Constructing a 0+1 field theory: the quartic oscillator and the Yang–Mills matrix model
Result, 2026-10-27; standard construction assembled here, not a novel existence theorem. For any supplied \(m,\lambda,\hbar>0\), the one-component \(\phi^4_1\) theory has a continuous, stationary Euclidean field on the full time line, a reflection-positive bounded-cylinder algebra, a non-Gaussian vacuum, and a reconstructed positive-energy quantum system with a unique vacuum and a positive spectral gap. The finite-time lattice measures with ground-state boundary weights converge under arbitrary time refinement. Their real-time counterpart is the nonrelativistic quartic oscillator. The same construction applies to the existing \(SU(2)\) bosonic matrix Hamiltonian, including its gauge-singlet sector, using the compact-resolvent result already proved in low-dimensional mass gap, Theorem 6.
The one-dimensional Euclidean field has only time as a base-space coordinate, so spacelike locality and Lorentz boosts impose no nontrivial condition. A relativistic particle may instead be described by a one-dimensional worldline theory with a spatial target and a different Hamiltonian. Neither construction derives \(\hbar>0\), proves the original 1998 consistency conjecture, or constructs four-dimensional Yang–Mills. The statements below use established Feynman–Kac, semigroup and spectral theorems; the assembly is written here, without a separate referee check.
1. Model and prior work
Put \[ V(q)=\frac{\lambda}{4}q^4,\qquad S_E[q]=\int\left(\frac m2\dot q^{\,2}+V(q)\right)d\tau. \tag{1} \] The symbolic \(e^{-S_E/\hbar}\mathcal Dq\) below means an actual Brownian-bridge tilt or its stationary extension, not a Lebesgue measure on path space. Brownian paths are almost surely nowhere differentiable; the kinetic integral in (1) is therefore a mnemonic for the reference measure, not an ordinary pathwise integral.
The Rivero 1998 test, Theorems 2(b) and 3, already proves the fixed-\(\hbar\) real-time product limit and the zero-endpoint Euclidean bridge limit for the potential \(\lambda_{\rm old}q^4\); here \(\lambda_{\rm old}=\lambda/4\). Its zero-endpoint bound also permits \(\hbar\downarrow0\), so it does not select a quantum action unit. The nonlinear cell is one exact elimination step for (1), with its full normalized interaction retained. Its one-step estimates alone were not a stationary continuum construction.
Feynman–Kac and the \(P(\phi)_1\) process are established literature: Simon 1999 (the bounded-below case cited and its theorem passage read in the 1998 test), and Rosen–Simon 1976 (bibliographic record). The real-time quartic oscillator also has an oscillatory-integral treatment by Albeverio–Mazzucchi 2005 (abstract). None of these references is a novelty claim.
2. A vacuum and a gap from the Hamiltonian
Define on \(L^2(\mathbb R,dq)\) the Friedrichs operator \[ H=-\frac{\hbar^2}{2m}\frac{d^2}{dq^2}+\frac{\lambda}{4}q^4.\tag{2}\] The form domain is \(H^1(\mathbb R)\cap L^2(\mathbb R,q^4dq)\). The form is closed and nonnegative. It has compact embedding into \(L^2\): Rellich compactness holds on every bounded interval, while its unit ball satisfies \[\int_{|q|>R}|\psi(q)|^2dq\le R^{-4}\int q^4|\psi(q)|^2dq.\tag{3}\] Thus \(H\) has compact resolvent. Feynman–Kac makes \(e^{-tH/\hbar}\) positivity improving for \(t>0\): the free heat kernel is strictly positive and every finite bridge has a strictly positive potential weight. By the positivity-improving ground-state theorem, the bottom eigenvalue \(E_0\) is simple and its normalized eigenfunction \(\Omega\) is strictly positive. Discreteness implies \(E_1>E_0\) and \[\Delta=E_1-E_0>0.\tag{4}\]
Writing \(\ell=(\hbar^2/(m\lambda))^{1/6}\) and \(q=\ell x\), unitary rescaling gives \[H=\hbar^{4/3}\lambda^{1/3}m^{-2/3} \left(-\tfrac12\frac{d^2}{dx^2}+\tfrac14x^4\right)\tag{5}\] up to that unitary conjugation. Thus the gap is a positive dimensionless eigenvalue difference times \(\hbar^{4/3}\lambda^{1/3}m^{-2/3}\); it vanishes in the indicated small-\(\hbar\) or small-\(\lambda\) limits. This does not imply positivity of \(\hbar\) from classical premises.
3. Arbitrary time refinement
For \(s>0\), let \[p_s(x,y)=\left(\frac{m}{2\pi\hbar s}\right)^{1/2} e^{-m(x-y)^2/(2\hbar s)}\] and define the symmetric single-step transfer kernel \[K_s^{\rm lat}(x,y)=p_s(x,y) e^{-s[V(x)+V(y)]/(2\hbar)}.\tag{6}\] For a partition \(\pi\) of \([0,T]\), their product and integration over internal nodes equal \(p_T(x,y)\) times the expectation, under a Brownian bridge of variance parameter \(\hbar/m\) from \(x\) to \(y\), of \(\exp[-W_\pi/\hbar]\), where \[W_\pi=\sum_j\frac{t_{j+1}-t_j}{2} [V(q(t_j))+V(q(t_{j+1}))].\tag{7}\] The bridge is continuous almost surely, so its polygonal interpolation converges uniformly and \(W_\pi\to\int_0^T V(q(t))dt\) for every sequence with \(|\pi|\to0\). Since \(0\le e^{-W_\pi/\hbar}\le1\), dominated convergence gives the Feynman–Kac bridge, including bounded continuous path observables. This is the argument of the 1998 test, extended from zero to arbitrary fixed endpoints; no change of \(\hbar\) accompanies the refinement.
To obtain vacuum rather than fixed-endpoint data, weight the two ends by \(\Omega(x)\Omega(y)\,dx\,dy\) and normalize. The free heat kernel bounds the tilted kernels, and the \(L^2\) contraction of the free semigroup gives an integrable bound against \(\Omega(x)\Omega(y)\). Dominated convergence therefore applies to these finite-interval distributions. Their continuum normalizer is \(\langle\Omega,e^{-TH/\hbar}\Omega\rangle=e^{-TE_0/\hbar}\). Ground-state boundary weights make the continuum restrictions consistent as the interval grows. Kolmogorov consistency and the locally continuous bridge laws give a stationary process \(q(\tau)\) on the full time line. This construction uses the continuum ground state at the boundary; it does not assert uniform interchange of lattice-spacing and infinite-time limits for arbitrary lattice boundary states.
Equivalently, with \(A=H-E_0\) and kernel \(K_t\) of \(e^{-tH/\hbar}\), the stationary one-time law and transition are \[\mu(dq)=\Omega(q)^2dq,\qquad P_t(x,dy)=e^{tE_0/\hbar}\frac{\Omega(y)}{\Omega(x)}K_t(x,y)dy.\tag{8}\] The eigenvalue equation makes \(P_t\) Markov; symmetry of \(K_t\) makes it reversible for \(\mu\). Times \(\tau_1\le\cdots\le\tau_n\) and bounded functions \(f_j\) give \[\mathbb E\!\left[\prod_{j=1}^n f_j(q(\tau_j))\right] =\langle\Omega,f_1(Q)e^{-(\tau_2-\tau_1)A/\hbar}f_2(Q) \cdots f_n(Q)\Omega\rangle.\tag{9}\]
4. Reflection positivity and non-Gaussianity
Let \(F\) be a finite linear combination of products of bounded functions of \(q(t)\) at nonnegative times, and let \(\Theta\) reflect time and complex-conjugate the observable. From (9), \[\mathbb E[(\Theta F)F]=\|v_F\|^2\ge0,\tag{10}\] where for a single term with \(0\le t_1\le\cdots\le t_n\), \[v_F=e^{-t_1A/\hbar}f_1(Q)e^{-(t_2-t_1)A/\hbar} \cdots f_n(Q)\Omega,\] and \(v_F\) is extended linearly. This is reflection positivity for the bounded cylinder algebra. The associated reconstruction gives the semigroup \(e^{-tA/\hbar}\) and Hilbert space \(L^2(\mathbb R)\): the vectors \(f(Q)\Omega\), for bounded \(f\), are dense because \(\Omega>0\) almost everywhere. The coordinate \(Q\) is self-adjoint as multiplication by \(q\) and \(e^{-itA/\hbar}\) gives real-time unitary dynamics.
The potential is even, so uniqueness and positivity make \(\Omega\) even. The field is not Gaussian. If the one-time density \(\Omega^2\) were Gaussian, positivity would make \(\Omega=C e^{-\alpha q^2/2}\), \(\alpha>0\). But substituting this into \(H\Omega=E_0\Omega\) leaves the nonzero quartic term \(\lambda q^4/4\), which cannot be cancelled by the constant and quadratic terms from the derivative. Thus even the one-time marginal is non-Gaussian. This is a continuum witness, unlike the unscaled shrinking cell cumulant in the Alpoge–Balaban peek.
For bounded real \(f,g\), spectral projection also gives \[|\operatorname{Cov}_\mu(f(q(0)),g(q(t)))| \le e^{-t\Delta/\hbar}\,\|f\Omega\|\,\|g\Omega\|.\tag{11}\] The gap is an energy gap of this explicitly constructed quantum mechanical Hamiltonian, not a four-dimensional field-theory mass gap. The claim here is the bounded-cylinder reconstruction; it does not separately verify a chosen full hierarchy of unbounded-field growth axioms.
5. The gauge-theory rung and relativity
A literal pure gauge connection on a one-dimensional base has no two-form curvature: \(F_{00}=0\). A time circle can retain a global holonomy. The nontrivial model connected to higher-dimensional Yang–Mills is instead its dimensional reduction: spatial gauge components survive as adjoint matrices \(X_i(t)\) and the temporal gauge field imposes Gauss’s law. For \(SU(2)\) and \(D\ge2\) matrices, the matrix theorem proves compact resolvent, a simple positive vacuum, and a gap on the full configuration space and its gauge-invariant subspace. Its potential is a sum of squared commutators and has unbounded commuting valleys, so the scalar argument \(V\to\infty\) cannot be reused. The multicomponent version of (6)–(10), restricted to bounded invariant functions, supplies the stationary reflection-positive gauge-singlet construction; positivity and simplicity make the ground state invariant. This finite-dimensional matrix model keeps only spatially constant modes; it does not justify discarding the other modes of the full theory or construct Yang–Mills on \(\mathbb R^4\). No \(SU(3)\) matrix gap is asserted here.
The field theory (1) is a scalar theory on a one-dimensional spacetime, and is ordinary nonrelativistic quantum mechanics when \(q\) is read as the coordinate of a particle. There are no nontrivial boosts in \(0+1\) spacetime dimensions. A relativistic particle in \(d\) target-space dimensions instead can use the positive-energy Hamiltonian \[H_{\rm rel}=\sqrt{c^2P^2+m^2c^4}-mc^2+V(Q).\tag{12}\] Self-adjointness gives a unitary time evolution when the operator is well-defined, and the positive semigroup can have a subordinate-Brownian, jump-process representation under suitable potential assumptions. See Hiroshima–Ichinose–Lorinczi (abstract and introduction) and Gheryan et al. (abstract), which treats both relativistic and nonrelativistic Nelson models. Their theorems have their own assumptions. A confining external \(V\) selects a frame; it does not establish full Poincare covariance. For \(V=0\) on \(\mathbb R^d\), the bottom of (12) is not a normalizable vacuum, so the stationary ground-state construction above cannot be applied without additional structure. A proper-time worldline and the Klein–Gordon/Feynman propagator require their own constraint and normalization conventions.
6. Consequence for STATE
The preliminary \(0+1\) existence statement is a standard theorem for specified quantum weights and is now assembled with vacuum, reflection positivity, non-Gaussianity and the continuum limit. The same architecture connects to the existing \(SU(2)\) matrix-model gap. The open Newton question remains whether independently justified premises require a positive universal action scale; the \(\hbar=0\) alternatives and the real-time/classical distinction remain essential.