Definition. Spatial link fields and their transfer kernel [fdip-007Z]

Let \(G\) be a finite group, \(U(1)\) or \(SU(2)\), with normalized Haar measure. Fix a finite incidence presentation with vertex set \(V_s\), oriented edge set \(E_s\) and source and target maps \(s,t:E_s\to V_s\). The spatial configuration space and link Hilbert space are \[X=G^{E_s},\qquad \mathcal H_{\mathrm {link}}=L^2(X,dU;\mathbb C),\] where \(dU\) is product Haar probability measure. The gauge action of \(K=G^{V_s}\) is \[(h\cdot U)_e=h_{t(e)}U_eh_{s(e)}^{-1},\qquad (R_hf)(U)=f(h^{-1}\cdot U).\] Haar invariance makes every \(R_h\) unitary.

Choose a real continuous central kernel \(k\) satisfying the positive coefficient conditions: \[\begin {gathered} k(g)>0,\qquad k(g^{-1})=k(g),\\ k(g)=\sum _{\rho \in \widehat G}d_\rho a_\rho \chi _\rho (g), \qquad a_\rho >0,\qquad \sum _{\rho \in \widehat G}d_\rho ^2a_\rho <\infty . \end {gathered}\] All irreducible representations here are complex and unitary. The expansion is uniformly absolutely convergent. Pointwise positivity and coefficient positivity are separate hypotheses.

Let \(V:X\to \mathbb R\) be continuous and gauge invariant, and set \(b(U)=e^{-V(U)/2}\). Compactness makes \(b\) and \(b^{-1}\) bounded. If \(M_b\) denotes multiplication by \(b\), define \[\begin {aligned} c(U,U')&=\prod _{e\in E_s}k(U_eU_e'^{-1}),\\ (C_kf)(U)&=\int _Xc(U,U')f(U')\,dU',\\ T&=M_bC_kM_b. \end {aligned}\] This is the pure-gauge transfer operator for the specified data. Its integral kernel is \(b(U)c(U,U')b(U')\). The spatial potential may be a sum of Wilson plaquette terms, but the definition only needs its stated invariance and continuity.

The factorization separates the two spatial half-weights from the temporal convolution. It parallels the gauge factors in [luscher1977transfer, Eqs. (17), (20)–(21), p. 287]; the source comparison distinguishes this pure-gauge model from that paper's gauge and fermion construction.