluctuations. – Propagator Bounds: The quadratic part of the expanded action defines the propagator. We derive rigorous position-space bounds on the propagator, showing exponential decay with a rate proportional to 1 /L ( a ) = a1/2 . This relies heavily on the 1 / ( kL ) 3 decay of the shape factor (Appendix D). The proof involves going to momentum space, using the shape factor, and carefully estimating the inverse Fourier transform. – Vertex Bounds: The higher-order terms in the expansion define the interaction vertices. We derive rigorous bounds on these vertices, showing they are sufficiently small, thanks to the shape factor and asymptotic freedom. • Faddeev-Popov Determinant Bound: We prove ∆ F P [ U ] ≤ exp ( C|C| ), where C is a constant. This, combined with the positivity (Appendix J), controls the determinant’s contribution. • Combinatorial Estimates: The number of ways to form a connected cluster of a given size is estimated. • Putting it All Together: Combining all these bounds (propagator, vertices, gauge fixing, Faddeev-Popov determinant, combinatorics) gives the final exponential decay bound |w ( C ) | ≤ exp ( −κ|C| ). The condition L ( a ) = a−1/2 is essential for ensuring κ remains positive as a→0. We prove that κ > 0 for sufficiently small gand large λ. 5) Kirkwood-Salsburg Equations: The cluster activities satisfy the Kirkwood-Salsburg (KS) equations, a set of integral equations relating the activity of a cluster to activities of smaller clusters. The exponential decay bound ensures that the ”kernel” of the KS equations has norm less than 1, guaranteeing convergence of the cluster expansion. 6) Tightness: To prove weak convergence, we establish tightness of the sequence of measures dµa [ U ]. This means showing that for any ϵ > 0, there’s a compact set K in the space of distributional gauge fields such that: dµa[U](K)>1−ϵ for all a . Tightness is proved by establishing uniform bounds (independent of a) on correlation functions, e.g.: Zdµa[U]|Tr(Uµ1(x1). . . Uµn(xn))| ≤ Cn,x1,...,xn and similarly for products of the non-local field strength. These bounds are derived from the cluster expansion, using the exponential decay of cluster activities and the properties of the action (shape factor decay, gauge fixing). 34A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 7) Prokhorov’s Theorem: Tightness, by Prokhorov’s theorem, implies the existence of a weakly convergent subsequence of measures. 8) Schwinger-Dyson Equations and Uniqueness: The limiting measure satisfies the Schwinger-Dyson equations: Zdµ[A]δ δAµ(x)F[A]e−S[A]= 0 We prove the uniqueness of the solution to the Schwinger-Dyson equations. This is done by showing that they define a contraction mapping in a suitable Banach space of correlation functions. The non-locality (decay of the shape factor) and gauge fixing are crucial for this. The contraction mapping argument guarantees a unique fixed point, hence a unique solution to the Schwinger-Dyson equations. The combination of all these steps, with their rigorous mathematical justification, demonstrates the weak convergence of the regularized measures to a unique continuum measure dµ[A]. Appendix G. Details on Reflection Positivity This appendix gives a detailed, rigorous proof of reflection positivity (OS3 axiom) for the constructed measure. Proof of Reflection Positivity. We want to prove that for any function F of the gauge fields in the positive time half-space (x0>0), we have: Z(ΘF)F dµ[A]≥0 where Θ is the time reflection operator. The proof uses a checkerboard estimate, adapting the methods from [17,18]. 1) Time Reflection Operator (Definition): The time reflection operator Θ acts on functions of the gauge field as follows: • For a spacetime point x = ( x0, x ), its time reflection is θx = ( −x0, x ). •For the gauge field, the action of Θ is: (ΘA)µ(x) = (−Aµ(θx) if µ= 0 Aµ(θx) if µ= 1,2,3 This transformation law ensures that the classical Yang-Mills action is invariant under time reflection. •For the holonomy: ΘU(γx, L) = U−1(γθx, L) = U†(γθx, L) This property is crucial. We derive this from the Dyson Series. Let us denote the time-reflection of the loop γas θγ. The Dyson series is U(γx, L) = 1 + ∞ X n=1 inIγx dyµ 1Iγy1 dyν 2···Iγyn−1 dyρ nAµ(y1)Aν(y2). . . Aρ(yn) A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-040335 Applying the time reflection: (ΘU)(γx, L) = 1+ ∞ X n=1 inIγx dyµ 1Iγy1 dyν 2···Iγyn−1 dyρ n(ΘAµ(y1))(ΘAν(y2)) . . . (ΘAρ(yn)) We change the variable yi→θyi and reverse the direction of integration, picking up a minus sign for each spatial component and no sign for the time component. The key property is that (ΘAµ)(y)dyµ=−Aµ(θy)†(θdy)µ. Therefore (ΘU)(γx, L) = 1 + ∞ X n=1 (−i)nIθγx d(θy1)µIθγy1 d(θy2)ν···Iθγyn−1 d(θyn)ρA† µ(θy1). . . A† ρ(θyn) = 1 + ∞ X n=1 (−i)n(−1)nIθγx dyρ n···Iθγy2 dyµ 1A† ρ(yn). . . A† µ(y1) =Pexp −iIθγx A† µ(y)dyµ=Pexp iIθγx Aµ(y)dyµ† =U(γθx, L)†=U(γθx, L)−1. 2) Checkerboard Estimate: We divide four-dimensional Euclidean spacetime into a lattice of hypercubes ∆ with side length δ . We choose δ to be larger than the non-locality scale L ( a ) but smaller than the correlation length 1 /m (where m is the mass gap). This choice is possible because L(a) = a−1/2→0 as a→0, while mremains positive. The checkerboard estimate is a crucial inequality that allows us to bound the functional integral over all of spacetime by a product of integrals over individual hypercubes. The key idea is that, due to the exponential decay of correlations (a consequence of the shape factor and gauge fixing), the interactions between fields in different hypercubes become weak when the hypercubes are sufficiently separated. The checkerboard estimate takes the following form: Zdµ[A]F[A]≤Y ∆Zdµ∆[A]|F∆[A]|p∆1/p∆ where: •dµ [ A ] is the continuum functional measure (the limit of the lattice measures dµa[U] as a→0). •F [ A ] is a functional of the gauge field. We will later take F [ A ] = (ΘF)F, where Fdepends only on fields at positive times. •The product Q∆is over all hypercubes ∆ in the lattice. •dµ∆ [ A ] represents the functional integral restricted to the gauge fields within the hypercube ∆. This is a crucial point: the integration is only over the link variables Uµ ( x ) that are associated with loops contained within (or intersecting) the hypercube ∆. •F∆ [ A ] is a function that depends only on the gauge fields within the hypercube ∆. 36A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 •F [ A ] = Q∆F∆ [ A ]. This means that we are considering functionals F [ A ] that can be written as a product of functions, each depending only on fields in a single hypercube. •p∆ are positive numbers (which we can choose to be positive integers) that satisfy the condition P∆1 p∆= 1. This inequality is a consequence of repeated applications of the H¨older inequality for integrals. 3) Reflection Positivity for Hypercube Integrals: We now show that for each individual hypercube ∆, the following inequality holds: Zdµ∆[A] (ΘF)F≥0 for any function F that depends only on the gauge fields within ∆ and at positive times (x0>0). This is the most intricate part of the proof. • Integration Order: We choose a specific order of integration over the link variables Uµ ( x ) within the hypercube ∆. We integrate first over the link variables on links that cross the time-reflection plane ( x0 = 0), and then over the link variables in the positive ( x0> 0) and negative ( x0< 0) time half-spaces. This ordering is essential for making the positivity manifest. • Change of Variables: We perform a change of variables within the hypercube integral. We express the integrand in terms of variables defined in the positive time half-space ( x0> 0) and their time-reflected counterparts. This change of variables is crucial for making the positivity manifest. The specific transformation depends on whether a link crosses the x0= 0 plane. • Holonomy Property: The key property of the holonomy under time reflection is: ΘU(γx, L) = U−1(γθx, L) = U†(γθx, L) where θx = ( −x0, x ) is the time-reflected point of x . This property follows directly from the definition of the holonomy as a path-ordered exponential (and its Dyson series representation) and the transformation properties of Aµ under time reflection. A complete, rigorous proof is provided in Appendix G. • Manifest Non-Negativity: After performing the change of variables and using the crucial time-reflection property of the holonomy, the integrand in the hypercube integral, Rdµ∆ [ A ] (ΘF)F can be written in a form that is manifestly non-negative. This relies on the specific form of our non-local action, the gauge-fixing term, and the carefully chosen integration order. The gauge-fixing term, which forces the holonomies to be close to diagonal matrices, is essential for this step. The specific loop structure allows for a clean separation of positive and negative time regions. • Faddeev-Popov Determinant: It is essential that the FaddeevPopov determinant, arising from the gauge-fixing procedure, is positive. A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-040337 This positivity is guaranteed by our proof of the absence of infinitesimal Gribov copies (Theorem 7.2 and Appendix I) and the explicit verification in Appendix J. 4) Combining the Steps: The checkerboard estimate, combined with the reflection positivity of the individual hypercube integrals, proves the overall reflection positivity of the (regularized) functional measure. Taking the limit a→ 0, we obtain the reflection positivity of the continuum measure. q.e.d. Appendix H. Gauge Fixing Validity This appendix provides the proof of Theorem 7.1. Proof of Theorem 7.1. We aim to show that for any smooth gauge field configuration, there exists a gauge transformation that brings it to a configuration satisfying the holonomy-based gauge-fixing condition (Definition 7.1). The condition requires that the holonomies around all loops in our ”sufficiently large” set Γa xare diagonal matrices. 1) Local Gauge Transformations: For each loop γa x,k in the set Γ a x (defined in Definition 2.1), we can find a local gauge transformation gx,k(y) that diagonalizes the discrete holonomy Ua(γa x,k): gx,k(x)Ua(γa x,k)gx,k(x)−1=D(γa x,k) where D ( γa x,k ) is a diagonal matrix in SU ( N ). This is always possible because any unitary matrix can be diagonalized by a unitary transformation. We choose this local gauge transformation gx,k ( y ) to have the following properties: •gx,k(y) is smooth (infinitely differentiable). •gx,k ( y ) is equal to the identity matrix ( gx,k ( y ) = 1) outside a small neighborhood of the loop γa x,k . The diameter of this neighborhood is of order L(a). •gx,k ( y ) is close to the identity matrix everywhere. This can be achieved because the original gauge field is smooth, and therefore the holonomies around the loops are close to the identity when the lattice spacing ais small (since L(a) = a−1/2). 2) Partition of Unity: We construct a smooth partition of unity {wx,k ( y ) } subordinate to a covering of the lattice by neighborhoods of the loops γa x,k. This means that: •wx,k(y) are smooth, non-negative functions. • The support of wx,k ( y ) is contained within a small neighborhood of the loop γa x,k (of size on the order of L(a)). •Px,k wx,k ( y ) = 1 for all lattice points y . This means that at any given point y , the sum of the values of all the partition of unity functions is equal to 1. 3) Global Gauge Transformation: We construct a global gauge transformation g ( y ) as a product of the local gauge transformations, weighted by the partition of unity: 38A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 g(y) = Y x,k exp (wx,k(y) log(gx,k(y))) The product is taken over all loops γa x,k in our ”sufficiently large” set. The logarithm is well-defined because we have chosen the local gauge transformations gx,k ( y ) to be close to the identity matrix. The exponential ensures that g ( y ) is an element of SU ( N ) for all y . This is a standard construction for ”gluing together” local gauge transformations to obtain a global one. 4) Holonomy Transformation: Under this global gauge transformation g ( y ), the discrete holonomy Ua ( γa x,k ) around the loop γa x,k transforms as: U′ a(γa x,k) = g(x)Ua(γa x,k)g(x)−1 where x is the center of the loop γa x,k . Since the support of wx,k ( y ) is localized near the loop γa x,k , and gx,k ( x ) diagonalizes Ua ( γa x,k ), the gauge transformation g ( x ) will be ”close” to gx,k ( x ). Therefore, U′ a ( γa x,k ) will be approximately diagonal. 5) Iterative Refinement: We can make U′ a ( γa x,k )arbitrarily close to a diagonal matrix by repeating this procedure iteratively. We define a sequence of gauge transformations g(n) ( y ), where each g(n) ( y ) is constructed from local gauge transformations that approximately diagonalize the holonomies obtained after applying the previous gauge transformation g(n−1) ( y ). We can also choose the supports of the partition of unity functions to shrink with each iteration. 6) Convergence: We prove that this sequence of gauge transformations g(n) ( y )converges (in a suitable topology, e.g., uniformly on compact sets) to a smooth gauge transformation g ( y ) that makes all the discrete holonomies U′ a ( γa x,k )exactly diagonal. This involves careful estimates on the errors introduced at each step of the iteration and uses the smoothness of the original (discrete) gauge field (link variables). The details of this convergence proof are crucial and are provided in Appendix H. q.e.d. Appendix I. Absence of Infinitesimal Gribov Copies This appendix provides the detailed proof of Theorem 7.2. Proof of Theorem 7.2. We want to show that the holonomy-based gauge-fixing condition eliminates infinitesimal Gribov copies. Suppose we have two gauge fields, Aµ ( x ) and A′ µ ( x ), that both satisfy the gauge-fixing condition (Definition 7.1). This means that the holonomies around all loops in our ”sufficiently large” set Γa xare diagonal for both gauge fields: Ua(γa x,k;A′ µ) = D′(γa x,k) Ua(γa x,k;Aµ) = D(γa x,k) for all loops γa x,k ∈ Γ a x , where D ( γa x,k ) and D′ ( γa x,k ) are diagonal matrices in SU(N). A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-040339 Now, assume that A′ µ ( x ) is related to Aµ ( x ) by an infinitesimal gauge transformation: A′ µ(x) = Aµ(x) + Dµω(x) where ω ( x ) is an infinitesimal su ( N )-valued gauge transformation parameter, and Dµω=∂µω+i[Aµ, ω] is the covariant derivative of ω. We can expand the holonomy Ua(γa x,k;A′ µ) to first order in ω: Ua(γa x,k;A′ µ)≈Ua(γa x,k;Aµ) + iUa(γa x,k;Aµ)Iγa x,k (Dµω)(y)dyµ The gauge-fixing condition requires that both Ua ( γ ; Aµ ) and Ua ( γ ; A′ µ ) are diagonal. Therefore, the term iUa ( γa x,k ; Aµ ) Hγa x,k ( Dµω )( y ) dyµ must also be a diagonal matrix. Since Ua ( γa x,k ; Aµ ) = D ( γa x,k ) is diagonal and close to the identity (due to the gauge fixing and the smoothness of the gauge field), it is invertible. Therefore, we conclude that the integral Hγa x,k ( Dµω )( y ) dyµ must be diagonal for all loops γa x,k in our ”sufficiently large” set. Now we use the fact that Dµω = ∂µω + i [ Aµ, ω ]. Integrating by parts along the closed loop γa x,k, we get: Iγa x,k Dµω dyµ=Iγa x,k ∂µω dyµ+iIγa x,k [Aµ, ω]dyµ=iIγa x,k [Aµ(y), ω(y)] dyµ since the boundary term Hγa x,k ∂µω dyµ vanishes for closed loops (because ω is a single-valued function on the lattice). Therefore, we have the condition that Iγa x,k [Aµ(y), ω(y)] dyµ must be a diagonal matrix for all loops γa x,k in our ”sufficiently large” set. We can approximate the integral around the small loop γa x,k as follows: Iγa x,k [Aµ(y), ω(y)] dyµ≈"Iγa x,k Aµ(y)dyµ, ω(x)#+O(L(a)×derivatives of ω) where x is the center of the loop. Since L ( a ) = a−1/2 , and we are considering smooth ω, the O(L(a)) term goes to zero in the continuum limit (a→0). The ”sufficiently large” condition on the loop set (Definition 2.1, made precise and proven in Appendix M) is now crucial. This condition states that the integrals Hγa x,k Aµ ( y ) dyµ , for the various loops γa x,k in our ”sufficiently large” set, generate (through linear combinations and commutators) the entire su ( N ) algebra. Therefore, if the commutator hHγa x,k Aµ(y)dyµ, ω(x)i is a diagonal matrix for all loops in our ”sufficiently large” set, this implies that [ α, ω ( x )] = 0 for all elements αof the Lie algebra su(N). If an element ω ( x ) of the Lie algebra su ( N ) commutes with all elements of su ( N ), then ω ( x ) must belong to the center of the Lie algebra. However, 40A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-0403 the Lie algebra su ( N ) has a trivial center, meaning that the only element that commutes with all other elements is the zero matrix (up to multiples of the identity, but those are excluded because ω ( x ) is traceless). Therefore, we conclude that ω(x) = 0. This completes the proof that our holonomy-based gauge-fixing condition eliminates infinitesimal Gribov copies. q.e.d. Appendix J. Faddeev-Popov Determinant This appendix addresses the Faddeev-Popov determinant arising from our holonomy-based gauge fixing. The Faddeev-Popov determinant appears when we insert a delta functional into the path integral to enforce the gauge-fixing condition. In the continuum, the gauge-fixing condition is (conceptually): Y xY γx∈Γx δ(U(γx, L)−D(γx)) where the product is over all points x and all loops γx in our ”sufficiently large” set, and D ( γx ) is a target diagonal matrix. On the lattice, the gauge fixing condition is: Y xY γ∈Γa x δ(Ua(γ)−D(γ)) To handle this, we introduce a gauge-fixing term in the action, SGF [ U ] (Definition 5.1), which has the effect of replacing the delta functional with a sharply peaked Gaussian (in the limit λ→ ∞ ). This makes the functional integral well-defined. The general formula for the Faddeev-Popov determinant is: ∆F P [A] = det δG[Aω] δω ω=0 where Aωis the gauge-transformed field: Aω µ(x) = ω(x)Aµ(x)ω(x)−1+iω(x)∂µω(x)−1 and G [ A ] represents our gauge-fixing condition, which in our case is a set of conditions for each loop γa x,k: Gx,k[A] = Ua(γa x,k)−D(γa x,k)=0. Thus we have, ∆F P [A] = det δGx,k[Aω] δω(y)ω=1. We need to compute the functional derivative of Gx,k [ Aω ] with respect to the gauge transformation parameter ω ( y ) and then evaluate it at ω = 1 (the identity). This involves the derivative of the holonomy, which has been discussed in Appendix C. The result will be a linear operator acting on ω ( y ), and the Faddeev-Popov determinant is the determinant of this operator. The indices on the matrix whose determinant we take are ( x, k ) labelling the loop and y labelling the space-time point. A RIGOROUS PROOF OF THE MASS GAP IN SU(N)YANG-MILLS THEORY PREPRINT - VERSION V2; DOI: 10.5281/ZENODO.14975444 YUTA AGAWA UNAFFILIATED; ORCID ID: 0009-0005-6336-040341 The crucial point is that because we are using a gauge-fixing condition that forces the holonomies to be diagonal, and because we have proven that there are no infinitesimal Gribov copies (Theorem 7.2 and Appendix I), the resulting Faddeev-Popov operator is positive definite. This, in turn, implies that the Faddeev-Popov determinant is positive. This positivity is essential for the reflection positivity property (OS3 axiom) of the functional measure. While a completely explicit calculation of the determinant is complicated due to the non-local nature of the gauge fixing, the following key points ensure its positivity: 1) Absence of Infinitesimal Gribov Copies: Theorem 7.2 guarantees that there are no infinitesimal gauge transformations that leave the gaugefixed holonomies unchanged. This means that the operator δGx,k[Aω] δω(y) has no zero eigenvalues (except for the trivial zero eigenvalue corresponding to global gauge transformations, which are factored out). 2) Diagonal Gauge Fixing: The fact that we are forcing the holonomies to be diagonal matrices significantly simplifies the structure of the FaddeevPopov operator. It effectively decouples the different color components. 3) Smoothness: The underlying gauge fields are assumed to be smooth (in the Sobolev space Hs with s > 2), which ensures that the holonomies are well-behaved. These properties, combined, guarantee that the Faddeev-Popov determinant is strictly positive and can be bounded from above by an exponential of the volume of the cluster. This ensures it does not spoil the cluster expansion bounds. Appendix K. Strong Coupling Analysis This appendix provides more details for the strong-coupling analysis (Section 8). We work in the temporal gauge ( A0 = 0) and use the Hamiltonian formulation of the theory. The Hamiltonian for our non-local formulation is: H=g2 2Zd3xTr(Ei(x;γx, L)Ei(x;γx, L))−1 2g2Zd3xTr( ˆ Fij(x;γx, L)ˆ Fij(x;γx, L)) where Ei is the non-local electric field operator (conjugate to Ai ), and ˆ Fij is the non-local magnetic field strength operator (Definition 2.4). In the strong-coupling limit ( g→ ∞ ), the electric field term dominates the Hamiltonian. This term favors configurations where the electric field (and therefore the fluctuations of the gauge field) are large. We use a variational ansatz for the ground state wave functional: Ψ[A] = Y γx ψ(U(γx, L)), ψ(U) = exp αTr(U+U†) where α is a variational parameter, and the product is over all loops γx in our ”sufficiently large” set (Definition 2.1). This trial wave functional is: