The operator product expansion converges in massless \(\varphi_4^4\)-theory (Q261590)

In [the second and the third author, ibid. 313, No. 1, 257--290 (2012; Zbl 1332.81119)] for the massive Euclidean \(\phi^4_4\)-theory in perturbative loop expansion, the operator product expansion (OPE) is shown to converge at any loop order when considering composite operator insertions into correlation functions. The present paper proves the same result for massless theory. The authors say the proof of the massless case requires entirely new techniques, because we have to control with sufficient precision the exceptional momentum singularities of the massless correlation functions. Let \(\{\mathcal{O}_A\}\) be the collection of all local operators of the theory. OPE states \[ \mathcal{O}_A(x)\mathcal{O}_B(y)\sim \sum_C\mathcal{C}^C_{AB}(x-y)\mathcal{O}_C(y). \] It is a short-distance expansion and expected as good as \(O(|x-y|^\Delta)\) as \(|x-y|\to 0\) when the sum up includes a sufficiently high operator dimension. Throughout in this paper, theory means massless Euclidean \(\phi^4_4\)-theory. Then taking test functions \(F_i\) from the Schwartz space \(\mathcal{S}(\mathbb{R}^n)\) and set \(\phi(F)=\int d^4x\phi(x)F(x)\), the first main theorem of this paper is \[ |\langle\mathcal{O}_A(x)\mathcal{O}_B(0)\phi(F_1)\cdots\phi(F_N)\rangle-\sum_{C;[C]\leq D}\mathcal{C}^C_{AB}(x)\langle \mathcal{O}_C(0)\phi(F_1)\cdots\phi(F_N)\rangle|_{L-loops} \leq C_{AB}^\Delta |x|^\Delta \] with explicit description of \(C^\Delta_{AB}\)(\S1 (3)). Here \(D=[A]+[B]+\Delta\) and \([A]\) is the dimension of a composit operator \(\mathcal{O}_A\) (Theorem1). As a consequence, if the Fourier transform \(\hat{F}\) of \(F\) has compact support, or more generally, if \(|\hat{F}_i(p)|\) decays more rapidly than \(e^{-|p|^a}\), \(a>2(N+2L+3)\), then OPE converges. The second main theorem (Theorem 2) of this paper takes test functions to be localized in momentum space away from exceptional configurations and have bounded total momentum. Let \(P\) be the admitted maximum momentum, and \(\epsilon\) be the minimal distance to the set of exceptional momentum admitted in the support of \(\hat{F}_1(p_1),\ldots,\hat{F}_N(p_N)\) (For the detailed definition of the class of test functions, see \S5, p. 429). Then, the following estimate is proved \[ |\langle\mathcal{O}_A(x)\mathcal{O}_B(0)\phi(F_1)\cdots\phi(F_N)\rangle -\sum_{C:[C]\leq D}\mathcal{C}^C_{AB}\langle\mathcal{O}_C(0)\phi(F_1)\cdots\phi(F_N)\rangle|_{L-loops}\leq C^{\Delta,P,\epsilon}_{AB}|x|^\Delta, \] with the explicit description of \(C^{\Delta,P,\epsilon}_{AB}\) (\S1 (4)). It shows the remainder behaves as \(P^{\Delta(N+2L+1)}/\sqrt{\Delta!}\) if \(\Delta\) is large. Therefore OPE converges but it converges more slowly for large \(P\). The authors say that this is physically plausible, since \(C^{\Delta,P,\epsilon}_{AB}\) involving the terms \(\epsilon^{-3n}\) and powers of \(\log\epsilon\). Hence, the bound on the remainder is large if \(\epsilon\) becomes small, which is also physically plausible. Proofs of these theorems use the renormalisation group flow framework. It endows the theory with UV cutoff \(\Lambda_0\) and IR cutoff \(\Lambda\). The regularized momentum space propagator is defined to be \[ C^{\Lambda,\Lambda_0}(p)=\frac{1}{p^2}\left[\exp\left(-\frac{p^2}{\Lambda_0^2}\right)-\exp\left(-\frac{p^2}{\Lambda^2}\right)\right]. \] Then the flow equations are written in terms of the hierarchy of ``connected amputated Schwinger functions'' (CAS's), whose generating functional is given by \[ -L^{\Lambda,\Lambda_0}=h\log[\mu^{\Lambda,\Lambda_0}\ast\exp(-\frac{1}{h}L^{\Lambda_0})]-h\log Z^{\lambda,\Lambda_0}. \] The correlation (Schwinger) functions of \(n\) basic fields can be recovered from the CAS's at the end. Set formal power series expansion of \(L^{\Lambda,\Lambda_0}\) in terms of of Feynman diagrams as \[ L^{\Lambda,\Lambda_0}(\phi,M) =\sum_{N>0}^\infty\sum_{L=0}^\infty\int\frac{d^4p_1}{(2\pi)^4}\cdots\frac{d^4p_N}{(2\pi)^4}\bar{\mathcal{L}}^{\Lambda,\Lambda_0}_{N,L}(p_1,\ldots,p_N;M)\hat{\phi}(p_1)\cdots\hat{\phi}(p_N), \] where \(M\) is the renormalisation scale, estimates of \(\partial^w_{\overrightarrow{p}}\mathcal{L}_{N,L}^{\Lambda,\Lambda_0}(\overrightarrow{p};M)\) are given as (Theorem 3 (69 and 70)). They are key of the proofs of main Theorems (Theorem 1 and 2). Proof is done via the induction about \(N\) and detailed analysis of weighted trees (cf. Definition 1). The authors say to control the (physical) IR-singularities present in massless theories, it needs to extend the techniques of ``tree bounds'' (cf. \textit{G. Keller} and the third author [ibid. 161, No. 3, 515--532 (1994; Zbl 0793.60113)]).. These estimates safe to take limits \(\Lambda\to 0\), \(\Lambda_0\to \infty\). Since the CAS's with insertion of the composition operator \(\mathcal{O}_A\), \(A=[N',w']\) is a multi-index, are defined through the flow equation (\S3.2,. \S4 (116)), bounds on the CAS's with one and two insertions are computed from Theorem 3 (Theorem 4, Theorem 5). Proofs of Theorem 1 and Theorem 2 are given in \S5 by using these results. Some explicit bounds which are useful in the inductive proof of Theorem 3 are corrected in Appendix A. In Appendix B, a bound on the smeared connected Schwinger functions with two insertions (\S4. Corollary 6) is proved.